69530 v2 Middle East and North Africa Region Labor Migration from North Africa Development Impact, Challenges, and Policy Options Volume 2 Statistical Appendix A project implemented A project funded by the by the World Bank European Union This report as well as the background research underlying the analysis and conclusions of this report constitute part of an EC- Funded World Bank Program of International Migration from Middle East and North Africa and Poverty Reduction Strategies, a program of migration-related research and activities to identify and support the implementation of projects, policies, regional arrangements, and institutional reforms that will maximize the benefits of international migration flows and reduce their costs. The views herein are those of the authors and should not be attributed to the World Bank, the European Commission, or the institutions and countries they represent. Keller MNA 5-27-10vol2.indd 1 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Keller MNA 5-27-10vol2.indd 2 5/27/10 2:41 PM Table of Contents Appendix 1: Measuring Growth, Accumulation, and TFP Growth........................................1 Appendix 2: Description of Migration/Remittances and Poverty Analysis in Egypt: Methodology and Results.......................................................................5 Appendix 3: Description of Remittances and Poverty Analysis in Morocco: Methodology and Results.......................................................................................19 Appendix 4: Description of Migration/Remittances and Labor Market/ Employment Analysis in Egypt: Methodology and Results. ..........................21 Appendix 5: Description of Migration/Remittances and Labor Market/ Employment Analysis in Morocco: Methodology and Results......................69 Appendix 6: Description of Migration/Remittances and Decisions Affecting Children in Egypt: Methodology and Results. ..................................................77 Appendix 7: Description of Migration/Remittances and Decisions Affecting Children in Morocco: Methodology and Results............................89 Appendix 8: Impact of Remittances on Growth: Methodology and Results. ....................93 Appendix 9: Return Migration and Occupational Mobility...................................................97 Appendix 10: Return Migration and Entrepreneurship Analysis........................................107 Appendix 11: Review of Institutional And Legal Framework for Migration in Spain and the Netherlands. ............................................................................115 Appendix 12: Computable General Equilibrium Analysis of Impact of Increasing MENA to Europe Migration.......................................................125 iii Keller MNA 5-27-10vol2.indd 3 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Keller MNA 5-27-10vol2.indd 4 5/27/10 2:41 PM Appendix 1: Measuring Growth, Accumulation, and TFP Growth (From Keller and Nabli, 2007, with estimates updated by Keller 2009) To examine how the MENA1 region’s growth observed peak level) to maintain the assumption has changed since it began its comprehensive that “trueâ€? productivity can only improve and structural reform process, we made simple calcu- that measured reductions in TFP can only reflect lations of the change in both the region’s rate of short-term fluctuations. accumulation, as well as the region’s total factor productivity (TFP) growth. For our purposes, we have adopted a more casual approach about our measurements. Our TFP growth is the residual of what cannot interest is to explore how MENA and North Afri- be explained by investments if we assume those ca’s overall growth has improved or deteriorated investments (both physical and human) earn a since it began the structural reform process. In reasonable rate of return. TFP growth is often the end, growth will be determined by both ac- thought of as “technical progress,â€? but in fact, cumulation of physical and human capital, as well as the residual of a growth accounting estima- as the overall manner in which those factors are tion, it not only embodies the differences across put to production. For the MENA region, things countries in their progress in the adoption of such as improved capacity utilization of capital better technology, but also reflects a host of non- and human capital by the region are precisely the technological differences, including changes in elements we believe may be heavily affected by the utilization of both capital and labor, changes structural reform, and thus we would like to have in schooling quality, and changes in the overall this effect reflected in our estimates. At the same efficiency with which factors are allocated in the time, as we discuss in the subsequent section, we production process. Because of the many other have controlled for global shocks. factors that can potentially affect the growth residual, much empirical work has focused on Under many circumstances, the environment reducing those elements of the residual (TFP) created to encourage investment would also that do not reflect actual shifts in technological correspond to an environment in which those in- opportunities in the economy. For example, vestments could be productive. But in the MENA adjustments for the business cycle have been region, accumulation and productivity have often introduced to account for the short-term fluc- gone in opposite directions, such as during the tuations in capacity utilization (Griliches, 1979; period of massive public sector investments that Lefort and Solimano; 1994; Fajnzylber and Leder- yielded rates of return well below international man, 2000). An alternative procedure employed by Griliches and Lichtenberg (1984) has been to estimate growth over five-year periods, and 1 The Middle East and North Africa region (MENA) comprises Algeria, Bahrain, Djibouti, Egypt, Iran, Iraq, Jordan, Kuwait, to only allow the TFP series to increase or stay Lebanon, Libya, Morocco, Oman, Qatar, Saudi Arabia, Syria, constant (resetting any values to the previously Tunisia, United Arab Emirates, West Bank and Gaza, and Yemen. 1 Keller MNA 5-27-10vol2.indd 1 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options norms. Examining growth alone will mask these international evidence, a reasonable approxima- very different effects, and the somewhat anemic tion of that rate of return is 10 percent, which we growth that has characterized the region since have assumed for the purposes of our analysis. reform may be more a reflection of significantly lower public investments than of continuing poor TFP growth was calculated over ten-year pe- productivity performance. And from the stand- riods from 1960–2000, and then from 2000–2005, point of evaluating the impact of the region’s rather than on an annual basis, to minimize the structural reform, it is precisely TFP growth error that is inherent in current capital stock that we would expect to be most influenced by measurements. National accounts would attri- changes in national policies that enhance the bute any investment expenditures made over efficiency of capital and labor. the year, even the last day of the year, to that year’s capital stock. However, it is unlikely that Data and Methodology that investment expenditure would contribute to economic growth immediately, but rather TFP growth estimates were made utilizing panel would only create the potential to contribute to data of capital stock accumulation, human capi- growth into the future. To reduce this lag-effect tal stock accumulation, and GDP growth from that physical capital exhibits, we calculated TFP 1960–2005. Estimates of the physical capital stock growth based on ten-year averages (except for for a sample of 83 economies from 1960 to 1990 the final period, which is a five-year average). come from Nehru and Dhareshwar (1993), which was created by a perpetual inventory method Production was assumed to follow a Cobb- from investment rates from 1950 forward, with Douglas specification with constant returns initial assumptions about the capital/output ratio, to scale between physical and human-capital- and assuming a common fixed annual geometric augmented labor: depreciation rate of 0.04. These capital stock data were extended to 2005 using the growth rates of Yt = A (t) * Kta*Ht(1–a) constant price local currency investment from the World Bank’s World Development Indicators where Y is output, A is an index of total factor database,2 and applying similar assumptions on productivity, and K and H are the stocks of physi- the depreciation rate. Capital stock estimates for cal and human-augmented labor, respectively. another 12 economies, including four economies Dividing both sides by the work force, taking in the MENA region of particular interest to us,3 logs, and first-differencing, growth of output per were created according to a similar methodology, laborer can be related as follows: using investment rates from 1960 forward. ln (yi / yi–1,) = a ln (kt / kt–1) + (1–a) Real GDP in constant local currency also ln (ht / ht–1) + ln (At / At–1) comes from World Bank data. The human capital-augmented labor stock was estimated, To determine the coefficients on capital and using both labor force estimates from the World human-capital augmented z, a and (1-a), the Bank’s World Development Indicators, and esti- average annual rate of GDP per capita growth mates of the educational attainment of the adult over the decade was regressed on average growth population from Barro and Lee.4 The functional of physical capital per worker and human-capital form of human capital augmented labor has been assumed as: 2 In the case of MENA economies, where there were inconsis- tencies, the World Bank MENA regional database investment H = L e (r * S) series was preferred. 3 The four focus countries of this study include Algeria, Morocco, Tunisia and Egypt. where L is the labor force and S is the average 4 Barro and Lee, 2000. Educational attainment data (available years of schooling of the adult population, and r until 1999) were extended to 2000 assuming constant growth is the rate of return to schooling. According to between 1995–2000. 2 Keller MNA 5-27-10vol2.indd 2 5/27/10 2:41 PM Appendix 1: Measuring growth, accumulation, and TFP growth per worker with a least squares trend over the purpose here is not to break new ground in mea- entire period of availability (1960–2000). suring TFP, but to evaluate the region’s perfor- mance in factor allocation and efficiency. Thus, From our estimation, the elasticity of output we have calculated the TFP using three distinct of physical capital was estimated to be 0.49, calculation of factor shares—ak=0.3, ak=0.4, and somewhat higher than the average estimated ak=0.5, to check the sensitivity of the region’s coefficient from previous research, but within growth performance to the assumptions made the range of accepted parameters. This may be on the output elasticities. The results, in terms due to the inclusion of several more developing of orders of magnitude, are robust to changes in countries than in the original Nehru-Dhareshwar these elasticities. Final TFP estimates utilized physical capital stock dataset, made possible an assumed output elasticity with respect to using World Bank data. At the same time, our capital of 0.4 3 Keller MNA 5-27-10vol2.indd 3 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Keller MNA 5-27-10vol2.indd 4 5/27/10 2:41 PM Appendix 2: Description of Migration/ Remittances and Poverty Analysis in Egypt: Methodology and Results (The following description is taken from Roushdy, Assaad, and Rashed, 2009) The analysis mainly relies on data from the Egypt revealed that there are two distinct attrition Labor Market Panel survey of 2006 (ELMPS 06), processes at play. The first is if the entire house- which is one of the first true nationwide longitudi- hold could not be located in 2006 and the second nal surveys to be carried out in Egypt. It attempted is when an individual who split from one of the to track households and individuals first inter- households that were successfully tracked could viewed in 1998 as part of the Egypt Labor Market not be found. The rate of the first type of attri- Survey of 1998 (ELMS 98) and re-interview them tion was about 23.6 percent (1,138 households) in 2006. Both the ELMPS 06 and ELMS 98 were at the household level. More than 54 percent conducted by the Economic Research Forum (615 households) of this first stage attrition cases (ERF) in cooperation with CAPMAS. The ELMS resulted from the loss of identifying records of 98 was carried out on a nationally representative the households between 1998 and 2006, but, sample of 4,816 households. The ELMPS 06 tracks luckily, the process by which they were lost was the labor market and demographic characteristics almost entirely random (see Barsoum, 2008). of the households and individuals interviewed in The remaining attrition cases were due to the 1998, and any new households that might have total relocation of the household, the death of all formed as a result of splits from the original household members, or, in a few cases, refusal to households. The ELMPS 06 sample consists of a participate in the survey. On the other hand, the total of 8,349 households distributed as follows: second attrition process results from the inability (i) 3,684 households from the original ELMS 98 to locate individuals who split from their original survey, (ii) 2,167 new households that emerged households, conditional on finding the original from these households as a result of splits, and households in the first stage. The rate of this (iii) a refresher sample of 2,498 households. Of the second type of attrition was about 15.4 percent. 23,997 individuals interviewed in 1998, 17,357 (72 Of the 18,856 members of the 1998 households percent) were successfully re-interviewed in 2006, found in 2006, 14,661 were still in their original forming a panel that can be used for longitudinal households, 790 had died, 220 had left the country, analysis. The 2006 sample contains an additional and 3,185 had split off to form separate house- 19,743 “newâ€? individuals. Of these, 2,663 indi- holds within Egypt. Of those splits, we success- viduals joined the original 1998 households, 4,880 fully located 2,694 individuals, implying that the joined the split households, and 12,200 were part remaining 491 of the splits could not be located. of the refresher sample of households.5 Our analysis of the attrition process that oc- 5 The data description and attrition analysis presented here is curred in the panel tracked from 1998 to 2006 based on Assaad (2007) and Assaad & Roushdy (2008). 5 Keller MNA 5-27-10vol2.indd 5 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options An examination of the household and indi- members living abroad, the amount and type of vidual correlates in 1998 of those two attrition these remittances, and which household member processes revealed that some household char- receives the remittances. ELMPS 06 also includes acteristics in 1998 were in fact systematically information on the place and reason of migration associated with the first type of attrition, but no for individuals who were in the household in 1998 individual characteristics in 1998 were associated but were not found in 2006 because they migrated with the second type of attrition (see Assaad between the 1998 and 2006. and Roushdy 2008 for a detailed comparison of individuals/households who left versus those who Although the two ELMSs are rich sources stayed in the sample). We expect that very few of information on labor market dynamics and cases of those missing households were due to individual and household characteristics, the migration of all of the household members, since ELMSs samples were not designed to measure migration in Egypt is often of a short-term nature migration. Accordingly, the number of migrants by a single member in the household. However, appearing in each of the ELMSs is fairly small. we expect that households that migrate in their The ELMPS 06 sample contains about 603 return entirety would tend to be richer, since the whole migrants (who migrated and returned before the family can afford to relocate. Hence, not correct- 2006 survey interview) and 396 current migrants ing for this household-level attrition, when using (who were still living abroad during the 2006 the panel data, might lead to a downward biased interview). While in the ELMS 98 there are only estimate of the effect of migration on poverty. about 471 return migrants and no information Accordingly, weights based on the probability was collected about current migrants. Hence, of non-response were constructed to adjust the we do not expect to obtain accurate trends of cross-sectional and panel samples from the migration and remittances flows from the ELMSs ELMPS 06 for these attrition processes. Only the data that would coincide with official estimates. variables that were found to impact the probabil- However, to the best of our knowledge, the ity of the first type of attrition in a significant way ELMPS 06 is the only recent national household were used to predict the weights that correct for survey that collects information on incidence of attrition. Those weights are applied whenever international migration and remittances. panel data is used in the analysis of this paper. The focus of the anlaysis uses the ELMPS The ELMPS 06 and ELMS 98 provide detailed 06 sample in the cross-sectional analysis, since information on household housing conditions, it provides richer information, relative to ELMS ownership of durables, access to basic services 98, on international migration and remitances. and the neighborhood infrastructure. It also con- tains a great deal of information on the household Determinants of Migration and members’ education, employment status, time Remittances allocation, job mobility, earnings, migration, and household enterprises. With regard to migration Before investigating the effect of migration questions, each round of the Egypt Labor Market and remittances on household poverty status, Surveys (ELMSs) contains information on inter- we are interested in exploring the household nal and international migration history (e.g., place characteristics that might motivate the decision of birth, year leaving place of birth, and the place to migrate and remit. In this section, a Probit and date of the previous two moves if different specification is used to model the likelihood from the current place of residence). ELMS 98 in- of migration (receiving remittances) at the cludes only one (yes/no) question on whether the household level. The dependent variable takes household receives remittances from relative(s) the value 1 if the household, h, is a migrant living abroad. However, in ELMPS 06, a new household (remittances-recipient household) module on current migrants and remittances was and zero otherwise. The explanatory variables added and it includes questions on whether the consist of a set of the household and household household receives remittances from household head characteristics. It is worth mentioning here 6 Keller MNA 5-27-10vol2.indd 6 5/27/10 2:41 PM Appendix 2: Description of Migration/Remittances and Poverty Analysis in Egypt: Methodology and Results that we need to restrict the analysis to variables three dummies variables (illiterate or no degree that are less likely to be caused by the migration is the omitted category): primary or preparatory decision per se. For instance, one should try to degree, secondary degree, and above secondary avoid variables such as: the number of children degree. Marital status is captured by the two in the household below age 5, household wealth, dummies (not married is the omitted category): residence, and current household head charac- married, and divorced or widowed. teristics. Such variables are arguably endogenous to migration decision. The number of newly born/ Moreover, since migration is a chain phenom- young children is obviously affected by the spouse enon, it is often expected that households belong- absence from the household. Household wealth ing to traditionally migrant sender communities and residence often change after migration. Also, are more likely to have better social networks the household head and his/her characteristics abroad which can potentially help in the migration change if the original head is the migrant member. process of other household members. Accord- In the regression analysis of this section we try ingly, in this analysis we include the following two to avoid such variables. Instead of using current variables to proxy for migration networks: the household head’s characteristics in the regression percent of households with at least one current analysis, we introduce a migration-neutral head migrant in the village/shiakha of the household as a substitute. If the current head is a male, we and its interaction with the average years of use the household head’s spouse characteris- schooling of adult members of the household. The tics—regardless of whether the household has a percent of households with at least one current migrant or not. If the head is not married we use migrant in the village/shiakha of the household the characteristics of the oldest female (above is obtained from the 2006 Census.8 Such proxies age 15) living in the household. Only when the have been frequently suggested in the literature. head is a male living alone, we use his own char- We believe that, in Egypt, these variables are good acteristics. We are aware that the characteristics proxies of the size of the household’s migration of the migration-neutral head would have less network abroad. We also expect that the adult explanatory power in comparison to that of the members of the households, specifically those current household head, since under this defini- who are more educated, would make better use of tion the substitute head might have a marginal the information available through their networks. role in household decisions.6 However, contrary to the current household head, we believe that Results the characteristics of this migration-neutral head are arguably exogenous to migration decision, Table A1 shows the regression results of the since our sample shows that women generally do migration and remittances decisions. In this pa- not migrate alone. Also, in Egypt generally, there per we report marginal effects as well as Huber- exists a correlation between the characteristics of White adjusted standard errors to account for the household members; and hence we expect the heteroskedasticity in all tables.9 In both tables, characteristics of the migration-neutral head to be similar to that of the current household head. 6 A better alternative for the migration-neutral head is to use the characteristics of the household head before migration. In the regression analysis, the household Unfortunately, this information is not available in the data. 7 As discussed in the previous section, migration in Egypt is composition is captured by five variables: number often of a short term nature; hence, the number of children above of children age 6–157, number of unmarried males age 6 (relative to the number of children less than age 5) are less age 16–30, number of unmarried females age likely to be affected by the spouse absence from the household. 8 As has been suggested in the literature (Section 4.2), it would 16–30, number of elderly aged 64+, average years have been better to use the lagged/historical migration levels of schooling of males above age 18, and average instead of the same year of the household survey, but, unfor- years of schooling of females above age 18 in the tunately, migration information was not collected in censuses prior to that conducted in 2006. household. The substitute head’s characteristics 9 Marginal effects are based on marginal change for continuous include: age, marital status and education. The variables and change from 0 to 1 for dummy variables. Coef- substitute head education is measured by the ficients are available upon request. 7 Keller MNA 5-27-10vol2.indd 7 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Table A1: Determinants of Egyptian migration and receiving remittances – Household level, 2006 Migration Remittances Variables (1) (2) (3) (4) No Children 6–14 0.003 0.002 0.002 0.001 (0.002) (0.002) (0.001) (0.001) No. Males 15–29 0.009*** 0.008*** 0.003* 0.003* (0.003) (0.002) (0.002) (0.002) No. Females 15–29 0.008*** 0.006** 0.003* 0.002 (0.003) (0.003) (0.002) (0.002) No. Elderly 64+ 0.001 0.002 –0.001 0.000 (0.006) (0.005) (0.004) (0.003) Avg. Male 18+ Years of schooling –0.006*** –0.005*** –0.004*** –0.004*** (0.000) (0.000) (0.000) (0.000) Avg. Female 18+ Years of schooling 0.001 0.001 0.002*** 0.002*** (0.001) (0.001) (0.001) (0.000) Household substitute head characteristics Age –0.001 –0.000 –0.000 –0.000 (0.001) (0.001) (0.001) (0.001) Age square 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) Married(d)1 –0.008 –0.006 –0.008 –0.008 (0.016) (0.014) (0.011) (0.010) Divorced or Widowed(d)1 –0.033*** –0.029*** –0.025*** –0.021*** (0.009) (0.008) (0.004) (0.003) Primary or preparatory degree(d)2 0.002 0.002 –0.008 –0.006 (0.010) (0.009) (0.005) (0.004) Secondary degree(d)2 0.023* 0.017 –0.002 –0.004 (0.013) (0.012) (0.007) (0.006) Above secondary degree(d)2 0.037** 0.028* –0.002 –0.005 (0.017) (0.015) (0.008) (0.006) % of HHs with Migrants in Shiakha/village from Census 2006 0.584*** 0.242*** (0.132) (0.080) % of HHs with Migrants in Shiakha/village x Avg. Yrs 0.076*** 0.041*** of schooling of 18+ (0.015) (0.009) Observations 8345 8345 8345 8345 Pseudo R-squared 0.0621 0.117 0.124 0.170 Notes: Marginal effects are reported and robust standard errors in parentheses (d) for discrete change of dummy variable from 0 to 1 *** p<0.01, ** p<0.05, * p<0.1 1 reference category: never married 2 reference category: no educational certificate column 1 and 3 control for the household compo- 10 Unfortunately, the ELMPS data does not provide the remit- sition and the substitute head’s characteristics, tances senders’ characteristics or the type of relationship of the while column 2 and 4 investigate the effect of sender to his/her home family. Thus, it is important to note here that in the absence of such variables, it is difficult to interpret the network variables.10 these results as different motives for sending remittances (see Acosta 2006 for a discussion). The table shows that, in both specifications, households with larger numbers of males and 8 Keller MNA 5-27-10vol2.indd 8 5/27/10 2:41 PM Appendix 2: Description of Migration/Remittances and Poverty Analysis in Egypt: Methodology and Results females (age 15–29) are more likely to have a mi- the regressions explaining the likelihood of be- grant member and receive remittances. Also, in ing a migrant household, since both dependent all models, adult males’ average years of school- variables are highly correlated (R2 = 0.686). ing decreases the likelihood of migration and In fact, as mentioned above, 66 percent of the receiving remittances, while the females’ average households with at least one current migrant years of schooling only increases the likelihood member receive remittances; although the data of receiving remittances. These results should be does not show whether remittances are actually taken with caution, since these might be the re- received from those migrant family members. sults of migration per se. As mentioned earlier, if On the other hand, 86 percent of households migration selects on education and gender, adult receiving remittances have at least one current males with higher education levels would be the member abroad. ones who are more likely to migrate—which in turn would lead to poorer endowment of human Impact of migration and remittances capital among males who stay in the household. on household poverty Incidents of migration and remittances are less common among households with widowed or This section investigates the effect of migration divorced substitute heads. The household substi- and receiving remittances on household poverty tute head education is only significant in the mi- status. The variable used to investigate the ef- gration regression. A household whose substitute fect of remittances in the regression analysis is head has above secondary education, relative to whether the household receives transfers from illiterate heads, has a higher likelihood of having a abroad, instead of the amount of remittances in migrant member by about four percentage points order to avoid possibilities of recall bias.11 in specification 1 and by about three percentage points in specification 2. The outcome variable of interest in this analysis is whether the household is poor or not. Controlling for the network variables im- The following Probit regression is estimated to proves the fit of the migration and remittances explain the poverty status of the household: models. In both models, the migration network variables increase the likelihood of being a mi- ′ β + I h γ + eh ) ′ , I h ) = Φ( X h Pr( Poorh = 1 | X h grant and a remittances recipient household. Belonging to a village/shiakha that is traditionally The outcome is a binary variable which takes migrant-sending increases the likelihood of migra- the value 1 if the household h belongs to the low- tion and receiving remittances. More specifically, est quintile of the wealth distribution and zero a one percent increase in the fraction of migrants otherwise. Xh is a vector of the household and in the village/shiakha increases the probability of the household head characteristics. The set of migration by 58 percentage points (column 2) household and household head characteristics and the probability of receiving remittances by 24 included in this poverty equation consists of: percentage points (column 4). While, the interac- the household region of residence, number of tion term of the percent of migrants and average children age 0–5, number of children age 6–15, years of schooling further increase the likelihood number of unmarried male age 16–30, number of of migration by 7.6 percent and the likelihood of unmarried females age 16–30, number of elderly receiving remittances by 4 percent. This fits with age 64+, average years of schooling of males age our expectation that the more educated members 18+, average years of schooling of females age of the household are those who are more likely to 18+ in the household, and the substitute head make use of the migration information available through their network. 11 Since international transfers are generally considered another source of income, they traditionally tend to be underreported in It is not surprising that the results of the household surveys in comparison to macroeconomics balance regressions explaining the likelihood of receiving of payment figures. For a detailed discussion of this issue, see remittances are remarkably similar to those of Freund and Spatafora (2005) and Acosta et al. (2006). 9 Keller MNA 5-27-10vol2.indd 9 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options age, age square, marital status, and education. In the first-stage of each of the two-equa- Four interaction terms are also included: the in- tion model estimations, we estimate the full teraction of migration (remittances) with a rural model specification of the migration (remit- dummy of the household residence, and with tances) equation presented in column 2 (4) the household head education dummies. Those of Table 10.12 We use the two migration social interaction terms would allow us to investigate network variables discussed above (the percent whether poverty alleviation impact of migration of households with at least one current migrant and remittances are higher for migrants from in the village/shiakha of the household and its urban household versus those from rural house- interaction with the average years of schooling of holds and whether this impact differs depending adult members of the household) to instrument on the education status of the household. Ih is an for migration and remittances. We believe that indicator of whether the household has a migrant these instruments are good proxies of the local member (receive remittances, respectively) and migration network, since households belonging eh is the error term. to traditionally migrant sender communities are more likely to have better social networks Migration and remittances may be endog- abroad, which can potentially help in the migra- enous to household poverty. Also households tion process of other members. However, it is may not be randomly selected into being migrant not easy to defend that the number of migrants households or remittances recipient households. at the community level impacts household living The literature has often depended on instru- standard only through affecting migration; since, mental variables (IV) techniques to overcome for instance, among the most important deter- such endogeneity and selection bias problems. minants of migration are labor market opportu- However, since both poverty and migration nities which affect both migration and poverty. (receiving remittances, respectively) are binary One possible improvement, to reduce the effect variables, the model estimation strategy is not a of this potential problem, is to include others trivial choice. Newey (1987) argues that using a controls at the household community-level in two-stage least square (2SLS) in case of a binary the poverty equation. Accordingly, we include dependent outcome and a binary endogenous the following five variables to control for labor variable might lead to inconsistent estimates, and market structure at the cluster-level: the percent instead suggests the use of Amemiya’s general- of unemployed adult males age 18–64, percent of ized least square (GLS) estimator (provided un- males age 18–64 working in agriculture, percent der the IVprobit command in STATA packages) of males age 18–64 working in the public sector, in such occasions. Nevertheless, later on, Angrist percent of males age 18–64 working in private (1991) provided certain conditions under which wage work, and the percent of males age 18–64 a two-stage linear model (2SLS) can perform with secondary or higher education. well with binary endogenous variables models (Acosta, 2006). Moreover, for each specification of the bivari- ate and the ivprobit (corrected) models, we test In this analysis, as a robustness check, we the exogeneity of migration (remittances) to estimate a simple one equation Probit, a 2SLS household poverty. The null hypothesis here is and a GLS models. We also estimate a bivariate that the correlation between the error terms of Probit (two equation Probit) model using the the poverty and migration (remittances) equa- biprobit command in STATA but implement it tions, rho, is zero. If we cannot reject this null as an IV estimation. This specification allows us hypothesis, than we cannot reject that migration to account for the binary nature of poverty and is exogenous to household poverty, that is, migra- migration and, at the same time, deal with self- tion (remittances) is uncorrelated to the error selection and endogeneity of migration (remit- tances) by allowing the error terms in both the 12 We also investigated other specifications and found that similar poverty and migration (remittances) equations results are obtained for the poverty equation, when using any to be correlated. of those specifications. 10 Keller MNA 5-27-10vol2.indd 10 5/27/10 2:41 PM Appendix 2: Description of Migration/Remittances and Poverty Analysis in Egypt: Methodology and Results term of the poverty equation. In such case, the ent the 2SLS results. The GLS estimates are not results of the single equation Probit model would reported in the tables as they yield similar results be more efficient than those of the bivariate Pro- to that of the biprobit model. At the bottom of bit model. On the other hand, if the error terms the tables, the goodness of fit measures, the p- are strongly correlated (i.e., we cannot reject value of Sargen’s test for over-identification of that the unobservables that affect the poverty the instruments and the statistics of the weak status also influence the decision to migrate), instrument test are reported. we expect the size of coefficient of the migration (remittances) variable to be substantially larger The first stage results of the migration (re- in the corrected models than in the uncorrected mittances) equation closely resemble the results single equation model. presented in column 2 (4) of Table 10. Both instrumental variables are individually strongly Additionally, the 2SLS estimation allows us significant (at 1 percent level of significant). to perform both an over-identification test and Also, based on all the 2SLS models specifications a weak instruments test. The Sargen’s test for of both the migration and remittances effects on over-identification of the instrumental variables poverty, Sargen’s test for over-identification of tests the null hypothesis that both instruments the instruments does not reject the null hypoth- are valid; i.e. could be excluded from the poverty esis that both instruments are valid (p-values equation. A statistically significant Sargen’s test are substantially higher than the 10 percent statistic indicates that the instruments may not level). Additionally, the weak identification be valid. On the other hand, a test based on the test provides an F-statistic that is substantially Cragg Donald minimum eigenvalue statistic cre- higher than the threshold rule of thumb of 10. ated by Cragg and Donald (1993) can be used to All the R2 statistics of the first-stage regression test for the weakness of instruments. The value are also relatively high, so they do not imply a of this statistic is compared to critical values weak-instrument problem. Hence, we can reject provided by Stock and Yogo (2005). It provides the null hypothesis that our two instruments measures of goodness of fit of the first-stage are weak. equation (migration and remittances). It also uses an F-statistic to test the null hypothesis that On the other hand, for each of the biprobit the coefficients on the instruments are equal zero specifications in Table A2 and A3, the value of in the first-stage equation. The F-statistic is of- the correlation between the error terms of the ten compared, in the literature, to the threshold poverty and migration (remittances) equations, of 10, which is suggested by Staiger and Stock rho, and its significance level are reported. In (1997). An F-statistic below the threshold of the remittances and migration analysis, both the 10 suggests the existence of a weak-instrument biprobit and ivprobit (GLS) model specifications problem. lead to a p-value larger than 0.1 for the Wald-test of significance of rho (except for the biprobit Results specifications in the migration table the p-value is 0.07).13 Hence, we cannot reject the null-hypoth- Table A2 and A3 present the regression results eses that rho=0 at 5 percent significance level. of the effect of migration and receiving remit- In other words, we cannot reject that the error tances on household poverty. Once again both term of the migration (remittances) equation is tables report the marginal effects and the Huber- uncorrelated to the error term of the poverty White standard errors. Column 2, 4, and 6 add equation. Accordingly, in this case we expect the rural and education interaction terms to the coefficient results of both the corrected and investigate whether those interactions have ad- uncorrected models to be considerably close. ditional significant effects on household poverty. Column 1 and 2 present the uncorrected single equation Probit results, column 3 and 4 present 13 The results of the ivprobit (GLS) estimation lead to p-values the biprobit results, while column 5 and 6 pres- over 0.8 in all models. 11 Keller MNA 5-27-10vol2.indd 11 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Table A2: The impact of Egyptian migration on poverty status of the household (1) (2) (3) (4) (5) (6) Variables Probit Probit Biprobit Biprobit 2SLS 2SLS International migrant in HH –0.075*** –0.077*** –0.091*** –0.087*** –1.769*** 1.554** (0.011) (0.011) (0.010) (0.009) (0.568) (0.637) Community controls % unemployed males age 18–64 –0.136* –0.138* –0.135* –0.130* –0.170 –0.131 (0.075) (0.075) (0.075) (0.071) (0.132) (0.080) % males age 18–64 working in agriculture 0.195*** 0.195*** 0.200*** 0.191*** 0.402*** 0.342*** (0.024) (0.024) (0.025) (0.024) (0.044) (0.041) % males age 18–64 working in public sector 0.024 0.024 0.021 0.020 –0.017 0.027 (0.030) (0.030) (0.030) (0.029) (0.042) (0.035) % males age 18–64 working in private wage 0.121*** 0.121*** 0.121*** 0.115*** 0.114*** 0.133*** work (0.026) (0.026) (0.026) (0.025) (0.037) (0.031) % males age 18–64 with secondary + education –0.002 –0.002 –0.000 –0.000 0.097** 0.002 (0.021) (0.021) (0.022) (0.021) (0.038) (0.028) Household characteristics No Children 0–5 –0.002 –0.002 –0.002 –0.002 –0.002 –0.009 (0.004) (0.004) (0.004) (0.004) (0.006) (0.006) No Children 6_14 0.004 0.004 0.005 0.004 0.014*** 0.006 (0.003) (0.003) (0.003) (0.003) (0.005) (0.005) No. Males 15–29 –0.005 –0.005 –0.004 –0.004 0.002 –0.014*** (0.003) (0.003) (0.003) (0.003) (0.007) (0.006) No. Females 15–29 –0.010** –0.010** –0.009** –0.009** –0.016** –0.011* (0.004) (0.004) (0.004) (0.004) (0.007) (0.006) No. Elderly 64+ –0.004 –0.004 –0.004 –0.004 –0.013 –0.013 (0.007) (0.007) (0.007) (0.007) (0.014) (0.012) Avg. Male 18+ Years of schooling –0.007*** –0.007*** –0.007*** –0.007*** –0.017*** –0.010*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.001) Avg. Female 18+ Years of schooling –0.009*** –0.009*** –0.009*** –0.008*** –0.012*** –0.015*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) Alexandria and Suez(d)2 0.002 0.003 0.002 0.003 0.005 –0.008 (0.017) (0.017) (0.017) (0.016) (0.015) (0.012) Urban Lower Egypt(d)2 0.070*** 0.070*** 0.072*** 0.069*** 0.071*** 0.004 (0.020) (0.020) (0.020) (0.019) (0.023) (0.015) Urban Upper Egypt(d)2 0.179*** 0.178*** 0.179*** 0.172*** 0.118*** 0.076*** (0.024) (0.024) (0.024) (0.023) (0.018) (0.016) Rural Lower Egypt(d)2 0.105*** 0.105*** 0.106*** 0.101*** –0.007 0.081*** (0.019) (0.019) (0.019) (0.019) (0.022) (0.021) Rural Lower Egypt(d)2 0.248*** 0.247*** 0.250*** 0.241*** 0.164*** 0.252*** (0.027) (0.027) (0.027) (0.027) (0.025) (0.024) Rural x Migrant HH 0.031 0.036 0.038 0.042 1.562*** –1.362*** (0.049) (0.050) (0.050) (0.049) (0.564) (0.489) Household substitute head characteristics Age –0.003** –0.003** –0.003** –0.003** –0.003 –0.006*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) (continued on next page) 12 Keller MNA 5-27-10vol2.indd 12 5/27/10 2:41 PM Appendix 2: Description of Migration/Remittances and Poverty Analysis in Egypt: Methodology and Results Table A2: The impact of Egyptian migration on poverty status of the household (continued) (1) (2) (3) (4) (5) (6) Variables Probit Probit Biprobit Biprobit 2SLS 2SLS Age square 0.000** 0.000** 0.000** 0.000** 0.000 0.000*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Married(d)1 –0.110*** –0.108*** –0.112*** –0.106*** –0.080** –0.071** (0.032) (0.032) (0.032) (0.031) (0.040) (0.034) Divorced or Widowed(d)1 –0.050*** –0.049*** –0.053*** –0.050*** –0.083** –0.045 (0.015) (0.015) (0.015) (0.014) (0.042) (0.038) Primary or preparatory degree(d)3 –0.005 –0.007 –0.005 –0.007 –0.011 0.028 (0.012) (0.012) (0.012) (0.011) (0.020) (0.027) Secondary degree(d)3 –0.023* –0.023* –0.022* –0.021* 0.017 0.051* (0.013) (0.013) (0.013) (0.013) (0.023) (0.030) Above secondary degree(d)3 –0.097*** –0.097*** –0.096*** –0.091*** 0.050* 0.066** (0.008) (0.008) (0.009) (0.008) (0.030) (0.033) Primary or Prep degree x Migrant HH 0.059 0.061 –1.012** (0.077) (0.073) (0.434) Secondary degree x Migrant HH4 –0.006 0.009 –0.950** (0.041) (0.045) (0.378) Above secondary degree x Migrant HH –0.085*** –1.404** (0.006) (0.559) Observations 8338 8338 8338 8338 8338 8338 Pseudo R-squared 0.317 0.317 rho 0.236* 0.245* Wald-test 0f rho=0 (p-value) 0.067 0.069 Sargen’s test of over-identification (p-value) 0.991 0.227 Test of weak Instruments 21.637*** min eigenvalue statistic 11.820*** 0.7749 R2 0.5721 0.7741 Adjusted R2 0.5707 16.555*** F-test 13.919*** Notes: Marginal effects are reported and robust standard errors in parentheses (d) for discrete change of dummy variable from 0 to 1 *** p<0.01, ** p<0.05, * p<0.1 1 reference category: never married 2 reference category: Greater Cairo Region 3 reference category: no educational certificate 4 The interaction term “Above secondary degree x Migrant HHâ€? predicted failure perfectly in 64 observations in the probit estimation. Hence, to avoid STATA dropping those cases, the interaction of secondary degree and above secondary degree has been combined in the probit specification of column (2). As shown in Table A2, the coefficient of inter- points in the corrected models (column 3 and est, the effects of migration on household poverty 4). Similarly, both the corrected and uncorrected in both the corrected and uncorrected Probit models specification (Table 12) show that receiv- models are very close. Migration significantly ing remittances has the same effect on reducing decreases the likelihood of household poverty poverty (around eight percentage points) as that by about 8 percentage points in the uncorrected of migration. Hence, this analysis suggests that, in models (column 1 and 2) and by 9 percent Egypt, migration and receiving remittances have 13 Keller MNA 5-27-10vol2.indd 13 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Table A3: The impact of remittances on poverty status of the Egyptian household (1) (2) (3) (4) (5) (6) Variables Probit Probit Biprobit Biprobit 2SLS 2SLS Household receive remittances –0.083*** –0.086*** –0.088*** –0.088*** –2.071*** 2.318** (0.008) (0.008) (0.008) (0.007) (0.672) (0.914) Community controls % unemployed males age 18–64 –0.134* –0.137* –0.134* –0.132* –0.209** –0.114 (0.075) (0.076) (0.075) (0.073) (0.085) (0.084) % males age 18–64 working in agricul- 0.194*** 0.197*** 0.196*** 0.192*** 0.411*** 0.329*** ture (0.024) (0.025) (0.025) (0.024) (0.045) (0.043) % males age 18–64 working in public 0.031 0.032 0.030 0.029 0.031 0.022 sector (0.030) (0.030) (0.030) (0.029) (0.039) (0.035) % males age 18–64 working in private 0.126*** 0.128*** 0.126*** 0.123*** 0.160*** 0.121*** wage work (0.026) (0.027) (0.026) (0.026) (0.036) (0.033) % males age 18–64 with secondary + –0.003 –0.003 –0.002 –0.002 0.093** –0.012 education (0.021) (0.022) (0.021) (0.021) (0.038) (0.031) Household characteristics No Children 0–5 –0.003 –0.003 –0.003 –0.002 –0.004 –0.008 (0.004) (0.004) (0.004) (0.004) (0.006) (0.006) No Children 6_14 0.005 0.005 0.005 0.005 0.013** 0.006 (0.003) (0.003) (0.003) (0.003) (0.005) (0.006) No. Males 15–29 –0.005 –0.005 –0.005 –0.005 –0.004 –0.012** (0.003) (0.003) (0.003) (0.003) (0.006) (0.005) No. Females 15–29 –0.010** –0.010** –0.010** –0.010** –0.016** –0.013** (0.004) (0.004) (0.004) (0.004) (0.006) (0.006) No. Elderly 64+ –0.004 –0.004 –0.004 –0.004 –0.024* –0.015 (0.007) (0.007) (0.007) (0.007) (0.012) (0.012) Avg Male 18+ Years of schooling –0.007*** –0.007*** –0.007*** –0.007*** –0.018*** –0.009*** (0.001) (0.001) (0.001) (0.001) (0.003) (0.001) Avg Female 18+ Years of schooling –0.009*** –0.009*** –0.009*** –0.008*** –0.010*** –0.015*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) Alexandria and Suez(d)1 0.004 0.004 0.003 0.003 0.010 –0.017 (0.017) (0.017) (0.017) (0.017) (0.014) (0.014) Urban Lower Egypt(d)1 0.070*** 0.071*** 0.071*** 0.069*** 0.073*** –0.003 (0.020) (0.020) (0.020) (0.019) (0.022) (0.017) Urban Upper Egypt(d)1 0.177*** 0.179*** 0.177*** 0.174*** 0.107*** 0.082*** (0.024) (0.024) (0.024) (0.024) (0.016) (0.015) Rural Lower Egypt(d)1 0.104*** 0.105*** 0.104*** 0.102*** 0.002 0.070*** (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) Rural Lower Egypt(d)1 0.245*** 0.247*** 0.245*** 0.241*** 0.173*** 0.242*** (0.027) (0.027) (0.027) (0.027) (0.022) (0.023) Rural x HH receive remittances 0.107 0.117 0.113 0.124 1.849*** –1.956*** (0.096) (0.100) (0.097) (0.098) (0.664) (0.713) (continued on next page) 14 Keller MNA 5-27-10vol2.indd 14 5/27/10 2:41 PM Appendix 2: Description of Migration/Remittances and Poverty Analysis in Egypt: Methodology and Results Table A3: The impact of remittances on poverty status of the Egyptian household (continued) (1) (2) (3) (4) (5) (6) Variables Probit Probit Biprobit Biprobit 2SLS 2SLS Household substitute head characteristics Age –0.003** –0.003** –0.003** –0.003** –0.003 –0.006*** (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) Age square 0.000** 0.000** 0.000** 0.000** 0.000* 0.000*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Married(d)2 –0.112*** –0.113*** –0.113*** –0.111*** –0.085** –0.055 (0.032) (0.032) (0.032) (0.032) (0.039) (0.038) Divorced or Widowed(d)2 –0.050*** –0.051*** –0.052*** –0.051*** –0.102** –0.026 (0.015) (0.015) (0.015) (0.014) (0.043) (0.042) Primary or preparatory degree(d) 3 –0.007 –0.008 –0.008 –0.008 –0.014 0.025 (0.011) (0.012) (0.011) (0.011) (0.019) (0.026) Secondary degree(d)3 –0.024* –0.026** –0.024* –0.025** 0.009 0.048 (0.013) (0.013) (0.013) (0.012) (0.021) (0.030) Above secondary degree(d)3 –0.097*** –0.098*** –0.097*** –0.095*** 0.025 0.062* (0.008) (0.008) (0.008) (0.008) (0.024) (0.033) Primary/Prep degree x HH receive remit- 0.040 0.044 –1.757** tances (0.090) (0.089) (0.709) Secondary degree x HH receive remit- 0.044 0.065 –1.441*** tances4 (0.064) (0.074) (0.559) Above secondary degree x HH receive –0.084*** –2.144*** remittances (0.005) (0.820) Observations 8338 8303 8338 8338 8338 8338 Pseudo R-squared 0.316 0.315 rho 0.108 0.178 Wald-test 0f rho=0 (p-value) 0.510 0.293 Sargen’s test of over-identification 0.4739 0.3164 (p-value) Test of weak instruments 18.209*** min eigenvalue statistic 13.838*** 0.7903 R2 0.5730 0.7896 Adjusted R2 0.5716 10.964*** F-test 14.381*** Notes: Marginal effects are reported and robust standard errors in parentheses (d) for discrete change of dummy variable from 0 to 1 *** p<0.01, ** p<0.05, * p<0.1 1 reference category: Greater Cairo Region 2 reference category: never married 3 reference category: no educational certificate 4 The interaction term “Above secondary degree x Migrant HHâ€? predicted failure perfectly in 64 observations in the probit estimation. Hence, to avoid STATA dropping those cases, the interaction of secondary degree and above secondary degree has been combined in the probit specification of column (2). 15 Keller MNA 5-27-10vol2.indd 15 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options fairly moderate effects on household poverty. secondary education substitute head decreases It is worth mentioning here that similar results the likelihood of falling in poverty by 8.5 (8.4) have been highlighted in a recent study within percentage points. the MENA region. In Morocco, Sasin (2008) finds that migration (as proxied by remittances In contrast, the rural and the education receipt) decreases the likelihood of poverty by interactions are strongly significant in the 2SLS about 7 percentage points. specifications. However, in Table A2, the migra- tion indicator and rural interaction term have The effects of the community-level variables opposite signs in each of the 2SLS specifica- and the household and household substitute tions. In model 5, the migration coefficient is head characteristics on poverty are very similar negative while the rural interaction term has an in all models of Table A2 and A3. In all model adverse effect on migration. Hence, this model specifications, the household neighborhood labor shows that being a migrant household in a ru- market structure has an interesting impact on ral area decreases the likelihood of poverty by the household’s own poverty status. Households 1.849–2.071= 0.222 percent. On the other hand, residing in neighborhoods with high percent- including the interaction term of the household age of agriculture and private wage work are head education in the model (column 6) causes more likely to be poor then their counterparts. the migration and the rural interaction coef- However, the neighborhood unemployment level ficients to switch signs. Nevertheless, column of adult males has a negative impact on house- 6 shows that being a migrant household in a hold poverty status. This might be due to the rural area with a substitute household head with relatively higher reservation wage of adult males some primary education decreases poverty by living in rich households/neighborhoods, since at least 1.4 percent (2.318–1.9560–1.7570), by they can afford to stay unemployed longer than 1.1 percent if the head has a secondary educa- their poor counterparts. On the household own tion, and by 1.8 percent if the household head characteristics front, the number of females age has above secondary education. Similar results 15–29 and the average years of schooling of both are observed in Table 12 for remittances effects. males and females above age 18 significantly but weakly (by less than 2 percent points) decrease To sum up, there is weak evidence (based the likelihood of falling in poverty. Also, as ex- on only the 2SLS estimation) that the impact of pected, poverty is significantly higher among migration and remittances is urban/rural specific. households residing in Lower and Upper Egypt However, there is stronger evidence that poverty in comparison to Greater Cairo. Poverty declines alleviation impact of migration and remittances with the household substitute head age. Also, increases with the household education status. poverty is substantially lower among households with heads who are married and has secondary In an additional set of regressions (not pre- or above education. sented here), we were also interested in investi- gating the effect of the migrant’s own character- The rural interaction terms do not show any istics (specifically, education status, country and additional significant effect in all the Probit and occupation) on the poverty alleviation potential biprobit estimations of Table A2 and A3. In other of remittances. We performed the same analysis words, there is no evidence that the impact of presented in Table A2 and A3, but we had to limit migration or remittances on poverty alleviation the sample to those households with at least one differs based on the household rural or urban current migrant member. This left us with only 362 residence. Also, the interaction between the households of which only 47 households were poor household substitute head and migration (remit- and 238 households were receiving remittances. tances) has some significant effects only in the The results did not show any additional significant biprobit full specification model. Column 4 in effects of the migrant’s own characteristics. We Table 11 (Table A3), shows that, being a migrant suspect that this might be mainly due to the small (remittances recipient) household with an above sample size of current migrants. 16 Keller MNA 5-27-10vol2.indd 16 5/27/10 2:41 PM Appendix 2: Description of Migration/Remittances and Poverty Analysis in Egypt: Methodology and Results As mentioned above, beside instrumental return migrants is that the 1998 round contains variables estimation techniques, another often no information on current migrants (see Section used remedy to correct for both the potential 5). For robustness checking, we also estimate endogeneity and the selection bias problem of both a linear fixed effects (fe) model and a lin- migration and remitting is to use panel data to ear random effects model (re) (using the xtreg control for time-invariant unobserved hetero- command in STATA). In this linear specification, geneity among households. In the following we use the value of the wealth index instead of we exploit the panel nature of the ELMPS to the dummy variable of whether the household further investigate the effect of migration and is poor (column 2 and 3). The same analysis is remittances on poverty. In Table A4, we estimate conducted in columns 4–6 to investigate the a one-equation random effect Probit model of effect of remittances on household poverty and the poverty status of the household (using the wealth. The table shows that, once again, in all xtprobit command in STATA) where the main model specifications (except for the migration explanatory variable is whether the household fixed effect specification) migration and remit- had at least one return migrant (column 1) in the tances significantly decrease the likelihood of last five years. The reason for only focusing on falling in poverty. Table A4: Panel analysis of the impact of migration and remittances on Egyptian household poverty and wealth (1) (2) (3) (4) (5) (6) Variable Probit re Linear re Linear fe Probit re Linear re Linear fe Return migrant in the last 5~a –0.817*** 0.122** 0.033 HH receive remittances –0.352* 0.205*** 0.144** Constant –2.074*** 0.016 0.018*** –2.080*** 0.012 0.014** Observations 7368 7368 7368 7368 7368 7368 R-squared within model 7.8e–05 7.8e–05 .00186 .00186 R-squared overall model .0023 .0023 .00339 .00339 R-squared between model .00483 .00483 .00451 .00451 17 Keller MNA 5-27-10vol2.indd 17 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Keller MNA 5-27-10vol2.indd 18 5/27/10 2:41 PM Appendix 3: Description of Remittances and Poverty Analysis in Morocco: Methodology and Results (The following description is taken from Sasin, 2008) This remittance/poverty analysis in Morocco consumption and on poverty, regression ap- makes use of the 2000/01 ENCDM survey, which proach and instrumental variables techniques is a multipurpose household survey, covering are used to correct for the fact that remittances 0.3 percent of the population. The module on are (i) not the only outcome of migration, and household transfers contains a question about (ii) endogenous. Notably, there is income lost the source of transfers, for which “family living because of the absence of a migrant, which may abroadâ€? is an option. More information on how not necessarily be compensated for by the money migration and remittances variables have been sent home plus lower household consumption derived can be found in the footnote.14 The needs. In such a case, migration can plausibly be poverty variable is derived using the consump- poverty-increasing. The regression framework tion aggregate and the poverty lines computed that tries to explain the consumption level and officially by the statistical office. poverty status of Moroccan households by their characteristics, including the migration status How accurate is the ENCDM? For many (proxied by remittance receipt) takes care about purposes it is relatively good. For instance, total the lost income and, additionally, controls for household consumption recorded in the survey other factors, such as health, gender or family is 253 billion dirham, compared with 299 bil- situation. Additionally, as the ordinary regres- lion dirham recorded by the National Account sion framework assumes that some households Statistics for the period of the survey. However, just happen to migrate or receive remittances remittances in the survey add up to only 2.3 per- completely exogenously, just like manna from cent of total consumption (or 2.8 percent, when including foreign pension) against the Balance of Payment figure of 12 percent. Even assuming 14 The survey allows distinguishing between “private domesticâ€?, that not all private transfers from abroad “passâ€? “private foreignâ€? (“theâ€? remittances), “public domesticâ€?, “public foreignâ€? and “other sourcesâ€? transfers. In addition, there is a through households budgets (e.g., some can be sizable class of “exceptional event transfersâ€? (accounting for 30 directly invested in real estate or stock exchange percent of all transfers) for which the source cannot be deter- and some may be money of migrants currently mined. These transfer have been distributed between “private domesticâ€? and “private foreignâ€? classes using multinomial logit residing abroad, who are not included in the technique (i.e., according to the probability that a household sample) it seems that remittances in the survey receive domestic or foreign remittances or both). As a result, are underestimated. 30 percent of these unallocated transfers have been allocated to the “private foreignâ€? class, the share of people in foreign re- mittance receiving households increased from 10.6% to 11.3%, To measure the impact of migration (as while those in domestic remittance receiving households – from proxied by remittance receipt) on the level of 33.9% to 42.3%. 19 Keller MNA 5-27-10vol2.indd 19 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options heaven, while, as discussed above, migration share approximates local migration networks is “endogenous,â€? we use instrumental variable and we argue that—migration being chain phe- techniques, to correct for the possible bias. nomenon—these networks help people migrate, so they are correlated with household migration As the instrument we have used the share of status in 2001. On the other hand, we assume return migrants in the population at the province that these networks do not impact individual level in the 1994 Census. We believe that this level living standards seven years later. Table A5: Impact of migration on Moroccan living standards, regression results Uncorrected results Instrumental variable Log consum (OLS) Poverty (probit) Log cons (OLS) Poverty (probit) Dependent variable dF/dx t-stat dF/dx z-stat dF/dx t-stat dF/dx z-stat Gender (female) 0.054 9.8 –0.012 –3.7 0.053 6.5 –0.013 –2.8 Age 0.012 18.8 –0.004 –10.3 0.012 17.9 –0.004 –10.2 Age-squared 0.000 –9.7 0.000 5.8 0.000 –9.7 0.000 6.7 Urban 0.395 65.6 –0.119 –33.4 0.394 30.2 –0.103 –13.0 Employed in agriculture –0.049 –6.3 0.015 3.2 –0.049 –5.8 0.015 3.5 Education: basic 1 cycle st 0.192 27.6 –0.078 –19.1 0.191 13.4 –0.056 –7.7 Education: basic 2 cycle nd 0.344 41.7 –0.111 –22.8 0.342 17.9 –0.076 –8.8 Education: secondary 0.485 48.8 –0.125 –21.4 0.483 18.5 –0.089 –8.5 Education: tertiary 0.583 46.3 –0.115 –15.4 0.582 22.6 –0.083 –7.9 Education: other 0.130 10.0 –0.039 –5.1 0.129 9.7 –0.028 –4.6 Unemployed –0.021 –2.1 –0.009 –1.5 –0.021 –2.1 0.005 0.9 Inactive 0.048 6.9 –0.028 –6.8 0.048 6.4 –0.017 –4.0 (Inverse) dependency ratio 0.381 31.8 –0.180 –25.4 0.383 11.6 –0.163 –8.2 Sickness in family 0.196 30.3 –0.050 –14.5 0.193 24.7 –0.049 –12.0 Senior position 0.558 48.5 –0.024 –3.5 0.558 37.2 –0.076 –8.3 Public sector employment 0.053 6.4 –0.014 –2.9 0.053 4.6 –0.019 –2.8 Migration 0.320 43.0 –0.088 –20.0 0.342 1.0 –0.096 –0.7 N= 58,512 58,512 58,512 58,512 R= 2 0.342 0.114 0.343 0.101 Source: Calculations from ENCDM (2001) Note: for dummy variables dF/dx is for discrete change from 0 to 1. Base category: no education, employed, male 20 Keller MNA 5-27-10vol2.indd 20 5/27/10 2:41 PM Appendix 4: Description of Migration/ Remittances and Labor Market/ Employment Analysis in Egypt: Methodology and Results (The following description is taken from Binzel and Assaad, 2009) Data of remittance income a household receives, i.e., the sum over the transfers made by all remitters We are using the Egypt Labor Market Panel whether or not any of these have been sent to any Survey of 2006 (ELMPS 06). The survey was ad- particular household member or not. We provide ministered to a nationally representative sample summary statistics below. of 8,349 households, of which 3,684 were among the original 4,816 households interviewed in the Model Egypt Labor Market Survey of 1998 (ELMS 98). An additional 2,167 new households emerged The literature has dealt in various ways with from these 3,684 households as a result of splits, the problem of self-selection, i.e. the fact that and a refresher sample of 2,498 households was migrant and non-migrant households signifi- added in 2006. cantly differ along unobservable characteristics (Funkhauser 1995), and endogeneity of migra- In our analysis, we focus on young men and tion and remittances, that is, the fact that the women aged 16–34 who had left school by the outcome of interest (e.g., labor force participa- time of the survey in 2005/2006. This leaves us tion) is likely to affect the migration decision. To with a working sample of 5,551 young women and give an example, both decisions may be part of a 5,467 young men. With “migrant householdsâ€? we wider household strategy where the household refer to households in our sample that list at least decides to send one member abroad in order one household member as working abroad at the to accelerate capital accumulation, while the time of the survey. Out of the 361 households non-migrant members use the capital to set up with an international migrant (4% of all house- a family business. holds in our sample), only 29 households mention a second and 5 mention a third household mem- In line with recent studies, we estimate the ber working abroad. By “remittance-receiving following model using an instrumental variable householdsâ€? we refer to households who receive approach: international remittances both monetary and in kind (which are also expressed in monetary Yi = α0 + α1 Ri + α2 Zi + ε i (structural equation) terms by the respondent). Average monthly remittance income refers to the total amount Zi = β1 + β 2 Ri + β 3 IVi + δi (reduced-form for Z) 21 Keller MNA 5-27-10vol2.indd 21 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options for all i =1, 2,...n individuals. stands for the out- share of migrant households at the governorate come variable (e.g., labor force participation) level for clusters with fewer than 10 households which is assumed to depend on a set of individual in the sample. We interact our instrument with characteristics (R) and (Z) which stands for the number of adults in the household and with (i) whether the individual is living in a migrant the average years of schooling of all adults in household, (ii) whether the individual is living in the household. We use the same instruments for a remittance-receiving household, or (iii) average remittances since migration typically precedes monthly remittance income (in 100 LE).15 the sending of remittances. In fact, as shown in Table A7, very few households in our sample Our outcome variables are participation in receive remittances but do not list having any the labor market and hours worked so we esti- current household member working abroad. mate our main equation as a Probit and a Tobit Note, however, that a number of households with model, respectively. For cases (i) and (ii), Z is a migrant at the time of the survey do not receive a dichotomous variable so we estimate the IV any remittances. equation as a Probit model. For the IV Probit es- timation, we use the biprobit command in STATA In many of our estimated models, it turns out but implement it as an IV estimation. For the IV that the correlation between the two equations is Tobit estimation, we use the user-written STATA insignificant, i.e. that migration (remittances) is program “cmpâ€? (Roodman 2007). For case (iii), exogenous to the outcome of interest. Therefore, we can draw on the ivprobit and ivtobit com- we also present results using the same model mands in STATA since Z is a continuous variable. specification but without instrumenting for the Z variable. To deal with self-selection and endogeneity (i.e. that Z and e may be correlated), we instru- Our outcome variable is whether the indi- ment for the migration (remittances) variable and vidual was engaged in any specific work in the allow the error terms in both equations (e and reference period (i.e., referring to the main job d) to be correlated. In this version of the paper in the three months prior to the interview) and, we were unable to get external data to estimate respectively, how many hours he or she spent on this instrument, so we use internal cluster-level this activity in the past 7 days. We differentiate estimates from the ELMPS 06. As the instrument between several employment states: labor force we use the percentage of households with at least participation, unemployment, employment in one current international migrant at the cluster wage and salary work, being self-employed or level, excluding the household itself. Since the an employer17 and working unpaid for the family. ELMPS 2006 includes also split households from While unemployment relates only to individuals households surveyed in 1998, some districts con- in the labor force, the employment categories, sist of very few households. Therefore, we use the such as wage and salary work, relate to the entire population. Labor force participation refers to participation in any market work (i.e., excluding Table A7: Households receiving international subsistence work) or being unemployed, using remittances and/or households with at least the standard definition of unemployment (active search criterion). For women, we additionally one current international migrant. consider domestic and subsistence work. Here, Households receiving remittances 15 The exchange rate between LE and US $ was 5.5 LE/$ at the No Yes Total end of October 2008. 16 In the next version of the paper we will be using the share Households with at no 7,955 35 7,990 of the adult population of the village or neighborhood who are least one current migrants as measured by the 2006 Population Census. yes 121 240 361 international migrant 17 We have lumped together these two categories – employer and self-employed because of the very limited number of employers Total 8,076 275 8,351 among young workers. 22 Keller MNA 5-27-10vol2.indd 22 5/27/10 2:41 PM Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results the reference period are the seven days prior to variable to the outcome variable. We report the the interview.18 With domestic work being nearly p-value of either the Wald test of exogeneity in universal among young women aged 16–34 who the ivprobit and ivtobit commands, or of the sig- have left school (97%), we only estimate the ef- nificance of the correlation of the disturbances fects on the time spent on domestic chores, but of the two equations shown in section 4.2 (rho not on the participation decision. in the biprobit command and althanhrho in the cmp command). See Wooldridge (2002: In the regressions, we include individual 472–478) for a discussion. The null hypothesis characteristics (age, marital status and educa- is that the correlation is zero, meaning that tional attainment) and household-level charac- migration (remittance-income) is uncorrelated teristics (number of females in the household, with the error term in the employment equation, number of children aged 0–5, number of chil- i.e. is exogenous to our outcome of interest.19 dren aged 6–14 and presence of elderly in the Hence, we need a p-value smaller than 0.1 (10% household) as exogenous explanatory variables. significance level) or 0.05 (5% significance level) Furthermore, we control for regional differences in order to reject the null. If we cannot reject in the labor market. As one would expect for the null, there is no need for an IV estimation Egypt, far more women than men aged 16–34 and estimates of a Probit or Tobit model will be are married (Binzel and Assaad 2008). Moreover, more efficient. Since this happens quite often, more women are left without any educational we always report estimates with and without certificate and fewer women have a general or instrumenting for Z. Note that if the two error technical secondary degree while the number of terms are strongly correlated, the size of the above secondary graduates is roughly the same migration (remittance income) effect in the IV for young men and women. estimation may be far larger compared to the coefficient we receive without instrumenting for Estimation Results migration (remittance income) and it may be of opposite sign, depending also on whether they We first present estimation results for labor force are positive or negatively correlated. participation and employment in different work types separately for young men and women The Impact of Migration and (Section 5.1) and then discuss the effect migra- Remittances on Employment tion and remittance-income has on the number of hours they work (Section 5.2). The impact International Migration of other explanatory variables such as age and education on employment behavior should be, Throughout this section, we refer to the margin- and in fact is, roughly the same throughout the al effects. Looking at the impact of migration on different models. We therefore only discuss them female employment, we can first of all note that once when looking at the impact of migration. only the decision to enter the labor force and to work unpaid for the family are correlated with On a technical note, regression results of the the (household) decision to migrate. For these first step, i.e., the IV equation, are reported in Table A9 in the Appendix. We also report test statistics and p-values of a Wald test on the joint 18 Based on the ELMPS 06 questionnaire, subsistence work significance of our three instruments in our IV includes agricultural activities, raising poultry/livestock, and equations (bottom of Table A9). Our instru- producing ghee/butter/cheese for domestic consumption. Do- mestic work includes time spent on cooking, washing dishes, ments are highly significant with respect to each doing laundry and ironing, cleaning the house, collecting water, instrumented variable, i.e., living in a migration collecting firewood or other fuel, shopping for food, clothing, and household, living in a remittance-receiving household items, helping in caring for the sick or the elderly, and taking care of children. household and average monthly remittance 19 Note, however, that we can only conclude that Z is not endog- income (in 100 LE). Moreover, in all our estima- enous if our instruments themselves are exogenous (Wooldridge tions we test for exogeneity of the instrumented 2002: 472–478). 23 Keller MNA 5-27-10vol2.indd 23 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options two models, we therefore refer to the biprobit With labor force participation being very estimation results while for all other models, we high among men, young men’s employment deci- refer to the Probit estimation (see Table A8). sions seem to be independent from the migration In general, the pattern we found earlier in the decision as the correlation of the error terms is descriptive analysis is confirmed by our regres- insignificant throughout all specifications (see sion results albeit not all differences between Table A26). Looking at the results of the Probit women living in migrant households and not are specifications, we find that men living in a mi- also significant. Migration significantly increases grant household have a significantly lower prob- the probability of a young woman to enter the ability—albeit small in size—to work for a wage labor market by 37% albeit the probability and a higher probability to work self-employed of being employed in wage and salary work / as employer. Overall, employment types seem significantly decreases for women living in a to be mainly affected by regional labor market migrant household. Instead, living in a migrant patterns. In less metropolitan areas, men’s prob- household significantly increases the probability ability to engage in wage work is lower while the of women to engage in subsistence work and probability of being self-employed or working especially in unpaid family work. For instance, unpaid for the family is higher compared to women living in a migrant household have a 23% their counterparts living in the Greater Cairo higher probability of ending up in unpaid family Region. Our estimations also confirm findings work compared to their counterparts. Together from earlier studies that not only women but also with the fact that for unpaid family work there men with above secondary degree have a higher is sizeable correlation between the error terms probability of finding themselves unemployed of the two equations, it may suggest that men than men with less educational attainment (As- who migrated worked in a family business and saad 2007, Amer 2007). As mentioned above, that young women help replace the labor of the Assaad (2007) attributes this to the decline in migrant. government sector jobs. Furthermore, married men are more likely to have work for wages and Educational attainment has a strong and less likely to work unpaid for the family. This significant positive effect on labor force par- is in line with earlier studies that show that ticipation and on being unemployed as well as young men are often forced to take up informal on having a wage and salary job. This reflects jobs when entering the labor market, and only the tendency of educated women in Egypt to manage to obtain better jobs after a few years work for the government / in the public sector, (Assaad 2007). By the time they marry, they since it is associated with short working hours are more likely to have a job that is perceived as and social security and other benefits that allow respectable by the potential bride and her family women to combine market work with domestic (Binzel and Assaad 2008). work including child care. Yet, job opportunities in the public sector have decreased over the last To sum up, migration does affect the alloca- decade without an increase in similar jobs in the tion of labor within the household but is gendered private sector, which has left many young and in that it mainly affects young women, at least, educated women unemployed (Assaad 2007). i.e., without considering hours worked. As expected, being married has a strong negative effect on the probability of young women to be International Remittance Income in the labor force and to have a wage and salary work.20 Regional variation in female employment Whereas migration as such has a negative effect is reflected in our results, too. Unpaid family and on households as it increases their dependency particularly subsistence work but also labor force ratio, remittance-income that may result from mi- participation in general, is more likely among women in Lower and Upper Egypt where wage work is less available and agriculture plays a 20 Only few women in our sample aged 16 to 34 are divorced greater role. or widowed. 24 Keller MNA 5-27-10vol2.indd 24 5/27/10 2:41 PM Table A8: The Impact of Migration on Female Employment (Young Women Aged 16–34, Out of School) Keller MNA 5-27-10vol2.indd 25 Results from an IV Probit estimation with endogenous binary variable (to control for migration) and a binary outcome variable reflecting a particular employment status, e.g., participating in the labor force, using the biprobit command in stata. The alternative specification is a probit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported with standard errors in parentheses. Empl. in wage & Self-employed / In the labor force Unemployed salary work employer Unpaid family work Subsistence work biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit Migrant hh3 (d) 0.373*** 0.035 0.438** 0.075 0.012 –0.035** 0.051 0.005 0.227** 0.018* 0.187* 0.049* (0.077) (0.033) (0.209) (0.053) (0.109) (0.014) (0.080) (0.006) (0.093) (0.011) (0.111) (0.027) Age 0.061*** 0.064*** 0.067** 0.069** –0.005 –0.004 0.001 0.001 0.007 0.007* –0.001 –0.000 (0.016) (0.016) (0.032) (0.031) (0.009) (0.009) (0.002) (0.002) (0.005) (0.004) (0.013) (0.012) Age squared –0.001*** –0.001*** –0.002** –0.002*** 0.000 0.000 –0.000 –0.000 –0.000 –0.000* 0.000 0.000 (0.000) (0.000) (0.001) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Primary or preparatory –0.058*** –0.058** 0.280*** 0.298*** 0.036* 0.036* –0.006*** –0.006*** –0.013*** –0.012*** –0.043*** –0.043*** degree1 (d) (0.022) (0.023) (0.106) (0.108) (0.021) (0.021) (0.002) (0.002) (0.004) (0.004) (0.014) (0.014) 1 Secondary degree (d) 0.129*** 0.137*** 0.635*** 0.651*** 0.112*** 0.113*** –0.007** –0.006** –0.037*** –0.035*** –0.102*** –0.100*** (0.018) (0.019) (0.054) (0.051) (0.015) (0.015) (0.003) (0.003) (0.006) (0.005) (0.012) (0.012) Above secondary de- 0.338*** 0.349*** 0.600*** 0.615*** 0.291*** 0.293*** –0.007*** –0.007*** –0.047*** –0.044*** –0.176*** –0.175*** gree1 (d) (0.024) (0.023) (0.061) (0.060) (0.026) (0.026) (0.002) (0.002) (0.005) (0.004) (0.012) (0.012) Ever-married (d) –0.281*** –0.282*** –0.102*** –0.092** –0.126*** –0.125*** 0.005 0.005 –0.006 –0.005 0.034** 0.035** (0.022) (0.022) (0.039) (0.038) (0.017) (0.017) (0.004) (0.004) (0.007) (0.006) (0.016) (0.016) Nr of females in hh 0.005 0.010 –0.029* –0.022 0.006 0.006 –0.006*** –0.006*** –0.000 0.001 0.014* 0.016** (0.009) (0.009) (0.015) (0.015) (0.005) (0.005) (0.002) (0.002) (0.002) (0.002) (0.007) (0.007) Nr of children aged 0–5 –0.018** –0.016* 0.020 0.025 –0.021*** –0.021*** –0.004*** –0.004*** 0.004* 0.004** 0.010 0.011* in hh (0.008) (0.008) (0.017) (0.016) (0.005) (0.005) (0.001) (0.001) (0.002) (0.002) (0.007) (0.006) Nr of children aged 6–14 0.030*** 0.030*** 0.009 0.012 –0.009* –0.009* 0.003** 0.002** 0.008*** 0.008*** 0.021*** 0.022*** in hh (0.008) (0.008) (0.014) (0.014) (0.005) (0.005) (0.001) (0.001) (0.002) (0.002) (0.006) (0.006) Presence of elderly in 0.022 0.032 0.045 0.057* –0.019* –0.017* –0.005* –0.005* 0.009 0.011 0.022 0.026 hh (d) (0.021) (0.021) (0.035) (0.034) (0.010) (0.010) (0.003) (0.003) (0.007) (0.007) (0.018) (0.018) (continued on next page) Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 25 5/27/10 2:41 PM 26 Keller MNA 5-27-10vol2.indd 26 Table A8: The Impact of Migration on Female Employment (Young Women Aged 16–34, Out of School) (continued) Results from an IV Probit estimation with endogenous binary variable (to control for migration) and a binary outcome variable reflecting a particular employment status, e.g., participating in the labor force, using the biprobit command in stata. The alternative specification is a probit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported with standard errors in parentheses. Empl. in wage & Self-employed / In the labor force Unemployed salary work employer Unpaid family work Subsistence work biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit 2 Alexandria and Suez (d) 0.045 0.047 0.043 0.041 0.008 0.008 –0.001 –0.001 0.011 0.011 0.132* 0.132* (0.029) (0.029) (0.051) (0.050) (0.015) (0.015) (0.006) (0.006) (0.024) (0.022) (0.071) (0.071) 2 Urban Lower Egypt (d) 0.092*** 0.104*** 0.198*** 0.215*** –0.016 –0.014 –0.001 –0.001 0.038 0.037 0.490*** 0.495*** (0.028) (0.028) (0.051) (0.050) (0.013) (0.012) (0.006) (0.006) (0.026) (0.025) (0.063) (0.063) Rural Lower Egypt2 (d) 0.070*** 0.081*** 0.135*** 0.148*** 0.004 0.006 0.001 0.001 0.050** 0.047** 0.673*** 0.677*** (0.024) (0.024) (0.045) (0.045) (0.013) (0.013) (0.005) (0.005) (0.021) (0.020) (0.044) (0.044) 2 Urban Upper Egypt (d) 0.074*** 0.080*** 0.042 0.040 –0.006 –0.006 0.012 0.012 0.113*** 0.109*** 0.478*** 0.480*** (0.027) (0.027) (0.045) (0.046) (0.012) (0.012) (0.008) (0.008) (0.037) (0.035) (0.062) (0.062) 2 Rural Upper Egypt (d) 0.154*** 0.178*** –0.039 –0.015 –0.029** –0.027** 0.032*** 0.033*** 0.135*** 0.138*** 0.626*** 0.634*** (0.027) (0.027) (0.048) (0.048) (0.014) (0.013) (0.011) (0.012) (0.034) (0.034) (0.049) (0.048) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5551 5551 1713 1713 5551 5551 5551 5551 5551 5551 5551 5551 rho=0: Prob > chi2 0.0000 0.1611 0.5722 0.2784 0.0002 0.1405 Note: Since there may be more than one young woman living in a household, we take clustering at the household level into consideration. (d) for discrete change of dummy variable from 0 to 1, * p<0.10, ** p<0.05, *** p<0.01 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. 5/27/10 2:41 PM Table A9: The Impact of Migration on Male Employment (Young Men Aged 16–34, Out of School) Keller MNA 5-27-10vol2.indd 27 Results from an IV Probit estimation with endogenous binary variable (to control for migration) and a binary outcome variable reflecting a particular employment status, e.g., participating in the labor force, using the biprobit command in stata. The alternative specification is a probit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported with standard errors in parentheses. Empl. wage & salary Self-employed / In the labor force Unemployed work employer Unpaid family work biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit Migrant hh3 (d) 0.022 –0.015 –0.020 0.001 –0.167 –0.111** 0.274 0.057* –0.028 0.035 (0.037) (0.020) (0.034) (0.015) (0.163) (0.045) (0.203) (0.033) (0.031) (0.023) Age –0.006 –0.006 –0.006 –0.006 0.030* 0.030* 0.069*** 0.069*** –0.007 –0.006 (0.009) (0.009) (0.006) (0.006) (0.016) (0.016) (0.012) (0.012) (0.007) (0.007) Age squared 0.000* 0.000* 0.000 0.000 –0.000 –0.000 –0.001*** –0.001*** 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Primary or preparatory degree1 0.019** 0.019** –0.006 –0.006 0.049** 0.049** –0.014 –0.014 0.001 0.001 (d) (0.009) (0.009) (0.012) (0.012) (0.023) (0.023) (0.014) (0.014) (0.011) (0.011) 1 Secondary degree (d) 0.039*** 0.040*** 0.058*** 0.057*** –0.004 –0.004 –0.029** –0.028** 0.017* 0.016* (0.008) (0.008) (0.013) (0.013) (0.021) (0.021) (0.012) (0.012) (0.009) (0.009) 1 Above secondary degree (d) 0.031*** 0.032*** 0.152*** 0.149*** 0.021 0.020 –0.087*** –0.085*** –0.012 –0.015 (0.009) (0.009) (0.027) (0.026) (0.024) (0.024) (0.011) (0.011) (0.011) (0.011) Ever-married (d) 0.091*** 0.090*** –0.067*** –0.066*** 0.087*** 0.089*** 0.089*** 0.086*** –0.036*** –0.033*** (0.010) (0.010) (0.009) (0.009) (0.021) (0.020) (0.014) (0.014) (0.010) (0.010) Nr of females in hh 0.001 0.001 0.009*** 0.009*** –0.019** –0.019** –0.020*** –0.019*** 0.023*** 0.022*** (0.004) (0.004) (0.004) (0.003) (0.009) (0.009) (0.006) (0.006) (0.004) (0.004) Nr of children aged 0–5 in hh –0.005 –0.005 –0.014*** –0.014*** –0.001 –0.001 0.001 0.002 –0.002 –0.003 (0.004) (0.004) (0.004) (0.004) (0.010) (0.010) (0.006) (0.006) (0.005) (0.005) Nr of children aged 6–14 in hh 0.002 0.002 –0.002 –0.002 –0.023** –0.023** –0.005 –0.005 0.018*** 0.017*** (0.004) (0.004) (0.004) (0.004) (0.010) (0.010) (0.007) (0.007) (0.004) (0.004) Presence of elderly in hh (d) –0.006 –0.005 –0.000 –0.001 –0.021 –0.022 –0.032** –0.031** 0.042*** 0.040*** (0.009) (0.009) (0.007) (0.007) (0.021) (0.021) (0.013) (0.013) (0.013) (0.013) (continued on next page) Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 27 5/27/10 2:41 PM 28 Keller MNA 5-27-10vol2.indd 28 Table A9: The Impact of Migration on Male Employment (Young Men Aged 16–34, Out of School) (continued) Results from an IV Probit estimation with endogenous binary variable (to control for migration) and a binary outcome variable reflecting a particular employment status, e.g., participating in the labor force, using the biprobit command in stata. The alternative specification is a probit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported with standard errors in parentheses. Empl. wage & salary Self-employed / In the labor force Unemployed work employer Unpaid family work biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit 2 Alexandria and Suez (d) –0.001 –0.001 0.014 0.014 –0.041 –0.041 0.045* 0.046* –0.029* –0.029* (0.014) (0.014) (0.012) (0.012) (0.031) (0.031) (0.025) (0.025) (0.017) (0.017) 2 Urban Lower Egypt (d) –0.012 –0.011 0.001 –0.000 –0.128*** –0.129*** 0.105*** 0.106*** 0.053* 0.050* (0.014) (0.014) (0.010) (0.010) (0.029) (0.029) (0.025) (0.026) (0.029) (0.028) Rural Lower Egypt2 (d) –0.007 –0.006 –0.008 –0.008 –0.135*** –0.136*** 0.067*** 0.069*** 0.116*** 0.112*** (0.012) (0.012) (0.008) (0.008) (0.026) (0.026) (0.020) (0.020) (0.028) (0.028) 2 Urban Upper Egypt (d) –0.021 –0.021 0.000 –0.000 –0.153*** –0.154*** 0.105*** 0.106*** 0.078*** 0.076*** (0.015) (0.015) (0.009) (0.009) (0.028) (0.028) (0.025) (0.025) (0.030) (0.029) 2 Rural Upper Egypt (d) –0.020 –0.018 –0.025*** –0.026*** –0.121*** –0.123*** 0.069*** 0.073*** 0.131*** 0.123*** (0.014) (0.013) (0.007) (0.007) (0.028) (0.027) (0.022) (0.022) (0.031) (0.030) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5467 5467 4905 4905 5467 5467 5467 5467 5467 5467 rho=0: Prob > chi2 0.4195 0.6428 0.7139 0.2082 0.1804 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. (d) for discrete change of dummy variable from 0 to 1, * p<0.10, ** p<0.05, *** p<0.01 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. 5/27/10 2:41 PM Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results gration should positively affect households as it The Impact of Migration and increases their income. Yet, whether households Remittances on Hours Worked use this extra cash in a productive way, and if so, how exactly, is disputable as our literature review In this section, we discuss results from estimating above has shown. IV Tobit estimations with endogenous variable and simple Tobit estimations on hours worked by For women aged 16–34, living in a remit- young men and, respectively, by young women. tance-receiving household significantly and Now, we generally refer to the expected value strongly affects market labor force participation conditional on that the individual engages in a but not the type of employment chosen (see specific type of work, e.g. wage work. Coefficients Table A10). Also unemployment is not signifi- are reported below. cantly affected. Note that, once again, only for labor force participation and for unpaid family International Migration work, living in a remittance-receiving household turns out to be endogenous. Table A18 shows As before, we first discuss the impact of migration the result when we instrument for the value on hours worked by young women. Estimation of remittances sent instead of whether or not results show that hours worked in the market or the household is receiving remittances. Results unpaid for the family as well as hours allocated to suggest that the average monthly amount of subsistence work are correlated with the migra- remittances sent has no significant impact on tion decision (see Table A12). When taking care young women’s time allocation. of endogeneity, it turns out that women living in a migrant household spend more time in market Coming back to our results in Table A10, it activities compared to their counterparts. This is surprising that a woman who lives in a house- seems to be mainly due to more hours spent hold that receives remittances has a 36% higher unpaid working for the family as time spent in probability to be in the labor force compared to wage and salary work is significantly lower while her counterparts. This suggests that remittances those self-employed do not change their behav- may be used in a productive way and not in order ior. Women living in a migrant household also to substitute work for leisure. Yet, none of the allocate more time to subsistence work while do- employment categories seem to be particularly mestic work remains unaffected suggesting that affected, so we cannot infer where the increase the entire workload of young women increases in the labor supply goes. in these households. Since women can combine domestic work with subsistence and unpaid Looking at the effects of remittance income family work relatively well, households seem to on male employment, we find that, similarly to spend remittance income elsewhere, but not on migration, the effect is negligible (see Tables A11 the consumption of durable goods and services and A19). Living in a remittance-receiving house- that may free women from domestic work and hold and, respectively, the value of remittances enable them to take up a wage work (in fact, in sent has only a statistically significant negative the previous section, we did not find any signifi- effect on being employed in wage and salary cant and positive impact on wage employment). work. Again, our results suggest that young men’s This could also be driven by the comparatively decisions to work are independent from nonlabor low wages women often earn. income sources—they all work anyways. Differ- ences in the descriptive statistics are thus likely Highly educated women work many more to be due to other, migration-related factors, hours (roughly 15 hours per week) in wage such as education and region of residence, that and salary work and less in the other types of are correlated with migration. We will see in the employment. Married women strongly reduce next section whether we find any impact on the the number of hours spent on market work and hours young men and women work in the differ- experience much higher levels of domestic work. ent activities. Children aged 0–5 in the household have a simi- 29 Keller MNA 5-27-10vol2.indd 29 5/27/10 2:41 PM 30 Keller MNA 5-27-10vol2.indd 30 Table A10: The Impact of Remittances on Female Employment (Young Women Aged 16–34, Out of School) Results from an IV Probit estimation with endogenous binary variable (to control for remittances) and a binary outcome variable reflecting a particular employment status, e.g., participating in the labor force, using the biprobit command in stata. The alternative specification is a probit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported with standard errors in parentheses. Empl. in wage & Self-employed / In the labor force Unemployed salary work employer Unpaid family work Subsistence work biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit 3 Remittance-receiving hh (d) 0.357*** 0.014 0.423* 0.105 0.109 -0.028 0.051 0.009 0.110 0.001 0.097 0.041 (0.096) (0.036) (0.246) (0.069) (0.148) (0.018) (0.068) (0.008) (0.073) (0.009) (0.110) (0.031) Age 0.062*** 0.064*** 0.066** 0.070** –0.005 –0.004 0.001 0.001 0.007 0.007* –0.000 –0.000 (0.016) (0.016) (0.031) (0.031) (0.009) (0.009) (0.002) (0.002) (0.005) (0.004) (0.013) (0.012) Age squared –0.001*** –0.001*** –0.002** –0.002*** 0.000 0.000 –0.000 –0.000 –0.000* –0.000* 0.000 0.000 (0.000) (0.000) (0.001) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 1 Primary or preparatory degree –0.059*** –0.058** 0.294*** 0.303*** 0.036* 0.037* –0.006*** –0.006*** –0.013*** –0.012*** –0.043*** –0.043*** (d) (0.023) (0.023) (0.108) (0.109) (0.021) (0.021) (0.002) (0.002) (0.004) (0.004) (0.014) (0.014) 1 Secondary degree (d) 0.132*** 0.138*** 0.644*** 0.653*** 0.112*** 0.113*** –0.007** –0.006** –0.036*** –0.034*** –0.100*** –0.099*** (0.019) (0.019) (0.053) (0.051) (0.015) (0.015) (0.003) (0.003) (0.005) (0.005) (0.012) (0.012) Above secondary degree1 (d) 0.343*** 0.350*** 0.612*** 0.617*** 0.291*** 0.293*** –0.007*** –0.007*** –0.045*** –0.044*** –0.175*** –0.174*** (0.024) (0.023) (0.060) (0.060) (0.026) (0.026) (0.002) (0.002) (0.005) (0.004) (0.012) (0.012) Ever-married (d) –0.281*** –0.281*** –0.093** –0.090** –0.127*** –0.125*** 0.005 0.005 –0.005 –0.005 0.035** 0.036** (0.022) (0.022) (0.038) (0.038) (0.017) (0.017) (0.004) (0.004) (0.006) (0.006) (0.016) (0.016) Nr of females in hh 0.008 0.010 –0.022 –0.021 0.006 0.006 –0.006*** –0.006*** 0.001 0.001 0.016** 0.016**   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options (0.009) (0.009) (0.014) (0.015) (0.005) (0.005) (0.002) (0.002) (0.002) (0.002) (0.007) (0.007) Nr of children aged 0–5 in hh –0.018** –0.015* 0.023 0.025 –0.022*** –0.021*** –0.004*** –0.004*** 0.004* 0.004** 0.010 0.011 (0.008) (0.008) (0.017) (0.016) (0.005) (0.005) (0.001) (0.001) (0.002) (0.002) (0.007) (0.007) Nr of children aged 6–14 in hh 0.031*** 0.030*** 0.011 0.012 –0.009* –0.009* 0.003*** 0.002** 0.008*** 0.008*** 0.022*** 0.022*** (0.008) (0.008) (0.014) (0.014) (0.005) (0.005) (0.001) (0.001) (0.002) (0.002) (0.006) (0.006) Presence of elderly in hh (d) 0.031 0.033 0.053 0.059* –0.019* –0.018* –0.005* –0.005* 0.012* 0.011* 0.027 0.027 (0.021) (0.021) (0.034) (0.034) (0.010) (0.010) (0.003) (0.003) (0.007) (0.007) (0.018) (0.018) (continued on next page) 5/27/10 2:41 PM Table A10: The Impact of Remittances on Female Employment (Young Women Aged 16–34, Out of School) (continued) Keller MNA 5-27-10vol2.indd 31 Results from an IV Probit estimation with endogenous binary variable (to control for remittances) and a binary outcome variable reflecting a particular employment status, e.g., participating in the labor force, using the biprobit command in stata. The alternative specification is a probit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported with standard errors in parentheses. Empl. in wage & Self-employed / In the labor force Unemployed salary work employer Unpaid family work Subsistence work biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit Alexandria and Suez2 (d) 0.044 0.047 0.039 0.040 0.007 0.008 –0.001 –0.001 0.011 0.011 0.131* 0.132* (0.029) (0.029) (0.050) (0.050) (0.015) (0.015) (0.006) (0.006) (0.023) (0.023) (0.071) (0.071) 2 Urban Lower Egypt (d) 0.093*** 0.105*** 0.202*** 0.214*** –0.018 –0.015 –0.001 –0.001 0.038 0.039 0.493*** 0.495*** (0.028) (0.028) (0.051) (0.050) (0.012) (0.012) . (0.006) (0.026) (0.026) (0.063) (0.063) 2 Rural Lower Egypt (d) 0.073*** 0.082*** 0.138*** 0.147*** 0.003 0.005 0.001 0.001 0.049** 0.048** 0.676*** 0.678*** (0.024) (0.024) (0.045) (0.045) (0.013) (0.013) (0.005) (0.005) (0.021) (0.020) (0.044) (0.044) Urban Upper Egypt2 (d) 0.076*** 0.080*** 0.040 0.040 –0.007 –0.006 0.012 0.011 0.112*** 0.110*** 0.480*** 0.481*** (0.027) (0.027) (0.045) (0.046) (0.012) (0.012) (0.008) (0.008) (0.036) (0.036) (0.062) (0.061) 2 Rural Upper Egypt (d) 0.163*** 0.180*** –0.023 –0.013 –0.031** –0.027** 0.032*** 0.033*** 0.139*** 0.141*** 0.633*** 0.636*** (0.027) (0.027) (0.047) (0.048) (0.013) (0.013) (0.011) (0.011) (0.034) (0.034) (0.048) (0.048) p 0.000 0.000 0.000 0.000 0.000 0.000 N 5551 1713 5551 5551 5551 5551 rho=0: Prob > chi2 0.0004 0.2540 0.1681 0.2696 0.0110 0.5545 Note: Since there may be more than one young woman living in a household, we take clustering at the household level into consideration. (d) for discrete change of dummy variable from 0 to 1, * p<0.10, ** p<0.05, *** p<0.01 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 31 5/27/10 2:41 PM 32 Keller MNA 5-27-10vol2.indd 32 Table A11: The Impact of Remittances on Male Employment (Young Men Aged 16–34, Out of School) Results from an IV Probit estimation with endogenous binary variable (to control for remittances) and a binary outcome variable reflecting a particular employment status, e.g. participating in the labor force, using the biprobit command in stata. The alternative specification is a probit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported with standard errors in parentheses. Empl. in wage & salary In the labor force Unemployed work Self-employed / employer Unpaid family work Biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit 3 Remittance-receiving hh (d) 0.019 –0.015 –0.029 0.005 –0.261 –0.114* 0.288 0.030 –0.024 0.054 (0.057) (0.026) (0.026) (0.023) (0.246) (0.059) (0.286) (0.040) (0.071) (0.034) Age –0.006 –0.006 –0.007 –0.006 0.029* 0.030* 0.069*** 0.069*** –0.007 –0.006 (0.009) (0.009) (0.006) (0.006) (0.016) (0.016) (0.012) (0.012) (0.007) (0.007) Age squared 0.000* 0.000* 0.000 0.000 –0.000 –0.000 –0.001*** –0.001*** 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Primary or preparatory 0.019** 0.019** –0.005 –0.006 0.050** 0.049** –0.015 –0.014 0.001 0.000 degree1 (d) (0.009) (0.009) (0.013) (0.012) (0.023) (0.024) (0.014) (0.014) (0.011) (0.011) 1 Secondary degree (d) 0.039*** 0.040*** 0.059*** 0.057*** –0.003 –0.004 –0.029** –0.028** 0.017* 0.015 (0.008) (0.008) (0.013) (0.013) (0.021) (0.021) (0.012) (0.012) (0.010) (0.009) Above secondary degree1 (d) 0.031*** 0.032*** 0.153*** 0.149*** 0.021 0.019 –0.086*** –0.084*** –0.013 –0.014 (0.009) (0.009) (0.027) (0.026) (0.024) (0.024) (0.011) (0.011) (0.011) (0.011) Ever-married (d) 0.091*** 0.090*** –0.068*** –0.066*** 0.088*** 0.090*** 0.087*** 0.085*** –0.035*** –0.033*** (0.010) (0.010) (0.009) (0.009) (0.021) (0.020) (0.014) (0.014) (0.010) (0.010) Nr of females in hh 0.001 0.001 0.009*** 0.009*** –0.020** –0.020** –0.019*** –0.018*** 0.023*** 0.023***   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options (0.004) (0.004) (0.004) (0.003) (0.009) (0.009) (0.006) (0.006) (0.004) (0.004) Nr of children aged 0–5 in hh –0.005 –0.005 –0.014*** –0.014*** –0.001 –0.002 0.002 0.002 –0.003 –0.003 (0.004) (0.004) (0.004) (0.004) (0.010) (0.010) (0.006) (0.006) (0.005) (0.005) Nr of children aged 6–14 in hh 0.002 0.002 –0.002 –0.002 –0.023** –0.023** –0.005 –0.005 0.018*** 0.017*** (0.004) (0.004) (0.004) (0.004) (0.010) (0.010) (0.007) (0.007) (0.004) (0.004) Presence of elderly in hh (d) –0.005 –0.005 –0.001 –0.001 –0.024 –0.024 –0.030** –0.030** 0.041*** 0.041*** (0.009) (0.009) (0.007) (0.007) (0.021) (0.021) (0.013) (0.013) (0.013) (0.013) (continued on next page) 5/27/10 2:41 PM Table A11: The Impact of Remittances on Male Employment (Young Men Aged 16–34, Out of School) (continued) Keller MNA 5-27-10vol2.indd 33 Results from an IV Probit estimation with endogenous binary variable (to control for remittances) and a binary outcome variable reflecting a particular employment status, e.g. participating in the labor force, using the biprobit command in stata. The alternative specification is a probit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported with standard errors in parentheses. Empl. in wage & salary In the labor force Unemployed work Self-employed / employer Unpaid family work Biprobit probit biprobit probit biprobit probit biprobit probit biprobit probit 2 Alexandria and Suez (d) –0.001 –0.001 0.014 0.014 –0.040 –0.041 0.045* 0.046* –0.029* –0.030* (0.014) (0.014) (0.012) (0.012) (0.031) (0.031) (0.025) (0.025) (0.017) (0.017) 2 Urban Lower Egypt (d) –0.012 –0.011 0.001 –0.000 –0.126*** –0.129*** 0.105*** 0.107*** 0.052* 0.049* (0.014) (0.014) (0.010) (0.009) (0.029) (0.029) (0.026) (0.026) (0.029) (0.028) Rural Lower Egypt2 (d) –0.007 –0.006 –0.007 –0.008 –0.132*** –0.135*** 0.067*** 0.069*** 0.115*** 0.111*** (0.012) (0.012) (0.008) (0.008) (0.026) (0.026) (0.020) (0.020) (0.029) (0.027) 2 Urban Upper Egypt (d) –0.021 –0.021 0.001 –0.000 –0.152*** –0.154*** 0.106*** 0.107*** 0.077*** 0.075** (0.015) (0.015) (0.009) (0.009) (0.029) (0.028) (0.025) (0.025) (0.030) (0.029) 2 Rural Upper Egypt (d) –0.019 –0.018 –0.025*** –0.026*** –0.123*** –0.126*** 0.073*** 0.075*** 0.127*** 0.123*** (0.014) (0.013) (0.007) (0.007) (0.028) (0.027) (0.022) (0.022) (0.032) (0.030) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5467 5467 4905 4905 5467 5467 5467 5467 5467 5467 rho=0: Prob > chi2 0.6120 0.4895 0.5571 0.2577 0.5114 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. (d) for discrete change of dummy variable from 0 to 1, * p<0.10, ** p<0.05, *** p<0.01 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 33 5/27/10 2:41 PM 34 Keller MNA 5-27-10vol2.indd 34 Table A12: The Impact of Migration on Hours Worked by Young Women (Aged 16–34, Out of School) Results from an IV Tobit estimation with endogenous binary variable (to control for migration) and the left-censored outcome variable hours worked of women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported conditional on working with standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit 3 Migrant hh (d) 12.192** –0.165 –0.844 –2.677** 11.033 1.115 14.454*** 1.467* 3.663*** 0.474** –2.543 2.408 (5.528) (1.122) (7.033) (1.197) (13.212) (1.216) (4.445) (0.831) (1.298) (0.238) (4.513) (1.884) Age 0.507 0.555 –0.314 –0.305 0.255 0.281 0.831* 0.875* –0.001 0.005 3.160*** 3.136*** (0.606) (0.600) (0.637) (0.637) (0.713) (0.701) (0.490) (0.474) (0.127) (0.124) (0.796) (0.796) Age squared 0.001 0.000 0.019 0.019 0.000 –0.000 –0.017* –0.018* 0.001 0.000 –0.058*** –0.057*** (0.012) (0.012) (0.012) (0.012) (0.014) (0.013) (0.010) (0.009) (0.003) (0.002) (0.016) (0.016) Primary or preparatory –2.504*** –2.466*** 2.326* 2.332* –2.177** –2.121** –1.565*** –1.563*** –0.583*** –0.583*** 0.677 0.664 degree1 (d) (0.811) (0.809) (1.347) (1.346) (0.944) (0.925) (0.533) (0.522) (0.150) (0.148) (1.140) (1.136) 1 Secondary degree (d) –0.830 –0.648 7.323*** 7.364*** –2.014** –1.888** –4.302*** –4.109*** –1.166*** –1.113*** 2.271** 2.154** (0.661) (0.654) (0.984) (0.985) (0.811) (0.766) (0.508) (0.496) (0.134) (0.129) (0.916) (0.912) Above secondary degree1 5.435*** 5.564*** 15.143*** 15.192*** –2.390** –2.275** –7.377*** –7.079*** –2.249*** –2.199*** –1.812 –1.914* (d) (0.869) (0.860) (1.288) (1.289) (0.956) (0.930) (0.870) (0.844) (0.169) (0.163) (1.119) (1.108) Ever-married (d) –6.796*** –6.676*** –7.807*** –7.777*** 1.741 1.745 –0.771 –0.754 0.350** 0.365** 15.352*** 15.314*** (0.861) (0.851) (0.913) (0.908) (1.321) (1.299) (0.659) (0.641) (0.164) (0.161) (0.904) (0.896)   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Nr of females in hh 0.279 0.418 0.462 0.493 –1.596*** –1.504*** –0.067 0.035 0.052 0.086 –4.835*** –4.923*** (0.334) (0.327) (0.349) (0.329) (0.583) (0.547) (0.232) (0.221) (0.070) (0.065) (0.431) (0.426) Nr of children aged 0–5 –0.761** –0.680** –1.442*** –1.427*** –1.204*** –1.141*** 0.313 0.396* 0.150** 0.169*** 6.261*** 6.223*** in hh (0.302) (0.302) (0.327) (0.324) (0.389) (0.369) (0.215) (0.206) (0.063) (0.063) (0.445) (0.441) Nr of children aged 6–14 0.780*** 0.775*** –0.632* –0.629* 0.727*** 0.710*** 0.808*** 0.786*** 0.149*** 0.148*** –0.026 –0.028 in hh (0.273) (0.269) (0.347) (0.346) (0.271) (0.267) (0.195) (0.189) (0.057) (0.055) (0.367) (0.366) Presence of elderly in hh –0.243 0.015 –1.282* –1.226* –1.530 –1.393 0.929 1.076* 0.240 0.315* 2.724** 2.557** (0.763) (0.749) (0.754) (0.730) (1.011) (1.009) (0.610) (0.588) (0.173) (0.168) (1.082) (1.096) (continued on next page) 5/27/10 2:41 PM Table A12: The Impact of Migration on Hours Worked by Young Women (Aged 16–34, Out of School) (continued) Keller MNA 5-27-10vol2.indd 35 Results from an IV Tobit estimation with endogenous binary variable (to control for migration) and the left-censored outcome variable hours worked of women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are reported conditional on working with standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit Alexandria and Suez2 (d) 0.230 0.253 0.258 0.267 –0.550 –0.534 0.817 0.768 1.395* 1.371* 1.959 1.935 (1.077) (1.062) (0.935) (0.933) (1.967) (1.923) (2.172) (2.082) (0.742) (0.730) (1.757) (1.761) 2 Urban Lower Egypt (d) –1.630 –1.356 –1.277 –1.207 –0.298 –0.189 2.951 2.918 5.510*** 5.519*** –7.456*** –7.631*** (0.998) (0.988) (0.901) (0.871) (1.832) (1.802) (1.895) (1.832) (0.940) (0.932) (1.408) (1.404) 2 Rural Lower Egypt (d) –1.060 –0.803 –0.251 –0.187 0.257 0.325 4.266*** 4.091*** 8.208*** 8.208*** –2.897** –3.064** (0.901) (0.889) (0.853) (0.818) (1.414) (1.384) (1.621) (1.558) (0.893) (0.884) (1.450) (1.461) Urban Upper Egypt2 (d) 0.554 0.687 –1.017 –0.987 2.524 2.529 7.509*** 7.235*** 5.839*** 5.790*** –10.881*** –10.962*** (0.960) (0.946) (0.800) (0.794) (1.573) (1.541) (1.991) (1.912) (0.935) (0.921) (1.175) (1.176) 2 Rural Upper Egypt (d) 3.062*** 3.657*** –2.803*** –2.684*** 5.768*** 5.993*** 8.908*** 8.958*** 7.784*** 7.928*** –12.766*** –13.075*** (0.990) (0.979) (1.028) (0.947) (1.527) (1.505) (1.842) (1.792) (0.900) (0.901) (1.239) (1.221) atanhrho_12 (coefficient) –0.396*** –0.091 –0.476 –0.661*** –0.441*** 0.099 (0.136) (0.323) (0.459) (0.156) (0.135) (0.072) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5551 5551 5551 5551 5551 5551 5551 5551 5551 5551 5551 5551 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 35 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options lar, albeit weaker, effect. Interestingly, children fact that they earn a higher income, which allows aged 6–14 in the household have a similar impact them to allocate more time to leisure. (As regards on time allocated to market and subsistence the number of hours allocated to unpaid family work whereas domestic work is not affected. work, we do not find any significant change. Yet, This suggests that childcare plays an important we do find a negative and significant effect when role for young women as long as children have we estimate the impact of remittances. Since we not entered school. Finally, as in all our estima- have very few observations for some variables, tions, we find regional differences in labor supply e.g., very few young men living in a migrant behavior. Surprisingly though, young women in household in the Greater Cairo Region and who less metropolitan areas engage less in domestic are working unpaid for the family, we probably work compared to their counterparts. This may need to be cautious about the estimation results suggest that in rural areas and especially in even though the model converged after reducing Upper Egypt, where extended family living is the problematic variables.) Being married gener- very common, domestic work is allocated to the ally increases the hours worked by young men. older women in the household whereas young Albeit young men need to raise large amounts of women are obliged to help in subsistence work. capital in order to marry, these estimates suggest In contrast, nuclear living arrangements are that they are faced with even greater demands more common in metropolitan areas and have once they have married. That we do not find any increased among young couples over the last significant impact of the number of children aged decade (Binzel and Assaad 2008). 0–5 in the household may be due to high correla- tion between marriage and childbirth. Typically, What about young men? Earlier, when look- the first child is on average born relatively soon ing at the impact of migration on the decision of after marriage, so that the effect of childbirth young men to enter the labor force and take up may be similar to a lagged effect of marriage and a particular type of job, we did not find any sig- thus turns out to be insignificant.21 As one would nificant impact. Now, we find that the number of expect, men living outside the Greater Cairo Re- hours worked in market work in general is not sig- gion work more hours self-employed and fewer nificantly affected, neither positive nor negative hours in a waged job than their counterparts. For (see Table A13). Yet, young men’s decisions about unpaid family work, results show that young men the number of hours spent in a particular employ- living in urban areas spend significantly less time ment type are affected. Young men living in a mi- in unpaid family work. grant household work about 6 hours per week less in wage and salary work and 20 hours per week International Remittance Income less in unpaid family work. Employers and young men working self-employed, in contrast, work Remittances have a similar effect on women’s about 4 hours more per week. The decrease in time allocation as migration, although the effect hours worked unpaid for the family may suggest on time allocated to subsistence work is not that young men do not need to replace the labor significant anymore (see Table A14). As before, of the migrant who went abroad—in contrast to women having a wage and salary job do not what we found earlier for young women.While change their behavior. Also, the number of hours age did not affect men’s decision about whether young women allocate to domestic work, such or not to work, the time spent in any market as childcare, is not affected. One may think that activity increases with age (though decreases at the value of remittances sent to the household is some point again, see age squared). Moreover, likely to play a role. Estimation results are pro- earlier results showed that men with second- vided in Table A15. Only wage and salary work ary degree and above have a lower probability to be self-employed or work as employer. Now, 21 In 1995, the median interval between marriage and first birth we see that even those who are self-employed was 13.9 months (Eltigani 2000). Based on the ELMPS 06, almost spent less time working compared to those with 60% of the women gave birth to their first child in the year after no educational certificate. This may hint to the marriage and 21% in the second year after marriage. 36 Keller MNA 5-27-10vol2.indd 36 5/27/10 2:41 PM Table A13: The Impact of Migration on Hours Worked by Young Men (Aged 16–34, Out of School) Keller MNA 5-27-10vol2.indd 37 Results from an IV Tobit estimation with endogenous binary variable (to control for migration) and the left-censored outcome variable hours worked of men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work cmp tobit cmp tobit cmp tobit cmp tobit Current international migrant in –3.107 –1.575 –8.210** –5.504*** 14.136 3.719* –19.590*** 2.909* hh3 (d) (2.886) (2.090) (3.861) (1.808) (11.296) (2.036) (1.019) (1.752) Age 2.937*** 2.942*** 2.005*** 2.015*** 5.033*** 5.016*** –0.426 –0.326 (0.688) (0.687) (0.697) (0.696) (0.910) (0.907) (1.041) (0.666) Age squared –0.042*** –0.042*** –0.029** –0.029** –0.081*** –0.081*** –0.009 –0.004 (0.013) (0.013) (0.013) (0.013) (0.017) (0.017) (0.021) (0.013) Primary or preparatory degree1 (d) 2.310** 2.304** 2.599** 2.586** –1.070 –1.069 0.076 0.071 (0.997) (0.998) (1.064) (1.074) (1.021) (1.016) (1.658) (1.069) 1 Secondary degree (d) –0.334 –0.349 0.434 0.401 –2.196** –2.161** 2.940** 1.214 (0.876) (0.877) (0.903) (0.905) (0.861) (0.858) (1.398) (0.894) 1 Above secondary degree (d) –6.364*** –6.404*** –0.003 –0.090 –6.758*** –6.635*** 1.347 –2.005* (0.944) (0.944) (0.998) (0.993) (0.939) (0.932) (1.888) (1.144) Ever-married (d) 9.239*** 9.280*** 3.804*** 3.888*** 6.297*** 6.124*** –7.222*** –3.298*** (0.790) (0.789) (0.866) (0.857) (0.985) (0.976) (1.440) (0.968) Nr of females in hh –0.241 –0.255 –0.655 –0.685* –1.365*** –1.314*** 3.720*** 2.336*** (0.408) (0.408) (0.413) (0.411) (0.448) (0.443) (0.584) (0.378) Nr of children aged 0–5 in hh 0.153 0.139 –0.099 –0.128 0.109 0.162 0.452 –0.387 (0.356) (0.356) (0.415) (0.413) (0.404) (0.403) (0.675) (0.446) Nr of children aged 6–14 in hh 0.067 0.064 –1.197*** –1.204*** –0.302 –0.294 2.617*** 1.687*** (0.413) (0.414) (0.436) (0.434) (0.572) (0.570) (0.552) (0.367) Presence of elderly in hh (d) –0.785 –0.812 –1.238 –1.294 –2.640*** –2.557*** 6.324*** 3.548*** (0.925) (0.924) (0.917) (0.912) (0.975) (0.969) (1.609) (1.018) (continued on next page) Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 37 5/27/10 2:41 PM 38 Keller MNA 5-27-10vol2.indd 38 Table A13: The Impact of Migration on Hours Worked by Young Men (Aged 16–34, Out of School) (continued) Results from an IV Tobit estimation with endogenous binary variable (to control for migration) and the left-censored outcome variable hours worked of men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work cmp tobit cmp tobit cmp tobit cmp tobit 2 4 Alexandria and Suez (d) –2.645** –2.649** –2.318** –2.326** 2.980* 3.002* –8.591*** –3.454 (1.280) (1.280) (1.082) (1.081) (1.716) (1.709) (1.222) (2.376) 2 Urban Lower Egypt (d) –3.848*** –3.871*** –5.807*** –5.851*** 6.338*** 6.403*** 4.348* (1.182) (1.181) (0.998) (0.991) (1.584) (1.583) (2.355) 2 Rural Lower Egypt (d) –3.191*** –3.217*** –5.967*** –6.017*** 4.140*** 4.223*** 8.950*** (1.056) (1.057) (0.936) (0.927) (1.342) (1.339) (2.083) Urban Upper Egypt2 (d) –5.310*** –5.327*** –7.346*** –7.379*** 6.466*** 6.522*** 6.298*** (1.127) (1.127) (0.954) (0.949) (1.549) (1.545) (2.273) 2 Rural Upper Egypt (d) –4.301*** –4.364*** –6.897*** –7.018*** 4.304*** 4.560*** 9.388*** (1.077) (1.073) (0.974) (0.954) (1.472) (1.457) (2.200) atanhrho_12 (coefficient) 0.037 0.082 –0.238 3.803*** (0.049) (0.106) (0.209) (0.430) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5467 5467 5467 5467 5467 5467 5467 5467 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. 4 Due to too few ob- servations in some of the region categories, urban and rural regions are lumped together. The coefficient hence refers to an urban dummy. 5/27/10 2:41 PM Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results is affected: every 100 LE a remittance-receiving and work 17 hours less per week unpaid for the household obtains decreases the number of family. In contrast to migration, those working hours worked by a young woman by roughly self-employed or as employer do not change the six hours per week. This provides evidence that number of hours worked. The picture changes remittance-income has a mixed effect depending when we instrument for the average monthly on the type of household. If households have their value of remittances sent to the household. Every own business and/or are engaged in subsistence 100 LE that a household receives in form of re- economy, we find an increase in productivity for mittances decreases the hours worked by young women by either taking up work or by increasing men in any market work by 13 hours per week. the number of hours worked. On the contrary, if The effect on hours allocated to wage and salary women earn a wage income, they are more likely work is similarly high and thus appears stronger to reduce their workload and substitute work for compared to when we instrument for whether leisure. In part, this may be driven by the rela- or not the household is receiving remittances. In tively low wage women earn. Households may contrast, young men slightly increase the hours therefore prefer that women stay at home, which they work per week self-employed or unpaid for conforms to gender roles and may be particularly the family which could provide evidence for the preferred for unmarried women. Since domestic hypothesis that remittance-income may be spent work does not increase, we cannot observe any in a productive way even if the effect is not very specialization of women on domestic chores. large. Yet, remittance income is not significantly correlated with these two employment catego- How does young men’s time allocation ries; it is instead significantly correlated with the change in response to living in a household hours worked decision for any market work and that receives remittances? The effect on time for wage and salary work. All in all, hence, we do allocated to wage and salary is very similar to not obtain any clear-cut results on the effect of our findings for migration: They spend around average monthly remittance income on employ- six hours less per week in wage and salary work ment behavior. 39 Keller MNA 5-27-10vol2.indd 39 5/27/10 2:41 PM 40 Keller MNA 5-27-10vol2.indd 40 Table A14: The Impact of Remittances on Hours Worked by Young Women (Aged 16–34, Out of School) Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work cmp Tobit cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit 3 Remittance-receiving hh (d) 10.260* –1.146 3.920 –2.012 11.821 1.786 8.583** –0.156 2.482 0.327 –4.759 1.487 (5.873) (1.313) (6.260) (1.490) (13.867) (1.328) (4.081) (0.909) (1.584) (0.262) (3.699) (2.236) Age 0.528 0.562 –0.328 –0.309 0.233 0.256 0.880* 0.887* 0.002 0.005 3.170*** 3.141*** (0.605) (0.600) (0.639) (0.636) (0.710) (0.702) (0.481) (0.473) (0.125) (0.124) (0.796) (0.796) Age squared 0.001 0.000 0.019 0.019 0.001 0.000 –0.018* –0.018* 0.000 0.000 –0.058*** –0.057*** (0.012) (0.012) (0.012) (0.012) (0.014) (0.013) (0.009) (0.009) (0.002) (0.002) (0.016) (0.016) Primary or preparatory –2.507*** –2.460*** 2.340* 2.359* –2.203** –2.137** –1.557*** –1.528*** –0.587*** –0.583*** 0.692 0.665 degree1 (d) (0.811) (0.810) (1.354) (1.350) (0.939) (0.925) (0.530) (0.525) (0.149) (0.149) (1.141) (1.137) 1 Secondary degree (d) –0.759 –0.636 7.312*** 7.369*** –2.011** –1.896** –4.162*** –4.052*** –1.136*** –1.106*** 2.297** 2.184** (0.658) (0.653) (0.989) (0.987) (0.804) (0.766) (0.507) (0.498) (0.132) (0.129) (0.909) (0.909) Above secondary degree1 (d) 5.508*** 5.567*** 15.153*** 15.187*** –2.354** –2.277** –7.204*** –7.065*** –2.216*** –2.195*** –1.806 –1.883* (0.866) (0.860) (1.293) (1.291) (0.944) (0.931) (0.860) (0.846) (0.166) (0.164) (1.110) (1.107) Ever-married (d) –6.744*** –6.668*** –7.836*** –7.769*** 1.765 1.751 –0.738 –0.724 0.355** 0.365** 15.357*** 15.324*** (0.863) (0.850) (0.915) (0.909) (1.318) (1.306) (0.652) (0.640) (0.163) (0.162) (0.904) (0.896) Nr of females in hh 0.362 0.423 0.440 0.477 –1.541*** –1.507*** 0.046 0.085 0.080 0.093 –4.837*** –4.894***   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options (0.332) (0.325) (0.335) (0.330) (0.550) (0.548) (0.226) (0.220) (0.068) (0.065) (0.431) (0.427) Nr of children aged 0–5 in –0.756** –0.668** –1.487*** –1.438*** –1.217*** –1.153*** 0.341 0.408** 0.152** 0.170*** 6.288*** 6.227*** hh (0.302) (0.302) (0.327) (0.325) (0.383) (0.369) (0.211) (0.207) (0.063) (0.063) (0.447) (0.443) Nr of children aged 6–14 0.804*** 0.775*** –0.627* –0.631* 0.742*** 0.710*** 0.818*** 0.785*** 0.155*** 0.149*** –0.039 –0.023 in hh (0.271) (0.269) (0.348) (0.347) (0.273) (0.266) (0.193) (0.189) (0.055) (0.055) (0.366) (0.365) Presence of elderly in hh (d) –0.013 0.020 –1.311* –1.284* –1.393 –1.384 1.141* 1.130* 0.322* 0.332** 2.668** 2.628** (0.754) (0.747) (0.735) (0.730) (1.019) (1.012) (0.596) (0.587) (0.171) (0.168) (1.108) (1.104) (continued on next page) 5/27/10 2:41 PM Table A14: The Impact of Remittances on Hours Worked by Young Women (Aged 16–34, Out of School) (continued) Keller MNA 5-27-10vol2.indd 41 Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work cmp Tobit cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit Alexandria and Suez2 (d) 0.209 0.264 0.239 0.283 –0.571 –0.556 0.786 0.776 1.375* 1.373* 1.997 1.931 (1.069) (1.062) (0.937) (0.933) (1.949) (1.923) (2.125) (2.084) (0.734) (0.730) (1.755) (1.760) 2 Urban Lower Egypt (d) –1.575 –1.323 –1.399 –1.230 –0.330 –0.213 2.949 2.988 5.483*** 5.534*** –7.373*** –7.600*** (0.993) (0.989) (0.878) (0.872) (1.815) (1.801) (1.863) (1.838) (0.932) (0.933) (1.405) (1.405) 2 Rural Lower Egypt (d) –0.985 –0.778 –0.356 –0.220 0.229 0.308 4.200*** 4.162*** 8.189*** 8.227*** –2.834* –3.029** (0.897) (0.889) (0.830) (0.819) (1.406) (1.390) (1.590) (1.562) (0.887) (0.886) (1.453) (1.462) Urban Upper Egypt2 (d) 0.618 0.695 –1.050 –1.003 2.514 2.512 7.432*** 7.295*** 5.816*** 5.806*** –10.865*** –10.941*** (0.952) (0.945) (0.799) (0.795) (1.564) (1.542) (1.954) (1.916) (0.926) (0.922) (1.175) (1.177) 2 Rural Upper Egypt (d) 3.296*** 3.702*** –2.972*** –2.751*** 5.764*** 5.960*** 9.056*** 9.154*** 7.868*** 7.972*** –12.699*** –12.995*** (0.986) (0.979) (0.956) (0.946) (1.523) (1.506) (1.824) (1.805) (0.901) (0.905) (1.220) (1.219) atanhrho_12 (coefficient) –0.362** –0.242 –0.442 –0.493*** –0.304* 0.123** (0.145) (0.211) (0.442) (0.173) (0.177) (0.048) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5551 5551 5551 5551 5551 5551 5551 5551 5551 5551 5551 5551 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 41 5/27/10 2:41 PM 42 Keller MNA 5-27-10vol2.indd 42 Table A15: The Impact of the Value of Remittances Sent on Hours Worked by Young Women (Aged 16–34, Out of School) Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit Average monthly remittance- 1.344 –0.018 –5.453* –0.295 0.062 –0.050 2.344 0.077 0.869 0.038 –0.299 0.124 income (in 100 LE)3 (2.630) (0.179) (2.965) (0.225) (1.791) (0.141) (2.276) (0.101) (0.764) (0.045) (1.574) (0.158) Age 0.550 0.551 –0.288 –0.326 0.296 0.293 0.982* 0.870* –0.002 0.001 3.171*** 3.168*** (0.607) (0.600) (0.777) (0.637) (0.708) (0.705) (0.541) (0.474) (0.135) (0.124) (0.798) (0.797) Age squared 0.000 0.000 0.020 0.019 –0.000 –0.000 –0.020* –0.018* 0.001 0.001 –0.058*** –0.058*** (0.012) (0.012) (0.015) (0.012) (0.013) (0.013) (0.011) (0.009) (0.003) (0.002) (0.016) (0.016) Primary or preparatory –2.662*** –2.466*** 3.536** 2.390* –2.129** –2.116** –2.014*** –1.540*** –0.713*** –0.591*** 0.710 0.653 degree1 (d) (0.886) (0.810) (1.591) (1.353) (0.975) (0.929) (0.682) (0.527) (0.193) (0.149) (1.168) (1.139) 1 Secondary degree (d) –0.817 –0.644 9.161*** 7.382*** –1.885** –1.871** –4.810*** –4.061*** –1.279*** –1.106*** 2.269** 2.210** (0.732) (0.654) (1.157) (0.988) (0.805) (0.766) (0.620) (0.500) (0.168) (0.129) (0.945) (0.913) Above secondary degree1 (d) 5.282*** 5.578*** 18.405*** 15.249*** –2.281** –2.261** –8.303*** –7.088*** –2.461*** –2.191*** –1.801 –1.895* (1.032) (0.861) (1.646) (1.293) (1.068) (0.933) (0.978) (0.848) (0.193) (0.164) (1.166) (1.109) Ever-married (d) –6.848*** –6.695*** –8.605*** –7.764*** 1.711 1.717 –0.955 –0.700 0.358** 0.375** 15.378*** 15.358*** (0.875) (0.856) (1.075) (0.910) (1.312) (1.309) (0.736) (0.643) (0.180) (0.162) (0.926) (0.896) Nr of females in hh 0.225 0.408 1.022** 0.487 –1.523** –1.504*** –0.300 0.075 –0.023 0.094 –4.838*** –4.897***   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options (0.412) (0.326) (0.474) (0.331) (0.685) (0.552) (0.285) (0.222) (0.083) (0.065) (0.483) (0.431) Nr of children aged 0–5 in hh –0.861** –0.676** –1.041* –1.436*** –1.149*** –1.134*** 0.157 0.396* 0.074 0.166*** 6.287*** 6.235*** (0.429) (0.303) (0.575) (0.325) (0.419) (0.368) (0.344) (0.209) (0.107) (0.063) (0.495) (0.445) Nr of children aged 6–14 in 0.754*** 0.773*** –0.715* –0.631* 0.713*** 0.713*** 0.815*** 0.783*** 0.141** 0.149*** –0.026 –0.034 hh (0.273) (0.270) (0.414) (0.347) (0.270) (0.268) (0.222) (0.190) (0.060) (0.055) (0.368) (0.366) Presence of elderly in hh (d) 0.056 –0.006 –1.855* –1.321* –1.340 –1.339 1.371** 1.125* 0.394** 0.341** 2.629** 2.648** (0.759) (0.747) (1.050) (0.729) (1.020) (1.013) (0.694) (0.589) (0.197) (0.168) (1.120) (1.110) (continued on next page) 5/27/10 2:41 PM Table A15: The Impact of the Value of Remittances Sent on Hours Worked by Young Women (Aged 16–34, Out of School) (continued) Keller MNA 5-27-10vol2.indd 43 Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit Alexandria and Suez2 (d) 0.257 0.238 0.199 0.247 –0.523 –0.523 0.844 0.780 1.442* 1.379* 1.908 1.911 (1.074) (1.062) (1.148) (0.934) (1.922) (1.924) (2.312) (2.088) (0.756) (0.730) (1.753) (1.760) 2 Urban Lower Egypt (d) –1.539 –1.368 –0.903 –1.291 –0.150 –0.138 2.911 2.984 5.434*** 5.550*** –7.517*** –7.563*** (1.040) (0.988) (1.130) (0.869) (1.818) (1.806) (2.016) (1.840) (0.930) (0.934) (1.438) (1.405) 2 Rural Lower Egypt (d) –0.813 –0.799 –0.200 –0.261 0.398 0.399 4.509*** 4.172*** 8.263*** 8.221*** –2.985** –2.992** (0.905) (0.889) (0.995) (0.817) (1.393) (1.392) (1.734) (1.565) (0.870) (0.885) (1.473) (1.469) Urban Upper Egypt2 (d) 0.705 0.699 –1.253 –1.023 2.567 2.567* 7.820*** 7.319*** 5.839*** 5.794*** –10.978*** –10.976*** (0.963) (0.947) (0.967) (0.794) (1.565) (1.542) (2.082) (1.920) (0.908) (0.922) (1.221) (1.178) 2 Rural Upper Egypt (d) 3.265*** 3.661*** –2.108* –2.792*** 6.074*** 6.102*** 8.897*** 9.106*** 7.583*** 7.969*** –12.907*** –13.014*** (1.199) (0.979) (1.278) (0.944) (1.630) (1.504) (2.072) (1.803) (0.915) (0.904) (1.333) (1.220) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5541 5541 5541 5541 5541 5541 5541 5541 5541 5541 5541 5541 Wald test of exogeneity 0.5840 0.0817 0.9494 0.2853 0.2339 0.7889 (Prob >chi2) Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 43 5/27/10 2:41 PM 44 Keller MNA 5-27-10vol2.indd 44 Table A16: The Impact of Remittances on Hours Worked by Young Men (Aged 16–34, Out of School) Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work Cmp tobit cmp tobit cmp tobit cmp4 tobit 3 Remittance-receiving hh (d) –2.667 –1.270 –10.763 –6.416*** 13.680 1.924 –17.353*** 4.602* (4.272) (2.986) (9.771) (2.359) (15.017) (2.642) (0.930) (2.432) Age 2.935*** 2.942*** 1.978*** 2.006*** 5.019*** 4.997*** –0.615 –0.328 (0.688) (0.687) (0.697) (0.697) (0.906) (0.906) (0.950) (0.666) Age squared –0.042*** –0.042*** –0.029** –0.029** –0.081*** –0.080*** –0.004 –0.004 (0.013) (0.013) (0.013) (0.013) (0.017) (0.017) (0.019) (0.013) Primary or preparatory de- 2.317** 2.307** 2.654** 2.613** –1.115 –1.081 0.024 gree1 (d) (0.999) (0.999) (1.077) (1.076) (1.020) (1.016) (1.067) 1 Secondary degree (d) –0.341 –0.354 0.454 0.399 –2.198** –2.146** 1.191 (0.877) (0.877) (0.914) (0.905) (0.862) (0.859) (0.892) 1 Above secondary degree (d) –6.406*** –6.426*** –0.057 –0.146 –6.684*** –6.596*** –3.580*** –1.993* (0.945) (0.944) (1.019) (0.992) (0.935) (0.932) (1.220) (1.142) Ever-married (d) 9.282*** 9.302*** 3.868*** 3.946*** 6.168*** 6.066*** –5.223*** –3.277*** (0.789) (0.788) (0.882) (0.856) (0.977) (0.975) (1.272) (0.966) Nr of females in hh –0.261 –0.266 –0.695* –0.715* –1.300*** –1.279*** 3.573*** 2.356*** (0.407) (0.407) (0.413) (0.410) (0.444) (0.442) (0.530) (0.376)   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Nr of children aged 0–5 in hh 0.140 0.132 –0.107 –0.140 0.139 0.176 0.718 –0.385 (0.356) (0.356) (0.423) (0.413) (0.403) (0.404) (0.612) (0.445) Nr of children aged 6–14 in hh 0.066 0.063 –1.187*** –1.198*** –0.302 –0.294 2.784*** 1.687*** (0.415) (0.415) (0.437) (0.434) (0.571) (0.570) (0.501) (0.364) Presence of elderly in hh –0.837 –0.838 –1.380 –1.386 –2.510*** –2.506*** 3.606*** (0.923) (0.923) (0.911) (0.911) (0.973) (0.970) (1.018) (continued on next page) 5/27/10 2:41 PM Table A16: The Impact of Remittances on Hours Worked by Young Men (Aged 16–34, Out of School) (continued) Keller MNA 5-27-10vol2.indd 45 Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work Cmp tobit cmp tobit cmp tobit cmp4 tobit Alexandria and Suez2 (d) –2.637** –2.646** –2.275** –2.310** 2.950* 2.985* –3.517 (1.280) (1.281) (1.090) (1.083) (1.713) (1.709) (2.369) 2 Urban Lower Egypt (d) –3.854*** –3.875*** –5.759*** –5.841*** 6.354*** 6.440*** 4.246* (1.182) (1.180) (1.018) (0.992) (1.587) (1.586) (2.347) 2 Rural Lower Egypt (d) –3.194*** –3.220*** –5.901*** –5.999*** 4.141*** 4.238*** 8.842*** (1.057) (1.057) (0.964) (0.928) (1.342) (1.340) (2.076) Urban Upper Egypt2 (d) –5.313*** –5.329*** –7.315*** –7.375*** 6.467*** 6.536*** 6.191*** (1.128) (1.127) (0.966) (0.950) (1.548) (1.545) (2.268) 2 Rural Upper Egypt (d) –4.376*** –4.403*** –7.020*** –7.121*** 4.528*** 4.657*** 9.352*** (1.073) (1.071) (0.988) (0.953) (1.460) (1.455) (2.193) atanhrho_12 (coefficient) 0.030 0.129 –0.249 3.698*** (0.062) (0.308) (0.243) (0.392) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5467 5467 5467 5467 5467 5467 5467 5467 Note: Since there may be more than one young man living in a HH, we take clustering at the HH level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 Instrumental ariable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. 4 Due to too few observations, the educational categories have been lumped together (the reference category now includes all men with secondary degree and less) and the variable “presence of elderly in hhâ€? and the regional dummies have been dropped. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 45 5/27/10 2:41 PM 46 Keller MNA 5-27-10vol2.indd 46 Table A17: The Impact of the Value of Remittances Sent on Hours Worked by Young Men (Aged 16–34, Out of School) Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit Average monthly remittance-income –12.856** –0.359 –13.284** –1.654*** 3.536 0.495* 5.112 0.498** (in 100 LE)3 (6.542) (0.443) (6.458) (0.472) (8.491) (0.256) (5.934) (0.241) Age 2.821*** 2.954*** 2.045*** 1.999*** 5.034*** 5.002*** –0.365 –0.347 (0.770) (0.688) (0.789) (0.697) (0.923) (0.908) (0.722) (0.669) Age squared –0.042*** –0.043*** –0.031** –0.029** –0.081*** –0.081*** –0.004 –0.004 (0.015) (0.013) (0.015) (0.013) (0.017) (0.017) (0.014) (0.013) 1 Primary or preparatory degree (d) 2.244** 2.283** 2.648** 2.588** –1.061 –1.025 –0.082 –0.043 (1.065) (0.999) (1.146) (1.076) (1.042) (1.019) (1.139) (1.070) 1 Secondary degree (d) 0.352 –0.337 1.015 0.402 –2.337** –2.158** 1.020 1.201 (1.028) (0.877) (1.068) (0.905) (0.979) (0.861) (1.022) (0.893) 1 Above secondary degree (d) –4.332*** –6.405*** 1.474 –0.088 –7.034*** –6.611*** –2.675* –2.008* (1.438) (0.944) (1.522) (0.994) (1.310) (0.934) (1.451) (1.146) Ever-married (d) 7.496*** 9.287*** 2.830*** 3.927*** 6.417*** 6.095*** –3.091*** –3.295*** (1.092) (0.788) (1.085) (0.856) (1.205) (0.973) (1.115) (0.966) Nr of females in hh 0.182 –0.253 –0.298 –0.706* –1.397*** –1.277*** 2.326*** 2.363*** (0.476) (0.408) (0.481) (0.410) (0.531) (0.443) (0.449) (0.376)   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Nr of children aged 0–5 in hh 0.868 0.153 0.566 –0.117 –0.027 0.148 –0.690 –0.401 (0.676) (0.356) (0.658) (0.413) (0.632) (0.402) (0.611) (0.447) Nr of children aged 6–14 in hh 0.191 0.056 –1.009** –1.176*** –0.339 –0.303 1.712*** 1.684*** (0.485) (0.416) (0.501) (0.434) (0.584) (0.571) (0.409) (0.365) Presence of elderly in hh (d) –0.772 –0.836 –1.365 –1.417 –2.533** –2.511*** 3.817*** 3.630*** (0.992) (0.923) (0.977) (0.909) (0.987) (0.970) (1.081) (1.022) (continued on next page) 5/27/10 2:41 PM Table A17: The Impact of the Value of Remittances Sent on Hours Worked by Young Men (Aged 16–34, Out of School) (continued) Keller MNA 5-27-10vol2.indd 47 Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Marginal effects are report conditional on working with standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit Alexandria and Suez2 (d) –1.937 –2.641** –1.775 –2.294** 2.885 2.981* –3.861 –3.510 (1.375) (1.280) (1.284) (1.083) (1.771) (1.708) (2.512) (2.375) 2 Urban Lower Egypt (d) –2.888** –3.877*** –5.114*** –5.846*** 6.313*** 6.448*** 4.204* 4.334* (1.310) (1.180) (1.192) (0.991) (1.702) (1.586) (2.488) (2.355) 2 Rural Lower Egypt (d) –2.461** –3.223*** –5.411*** –5.996*** 4.111*** 4.201*** 9.064*** 8.968*** (1.122) (1.057) (1.027) (0.927) (1.409) (1.339) (2.221) (2.086) Urban Upper Egypt2 (d) –4.751*** –5.327*** –7.140*** –7.422*** 6.602*** 6.591*** 6.453*** 6.284*** (1.122) (1.128) (1.025) (0.949) (1.583) (1.546) (2.400) (2.275) 2 Rural Upper Egypt (d) –2.694** –4.404*** –5.735*** –7.082*** 4.301** 4.632*** 9.064*** 9.369*** (1.226) (1.071) (1.150) (0.954) (1.817) (1.455) (2.440) (2.200) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5460 5460 5460 5460 5460 5460 5460 5460 Wald test of exogeneity (Prob >chi2) 0.0570 0.0685 0.7190 0.4326 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 47 5/27/10 2:41 PM 48 Keller MNA 5-27-10vol2.indd 48 Table A18: The Impact of the Value of Remittances Sent on Female Employment (Young Women Aged 16–34, Out of School) Results from an IV Probit estimation with a continuous endogenous variable (to control for remittance income) and a binary outcome variable reflecting a particular employment status, e.g. participating in the labor force, using the ivprobit command in stata. The alternative specification is a probit model assuming that remittance- income is exogenous to the outcome variable. Standard errors in parentheses. Empl. in wage & Self-employed / In the labor force Unemployed salary work employer Unpaid family work Subsistence work ivprobit probit ivprobit probit ivprobit probit ivprobit probit ivprobit probit ivprobit Probit Average monthly remittance- 0.083 0.001 0.050 0.005 –0.069 –0.005 0.000 –0.000 0.031 0.001 0.069 0.002 income (in 100 LE)3 (0.098) (0.004) (0.035) (0.007) (0.043) (0.003) (0.006) (0.000) (0.041) (0.001) (0.060) (0.005) Age 0.060*** 0.064*** 0.075** 0.074** –0.004 –0.005 0.001 0.001 0.010 0.007 –0.001 –0.001 (0.018) (0.016) (0.031) (0.031) (0.011) (0.009) (0.002) (0.002) (0.007) (0.004) (0.013) (0.012) Age squared –0.001*** –0.001*** –0.002*** –0.002*** 0.000 0.000 –0.000 –0.000 –0.000 –0.000* 0.000 0.000 (0.000) (0.000) (0.001) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 1 Primary or preparatory degree –0.065*** –0.058** 0.270** 0.301*** 0.051** 0.037* –0.006*** –0.006*** –0.021 –0.012*** –0.051*** –0.044*** (d) (0.023) (0.023) (0.111) (0.109) (0.024) (0.021) (0.002) (0.002) (0.014) (0.004) (0.017) (0.014) 1 Secondary degree (d) 0.118*** 0.137*** 0.625*** 0.651*** 0.131*** 0.113*** –0.006** –0.006** –0.052** –0.034*** –0.107*** –0.099*** (0.036) (0.019) (0.061) (0.051) (0.018) (0.015) (0.003) (0.003) (0.025) (0.005) (0.014) (0.012) Above secondary degree1 (d) 0.305*** 0.349*** 0.586*** 0.613*** 0.308*** 0.294*** –0.007*** –0.007*** –0.069* –0.044*** –0.186*** –0.173*** (0.075) (0.024) (0.068) (0.060) (0.025) (0.026) (0.002) (0.002) (0.037) (0.004) (0.016) (0.012) Ever-married (d) –0.269*** –0.282*** –0.101*** –0.091** –0.129*** –0.125*** 0.005 0.005 –0.009 –0.005 0.033* 0.037** (0.039) (0.022) (0.038) (0.038) (0.018) (0.017) (0.004) (0.004) (0.010) (0.006) (0.017) (0.016) Nr of females in hh –0.002 0.010 –0.030** –0.022 0.013* 0.006 –0.006** –0.006*** –0.003 0.001 0.006 0.016**   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options (0.012) (0.009) (0.015) (0.015) (0.007) (0.005) (0.002) (0.002) (0.004) (0.002) (0.008) (0.007) Nr of children aged 0–5 in hh –0.025** –0.015* 0.019 0.026 –0.016** –0.021*** –0.004** –0.004*** 0.002 0.004** 0.002 0.011 (0.012) (0.008) (0.018) (0.016) (0.007) (0.005) (0.002) (0.001) (0.004) (0.002) (0.010) (0.007) Nr of children aged 6–14 in hh 0.027*** 0.030*** 0.009 0.012 –0.010* –0.009* 0.002** 0.002** 0.010** 0.008*** 0.020*** 0.022*** (0.009) (0.008) (0.014) (0.014) (0.006) (0.005) (0.001) (0.001) (0.005) (0.002) (0.006) (0.006) Presence of elderly in hh (d) 0.035 0.033 0.063* 0.062* –0.025* –0.019* –0.004* –0.004* 0.017 0.011* 0.031 0.028 (0.022) (0.021) (0.035) (0.034) (0.014) (0.010) (0.003) (0.003) (0.011) (0.007) (0.020) (0.018) (continued on next page) 5/27/10 2:41 PM Table A18: The Impact of the Value of Remittances Sent on Female Employment (Young Women Aged 16–34, Out of School) (continued) Keller MNA 5-27-10vol2.indd 49 Results from an IV Probit estimation with a continuous endogenous variable (to control for remittance income) and a binary outcome variable reflecting a particular employment status, e.g. participating in the labor force, using the ivprobit command in stata. The alternative specification is a probit model assuming that remittance- income is exogenous to the outcome variable. Standard errors in parentheses. Empl. in wage & Self-employed / In the labor force Unemployed salary work employer Unpaid family work Subsistence work ivprobit probit ivprobit probit ivprobit probit ivprobit probit ivprobit probit ivprobit Probit 2 Alexandria and Suez (d) 0.044 0.046 0.038 0.039 0.007 0.007 –0.001 –0.001 0.014 0.011 0.127* 0.132* (0.028) (0.029) (0.050) (0.050) (0.017) (0.015) (0.006) (0.006) (0.030) (0.023) (0.067) (0.071) 2 Urban Lower Egypt (d) 0.087** 0.105*** 0.209*** 0.217*** –0.011 –0.016 –0.000 –0.000 0.043 0.038 0.451*** 0.497*** (0.038) (0.028) (0.049) (0.050) (0.016) (0.012) (0.006) (0.006) (0.030) (0.026) (0.079) (0.063) Rural Lower Egypt2 (d) 0.075*** 0.082*** 0.144*** 0.148*** 0.007 0.005 0.002 0.002 0.060** 0.048** 0.636*** 0.678*** (0.027) (0.024) (0.044) (0.045) (0.014) (0.012) (0.005) (0.005) (0.028) (0.020) (0.067) (0.044) 2 Urban Upper Egypt (d) 0.074*** 0.079*** 0.041 0.036 –0.008 –0.006 0.012 0.012 0.125*** 0.110*** 0.446*** 0.481*** (0.028) (0.027) (0.045) (0.045) (0.014) (0.012) (0.008) (0.008) (0.041) (0.036) (0.071) (0.062) 2 Rural Upper Egypt (d) 0.142** 0.180*** –0.031 –0.013 –0.018 –0.028** 0.034*** 0.034*** 0.140*** 0.140*** 0.575*** 0.637*** (0.055) (0.027) (0.046) (0.047) (0.017) (0.013) (0.012) (0.012) (0.034) (0.034) (0.077) (0.048) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5541 5541 1711 1711 5541 5541 5541 5541 5541 5541 5541 5541 rho=0: Prob > chi2 0.3623 0.1630 0.0598 0.9585 0.1610 0.1233 Note: Since there may be more than one young woman living in a household, we take clustering at the household level into consideration. (d) for discrete change of dummy variable from 0 to 1, * p<0.10, ** p<0.05, *** p<0.01 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 49 5/27/10 2:41 PM 50 Keller MNA 5-27-10vol2.indd 50 Table A19: The Impact of the Value of Remittances Sent on Male Employment (Young Men Aged 16–34, Out of School) Results from an IV Probit estimation with a continuous endogenous variable (to control for remittance income) and a binary outcome variable reflecting a particular employment status, e.g. participating in the labor force, using the ivprobit command in stata. The alternative specification is a probit model assuming that remittance- income is exogenous to the outcome variable. Standard errors in parentheses. Empl. in wage & salary Self-employed / In the labor force Unemployed work employer Unpaid family work ivprobit probit ivprobit probit ivprobit probit ivprobit probit ivprobit probit Average monthly remittance- –0.045 –0.003 0.045 0.003 –0.223** –0.032*** 0.090 0.008* 0.077 0.005** income (in 100 LE)3 (0.048) (0.003) (0.080) (0.003) (0.098) (0.009) (0.136) (0.004) (0.093) (0.003) Age –0.006 –0.006 –0.007 –0.006 0.028* 0.030* 0.069*** 0.069*** –0.007 –0.006 (0.009) (0.009) (0.007) (0.006) (0.016) (0.016) (0.013) (0.012) (0.009) (0.007) Age squared 0.000 0.000* 0.000 0.000 –0.000 –0.000 –0.001*** –0.001*** 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 1 –0.000 Primary or preparatory degree 0.021** 0.019** –0.009 –0.006 0.046** 0.049** –0.015 –0.014 –0.001 (d) (0.010) (0.009) (0.015) (0.012) (0.022) (0.024) (0.015) (0.014) (0.013) (0.011) 1 Secondary degree (d) 0.044*** 0.040*** 0.059*** 0.057*** 0.007 –0.004 –0.033** –0.028** 0.014 0.016* (0.011) (0.008) (0.018) (0.013) (0.021) (0.021) (0.015) (0.012) (0.011) (0.009) Above secondary degree1 (d) 0.038*** 0.032*** 0.143*** 0.148*** 0.044* 0.020 –0.095*** –0.085*** –0.025 –0.015 (0.013) (0.009) (0.026) (0.026) (0.026) (0.024) (0.024) (0.011) (0.020) (0.011) Ever-married (d) 0.092*** 0.090*** –0.072*** –0.066*** 0.063** 0.089*** 0.094*** 0.086*** –0.033*** –0.033*** (0.012) (0.010) (0.020) (0.009) (0.026) (0.020) (0.019) (0.014) (0.012) (0.010) Nr of females in hh 0.003 0.001 0.009** 0.009*** –0.011 –0.020** –0.021*** –0.018*** 0.024*** 0.023***   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options (0.004) (0.004) (0.004) (0.003) (0.010) (0.009) (0.008) (0.006) (0.005) (0.004) Nr of children aged 0–5 in hh –0.002 –0.004 –0.019 –0.014*** 0.011 –0.001 –0.003 0.002 –0.008 –0.003 (0.005) (0.004) (0.012) (0.004) (0.012) (0.010) (0.010) (0.006) (0.008) (0.005) Nr of children aged 6–14 in hh 0.002 0.002 –0.003 –0.002 –0.018* –0.023** –0.006 –0.005 0.019*** 0.017*** (0.004) (0.004) (0.005) (0.004) (0.010) (0.010) (0.008) (0.007) (0.005) (0.004) Presence of elderly in hh (d) –0.005 –0.005 –0.001 –0.001 –0.022 –0.025 –0.031** –0.030** 0.046*** 0.041*** (0.009) (0.009) (0.008) (0.007) (0.020) (0.021) (0.014) (0.013) (0.015) (0.013) (continued on next page) 5/27/10 2:41 PM Table A19: The Impact of the Value of Remittances Sent on Male Employment (Young Men Aged 16–34, Out of School) (continued) Keller MNA 5-27-10vol2.indd 51 Results from an IV Probit estimation with a continuous endogenous variable (to control for remittance income) and a binary outcome variable reflecting a particular employment status, e.g. participating in the labor force, using the ivprobit command in stata. The alternative specification is a probit model assuming that remittance- income is exogenous to the outcome variable. Standard errors in parentheses. Empl. in wage & salary Self-employed / In the labor force Unemployed work employer Unpaid family work ivprobit probit ivprobit probit ivprobit probit ivprobit probit ivprobit probit 2 Alexandria and Suez (d) 0.001 –0.001 0.015 0.014 –0.028 –0.040 0.042 0.045* –0.037 –0.029* (0.015) (0.014) (0.014) (0.012) (0.031) (0.031) (0.026) (0.025) (0.024) (0.017) 2 urban Lower Egypt (d) –0.009 –0.011 –0.004 –0.001 –0.102*** –0.129*** 0.100*** 0.107*** 0.049* 0.050* (0.015) (0.014) (0.012) (0.009) (0.035) (0.029) (0.028) (0.026) (0.030) (0.028) 2 rural Lower Egypt (d) –0.004 –0.006 –0.012 –0.008 –0.111*** –0.135*** 0.065*** 0.068*** 0.115*** 0.113*** (0.012) (0.012) (0.011) (0.008) (0.031) (0.026) (0.021) (0.020) (0.029) (0.028) urban Upper Egypt2 (d) –0.021 –0.021 –0.001 –0.000 –0.134*** –0.155*** 0.105*** 0.107*** 0.080** 0.076** (0.015) (0.015) (0.010) (0.009) (0.033) (0.028) (0.025) (0.025) (0.031) (0.029) 2 0.123*** rural Upper Egypt (d) –0.014 –0.018 –0.034 –0.026*** –0.088*** –0.125*** 0.064** 0.075*** 0.117*** (0.014) (0.013) (0.021) (0.007) (0.034) (0.027) (0.027) (0.022) (0.030) (0.030) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5460 5460 4898 4898 5460 5460 5460 5460 5460 5460 rho=0: Prob > chi2 0.3125 0.4873 0.0707 0.5083 0.2991 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. (d) for discrete change of dummy variable from 0 to 1, * p<0.10, ** p<0.05, *** p<0.01 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 51 5/27/10 2:41 PM 52 Keller MNA 5-27-10vol2.indd 52 Table A20: The Impact of Migration on Hours Worked by Young Women (Aged 16–34, Out of School) — Coefficients Results from an IV Tobit estimation with endogenous binary variable (to control for migration) and the left-censored outcome variable hours worked of women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit 3 Migrant hh 44.751*** –0.758 –5.301 –17.606** 85.845 11.196 80.921*** 11.669* 13.487*** 2.221** –3.461 3.171 (16.470) (5.172) (45.090) (8.426) (77.665) (11.854) (17.178) (6.307) (3.684) (1.068) (6.245) (2.450) Age 2.305 2.543 –1.934 –1.880 2.602 2.916 6.793* 7.315* –0.007 0.023 4.227*** 4.193*** (2.758) (2.746) (3.918) (3.917) (7.268) (7.269) (4.003) (3.962) (0.618) (0.608) (1.064) (1.064) Age squared 0.005 0.001 0.117 0.116 0.005 –0.001 –0.140* –0.149* 0.003 0.002 –0.077*** –0.077*** (0.054) (0.053) (0.075) (0.075) (0.139) (0.139) (0.079) (0.078) (0.012) (0.012) (0.021) (0.021) Primary or preparatory –11.896*** –11.797*** 13.695* 13.737* –23.436** –23.188** –13.379*** –13.681*** –2.970*** –2.992*** 0.903 0.885 degree1 (4.033) (4.051) (7.639) (7.642) (10.670) (10.672) (4.751) (4.777) (0.796) (0.794) (1.514) (1.509) 1 Secondary degree –3.792 –2.979 42.875*** 43.125*** –20.857** –19.852** –36.363*** –35.499*** –5.825*** –5.596*** 3.026** 2.870** (3.029) (3.014) (5.727) (5.726) (8.333) (8.127) (4.313) (4.325) (0.677) (0.660) (1.218) (1.212) Above secondary degree1 23.016*** 23.651*** 75.193*** 75.415*** –25.536** –24.673** –71.288*** –69.929*** –12.640*** –12.439*** –2.444 –2.580* (3.455) (3.420) (5.521) (5.516) (10.919) (10.860) (10.756) (10.713) (1.107) (1.095) (1.520) (1.506) Ever-married –29.012*** –28.693*** –44.172*** –44.040*** 18.169 18.513 –6.234 –6.233 1.730** 1.817** 21.612*** 21.544*** (3.465) (3.448) (4.695) (4.683) (14.187) (14.130) (5.273) (5.244) (0.824) (0.817) (1.351) (1.337) Nr of females in hh 1.268 1.914 2.842 3.038 –16.287*** –15.601*** –0.546 0.290 0.253 0.423 –6.468*** –6.582***   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options (1.516) (1.493) (2.149) (2.025) (5.901) (5.808) (1.901) (1.851) (0.339) (0.320) (0.579) (0.570) Nr of children aged 0–5 –3.462** –3.114** –8.872*** –8.787*** –12.292*** –11.829*** 2.558 3.308* 0.728** 0.830*** 8.375*** 8.320*** in hh (1.374) (1.383) (2.017) (2.002) (3.876) (3.807) (1.763) (1.726) (0.305) (0.307) (0.593) (0.588) Nr of children aged 6–14 3.550*** 3.550*** –3.891* –3.874* 7.422*** 7.363*** 6.606*** 6.573*** 0.726*** 0.727*** –0.035 –0.037 in hh (1.243) (1.235) (2.140) (2.137) (2.779) (2.779) (1.588) (1.571) (0.276) (0.269) (0.491) (0.489) Presence of elderly in hh –1.109 0.070 –8.089* –7.728 –16.190 –14.937 7.412 8.737* 1.149 1.509* 3.594** 3.374** (3.499) (3.429) (4.867) (4.711) (11.143) (11.217) (4.763) (4.641) (0.814) (0.787) (1.413) (1.434) (continued on next page) 5/27/10 2:41 PM Table A20: The Impact of Migration on Hours Worked by Young Women (Aged 16–34, Out of School) — Coefficients (continued) Keller MNA 5-27-10vol2.indd 53 Results from an IV Tobit estimation with endogenous binary variable (to control for migration) and the left-censored outcome variable hours worked of women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit 2 Alexandria and Suez 1.042 1.153 1.579 1.635 –5.693 –5.613 6.523 6.271 6.086** 6.020** 2.591 2.559 (4.857) (4.817) (5.694) (5.679) (20.692) (20.549) (16.909) (16.593) (2.945) (2.913) (2.293) (2.298) 2 Urban Lower Egypt –7.628 –6.360 –8.061 –7.617 –3.068 –1.967 22.274* 22.466* 19.022*** 19.063*** –10.432*** –10.683*** (4.806) (4.753) (5.832) (5.632) (18.988) (18.873) (13.192) (12.971) (2.563) (2.533) (2.084) (2.081) Rural Lower Egypt2 –4.878 –3.712 –1.548 –1.154 2.614 3.357 32.299*** 31.667*** 27.940*** 27.954*** –3.916** –4.142** (4.196) (4.147) (5.280) (5.061) (14.309) (14.199) (11.511) (11.309) (2.481) (2.450) (1.986) (2.002) 2 Urban Upper Egypt 2.496 3.110 –6.381 –6.192 24.335* 24.732* 51.100*** 50.295*** 20.085*** 19.979*** –15.527*** –15.643*** (4.287) (4.233) (5.112) (5.073) (14.383) (14.266) (11.610) (11.389) (2.548) (2.515) (1.837) (1.838) 2 Rural Upper Egypt 13.423*** 16.014*** –17.944*** –17.159*** 53.404*** 56.041*** 61.478*** 62.714*** 26.408*** 26.768*** –18.051*** –18.507*** (4.196) (4.110) (6.839) (6.312) (13.036) (12.716) (11.007) (10.783) (2.464) (2.441) (1.906) (1.880) _cons –95.241*** –99.848*** –103.845** –105.080** –225.968** –230.207** –196.723*** –202.875*** –32.384*** –32.950*** –12.833 –12.043 (34.058) (34.014) (50.151) (50.108) (90.808) (90.837) (49.252) (48.870) (7.815) (7.714) (12.546) (12.561) lnsig_1 4.069*** 57.860*** 4.187*** 65.748*** 4.355*** 76.981*** 4.050*** 55.963*** 2.615*** 13.467*** 3.520*** 33.768*** _cons (0.017) (0.965) (0.022) (1.402) (0.063) (4.672) (0.039) (2.178) (0.033) (0.426) (0.020) (0.686) atanhrho_12 –0.396*** –0.091 –0.476 –0.661*** –0.441*** 0.099 _cons (0.136) (0.323) (0.459) (0.156) (0.135) (0.072) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5551 5551 5551 5551 5551 5551 5551 5551 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 53 5/27/10 2:41 PM 54 Keller MNA 5-27-10vol2.indd 54 Table A21: The Impact of Migration on Hours Worked by Young Men (Aged 16–34, Out of School) – Coefficients Results from an IV Tobit estimation with endogenous binary variable (to control for migration) and the left-censored outcome variable hours worked of men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work cmp tobit cmp tobit cmp tobit cmp tobit 3 Current international migrant in hh –4.154 –2.078 –20.218* –12.959*** 61.609 18.903* –166.678*** 18.632* (3.964) (2.794) (10.878) (4.630) (39.986) (9.743) (9.869) (10.580) Age 3.827*** 3.833*** 4.356*** 4.377*** 27.186*** 27.162*** –2.653 –2.218 (0.896) (0.895) (1.511) (1.511) (4.976) (4.979) (6.476) (4.532) Age squared –0.055*** –0.055*** –0.063** –0.063** –0.436*** –0.437*** –0.053 –0.028 (0.017) (0.017) (0.029) (0.029) (0.092) (0.092) (0.129) (0.091) 1 Primary or preparatory degree 2.973** 2.966** 5.513** 5.486** –5.856 –5.869 0.471 0.485 (1.270) (1.271) (2.224) (2.227) (5.666) (5.656) (10.300) (7.255) 1 Secondary degree –0.435 –0.455 0.942 0.870 –11.940** –11.782** 18.155** 8.215 (1.142) (1.143) (1.962) (1.962) (4.718) (4.714) (8.593) (6.023) 1 Above secondary degree –8.563*** –8.618*** –0.006 –0.197 –39.239*** –38.590*** 8.292 –13.989* (1.312) (1.314) (2.167) (2.159) (5.898) (5.877) (11.503) (8.203) Ever-married 11.917*** 11.968*** 8.193*** 8.371*** 33.326*** 32.513*** –45.538*** –22.660*** (1.041) (1.040) (1.861) (1.841) (5.138) (5.126) (9.161) (6.689) Nr of females in hh –0.314 –0.333 –1.422 –1.488* –7.375*** –7.116*** 23.141*** 15.887*** (0.532) (0.532) (0.899) (0.893) (2.423) (2.407) (3.623) (2.545)   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Nr of children aged 0–5 in hh 0.200 0.181 –0.214 –0.277 0.591 0.878 2.812 –2.630 (0.464) (0.464) (0.902) (0.896) (2.181) (2.182) (4.199) (3.033) Nr of children aged 6–14 in hh 0.088 0.083 –2.600*** –2.615*** –1.631 –1.590 16.283*** 11.472*** (0.539) (0.540) (0.945) (0.944) (3.089) (3.090) (3.453) (2.488) Presence of elderly in hh –1.028 –1.063 –2.723 –2.848 –14.764*** –14.325** 37.092*** 22.856*** (1.216) (1.214) (2.040) (2.033) (5.646) (5.622) (8.926) (6.213) (continued on next page) 5/27/10 2:41 PM Table A21: The Impact of Migration on Hours Worked by Young Men (Aged 16–34, Out of School) – Coefficients (continued) Keller MNA 5-27-10vol2.indd 55 Results from an IV Tobit estimation with endogenous binary variable (to control for migration) and the left-censored outcome variable hours worked of men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work cmp tobit cmp tobit cmp tobit cmp tobit Alexandria and Suez2 –3.510** –3.516** –5.175** –5.193** 15.402* 15.547* –53.057***4 –25.083 (1.730) (1.731) (2.482) (2.482) (8.496) (8.475) (7.618) (18.462) 2 Urban Lower Egypt –5.143*** –5.174*** –13.481*** –13.589*** 31.435*** 31.805*** 27.524** (1.621) (1.620) (2.473) (2.468) (7.251) (7.248) (13.966) 2 Rural Lower Egypt –4.214*** –4.248*** –13.526*** –13.644*** 21.537*** 22.003*** 55.185*** (1.412) (1.413) (2.206) (2.202) (6.739) (6.725) (11.850) Urban Upper Egypt2 –7.155*** –7.178*** –17.280*** –17.361*** 32.173*** 32.507*** 38.895*** (1.571) (1.571) (2.431) (2.428) (7.136) (7.117) (12.897) 2 Rural Upper Egypt –5.724*** –5.809*** –15.874*** –16.169*** 22.226*** 23.545*** 56.857*** (1.459) (1.455) (2.353) (2.332) (7.296) (7.186) (12.096) Constant –18.732 –18.835 –39.547** –39.916** –501.152*** –499.977*** –76.293 –107.126* (11.781) (11.768) (19.676) (19.670) (66.515) (66.546) (79.031) (55.217) lnsig_1 3.302*** 27.161*** 3.755*** 42.736*** 4.414*** 82.397*** 4.826*** 82.011*** _cons (0.013) (0.364) (0.013) (0.551) (0.017) (1.416) (0.045) (2.122) atanhrho_12 0.037 0.082 –0.238 3.803*** _cons (0.049) (0.106) (0.209) (0.430) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5467 5467 5467 5467 5467 5467 5467 5467 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. 4 Due to too few ob- servations in some of the region categories, urban and rural regions are lumped together. The coefficient hence refers to an urban dummy. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 55 5/27/10 2:41 PM 56 Keller MNA 5-27-10vol2.indd 56 Table A22: The Impact of Remittances on Hours Worked by Young Women (Aged 16–34, Out of School) – Coefficients Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit 3 Remittance-receiving hh 38.635** –5.382 21.986 –13.025 90.146 17.568 54.798*** –1.310 9.845* 1.550 –6.591 1.969 (d) (18.361) (6.326) (32.050) (10.164) (78.319) (12.416) (20.099) (7.678) (5.140) (1.206) (5.300) (2.935) Age 2.404 2.572 –2.013 –1.899 2.393 2.657 7.267* 7.412* 0.011 0.026 4.242*** 4.200*** (2.758) (2.748) (3.917) (3.912) (7.289) (7.283) (3.972) (3.949) (0.613) (0.608) (1.065) (1.064) Age squared 0.003 0.000 0.118 0.116 0.008 0.004 –0.149* –0.151* 0.002 0.002 –0.077*** –0.077*** (0.053) (0.053) (0.075) (0.075) (0.139) (0.139) (0.078) (0.078) (0.012) (0.012) (0.021) (0.021) Primary or preparatory –11.937*** –11.765*** 13.734* 13.885* –23.857** –23.387** –13.450*** –13.354*** –3.007*** –2.991*** 0.922 0.885 degree1 (d) (4.042) (4.054) (7.660) (7.654) (10.689) (10.685) (4.779) (4.793) (0.797) (0.795) (1.515) (1.509) 1 Secondary degree (d) –3.473 –2.923 42.686*** 43.125*** –20.932** –19.951** –35.515*** –34.977*** –5.695*** –5.560*** 3.061** 2.910** (3.022) (3.010) (5.740) (5.737) (8.344) (8.141) (4.343) (4.336) (0.674) (0.660) (1.209) (1.209) Above secondary degree1 23.337*** 23.666*** 75.079*** 75.368*** –25.280** –24.705** –70.264*** –69.702*** –12.498*** –12.415*** –2.436 –2.538* (d) (3.437) (3.417) (5.528) (5.522) (10.886) (10.885) (10.743) (10.714) (1.100) (1.095) (1.508) (1.504) Ever-married (d) –28.851*** –28.663*** –44.202*** –43.974*** 18.522 18.595 –6.024 –5.981 1.762** 1.815** 21.623*** 21.560*** (3.465) (3.444) (4.699) (4.690) (14.202) (14.215) (5.271) (5.238) (0.822) (0.818) (1.350) (1.337)   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Nr of females in hh 1.652 1.939 2.702 2.937 –15.812*** –15.635*** 0.380 0.711 0.392 0.453 –6.472*** –6.544*** (1.508) (1.485) (2.053) (2.032) (5.784) (5.814) (1.864) (1.834) (0.332) (0.320) (0.578) (0.571) Nr of children aged 0–5 –3.443** –3.061** –9.121*** –8.850*** –12.488*** –11.961*** 2.813 3.407** 0.742** 0.832*** 8.413*** 8.327*** in hh (1.375) (1.385) (2.011) (2.006) (3.853) (3.798) (1.747) (1.735) (0.310) (0.309) (0.595) (0.591) Nr of children aged 6–14 3.665*** 3.547*** –3.846* –3.884* 7.608*** 7.366*** 6.759*** 6.560*** 0.758*** 0.729*** –0.053 –0.031 in hh (1.239) (1.234) (2.143) (2.140) (2.779) (2.777) (1.586) (1.576) (0.271) (0.269) (0.490) (0.488) Presence of elderly in hh –0.059 0.094 –8.245* –8.100* –14.770 –14.848 9.138** 9.149** 1.534* 1.585** 3.522** 3.467** (d) (3.438) (3.420) (4.734) (4.716) (11.222) (11.253) (4.639) (4.614) (0.795) (0.785) (1.449) (1.444) (continued on next page) 5/27/10 2:41 PM Table A22: The Impact of Remittances on Hours Worked by Young Women (Aged 16–34, Out of School) – Coefficients (continued) Keller MNA 5-27-10vol2.indd 57 Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit cmp tobit Alexandria and Suez2 (d) 0.948 1.201 1.460 1.729 –5.946 –5.858 6.341 6.334 6.022** 6.025** 2.642 2.554 (4.836) (4.816) (5.689) (5.674) (20.637) (20.580) (16.724) (16.589) (2.927) (2.915) (2.290) (2.297) 2 Urban Lower Egypt (d) –7.379 –6.201 –8.825 –7.761 –3.412 –2.223 22.444* 22.952* 18.964*** 19.102*** –10.313*** –10.638*** (4.784) (4.752) (5.692) (5.643) (18.939) (18.899) (13.058) (12.966) (2.544) (2.535) (2.077) (2.081) 2 Rural Lower Egypt (d) –4.540 –3.595 –2.197 –1.358 2.340 3.183 32.102*** 32.158*** 27.903*** 28.003*** –3.831* –4.094** (4.175) (4.144) (5.132) (5.067) (14.321) (14.276) (11.398) (11.313) (2.462) (2.452) (1.989) (2.002) 2 Urban Upper Egypt (d) 2.788 3.148 –6.569 –6.289 24.357* 24.591* 50.988*** 50.621*** 20.038*** 20.023*** –15.505*** –15.611*** (4.253) (4.231) (5.093) (5.078) (14.361) (14.290) (11.475) (11.385) (2.528) (2.516) (1.836) (1.838) 2 Rural Upper Egypt (d) 14.433*** 16.202*** –19.002*** –17.597*** 53.617*** 55.797*** 62.735*** 63.824*** 26.619*** 26.876*** –17.955*** –18.388*** (4.161) (4.106) (6.380) (6.314) (13.009) (12.751) (10.886) (10.790) (2.452) (2.444) (1.875) (1.875) Constant –97.197*** –100.318*** –102.520** –104.808** –223.100** –226.520** –203.523*** –205.568*** –32.741*** –33.091*** –13.083 –12.226 (34.099) (34.033) (50.078) (50.027) (91.268) (91.136) (49.028) (48.790) (7.764) (7.715) (12.559) (12.557) lnsig_1 constant 4.065*** 57.844*** 4.190*** 65.807*** 4.352*** 76.992*** 4.036*** 56.026*** 2.607*** 13.475*** 3.521*** 33.773*** (0.017) (0.965) (0.022) (1.403) (0.062) (4.684) (0.039) (2.172) (0.033) (0.426) (0.020) (0.687) atanhrho_12 constant –0.362** –0.242 –0.442 –0.493*** –0.304* 0.123** (0.145) (0.211) (0.442) (0.173) (0.177) (0.048) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5551 5551 5551 5551 5551 5551 5551 5551 5551 5551 5551 5551 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1. 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 57 5/27/10 2:41 PM 58 Keller MNA 5-27-10vol2.indd 58 Table A23: The Impact of the Value of Remittances Sent on Hours Worked by Young Women (Aged 16–34, Out of School) – Coefficients Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit Average monthly 6.105 –0.082 –29.392* –1.818 0.639 –0.523 17.521 0.643 4.029 0.187 –0.400 0.166 remittance-income (in (11.939) (0.820) (16.090) (1.385) (18.551) (1.460) (16.902) (0.845) (3.510) (0.220) (2.105) (0.211) 100 LE)3 Age 2.498 2.523 –1.554 –2.005 3.064 3.035 7.342* 7.269* –0.009 0.003 4.241*** 4.237*** (2.759) (2.745) (4.188) (3.914) (7.328) (7.306) (4.057) (3.960) (0.626) (0.608) (1.066) (1.066) Age squared 0.002 0.001 0.110 0.118 –0.003 –0.003 –0.149* –0.148* 0.003 0.003 –0.078*** –0.077*** (0.054) (0.053) (0.081) (0.075) (0.140) (0.139) (0.080) (0.078) (0.012) (0.012) (0.021) (0.021) Primary or preparatory –12.658*** –11.785*** 18.139** 14.054* –23.255** –23.109** –15.807*** –13.463*** –3.471*** –3.041*** 0.946 0.870 degree1 (d) (4.408) (4.051) (7.803) (7.660) (11.145) (10.694) (5.578) (4.805) (0.977) (0.797) (1.551) (1.513) Secondary degree1 (d) –3.728 –2.955 47.166*** 43.185*** –19.806** –19.655** –37.107*** –35.051*** –6.079*** –5.564*** 3.024** 2.944** (3.352) (3.011) (5.801) (5.735) (8.605) (8.120) (4.893) (4.358) (0.813) (0.662) (1.255) (1.215) Above secondary degree1 22.384*** 23.689*** 81.700*** 75.601*** –24.713** –24.496** –72.897*** –69.931*** –13.103*** –12.402*** –2.429 –2.555* (d) (4.089) (3.419) (6.183) (5.523) (11.859) (10.873) (10.948) (10.752) (1.232) (1.096) (1.584) (1.507) Ever-married (d) –29.186*** –28.749*** –43.319*** –43.941*** 18.124 18.194 –7.053 –5.781 1.681* 1.871** 21.649*** 21.614*** (3.510) (3.446) (5.019) (4.690) (14.386) (14.218) (5.401) (5.269) (0.859) (0.820) (1.352) (1.337) Nr of females in hh 1.021 1.868 5.509** 2.994 –15.781** –15.582*** –2.239 0.624 –0.107 0.459 –6.472*** –6.549***   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options (1.869) (1.488) (2.549) (2.034) (6.986) (5.852) (2.138) (1.850) (0.383) (0.320) (0.645) (0.575) Nr of children aged 0–5 –3.910** –3.091** –5.613* –8.836*** –11.904*** –11.749*** 1.170 3.312* 0.343 0.816*** 8.409*** 8.339*** in hh (1.945) (1.386) (3.098) (2.009) (4.560) (3.792) (2.575) (1.749) (0.496) (0.310) (0.652) (0.594) Nr of children aged 6–14 3.424*** 3.536*** –3.854* –3.881* 7.386*** 7.387*** 6.093*** 6.539*** 0.654** 0.730*** –0.035 –0.046 in hh (1.239) (1.234) (2.237) (2.137) (2.786) (2.788) (1.644) (1.582) (0.275) (0.269) (0.492) (0.489) Presence of elderly in 0.252 –0.029 –10.268* –8.337* –14.344 –14.333 9.929** 9.111** 1.781** 1.630** 3.470** 3.495** hh (d) (3.441) (3.419) (5.964) (4.719) (11.217) (11.218) (4.891) (4.630) (0.868) (0.785) (1.457) (1.452) (continued on next page) 5/27/10 2:41 PM (continued Table A23: The Impact of the Value of Remittances Sent on Hours Worked by Young Women (Aged 16–34, Out of School) – ) Coefficients Keller MNA 5-27-10vol2.indd 59 Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for women aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Self-employed / Any market work Wage & salary work employer Unpaid family work Subsistence work Domestic work ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit 2 Alexandria and Suez (d) 1.164 1.085 1.070 1.514 –5.495 –5.499 6.176 6.362 6.047** 6.056** 2.524 2.527 (4.833) (4.817) (6.149) (5.684) (20.528) (20.523) (16.527) (16.617) (2.911) (2.916) (2.296) (2.298) Urban Lower Egypt2 (d) –7.182 –6.415 –4.935 –8.153 –1.559 –1.439 20.322 22.923* 18.556*** 19.163*** –10.521*** –10.588*** (4.986) (4.749) (6.251) (5.635) (18.972) (18.854) (13.131) (12.983) (2.599) (2.536) (2.101) (2.080) Rural Lower Egypt2 (d) –3.727 –3.689 –1.078 –1.612 4.094 4.107 31.421*** 32.230*** 27.752*** 28.008*** –4.036** –4.045** (4.183) (4.142) (5.384) (5.061) (14.228) (14.235) (11.378) (11.328) (2.476) (2.453) (2.013) (2.012) 2 Urban Upper Egypt (d) 3.166 3.160 –6.873 –6.416 25.062* 25.070* 49.537*** 50.769*** 19.807*** 20.009*** –15.676*** –15.670*** (4.274) (4.232) (5.394) (5.075) (14.244) (14.246) (11.407) (11.403) (2.520) (2.519) (1.841) (1.840) 2 Rural Upper Egypt (d) 14.260*** 16.016*** –11.616 –17.867*** 56.675*** 56.916*** 57.460*** 63.564*** 25.609*** 26.894*** –18.263*** –18.421*** (5.037) (4.102) (7.195) (6.307) (13.299) (12.672) (11.746) (10.801) (2.591) (2.445) (1.969) (1.874) Constant –96.828*** –99.473*** –116.926** –103.457** –231.500** –231.580** –196.026*** –204.001*** –31.054*** –32.897*** –12.964 –12.693 (34.380) (33.983) (53.610) (50.045) (91.248) (91.256) (50.243) (48.901) (7.966) (7.720) (12.618) (12.584) alpha _cons / sigma –6.411 57.826*** 27.866* 65.776*** –1.197 76.983*** –17.608 56.127*** –3.997 13.481*** 0.586 33.783*** constant (11.709) (0.965) (16.005) (1.403) (18.873) (4.662) (16.480) (2.178) (3.358) (0.428) (2.188) (0.689) lns_cons 4.058*** 4.182*** 4.344*** 4.021*** 2.595*** 3.520*** (0.017) (0.021) (0.061) (0.039) (0.032) (0.020) lnv _cons 0.487*** 0.487*** 0.487*** 0.487*** 0.487*** 0.487*** (0.154) (0.154) (0.154) (0.154) (0.154) (0.154) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5541 5541 5541 5541 5541 5541 5541 5541 5541 5541 5541 5541 Wald test of exogeneity 0.5840 0.0817 0.9494 0.2853 0.2339 0.7889 (Prob >chi2) Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 59 5/27/10 2:41 PM 60 Keller MNA 5-27-10vol2.indd 60 Table A24: The Impact of Remittances on Hours Worked by Young Men (Aged 16–34, Out of School) – Coefficients Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work cmp tobit cmp tobit cmp tobit cmp4 tobit 3 28.447** Remittance-receiving hh (d) –3.555 –1.673 –27.949 –15.382** 59.664 10.062 –149.791*** (5.831) (3.975) (30.793) (6.275) (53.312) (13.354) (10.150) (13.697) Age 3.825*** 3.833*** 4.297*** 4.358*** 27.107*** 27.042*** –3.837 –2.233 (0.896) (0.896) (1.514) (1.512) (4.958) (4.966) (5.924) (4.530) Age squared –0.055*** –0.055*** –0.062** –0.063** –0.436*** –0.435*** –0.024 –0.028 (0.017) (0.017) (0.029) (0.029) (0.092) (0.092) (0.118) (0.091) Primary or preparatory de- 2.982** 2.970** 5.627** 5.543** –6.106 –5.932 0.164 gree1 (d) (1.272) (1.272) (2.232) (2.230) (5.665) (5.652) (7.254) 1 –11.690** Secondary degree (d) –0.445 –0.462 0.984 0.867 –11.953** 8.060 (1.144) (1.144) (1.981) (1.961) (4.720) (4.714) (6.014) 1 Above secondary degree (d) –8.620*** –8.649*** –0.124 –0.317 –38.777*** –38.318*** –23.055*** –13.905* (1.315) (1.313) (2.216) (2.159) (5.874) (5.868) (8.145) (8.183) Ever-married (d) 11.972*** 11.997*** 8.331*** 8.496*** 32.656*** 32.190*** –32.954*** –22.516*** (1.041) (1.040) (1.893) (1.837) (5.112) (5.119) (8.096) (6.675) Nr of females in hh –0.340 –0.346 –1.510* –1.553* –7.020*** –6.921*** 22.294*** 16.024*** (0.530) (0.531) (0.897) (0.891) (2.403) (2.398) (3.299) (2.534)   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Nr of children aged 0–5 in hh 0.183 0.172 –0.232 –0.303 0.753 0.950 4.483 –2.619 (0.464) (0.463) (0.920) (0.898) (2.178) (2.184) (3.820) (3.026) Nr of children aged 6–14 in hh 0.086 0.083 –2.579*** –2.603*** –1.628 –1.593 17.370*** 11.473*** (0.541) (0.541) (0.949) (0.943) (3.084) (3.086) (3.154) (2.473) Presence of elderly in hh –1.095 –1.097 –3.039 –3.052 –14.013** –14.019** 23.209*** (1.214) (1.214) (2.034) (2.034) (5.618) (5.611) (6.200) (continued on next page) 5/27/10 2:41 PM Table A24: The Impact of Remittances on Hours Worked by Young Men (Aged 16–34, Out of School) – Coefficients (continued) Keller MNA 5-27-10vol2.indd 61 Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work cmp tobit cmp tobit cmp tobit cmp4 tobit 2 15.456* Alexandria and Suez (d) –3.500** –3.512** –5.077** –5.156** 15.255* –25.582 (1.731) (1.731) (2.498) (2.485) (8.484) (8.469) (18.459) 2 31.953*** Urban Lower Egypt (d) –5.151*** –5.180*** –13.363*** –13.565*** 31.508*** 26.921* (1.621) (1.620) (2.528) (2.470) (7.263) (7.250) (13.961) 2 22.065*** Rural Lower Egypt (d) –4.218*** –4.252*** –13.370*** –13.602*** 21.539*** 54.584*** (1.413) (1.413) (2.285) (2.205) (6.738) (6.726) (11.836) Urban Upper Egypt2 (d) –7.160*** –7.182*** –17.202*** –17.354*** 32.179*** 32.551*** 38.293*** (1.572) (1.572) (2.466) (2.430) (7.134) (7.116) (12.901) 2 24.007*** Rural Upper Egypt (d) –5.826*** –5.862*** –16.177*** –16.421*** 23.331*** 56.668*** (1.454) (1.452) (2.412) (2.335) (7.198) (7.163) (12.071) Constant –18.741 –18.859 –38.939** –39.765** –499.239*** –497.976*** –63.395 –106.348* (11.779) (11.770) (19.713) (19.688) (66.253) (66.356) (72.066) (55.229) lnsig_1 constant / signa cons 3.302*** 27.163*** 3.756*** 42.747*** 4.414*** 82.431*** 4.714*** 81.926*** (0.013) (0.364) (0.013) (0.552) (0.017) (1.413) (0.042) (2.124) atanhrho_12 constant 0.030 0.129 –0.249 3.698*** (0.062) (0.308) (0.243) (0.392) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5467 5467 5467 5467 5467 5467 5467 5467 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. 4 Due to too few observations, the educational categories have been lumped together (the reference category now includes all men with secondary degree and less) and the variable “presence of elderly in hhâ€? and the regional dummies have been dropped. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 61 5/27/10 2:41 PM 62 Keller MNA 5-27-10vol2.indd 62 Table A25: The Impact of the Value of Remittances Sent on Hours Worked by Young Men (Aged 16–34, Out of School) – Coefficients Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit Average monthly remittance- –18.220** –0.468 –29.759** –3.593*** 18.936 2.680* 33.204 3.386** income (in 100 LE)3 (9.294) (0.578) (14.525) (1.026) (45.462) (1.391) (38.486) (1.638) Age 3.998*** 3.851*** 4.580*** 4.342*** 26.955*** 27.083*** –2.373 –2.363 (1.092) (0.897) (1.766) (1.512) (4.995) (4.979) (4.686) (4.553) Age squared –0.060*** –0.056*** –0.069** –0.062** –0.432*** –0.436*** –0.024 –0.026 (0.021) (0.017) (0.034) (0.029) (0.093) (0.092) (0.094) (0.091) 1 –0.290 Primary or preparatory degree (d) 3.140** 2.940** 5.805** 5.489** –5.753 –5.618 –0.534 (1.473) (1.273) (2.460) (2.230) (5.729) (5.662) (7.412) (7.285) 1 Secondary degree (d) 0.499 –0.440 2.269 0.872 –12.599** –11.760** 6.602 8.132 (1.455) (1.145) (2.382) (1.962) (5.311) (4.728) (6.590) (6.022) 1 Above secondary degree (d) –6.280*** –8.623*** 3.268 –0.192 –40.539*** –38.426*** –17.924* –14.014* (2.136) (1.314) (3.340) (2.161) (8.086) (5.882) (10.044) (8.217) Ever-married (d) 10.522*** 11.982*** 6.304*** 8.454*** 33.672*** 32.357*** –20.256*** –22.648*** (1.530) (1.039) (2.406) (1.837) (6.178) (5.107) (7.363) (6.676) Nr of females in hh 0.258 –0.330 –0.668 –1.533* –7.481*** –6.914*** 15.108*** 16.077*** (0.675) (0.532) (1.078) (0.892) (2.845) (2.405) (2.915) (2.534)   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Nr of children aged 0–5 in hh 1.229 0.200 1.268 –0.254 –0.143 0.803 –4.481 –2.729 (0.958) (0.465) (1.475) (0.897) (3.387) (2.175) (3.966) (3.044) Nr of children aged 6–14 in hh 0.271 0.073 –2.261** –2.554*** –1.813 –1.638 11.120*** 11.453*** (0.687) (0.542) (1.123) (0.943) (3.128) (3.093) (2.636) (2.477) Presence of elderly in hh (d) –1.100 –1.095 –3.096 –3.122 –14.013** –14.051** 23.511*** 23.367*** (1.419) (1.215) (2.240) (2.031) (5.648) (5.615) (6.278) (6.224) (continued on next page) 5/27/10 2:41 PM Table A25: The Impact of the Value of Remittances Sent on Hours Worked by Young Men (Aged 16–34, Out of School) – Coefficients (cont.) Keller MNA 5-27-10vol2.indd 63 Results from an IV Tobit estimation with endogenous binary variable (to control for remittances) and the left-censored outcome variable hours worked for men aged 16–34 who left school using the user-written program cmp (Roodman 2007). The alternative specification is a tobit model assuming that migration is exogenous to the outcome variable. Standard errors in parentheses. Any market work Wage & salary work Self-employed / employer Unpaid family work ivtobit tobit ivtobit tobit ivtobit tobit ivtobit tobit 2 Alexandria and Suez (d) –2.785 –3.507** –4.052 –5.120** 14.816* 15.442* –26.819 –25.530 (2.006) (1.731) (2.987) (2.483) (8.726) (8.473) (18.711) (18.500) 2 Urban Lower Egypt (d) –4.176** –5.185*** –12.070*** –13.576*** 31.110*** 32.006*** 25.634* 27.455** (1.933) (1.619) (2.967) (2.468) (7.716) (7.252) (14.284) (13.979) 2 Rural Lower Egypt (d) –3.527** –4.258*** –12.550*** –13.592*** 21.222*** 21.893*** 53.790*** 55.312*** (1.626) (1.414) (2.469) (2.203) (7.003) (6.731) (12.036) (11.868) Urban Upper Egypt2 (d) –6.951*** –7.181*** –17.138*** –17.472*** 32.574*** 32.820*** 38.300*** 38.828*** (1.695) (1.573) (2.643) (2.430) (7.189) (7.115) (13.041) (12.920) 2 Rural Upper Egypt (d) –3.871** –5.866*** –13.410*** –16.325*** 22.046** 23.898*** 53.148*** 56.773*** (1.786) (1.453) (2.808) (2.335) (8.907) (7.167) (12.952) (12.106) Constant –21.329 –19.141 –43.175* –39.698** –496.536*** –498.432*** –103.897* –104.998* (14.301) (11.784) (23.020) (19.690) (66.743) (66.514) (57.511) (55.511) alpha _cons / sigma_cons 17.887* 27.180*** 26.360* 42.718*** –16.375 82.488*** –30.085 82.087*** (9.397) (0.364) (14.468) (0.552) (45.517) (1.417) (38.337) (2.123) lns_cons 3.301*** 3.753*** 4.413*** 4.408*** (0.013) (0.013) (0.017) (0.026) lnv _cons –0.058 –0.058 –0.058 –0.058 (0.121) (0.121) (0.121) (0.121) p 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 N 5460 5460 5460 5460 5460 5460 5460 5460 Wald test of exogeneity (Prob >chi2) 0.0570 0.0685 0.7190 0.4326 Note: Since there may be more than one young man living in a household, we take clustering at the household level into consideration. * p<0.10, ** p<0.05, *** p<0.01, (d) for discrete change of dummy variable from 0 to 1 1 reference category: no educational certificate, 2 reference category: Greater Cairo Region, 3 instrumented variable. Instruments are the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household. Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 63 5/27/10 2:41 PM 64 Keller MNA 5-27-10vol2.indd 64 Table A26: Results of the First Step of the various IV estimations Results from estimating the impact of each migration, remittance-income and average monthly remittance income on young women’s and, respectively, young men’s labor force participation as example. For all three instrumented variables, we use the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household as instruments (see Section 4.2). Coefficients are reported with standard errors in parentheses. At the bottom of the table, Wald test statistics on the joint significance of the three instruments are reported. Young women’s labor force participation Young men’s labor force participation Average monthly Average monthly Migration Remittance- remittance income Migration Remittance- remittance income Instrumenting for household receiving household (in 100 LE) household receiving household (in 100 LE) Outcome equation Migrant household3 0.978*** 0.200 (0.206) (0.397) Remittance-receiving household 0.934*** 0.173 (0.254) (0.593) Average monthly remittance income 0.240 –0.328 (in 100 LE) (0.270) (0.286) Age 0.179*** 0.182*** 0.172*** –0.049 –0.049 –0.043 (0.048) (0.048) (0.057) (0.072) (0.072) (0.070) Age squared –0.003*** –0.003*** –0.003** 0.002 0.002 0.002 (0.001) (0.001) (0.001) (0.002) (0.002) (0.002) 1 Primary or preparatory degree –0.178** –0.180** –0.196*** 0.165** 0.164** 0.162** (0.071) (0.072) (0.069) (0.084) (0.084) (0.082)   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Secondary degree1 0.371*** 0.379*** 0.334*** 0.326*** 0.327*** 0.333*** (0.053) (0.053) (0.114) (0.072) (0.072) (0.070) 1 Above secondary degree 0.913*** 0.926*** 0.816*** 0.273*** 0.279*** 0.315*** (0.064) (0.064) (0.219) (0.091) (0.090) (0.086) Ever-married –0.779*** –0.780*** –0.738*** 0.767*** 0.765*** 0.702*** (0.061) (0.062) (0.129) (0.085) (0.085) (0.121) (continued on next page) 5/27/10 2:41 PM Table A26: Results of the First Step of the various IV estimations (continued) Keller MNA 5-27-10vol2.indd 65 Results from estimating the impact of each migration, remittance-income and average monthly remittance income on young women’s and, respectively, young men’s labor force participation as example. For all three instrumented variables, we use the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household as instruments (see Section 4.2). Coefficients are reported with standard errors in parentheses. At the bottom of the table, Wald test statistics on the joint significance of the three instruments are reported. Young women’s labor force participation Young men’s labor force participation Average monthly Average monthly Migration Remittance- remittance income Migration Remittance- remittance income Instrumenting for household receiving household (in 100 LE) household receiving household (in 100 LE) Nr of females in hh 0.014 0.023 –0.006 0.007 0.009 0.021 (0.025) (0.026) (0.035) (0.030) (0.030) (0.029) Nr of children aged 0–5 in hh –0.052** –0.053** –0.072** –0.041 –0.039 –0.015 (0.024) (0.024) (0.033) (0.033) (0.033) (0.040) Nr of children aged 6–14 in hh 0.088*** 0.091*** 0.078*** 0.013 0.013 0.016 (0.023) (0.023) (0.027) (0.029) (0.029) (0.028) Presence of elderly in hh 0.064 0.090 0.098 –0.043 –0.038 –0.037 (0.061) (0.060) (0.063) (0.069) (0.068) (0.066) Alexandria and Suez2 0.129 0.126 0.123 –0.009 –0.009 0.010 (0.080) (0.080) (0.079) (0.113) (0.113) (0.108) 2 Urban Lower Egypt 0.256*** 0.260*** 0.241** –0.090 –0.088 –0.062 (0.075) (0.075) (0.108) (0.104) (0.104) (0.104) 2 Rural Lower Egypt 0.201*** 0.208*** 0.209*** –0.053 –0.052 –0.030 (0.067) (0.067) (0.081) (0.091) (0.092) (0.089) Urban Upper Egypt2 0.208*** 0.214*** 0.206*** –0.157 –0.157 –0.144 (0.074) (0.073) (0.079) (0.099) (0.100) (0.097) 2 Rural Upper Egypt 0.430*** 0.453*** 0.391** –0.150 –0.143 –0.096 (0.073) (0.073) (0.161) (0.098) (0.096) (0.102) _cons –3.399*** –3.453*** –3.177*** 0.646 0.641 0.558 (0.594) (0.596) (0.889) (0.829) (0.830) (0.814) (continued on next page) Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 65 5/27/10 2:41 PM 66 Keller MNA 5-27-10vol2.indd 66 Table A26: Results of the First Step of the various IV estimations (continued) Results from estimating the impact of each migration, remittance-income and average monthly remittance income on young women’s and, respectively, young men’s labor force participation as example. For all three instrumented variables, we use the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household as instruments (see Section 4.2). Coefficients are reported with standard errors in parentheses. At the bottom of the table, Wald test statistics on the joint significance of the three instruments are reported. Young women’s labor force participation Young men’s labor force participation Average monthly Average monthly Migration Remittance- remittance income Migration Remittance- remittance income Instrumenting for household receiving household (in 100 LE) household receiving household (in 100 LE) IV equation Age 0.065 0.080 0.025 0.005 –0.059 0.012 (0.079) (0.091) (0.045) (0.089) (0.109) (0.035) Age squared –0.001 –0.001 –0.000 –0.001 0.001 –0.000 (0.002) (0.002) (0.001) (0.002) (0.002) (0.001) 1 Primary or preparatory degree 0.134 0.115 0.185** 0.089 0.169 0.017 (0.112) (0.119) (0.087) (0.134) (0.166) (0.032) secondary degree1 0.328*** 0.259** 0.160*** 0.229** 0.260* 0.055** (0.096) (0.102) (0.061) (0.114) (0.148) (0.027) 1 Above secondary degree 0.325*** 0.211* 0.244*** 0.519*** 0.426** 0.136*** (0.115) (0.124) (0.083) (0.142) (0.180) (0.048) Ever-married 0.114 0.129 0.022 –0.467*** –0.413*** –0.084* (0.093) (0.108) (0.078) (0.106) (0.111) (0.050)   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Nr of females in hh 0.201*** 0.158** 0.065 0.163*** 0.135* 0.031 (0.054) (0.065) (0.078) (0.054) (0.069) (0.020) Nr of children aged 0–5 in hh 0.060 0.099** 0.112* 0.085* 0.105* 0.055 (0.037) (0.041) (0.062) (0.048) (0.059) (0.036) Nr of children aged 6–14 in hh 0.004 –0.018 0.002 0.020 0.047 0.009 (0.039) (0.043) (0.024) (0.050) (0.051) (0.022) (continued on next page) 5/27/10 2:41 PM Table A26: Results of the First Step of the various IV estimations (continued) Keller MNA 5-27-10vol2.indd 67 Results from estimating the impact of each migration, remittance-income and average monthly remittance income on young women’s and, respectively, young men’s labor force participation as example. For all three instrumented variables, we use the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household as instruments (see Section 4.2). Coefficients are reported with standard errors in parentheses. At the bottom of the table, Wald test statistics on the joint significance of the three instruments are reported. Young women’s labor force participation Young men’s labor force participation Average monthly Average monthly Migration Remittance- remittance income Migration Remittance- remittance income Instrumenting for household receiving household (in 100 LE) household receiving household (in 100 LE) Presence of elderly in hh 0.331*** 0.149 –0.056 0.237** 0.055 –0.002 (0.096) (0.112) (0.152) (0.117) (0.146) (0.042) Alexandria and Suez2 0.080 0.259 –0.010 0.032 0.315 0.043 (0.178) (0.194) (0.077) (0.225) (0.297) (0.060) 2 Urban Lower Egypt 0.311** 0.454*** –0.006 0.206 0.511** 0.023 (0.148) (0.164) (0.088) (0.190) (0.247) (0.056) 2 Rural Lower Egypt 0.253* 0.357** –0.103 0.203 0.541** 0.009 (0.139) (0.155) (0.085) (0.168) (0.221) (0.042) Urban Upper Egypt2 0.124 0.160 –0.105 0.161 0.444* –0.016 (0.154) (0.171) (0.077) (0.188) (0.249) (0.043) 2 Rural Upper Egypt 0.409*** 0.425*** 0.021 0.316* 0.410* 0.036 (0.140) (0.157) (0.083) (0.169) (0.235) (0.046) IV (percentage of households with 0.094*** 0.085*** 0.033 0.091*** 0.082*** 0.021 current migrants in the district) (0.014) (0.015) (0.042) (0.020) (0.024) (0.023) IV interacted with the number of –0.010*** –0.010*** 0.006 –0.006** –0.008* –0.000 adults in hh (0.002) (0.003) (0.016) (0.003) (0.004) (0.003) IV interacted with average years of –0.001 –0.000 0.000 –0.002* –0.001 –0.000 schooling of adults in hh (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (continued on next page) Appendix 4: Description of Migration/Remittances and Labor Market/Employment Analysis in Egypt: Methodology and Results 67 5/27/10 2:41 PM 68 Keller MNA 5-27-10vol2.indd 68 Table A26: Results of the First Step of the various IV estimations (continued) Results from estimating the impact of each migration, remittance-income and average monthly remittance income on young women’s and, respectively, young men’s labor force participation as example. For all three instrumented variables, we use the percentage of households with at least one current international migrant at the district level and its interaction with the number of adults in the household and the average years of schooling of all adults in the household as instruments (see Section 4.2). Coefficients are reported with standard errors in parentheses. At the bottom of the table, Wald test statistics on the joint significance of the three instruments are reported. Young women’s labor force participation Young men’s labor force participation Average monthly Average monthly Migration Remittance- remittance income Migration Remittance- remittance income Instrumenting for household receiving household (in 100 LE) household receiving household (in 100 LE) Constant –3.851*** –4.194*** –0.694 –2.578** –2.254 –0.206 (0.978) (1.104) (0.518) (1.105) (1.371) (0.446) athrho constant –0.472*** –0.451*** –0.423 –0.156 –0.128 0.297 (0.107) (0.127) (0.464) (0.194) (0.252) (0.294) lnsigma constant 0.487*** –0.058 (0.154) (0.121) p 0.000 0.000 0.000 0.000 0.000 0.000 N 5551 5551 5541 5467 5467 5460 Wald Test statistics on the 127.00 107.81 20.79 61.90 40.68 8.96 instruments Prob>chi2 0.0000 0.0000 0.0001 0.0000 0.0000 0.0299 * p<0.10, ** p<0.05, *** p<0.01 1 reference category: no educational certificate, ² reference category: Greater Cairo Region.   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options 5/27/10 2:41 PM Appendix 5: Description of Migration/ Remittances and Labor Market/ Employment Analysis in Morocco: Methodology and Results (The following description is taken from Silva, 2009) Econometric Methodology where pij is the probability that individual i in household j participates in the labor market, Remittances and labor supply Remittancesj is the variable of interest related to remittance receipts (equal to 1 if household Using the 2000/01 ENCDM data, we estimate j receives remittances and 0 otherwise), is a set the relationship between remittances and the of demographic characteristics for individual i in probability of participating in labor market. Two household j (including, a quadratic term in age, different estimation techniques are used. First, a years of schooling and marital status), is a set of Probit model is estimated. Second, the analysis is household characteristics (including, number of complemented by applying the “propensity score children in the household, presence of zero- to matchingâ€? method, which does not impose a lin- five-year-old child, number of adult males and ear relationship between the variables of interest. females in the household, family home owner- ship), Pi' is a set labor market indicators for the The decision of whether to participate in province of residence of individual i (including, labor markets depends on, among other factors, proportion of households with sanitary services demographic characteristics of the individual and proportion of household heads working in (that affect his potential market wage), family agricultural activities), reg_dummies are re- attributes (including, the number of dependents, gional dummies and uij is an unobserved error their age structure, non labor income/wealth, that term. Since this probability (pij) is not directly affect the reservation wage), labor market indi- observed, the propensity equation is revised as cators for the province and region of residence. a Probit model: Our hypothesis is that it also depends on whether the individual belongs to remittances recipient Pr( pij = 1) = Φ( αRemittances j (2) or non-recipient households. Following Acosta 2006, we state the probability that an individual +Χ β + Z θ + P δ + reg _ dummies + ε0 j ) ' ij ' j i ' participates in the labor market as: where pi = 1 [pi = 0] in the event that individual i of household j [does not] participates in the pij = αRemittances j + Χ β + Z θ + P δ + (1) ' ' ' ij j i labor market. F is the standard normal distribu- reg _ dummies + µ ij tion function. 69 Keller MNA 5-27-10vol2.indd 69 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Second, the analysis is complemented by ap- that they constitute a good proxy for migration plying the “propensity score matchingâ€? method networks. The use of this instrument is supported that does not superimpose a linear relationship by the fact that historical networks are one of the between the dependent variable and the variable principal determinants of migration and, there- of interest. In this case, individuals in the two fore correlate with the probability of receiving groups (participation and non-participating) are remittances; however this has a small impact on “matchedâ€? in terms of all common characteristics each individual’s participation many years later. except one (belonging to a remittances receiving In this paper we argue that a similar reasoning household), and this allows to see what differ- should be valid for the share of return migrants in ence that one characteristic makes. Comparing the population at the province level. Our assump- the difference in the average propensity to par- tion is that this share is also a good proxy for the ticipate between these two otherwise similar strength of migration networks that would help groups of individuals provides a measure of the people to migrate, and therefore, be correlated impact of remittances. with the household remittance-receiving status in 2001 but would not have an impact on each Third, we use an instrumental variable ap- individual’s propensity to participate in labor proach to investigate whether self-selection markets seven years later. In addition, and as a (unobservable characteristics such as social skills robustness check, we experiment with an alter- might influence both the likelihood of having a native instrument to prevent reverse causality: migrant member in the household, and there- we use province-location (rural/urban) share fore, receiving remittances, and the decision to of remittance-receiving households (minus the be inactive) and reverse causality (a household individual household’s own responses) in place member may have to migrate, and send remit- of the household own response to whether it tances, because non-migrant members are inac- received remittances. This captures the broader tive). We follow the recent empirical literature environment in which the household lives and (e.g., Barham and Boucher 1998, and McKensie allows the household’s own contribution to the and Rapoport 2006) and use maximum likeli- average to be excluded. hood to estimate a bivariat Probit model of the following form: The correlation between the participation decision and living in a remittance-receiving Pr( pij = 1) = Φ( αRemittances j (3.1) household is given by Ï? = Cov( ε1i , ε2 i ) . If Ï? is significantly different from zero (low value for +Χ 'ijβ + Z 'j θ + Pi'δ + reg _ dummies + ε1 j ) rho=0: prob>chi2) results using bivariate Pro- bit should be more precise than those of simple Pr( Remittances j = 1) = Φ( IV j' γ + (3.2) Probit models. Χ 'ij β + Z 'j θ + Pi' δ + reg _ dummies + ε2 j ) Remittances and entrepreneurship Where IV is the instrument used to identify re- A similar analysis is conducted for entrepre- mittance receiving households (or more broadly neurship. In this case, pi = 1 [pi = 0] in the households with international migrant mem- event that individual i in household j is [is not] bers). Like others (e.g., Woodruff and Zenteno self-employed (conditional on participating in 2001, Hanson and Woodruff 2003, and McKenzie the labor force). In our analysis we interpret and Rapport 2006), we draw on past information self-employment as “entrepreneurialâ€? activities. on migration rates to instrument for migration. However, use the share of return migrants in the Remittances and labor allocation population at the province level, for urban and between different types of employment rural areas separately, from the 1994 Census, instead of migration rates. The use of migration To analyze the impact of remittances on labor rates as an instrument was framed on the idea allocation between different types of employ- 70 Keller MNA 5-27-10vol2.indd 70 5/27/10 2:41 PM Appendix 5: Description of Migration/Remittances and Labor Market/Employment Analysis in Morocco: Methodology and Results ment we estimate a multinomial logit where the force among urban counterparts is around 2.2 %. states of the world considered for the dependent Males have lower disincentive effects compared variable are: inactivity (reference category), to females, ranging from 1.4 to 2.8 %, depending wage-earner, self-employed, unpaid work and on the area of residence. Among both males and unemployment. females, the stronger negative impact in terms of labor supply for individuals in households re- Results porting remittances are found in rural areas for females and urban areas for males. Labor force participation We complement the Probit analysis using Table A27 reports a series of four Probit regres- propensity score matching methods. These sions, according to equation (2). Results are results can be viewed as a robustness check. robust and stable, and support the hypothesis Table A28 reports these results. They indeed on which the development of the model was show that remittances lower participation, par- based: remittances are significantly associated ticularly among rural female households. with lower labor force participation among males and females, in both rural and urban areas. This Table A29 presents our results for the mar- result is robust across all 4 groups. ginal effects22 of living in a remittance-receiving household on labor force participation. The esti- Rural females from remittances recipient mates clearly show that, when employing bivari- households are 4.2 % more likely to be out of the labor market, while the corresponding decline 22 Note that differently from Probit findings, these results are in the probability of participating in the labor in percentage points calculated at the mean of each variable. Table A27: Results for labor force participation: Probit analysis PROBIT RESULTS: Mg effects reported Labor force participation Rural Urban Sample Males Females Males Females Household receives remittances –0.014*** –0.042*** –0.028*** –0.022*** (0.0005) (0.001) (0.0005) (0.0005) Control variables age 0.020*** 0.022*** 0.071*** 0.060*** age square –0.000*** –0.000*** –0.001*** –0.001*** years edu –0.010*** –0.018*** –0.015*** 0.010*** maried dummy 0.036*** –0.095*** 0.086*** –0.298*** nr kids in the hh 0.001*** 0.005*** 0.016*** –0.014*** young kids in hh dummy –0.003** –0.032*** –0.017*** 0.002 nr adult male in hh –0.001*** –0.024*** 0.003*** –0.028*** nr adult female in hh 0.000*** –0.000*** 0.000*** 0.012*** house owner hh dummy 0.005*** 0.109*** –0.009*** –0.034*** share hh with sanit in province –0.126*** –0.996*** 0.091*** –0.497*** share hh agricult in prov –0.071*** 0.040*** 0.010*** –0.119*** region dummies yes yes yes yes Nr obs. 11,273 12,462 13,909 15,580 Predicted propensity for participation 0.95 0.52 0.86 0.28 R-square 0.23 0.10 0.28 0.15 ***Significant at 1% level. Robust standard errors are reported in parenthesis. 71 Keller MNA 5-27-10vol2.indd 71 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Table A28: Results for labor force participation: Propensity score matching PROPENSITY SCORE MATCHING RESULTS Propensity to participate Rural Urban Sample Males Females Males Females Household receives remittances 84.7% 43.9% 74.2% 31.1% Household does not receives remittances 89.7% 48.0% 76.3% 33.1% Difference –3.1%*** –4.1%** –2.1%** –1.9%** ***Significant at 1 percent level. Robust standard errors are reported in parenthesis. Table A29: Results for labor force participation: Instrumental variable method Rural Urban Sample Males Females Males Females Dependent variable: LF Participation Household receives remitances+ –0.191** –0.331*** 0.080*** –0.209*** [0.435] [0.197] [0.146] [0.184] Control variables age 0.021*** 0.021*** 0.071*** 0.058*** age square –0.000*** –0.000*** –0.001*** –0.001*** years edu –0.010 –0.016*** –0.016*** 0.011*** maried dummy 0.036*** –0.090*** 0.092*** –0.292*** nr kids in the hh 0.001 0.005 0.017*** –0.017*** young kids in hh dummy –0.002 –0.033** –0.018* 0.002 nr adult male in hh –0.002 –0.026*** 0.004 –0.032*** nr adult female in hh 0.002 0.004 0.001 0.013*** house owner hh dummy 0.005 0.110*** –0.008 –0.032*** share hh with sanit in province –0.121* –0.925*** 0.062 –0.420*** share hh agricult in prov –0.076*** 0.034*** 0.003 –0.105*** region dummies yes yes yes yes Dependent variable: Being in remmitances receiving household retur_mig_rate 7.298*** 6.570*** –1.100 –3.776*** return_mig_rate*average eduction of hh 1.108*** 1.663*** 0.225** 0.356*** return_mig_rate*nr adult male in hh –1.403** –2.252*** –0.492* –0.518 Nr. Obs 10,232 12,357 13,425 15,488 rho=0:prob>chi2 0.083 0.000 0.000 0.001 ***, **, * Significant at 1%, 5% and 10% level, respectively. Standard errors are reported in parenthesis. +Instrumented variable. Instruments are rates of return migrants at the province level and divided in urban and rural from Census 1994 and its interaction with the number of adults in the household and the average years of schooling of adults in the household. Results estimated using biprobit command in stata. ate Probit models, individuals living in remittance labor markets, particularly if those individuals receiving/migrant households in rural areas are are women. In urban areas, not only does the also significantly more likely to be inactive in impact of receiving remittances on participation 72 Keller MNA 5-27-10vol2.indd 72 5/27/10 2:41 PM Appendix 5: Description of Migration/Remittances and Labor Market/Employment Analysis in Morocco: Methodology and Results appear to be smaller but, if any, it appears to be The role of age, education and level positive among men. of expenditure on the impact of remittances on participation A similar analysis that also employees bi- variat Probit models but uses the percentage To check how the above results vary across indi- of households that receive remittances at the vidual characteristics, we split our data into sub- district level, excluding the household itself, as samples, according to age, education and level instrument for remittances/migration was also of expenditure, as a proxy for household wealth. performed. Results confirm that receiving remit- Table A30 reports the results. The coefficient on tances reduces labor force participation in rural belonging to a remittance-receiving household areas, particularly among women, while having remains negative and significant for all catego- a smaller impact in urban areas (the effect in ries, as expected. However, the magnitude of the the bivariate Probit model for urban areas is not estimates obtained change considerably across significant both for women and men). sub-samples. In particular, among women, the Table A30: The probability of participation in the labor market by age, education and level of expenditure Participation Probit Results: Mg effects All Rural reported Female Male Female Male 15–24 –0.01*** –0.05*** –0.02*** –0.03*** (0.0009) (0.0009) (0.0018) (0.0012) 25–44 –0.05*** –0.01*** –0.04*** 0.01*** (0.0008) (0.0003) (0.0016) (0.0003) 45–65 –0.08*** –0.04*** –0.07*** –0.05*** (0.0009) (0.0009) (0.0022) (0.0014) None –0.07*** –0.03*** –0.04*** –0.03*** (0.0007) (0.0007) (0.0012) (0.0009) Primary –0.01*** –0.02*** –0.01*** –0.01*** (0.0009) (0.0005) (0.0023) (0.0004) Seconday –0.01*** –0.06*** –0.16*** –0.12*** (0.002) (0.002) (0.0027) (0.0055) Tertiary 0.005* –0.02*** –0.04*** (0.003) (0.001) (0.0085) Bottom 20% 0.04*** –0.05*** 0.01*** –0.06*** (0.002) (0.002) (0.0026) (0.0021) 20%-80% –0.03*** –0.03*** –0.04*** –0.02*** (0.0007) (0.0005) (0.0013) (0.0007) Top 20% –0.04*** –0.02*** –0.04*** 0.02*** (0.0008) (0.0007) (0.0028) (0.0007) ***Significant at 1% level. *Significant at 10% level. Robust standard errors are in parenthesis. 73 Keller MNA 5-27-10vol2.indd 73 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options magnitude of the negative association between Interestingly, among poorer women, being being in a remittance-receiving household and in a household that received remittances is as- labor force participation increases with age and sociated with a higher probability of participa- reaches 8% for 45 to 65 years old women. In con- tion. The opposite occurs among richer women. trast, for men the magnitude of this association Among men, this association is negative for all has an “Uâ€? shape: higher (and similar) among expenditure groups, and particularly high among 15 to 24 years old and 45 to 65 years old, than those in the poorest quintile (reaching –5%).23 among 25 to 44 years old, which may suggest that remittances allow for extended schooling Remittances and entrepreneurship than otherwise (in the case of younger men) and earlier retirement. Table A31 reports the results on the association between remittances and entrepreneurship. We With respect to education, while for women find that remittances receipts positively and sig- the negative association between receiving nificantly relate with self-employment activities remittances and participating in labor markets in all groups. is stronger among those without schooling and those with tertiary education (compared to the Urban females from remittances recipient other education levels), for men this association households are 6.6 % more likely to be self- (although statistically significant) is relatively low in these education levels but high among 23 Note that given the differences in sample size across demo- individuals with secondary education. graphic groups, the conclusions should remain tentative. Table A31: Probit analysis: results for self-employment PROBIT RESULTS: Mg effects reported Self-employment Rural Urban Sample Males Females Males Females Household receives remittances 0.056*** 0.023*** 0.019*** 0.066*** (0.001) (0.001) (0.0006) (0.001) Control variables age 0.025*** 0.019*** 0.007*** 0.003*** age square 0.000*** 0.000*** 0.000*** 0.000*** years edu –0.007*** 0.000*** –0.012*** –0.012*** maried dummy 0.136*** –0.040*** 0.052*** 0.069*** nr kids in the hh 0.003*** –0.005*** 0.006*** 0.005*** young kids in hh dummy –0.042** 0.008*** –0.009*** –0.006 nr adult male in hh –0.030*** –0.042*** –0.005*** 0.005*** nr adult female in hh 0.001*** –0.016*** –0.003*** –0.019*** house owner hh dummy 0.085*** –0.027*** 0.043*** 0.043*** share hh with sanit in province –0.328*** 0.165*** 0.005*** –0.345*** share hh agricult in prov 0.623*** –0.064*** 0.018*** –0.067*** region dummies yes yes yes yes Nr obs. 11,608 7,723 11,130 4,929 Predicted propensity for participation 0.30 0.14 0.22 0.14 R-square 0.30 0.15 0.09 0.12 ***Significant at 1 percent level. ***Significant at 5 percent level. Robust standard errors are reported in parenthesis. 74 Keller MNA 5-27-10vol2.indd 74 5/27/10 2:41 PM Appendix 5: Description of Migration/Remittances and Labor Market/Employment Analysis in Morocco: Methodology and Results Table A32: Results for informality: Multinomial Logit analysis Self- Inactivity Wage earner employment Unpaid work Unemployment Nr Obs. 20,775 11,223 8,062 8,083 4,439 Share 40% 23% 15% 13% 9% Urban Male 0.030*** –0.057*** 0.010*** –0.005*** 0.022*** (0.0005) (0.0007) (0.0006) (0.0001) (0.0005) Female 0.019*** –0.032*** 0.014*** 0.001*** –0.002*** (0.0005) (0.0004) (0.0003) (0.0001) (0.0001) Rural Male 0.015*** –0.093*** 0.057*** 0.008*** 0.013*** (0.0005) (0.0013) (0.0015) (0.0011) (0.0005) Female 0.041*** –0.013*** 0.057*** –0.003*** 0.000*** (0.0011) (0.0002) (0.00149) (0.0006) (0.0001) ***Significant at 1 percent level. ***Significant at 5 percent level. Robust standard errors are reported in parenthesis. employed then counterparts from non-recipient address the question: do all types of participa- household. Among rural females, receiving tion decrease or is there labor reallocation across remittances is associated with an increase in various types of employment? the propensity for self-employment of 2.3 %. Males have lower incentive effects compared to Our results indicate that receiving remit- females, ranging from 1.9 to 5.6 %, depending tances is associated with labor reallocation on the area of residence. Among both males and across various types of employment. In par- females, the stronger positive impact in terms ticular, it reduces propensity to be a wage of self-employment for individuals in household earner while increasing the propensity for reporting remittances is found in urban areas for self-employment in all demographic groups. females and rural areas for males. Moreover, among women in urban areas and men in rural areas it is also associated with a Multinomial logit analysis for higher propensity for being engaged in unpaid informality work, increasing the overall propensity for being engaged in the informal, instead of formal sec- Table A32 presents the results on the relation- tor. In contrast, among men in urban areas and ship between remittances receipt and labor women in rural areas, beside the described ef- allocation across various types of employment. fect the propensity for being a wage earner and The base category in our analysis is inactivity. self-employed, it decreases the propensity for As discussed, receiving remittances increases unpaid work while increasing significantly the the propensity for inactivity (i.e., decreases propensity for being unemployed (particularly participation). The methodology allows us to among urban male). 75 Keller MNA 5-27-10vol2.indd 75 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Keller MNA 5-27-10vol2.indd 76 5/27/10 2:41 PM Appendix 6: Description of Migration/ Remittances and Decisions Affecting Children in Egypt: Methodology and Results (The following description is taken from Elbadawy and Assaad, 2009) Data work for boys and girls, in addition to hours of subsistence and market work. Therefore, ELMPS We will use the nationally-representative Egypt 06 enables us to capture girls’ work, which is Labor Market Panel Survey (ELMPS) 06. It is a often in the form of household chores. longitudinal survey that tracks and re-interviews households in the Egypt Labor Market Survey Methodology ELMS 98 households.24 ELMPS 06 contains infor- mation on international migration history as well Endogeneity of Migration and as a module on current migrants and remittances. Remittances Particularly, it has information on whether the household receives remittances from household As already mentioned, the main estimation prob- members living abroad as well as the amount and lem we need to overcome is the endogeneity of type of these remittances, and which household the migration and remittance decisions. Sasin member receives remittances. In the paper, our and McKenzie (2007) outline the methodological definition of a migrant household is a household difficulties that may be encountered when esti- that has member(s) who migrated in the last five years whether or not they returned. Our analysis excludes migrants taking their families with them 24 The ELMPS 06 sample contains 8,351 households. 3,685 of these households are original ELMS 98 survey households. 2,168 (since they could not be interviewed) and hence households split from original ELMS 98 households (for example, excludes effects on their children. due to marrying and forming a separate household). In addition, 2,498 new households were added to the sample to form a re- fresher sample. For more information on ELMS 98 and ELMPS ELMPS 06 is rich in education variables and 06, please refer to Assaad (2002a, b), Assaad (forthcoming) and enables us to examine the impact of migration Barsoum (2006). 72% of individuals interviewed in 1998 were and remittances on several education indicators: successfully re-interviewed in 2006, forming a panel that can be used for longitudinal analysis. A detailed analysis of sources of school attendance, staying in school, school at- attrition was undertaken in Assaad and Roushdy (Forthcoming). tendance for university-aged individuals, and They show that attrition was mainly caused by the random loss of taking lessons from and spending on private identifying records rather than by a systematic attrition process. tutoring. It also has detailed information on child They found no significant association between the probability of attrition and household and individual characteristics in 1998. work activities. Our data has the advantage of Weights based on the probability of non-response were used to having information on weekly hours of domestic correct for attrition. 77 Keller MNA 5-27-10vol2.indd 77 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options mating migration effects: there may be reverse ture: the instrumental variable approach. This en- causality, self-selection into migration and an tails finding a variable that affects migration and omitted variable problem. Reverse causality takes remittance behavior but that does not directly place if the outcome studied influences migration affect the outcome studied (except through its and remittances and not migration and remit- effect on migration and remittances). Migration tances influencing the outcome. In the schooling history and networks have been used repeatedly outcome case, it may be that households that as shown in the literature review. This choice is value schooling or who are income-constrained based on the hypothesis that networks affect mi- and need to raise the needed income for educa- gration prospects but not outcomes of left-behind tion spending are the ones that decide to send a household members. Social networks of friends member abroad in order that he/she remits the and relatives who already migrated facilitate the needed funds. Similarly, in the case of child work, migration of more individuals belonging to the it may be that households that do not want to same social network, as emphasized by sociolo- expose their children to work resort to migration gists (McKenzie 2005). in order to avoid depending on income from child work. Attributes such as concern for children’s In line with the literature, we use the per- education and child work aversion are hard to centage of households with migrants in the measure and are unobservable to the researcher locality (Shiakha/village level). We acquired resulting in an overestimate of a potentially posi- this data from the Egyptian statistical agency, tive effect of migration on schooling and poten- which they collected as part of Census 2006. tially negative effect of migration on child work. This instrument was also used in concurrent migration impact work (Roushdy et al. 2008, Self-selection into migration and remitting and Binzel and Assaad 2008). This measure is occurs as migrants/migrant households may not expected to capture the strength of social migra- randomly sort themselves into migration. This be- tion networks and is appropriate in the case of comes an issue when sorting is based on charac- Egypt since a lot of temporary migrants are low- teristics that are unobservable to the researcher skilled labor and, as noted by Hoodfar (1997), such as the propensity to take risk and the level their migration is highly affected by the social of motivation and ambition. Self-selection in such support they find in host countries as reflected cases is essentially an omitted variable problem in the network of relatives and people from the that can result in finding a correlation between same community. migration and schooling outcomes that is not causal in nature. The instrumental variable model used is as follows: Roushdy et al. (2008), using ELMPS 06, Y = f ( X1β1 , Zβ 2 , ε ) provides descriptive evidence that households in Egypt are not randomly selected into migration (Structural equation) (1) and receipt of remittances. Comparing migrant Z = f ( X 2 γ1 , IV γ 2 , δ) to non-migrant households and remittance- recipient to non-remittance-recipient households (Reduced-form for) (2) show that they differ along several demographic and socioeconomic characteristics. This confirms Equation (1) relates Y, the vector of child out- that, in order to get reliable estimates, factors comes (dependent variables) to the vector of determining migration and receipt of remittances explanatory variables consisting of X1, a vector need to be controlled for. of child and household characteristics and Z, the potentially endogenous vector of migration Model indicators as well as to e, the vector of error terms. We refer to equation (1) as the outcome To get around the endogeneity issue, we employ equation and equation (2) as the migration equa- the approach that is used most often in the litera- tion. Details on Y, X1, Z, X2 are provided below. 78 Keller MNA 5-27-10vol2.indd 78 5/27/10 2:41 PM Appendix 6: Description of Migration/Remittances and Decisions Affecting Children in Egypt: Methodology and Results Endogeneity is reflected in the potential a migrant household or not, where we define correlation of Z and e. To overcome endogeneity, a migrant member as migrated within the last equation (2) is used where IV is the instrument five years whether or not he/she returned; (2) a variable for migration and remittances that we binary variable indicating whether the child lives discussed above and the error terms (e and d) are in a household that receives remittances from a allowed to be correlated. A regression equation is (current) migrant household member and (3) fitted separately for each child outcome and each a left-censored variable (i.e., unconditional on migration indicator. The regression model used receiving remittances) indicating the average and the exact functional form in equations (1) monthly amount of remittances received by the and (2) depend on the nature of the dependent household expressed in 100 L.E.25 The variables and endogenous variables examined. are referred to in short as: migration, remittance- receipt and remittance-level. As will be shown in the following sub-sec- tions, our dependent variables and migration While the remittance-receipt measure is used indicators will either be binary or left-censored more often in the literature compared to the mi- resulting in various binary-censored combina- gration measure, we prefer including migration tions. Using 2-SLS will be inconsistent in these as a separate measure because remittances is cases and employing 2-stage estimators that not the only outcome of migration. For example, take into account that the dependent and en- migration affects households through the effect dogenous variables are not continuous will result of absence of a household member, which may in efficiency gains. Such models, however, are result in a disruption of family life. Also, as noted not implementable using built-in commands in by Sasin (2008), the income lost due to the ab- statistical packages. We use the user-written sence of a migrant household member may not Conditional Mixed Process (CMP) command in be fully offset by remittances he/she sends. In Stata (Roodman 2007) in all regressions. CMP addition, not all international migrants choose fits a host of multi-equation estimators that ac- to remit money. commodate various types of dependent variables in each estimated equation. We jointly estimate We include both remittance level and remit- a reduced-form migration/remittance equation tance receipt because the remittance level mea- and a structural schooling/child work equation sure may suffer from recall bias as respondents as a joint Tobit/Probit and Probit/Tobit model may not recall exact amounts especially that depending on the nature of the migration variable remittances are usually pooled with other income and the outcome variable examined. sources. On the other hand, respondents should recall whether they ever received remittances We also report results based on (single-equa- during the year (Acosta 2006). Caution should be tion) models that assume exogeneity of migration taken in interpreting the impact of the level of re- and remittances. We use Probit models for binary mittances since as per Acosta (2006), the level of outcomes and Tobit models for left-censored remittances may be underreported in household outcomes. These models are particularly rel- surveys thereby creating an underestimate of the evant in the cases where we find no evidence impact of remittances on outcomes. Despite the that the correlation between the two equations concerns outlined above, we do not exclude the is significant or in other words no evidence that remittance level indicator since it may not be migration or remittances are endogenous to the just receiving remittances that have an effect but outcome of interest. rather how much a household receives. Migration Indicators We use the intensity of migration network at the local level as an instrument variable IV for The vector Z in the outcome equation contains three migration-related measures: (1) a binary 25 The exchange rate at the time of data collection for ELMPS variable indicating whether the child lives in 06 was US$1 = 5.7 L.E. 79 Keller MNA 5-27-10vol2.indd 79 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options the three migration indicators. While migration dren’s competitiveness via facilitating obtaining networks are thought to facilitate migrating and higher scores.27 Concurrent to the improvements not necessarily remitting, the instrument would in several education indicators, there have been still capture remittances since migration and considerable reductions in the incidence of child remittances are correlated. In Egypt, Binzel and labor (Assaad 2002b, Zibani 2002). Assaad et al. Assaad (2008) found using ELMPS 06 that 66% (2008) present descriptive analysis of Egyptian of migrant households do receive remittances. children’s work experience. In addition, alternative variables used in the lit- erature to instrument for remittances would not The Y vector consists of schooling outcomes be relevant in the case of Egypt. For example, as well as child work outcomes. The schooling Amuedo-Dorantes and Pozo (2006) use the per outcomes are mainly school attendance and capita count of Western Union offices in the state private tutoring outcomes.28 For the bulk of the (interacted with a household education variable) analysis, we restrict our analysis to the school- as an instrument for remittances. This variable aged group 6–17. Specifically, we use the follow- is linked to remittances as it captures the acces- ing schooling variables: (1) school attendance sibility and extent of use of money transfer pro- for school-aged children (binary), (2) school at- viders. This variable, however, will not capture tendance conditional on school entry for school- remittance flows in Egypt because remittances aged children (binary), (3) school attendance for tend to be transferred informally and often by the university-aged children 19–21, (4) private hand. Wahba (2007) found, using ELMPS 06, that tutoring (binary), and (5) the annual level of only 22% of remittances were received through spending on private tutoring (left-censored i.e., the banking system. unconditional on taking tutoring). Schooling and Child Work Outcomes The unconditional school attendance out- come captures attendance resulting from re- A general improvement in education took place maining in school and from entering school to over the last two decades partly because of the start with. The school attendance conditional on implementation of an enormous school-building school entry captures survival in school. School plan in rural areas. 26 Consequently, groups attendance for those 19–21 is meant to capture traditionally suffering from lower educational the more discretionary university education. Due attainment such as girls in rural areas and more to its discretionary nature, university attendance generally residents of rural areas witnessed is expected to be more elastic to remittance a relatively larger improvement. As a result, income. While the broader 18–22 age-group is gender and urban/rural gaps in enrollment and relevant for university attendance, we excluded attainment are closing (Elbadawy, forthcoming). the 18 year olds since they may still be in high However, girls remain disadvantaged in terms of school, and excluded the 22 year olds since they school entry (Elbadawy and Assaad 2008). The could have already completed their university disparity in school entry is especially evident education. Examining the effect of migration in rural areas of Upper Egypt. Conditional on and remittances on a discretionary yet important entering school, girls are not disadvantaged in terms of remaining in school (Elbadawy and 26 Elbadawy (forthcoming) describes the education system in Assaad 2008, Lloyd et al. 2003). A key feature Egypt in some detail and provides an overview of education of the education system in Egypt, as in many de- trends in Egypt. veloping countries, is the use of private tutoring 27 Elbadawy et al. (2005) examines private tutoring determi- nants in Egypt. classes to supplement formal schooling. Based 28 The school attendance variables reflect actual school atten- on ELMPS 06, 45% of pre-university students dance rather than just being enrolled in school. Actual atten- use private tutoring services. Exam scores are dance is expected to be more accurate since it is not uncommon in developing countries that some of the children that enroll at the only criteria determining entry into higher school end up not attending classes. education levels. Private tutoring is, therefore, a 29 ELMPS 06 does not have information on academic perfor- strategy adopted by households to enhance chil- mance and exam scores. 80 Keller MNA 5-27-10vol2.indd 80 5/27/10 2:41 PM Appendix 6: Description of Migration/Remittances and Decisions Affecting Children in Egypt: Methodology and Results education investment such as private tutoring To sum up, we have the following child work contributes to the literature. As in the case of variables, (1) 14-hour cut-off variable based on university-aged attendance outcome, we expect the market definition of work, (2) 1-hour cut-off tutoring to be more elastic to remittance income variable based on the market definition of work, than school attendance. (3) 14-hour cut-off variable based on the inclu- sive definition of work, (4) 1-hour cut-off variable We use multiple measures for child work based on the inclusive definition of work, (5) reflecting the multiple definitions of what consti- 14-hour cut-off variable for domestic work, and tutes work, and the amount of work. There are (6) 1-hour cut-off variable for domestic work. three types of work activities: market, subsis- tence, and domestic. Labor market work is work For all outcomes, we run separate regres- involving productive activities for the purpose sions for boys and girls. This is because migra- of market exchange. Subsistence work includes tion/remittances and the other explanatory activities involving the production and process- variables can affect boys’ and girls’ schooling and ing of primary goods for purposes of household child work differently. consumption. Domestic work includes activities such as cooking, errands, house cleaning, col- Explanatory Variables lecting water, laundry, and childcare. Given that a considerable number of hours can be spent The migration variables were explained in detail on domestic work, conventional definitions that above. In the regressions, we interact the migra- exclude such activities can significantly under- tion variable whose impact is examined with the estimate estimates of child work especially for age groups: 6–11, 12–14 and 15–17. This is to girls. ELMPS 06 has the advantage of having allow for a differential impact by age groups as information on weekly domestic hours for boys found in the literature. While it was found in the and girls. Following Assaad, Levison and Zibani literature that the migration effect was weaker (2008), we use a broad and a narrow definition or sometimes negative for older age groups, we of work. The “marketâ€? definition limits work to expect that migration in Egypt may have a stron- labor market work only while the “inclusiveâ€? defi- ger effect on the older age group. For example, nition includes the three types of work: market, with respect to school attendance, younger chil- subsistence, and domestic. dren are more likely to be in school because of a cohort effect; their cohort would benefit more For both the “marketâ€? and the “inclusiveâ€? from the general improvement in education as definition, we use two variables that are based mentioned above. They are also more likely to on the number of hours spent on the relevant be in school because older children are more work activities. Following international recom- likely to have dropped out of schooling already mendations, we use a 1-hour cut-off where a because surviving to higher education levels is child is counted as worker when he/she engages generally more optional. in a work activity for at least 1 hour per week. Following Assaad et al. (2008), we also use a Migration variables are not interacted in higher cut-off level of 14 hours per week.30 In the regressions of university-aged school at- addition, we employ two child work measures tendance. In some child work regressions, par- that exclusively focus on domestic work. One ticularly under the market work definition for is a binary variable taking the value one if the younger groups, some age-interactions with the child did 1-hour of domestic work and the second migration variable were dropped during estima- is a binary variable taking the value one if the tion because there were not enough observations child did 14-hour of domestic work. Examining for children that are working. Therefore, the age domestic work allows us to capture the possible substitution effect for time within the household 30 Assaad et al. (2008) experimented with cut-off levels between because of absence of household members be- 8 and 14 hours and found the percentage of child work not to cause of migration (directly or indirectly). significantly vary by cut-off level. 81 Keller MNA 5-27-10vol2.indd 81 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options groups had to be lumped into 6–14 and 15–17. age of 5 can be affected by a spouse’s absence In the few cases, where the younger 6–14 age due to migration.33 Similarly, the current level of group was dropped, age-interactions with the household wealth can be affected by remittance migration variable were removed from the re- inflows to the household. gression and only the migration variable itself was included. Results A uniform set of independent variables X1 The unconditional probability of the binary and X2 is used in all regressions. Broadly speak- schooling/work outcomes for the reference boy ing, the vector of independent variables in the and girl are listed in the top row of each table.34 outcome equation X1, include child, parents, and The reference school-aged boy (girl) is a 12 year household characteristics. Specifically, it con- old boy (girl) who lives in the Greater Cairo re- sists of the following variables: child age, child gion in a nuclear household. He (she) is not the age squared, whether the child is the eldest child eldest child of the household head. His (her) of the household head, whether the child lives in father and mother have mean years of schooling an extended household, father’s age when child and have mean age (among parents whose child was 6, mother’s age when child was 6,31 whether is 6 years of age). His (her) father is not absent the father is absent permanently, father’s years permanently. The reference university-aged boy of schooling, mother’s years of schooling, and (girl) only differs in terms of age. His (her) age a set of dummy variables indicating the region is 20 years. First-stage results show that the where the child resides.32 While we largely fol- instrument variable we use performs reasonably lowed the specification in Assaad et al. (2008), well as it is consistently positive and statistically we excluded variables that are arguably as- significant at the 1% level for each migration sociated with migration such as household indicator instrumented for. wealth, household composition, relationship to household head, and temporary absence of In all models, we test for exogeneity of the parents. Similarly, we did not use variables like instrumented variable to the outcome variable current household head characteristics (used using the estimate of athanhrho that is part by Acosta 2006) since the head (and therefore of the output of the CMP command. It is linked his/her characteristics) can directly change as to the correlation between the error terms of a result of migration. the outcome and migration equations. The null hypothesis is that this correlation is zero, which We use the same migration equation speci- signifies that the migration variable is exogenous fication used in concurrent work by Roushdy to the error term in the outcome equation, in et al. (2008) and Binzel and Assaad (2008). which case models assuming exogeneity (in our In addition to IV the variable explained above, the migration equation vector of independent 31 Younger parents are more likely to be resource-constrained variables in the migration equation X2 contains and therefore are more likely not to afford schooling. Because household composition and household educa- older kids are more likely to have older parents, we fix parents’ tion variables: the number of children 6–14, age at a given child age to allow for a meaningful comparison. number of males 15–29, number of females We chose to use parents’ age when the child was age 6 because this is the age at which school decisions are made. 15–29, number of elderly 64 and above, aver- 32 The region definition we use incorporates an urban/rural age years of schooling for males 18 and above, breakdown. The regional categorization is as follows: Greater average years of schooling for females 18 and Cairo region, Alexandria and Suez Canal governorates, Urban Lower Egypt, Rural Lower Egypt, Urban Upper Egypt and Rural above. We were careful not to include variables Upper Egypt. Greater Cairo is treated as the reference category. that can possibly be affected by migration such 33 As mentioned earlier in the paper, migration in Egypt is mostly as the number of children in the household temporary. Therefore, we do not expect the number of children above age 6 to be affected by spousal absence due to migration. below age 5, household wealth, current region 34 For binary outcomes, the marginal effects are calculated based of residence and current household head char- on the probability of a positive outcome. For censored outcomes, acteristics. The number of children under the the marginal effects are calculated based on linear predictions. 82 Keller MNA 5-27-10vol2.indd 82 5/27/10 2:41 PM Appendix 6: Description of Migration/Remittances and Decisions Affecting Children in Egypt: Methodology and Results case Probit/Tobit models) will be more efficient having a migrant member on school attendance and there would be no need to use an IV estima- of older boys 15–17 and 19–21. This suggests tion. Estimates of athanrho are provided in the that the positive remittance effect is offset by the bottom row of each table in the appendices. disruptive effect of the migrant absence resulting in an insignificant effect of migration. In this section, we only provide summary result tables focusing on migration indicators With respect to the effect of migration and effects (based on the marginal effects tables remittances on schooling of girls, we found some A1.1–A1.24). We focus on the level of statistical evidence that migration and remittance receipt significance of the migration variables. If the increase the likelihood of attending school. migration variable is statistically significant, its For young girls 6–11, the positive migration sign as well as its marginal effect estimate are also effect can be interpreted as a speeding-up of provided. In indicating the level of significance, school attendance since we observe no effect a “0â€? denotes that the migration indicator has a on conditional school attendance: girls 6–11 statistically insignificant effect, “+ + +â€? / “– – –“ in migrant households are less likely to be late denotes that the migration variable has a posi- school-entrants. There is also some positive tive/negative effect that is significant at the 1% effect of migration and receiving remittances level, “+ +â€? / “– –“ sign indicates that the migra- on school attendance for older girls. However, tion variable has a positive/negative effect that according to the CMP model, remittance level is significant at the 5% level, while “+â€? or “–“ has a negative effect on attendance for girls. The sign indicates that the migration variable has a non-robustness in the result for girls schooling positive/negative effect that is significant only at makes it non-trustable especially that this is one the 10% level. of the few schooling regressions where athanrho is statistically significant. We report results based on both the CMP and the Probit/Tobit models in the summary tables. As in the case of boys, we found that re- When we have no evidence against exogeneity mittances receipt may increase the probability (i.e., when athanrho is statistically insignificant), of school attendance for university-aged girls. we mainly use the Probit/Tobit models’ estimates However, the result is not as robust as for boys. for interpretation. For easy reference, we high- Another result we found with respect to girls lighted the columns used in interpretation. schooling, and which we discount, is that related to the impact on private tutoring variables. Table A33 shows the effects on schooling of Under CMP, migration variables consistently boys. With respect to the effects on schooling at- have a negative impact on tutoring. We find this tendance, we found that there is not much of an unbelievable particularly that the Probit/Tobit effect of migration indicators on young boys 6–11 models are generally producing insignificant and 12–14. In addition, there was consistently no impacts on taking private tutoring and tutoring impact on tutoring variables for school-aged boys. spending. We found a mild positive effect of remittances receipt and remittance level on school attendance Now moving to the impacts on child work for for boys aged 15–17. More importantly, remit- boys, we found a very large effect of migration tances have a strong positive effect on attendance and remittances on young boys’ market work. for university-aged boys. The unconditional Migration practically reduces the probability of probability of attending school for the reference market work to zero in the case of boys 6–14. university-aged boy is 30 percent. This probabil- Interestingly, living in a migrant household seems ity increases by 20 percentage points when the to increase the likelihood of light work for older household receives remittances. It also increases boys 15–17 (1-hour cut-off under the inclusive by 3.6 percentage points per every 100 L.E of definition). This seems to be driven by an in- monthly remittances. It is worth noting that, un- creased likelihood of performing light domestic like remittances, there was no positive effect of work (1-hour cut-off of domestic work). 83 Keller MNA 5-27-10vol2.indd 83 5/27/10 2:41 PM 84 Keller MNA 5-27-10vol2.indd 84 Table A33: Effects of Migration and Remittances on Schooling of Egyptian Boys University-Aged School-Aged 6–17 19–21 Conditional School Private Tutoring School Attendance Attendance Private Tutoring Spending School Attendance CMP CMP CMP CMP CMP Probit Probit Probit Probit Probit Probit Tobit Tobit Probit Probit Uncond. Prob.* 0.95 0.99 0.40 0.30 Migration 6–11 0 +++ 0 0 0 0 0 0 N/A N/A (0.038) 12–14 0 0 0 0 0 0 0 0 N/A N/A 15–17 0 0 0 0 0 0 0 0 N/A N/A 19–21 N/A N/A N/A N/A N/A N/A N/A N/A 0 0 Exog.: Y/N* N ++ Y Y N ++ Y Remittance 6–11 0 0 0 0 0 0 0 0 N/A N/A Receipt 12–14 0 0 0 0 0 0 0 0 N/A N/A 15–17 0 + 0 + 0 0 0 0 N/A N/A (0.028) (0.006) 19–21 N/A N/A N/A N/A N/A N/A N/A N/A 0 ++ 0.225 Exog.: Y/N* N+ Y Y Y Y Remittance 6–11 0 0 0 0 0 0 0 0 N/A N/A   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Level 12–14 0 0 0 0 0 0 0 0 N/A N/A 15–17 0 + + + 0 0 0 0 N/A N/A (0.010) (0.003) (0.003) 19–21 N/A N/A N/A N/A N/A N/A N/A N/A 0 ++ (0.036) Exog.: Y/N* Y Y Y Y Y * Unconditional probability for the reference boy, Exogeneity (Y/N) is based on the athanrho estimate, marginal effects values in parenthesis 5/27/10 2:41 PM Table A34: Effects of Migration and Remittances on Schooling of Egyptian Girls Keller MNA 5-27-10vol2.indd 85 University-Aged School-Aged 6–17 19–21 Conditional School Private Tutoring School Attendance Attendance Private Tutoring Spending School Attendance CMP CMP CMP CMP CMP Probit Probit Probit Probit Probit Probit Tobit Tobit Probit Probit Uncond. Prob.* 0.94 0.99 0.43 0.2 Migration 6–11 0 ++ + 0 ––– – ––– 0 N/A N/A (0.032) (0.009) (–0.405) (–0.106) (–461.639) 12–14 0 0 ++ 0 ––– 0 – 0 N/A N/A (0.009) (–0.347) (–242.675) 15–17 0 + ++ + ––– 0 0 0 N/A N/A (0.026) (0.009) (0.005) (–0.357) 19–21 N/A N/A N/A N/A N/A N/A N/A N/A –– 0 (–0.147) Exog.: Y/N* Y Y N +++ N +++ Y Remittance 6–11 0 0 0 0 ––– 0 – 0 N/A N/A Receipt (–0.397) (–344.199) 12–14 0 ++ 0 0 ––– 0 ––– 0 N/A N/A (0.043) (–0.395) (–379.374) 15–17 0 +++ 0 ++ ––– 0 0 0 N/A N/A (0.037) (0.005) (–0.391) 19–21 N/A N/A N/A N/A N/A N/A N/A N/A ––– + (–0.224) (0.128) Exog.: Y/N* Y Y N+ N ++ Y Remittance 6–11 ––– 0 0 0 0 0 ––– – N/A N/A Level (–0.011) (53.159) (–25.932) 12–14 0 0 0 0 0 0 ––– 0 N/A N/A (–52.984) 15–17 – 0 0 0 0 ++ 0 0 N/A N/A (–0.012) (0.052) 19–21 N/A N/A N/A N/A N/A N/A N/A N/A 0 ++ (0.008) Exog.: Y/N* N +++ Y Y N ++ N+ *Unconditional probability for the reference girl, Exogeneity (Y/N) is based on the athanrho estimate, marginal effects values in parenthesis Appendix 6: Description of Migration/Remittances and Decisions Affecting Children in Egypt: Methodology and Results 85 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options With regard to impact on girls’ work (Table 4), tained under the market definition. Another the results on girls’ market work are unstable. interesting result is that unlike boys 15–17, girls The probability of market work is very low for 15–17 seem to enjoy a reduction in the heavy girls, which seems to result in contradictory domestic hours category as a result of the remit- results. We, therefore, discount the results ob- tance income. Table A35: Effects of Migration and Remittances on Child Work Activities of Egyptian Boys 6–17 Market Definition Inclusive Definition Domestic Work 14 Hour Cutoff 1-Hour Cutoff 14-Hour Cutoff 1-Hour Cutoff 14-Hour Cutoff 1-Hour Cutoff CMP CMP CMP CMP CMP CMP Probit Probit Probit Probit Probit Probit Probit Probit Probit Probit Probit Probit Uncond. 0.026 0.028 0.041 0.25 0.008 0.25 Prob.* Migration 6–14 0 ––– 0 ––– – –– – 0 – 0 – 0 (–0.022) (–0.022) (–0.031) (–0.024) (–0.152) (–0.008) (–0.152) 15–17 0 0 0 0 0 0 0 ++ 0 0 0 ++ (0.140) (0.140) Exog.: Y Y Y Y N +++ Y Y/N* Remit- 6–14 0 ––– 0 ––– 0 –– 0 0 0 0 0 0 tance (–0.018) (–0.017) (–0.025) Receipt 15–17 0 0 0 0 0 0 0 0 0 0 0 0 Exog.: Y Y Y Y Y Y Y/N* Remit- 6–14 0 – 0 0 – –– 0 0 –– 0 0 0 tance (–0.012) (–0.026) (–0.020) (–0.007) Level 15–17 0 0 0 0 0 0 0 0 0 0 0 0 Exog.: Y Y Y Y N+ Y Y/N* *Unconditional probability for the reference boy, Exogeneity (Y/N) is based on the athanrho estimate, marginal effects values in parenthesis 86 Keller MNA 5-27-10vol2.indd 86 5/27/10 2:41 PM Appendix 6: Description of Migration/Remittances and Decisions Affecting Children in Egypt: Methodology and Results Table A36: Effects of Migration and Remittances on Child Work Activities of Egyptian Girls 6–17 Market Definition Inclusive Definition Domestic Work 14-Hour 14-Hour 14-Hour Cutoff 1-Hour Cutoff Cutoff 1-Hour Cutoff Cutoff 1-Hour Cutoff CMP CMP CMP CMP CMP CMP Probit Probit Probit Probit Probit Probit Probit Probit Probit Probit Probit Probit Uncond. 0.005 0.006 0.11 0.41 0.093 0.41 Prob.* Migration 6–14 N/A N/A N/A N/A 0 0 0 0 0 0 0 0 15–17 N/A N/A N/A N/A 0 0 0 0 0 0 0 0 6–17 +++ – +++ 0 N/A N/A N/A N/A N/A N/A N/A N/A (0.349) (–0.004) (0.454) Exog.: Y/N* Y N+ Y Y Y Y Remit- 6–14 N/A N/A N/A N/A 0 0 0 0 0 0 0 0 tance 15–17 N/A N/A N/A N/A 0 ––– 0 0 0 ––– 0 0 Receipt (–0.068) (–0.056) 6–17 +++ –– 0 –– N/A N/A N/A N/A N/A N/A N/A N/A (0.492) (–0.004) (–0.006) Exog.: Y/N* Y Y Y Y Y Y Remit- 6–14 N/A N/A N/A N/A 0 0 0 + 0 0 0 + tance (0.014) (0.014) Level 15–17 N/A N/A N/A N/A 0 0 0 0 0 0 0 0 6–17 +++ 0 +++ 0 N/A N/A N/A N/A N/A N/A N/A N/A (0.003) (0.004) Exog.: Y/N* N ––– N ––– N –– Y N –– Y *Unconditional probability for the reference girl, Exogeneity (Y/N) is based on the athanrho estimate, marginal effects values in parenthesis 87 Keller MNA 5-27-10vol2.indd 87 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Keller MNA 5-27-10vol2.indd 88 5/27/10 2:41 PM Appendix 7: Description of Migration/ Remittances and Decisions Affecting Children in Morocco: Methodology and Results (The following description is taken from Herrera, Dudwick, and Murrugarra 2008) Methodology: Intra-household bution. Although the differences in the shares Expenditures Allocation of education and health are small, female heads in remittance receiving households spend more The econometrical analysis is done by modeling on those expenses than those in non-receiving the share of food and education expenditures, households. Even more, graph 1 shows that over the total household budget, as a function of among remittance receiving households, female the following key variables: i) remittances, ii) the heads spend more in food than their male coun- female head, and iii) the demographic share of terparts along the income distribution. boys and girls; as well as their combined effect, and iv) household socioeconomic characteristics Remittances and female headship have as control variables differentiated effects on intra household allocation by urban and rural areas. The Results econometric results for the food share show that remittances neither its combined effect with fe- Remittance-receiving households spend less male headship do not affect the allocation on this on food and slightly more in health, educa- expenditure in urban areas. Opposite results are tion, housing, and transport. Table A37 shows found in rural areas, where remittances increase the average expenditure shares among the remit- the food share only in female-headed households. tance-receiving and non-receiving households. Remittances and female headship do not affect Except for the food expenses, in which receiving the education share. households spend 5 percent less of their budget compared to the non-receiving households; the Remittances trigger a more equal intra- difference in the rest of the expenditure shares household allocation on education and food is smaller, less than 2 percent. These patterns between boys and girls, closing the existing are similar among female-headed households. gender gaps in rural areas. The econometric (See Panel b, Table A37). However, remittances results for the food share expenses in rural areas seem to enhance women’s allocation preferences indicate that boys’ demographic share increases on human capital expenses as the empirical the allocation of food while the girls’ do not af- evidence discussed above suggests a potential fect at all. However, in presence of remittances role of the female head in the resources distri- this boys’ effect is offset closing the gender gap 89 Keller MNA 5-27-10vol2.indd 89 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Table A37: Main Expenditures Patterns for Remittance-Receiving and Non-receiving Households, 2001 Average Shares of the Expenditures Panel a Panel b Total HH Female Headed HH Shares of Expenditures on Total Consumption Receiving HH Non Receiving Receiving HH Non Receiving Food 42.2% 46.9% 42.5% 46.2% Clothing 4.9% 4.0% 4.9% 3.7% Housing 23.9% 24.7% 26.2% 28.1% Health 5.0% 4.2% 5.5% 4.8% Education 1.4% 1.2% 1.3% 0.9% Transport and Communication 6.6% 4.9% 5.7% 4.3% Source: Calculation from ENCDM (2001) 1 Education share excludes the leisure type of expenses 2 Health excludes the hygiene type of expenses in this allocation. Since female headship affects as a function of the key variables: remittances, the food share among remittance receiving female head, and their combined effect; as well as households in rural areas, these results could his/her individual and household socioeconomic suggest that women lead a more even distribution characteristics. Regressions are at the individual of remittances on food expenses between girls level for children 6–11 years old and by urban and and boys without gendered preferences. There- rural areas. Given the limited data on nutritional fore, women attempt to deal with the inequality outcomes, the nutrition analysis is focused on in allocation resources induced by male heads, the anthropometric z-score indicator height for which is documented in MOP35 “Fathers prefer age.37 This indicator measures children’s stunt- to give education and money to sons over ing (if his/her z-score is below three standard daughters because the first ones will have a deviations of the mean) and shortness (if his/her higher return, girls will marry at any point.â€? z-score is below two standard deviation of the mean) conditions. The children’s probability of Similar evidence is found for the education being stunted or short (by the above definitions share. Among non-remittance-receiving house- of those terms) is modeled in a similar way as the holds, boys’ demographic share increases the primary school enrollment. Regressions are at education household expenses more than girls’ the individual level for children from 2 to 5 years do, resulting in a gender gap in this allocation; old. This section discusses the main findings of however, this difference is smaller among re- ceiving households. These findings indicate that remittances offset the gendered bias allocation in 35 World Bank (2007) Moving Out of Poverty in Morocco, Middle education and food expenses, triggering gender East and North Africa Region Report 36 Sadiqui (2004) shows that women in rural areas have been parity in children’s human capital accumulation disadvantaged in relation to urban modernization, For instance, in rural areas where girls are more marginalized.36 their education and health access has traditionally been limited. 37 The Z-scores (Standard deviation score) is defined as the difference between the value for an individual and the median Children’s educational and nutritional value of the reference population for the same age or height, outcomes. This section investigates whether the divided by the standard deviation of the reference population. impact of remittances on the food and education The nutritional analysis only includes height for age because the allocations are reflected in children’s educational data quality of the other anthropometric indicators (weight for age and weight for height) was not enough to implement the and nutritional outcomes. For education, we ana- models; moreover, this data was not consistent with other sources lyze the probability of a child’s school enrollment of information such as the Demographic and Health Survey, 2003. 90 Keller MNA 5-27-10vol2.indd 90 5/27/10 2:41 PM Appendix 7: Description of Migration/Remittances and Decisions Affecting Children in Morocco: Methodology and Results the analyses addressing the following questions This difference is considerable larger in rural areas formulated in the conceptual framework: Are where primary school rates are 80 percent among the potential differences in allocation resources households with remittances and 10 percent less reflected in children’s educational and nutritional in non-receiving households. In urban areas both outcomes? Do the remittance-receiving female- groups of households have similar school enroll- headed households have different educational ment rates The school enrollment rates, disag- and nutritional outcomes? Do the receiving- gregated by gender, also evidence a more even remittance households have different impacts education access for girls in rural areas. While on school enrollment o between girls and boys? among remittance-receiving households in rural areas, the boys and girls’ primary school rates are Despite the fact that Morocco has made almost the same, in the non-receiving households some progress in improving the education the boy’s primary school rate is 85 percent and the sector, there is still room for increasing girls’ is 78 percent. These gender differences are access and gender parity, mainly in rural not significant in urban areas. areas. During the last decade Morocco has made some progress to improve the educational indica- Children in remittance-receiving house- tors: the number of enrolled students from 6 to holds with a female head are more likely 17 years old increased from 4.1 to 5.5 million and to go to primary school in rural areas. The the primary school rate boosted from 54 percent results from the econometric models show that to 92 percent between 1992 and 2005; however, remittances do not affect the children’s likeli- the country is still lagged compared to the results hood of going to primary school in urban areas, of other countries in MENA.38 According to the while they do in rural areas but only in female- 2001 ENCDM survey, around 60 percent of the headed households. These results corroborate household heads do not have any schooling and similar evidence coming from qualitative studies this percentage increased to 73 percent in rural in Morocco which show that women in remit- areas. Furthermore, the illiterate population tance-receiving households try to improve their (over 10 years old) represents 55 percent of the children’s education through supporting their total population. Although the government has daughters’ school enrollment or choosing a more promoted a more equitable education, 83 percent modern education system than the traditional of women remain illiterate and the urban boys’ for their children (Fadiqui 2004, and De Hass school enrollment doubles that of rural girls.39 Ac- 2003). In addition, this evidence corroborates cording to the 2005 MOP evidence, limited access other empirical findings in developing countries, to education for rural girls is in part explained exposing the importance of the role of Moroccan by the lack of transportation to schools, which women in reaping the potential remittance gains are located far away from the communities, and in a household’s human capital accumulation. enforced by gender norms, which do not allow girls to go by themselves to the school; in this Remittances are associated with a reduc- sense, boys are freer than girls: “The girl of the tion of the gender gap in the primary school rural areas is not free, she can not continue enrollment in rural areas. The econometric her studies far from her parents. She is al- evidence indicates that the probability of attend- ways supervised. After the primary studies, ing school increased more for girls than for boys the girls are obliged to stay home and do the in rural areas where remittances are received in houseworkâ€?(MOP, 2005). female-headed households. Since these female gains are larger, remittances are associated with Remittance receiving households have a reduction in the primary school enrollment higher primary school enrollment rates and less gender disparity. While 91 percent of 38 For 2005 the primary school enrollment in Tunisia was 97%, children in remittance-receiving households are 98.3% in Egypt, and 100% in Jordan. enrolled in primary school, this percentage de- 39 School enrollment for children between 12 and 14 years old, creases to 82 percent in non-receiving households. source World Bank 2007. 91 Keller MNA 5-27-10vol2.indd 91 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options gender gap. These results could be linked with ness, and an extreme low value (below 3 standard the findings for the intra-household allocation in deviations) to stunting. Table A38 shows that the education where remittances help to close the percentage of children, between 2 and 5 years gender distribution gap in female-headed house- old, with shortness and stunting conditions in holds in rural areas. Thus, a more equal distribu- non remittance receiving households doubles the tion of these resources between girls and boys is percentage in receiving households. also reflected in their parity of school enrollment. One possible explanation of why rural women are Children in remittance-receiving house- promoting their daughters’ school literacy could holds are less likely to have malnutrition be found in the MOP evidence which suggests in rural areas. The econometric evidence indi- that women feel extremely constrained in their cates that remittances decrease the probability life chances, including their potential to earn, of being short in children from 2 to 5 years old.40 by their lack of education. Therefore, combin- The analysis also shows that in urban areas remit- ing these two pieces of evidence remittances tances do not affect the height for age indicator could enhance women’s social and economic while in rural areas they do. These results in rural participation in rural areas where their inclusion areas could be linked to the findings of the in- is more lagged. crease of food allocation in remittance-receiving households with female heads, suggesting that Children in remittance receiving house- women’s motivation for increasing the food al- holds present better nutritional indicators location could be to improve their children’s than those in non-receiving households. nutrition. This evidence goes well together with Given the limited nutritional information in the the education results; female heads who have 2001 ENCDM survey, the analysis is focused on some control over the household budget allocate the anthropometric z-score indicator height for the remittances to boost their children’s human age, which measures chronic inadequacies in capital investments. nutrition and/or chronic illness. A low height for age indicator related to the population of refer- ence (below 2 standard deviations of the mean 40 Given the limited data, the nutritional analysis cannot be population of reference) is associated with short- extended by children’s gender. Table A38: Height for age indicator in Children between 2 and 5 years old, 2001 Remittance-Receiving HH Remittance-Non receiving HH % of Children Shortness (Below 2sd) 8.18 15.9 Stunting (Below 3) 3.63 6.71 Source: ENCDM, 2001 92 Keller MNA 5-27-10vol2.indd 92 5/27/10 2:41 PM Appendix 8: Impact of Remittances on Growth: Methodology and Results (From Keller, Mottaghi, van den Bosch, and Adams 2009) This section describes the data on remittances, Empirical Analysis financial development, and economic growth, as well as the control variables used in the growth Estimation Methodology regressions. The data are obtained from the World Bank’s World Development Indicators To test this hypothesis, the paper analyzes the (WDI), and from various recent publications. The impact of remittances on economic growth in two estimated model utilizes data from a sample of models, one without accounting for the financial four developing countries with annual data for sector variable and the second model where the the period of 1980–2007 with 112 observations. interaction of remittances and financial develop- ment variables is taken into account. Standard The estimated model uses M2/GDP, Loan/ financial market indicators were used in growth GDP, Credit/GDP, Deposit/GDP variables as a regressions as proxies for financial development. measure of financial development. All those vari- ables are related to the banking sector.41 The regression to be estimated is as follow- ing: The set of control variables include the fol- lowing: openness to international trade, defined Growthit= α0 + α1 Growthi t–1 + α2 Remittances it as the ratio of the sum of exports plus imports of goods to GDP, gross fixed capital formation to + α3* Xit+ mi + ni+ uit GDP, money supply (M2) to GDP, credit provided by the banking sector, and total labor force.42 All control variables are specified in natural logs. 41 For the estimation purposes, Credit/GDP ratio was used as a proxy for financial development in these countries. Table A39 provides descriptive statistics of 42 The ratio of (M2/GDP) equals currency plus demand and inter- the variables of interest. est bearing liabilities of banks and nonfinancial intermediaries, divided by GDP. It is considered the broadest measure of financial intermediation and includes three types of financial institutions: Table A40 shows bivariate correlations the central bank, deposit money banks, and other financial among the variables of interest. Growth is posi- institutions, and reflects the sum of demand, time, saving, and tively correlated with domestic credit to banking foreign currency deposits to GDP (Deposit/GDP). It measures the ability of banks to attract financial savings and provide a sector, remittances/GDP, openness/GDP, and the liquid store of value. Claims on the private sector divided by measures of financial development and is posi- GDP (LOAN/GDP) measures the extent to which the private tive. Remittances are also positively correlated sector relies on banks to finance consumption, working capital, and investment. Finally, credit provided by the banking sector with openness/GDP, and other measures of finan- to GDP (CREDIT/GDP), measures how much intermediation is cial development. (However causality in either performed by the banking system, including credit to the public direction cannot be concluded) and private sectors. 93 Keller MNA 5-27-10vol2.indd 93 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Table A39: Summary Statistics of Variables Number of Mean Median Maximum Minimum Std. Dev. Observations GDP Growth 1.9 2.2 10.4 –8.1 3.3 108 Credit/GDP 73.1 71.7 120.0 –3.5 23.5 108 Fixed capital/GDP 24.8 24.4 34.5 16.3 4.3 108 remittances/GDP 5.0 4.7 14.6 0.5 3.0 108 Openness/GDP 63.4 59.0 110.7 32.7 17.7 108 RemCredit 398.0 348.0 1480.8 –5.4 313.4 108 RemMoney 324.7 270.9 1132.6 36.6 256.8 108 M2/GDP 60.8 57.1 97.5 31.0 17.1 108 Table A40: Correlation matrix of variables Fixed GDP Credit/ capital/ Remittances/ Openness/ M2/ Growth GDP GDP GDP GDP RemCredit RemMoney GDP GDP Growth 1 Credit/GDP 0.02 1 Fixed capital/GDP –0.10 –0.01 1 remittances/GDP 0.16 0.43 –0.14 1 Openness/GDP 0.23 –0.22 0.12 0.02 1 RemCredit 0.15 0.63 –0.07 0.94 –0.06 1 RemMoney 0.17 0.56 –0.12 0.94 –0.07 0.97 1 M2/GDP 0.08 0.73 –0.12 0.37 –0.32 0.54 0.63 1 Where Growth it is the change in the real per possible to determine the impact of remittances capita GDP in constant dollar, Xit is the control on growth through the financial development variables, Mi is the time specific effect, ni is the variables. country specific fixed effects, uit is the error term, where remittances refer to log of remittances In the second regression, the role of remit- over GDP. We are interested in testing whether tances on growth through financial markets is the marginal impact of remittances on growth, examined. The hypothesis is whether the recipi- a2 and thus a3 are statistically significant. ent country’s financial depth could influence the impact of remittances on growth. To this end, the To explore the relationship between remit- remittance variable is allowed to interact with tances, financial development and growth, the an indicator of financial depth and test for the estimated model utilizes a panel (cross-country, significance of the interacted coefficient. A nega- time series) dataset consisting of four developing tive coefficient would indicate that remittances countries with annual data for 28 years. Two mod- are more effective in countries with shallower els are estimated. First the model is estimated financial systems; in other words, evidence of without the interaction of financial development substitutability between remittances and finan- variables. However in the second model, remit- cial instruments. On the other hand, a positive tances are allowed to interact with one of the interaction would imply that the growth effects financial development variables. This makes it of remittances are enhanced in deeper financial 94 Keller MNA 5-27-10vol2.indd 94 5/27/10 2:41 PM Appendix 8: Impact of Remittances on Growth: Methodology and Results systems, supporting complementarity of remit- a positive and significant impact on economic tances and other financial flows. growth in both models. However the impact of remittances on growth is more pronounced when The method used is Panel system General- the financial development variable is included, ized Methods of Moments regressions (GMM) which is the case in the second model. to control for endogeneity and serial correlation following Arellano and Bover (1995) taking into In terms of magnitude, one percentage account time specific effects.43 Use of GMM es- point increase (decrease) in remittances as a timator overcomes the problem of endogeneity percentage of GDP, other things being equal, of the columns of X as well as the correlation increases (decreases) GDP growth by 0.05 between the new error term and the lagged (0.01) percentage points in the second (first) difference of the dependent variable. Taking model. The coefficient sign of the interaction advantage of the panel nature of the data, GMM term (remittances *domestic credit) is negative estimators are based on differencing repressors implying that remittances can act as a substitute to control for unobserved effects. for the financial sector variable. By offering a response to the needs for credit and insurance In all regressions, two lags of all endogenous that the market has failed to provide. In other variables are used as instruments for all strictly words remittances have contributed to promote non-exogenous variables, including the remit- growth in countries with shallower financial sys- tances and financial depth indicators. In addition, tems. These models estimate the overall effect autocorrelation tests and the Hansen test of over of remittances on growth in a panel system. It is identifying restrictions are performed to assess important to note that the results for individual the validity of the instruments employed. countries might be different The model shows that all coefficients are sig- nificant at 5 percent, except for the labor force in 43 Arellano, M., and, O. Bover. 1995. “Another Look at the the second model that is significant at 10 percent. Instrumental-Variable Estimation of Error-Components Models.â€? The results suggest that remittance flows have Journal of Econometrics 68: 29–52. Table A41: Estimated Results (Dependent variable is GDP per capita growth) w/o Financial Dev. Variable w/ Financial Dev. Variable Variable Coefficient t-stat Coefficient t-stat Growth(–1) –0.40 –3.19 –0.48 –2.93 Remittances (–1) 0.01 2.46 0.05 2.01 Capital 0.06 2.96 0.08 2.38 Labor 0.07 8.24 0.27 1.68 Credit 0.07 2.41 Remittances *credit –0.06 –2.31 Note: All variables are represented in natural logs. The model Include common time dummies. 95 Keller MNA 5-27-10vol2.indd 95 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Keller MNA 5-27-10vol2.indd 96 5/27/10 2:41 PM Appendix 9: Return Migration and Occupational Mobility (From Gubert and Nordman 2008a) Data three modules relating to the different migratory stages: the returnees’ conditions before they left The data used in this study are drawn from for abroad; the returnees’ experience of migra- the three recent surveys that were simultane- tion lived abroad; and the returnees’ post-return ously conducted in 2006 on returned migrants conditions in the country of origin. in Algeria, Morocco, and Tunisia as part of the Return Migration to the Maghreb (MIREM) proj- The Determinants of Occupational ect (see Cassarino, 2008, and www.mirem.eu/ Mobility Between Pre- and Post- mirem?set_language=en, for further details on the Migration Periods whole project). About 330 returned migrants were interviewed in each country using a common ques- The objective of this section is to identify the tionnaire. In each country, the sampling procedure characteristics of those migrants who moved was based on a geographical stratification process. in status between the pre- and post-migration A few specific regions were selected using official periods, and the various factors that explain this statistics on return flows, so that the survey data mobility. Here, mobility is defined as mobility cannot be viewed as reflecting national trends. between occupational groups. An alternative ap- proach would be to rely on a continuous variable According to the team of the MIREM project, capturing, for instance, the migrants’ labor mar- a returnee is defined as “any person returning ket outcomes or their living standards as mea- to his/her country of origin, in the course of sured by earnings or household consumption. the last ten years, after having been an in- Unfortunately, such indicators are not available ternational migrant (whether short-term or in the MIREM database. Nonetheless, there is long-term) in another country. Return may also one objective reason for favoring a measure be permanent or temporary. It may be inde- of mobility based on occupation rather that on pendently decided by the migrant or forced a continuous economic variable: occupational by unexpected circumstances.â€? This definition status is indeed generally considered as a good partially draws on the one recommended by the indicator of social status as it encompasses other United Nations. It refers specifically to migrants dimensions including living standards, prestige, who returned to their country of origin in the and eventually power. course of the last ten years because this time limit allows the impact of the experience of migration Methodological issues on the interviewee’s pattern of reintegration to be assessed. It also allows the respondents to In order to measure the extent to which migrants recount their migratory experiences more pre- shift jobs within the occupational distribution, cisely. The questionnaire is structured around we first need to classify all occupations in a 97 Keller MNA 5-27-10vol2.indd 97 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options hierarchical order. To this end, we rely on the housing, one dummy for land ownership and methodology presented in Pasquier-Doumer one dummy taking the value 1 if the migrant (2005), which is based on a hierarchical cluster owns other durable goods. Subjective mea- analysis of occupational groups. The objective sures are dummies indicating whether the of this method is twofold: (1) to classify all the individual considers him/herself in a good, occupations that are listed in the MIREM data- medium or bad financial situation. The level base into a small number of mutually exclusive of education is measured in the pre-migration groups or clusters based on their similarities or period and comprises 9 dummies. dissimilarities; and (2) to establish a hierarchy • The Xs are then averaged over the 9 initial among the clusters. occupational categories. These indicators are then introduced in the hierarchical The association method retained is agglom- cluster analysis of these categories. We use erative and relies on a similarity index measuring the ‘cluster completelinkage’ command of the distance between the different occupations STATA. This command performs a hierarchi- on the basis of different criteria (X). Formally, cal agglomerative complete linkage cluster the retained measure of similarity between two analysis. Due to small sample size, all the occupations i and j is the Euclidean distance L2 data were pooled so that the analysis is (or dissimilarity index) given by: conducted on a single sample of returnees (n = 731). • The results of this analysis allow us to define the following mutually exclusive groups or clusters: • Group 1: Employers or Employees with where k is one of the p variables used to establish secured job : Occupations 5 & 1 the classification. • Group 2: Employees with secured part- time/short term jobs : Occupations 2 & 3 The whole procedure can be summarized in • Group 3: Small self-employed : Occupa- the following way: tion 6 • Group 4: Employees with unsecured jobs • We use information on occupational status or unemployed : Occupations 4, 7 & 9 prior to migration as our unit of analysis All • Group 5: Family workers : Occupation 8 those returnees who were out of the labor • In a last stage, we use information on occu- market prior to migration (i.e., who were pational status after return. By comparing either students, retired/inactive or home- the groups (1, 2, 3, 4 or 5) to which he/she makers) were dropped from the analysis, so belongs to before and after the period of we are left with 9 occupations. The purpose migration, we are able to determine whether is then to define broader groups of these 9 the migrant experienced an upward or a occupations so as to regroup similar occupa- downward occupational mobility. Three tions on the basis of the variables X. These variables are then constructed: an upward new groups, mutually exclusive, will hence mobility dummy (taking value 1 if the mi- be made of similar occupations according to grant experienced an upward mobility be- these Xs. tween the pre- and post-migration periods, • The different criteria X retained for the 0 otherwise); a downward mobility dummy cluster analysis are variables measuring the (taking value 1 if the migrant experienced migrant’s living standards and level of educa- a downward mobility between the pre- and tion. In order to assess the migrant’s living post-migration periods, 0 otherwise); and a standards (during the pre-migration period), variable taking value –1 if the migrant moved we use three objective and one subjective downward in occupation, 0 if he/she stayed measures of his/her financial situation. Objec- in the same occupational group and 1 if he/ tive measures include 8 dummies for type of she moved upward. 98 Keller MNA 5-27-10vol2.indd 98 5/27/10 2:41 PM Appendix 9: Return Migration and Occupational Mobility Econometric model The second block includes variables con- trolling for family status prior to migration, i.e., With our alternative measures of occupational whether the individual was married and the size mobility, our intention now is to disentangle the of his/her household (prior to migration). possible determinants of occupational mobility between pre- and post-migration periods. To The third block comprises controls for the this end, we first estimate two Probit models of occupational status of the migrant prior to migra- upward and downward occupational mobility tion. It is likely indeed that occupational mobility, respectively. In the first model, the dependent either upward or downward, is partly conditioned variable is a dichotomous variable taking value by the occupational level from which the migrant 1 if the migrant moved upward in occupation, starts. Of course, in the Probit model of upward 0 otherwise; in the second one, the dependent mobility, those returnees who were employers or variable is a dichotomous variable taking value employees with secured jobs prior to migration 1 if the migrant moved downward in occupa- (i.e., those who were in Group 1) are excluded tion, 0 otherwise. The difficulty, however, is from the analysis since, by construction of the that occupational mobility is not observed for mobility variable, it is impossible for them to those returnees who are out of the labor market. experience an upward mobility. Conversely, in Restricting our analysis to the sole returnees the Probit model of downward mobility, it is having (or seeking) a job after return would those returnees who were in the lowest group be fine if occupational data were missing at that are excluded since it is impossible for them random. This is doubtful, though, since those to experience a downward mobility. returnees who are out of the labor market are likely to constitute a self-selected sample. It is A fourth block of determinants includes indeed likely that some of the returnees who characteristics of the migrants’ overseas stay. would have moved downward in occupational These are important covariates deemed to in- status after return chose not to integrate the fluence the probability of professional success labor market. We thus need to take this selectiv- or failure after return. Among them, we include ity issue into account in order to get unbiased proxies of human capital accumulated abroad results. To this end, we use a Heckman selection such as whether the migrant worked when he model. We first estimate a dichotomous model was abroad or whether he/she received voca- predicting whether the individual is active (se- tional training. We also include one variable lection model) and then estimate a dichotomous measuring migration duration as a proxy for model of upward (or downward) occupational professional experience in the labor market of mobility. We use age at the time of the survey the receiving country and for skill acquisition. A to predict selection. dummy variable indicating whether the diploma held by the migrant (if any) has been recognized Estimation results are reported in Tables in the destination country is introduced in order A41 and A42. Column 1 shows estimation results to account for misuse of human capital abroad obtained on a pooled sample while Columns 2 to 4 (the so-called “brain wasteâ€?) or, on the contrary, show estimation results obtained on the Algerian, appropriate human capital transfer from origin to Moroccan and Tunisian samples respectively. destination countries. Last, a set of destination country dummies is considered. These variables Six blocks of independent variables are in- may capture environmental, institutional, or troduced in the model. network effects in the last immigration country that may affect the migrants’ success or failure The first block includes demographic char- after return. acteristics of the migrant, such as being female, age measured at time of migration, whether the A fifth block of independent variables that individual was born in an urban area, and whether includes three dummies scales the amount of he/she is bi-national. remittances the migrant used to send when he/ 99 Keller MNA 5-27-10vol2.indd 99 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Table A41: Probit Model of Upward Occupational Mobility After Migration with Selection Pooled sample Algeria Morocco Tunisia Individual characteristics Female 0.139 0.062 –0.153 –0.589 (0.42) (0.14) (0.14) (0.66) Age (at time of migration) 0.016 –0.003 0.004 0.015 (1.23) (0.11) (0.15) (0.50) Born in urban area 0.048 –0.069 0.101 0.240 (0.29) (0.23) (0.25) (0.81) Bi-national 0.144 –0.425 0.157 0.974* (0.53) (0.78) (0.16) (1.93) Family status before migration Married before migration –0.233 0.099 –0.725 –0.906** (1.21) (0.26) (1.42) (2.43) Household size before migration 0.005 0.006 0.018 –0.001 (0.21) (0.15) (0.28) (0.03) Occupational status prior to migration (ref. is Group 2) Was in Group 3 –0.440* –0.531 –1.233* –0.352* (1.84) (1.27) (1.69) (0.72) Was in Group 4 0.529*** 0.608* 0.328 0.741* (2.79) (1.83) (0.66) (1.92) Was in Group 5 2.705*** — 3.038*** — (4.52) (4.20) Characteristics of overseas stay Worked during last migration –0.044 –0.211 –0.195 –0.477 (0.26) (0.66) (0.37) (1.48) Trained during migration 0.158 0.287 0.026 –0.068 (0.78) (0.66) (0.05) (0.20) Diploma recognized (1: yes) 0.229 0.186 0.210 0.255 (1.09) (0.47) (0.33) (0.66) Migration duration (in years) 0.016 –0.045*** –0.055** 0.017 (1.55) (3.17) (2.00) (1.18) Past immigration country (dummies included but not shown) Past remittance behavior (ref. is sent nothing) Sent less than 500 per year –0.057 –0.377 –0.476 –0.693* (0.25) (0.78) (0.77) (1.70) Sent between 501 and 1000 per year 0.111 0.532 –0.249 –0.100 (0.54) (1.43) (0.47) (0.22) Sent more than 1000 per year 0.393** 0.726 0.744 0.346 (1.98) (1.49) (1.47) (1.03) Return conditions Time elapsed since return 0.048*** –0.024 –0.005 0.056 (2.70) (0.51) (0.17) (1.37) Back to birth place –0.002 0.376 –0.229 –0.263 (0.01) (1.09) (0.58) (0.90) Returned for administrative reasons –0.719*** –0.583 –0.774 –1.043*** (3.81) (1.55) (1.59) (2.76) (continued on next page) 100 Keller MNA 5-27-10vol2.indd 100 5/27/10 2:41 PM Appendix 9: Return Migration and Occupational Mobility Table A41: Probit Model of Upward Occupational Mobility After Migration with Selection (cont.) Pooled sample Algeria Morocco Tunisia Location after return is capital city 0.113 –0.950* 1.558* –0.711* (ref. is small city) (0.61) (1.75) (2.11) (1.76) Location after return is secondary city 0.042 –0.339 1.458** –0.514 (0.23) (0.97) (2.24) (1.36) Algerian returnees –0.073 — — — (0.35) Moroccan returnees –0.312 — — — (1.43) Constant –0.824 0.161 0.081 0.098 (1.52) (0.19) (0.06) (0.09) Observations 489 184 139 166 Absolute value of z statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. Table A42: Probit Model of Downward Occupational Mobility After Migration with Selection Pooled sample Algeria Morocco Tunisia Individual characteristics Female –0.299 –0.997 –0.092 — (0.81) (1.54) (0.19) Age (at time of migration) –0.005 –0.027 0.013 0.004 (0.41) (1.30) (0.46) (0.11) Born in urban area –0.093 –0.562* 0.527* –0.438 (0.50) (1.93) (1.72) (1.23) Bi-national 0.115 –0.261 — 0.056 (0.34) (0.50) (0.11) Family status prior to migration Married before migration –0.204 –0.039 –0.755* –0.777* (0.95) (0.11) (1.82) (1.74) Household size before migration 0.031 0.058 0.019 0.004 (1.19) (1.42) (0.42) (0.08) Occupational status prior to migration (ref. is Group 5) Was in Group 1 2.025*** 1.420*** 1.483*** 1.623*** (6.94) (3.87) (2.74) (3.64) Was in Group 2 2.298*** 1.887*** 1.561*** 2.004*** (7.54) (4.65) (3.34) (4.01) Was in Group 3 1.523*** — — 2.159*** (5.00) (3.94) Characteristics of overseas stay Worked during last migration 0.776*** 1.363*** 1.591*** 0.442 (3.37) (3.38) (3.23) (1.10) Trained during migration –0.240 –0.495 –0.414 –0.351 (1.04) (1.36) (0.81) (0.82) Diploma recognized (1: yes) –0.579*** — — –0.734** (2.85) (1.96) (continued on next page) 101 Keller MNA 5-27-10vol2.indd 101 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Table A42: Probit Model of Downward Occupational Mobility After Migration with Selection (cont.) Pooled sample Algeria Morocco Tunisia Migration duration (in years) –0.003 –0.027 0.051*** 0.024 (0.28) (1.79) (2.61) (1.30) Past immigration country (dummies included but not shown) Past remittance behavior (ref. is sent nothing) Sent less than 500 per year 0.398 0.359 0.337 0.573 (1.55) (0.70) (0.65) (1.32) Sent between 501 and 1000 per year –0.000 0.408 –0.315 –0.082 (0.00) (1.07) (0.86) (0.17) Sent more than 1000 per year –0.309 0.083 –0.161 –0.537 (1.46) (0.24) (0.41) (1.29) Return conditions Time elapsed since return –0.040 –0.014 –0.035 –0.051 (1.59) (0.31) (0.84) (1.02) Back to birth place 0.047 –0.574 0.189 0.012 (0.26) (1.52) (0.46) (0.04) Returned for administrative reasons 0.634*** 0.293 –0.001 0.933*** (2.84) (0.79) (0.00) (1.98) Location after return is capital city 0.302 0.550 –0.650 0.408 (ref. is small city) (1.63) (1.60) (1.35) (0.97) Location after return is secondary city –0.036 0.339 –0.427 –0.143 (0.16) (1.12) (0.89) (0.28) Algerian returnees 0.151 — — — (0.69) Moroccan returnees 0.087 — — — (0.34) Constant –3.092*** –2.190** –3.642*** –2.670* (5.09) (2.49) (3.94) (1.95) Observations 673 278 152 243 Absolute value of z statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. she was abroad is included as well (the reference while conditions of return are captured by a being no remittances). Indeed, migrants may face dummy variable indicating whether the migrant capital market imperfections in their home coun- deliberately chose to return or was forced to do try so that overseas savings and remittances are so. In addition, three dummies controlling for the subsequently able to fuel productive investments potential effect of location after return are used: (Mc Cormick and Wahba 2001). For this reason, a dummy for being back to the birthplace, and this information may affect migrants’ professional two dummies for the size of the city (capital and trajectories. As there is no direct measure of secondary city, the reference being a small city). overseas savings in the MIREM survey, we use these remittances dummies to control for the Estimation results effect of savings. Let us first describe the effects of demographic A last block of independent variables is variables on the probability of experiencing an oc- included to control for conditions and timing of cupational mobility. Most variables (in particular return. Time elapsed since return controls for gender, age, household size prior to migration) are labor market experience in the home country actually insignificant in both groups of regressions 102 Keller MNA 5-27-10vol2.indd 102 5/27/10 2:41 PM Appendix 9: Return Migration and Occupational Mobility (Tables A41 and A42). There are a few excep- A high amount of remittances is significantly tions, however. Being married prior to migration is and positively associated with upward mobility. found to reduce the probability of moving upward This result suggests that the higher the remit- within the whole sample, this result being driven tances, the lesser the budget constraint after by Tunisian returnees for whom the effect of the return, an important determinant of entrepre- variable is marginally strong. An analogous result neurship behavior. is found for Moroccan returnees, for which being married prior to migration is found to increase the Conditions of return also seem to play an probability of moving downward. Being bi-national important role in the migrants’ professional has an impact on upward mobility but the effect trajectories. In particular, those migrants who is significant only in the Tunisian case. returned for administrative reasons are less likely to experience an upward mobility. The ef- Introducing controls for occupational status fect is particularly strong in the case of Tunisian prior to migration provides evidence that mobil- returnees. “Forcedâ€? returns are indeed likely to ity is in large part conditioned by the migrants’ be unprepared returns and to negatively affect initial position in the distribution of occupations. the migrants’ professional reintegration in their More specifically, those returnees who were home country. A symmetric result is found in family workers prior to migration (i.e., those the model of downward mobility. Surprisingly returnees who were in Group 5) and those who enough, residing in the capital city after return were in Group 4 (small self-employed) are much is negatively associated with upward mobility more likely to experience an upward mobility in both Algeria and Tunisia, and positively in after their overseas stay than those returnees the case of Morocco. The result for Algeria and who had secured part-time or short term jobs Tunisia may be a sign of a lack of “goodâ€? jobs in (Group 2). The effect is particularly strong for Algiers and Tunis labor markets. Moroccan migrants. With regard to downward mobility, it is those returnees who were previ- Finally, the set of dummies for destination ously in Group 1 and Group 2 (employees with countries display few significant coefficients secured part-time or short term jobs) that are (the reference being migration to France). This much more likely to experience a downward result suggests that the receiving country has mobility. Returnees originating from Group 2 in little effect on the probability of having upward/ particular appear much more at risk of experi- downward mobility after migration once demo- encing a downward mobility. graphics, education, professional status, overseas stay and return characteristics of the migrants Characteristics of overseas stay are not sig- are accounted for. nificant at all in the models of upward mobility, except for migration duration. In the case of Al- In order to both improve and check the gerian and Moroccan returnees, those who stayed robustness of our results, we ran alternative abroad for a short period of time are more likely regressions using other mobility measures and to experience an upward mobility. Turning to the other estimators. In particular, we estimated an models of downward mobility, our proxy for skill ordered Probit model in which the dependent acquisition abroad (measured by a dummy vari- variable ranged from 1 (high-quality jobs) to 5 able taking value 1 if the migrant worked when he (low-quality jobs). was abroad, and 0 otherwise) is significant in most regressions but with an unexpected sign (working Regression results are displayed in Table A43. abroad being positively associated with experienc- To ease interpretation, note that a positive sign of ing a downward mobility). This suggests that our the coefficient means that the variable increases proxy is not a proper measure of skill acquisition the likelihood of a downward mobility, while a and controls for something else. By contrast, those negative sign means that the variable increases migrants whose diploma was recognized are less the likelihood of an upward mobility. Overall, no likely to move downward after return. difference emerges as compared to the previous 103 Keller MNA 5-27-10vol2.indd 103 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options estimates. Here again, occupational status prior pational mobility. The same holds true for return to migration is found to strongly explain occu- conditions, past remittance behavior, etc. Table A43: Ordered probit of occupational mobility after migration (from high to low quality jobs) Pooled data Algeria Morocco Tunisia Individual characteristics Female –0.312 –0.398 0.560 –0.698 (1.28) (1.11) (1.00) (1.22) Age (at time of migration) –0.003 –0.016 –0.003 –0.002 (0.35) (0.90) (0.16) (0.11) Born in urban area –0.176 –0.174 –0.605** –0.163 (1.35) (0.69) (2.21) (0.69) Bi-national –0.331 0.641* –1.169 –0.915** (1.51) (1.70) (1.28) (2.32) Family status before migration Married before migration 0.070 –0.059 –0.362 0.270 (0.49) (0.21) (1.21) (0.98) Household size before migration 0.018 0.042 0.009 0.008 (0.98) (1.27) (0.23) (0.23) Occupational status prior to migration (ref. is Group 1) Was in Group 2 0.696*** 0.979*** –0.020 0.903*** (3.71) (2.92) (0.05) (2.60) Was in Group 3 0.719*** 0.866*** 0.382 0.914*** (4.53) (3.13) (0.92) (3.34) Was in Group 4 1.454*** 1.511*** 1.274*** 1.771*** (7.36) (4.27) (2.76) (4.30) Was in Group 5 0.902*** 0.432 0.625 –7.398 (2.82) (0.36) (1.18) (0.00) Characteristics of overseas stay Worked during last migration 0.467*** 0.930*** 0.729** 0.135 (3.32) (3.68) (2.15) (0.54) Trained during migration –0.190 –0.373 –0.280 –0.158 (1.23) (1.23) (0.69) (0.60) Diploma recognized (1: yes) –0.438*** –0.190 –1.360*** –0.121 (2.96) (0.78) (3.21) (0.47) Migration duration (in years) 0.006 –0.006 0.035** 0.005 (0.97) (0.50) (2.16) (0.45) Past immigration country (dummies included but not shown) Past remittance behavior (ref. is sent nothing) Sent less than 500 per year 0.249 0.957*** –0.204 0.464 (1.42) (2.66) (0.55) (1.56) Sent between 501 and 1000 per year 0.047 0.472 –0.500 –0.104 (0.28) (1.53) (1.56) (0.30) Sent more than 1000 per year –0.255* –0.082 –0.588* –0.437* (1.73) (0.26) (1.84) (1.73) (continued on next page) 104 Keller MNA 5-27-10vol2.indd 104 5/27/10 2:41 PM Appendix 9: Return Migration and Occupational Mobility Table A43: Ordered probit of occupational mobility after migration (from high to low quality jobs) Pooled data Algeria Morocco Tunisia Time elapsed since return 0.011 0.015 0.022 0.018 (0.70) (0.42) (0.82) (0.58) Returned for administrative reasons 0.410*** –0.583 –0.774 –1.043*** (2.86) (1.55) (1.59) (2.76) Back to birth place –0.085 –0.213 0.322 –0.098 (0.69) (0.79) (1.20) (0.45) Location after return is capital city 0.317** 0.548** –0.249 0.190 (ref. is small city) (2.24) (1.96) (0.69) (0.65) Location after return is secondary city 0.071 –0.067 –0.575 0.006 (0.46) (0.25) (1.32) (0.02) Algerian returnees 0.347** — — — (2.27) Moroccan returnees 0.248 — — — (1.44) Observations 506 183 129 194 Pseudo R-squared 0.17 0.21 0.25 0.20 Absolute value of z statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. 105 Keller MNA 5-27-10vol2.indd 105 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Keller MNA 5-27-10vol2.indd 106 5/27/10 2:41 PM Appendix 10: Return Migration and Entrepreneurship Analysis (From Gubert and Nordman 2008b) Descriptive statistics from the return migrant E = 1 if E* > 0 survey suggest that entrepreneurs among re- turnees are on average different in some ways E = 0 if E* ≤ 0 from non-entrepreneurs: they are more likely to be male, are younger, have neither low nor high where E* is a latent variable measuring the pay- education levels, etc. In addition, the probability off from becoming an entrepreneur after return. of becoming an entrepreneur after return seems We assume that E* = bX + ε , where X is a vec- to be higher for returnees with a first experience tor of independent variables and ε, a normally as employers or self-employed, for those who distributed error term. received vocational training whilst abroad and for those who independently and freely chose Six blocks of independent variables are in- to return. troduced in this model. The purpose in this section is to construct The first block includes demographic charac- an econometric model of the probability of a teristics of the migrants such as sex, age, region returnee to become an entrepreneur in order of origin (the reference being rural), and being to examine whether these correlations hold in a bi-national. multivariate analysis. In order to fuel the discus- sion, estimation results will be compared to those The second block contains five education found in other studies focusing on the same issue dummies reflecting schooling attainment at but in other countries (in particular McCormick the time of the survey, namely primary cycle, and Wahba 2001; Ilahi 1999; Ammassari 2003; secondary cycles (I and II), university level (till and Black, King, and Tiemoko 2003) the fourth year of higher education) and higher degrees above the fourth year of university (the Econometric Model reference being no schooling). We estimate the Probit version of a discrete The third block comprises controls for the choice econometric model where the dependent occupational situation of the migrant prior to variable is a dummy variable taking the value 1 if migration. More precisely, a dummy for be- the returnee has become an entrepreneur since ing an entrepreneur prior to migration (the return, and 0 otherwise, using the restricted reference being any other occupation) is in- definition for an entrepreneur. cluded. The idea is to find out whether being an entrepreneur before migration affects the Formally, the model may be written as fol- probability of taking up this occupation upon lows: return once sociodemographic characteristics 107 Keller MNA 5-27-10vol2.indd 107 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options of the returnees and conditions of their return vironmental, institutional or network effects in are accounted for. the last immigration country that may affect the migrants’ success or failure after return. A fourth block of determinants includes char- acteristics of the migrants’ overseas stay. These Estimation results are important covariates deemed to influence the probability of professional success or failure Estimation results are reported in Table A44 To after return. Among them, we include proxies ease their interpretation, only marginal effects of human capital accumulated abroad such as of the covariates are shown. Interesting features whether the migrant worked when he/she was emerge. abroad or whether he/she received vocational training. We also include one variable measuring First, in line with what was suggested by migration duration as a proxy for professional descriptive statistics, female migrants are sig- experience in the labor market of the receiving nificantly less likely to become entrepreneur country and for skill acquisition. Three dummies after return, all else being equal. The effect is scaling the amount of remittances the migrants particularly strong for Tunisian migrants (with used to send before returning to their home a marginal effect of –0.34 compared to –0.14 for countries are included as well (the reference be- Algeria). Turning to the age variable, its expected ing no remittances). Indeed, migrants may face effect on entrepreneurial behavior is unclear. As capital market imperfections in the origin coun- argued by Ilahi (1999), if age is synonymous with try so that overseas savings and remittances are labor market experience, and wages rise with subsequently able to fuel productive investments experience, then age should be negatively asso- (McCormick and Wahba 2001). For this reason, ciated with the probability for self-employment this information may affect migrants’ professional or, turning it the other way round, positively trajectories. As there is no direct measure of associated with waged work. On the other hand, overseas savings in the MIREM survey, we use age may have a positive influence on managerial these remittances dummies to control for the talent and hence on the likelihood of becoming effect of savings. an entrepreneur. Estimation results suggest that the latter effect dominates for Algerians and Mo- A fifth block of independent variables is roccans while the opposite is true in the Tunisian included to control for conditions and timing of sample where age appears to be detrimental to return. Time elapsed since return controls for becoming an entrepreneur. With regard to the labor market experience in the home country returnees’ other characteristics, originating while conditions of return are captured by a from an urban area is positively associated with dummy variable indicating whether the migrant the probability of taking up an entrepreneurial deliberately chose to return or was forced to do job in Tunisia. In the case of Algerian migrants, so. A dummy variable indicating whether the having double nationality is also strongly linked returnees plan to re-migrate is also introduced. to engagement in entrepreneurial activities. This variable is indeed likely to affect entrepre- neurial behavior if return migrants consider their Strong positive impacts of education are comeback as a transitory period. found for all countries. For Algerians for instance, the education dummies are all significant at the Finally, three dummies controlling for the 10 percent level and disclose an increasing mar- potential effect of location after return are used: ginal effect from the primary till the university a dummy for being back to the birthplace, and level: Algerian returnees holding a university de- two dummies for the size of the city (capital and gree are indeed 47 percent more likely to become secondary city, the reference being a small city). entrepreneurs after returning compared with the reference category of no-education against 20 Last, a set of destination country dummies percent only for Algerian returnees who dropped is considered. These variables may capture en- out after primary school. Interestingly enough, 108 Keller MNA 5-27-10vol2.indd 108 5/27/10 2:41 PM Appendix 10: Return Migration and Entrepreneurship Analysis the reverse holds true for Tunisian returnees. ing value 1 for illiterates, they find no significant Those with high university degrees do not have influence from longer periods overseas on the an entrepreneurial behavior significantly differ- likelihood of becoming an entrepreneur amongst ent from those with no schooling. For Moroccans, illiterates. Following this approach, similar inter- the impact of education is less pronounced, espe- acted terms were computed and introduced in cially at intermediate levels of schooling. Holding the regressions but they were ultimately dropped a high university degree actually exerts a positive for lack of significance. Of course, it could be ar- and significant impact on entrepreneurial behav- gued that migration duration and activity choice ior in the Algerian case only. after return are jointly chosen and hence that the regression results presented so far suffer from An expected positive effect of being an an endogeneity bias. This issue raised by Dust- entrepreneur before migration is found for all man (2002) is investigated more thoroughly in countries. The impact is more significant and the next section. of a greater magnitude for Algerian returnees: previous Algerian entrepreneurs are about 27 Turning to the other regressors relating to percent more likely to become entrepreneurs characteristics of overseas stay, past remittance after returning against 19 percent and 18 percent behavior is found to have a positive effect on respectively for Moroccans and Tunisians. This the probability of becoming an entrepreneur for result corroborates the idea according to which, Moroccans and Tunisians. This is an expected all else being equal, entrepreneurial engagement result as this information accounts for savings, upon return is conditioned by previous experi- which are clearly an important asset for being ence in related activities. McCormick and Wahba able to open a business. (2002) and by Ilahi (1999) report a similar result in the case of Egyptian and Pakistani returnees Conditions of return appear to be strong respectively. determinants of the probability to engage in entrepreneurial activities upon return. First, Among the characteristics of overseas stay time elapsed since return is always positively that are considered, vocational training overseas correlated to entrepreneurship. This finding is is positively and significantly associated with probably reflective of a positive effect of return- entrepreneurship for Moroccan and Tunisian ees’ human capital accumulation after return, returnees. As discussed before, however, any namely experience and knowledge gained of the causal relationship between these two variables local market conditions and rules for running a is risky to ascertain, for training may be endog- business. This may also reveal the existence of enously determined in this type of model. With a minimum required time for gathering financial regard to migration duration, the usual assump- resources once back. Second, a “forced returnâ€? tion is that the longer the time spent overseas, is negatively associated with the probability of the greater the opportunity for skill acquisition. setting up entrepreneurial activities, especially As a result, migration duration is expected to for Moroccan and Tunisian returnees. Another in- positively influence entrepreneurship. Surpris- teresting finding is that planning to re-migrate is ingly enough, regression results suggest that negatively correlated to entrepreneurship, for all migration duration discloses a positive impact in countries. This is somewhat an expected result the Tunisian case only. The influence of migration as re-migration is not compatible with a desire to duration is not found to be significantly different engage time and financial resources in the home from zero in the case of Morocco and Algeria. country’s labor market. Last, migrants’ location These results are in sharp contrast with those after return appears to be a significant determi- found by McCormick and Wahba (2002). Using a nant of entrepreneurial activities, especially for sample of Egyptian returnees, they find that time the sample of Moroccan returnees. Unlike their spent overseas has a positive and highly signifi- Algerian and Tunisian counterparts, Moroccan cant effect on being an entrepreneur. However, migrants engage more in businesses when they after interacting the variable with a dummy tak- go back to their birthplace, all else being equal, 109 Keller MNA 5-27-10vol2.indd 109 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options and when they return to relatively large cities. By return, in particular as a determinant of becom- contrast, as far as entrepreneurship is concerned, ing entrepreneur. The second approach follows Algerians and Tunisians do not seem to benefit previous work by Dustmann and Kirchkamp from returning to the capital city. As suggested (2002) who deal with the potential simultaneity by Ilahi (1999), this last result may be due to the of migration duration and activity choice deci- fact that urban areas offer better access to waged sions after return. employment and raise the opportunity cost of self-employment. Endogeneity of migration duration Finally, turning to the role of the last im- Is migration duration exogenous in the decision migration country, we find little evidence of a of becoming an entrepreneur? The potential decisive impact of the last destination country problem arises when unobservable character- on the probability of becoming an entrepreneur istics of migrants that affect their probability of after return. For Moroccan and Tunisian re- becoming entrepreneur after return influence, at turnees, however, having migrated to Italy and the same time, their migration duration. As un- Germany respectively plays a significant role in observed factors for instance, we could think of the probability of entrepreneurship after return. migrants’ ability (innate or acquired) to integrate This result somewhat conforms to previous in the labor markets in origin and host countries, statistical findings on the over-representation such as perseverance or talent. When such un- of entrepreneurs among migrants who went to observed individual endowment is at work in Italy and Germany in our returnee samples. We the determination of migration duration and in now find that these effects persist once socio- post-migration decisions, the migration duration demographics and conditions of overseas stay variable is endogenously determined in the Probit and return are accounted for. equation of becoming an entrepreneur. Ignoring this issue in the Probit equation can then result As a robustness check, we run the same in biased estimated determinants. Probit regressions using the extended defini- tion of being an entrepreneur after return as the A way to correct this potential endogeneity dependent variable. We observe that the pattern bias is to use instrumental variable techniques of the determinants of entrepreneurship is very that consist in identifying (and introducing) a similar with this extended definition, thus indi- set of variables that affect migration duration, cating that the main previous findings are robust but not activity choice upon return. From the to changing definition of entrepreneurship. The MIREM questionnaire, we identified three types few noticeable changes concern the facts that: of potential instruments: variables controlling for being an entrepreneur before leaving is no more entry conditions in the last immigration country significant for Moroccan returnees; planning to (whether the individual entered legally; the type re-migrate becomes insignificant for Tunisian mi- of visa used), the number of children born during grants; and coming back to their birthplace is no the migration period, and whether the migrant more significant for Moroccans. Also, the impact experienced a change in his/her matrimonial of migration to Italy becomes insignificant for status during this period. These variables are Moroccan returnees, as it was for their Algerian assumed to affect migration duration, but not and Tunisian counterparts. activity choice upon return independently from one another. Robustness checks and additional results The different tests we performed could not reject these necessary assumptions for Moroccan In this section, we report two types of robustness and Algerian returnees. However, the chosen checks and alternative modeling. One tackles instruments performed poorly in the Tunisian the potential endogeneity of migration duration case. In the Algerian and Morocco cases, the in the determination of activity choice after endogeneity-corrected results (not reported) 110 Keller MNA 5-27-10vol2.indd 110 5/27/10 2:41 PM Appendix 10: Return Migration and Entrepreneurship Analysis indicate that we cannot reject the exogeneity differ as well, depending on the anticipated assumption of migration duration in the Probit activity after return. of becoming entrepreneur. Following Dustmann and Kirchkamp (2002), This robustness check then confirms the the hypotheses to be investigated further are choice made in the previous section of consider- thus whether migration duration and after- ing migration duration as an exogenous regressor migration activity are linked and, if that is the for Algerian and Moroccan returnees. case, whether optimal migration duration differs across the range of activities the migrant may For Tunisian returnees, an additional test choose after returning. was then necessary. We performed a regres- sion after dropping the migration duration in Hypothesis and econometric modeling the set of variables determining the probability of becoming an entrepreneur. Our goal was to We estimated an econometric model of activity check whether they were significant differences choice and optimal migration using a two-step between these estimates (not reported) and procedure. In the first step, we specified a model those presented in Table A44. Few important of activity choice as a multinomial choice logit. differences appeared in terms of the magnitude To this end, we defined three different regimes of the estimated coefficients. Among the few after return: inactive, waged employment, and changes in coefficient significance, however, we self-employment (the latter category including noticed the following: being bi-national become employers and self-employed workers). In the positively significant (at the 10 percent level) for second step, we ran models of migration duration Tunisian returnees as for Algerians; and being for each of the three regimes while taking into an entrepreneur before migration and having no account the results of step one, i.e., the prob- plan to remigrate now affect the probability of ability of the selective decision to engage into a being an entrepreneur upon return. specific activity after return. The remainder of the results commented The main idea here is to allow the explana- on in the previous section were qualitatively tory variables to affect migration duration differ- unchanged. ently in these three regimes. According to the model used by Dustmann and Kirchkamp (2002) Simultaneity of migration duration and indeed, the way optimal migration duration is activity choice after return related to regressors may differ across regimes. Estimating a unique duration equation across We further investigated what determines the regimes would thus be inappropriate, as it would returnees’ optimal migration duration and impose invalid across-equation restrictions. whether and how this decision interacts with fu- ture activity choice. In the economic literature, A related argument in favor of this approach theoretical models that have investigated the is that optimal migration duration might be a determinants of return migration and optimal function of activity choice after return, and that migration duration (Dustmann 1997, Dustmann this should be taken into account in the econo- 2003, Stark et al. 1997) generally assume that metric analysis. As mentioned above, this is be- there is only one activity the migrant takes up cause we are unlikely to observe all the variables after return. However, if there is a range of ac- influencing migrants’ choices. In particular, it is tivities the migrant may choose once back, and likely that some migrants’ unobservable charac- if migration duration and after-migration activity teristics explain both activity choice and optimal are jointly chosen, then the optimal migration migration duration. Accordingly, conditional on duration may differ across activities. Moreover, observables characteristics, individuals in each the way economic and demographic variables regime may be nonrandomly selected from the are related to optimal migration duration may population of returning migrants. 111 Keller MNA 5-27-10vol2.indd 111 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Our estimation strategy took this into ac- sian returnees in order to increase the sample count by estimating a model of migration dura- size and efficiency of the estimations. tion for each of the three regimes that includes a selectivity-correction term stemming from the Results multinomial logit of activity choice after return. A variety of econometric procedures can be used The results of the different tested models are to this end (see Dustmann and Kirchkamp 2002, not presented for reason of space. Besides, we Bourguignon et al. 2007). We used that developed encountered an important methodological limit by Dahl (2002), which has the advantage to be a that was revealed in the course of the tests. This non-parametric method, therefore less demand- difficulty stems from the nature of the data used ing in terms of assumptions on the error terms in this paper. Recall that the samples of returnees of the equations of interest. The idea is to use refer specifically to migrants who returned to the results of the polychotomous choice model their country of origin in the course of the last to compute, for each observation, a set of choice ten years. This choice was made to allow the probabilities, and then to correct the migration respondents to recount their migratory experi- duration equations of endogenous selection by ences more precisely. In addition, this enables adding a polynomial of these probabilities in the an assessment of the impact of the migration list of explanatory variables. experience on the interviewee’s pattern of re- integration. Only variables that are determined before the migrant’s emigration qualify as regressors. The downside of this, however, is that, by Variables that are determined during or after construction, the samples tend to overestimate the migration period may be affected by activity migration duration for migrants who left their choice or/and duration, and are as such poten- home country a long time ago, for instance in the tially endogenous in both the regime choice and 1970s. Similarly, those migrants who left during duration equations. We included a parsimonious this period and returned back home in the late list of socio-demographic variables such as sex, 1980s or early 1990s are not represented. To the age at which the individual migrated, educa- estimate migration duration models with this tion dummies, two dummies reflecting family dataset, it is thus important to control for the and matrimonial status (number of children and migrants’ date of emigration to avoid the results whether the migrant was married before emigra- to be artificially driven by the structure of the tion), and dummies characterizing the type of samples. visa at entry (no visa, work visa, family visa, and tourist visa). Without controls for the date of emigration, the selection correction terms were significant We included the same variables in the activity in the migration duration equations estimated choice and duration models. To obtain a nonpara- on the sub-sample of migrants who were inactive metric identification, we needed an exclusion after return (regime 1) and on the sub-sample restriction on the duration equation. To be a valid of migrants who were self-employed after return instrument, the excluded variable should affect (regime 3). This preliminary result brought activity choice after return and optimal migration support to the idea that migration duration and duration only via activity choice. To this end, we activity choice after return were simultaneously used a dummy indicating whether the migrant determined. Estimates of the other coefficients has been self-employed before emigration assum- were also interesting and often significant. Being ing that previous self-employment experience female was negatively associated with optimal should reduce the fixed costs of becoming an migration duration, especially for those migrants entrepreneur after return. who chose to be inactive or self-employed after return. The effect of the entry age variable was Note that the models were estimated on a strongly negative and did not differ between the pooled sample of Algerian, Moroccan and Tuni- three regimes. Therefore, an increase in age at 112 Keller MNA 5-27-10vol2.indd 112 5/27/10 2:41 PM Appendix 10: Return Migration and Entrepreneurship Analysis migration was found to decrease optimal migra- 1995–2005, with the reference being before tion duration in all three regimes. Last, education 1974). As expected, introducing these pe- was negatively and significantly associated with riod dummies strongly affected our estimates. migration duration, especially for self-employed First, the selection-correction terms became returnees. This might be because the level of insignificant in all three regimes. Second, most schooling captures higher relative wages of mi- significant effects found previously on the socio- grants in the host country demographic variables disappeared. In fact, most of the variations in migration durations across (Dustmann and Kirchkamp 2002): if return individuals were absorbed by the period effects, to schooling is higher in the host country, indi- the latter being strongly significant in all the viduals with higher levels of schooling have a migration duration models. higher relative wage abroad and may need less time to accumulate savings in order to open a However, a few interesting effects persisted, business after return. This result is then com- which we might then consider as relatively ro- patible with the conjecture that higher wages in bust: being female was still negatively associated the host country decrease the optimal migration with migration duration for those who became duration. inactive after return. Similarly, age at migration appeared as a persistent negative predictor of We then added in the models a set of three migration duration. Finally, high levels of school- period dummies referring to the migrants’ first ing were detrimental to migration duration for date of emigration (1974–1985, 1985–1995, waged returnees. 113 Keller MNA 5-27-10vol2.indd 113 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Keller MNA 5-27-10vol2.indd 114 5/27/10 2:41 PM Appendix 11: Review of Institutional And Legal Framework for Migration in Spain and the Netherlands (From Conde-Ruiz, et al. 2008, and de Neubourg, et al. 2008) SPAIN porated to the Spanish Constitution until the year 1985. Despite its name, this law was not The political setting, external events including concerned with the rights and liberties of for- the EU accession process, and economic develop- eigners living and working in Spain, although ments have combined to shape migration policy it regulated and, in some cases, curtailed some in Spain over time. The key phases of this process of their basic rights like freedom of association, since the 1970s are briefly described below. access to education, and trade-union affiliation in comparison with nationals. The main purpose Franco’s dictatorship (to 1975). Immigrants of this law was to give the authorities more had no fundamental rights during this period. The power to control the entrance of foreigners and authorities acted at their own discretion when facilitate their expulsion if found in an irregular granting (or refusing) permits for immigrants. situation in the country. More specifically, the Similarly, residence permits for foreigner were LOE established as a prerequisite for all immi- given on a completely discretionary basis by the grants to have obtained a valid visa in their home relevant authorities. country. This visa was only issued on condition of having received a job offer in Spain. The LOE The first years of democracy (1978–1985). also established a regime of sanctions for all im- When the Spanish Constitution was approved migrants found in an irregular situation in the in the year 1978, the emigration of nationals to country. Given that at this time there were little Europe was a more pressing and relevant issue international business relationships and that for- than immigration to Spain, which at this time was eign labor recruitment was almost non-existent, hardly significant. Thus, in specific legislation any would-be immigrant had almost no chance to (Article 42), the State was called to look after enter the country legally; the only way was either the rights of Spanish immigrant workers abroad to apply for a tourist visa or cross the border il- and facilitate the eventual return to their home legally. In sum, the LOE (1985) was, on paper, country, while there was no mention whatsoever a strict legal framework for immigrants, clearly of immigrants living in Spain. distinguishing between legal and irregular migra- tion. But its main purpose was, in fact, to create The National Law for Foreigners (LOE) the administrative mechanisms for the speedy (1985) (“Ley Organica de Extranjeria 1985â€?). expulsion of any irregular immigrant found in A comprehensive law to regulate the granting the country without providing any alternative of work permits for immigrants was not incor- channel for legal and orderly migration. 115 Keller MNA 5-27-10vol2.indd 115 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options The Regularization of 1991 and the new passed a new set of regulations for immigrants Immigration Regulation of 1996 (Reglamento in the Official Decree 155/1996. The regulation 1996). Despite the severe restrictions imposed by of 1996 replaced all previous legislation and the LOE 85, the number of immigrants entering represented a key step towards the creation of irregularly in the country continued to increase a more stable model of immigration. Among the significantly from the late 1980s onwards. They new measures included were: i) the introduction were mainly attracted by the economic develop- of scale of offenses so that immigrants could not ment of the country and its admission into the be expulsed from the country for minor offences, European Community in 1986. For the period and ii) the granting of a permanent work and 1985–1990, the average growth rate of GDP was residence permit for all immigrants after more above 4 percent and the unemployment rate went than 5 years of residence in the country. This down from 21.5 percent in 1985 to 15.9 percent last measure eliminated the problem of annual in 1991, all of them unequivocal signs that the permit renovation, granted important guarantees Spanish economy was starting to take off, car- for the children of immigrants, and created a rying with it more demand for labor. For most more secure legal framework for all immigrants. immigrants, the easiest way to enter Spain was However, many of the regulations more beneficial either to enter directly if from a country with no for the immigrants were, in fact, inconsistent with visa requirement (several Latin American coun- the previous LOE 85. tries) or to obtain a three-month tourist visa; once in the country, they would look for employ- The Organic Law 4/2000 of Rights and ment in the informal economy with expectations Liberties of foreigners and their Social Inte- for an eventual access to a legal situation in the gration (LODYLE) and its counter-reform in near future. In the case of migrants coming from the LOE 8/2000. The reality of immigration made countries exempted from the tourist visa system, the reform of the old LOE 85 inevitable. Due to there was little that stricter border control could political rivalries between the two main parties, do to prevent them from entering Spain. The this reform was carried out in two different pieces situation was chaotic: while the official figures for of legislation, but in practical terms both laws are legal migration remained very low, there were no relatively similar and are discussed together here. reliable estimates about the increasing number of The key changes that the LOE 4/2000 introduced irregular immigrants already living and working were: i) the creation of a coherent and clear set of in the country. rights for immigrants, establishing the equality of rights between nationals and foreigners and pe- For this reason, the Parliament recom- nalizing discrimination against immigrants; ii) the mended the regularization of all unauthorized incorporation of the right to family reunification immigrants by an extraordinary process that in accordance with directives of the European took place in 1991 and resulted in the issuing Human Rights Tribunal, iii) the subjective rights of a one-year work and residence permit for of immigrants in an irregular situation, including 112,000 migrants. In addition, an annual policy the right to medical assistance, access to educa- of “indirectâ€? regularization was introduced by tion, legal counsel or an interpreter when dealing a quota system (contingente anual) for all with the authorities. those irregular immigrants already working in Spain. An average of 30,000 immigrants per year The Period of Regularizations (2000– obtained their working and residence permits 2005). Still without a comprehensive legal from 1993 to 1999 through this quota system. framework for labor immigration, Spain wit- Yet residence and work permits by no means nessed a dramatic increase in the number of kept up with the actual number of immigrants immigrants arriving to the country in the first coming into the country. years of the new millennium. The economy was booming: GDP growth averaged 3.5 per- With the increasing immigration tide render- cent, the unemployment rate fell from 14.9 in ing the previous LOE 85 obsolete, the government year 2000 to 8.5 percent in the year 2006, and 116 Keller MNA 5-27-10vol2.indd 116 5/27/10 2:41 PM Appendix 11: Review of Institutional and Legal Framework for Migration in Spain and the Netherlands the employment rate for the active population regulations from the previous laws of 2000. With (16–64 years old) increased by 8 percentage the implementation of this new, more liberal points to reach 65 percent. It is possible that, set of regulations, the government drastically unintentionally, the legislation passed in the reduced the number of irregular immigrants in year 2000 encouraged the dramatic increase of a short period of time and introduced a number immigration to Spain. As in the case of previous of mechanisms aimed at regulating immigration laws and regulations (LOE 85 and the Regula- policy in the future. tion of 1996), the passing of the LODYLEA 4/2000 and the LOE 8/2000 were accompanied The key element of this new regulation was by extraordinary regularization processes. The to initiate another extraordinary process of first law allowed 170,000 immigrants to regulate regularization (this time called “normalizationâ€?) their legal situation in the country. The second which affected around 600,000 foreign workers. law, intended to cover the remaining immigrants The main requirements to be eligible for this new still in an irregular situation, resulted in more process of regularization were: i) a minimum than 300,000 people obtaining their permits to of six-months stay in Spain prior to the public work and live in Spain. After these two processes announcement of the regularization process, in- of regularization, the problem of immigration tended to prevent a flooding of new immigrants, would be controlled with the strict application and ii) a working contract with a minimum six- of the contingente anual. From the year 1993 months duration to be presented to the relevant to the year 1999, as mentioned above, this authorities by the employer. kind of labor quota system was used to issue a limited number of work and residence permits In addition, the new regulation has intro- for immigrants already in the country but in an duced important modifications to the policy irregular situation. This practice was now made framework. In order to prevent any more ex- impossible. From now on, all foreigners had first traordinary regularizations as in the past, two to be registered in the quota of their respective legal concepts for regularization have been home countries before being allowed to enter strengthened: i) the “working integrationâ€? (ar- Spain. Any would-be immigrant had to make raigo laboral) that requires any foreigner to an application in its home country for a visa have stayed in Spain for two years and have and a personal work permit for a specific job. worked either legally or not for at least one year, For that application to be accepted, an official and ii) the “social integrationâ€? (arriago social) certificate was required stating that there were that requires any foreigner to have been resident neither national, EU, or foreign residents in the for three years, have either a working contract job-seekers lists of any Autonomous Regions or a job offer for a period of more than one year, currently applying for that specific job. This and have direct family links with other foreign labor quota system, however, did not succeed as residents already living in Spain. With these two an instrument to control and regulate the arrival criteria, the process of regularization will now be of foreign workers to Spain. The reason is that done on strictly individual basis. the labor quota (fixed by the Government) was too low and represented less than 7 percent of Finally, the failed contingente anual system total inflow. is now (2008) to be replaced by a more pragmat- ic, flexible and efficient one. There is now only The new regulations for immigration, a provisional “database of difficult-to-fill vacant the “last regularizationâ€? and the new im- positionsâ€? (catalogo de dificil cobertura), which migration policy. In the year 2004, according is renewed quarterly. As long as the applicant’s to estimates in this report, the number of ir- requested job is in this provisional database, regular immigrants in Spain was above 1 million there will be no requirements for further au- people. Instead of pressing for a whole new thorization for the immigrant to come to Spain piece of legislation, the recently-elected social- apart from being called by his/her prospective ist Government decided to amend the existing employer. 117 Keller MNA 5-27-10vol2.indd 117 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options The Netherlands The second driving force behind the chang- es in migration policies is the sequence of strong Disregarding strong peaks in immigration during and explicit views on the place of migrants in the and just after the World War I, the economic cri- Dutch society. These views have been changing ses in the 1930’s and the period just after WWII, and shifting from the desirability of intentionally the Netherlands has shown a net emigration since temporary labor-demand-driven migration in 1865. Starting in the early 1950’s, immigration the 1960s and 1970s, over-stimulating perma- shows a positive trend outnumbering emigra- nent migration of workers and their families in tion in absolute numbers from the early sixties the 1980s and 1990s, to encouraging circular onward, turning the country into an immigration migration of high-skilled workers in the last country until recently. years. The views on whether and how the mi- grant workers should and can integrate into the The history of immigration after WWII in Dutch society are relatively independent from the Netherlands has to be written in several the economic needs in terms of the demand for periods: labor. They are on the one hand influenced by the dominant ideology on the desirability of the • 1945–1960: immigration predominantly from integration of migrants and its implementation. former colonies On the other hand, the ideology itself is partly • 1960–1973: labor demand driven immigration defined by reactions to the very presence of from Southern Europe and North Africa the migrant population, its composition and the • 1974–1997: immigration of family members socio-economic behavior of its members. The of migrant workers and asylum seekers as resulting reactive character of Dutch migration quantitatively the most important groups policies is important. Policies tended to recog- • 1997–2007: curbing immigration on family re- nize rather than organize migration flows and unification grounds and asylum, lower immi- tried to adapt to the newly identified realities. gration of low skilled workers and increasing This was especially the case in the 20th century. recruitment of higher skilled migrant workers Over the last decade, policies tended to be more pro-active and aimed at curbing certain types The exact phasing of the post-WWII era into of migration, although a significant reactive ele- these 4 periods is crude and more details will be ment is maintained as well. added in order to subdivide the latter periods further, thus reflecting changes in migration and The interaction between the driving forces integration policy. Before exploring the relation- and the reactive character of the policy changes ship between migration flows and policy changes leads to an additional peculiarity of the Dutch in more detail, however, it is important to identify migration policy since the Second World War: the two important forces that drive migration the functional intertwining of migration policies policy in the Netherlands. and integration policy. The dominant view on the type of integration into the Dutch economy and The first driving force is the labor supply society that migrants are expected to show had needs of the economy, or more specifically, the an important impact on the design of subsequent interpretation of supply needs by the policy mak- migration policies and the choice of the instru- ers. Economic growth and recessions and the ments used to reach the objectives. Needless to related changes in the demand for labor have had say that the objectives themselves were subject an influence on the design of immigration policies to shifts and changes following the sequence in over the last six decades in the country. Dutch the views on the functions and roles of migrants migration policy for workers has been largely in the economy and the society. adaptive to the economic situation leading to a remarkable synchronization of the changes in The next paragraphs elaborate on the driving economic growth and labor immigration. forces and their chronological logic. 118 Keller MNA 5-27-10vol2.indd 118 5/27/10 2:41 PM Appendix 11: Review of Institutional and Legal Framework for Migration in Spain and the Netherlands The responsible ex-colonial power a relatively minor quantitative element in the (1945–1980) Dutch migration debate in the late 20th century and beyond. That debate was colored much The decolonization process starting in 1945 with more by the migration flows in the 1960s, 1970s Indonesia and ending in the 1980’s with Surinam, and 1980s. The latter migration flows inspired boosted immigration in the Netherlands in sev- migration policy shifts to a much larger extent. eral waves. In the former case, approximately 350.000 people used the right to “repatriate.â€? To- The reluctant host (1959–1974) gether with residents of Indonesia, many Moluc- can soldiers, having served in the Dutch army in Being late in recovering from the Second World Indonesia, immigrated with their families in that War (compared to Belgium, France and Ger- period (1945–1950) as well. The anticipation of many), the Netherlands started to feel the excess the independence of Surinam in the 1970s made demand for labor only at the end of the 1950’s many residents of that country, having doubts on when the neighboring countries had already the economic and political stability of the colony started to use immigrant labor to fill the gap in the becoming independent, move to the European labor market. As a result, the first migrant work- part of the Kingdom. When the Dutch govern- ers entered the Netherlands via the neighboring ment gave the Surinamese the right to opt for countries. In the absence of an organized public Dutch citizenship in 1980, eventually one third of policy, it has been the employers who organized the Surinamese immigrated in the Netherlands. the recruitment of migrant workers mainly from Clearly colonial links, political instability, and the Mediterranean countries. The government economic uncertainty acted as pushing factors responded by signing so-called “recruitment for the migration flows while Dutch citizenship agreementsâ€? with the sending countries; first and the legal and economic conditions in the with Italy in 1960 and later with most of the other receiving country worked as pulling factors for Mediterranean countries (Spain, 1961; Portugal, the decolonization-related migration wave. 1963; Turkey, 1964; Greece, 1966; Morocco, 1969; Yugoslavia and Tunisia, 1970). The recruitment During the early decolonization period, im- agreements formalized the practice that workers migration was quantitatively important though could enter the Netherlands officially after hav- not massive compared to the population size ing obtained a work permit for the Netherlands of the receiving country. The policies related in their country of origin. However, many work- to these groups of migrants, however, were not ers first came in as tourists and obtained the merely seen as migration policy but rather as necessary documents only upon finding a job. the consequences of the post-colonial era. This Moreover, the prevailing rules themselves were changed in the 1980s: the Netherlands adopted relatively easy for immigrants to fulfill, even after a more restrictive immigration policy in general entering the country in an inappropriate man- (see below) and introduced visa restrictions on ner. Competition for workers between the fast many countries, including Surinam. Rather than growing economies in Europe was severe, and reducing immigration from this ex-colony, the employers, labor unions, and the government announcement of the introduction of a manda- seemingly accepted that economic growth could tory visa resulted in a boost in immigration from only be sustained if foreign labor supply was Surinam. The fear that easy migration possibilities admitted in the country in a fairly smooth way. would not be available in the future provided a strong incentive for relatively large numbers of Su- This changed after the first (relatively minor) rinamese residents to use the chance to immigrate recession of 1966–1967. Under the pressure of prior to the introduction of the new visa policy. the labor unions, the granting of work permits to workers who did not use the official procedure Despite the fact that immigration from was restricted (officially stopped in 1968), as ex-colonies was not insignificant, it was only was the renewal of work permits. However, the 119 Keller MNA 5-27-10vol2.indd 119 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options demand for foreign workers remained largely and came without their families; in the early unchanged during that recession. Many laid-off 1970s there were 55.000 Turkish and Moroccan foreign workers easily found jobs in other com- guest workers and only 20.000 family members. panies and industries and the entry of migrant workers reached new peaks in 1970 and 1971 The slowdown in economic growth in the after the economy picked up again. 1970s and the economic downturn in the early 1980s did change the demand for (foreign) The aftermath of the first oil crisis in 1973, labor. It, however, did not change the migrant and especially the economic downturn after population in a proportional way. Many Euro- 1979, had much more identifiable effects on im- pean foreign workers returned to their country migration in the Netherlands; less so on the ab- of origin, but many workers, especially those solute numbers, but more on the composition of stemming from Turkey and Morocco (and also the migrant population. It had, however, a major the ex-Surinamese as they became Dutch citi- impact on the public (and political) view on im- zens) did not. The government recognized that a migration and its function and consequences. In significant portion of the workers did not return order to understand these changes it is important to their home country during a recession but to specify the prevailing positions on immigration expected them eventually to do so neverthe- and migration policies in the 1960’s and the first less. In the 1974 “Memorandum of Reply,â€? it half of the 1970’s. was explicitly stated that the Netherlands had responsibilities to the guest workers and that “The Netherlands is not an immigration therefore, a policy to accommodate them in the country and should not become one.â€? was the Dutch society as long as they would remain in official position of the Dutch government in the country was imperative. This accommodat- 1974. The first official document that formulated ing policy had two main elements: giving guest a coherent policy view on immigration dates workers improved access to public services/social from 1970 (Memorandum Foreign Employees— security and providing cultural support. Since, “Nota Buitenlandse Werknemersâ€?) and already however, immigration remained to be considered had specified that the Netherlands was not an as temporary, integration with preservation of the immigration country. It argued that the Dutch workers’ own (sending-country-related) cultural economy needed foreign labor to sustain eco- identity became the main policy stance. The nomic growth but formalized the largely shared Netherlands became the “reluctant hostâ€? for the unofficial view that migrant workers were ap- guest workers, needing them to fill the vacancies, preciated but temporary transients in the Dutch and accommodating them in a way that allowed society, the attitude being “There is no need to them to live decently in the country, but provid- integrate them since they will leave when they ing facilities that were believed to stimulate and are not needed anymore.â€? Recognizing, however, facilitate their expected return. that there might be a structural need for foreign workers, the Memorandum of 1970 introduced Cutting down the number of the idea of contracts of very limited duration immigrants but nevertheless (two years) guaranteeing that individual work- increasing immigration (1974–1983) ers would not stay in the country for protracted periods but allowing nevertheless the economy At the same time of the recognition of the respon- to use immigration to deal with shortages in sibilities that the Netherlands had towards the labor supply. The concept of circular migration immigrated workers, the country started a new was not yet very popular, but the official policy policy cutting down the recruiting of foreign work- note proposed to and aimed at “rotatingâ€? foreign ers and stimulating the return of those who were workers between their sending country and the in the country. By the mid-1970s it became the Netherlands. It should be noted that this was aim of the government policy to restrict labor im- consistent with the reality on the ground: the migration to a strict minimum, tightening controls majority of the workers in that period were male on entry and using quotas on the sending country 120 Keller MNA 5-27-10vol2.indd 120 5/27/10 2:41 PM Appendix 11: Review of Institutional and Legal Framework for Migration in Spain and the Netherlands and on the company level (still using working by the beginning of the 1980s the policy view on permits as the main instrument). By adopting migration again showed a major shift (see next these policies, the government hoped to reduce section). Return migration never led to large the pulling factors of immigration from less devel- changes in the numbers. oped countries. Moreover, the Dutch government also wanted to reduce the effect of the pushing Despite the policy efforts, the objective of factors on migration from the poorer countries cutting down the number of new immigrants was and adopted a broad internationally oriented ap- never reached, not due to an increasing number proach aiming at “improving the prospects of of labor immigrants, but because of an increasing potential migrants in their sending countries.â€? number of people immigrating in the country for As part of that policy, the government wanted to reason related to family reunification and family restructure the Dutch economy in order to reduce formation. The total number of Non-OECD im- its need for low skilled labor, while making an ef- migrants more than tripled between 1975 and fort internationally to lift trade barriers for less 1985 from less than 200,000 to approximately developed countries and by stimulating selective 600,000. The increase was, to a large extent, due industrialization in the latter economies. These to migrating partners and children joining labor policies were seen as the “moral compensationâ€? migrants from the 1960s and the first half of the for the restrictive entry policies. 1970s. As already indicated above, the case of Surinam is special and linked to the process of Different than in France and Germany where independency, the option to obtain Dutch citizen- formal recruitment stops for foreign workers ship, and the anticipation of the introduction of were introduced, Dutch policy makers never visa requirements in 1980. decided to stop recruitment of foreign workers completely. That does not mean that there were The increase in the immigrated population no changes. Starting in 1973, and formalized by has confronted the Netherlands with an entirely the Foreign Employee Act (“Wet Buitenlandse new reality: a large part of the labor migrants Werknemersâ€?) in 1979, it is stipulated that no for- from the previous decades and the migrants eign worker will be allowed to settle permanently from ex-colonies, were joined by their families into the country “unless there are compelling and there to stay more than ever. reason to allow it.â€? Despite the fact that admis- sion continued to be regarded as a temporary “The Netherlands is a ‘de facto’ matter and that employers had to prove that there immigration countryâ€? (1983–1994) were no suitable employees to be found within the European Economic Area (EEA), labor im- By the end of the 1970s and the beginning of migration was still allowed under the 1979 act (to the 1980s it became clear that many immigrants be replaced in 1995 by a new act) when it was would stay in the country. A couple of violent considered beneficial to the Dutch economy. In incidents related to a small and specific group addition to work permits as an instrument, the of immigrants drew public attention to the introduction of visa requirements for countries precarious position of immigrants, questioning with a high emigration potential (Turkey, Morocco in a brutal way the policy that assumed their and Surinam) were introduced; visas, however, return to the sending country. The wake-up call were (and are) closely linked to work permits led to the political debate that resulted in the and had little autonomous effect except in the understanding and acceptance on the part of case of Surinam (see above section on colonies). both the public and the policy makers that large groups of immigrants in the country should be Starting in 1974, the government also tried acknowledged as permanent residents, and that to stimulate the return of migrant workers who therefore, a policy that reconsidered the immi- settled in the country during the economic boom grants’ position was a necessity. The Netherlands in the 1960s and early 1970s. The programs were Scientific Council for Government Policy (WRR) dropped because of the costs and the fact that published in 1979 a highly influential report that 121 Keller MNA 5-27-10vol2.indd 121 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options was endorsed by the political leadership in 1983 as well (due to pillarization or “verzuilingâ€?). In in the new Memorandum Minorities Policy (“Nota return, the immigrants were expected to acquire Minderhedenbeleidâ€?). It was during the debate the basic (language) skill that would allow them about this Memorandum that the Prime Minis- to function well in the Dutch society. ter, Mr. R. Lubbers, stated: “The Netherlands has become a ‘de facto’ immigration country.â€? Since, at that time, the old immigrants were Although this policy change was mainly related expected to stay in the Netherlands, immigration to the integration of present migrants into the continued despite the depressed demand for Dutch society, it had serious consequences for (foreign) labor due to a severe economic crisis, migration flows and migration policies. and notwithstanding a labor immigration policy that restricted inflow of new migrants based on The new integration policy targeted all the the old idea that the country was overpopulated immigrant minority groups originating in the old and showed “signs of congestion.â€? Immigration colonies and from the main non-European emi- due to family unification and formation had gration countries in the 1960s and 1970s with an already became the most important reason for extensive catalogue of instruments and policies. immigration in the late 1970s; it peaked in 1983– Besides their recognition as permanent residents 1984 and stayed important afterwards. This also in the country, the policy had four main elements: was a consequence of the new integration policy that aimed at integrating migrants into Dutch • First, to strengthen the cultural identity and society; immigration for family reunification and the social organization of the minority groups formation reasons was considered functional through support for ethnic organizations, to the integration of the earlier labor migrant. culture, radio, television broadcasting and Families were considered to integrate more easily language training; in the local ethnic community than single men, • Second, to improve their legal status and to and these ethnic communities would become an fight discrimination; integral part of the Dutch society. • Third, to provide active and passive political rights on the local level, and; Keeping in mind that the economy was de- • Fourth, to provide easier access to Dutch citi- pressed for large periods of the 1980s, thereby zenship for second-generation immigrants. leading to high unemployment rates for both Dutch and migrant workers alike, it is not During the 1980s numerous projects on the surprising that the new integration policy was national as well as on the local level were started selectively successful in terms of social and to provide better educational opportunities, im- political participation of the migrants, but failed proved chances on the labor market and access in terms of employment. The strong emphasis to social housing (previously impossible). It is on ethnic cultural values and language was also remarkable that the emphasis was on strength- increasingly held responsible for another failure: ening the ethnic cultural and social identity the new policy did not produce the expected rather than on the full integration into the Dutch results in terms of educational achievements of society. In contrast with the earlier policy during the new and old migrants. Out of the criticism the seventies, preservation of their own cultural on the Memorandum Minorities Policies a new identity, however, was no longer aimed at facili- policy wave was about the born. tating return of immigrants to their country of origin, but at providing a strong identification The doors are open and difficult to with their ethnic group based on common values close (1994–2001) and principals. It was expected that this positive identification would contribute to a better inte- The liberal family migration policy as part of gration of the ethnic groups (and by implication the new integration effort continued to attract also of the individuals) into Dutch society, which between 20,000 and 30,000 new migrants yearly, was considered to consist of separate groups until very recently. At the end of the 1980s, and 122 Keller MNA 5-27-10vol2.indd 122 5/27/10 2:41 PM Appendix 11: Review of Institutional and Legal Framework for Migration in Spain and the Netherlands especially the beginning of the 1990s, an even big- and reconsidered social assistance in cash in ger flood of asylum seekers joined these migrants. 2001, the number of asylum application dropped While throughout the 1970s and the 1980s the significantly. In the case of the Netherlands the Netherlands was chosen by a small number of tightening of the access to asylum was real- asylum seekers every year (a couple of thousand), ized through an administrative action linking their number increased to 14,000 in 1988 and all social economic databases, electronically further exploded to yearly peaks of 53,000 and allowing the authorities to identify and allocate 45,000 respectively in 1995 and 1999–2001. Not all illegal immigrants whenever they would use of them were granted asylum, but migration due the administration, health services, schools, to asylum increased to levels comparable to family or parts of the social security administration. migration between 1995 and 2002. In these years The so-called “Linkage Actâ€? made it easier for between 15,000 and 20,000 new asylum seekers the administration to identify the illegal asylum immigrated in the Netherlands every year; these cases. Moreover, it discouraged illegal asylum- migrants added to the approximately 22,000 fam- seeking by making access to the social system of ily migrants. the country more difficult. The combination of more legal restrictions, a faster procedure, and Combined with the mounting criticism of the the linkage seem to have led to a serious drop in policies as depicted in the previous paragraphs the number of asylum seekers. However, since and the extremely high concentration of all the that coincided with the end of a series of violent migrants in the big cities of the country, the view conflicts abroad, it is hard to be sure whether on migration policy was changed again. And the decrease in applications was solely due to again the migration policy was directly linked to the policy change. the views on the integration of migrants into the Dutch society and not to a view on the needs of Curbing family migration proved to be more the Dutch economy or the Dutch labor market. difficult. In the 1990s and the beginning of the Immigration (of Non-OECD immigrants) for labor 21st century, the Netherlands changed its policy motives stayed at a modest couple of thousand by abandoning some elements of the previous of people per year. policy, especially those elements that subsidized and stimulated ethnic cultural integration within The absolute number of non-OECD migrants the sub-community of the migrants. Instead, (excluding all the immigrants from Europe, North the new Newcomers Integration Act (“Wet In- America and Oceania) increased to one and a half burgering Nieuwkomersâ€?) of 1998 emphasized million in 2003, up from less than 200,000 in the integration of the new migrants (especially 1975. Without exaggeration it can be said that those coming for family re-unification or forma- migration pressure (in terms of integration of a tion) into the Dutch society. This emphasis was large group of culturally different immigrants) in created by adding conditions to the residence the Netherlands has been increasing seriously. permit; new migrants were required to take a 600-hour language and civic orientation course, In the 1990s integration policy elements combined with training in finding their way on were more and more introduced into the migra- the labor market and in the educational opportu- tion policy instruments. As can be understood nities of the country. Combined with devolution from the previous account, the main policy in financing and enforcement responsibilities challenge has not been to curb labor migration, at the municipal level and heavy support for nor to reduce irregular and illegal migration (see the four big cities that host a large part of the below), but to bring family and asylum migration migrant population, it was hoped that the new under control. Asylum migration is of course Act would speed up the integration of the new linked to the existence of violent conflicts and migrant population into the Dutch society. The disasters that initiated the migration linked to additional course requirement, however, did not asylum. When the Netherlands and other coun- discourage many new migrants from applying for tries tightened the criteria for a refugee status a residence permit. 123 Keller MNA 5-27-10vol2.indd 123 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Employers were pushed to provide more early integration for those that pass the test and employment opportunities to (the resident) eth- migrate to the Netherlands. Although in 2007 still nic minorities largely by soft measures such as more than half of the total applications for new providing an annual report on the issue (started temporary resident permits were due to family 1994); it had no big impact on the labor market unification and formation reasons, the effect of opportunities of the migrant population. In 1998 the new integration test is measurable. After the government, employers and labor unions its introduction in March 2006, the number of agreed on a more comprehensive approach. applications for temporary resident permits Since that time, the employment situation of (the necessary first step in gaining access to the the members of the ethnic minorities improved, country) from the targeted population dropped but because that improvement coincided with an from approximately 2,000 per month to less economic upturn, again it is impossible to assess than 1,000 per month. In 2007 labor migration the effect of this policy. accounted for approximately 20 percent of the total applications (including 14 percent for high- Changing the guard (2003–2007) skilled workers). Additional integration tests are also imposed on the migrants already resident in A string of events—including the murder of a the country. politician and a moviemaker in the Netherlands plus the change in the worldview on ethnic The Netherlands recently started to redi- minorities and Muslim-related terrorism after rect its policy aim towards labor migration. It 9/11—further hardened the public opinion on is expected that the country will need the im- immigration issues. The main change in immigra- migration of high-skilled migrants and less so of tion policy has been the introduction of a civic low-skilled migrants except for a small number integration test abroad (“Inburgeringsexamen of professions. Buitenlandâ€?).44 Since 2006, this test must be taken in the country of origin before applying for a residence permit to come to the Netherlands 44 The civic integration test includes a basic language test, and for reasons related to family unification or fam- a test on knowledge about the history of and daily life in the ily formation. It is expected that the test would Netherlands. It is aimed at facilitating integration into the Dutch discourage family immigration and stimulate society once the immigrant arrives. 124 Keller MNA 5-27-10vol2.indd 124 5/27/10 2:41 PM Appendix 12: Computable General Equilibrium Analysis of Impact of Increasing MENA to Europe Migration (From Docquier and Marchiori 2009) Methodology used in DLM for which systematic statistics by education level and country of birth are not available,45 except in Emigration stocks. It is well documented that the United States. Demographic evidence indi- statistics provided by source countries do not cates most U.S. illegal residents are captured in provide a realistic picture of emigration. When the census. However, there is no accurate data available, which is very rare, they are incomplete about the educational structure of these illegal and imprecise. While detailed immigration data migrants. Hence, the number of unskilled in are not easy to collect on an homogeneous basis, the immigrant population is probably underes- information on emigration can only be captured timated, assuming that most illegal immigrants by aggregating consistent immigration data col- are uneducated. Nevertheless, this limitation lected in receiving countries, where information should not significantly distort the estimates of about the birth country, gender, and education the migration rate of highly skilled workers. of natives and immigrants is available from national population censuses and registers (or Here are the main methodological choices samples of them). More specifically, the receiving of DLM: country j’s census usually identifies individuals on the basis of age, gender g, country of birth • The term “source countryâ€? usually designates i, and skill level s. The Docquier, Lowell, and independent states. They distinguish 195 Marfouk (2007) method (herinafater refered to source countries: 191 UN member states, as DLM) consists in collecting (census or reg- Holy See, Taiwan, Hong Kong, Macao, and isters) gender-disaggregated data from a large Palestinian Territories. They aggregate North set of receiving countries, with the highest level and South Korea, West and East Germany, of detail on birth countries and three levels of and the Democratic Republic and the Repub- educational attainment: s=h for high-skilled, lic of Yemen. They consider the same set of s=m for medium-skilled and s=l for low-skilled. source countries in 1990 and 2000, although = ∑ j Mthe Let Mti,g,s denote i, j t , g ,s stock of adults 25+ born in some of them had no legal existence in 1990 j, of gender g, skill s, living in country j at time (before the secession of the Soviet block, t. Aggregating these numbers over destination former Yugoslavia, former Czechoslovakia, countries j gives the stock of emigrants from country i: Mti,g,s = ∑ j Mti,,gj,s . 45 Hatton and Williamson (2002) estimate that illegal immigrants By focusing on census and register data, the residing in OECD countries represent 10 to 15 percent of the methodology badly captures illegal immigration total stock. 125 Keller MNA 5-27-10vol2.indd 125 5/27/10 2:41 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options and the German and Yemen reunifications) fully homogeneous across OECD countries. or became independent after January 1, In most receiving countries, foreign-born 1990 (Eritrea, East-Timor, Namibia, Marshall are individual born abroad with foreign Islands, Micronesia, and Palau). In these citizenship at birth. In a couple of countries, cases, the 1990 estimated stock is obtained “foreign-bornâ€? means “overseas-born,â€? i.e., by multiplying the 1990 value for the pre- an individual simply born abroad. secession state by the 2000 country share in • DLM distinguishes three levels of educa- the stock of immigrants (the share is gender- tion. Medium-skilled migrants are those and skill-specific). with upper-secondary education com- • The set of receiving countries is restricted to pleted. Low-skilled migrants are those OECD nations. Generally speaking, the skill with less than upper-secondary education, level of immigrants in non-OECD countries including those with lower-secondary and is expected to be very low, except in a few primary education or those who did not go countries such as South Africa, the six mem- to school. High-skilled migrants are those ber states of the Gulf Cooperation Council, with post-secondary education.49 This as- and some Eastern Asian countries. To allow sumption is compatible with Barro and Lee’s comparisons between 1990 and 2000, they human capital indicators (based on the consider the same 30 receiving countries in 1976-ISCED classification). Some migrants 1990 and 2000. Consequently, Czechoslova- did not report their education level. As in kia, Hungary, Korea, Poland, and Mexico are Docquier and Marfouk (2006), they classify considered as receiving countries in 1990 these unknowns as low-skilled migrants.50 despite the fact that they were not members Educational categories are built on the ba- of the OECD. sis of country specific information and are • They only consider the adult population compatible with human capital indicators aged 25 and over. This excludes students available for all sending countries. A map- who temporarily emigrate to complete their ping between the countries’ educational education. In addition, as it will appear in the classifications is sometimes required to next section, it allows the comparison of the harmonize the data.51 numbers of migrants with data on educational attainment in source countries. Emigration rates. DLM count as migrants • Migration is defined on the basis of the all adult (25 and over) foreign-born individuals country of birth rather than citizenship. living in an OECD country. A more mea`ningful While citizenship characterizes the foreign measure can then be obtained by comparing population, the “foreign-bornâ€? concept the emigration stocks to the total number of better captures the decision to emigrate.46 Usually, the number of foreign-born is much 46 In some receiving countries such as Germany, immigrants’ higher than the number of foreign citizens children (i.e., the second generation) usually keep their foreign citizenship. (twice as large in countries such as Hungary, 47 By contrast, in other OECD countries with a restricted access the Netherlands, and Sweden).47 Another to nationality (such as Japan, Korea, and Switzerland), the for- reason is that the concept of country of eign population is important (about 20 percent in Switzerland). birth is time invariant (contrary to citizen- 48 The OECD statistics report that 14.4 million foreign-born individuals were naturalized between 1991 and 2000. Countries ship which changes with naturalization) with a particularly high number of acquisitions of citizenship and independent of the changes in policies are the U.S. (5.6 million), Germany (2.2 million), Canada (1.6 regarding naturalization.48 The number of million), and Australia and France (1.1 million each). 49 In the US case, this includes those with one year of college. foreign-born can be obtained for a large 50 Country specific data by occupation reveal that the occupa- majority of OECD countries, although in a tional structure of those with unknown education is very similar limited number of cases the national census to the structure of low-skilled workers (and strongly different from that of high-skilled workers). See Debuisson et al. (2004) only gives immigrants’ citizenship (Germany, on Belgium data. Hungary, Italy, Japan, and Korea). It is worth 51 For example, Australian data mix information about the high- noting that the concept of foreign-born is not est degree and number of years of schooling. 126 Keller MNA 5-27-10vol2.indd 126 5/27/10 2:41 PM Appendix 12: Computable General Equilibrium Analysis of Impact of Increasing MENA to Europe Migration people born in the source country and belonging The GA model with to the same gender and educational category. heterogeneous agents This method allows the evaluation of the pres- sure imposed on the labor market in the source The starting point of the generational account- country. ing technique is the government intertemporal budget constraint. At the base year t, the sum In the spirit of Carrington and Detragiache of the public net wealth and the present value (1998), Adams (2003), Docquier and Marfouk of prospective aggregate net payments by living (2006), or Dumont and Lemaitre (2006), the sec- and future generations must equalize the present ond step consists in calculating the brain drain as value of prospective public purchases: a proportion of the total educated population born in the source country. Although DLM’s analysis is PVLt + PVFt + NWt = PVGt basedi on stocks (rather than flows), they refer to Mt ,proportions mti,g,s = these g ,s as emigration rates. Denoting where PVLt measures the present value of net i +M Rt ,g,s as the i stock t , g ,s of resident individuals aged 25+, tax payments by living generations over the rest of skill s, gender g, living in source country i, at of their life, PVFt is the present value of net tax time t, emigration rates are defined as payments by future generations, PVGt stands Mti, g,s for the present value of prospective government mti, g,s = purchases of goods and services NWt and is the Rti, g,s + Mti,g,s net public wealth. Mti,g,s = ibe used In particular, mti,g,s can as a proxy of the + Mti,g,s i. Rt ,g,s country The net wealth at time t is observed. Two of brain drain in the source the remaining terms are projected using contem- poraneous observations and official projections, This step requires using data on the size PVGt and PVLt. The fourth term, PVFt, can thus and the skill and gender structure of the adult be calculated as the residual burden bequeathed population in the source countries. Population to future generations. data by age are provided by the United Nations.52 DLM focuses on the population aged 25 and The present value of government purchases more. Data are missing for a couple of countries is the discounted sum of public expenditures: but can be estimated using the CIA world Fact- book.53 Population data are split across educa- ∞ Gs tional group using international human capital PVGt = ∑ s =t (1 + i)s−t indicators. Several sources based on attainment and/or enrollment variables can be found in the where Gsis the amount of public purchases pro- literature. As in Docquier and Marfouk (2006), jected at time s ≥ t; i denotes the interest rate. human capital indicators are taken from De In practice, the path of can be partly projected La Fuente and Domenech (2002) for OECD on the basis of budgetary forecasts (i.e., between countries and from Barro and Lee (2001) for t and t*) and partly projected using balanced non-OECD countries. For countries where Barro growth assumptions (between t* and ∞). In and Lee measures are missing, they predict the long run, it is assumed that Gs grows at the the proportion of educated using Cohen-Soto’s same rate as the growth rate of the total factor measures (see Cohen and Soto 2007). In the productivity, g. remaining countries where both Barro-Lee and Cohen-Soto data are missing (about 70 countries The present value of net tax payments by liv- in 2000), they transpose the skill sharing of the ing generations can be obtained by summing the neighboring country with the closest enrolment present value of net taxes these generations will rate in secondary/tertiary education, the closest gender gap in enrollment rates, and/or the clos- 52 See http://esa.un.org/unpp. est GDP per capita. 53 See http://www.cia.gov/cia/publications/factbook. 127 Keller MNA 5-27-10vol2.indd 127 5/27/10 2:42 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options pay to the government over the rest of their life, n0X ,t i.e., summing the generational accounts of living LNR0X ,t = cohorts. We distinguish three educational levels W0X ,t (L = low level, M = mean level and H = high level) and suppose that each individual can live a maxi- The basic issue of the generational accounting mum of D years. The present value of payments is the financial sustainability of public policies. by living generations, PVLt, can be written as Given the generational account of the newborns X at time t ( n0, ), will it be possible to be so gener- ( ) D t PVLt = ∑ n jL,t p jL,t + n jM p M + n jH,t p jH,t ,t j ,t ous with future generations? The present value j =0 of net tax payment by future generations, PVFt, does not itself give an answer to this question. To where p jX,t stands for the size of type-X popula- go further, one needs to transform this aggregate tion (X=L,M,H) of age j at time t and n jX,t mea- burden into an individual amount, the average sures the generational account of these agents. account of future cohorts. Generational accounts sum up the amount One way to proceed is to compute the hypo- of net taxes to be paid by each type of individual thetical generational accounts of future cohorts over the rest of his life: under the current fiscal policy. Using the same methodology than for living cohorts, we write: 1 D θk X . pkX,t +k− j n = X X j ,t p j ,t ∑ k= j ,t + k− j (1 + i)k− j j = 0,..., D Min( s−t −1, D ) ∞ θL pL + θM pM + θH pH where θkX ,t + k− j is the amount of net tax payment PVFt = * ∑ s=t +1 ∑j =0 j ,s j ,s j ,s j ,s j ,s j ,s (1 + i)s−t by an agent of type X and age k at time t+k–j. where PVFt* measures the present value of X In practice, p can be projected using k ,t + k− j net payments by future generations under the demographic forecasts (including mortality and assumption that the current fiscal policy is sus- net immigration flows), data on schooling levels tainable. per age, and estimates for the educational at- tainment of the young living generations after This hypothetical value PVFt* can then be completion of their education. The net taxes compared to the residual value PVFt computed θk X ,t + k− j can be partly extrapolated on the basis from the intertemporal budget constraint: of short-run forecasts (taking account of official budgetary projections and potential fiscal re- • If PVFt* = PVFt, the policy is sustainable and forms between t and t*) and partly extrapolated there is no need of fiscal adjustment; using balanced growth assumptions (between t* • If PVFt* > PVFt, the government budget is in and ∞). Typically, different assumptions can be surplus and benefits could be increased for considered for the items of θk X ,t + k− j . the same levels of taxes; • If PVFt* < PVFt, the current policy is not It should be noted that the generational ac- sustainable or not generationally balanced: it counts of the newborns, measuring the present implies either that future generations must value of net taxes they can be expected to pay over pay different net taxes than current genera- their whole lifetime, need not to be of the same tions or current policy must be adjusted to sign. It can be negative for low skill individuals and restore sustainability. positive for the high skilled. These generational accounts can be expressed as percentage of the In case of unsustainability, the basic meth- discounted lifetime labor income, denoted by odology suggests the adjustment of taxes and/or W0,X t for a newborn agent of type X. In the line of transfers at some date. In this paper, we use an GPS, this defines the lifetime net tax rate of the adjustment method, which concerns all members newborns: of all generations. If a gap has to be financed (in 128 Keller MNA 5-27-10vol2.indd 128 5/27/10 2:42 PM Appendix 12: Computable General Equilibrium Analysis of Impact of Increasing MENA to Europe Migration case of deficit) or allocated (in case of surplus), government. At this point, we provide first evalu- we compute the proportional adjustment in all ations of the implications of increased migration taxes (or in all transfers) required to balance on destination and origin regions. A last step is to the budget.54 include in the model diverse side effects of skilled emigration on source countries. The elasticities Let us decompose the net taxes on all gen- used for such an analysis were estimated in vari- erations in two basic components, taxes and ous articles of the brain drain literature. benefits: θ X j ,s ≡ θT X , j ,s − θB X , j ,s . A time-invariant ad- justment factor can be applied to each of these Demography components (gT for taxes and gB for benefits) so as to restore sustainability. We then apply these The population structure relies on the UN Popu- proportional changes to both living generations lation projections that are available between 1950 (over the rest of their lifetime) and future gen- to 2050. We also use this data to identify the erations so as to balance the budget constraint. migrants from developing to developed regions. Our adjustment rule is then summarized by the following set of equations: Population. Individuals reaching age 0 at year t belong to the generation t. The size of the young D D θT X (1 − γ T ) − θ B X (1 − γ B ) pkX,t +k− j generation increases over time at an exogenous PVFt adj = ∑∑∑  ,k ,t + k− j ,k ,t + k− j  growth rate: j =0 k = j X (1 + i)k− j Min( s−t −1, D ) θT  X N 0,t = mt −1 N 0,t −1 ,  , j ,s (1 − γ T ) − θ B, j ,s (1 − γ B ) p j ,s X X ∞ PVFt adj = ∑ s=t +1 ∑j =0 ∑ X (1 + i)s−t where N0,t measures the initial size of generation PVLadj t + PVFt adj + NWt = PVGt t and mt–1 is one plus the demographic growth rate, including both fertility and migration. There is a continuum of pairs (gT, gB) re- storing the balance. Two specific pairs are usu- At every period, agents of the same age class ally considered, one with gT=0 if the balance is (a=0,1,...,7) face an identical cumulative survival achieved through transfer cuts and one with gB=0 probability, which decreases with age. Hence, the if the balance is achieved through tax increases. size of each generation declines deterministically For each scenario, the lifetime net tax rate of fu- over time: ture generations can be computed and compared to that of the current newborns. N 0,t = mt −1 N 0,t −1 , Description of the Computable where 0 ≤ Pa,t +a ≤ 1 is the fraction of generation General Equilibrium (CGE) model t alive at age a (at period t+a). Obviously, we have P0,t=1. The CGE analysis can be resumed in three dif- ferent steps: calibration of the demography, Denoting φt by the proportion of skilled construction of the CGE model, and introduc- (post-secondary educated) among the first co- ing the side effects of skilled migration. Before hort, the skilled and unskilled cohort sizes are focusing on the CGE model, we constructed the given by: evolution of the population over the 2000–2100 period based on UN population forecasts. The evolution of the population (as well as of inter- national migration) then enters the model and 54 Our strategy slightly differs from Gokhale J., B.R. Page and determines the path of the economy over the J.R. Sturrock (1999) (GPS) who compute the changes in taxes and/or benefits so as to equalize the lifetime net tax rates of 21st century. The CGE model is characterized current and future generations. It should be noted that, in the by overlapping generations of individuals with line of GPS, the balance can also be restored through changes skill heterogeneity, a production sector and a in government purchases 129 Keller MNA 5-27-10vol2.indd 129 5/27/10 2:42 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options N 0s,t = φt N 0,t explicitly track migrants from the seven devel- oping regions into the three developed regions. N 0u,t = (1 − φt ) N 0,t North-North and South-South migrants are implicitly dealt with through the UN population We assume an exogenous participation profile data and forecasts. Our calibration strategy is per age and education group, λ a j . Hence, labor based on immigrant-to-population ratios, or ,t supply of type j at time t is given by the proportions of stock of immigrants to total population observed in the three developed regions. To begin with, we use statistics on the Ltj = ∑ λ a j ,t N aj,t , j = u, s . number, age structure and education levels of a immigrants in 2000 (combining the UN and the Specifically, we assume full participation except Docquier-Marfouk datasets). From the gross for the following three groups. First, young number of immigrant stock in each region, skilled spend a fraction of their time in obtain- we subtract the number of 0 to 14 years old ing education and do not fully participate in the migrants, and then we subtract all North-to- labor market ( 0 ≤ λ 0 s ,t ≤ 1 ). Second, part of the North migrants. Based on the Docquier-Marfouk population aged 55 to 64 years old are retired dataset, we calibrate the shares of immigrants ( 0 ≤ λ4 j ,t ≤ 1 ). Lastly, all individuals aged above by education level and by region of origin. 65 are retired ( λ a j ,t = 0 for all a>0). To construct the number of migrants before 2000, we use survival probabilities as well as Demographic changes affect the economic the growth rate of the immigrant population. performances of the economy (GDP or GNI per For immigration forecasts, we start from the capita) through the support ratio, defined as the 2000 numbers and let migrants die according ratio of labor force to population: to the survival probability forecasts. Assuming that all future migrants are aged 15–24, we let the change in the stock of immigrants follow SRt = ∑∑λ N a j j a ,t j a ,t , the UN forecasts (from which we subtract the ∑∑ N a j j a ,t 0–14 years old and North-to-North migrants using the 2000 proportions). We assume that And through human capital, defined as the future migrants are distributed by educational proportion of skilled in the residents labor force: level and by origin as in 2000. HCt = ∑λ N a s a ,t s a ,t ∑∑λ N a j j a ,t j a ,t 55 We firstly aggregate this data set by region and then partition it to obtain shares of skilled per age group. We proceed as fol- In the baseline, we compute Pa,t +a , the prob- lows in order to disaggregate the Barro-Lee data by age group. ability for an individual of generation t of being First, it is reasonable to assume that, at each period, the share alive at time t+a and the population growth rate of skilled individuals is higher for the younger age class. In particular, we assume that the share of skilled individuals aged mt for the period 1950 to 2050 from the United 85 to 94 corresponds arbitrarily to 80% of the share of skilled Nations data of World Population Prospects, the aged 75 to 84, which in turn is equal to 80% of that of the next 2000 Revision. In order to compute the share of younger age class, and so forth. As all the shares of skilled per age class then depend on the share of skilled aged 25 to 34, we skilled individuals of a generation φt, we use the compute this share in order to matches the total share of skilled Barro-Lee Dataset (2001), which provides data in 1950, as given by the Barro-Lee Dataset. Second, we report on the educational attainment of individuals aged the values of the shares of the age classes 25-34 to 65-74 of the 25 to 74 for the years 1950 to 2000 per country.55 following years. For example, the share of skilled aged 35 to 44 in 1960 is equal to the share of skilled aged 25-34 in 1950, as we In the future, we assume that the young cohorts assume that the skilled and unskilled individuals have the same are educated as the young one in 2000. probability to be alive at the beginning of each period. Third, for all the following years, we compute the share of skilled aged 25 to 34 in the same way as for the year 1950. Lastly, the share South-North migration. In order to calibrate of skilled aged 15 to 24 in 1950 is simply equal to the share of migration stocks and flows for the baseline, we skilled aged 25 to 34 in 1960. 130 Keller MNA 5-27-10vol2.indd 130 5/27/10 2:42 PM Appendix 12: Computable General Equilibrium Analysis of Impact of Increasing MENA to Europe Migration The model’s structure process representing the Harrod neutral tech- nological progress. Total efficient labor force Preferences. The expected utility function of combines the demands of skilled ( Ls t ) and of u our agents is assumed to be time-separable and unskilled labor ( Lt ) according to the trans- logarithmic: formation function characterized by a constant elasticity of substitution (CES): 7 E(Utj ) = ∑ Pa,t +a ln( caj,t +a ) , j = s, u 1 a =0 Lt = [Ï…t ( Ls t )σ + (1 − Ï…)( Lu t )σ ]σ , σ<1 j where c represents expenditures of age a ,t + a where ut is an exogenous skill-biased technologi- class at time . For natives in both developing cal progress, and s is defined as s = –1(1/ e), with and developed regions, caj,t +a is equivalent to e being the elasticity of substitution between goods consumption. However, for immigrants in skilled and unskilled labor. The capital share in the developed regions, caj,t +a is a Cobb-Douglas M, j output a is set to one third, as estimated in the combination of goods consumption ( ca ) and M, j ,t + a growth accounting literature. We follow Acemo- remittances ( RMa ). ,t + a glu (2002) in fixing the elasticity of substitution to 1.4 and thus the parameter s equals to 0.2857 1− γ tj γj caj,t +a = ( ca M, j ,t + a ) M, j ( RMa ,t + a ) t , j = s, u in the CES function. where gtj is an age-invariant propensity to remit There is one leading regional economy, North that determines the proportion of expenditures America (NAM), in the sense that the Harrod a migrant of generation t and skill group j sends neutral technological progress of each region as remittance to his/her region of origin. More- is a fraction of AtNAM, namely that the leader is over, we assume that this parameter varies with always ahead of other regions in terms of pro- country of origin.56 duction technology. Exogenous paths for the Harrod neutral technological progress At, the Furthermore, following de la Croix and skill-biased technical change and ut growth of the Docquier (2003), we postulate the existence of leading economy are unobservable and/or must an insurance mechanism à la the Arrow-Debreu be properly calibrated. model. As an individual dies, her/his assets are equally distributed among individuals belonging To obtain At for non-leading regions, we use to the same age class. Individuals thus maxi- the observed paths of GDP ratio, Yt /YtNAM, where mize their expected utility subject to a budget YtNAM is the leader’s GDP. We proceed as in de la constraint requiring equality between the dis- Croix and Docquier (2007), who use a backsolv- counted expected value of expenditures and ing identification method to calibrate total factor the discounted expected value of income, which productivity. It consists of swapping the unob- consists of net labor income, pension benefits, served exogenous variables At for the Yt /YtNAM other welfare transfers, and/or net remittances. observed endogenous variables and then solving the identification step with the Dynare algorithm Firms. At each period of time and in each region, (Juillard, 1996). The ratio of GDP’s is computed a representative and profit-maximizing firm uses by employing the data of the GDP per purchasing efficient labor (Lt) and physical capital (Kt) to power parity from the World Development Indi- produce a composite good (Yt). We assume a cators (WDI) for the three years 1980, 1990 and Cobb-Douglas production function with constant 2000. We adopt the value of 1980 for the years returns to scale, preceding 1980 and the value of 2000 for those Yt = K tα ( At Lt )1−α , 56 We model remittances in this way so that migrants and na- tives have identical asset accumulations. The age-invariance of where a measures the share of capital returns propensity to remit comes from our assumption that there is in the national product, and At is an exogenous no remittances decay. 131 Keller MNA 5-27-10vol2.indd 131 5/27/10 2:42 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options following 2000. We apply the same procedures Equilibrium. A competitive equilibrium of the for skill-biased technological change by using skill economy with perfect capital mobility is char- wage premiums, ht = wts / wtu The skill premiums acterized by i) households’ and firms’ first order for each region in year 2000 are arbitrarily fixed.57 conditions, ii) market-clearing conditions on the goods and labor markets, iii) budget balance Then, we let these values vary according to for each regional government, iv) the equality the pattern of the U.S. college wage premium between the aggregate quantity of world assets for the period 1950–2000 in Acemoglu (2003). and the quantity of the world capital stock plus Finally, the leader’s growth of Harrod neutral the sum of public debts of all regions, and finally technological progress is calibrated on real v) the arbitrage condition of the rates of return observations, and for future years, the value is to capital. The equilibrium on the goods market calibrated such that the annual growth rate is is achieved by Walras’ law. equal to 1.84 percent. The arbitrage condition in an integrated Government. The government levies taxes on economy with perfect capital mobility requires labor earnings and on consumption expenditures the equality of the expected returns to capital in order to finance general public consumption, up to region specific risk premium. pension benefits and other welfare transfers. The government surplus (St) can be written as Side effects of migration (for j = s,u): Migration, and especially skilled migration, has 7 St = Ï„ w t ∑ L w +Ï„ j t t j c t ∑ ∑φ j t −a N c j a ,t a ,t been found to have diverse side effects on mi- j ={ s ,u } j ={ s ,u } a =0 grants’ source countries. In particular, human 7 capital formation, technology adoption, and − ∑ b ∑φ j ={ s ,u } t j a =0 j t −a N a,t (1 − etj )(1 − λ a j ,t ) informational costs can be affected by skilled 7 emigration. The following paragraphs we explain −ψ t ∑ wtj ∑ φtj−a N a,t (1 − etj )ζ a j − ctgYt how we integrate these effects in the “Brain Gainâ€? j ={ s ,u } a =0 scenario. where λ a j ,t is the labor participation rate for a j Human capital. A recent wave of theoretical type individual of age class a, wt is labor income, contributions demonstrates that skilled migra- Ï„tc is consumption tax, Ï„tw income tax, btj (indi- tion can raise the average of human capital in vidual) pension benefits, ζa j are other welfare the sending countries (Mountford 1997, Stark transfers received by an individual of type j and et al. 1997 and 1998, Vidal 1998, Beine et al. they are represented as a time-constant fraction 2001 and Stark and Wang 2002). These papers of labor income, the generosity factor ψ t is the assume that the return to education is higher factor by which these other welfare transfers abroad and that skilled workers have a much are multiplied at time t, ctg is a part of national higher probability to emigrate than unskilled income used to finance general public spending. workers (a hypothesis strongly supported by Education is exogenous and individuals spend a the data). Hence, migration prospects raise fraction etj of their total time (which is only posi- the expected return to human capital and tive in their first period of life), φtj is the propor- induce more people to invest in education at tion of individuals of skill type j among generation origin. Ex-ante, more people opt for education. t ( φtj = φt when j=s and φtj = 1 − φt j=u). Ex-post, some of them will be leaving. Under certain conditions detailed in these models, the The government also issues bonds and pays incentive effect (or brain effect) dominates that interests on public debt. The government’s of actual emigration (or drain effect), which budget constraint is satisfied at each period by adjusting the wage tax rate . Public debt is com- puted from the WDI Database. 57 h2000 is fixed at 2.35 in EU and 3 in MENA. 132 Keller MNA 5-27-10vol2.indd 132 5/27/10 2:42 PM Appendix 12: Computable General Equilibrium Analysis of Impact of Increasing MENA to Europe Migration creates the possibility of a net brain gain for modern technologies. Vandenbussche, Aghion the source country. and Meghir (2006), henceforth VAM, estimated a neo-Schumpeterian model using panel data Beine et al. (2008) found evidence that the on OECD countries. More recently, Lodigiani prospect of skilled migration is positively associ- (2008) has extended the framework by adding ated with gross (pre-migration) human capital a diaspora externality: skilled emigrants living levels in cross-country regressions. They used a in rich countries increase the capacity to adopt β-convergence empirical specification: modern technologies. She re-estimated the model on a larger sample of countries (including ln( HCta ) = α − β ln( HCta ) + γ ln( mts ) + ∑ ηi X i,t developing countries) and obtained the follow- ing specification: where ln( HCta ) is the growth rate of human capital between t and t+1, HCta denotes human  A capital measured as the proportion of skilled ln( At ) = .59 − .28 ln  t∗  + 1.43 HCt among natives at time t (superscript a stands for  At  natives, or human capital ex-ante, before emigra-  A tion occurs), mts is the skilled emigration rate, −0.10 ln( Mts ) + .87 ln  t∗  HCt is a vector of other control variables, (a,b,g,h)  At  is a set of parameters. The long-run elasticity of  A natives’ human capital to the skilled emigration −0.06 ln  t∗  ln( Mts ) + ε t  At  rate is equal to g / b. It amounts to 9.6 percent in the parsimonious IV model. where ∆InAt is the rate of technical progress, In our simulations, we build on Docquier, At∗ is the technological level of the leader (typi- Lowell and Marfouk’s data and compute the rela- cally, the NAM region), Mts is the stock of skilled tive change in skilled emigration rates resulting emigrants living in the leading economy and et from the rise in emigration flows to the North. is an exogenous component. Confirming VAM, We assume that all the countries belonging to the interaction effect between proximity and the region experience these relative changes. the proportion of workers with tertiary educa- Using the above long-run elasticity, we compute tion is positive, meaning that skilled workers are the change in human capital of natives and resi- more important for growth in economies closer dents (natives minus migrants). Assuming that to the frontier. On the contrary, the interaction the long-run level of human capital is reached effect between proximity and the log of skilled in 2050, we compute the proportion of skilled emigrants is negative, implying that skilled emi- among remaining residents, denoted by HCt. gration has a decreasing effect on growth when We first do it country-by-country and then ag- a country approaches the frontier, or that migra- gregate countries by region. The convergence tion is more important for countries far from the to the 2050 level of human capital level is linear. frontier. Backward countries, that rely more on Finally, we compute the change in φt required to adoption, can benefit more from skilled diaspora obtain the desired levels of human capital. The as it facilitates technology and knowledge transfer skilled emigration rates and the φt come back to from abroad. their baseline value after 2060. In the baseline, we plug the human capital Total Factor Productivity (TFP). Following and migration forecasts in the above equation Benhabib and Spiegel (2005), themselves to predict the evolution of the technology. On building on Nelson and Phelps (1966), we the period 1950–2000, we calibratee e so that consider an endogenous Harrod-neutral our baseline simulations perfectly match the technical progress of the neo Schumpeterian GDP observations (as percentage of the lead- type. Technical changes are determined by ing economy). The calibrated path for is rather the regional capacity to innovate and to adopt stationary and distributed around zero. 133 Keller MNA 5-27-10vol2.indd 133 5/27/10 2:42 PM   Labor Migration from North Africa – Development Impact, Challenges, and Policy Options Since our shock modifies human capital and of a complementarity between FDI and skilled the number of skilled emigrants abroad, it affects migration with a similar elasticity. the rate of technical progress as well. Given the specification above, ∆InAt increases in HCt when In our model, we assume that physical capital A  is mobile across regions, the optimal marginal ln  t∗  > –1.64, i.e. when the economy is not productivity of capital is equal to the international A   t  interest rate rt* augmented of country-specific pre- too far from the frontier. Moreover, ∆InAt in- mium pt reflecting informational costs or risks. The A  premium level is endogenous and depends on the creases ln  t∗  if < –1.67, i.e., when is far from size of the skilled diaspora abroad ( Mts ). We have: A   t  the frontier. rt∗(1 + Ï€t ) = αK α−1( A L )1−α − d t t t (1 + Ï€t ) = (1 + Ï€0,t )( Mts )− Risk premium . A large sociological litera- ture emphasizes that the creation of migrants’ networks facilitates the further movement of where d is the capital depreciation rate, p0,t is an persons, and the movement of goods, factors, exogenous variables used to calibrate the base- and ideas between the migrants’ host and home line level of the premium and –ϕ is the elasticity countries. Several studies investigated whether of the premium to the skilled diaspora size. FDI and migration are substitutes (as one would expect) or complements. Using cross-section Using panel data, Docquier and Lodigiani data, Docquier and Lodigiani (2008) find evidence (2008) have estimated that the long-run elastic- of significant network externalities in a dynamic ity of foreign direct investments to the skilled empirical model of FDI-funded capital accu- diaspora is equal to 0.75. Using the specifications mulation. Their analysis confirms that business above and relying on the fact that foreign direct networks are mostly driven by skilled migration. investments represent 12.5 percent of total in- Using bilateral FDI and migration data, Kugler vestments in developing countries, the calibrated and Rapoport (2007) also found strong evidence value for Ï• is equal to 0.05. 134 Keller MNA 5-27-10vol2.indd 134 5/27/10 2:42 PM Keller MNA 5-27-10vol2.indd 135 5/27/10 2:42 PM Keller MNA 5-27-10vol2.indd 136 5/27/10 2:42 PM