WPS5104 Policy Research Working Paper 5104 Poverty in Latin America Sources Of Welfare Disparities in Ecuador Gladys López-Acevedo Monica Tinajero The World Bank Europe and Central Asia Region Poverty Sector October 2009 Policy Research Working Paper 5104 Abstract This paper contributes to the analysis of spatial poverty difference. Comparing a leading and a lagging region, in Ecuador by deepening the understanding of the such as the coast versus the Amazon, the characteristics constraints faced by the poor in the country through explain about 90 percent of the welfare differential in an investigation of the role of portable characteristics urban areas, while the returns explain about 30 percent (human capital) and geography in explaining welfare. of the welfare differential in rural areas. Among the At the national level, the results indicate that these characteristics analyzed, education is the most important characteristics explain 72 percent of the differences in variable for explaining differences in living conditions welfare level between urban and rural areas, while returns between urban and rural areas in Ecuador. to these characteristics account for 28 percent of the This paper--a product of the Poverty Sector, Euorpe and Central Asia Region--is part of a larger effort in the department to increase knowledge on poverty. Policy Research Working Papers are also posted on the Web at http://econ.worldbank. org. The author may be contacted at gacevedo@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team POVERTY IN LATIN AMERICA: SOURCES OF WELFARE DISPARITIES IN ECUADOR Gladys López-Acevedo Senior Economist, LCSPP The World Bank gacevedo@worldbank.org Monica Tinajero Consultant The World Bank Tinajerobm@gmail.com JEL classification: Keywords: Ecuador, Welfare, Poverty Corresponding author: Gladys Lopez-Acevedo, The World Bank, 1818 H Street NW, Washington DC 20433-USA. E-mail: gacevedo@worldbank.org. This paper is part of a Latin America and Caribbean regional World Bank study on Sources of Welfare Disparities Within and Across Regions in LAC financed by the LAC Chief Economist office and the PREM Poverty anchor unit. The findings, interpretations, and conclusions in this paper are entirely those of the authors and do not necessarily reflect the view of the World Bank. Background paper for the regional study on poverty in Latin America at the World Bank. These are views of the authors and do not necessarily reflect those of the World Bank, its executive directors, or the countries they represent. Introduction Countries in Latin America continue to struggle to share the benefits of economic growth with the poor segments of their population. The progress towards achieving the first Millennium Development Goal of eradicating extreme poverty by 2015 is still far from being achieved in most Latin American countries. Another feature common in many countries in the region is geographic disparities in living standards that persist over time in spite of overall economic growth. A better understanding of which factors influence disparities in living standards is crucial to provide better guidance to poverty reduction strategies. Two opposing views present divergent explanations on which factors are likely to influence the spatial distribution of poverty. The "concentration view" holds that poor areas arise from the persistent concentration in these areas of individuals with personal attributes that inhibit growth in their living standards. According to this view, otherwise identical individuals will have the same growth prospects independently of where they live. Thus, geography does not play a causal role in explaining the level of and growth in living standards. The other view holds that "geography" itself is the cause of the high level of poverty and weak growth of living standards over time. In areas better endowed with local public goods, such as better access to infrastructure and other basic services (electricity, water and sanitation), geographic externalities may facilitate the exit of poor households from poverty. This paper provides an applied framework to analyze which of these two views better explains spatial disparities in poverty in Ecuador. Ecuador is located in northwestern South America, with an ethnically diverse population comprised of mestizos (62 percent), Amerindians (25 percent), immigrants from European countries as well as unmixed descendants of early Spanish colonist (10 percent) and Afro-Ecuadorians (3 percent). The population of over 13,750,000 people (in 2008) is growing at a rate of 1.1 percent per year,1 and one of every four Ecuadorians is between 15 and 29 years old. Nearly 60 percent of its population lives in urban areas and the rest in the rural sector. Ecuador's economy depends heavily on oil exports, which accounts for an average of 35 percent of GDP (1997-2001). In the past three decades, GDP per capita growth has been low, at an average of 1.7 percent annually during 1970-2002. GDP volatility has been high, mainly as a consequence of external shocks such as downturns in commodity prices, as well as unstable fiscal policy. At the end of 1999 Ecuador faced a severe economic crisis due to several factors: external economic shocks, "El Niño" weather phenomenon (1997) and poor economic management. The economy recovered after the 1999 crisis, with GDP per capita rising to 2.1, 6.5 and 4.5 percent in 2003, 2004 and 2005, respectively. Average inflation fell from an annual rate of 52.2 percent in 1999 to 7.9 percent in 2003, 2.7 percent in 2004 and 2.1 percent in 2005.2 Poverty and inequality remain a major concern for Ecuador. According to the World Bank's Poverty Assessment for Ecuador (2004), poverty increased from 40 percent in 1990 to 45 percent in 2001 (mainly as a consequence of the 1999 crisis), and fell to around 36 percent between 2001 and 2004 (mainly as a consequence of macroeconomic stabilization). While poverty declined in both urban and rural areas, in 2004 the rural poverty rate was still more than twice the urban poverty rate. With these differences in mind, the goal of this paper is to evaluate how much of the difference in standards for living across areas and regions in Ecuador can be attributed to disparities in the mobile non- geographic variables, and how much to location differences in the returns to those characteristics. The paper follows closely the methodological framework of Ravallion and Wodon (1999), in which they propose a methodology for decomposing the differentials in living standards between geographical areas into: a) differences in non-geographic characteristics and b) differences in returns to those characteristics. 1 World Development Indicators database, April 2008. 2 Idem. 2 The paper is structured as follows. Section 1 presents a brief description of the data, and Section 2 presents the model. Section 3 discusses the national-level results from the model using regression analysis to test the relationship between different household characteristics and welfare. Section 4 carries out the decompositions at the national level to analyze the relative importance of characteristics and returns to characteristics on welfare. Section 5 discusses the extension of the model to allow for comparisons within and across regions. Section 6 concludes. 1. Data Source This analysis uses data from the Living Conditions Survey (Encuesta de Condiciones de Vida--ECV) for 2005-2006. The ECV is a micro-level dataset collected by Ecuador's National Statistics and Census Institute (Instituto Nacional de Estadística y Censos--INEC). The ECV collects information about individual characteristics like gender, age, education, main activity, income, household's attributes, and detailed information on the expenses of the household. The survey is representative at the national level and by urban and rural areas. Other representative areas include Quito, Guayaquil, Cuenca and Machala. The survey design was stratified, multi-staged and clustered. At the first and second stages the units were selected with probability proportional to the number of households on them, while at the third stage households were selected with equal probability. Weights are then necessary to get suitable estimations. The final sample size was 13,581 households, 8,065 in urban area and 5,516 in rural areas (Table 1). Table 1. Households by Region Region Urban Rural Total Sierra 3,815 3,451 7,266 Costa 3,770 1,537 5,307 Oriente 480 528 1,008 Total 8,065 5,516 13,581 Source: Own calculations based on ECV. Nation-wide, the ECV indicates that urban households have an average consumption of one and a half times the poverty line, while the mean consumption of rural households is almost equal to the poverty line. 2. Model A linear model is used to estimate the contribution of location differences in living standards after controlling for non-geographic attributes. The dependent variable is the logarithm of the welfare ratio, which is a proxy for living standards, and the independent variables are household attributes. These are two different regression equations, one for each area (urban and rural). The model is as follows: ln Wi U U X i U G i Ui ' ' (i U ) , (1) ln Wi R 'R X i 'R G i Ri (i R) , (2) where Wi denotes the welfare ratio, defined as the per capita consumption deflated or divided by a national poverty line at the household i , X i is a vector of non-geographic variables for the household i , G i is a vector of geographical dummy variables (provinces), and U and R denote the urban and rural 3 areas. The error terms, Ui and Ri , are assumed to be independent and identically distributed with zero mean; U , R , U , R , U and R are the parameters to be estimated.3 Based on the above specification, differences in living standards between different geographic areas may be attributed to two main factors: 1. differences in the mobile non-geographic characteristics between urban and rural areas, holding all else constant, i.e. differences between X Ui and X Ri ; or 2. differences in the returns to characteristics, holding all else constant, i.e. differences between U and R . The vectors X i and G i of explanatory variables include the following: Demographics. The number of babies, children, teenagers and adults; age of the household head and their squared values; gender of the household head; household structure (household head married with spouse, single without spouse, separated/divorced/widowed without spouse, and married without spouse present); and ethnicity of household head (indigenous, mixed raced, white, black, other). Education. Level of education according to the household head and his/her spouse years of schooling (none, one to four, five to seven, eight to 10, 11 to 13 and 14 or more years). House ownership. Four categories: own, rent, own and paying, other. Occupation status. Whether the household head is an employee, employer, self-employed, employee without pay, farm laborer, owner-farmer, self-employed farm laborer or not working. Geography. In addition to urban and rural areas, there are dummies for each of the 21 provinces4 (the ECV excludes the Galápagos islands): Azuay, Bolívar, Cañar, Carchi, Cotopaxi, Chimborazo, El Oro, Esmeraldas, Guayas, Imbabura, Loja, Los Ríos, Manabí, Morona Santiago, Napo, Orellana, Pastaza, Pichincha, Tungurahua, Sucumbíos and Zamora. Several of the variables mentioned above are categorical, therefore it is necessary to leave one category as a reference group. Those categories are: Pichincha province (the location of Quito, Ecuador's capital), male household head, married with spouse, no education of household head, no education of spouse, owner of house, and employee. Analogous equations to (1) and (2) are estimated by region in order to allow for variation of coefficients across regions, as follows: ln Wi Uk Uk X i Uki ' (i Uk ) , (1') ln Wi Rk 'Rk X i Rki (i Rk ) , (2') where Uk and Rk denote the urban and rural areas of region k , respectively; k S , C , O denotes Sierra ("mountain"), Costa ("coast") and Oriente ("east" or "jungle"). These regions are conformed by the following provinces: 3 The omitted variables are assumed to be uncorrelated with place of residence, otherwise the estimates of geographic effects will be biased. 4 Note: The provinces of Santa Elena and Santo Domingo de los Tsachilas were created in 2007, and were not used in the survey data employed by this study. 4 Sierra: Azuay, Bolívar, Cañar, Carchi, Cotopaxi, Chimborazo, Imbabura, Loja, Tungurahua and Pichincha provinces, which account for 47 percent of households in Ecuador. Costa: El Oro, Esmeraldas, Guayas, Los Ríos and Manabí provinces, which account for 49 percent of households in Ecuador. Oriente: Morona Santiago, Napo, Orellana, Pastaza, Sucumbíos and Zamora provinces, which account for 4 percent of the households in the country. The urban-rural composition in the first two regions is similar, with 65 percent of the households in the urban area in the case of Sierra and 75 percent in the case of Costa. In the Oriente region, approximately 61 percent of households are in rural areas. 3. Estimates from the Model at the National Level This section estimates the returns to each one of the household characteristics at the national level, by urban and rural areas, using linear regression. This is the necessary first step before subsequently moving on to the decomposition analysis. The explanatory variables from equations (1) and (2) included in the model explain approximately 59 percent and 52 percent of the variability in the welfare ratio for urban and rural areas, respectively (Table 2).5 The estimated effects differ by areas. However, these effects are significant and with the expected sign by area. The analysis indicates the following: An increase in household size and having a female household head is correlated with lower consumption; the welfare ratio for a household head with spouse is less than the welfare ratio of a household head without spouse; an increase in the household head's age increases the welfare ratio; and the welfare ratio is lower for indigenous household heads. The welfare ratio of a household head with some education is greater than the welfare ratio of a household head with no education, and the same result applies to the education of the spouse. The returns to education of the household head's spouse are significant in both urban and rural areas, but higher in urban areas. Consumption is lower in households that do not own their house (rent, paying, other) compared to households with their own house. Welfare ratios are higher for employers or owner-farmers than for employees, and lower for the self-employed, employees without pay, farm laborers, self-employed farm laborers or those not working. On average, consumption in several provinces is lower than in Pichincha province, which includes the national capital. Estimated coefficients for rural and urban areas are different for some provinces. 5 In order to estimate the parameters and their standard errors, the survey design was taken into account. 5 Table 2. Regressions for Log Welfare Ratio Urban Rural Standard Standard Explanatory variables Coefficient Error Coefficient Error Constant -0.336 * 0.101 -0.279 * 0.107 Province Azuay 0.142 * 0.040 -0.001 0.085 Bolívar -0.265 * 0.077 -0.238 * 0.090 Cañar -0.027 0.049 0.071 0.091 Carchi -0.333 * 0.047 -0.438 * 0.095 Cotopaxi -0.169 * 0.055 -0.063 0.085 Chimborazo -0.198 * 0.056 -0.202 * 0.089 El Oro -0.186 * 0.039 -0.145 0.094 Esmeraldas -0.191 * 0.044 -0.174 * 0.084 Guayas -0.247 * 0.036 -0.057 0.089 Imbabura -0.200 * 0.042 -0.294 * 0.100 Loja -0.077 ** 0.043 -0.272 * 0.094 Los Ríos -0.382 * 0.039 -0.112 0.086 Manabí -0.358 * 0.047 -0.189 * 0.082 Morona Santiago 0.020 0.083 -0.601 * 0.166 Napo -0.221 0.139 -0.351 * 0.141 Pastaza -0.123 * 0.052 -0.374 * 0.158 Tungurahua -0.035 0.039 -0.089 0.077 Zamora -0.197 * 0.084 -0.298 * 0.092 Sucumbíos -0.002 0.044 -0.209 * 0.085 Orellana -0.039 0.055 -0.335 ** 0.175 Demographics Number of babies -0.296 * 0.036 -0.376 * 0.032 Number of babies squared 0.035 0.021 0.076 * 0.018 Number of children -0.272 * 0.014 -0.264 * 0.015 Number of children squared 0.024 * 0.004 0.024 * 0.004 Number of teenagers -0.209 * 0.022 -0.141 * 0.023 Number of teenagers squared 0.015 0.010 0.012 ** 0.007 Number of adults -0.109 * 0.016 -0.076 * 0.017 Number of adults squared 0.005 * 0.002 0.007 * 0.003 Sex of the head -0.022 0.027 -0.091 * 0.036 No spouse, single 0.253 * 0.052 0.092 ** 0.048 No spouse, separated/divorced/widowed 0.227 * 0.048 0.206 * 0.037 No spouse, married 0.422 * 0.065 0.459 * 0.070 Age of household head 0.028 * 0.003 0.020 * 0.003 Age of household head squared -0.0002 * 0.000 -0.0002 * 0.000 Mixed race (mestizo) household head 0.071 ** 0.040 0.214 * 0.038 White household head 0.167 * 0.049 0.261 * 0.054 Black household head 0.026 0.053 0.274 * 0.060 6 Mixed race (mulato) household head 0.041 0.050 0.262 * 0.068 Other ethnicity household head 0.093 0.350 ---- Education of Household Head 1 - 4 years 0.255 * 0.047 0.156 * 0.028 5 - 7 years 0.452 * 0.046 0.276 * 0.029 8 - 10 years 0.572 * 0.047 0.397 * 0.037 11 - 13 years 0.765 * 0.051 0.520 * 0.050 14 - + years 1.135 * 0.056 0.913 * 0.101 Education of Spouse 1 - 4 years 0.020 0.043 0.036 0.030 5 - 7 years 0.124 * 0.044 0.086 * 0.029 8 - 10 years 0.156 * 0.046 0.144 * 0.040 11 - 13 years 0.267 * 0.045 0.216 * 0.052 14 - + years 0.472 * 0.048 0.512 * 0.075 House Ownership Rent -0.130 * 0.018 0.042 0.060 Own and paying -0.004 0.036 0.269 * 0.127 Other -0.146 * 0.020 -0.057 * 0.027 Position in Occupation Employer 0.313 * 0.027 0.297 * 0.052 Self-employed -0.051 * 0.017 -0.062 * 0.031 Employee no pay 0.057 0.048 -0.154 * 0.054 Farm labourer -0.131 * 0.030 -0.169 * 0.026 Owner farmer 0.238 * 0.055 0.217 * 0.039 Self-employed farm labourer -0.232 * 0.057 -0.197 * 0.029 Not working -0.046 ** 0.027 -0.302 * 0.041 Source: Own calculations based on ECV. Note: Number of observations: 7950 (urban) and 5481 (rural). R2=0.59 (urban) and 0.52 (rural), * indicates that the coefficient is significant at the 5 percent level, and ** at the 10 percent level. The base categories are: Pichincha area (Quito belongs to that area), male household head, married with spouse, indigenous household head, no education of household head, no education of spouse, own house and employee. According to the F tests, the null hypothesis H o : l1 ,U l1 , R , , lM ,U lM , R , where lm is the coefficient for the category m of variable l and M is the number of categories of this variable--i.e., the hypothesis that the effects of all categories of variable l are the same in urban and rural areas--is rejected for almost all variables (Table 3). In other words, all the explanatory variables but education of the spouse have different effects in urban than in rural sectors. Table 3. Test of Equality of Coefficients Between Urban and Rural Regressions Number of T value Test (5% Variable restrictions /F value P value level) Constant 1 -2.61 0.01 Rejected Province 20 5.38 0.00 Rejected Demographics 18 5.87 0.00 Rejected Education variables 10 2.04 0.03 Rejected Education of Household Head 5 2.95 0.01 Rejected Education of Spouse 5 0.63 0.68 Not rejected 7 House Ownership 3 5.16 0.00 Rejected Position in occupation 7 5.58 0.00 Rejected Source: Own calculations based on ECV Survey. Annex 1 presents the results for the three regions chosen in the analysis using equations (1') and (2'). The Oriente region (mainly the rural part) is very different to the rest of the regions in characteristics and on returns. By contrast the Sierra (including Quito) and Costa (including Guayaquil) appear to be more similar on characteristics and returns. 4. Decompositions at the National Level In light of the salient differences in per capita income between rural-urban areas and across provinces, the following analysis considers the contribution of household characteristics to per capita income differences. These differences may come from differences in characteristics (for example, a lower level of education in rural areas) or from differences in the returns to characteristics (for example, a lower impact of education on earnings and thereby a lower per capita income in rural areas). In some cases, the differences in characteristics and in the returns to characteristics reinforce each other, but in other cases they might not. This is analyzed using the Oaxaca-Blinder decomposition of the income gap between assets and returns to those assets (Blinder 1973 and Oaxaca 1973). The estimates from equations (1) and (2) are used to analyze the comparisons. Three types of geographic comparisons are examined: i) the difference in mean welfare ratios between urban and rural areas, ii) the difference within the urban and rural sectors across provinces, and iii) the difference between urban and rural areas within a given province. The first questions concern the overall differential in living standards between urban and rural areas at the national level. This entails a comparison of urban-rural differentials in mean welfare ratios, given by (3): E log Wi i U , X i XU E log Wi i R, X i X R U R U XU 'R X R j pUj Uj p Rj Rj ' (3) where XU and X R are the sample means for urban and rural areas respectively, and pUj and p Rj are the proportions of province j 's population in each area. Table 4 shows the result obtained for equation (3) using the coefficients showed in Table 2 and the means reported in Annex Table A1. The difference in the intercepts, -0.057, gives the difference between the fitted log welfare ratio for a married couple, both illiterate, with a male indigenous household head who owns a house and is an employee living in the urban Pichincha area, and a household with the same characteristics located in the rural Pichincha area. Among the non-geographic variables that could cause the differential impact in urban and rural areas, house ownership is a minor factor. The next variable in order of significance is the occupation of household head, while most of the differential is due to demographic variables and education. In fact, demographics accounts for around 32 percent of the differential in log welfare ratio and education for 68.5 percent approximately, which is explained by both higher levels and higher returns to education in urban sector. The difference due to the geographic variable is small, which indicates that, controlling for other characteristics, the gap between Pichincha province and all other provinces in urban areas is almost of the same order in magnitude as the gap in rural areas. 8 Table 4. Contribution of Variables to Average Level of Log Welfare Ratio Urban- Urban Rural Rural Difference Mean log welfare ratio 0.689 -0.024 0.713 Decomposition Constant -0.336 -0.279 -0.057 Geographic variables -0.158 -0.141 -0.017 Demographic variables 0.418 0.185 0.233 Education variables 0.817 0.328 0.489 House ownership variable -0.054 -0.007 -0.046 Occupation variable 0.001 -0.110 0.111 Source: Own calculations based on ECV Survey. The comparison between urban areas or rural areas of two provinces is based simply on the comparison of the coefficients of province dummy variables in Table 2. For the urban areas, only in Azuay is the coefficient estimate positive, while in the rural areas all of the significant province coefficient estimates are negative. Households living in Pichincha appear to be better off or equal to their urban and rural counterparts from other provinces, after controlling for the non-geographic characteristics. The second component of the right side in equation (3) reflects differentials in urban and rural consumptions due to both differences in returns and in characteristics. It does not quantify structural differences, which is accomplished by comparing the expected gain in consumption from living in urban areas of a given province over rural areas, given the national means of all non-geographic characteristics, X* . For province j this is given by: E log Wi i U , G i G j , X i X* E log Wi i R, G i G j , X i X* U R U 'R X* Uj Rj ' (4) where G j is a vector with zeros in all its entries except for entry j if the household belongs to province j . The first term in equation (4) is the same as in equation (3), and the second term gives the effect of urban-rural differences in the returns to household characteristics. The sum of these two terms is 0.24, which can be obtained from coefficients in Table 2 and means at national level in Table A1, and accounts for the difference between the conditional log consumption of households living in urban and rural areas of the Pichincha province when conditioning on national means. The third term is close to zero for seven of the provinces, meaning that in these provinces the differences in expected log consumptions between urban and rural areas are similar to Pichincha province (Table 5). For most of the remaining provinces, the differences are moderate to large; this suggests that residence in a given province is related to differences in expected consumption, once we control for household characteristics. The province of Morona Santiago has the largest difference in consumption between urban and rural area. 9 Table 5. Differences Uj Rj and Expected Gain by Province Difference Expected Province Urban - Rural gain Azuay 0.142 0.380 Bolívar -0.027 0.211 Cañar -0.098 0.140 Carchi 0.105 0.343 Cotopaxi -0.106 0.132 Chimborazo 0.004 0.242 El Oro -0.041 0.197 Esmeraldas -0.017 0.221 Guayas -0.189 0.049 Imbabura 0.094 0.332 Loja 0.195 0.433 Los Ríos -0.270 -0.032 Manabí -0.169 0.069 Morona Santiago 0.621 0.859 Napo 0.130 0.368 Pastaza 0.250 0.488 Tungurahua 0.054 0.292 Zamora 0.101 0.339 Sucumbíos 0.207 0.445 Orellana 0.296 0.534 Source: Own calculations based on ECV. To estimate the importance of characteristics and returns obtained from these characteristics in explaining welfare differences between urban and rural areas, we analyze each of these factors separately. i) Geographic profile of living standards. This reflects the differentials in returns between urban and rural areas, isolating the structural component, i.e., controlling for all non-geographical attributes. The geographic log welfare ratio is defined as: ln GEOUj E log Wi i U , G i G j , X i X * U U X * Uj ' ' (5) ln GEORj E log W i R, G i i G j , Xi X * R 'R X * Rj . ' (6) Any differences between the log of welfare ratio in (5) and (6), for a given province, are due to differences in the returns to the common characteristics X * . ii) Concentration profile of living standards. This reflects the spatial concentration of non-geographic characteristics, by isolating the effects of non-geographic attributes controlling for geographic variables. The concentration log welfare ratio is defined as: log CON Uj E log Wi X i XU N 'N XU j j log CON Rj E log W X i i XR j N 'N X R , j where XU and X R represent the mean characteristics of the households living in the province j j j for the urban and rural areas, respectively; N and N are the weighted average parameters: 10 N pU U pUj Uj pR R p Rj Rj , N pU U p R R . For most of the provinces, as was expected, both geographic and concentration urban welfare ratios are lower than the unconditional ones (Table 6). However, the reasons are not the same. In the geographic profile, this is because urban households tend to have better characteristics than households at national level. In the concentration profile, this is because the returns to better characteristics tend to be larger in urban areas than nationally. By analogous reasoning, average consumption for geographic and concentration simulations in rural areas is larger compared to an unconditional profile. Table 6. Welfare Ratios by Province and by Urban/Rural Area (logarithms) Geographic Concentration Unconditional Province Profile Profile Profile Urban Rural Urban Rural Urban Rural Pichincha 0.69 0.45 0.68 0.26 0.91 0.25 Azuay 0.83 0.45 0.70 0.17 1.06 0.16 Bolívar 0.42 0.21 0.70 0.04 0.66 -0.23 Cañar 0.66 0.52 0.43 0.12 0.63 0.16 Carchi 0.36 0.01 0.59 0.18 0.49 -0.26 Cotopaxi 0.52 0.39 0.68 0.09 0.74 0.00 Chimborazo 0.49 0.25 0.77 -0.03 0.81 -0.26 El Oro 0.50 0.31 0.59 0.34 0.63 0.16 Esmeraldas 0.50 0.28 0.51 0.05 0.53 -0.07 Guayas 0.44 0.39 0.59 0.11 0.58 0.06 Imbabura 0.49 0.16 0.55 0.03 0.59 -0.26 Loja 0.61 0.18 0.75 0.15 0.90 -0.15 Los Ríos 0.31 0.34 0.49 0.22 0.33 0.11 Manabí 0.33 0.26 0.55 0.09 0.42 -0.08 Morona Santiago 0.71 -0.15 0.60 -0.12 0.84 -0.74 Napo 0.47 0.10 0.84 -0.13 0.86 -0.50 Pastaza 0.57 0.08 0.63 -0.24 0.72 -0.61 Tungurahua 0.65 0.36 0.68 0.18 0.88 0.07 Zamora 0.49 0.15 0.57 0.05 0.59 -0.25 Sucumbíos 0.69 0.24 0.64 0.18 0.86 -0.02 Orellana 0.65 0.12 0.26 -0.11 0.42 -0.43 Source: Own calculations based on ECV Survey. There are important and positive correlations between geographic and unconditional profiles (0.73 for urban, 0.91 for rural and 0.91 for national), and between concentration and unconditional profiles (0.72 for urban, 0.87 for rural and 0.96 for national), meaning that both structural differentials (returns) and disparities in characteristics (endowments) are important to explain urban-rural differences in welfare ratios within provinces. Mean consumption--conditional on non-geographic variables set at national level--is higher in urban than in rural areas, meaning that the returns are larger for the urban sector than for rural areas. When the welfare ratio is simulated setting all parameters at the national level and explanatory variables at the mean of the province, urban measures are also higher compared to the rural ones. This is due to household characteristics that tend to increase living standards, such as education, which is higher in urban than in rural sectors. 11 In order to calculate profiles at the national level, equations analogous to (1) and (2) were estimated without province dummies (Table 7). The results indicate that characteristics are more important in explaining differences between urban and rural sector at the national level. Characteristics explain around 72 percent of the differential in welfare ratio, while returns explain 28 percent. Table 7. Welfare Ratios at National Level by Urban/Rural Area (logarithms) Profile Urban Rural Geographic 0.527 0.326 Concentration 0.625 0.112 Unconditional 0.689 -0.024 Source: Own calculations based on ECV Survey. Poverty Measures This section presents the same analysis as above, but based on poverty rates instead of log welfare ratios. Assuming normally distributed errors, and conditioning on national sample means, the conditional probabilities of being poor for a household living in province j of urban and rural areas are represented as: P log Wi 0 | i U , G i G j , Xi X U U X Uj U P log Wi 0 | i R, G i G j , Xi X R R X Rj R where U and R are the standard deviations of the errors in the urban and rural regressions and is the cumulative density of the standard normal distribution. Similarly, to calculate the simulated poverty rates based on the concentration profile, the weighted average of urban and rural parameters are used. The sources of the differences between simulations and the unconditional (actual) poverty rates are characteristics or changes in returns to those characteristics, and the assumption that logarithm of welfare ratio follows a normal distribution. Thus, the poverty rates for the unconditional profile were calculated assuming a normal distribution. For most of the provinces, the concentration, geographic, and unconditional measures of poverty are all lower for urban than for rural areas (Table 8). The correlations between the geographic and unconditional (normal) profiles are 0.72 for urban and 0.91 for rural, and are very close to the correlations between concentration and unconditional (actual) profiles, which are 0.69 and 0.88 for urban and rural areas, respectively. The conclusions are similar to those discussed above, meaning that both returns and characteristics explain the differences in welfare ratios within each province, while at the national level characteristics are more important. 12 Table 8. Poverty Rates by Province and by Urban/Rural Area Geographic Concentration Unconditional Unconditional Province Profile Profile Profile (normal) Profile Urban Rural Urban Rural Urban Rural Urban Rural Pichincha 8.2 18.4 8.5 30.1 11.5 36.3 12.7 42.5 Azuay 4.7 18.4 8.0 36.8 7.9 41.2 5.0 43.0 Bolívar 19.6 33.5 8.0 47.0 19.1 62.3 11.6 65.9 Cañar 9.1 14.8 19.4 40.8 20.1 41.5 13.8 41.1 Carchi 23.6 48.9 11.7 35.9 25.9 64.2 28.0 64.3 Cotopaxi 14.7 21.9 8.5 43.0 16.5 49.9 14.6 48.5 Chimborazo 16.1 30.9 6.0 52.0 14.3 64.2 16.0 68.3 El Oro 15.5 27.1 11.9 24.9 20.2 41.0 15.3 41.6 Esmeraldas 15.8 29.0 15.0 46.0 24.1 53.6 25.9 53.3 Guayas 18.6 21.6 11.6 41.5 22.2 46.9 22.4 46.2 Imbabura 16.2 37.7 13.2 47.8 21.7 64.2 21.0 68.1 Loja 10.8 36.0 6.5 38.6 11.7 58.4 8.4 63.4 Los Ríos 26.7 24.9 16.2 33.1 33.0 44.0 33.7 45.5 Manabí 25.2 30.0 13.4 42.6 29.0 54.5 31.8 61.2 Morona Santiago 7.6 61.7 11.4 59.3 13.4 84.8 17.0 77.2 Napo 17.3 42.0 4.6 60.1 12.9 75.4 16.9 71.3 Pastaza 12.7 43.8 10.3 68.4 16.9 79.9 11.9 76.0 Tungurahua 9.3 23.5 8.4 36.4 12.1 46.2 11.7 46.9 Zamora 16.1 38.0 12.5 46.4 21.9 63.5 21.4 68.2 Sucumbíos 8.3 31.4 9.7 36.0 12.7 50.8 11.4 50.8 Orellana 9.5 40.8 30.2 58.8 28.9 72.6 23.9 66.5 Source: Own calculations based on ECV Survey. 5. Decompositions at the Regional Level This section expands the analysis from national to regional level. The analysis clearly reveals that for both the Sierra and the Costa, demographics and education account for the largest share in total log welfare ratio (Table 9). In the Costa, demographic factors account for about 28 percent of the differential in log welfare ratio, and education for more than 79 percent. In the Sierra, demographic factors account for about 32 percent of the differential in log welfare ratio, and education for more than 74 percent. In Oriente, the difference in the constant term accounts for the largest share in living standards, which means that a married couple, both illiterate, with male indigenous household head who owns a house and is an employee living in the urban Oriente has a mean log welfare ratio 0.72 higher than a household with the same characteristics located in the rural Oriente. Table 9. Contribution of Variables to Average Level of Log Welfare Ratio by Region Urban-Rural Difference Sierra Costa Oriente Mean log welfare ratio 0.868 0.516 1.142 Decomposition Constant term -0.090 -0.075 0.720 Demographic variables 0.242 0.166 -0.272 Education variables 0.682 0.383 0.494 House ownership variable -0.089 -0.024 -0.042 Position in occupation variable 0.122 0.065 0.242 Source: Own calculations based on ECV Survey. 13 Comparison between urban and rural areas within each of the regions is analogous to that of national level: the concentration profile uses weighted parameters from specifications (1') and (2') and the means for urban and rural areas of the respective region, while the geographic profile was obtained by setting the mean at the regional level and the parameters estimated using (1') and (2'). To compare across the regions of the same area (urban/rural), the concentration measure uses the means of each region only for the area of interest (urban/rural) and weighted returns for the two regions in the comparison from the regressions that were estimated separately for each region. The geographic effect uses the weighted means for the two regions in the comparison and the estimated coefficients for each region and area of interest. The results are presented below. In the Sierra and Costa, the disparities in the concentration profile between urban and rural areas are larger than the differences in the geographic profile, which means that characteristics are the major explanation behind welfare differentials in urban and rural areas. In the Oriente, which is mainly rural (approximately 61 percent of households), both returns and characteristics are important in order to explain the differences in welfare ratio between urban and rural sector (Figure 1). The same conclusions emerged using the poverty rates shown in Table 10. Figure 1. Decomposition of Welfare Ratios (logarithms) Within Regions Sierra Costa 1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0 Geo graphic Co ncentratio n Unco nditio nal Ge o gra phic C o nc e ntra tio n Unc o nditio na l Urban Rural Urban Rural Oriente 1.0 0.8 0.6 0.4 0.2 0.0 Ge o gra phic C o nc e ntra tio n Unc o nditio na l -0.2 -0.4 -0.6 Urban Rural Source: Own calculations based on ECV Survey. 14 Table 10. Poverty Rates by Region and Within Region Geographic Concentration Unconditional Unconditional Province Profile Profile Profile (normal) Profile Urban Rural Urban Rural Urban Rural Urban Rural Sierra 9.4 22.3 5.9 36.5 12.5 49.8 12.8 52.6 Costa 19.2 32.4 16.6 43.6 23.0 48.6 24.2 51.0 Oriente 23.6 59.4 12.0 43.6 16.5 68.7 16.3 67.6 Source: Own calculations based on ECV Survey. For the Oriente, characteristics and returns each explain around 50 percent of the difference in log-welfare ratios between urban and rural sector (Figure 2). In the Sierra and Costa, characteristics explain approximately 30 percent and 34 percent of the differentials, respectively. Figure 2. Decomposition of Welfare Differences (logaritms) Within Regions Oriente Costa Sierra 0% 20% 40% 60% 80% 100% Returns Characteristics Source: Own calculations based on ECV Survey. The urban-rural composition of the Sierra and Costa are similar, with 65 percent of the households in urban areas in the case of Sierra and 75 percent in the case of Costa, while Oriente is mainly rural, at 61 percent approximately. In terms of poverty, Costa is the lagging region in the urban sector while Oriente is behind in the rural area. The comparison between Oriente and Costa on the one hand and Sierra and Costa on the other in urban areas reveals that urban welfare differences are mostly due to returns, rather than characteristics. In the former case, concentration plays a role also, but in spite of the fact that Oriente has better characteristics than Costa, which yields a difference in the concentration profile, it is smaller than the difference in the returns. In the second comparison, both regions have similar characteristics, which lead to a differential only in returns (Figure 3). In the rural sector, both geographic and concentration effects explain the disparities in poverty rates between Sierra and Oriente. Characteristics explain the main differential between Costa and Oriente, although there is a difference in returns as well (Figure 4). 15 Figure 3. Decomposition of Poverty Rates. Urban Areas Oriente vs Costa Sierra vs Costa 30 30 25 25 20 20 Poverty Rate Poverty Rate 15 15 10 10 5 5 0 0 Ge o gra phic C o nc e ntra tio n Unc o nditio na l Unc o nditio na l Ge o gra phic C o nc e ntra tio n Unc o nditio na l Unc o nditio na l No rm a l No rm a l Oriente Urban Costa Urban Sierra Urban Costa Urban Figure 4. Decomposition of Poverty Rates. Rural Areas Sierra vs Oriente Costa vs Oriente 80 80 70 70 60 60 Poverty Rate Poverty Rate 50 50 40 40 30 30 20 20 10 10 0 0 Ge o gra phic C o nc e ntra tio n Unc o nditio na l Unc o nditio na l Ge o gra phic C o nc e ntra tio n Unc o nditio na l Unc o nditio na l No rm a l No rm a l Sierra Rural Oriente Rural Costa Rural Oriente Rural Source: Own calculations based on ECV Survey. In sum, returns explain about 90 percent of the differences in log welfare ratio between the urban areas of Oriente and Costa and about 70 percent between the urban areas of Sierra and Costa (Figure 5). The differential between the rural areas of Sierra and Oriente is explained by returns and characteristics, while the differences between Costa and Oriente are mainly due to characteristics. 16 Figure 5. Decomposition of Welfare Differences Between Regions Oriente vs Costa (Urban) Sierra vs Costa (Urban) Sierra vs Oriente (Rural) Costa vs Oriente (Rural) 0% 20% 40% 60% 80% 100% Returns Characteristics 6. Conclusions Overall, the results of the study indicate that at the national level, urban-rural differences in welfare ratios are mainly due to characteristics, which account for almost three-quarters of the difference. Comparisons in living standards within a given region across urban and rural areas reveal that in the Sierra and Costa regions, characteristics are also the main source of differences in welfare. In the Oriente, both characteristics and returns are important to explain the welfare level differences between urban and rural areas. The empirical results indicate that among the characteristics analyzed, education was the variable with the largest contribution to explain the differential in mean log welfare ratio between urban and rural areas, as was expected. Other non-geographic characteristics also contribute to the difference, but to a lesser extent. According to the results obtained from both urban and rural areas in the simulated welfare ratios and the simulated poverty rates within each province, differences in living standards between urban and rural areas can be attributed to disparities in non-geographic variables and to differences in the returns to those characteristics. This indicates that poor areas are poor not only because households with observable non- geographic characteristics that favor poverty are settled there, but because there are differences in the returns to those characteristics by location. Comparisons in living standards between a leading and a lagging region reveal that for urban areas, the differences between Oriente and Costa and between Sierra and Costa are due to returns. In the case of rural areas, the differentials between Costa and Oriente are explained by characteristics, while the differences between Sierra and Oriente are explained by both returns and characteristics. The findings suggest that government programs increasing the portable human capital endowments of poor people can be a major step forward towards increasing the equality of opportunities for welfare. The analysis also reveals that returns between regions may be substantially different, which suggests that migration is not equalizing returns. To accelerate this process, governments can try to eliminate impediments to labor mobility and facilitate the movement of people towards areas of economic opportunity. Finally, there may be cases of poor and relatively isolated regions inhabited by a significant number of people who for cultural reasons are more attached to land and thus less eager to migrate to 17 areas of better economic opportunities. In specific circumstances, territorial interventions aimed at turning around the economy of a region may be the best approach to improve the livelihoods of the poor. REFERENCES Bourguignon, Francois, Francisco Ferreira, and Phillippe Leite. 2002. "Beyond Oaxaca-Blinder: Accounting for Differences in Household Income Distributions across Countries." World Bank Policy Research Working Paper 2828. Jalan, Jyotsna and Martin Ravallion (2002) "Geographic Poverty Traps? A Micro Model of consumption Growth in Rural China," Journal of Applied Econometrics, 17: pp. 329-346. Kanbur, Ravi. 2006. "The policy significance of inequality decompositions" Journal of Economic Inequality, 4(3): pp 367-374. Nguyen, T.; J. Albrecht; S. Vroman; and M. Westbroo (2007): "A quantile regression decomposition of urban­rural inequality in Vietnam". In: Journal of Development Economics 83 (2007) 466­490. Ravallion, M. and Q. Woodon (1999): "Poor areas or only poor people? Journal of Regional Science", Vol 39, No. 4, 1999, pp. World Bank (2005) "Ecuador Poverty Assessment". Mimeo World Bank (2008) "Regional Study on Welfare Disparities Across and Within Regions in LAC". Office of the Chief Economist. 18 Annex Table A1. Means/Percentages and Standard Errors Urban Rural National Std. Std. Std. Mean / Err. of Mean / Err. of Mean / Err. of Percentage Mean / Percentage Mean / Percentage Mean / Variable Percent Percent Percent Province Azuay 4.0% 0.2% 7.0% 1.2% 4.9% 0.4% Bolívar 0.5% 0.2% 3.0% 0.6% 1.3% 0.2% Cañar 0.9% 0.2% 3.1% 0.6% 1.6% 0.3% Carchi 1.0% 0.2% 1.9% 0.5% 1.3% 0.2% Cotopaxi 1.1% 0.3% 6.3% 1.0% 2.7% 0.4% Chimborazo 2.0% 0.4% 6.2% 1.1% 3.3% 0.5% El Oro 5.3% 0.7% 3.7% 0.9% 4.8% 0.5% Esmeraldas 2.5% 0.4% 3.9% 0.8% 3.0% 0.4% Guayas 33.5% 1.3% 11.7% 2.4% 26.7% 1.2% Imbabura 2.8% 0.5% 3.1% 0.8% 2.9% 0.4% Loja 2.1% 0.4% 5.6% 1.1% 3.2% 0.4% Los Ríos 4.6% 0.7% 7.5% 1.4% 5.5% 0.7% Manabí 7.7% 1.1% 12.1% 2.0% 9.1% 1.0% Morona Santiago 0.4% 0.1% 2.1% 0.7% 0.9% 0.2% Napo 0.3% 0.1% 0.9% 0.4% 0.5% 0.2% Pastaza 0.5% 0.2% 1.1% 0.7% 0.7% 0.2% Tungurahua 2.7% 0.5% 6.2% 1.0% 3.8% 0.5% Zamora 0.2% 0.1% 1.1% 0.4% 0.5% 0.2% Sucumbíos 0.5% 0.2% 1.9% 0.6% 1.0% 0.2% Orellana 0.4% 0.2% 1.0% 0.4% 0.6% 0.2% Pichincha 26.8% 1.3% 10.4% 2.2% 21.7% 1.1% Demographics Number of babies 0.2 0.0 0.3 0.0 0.3 0.0 Number of babies squared 0.3 0.0 0.4 0.0 0.3 0.0 Number of children 0.8 0.0 1.1 0.0 0.9 0.0 Number of children squared 1.6 0.0 2.7 0.1 1.9 0.0 Number of teenagers 0.5 0.0 0.6 0.0 0.5 0.0 Number of teenagers squared 0.8 0.0 1.1 0.0 0.9 0.0 Number of adults 2.4 0.0 2.4 0.0 2.4 0.0 Number of adults squared 5.8 0.1 5.5 0.1 5.7 0.1 Female household head 23.3% 0.6% 15.7% 0.7% 20.9% 0.5% Male household head 76.7% 0.6% 84.3% 0.7% 79.1% 0.5% No spouse, single 7.2% 0.3% 5.8% 0.4% 6.8% 0.3% 19 No spouse, separate/divorced/widow 22.8% 0.6% 17.1% 0.6% 21.0% 0.5% No spouse, married 1.6% 0.2% 2.1% 0.2% 1.7% 0.1% Spouse, married 68.4% 0.7% 75.0% 0.8% 70.5% 0.5% Age of household head 45.9 0.2 48.1 0.3 46.6 0.2 Age of household head squared 2346.6 24.2 2585.3 35.7 2421.6 20.2 Mixed race (mestizo) household head 82.0% 0.7% 73.7% 1.6% 79.4% 0.7% White household head 8.7% 0.5% 5.6% 0.5% 7.7% 0.4% Black household head 2.9% 0.3% 3.1% 0.5% 3.0% 0.3% Mixed race (mulato) household head 2.9% 0.3% 1.5% 0.3% 2.5% 0.2% Other ethnicity household head 0.1% 0.0% 0.0% 0.0% 0.05% 0.02% Indigenous 3.4% 0.4% 16.2% 1.5% 7.4% 0.6% Education of Household Head 1 - 4 years 8.6% 0.4% 22.1% 0.8% 12.8% 0.4% 5 - 7 years 28.9% 0.8% 43.2% 1.0% 33.4% 0.6% 8 - 10 years 14.2% 0.5% 7.3% 0.4% 12.0% 0.4% 11 - 13 years 21.2% 0.6% 7.4% 0.5% 16.9% 0.4% 14 - + years 23.5% 1.1% 4.3% 0.6% 17.5% 0.8% None 3.6% 0.3% 15.7% 0.7% 7.4% 0.3% Education of Spouse 1 - 4 years 5.1% 0.3% 13.9% 0.6% 7.8% 0.3% 5 - 7 years 18.0% 0.6% 32.4% 0.9% 22.5% 0.6% 8 - 10 years 10.9% 0.5% 7.6% 0.5% 9.9% 0.4% 11 - 13 years 16.7% 0.5% 5.6% 0.5% 13.2% 0.4% 14 - + years 14.8% 0.7% 2.4% 0.3% 10.9% 0.5% None / No spouse 34.6% 0.7% 38.2% 1.0% 35.7% 0.6% House Ownership Rent 25.0% 0.8% 3.2% 0.5% 18.1% 0.6% Own and paying 3.6% 0.4% 0.8% 0.1% 2.7% 0.3% Other 14.5% 0.5% 19.6% 1.0% 16.1% 0.5% Own and paid 56.9% 0.9% 76.4% 1.1% 63.0% 0.7% Position in Occupation Employer 8.5% 0.4% 2.1% 0.2% 6.5% 0.3% Self-employed 27.8% 0.7% 10.3% 0.8% 22.3% 0.6% Employee without pay 1.5% 0.1% 1.6% 0.2% 1.5% 0.1% Farm labourer 3.3% 0.4% 21.7% 1.2% 9.1% 0.5% Owner farmer 1.0% 0.1% 6.2% 0.6% 2.7% 0.2% 20 Self-employed farm labourer 1.6% 0.2% 32.4% 1.3% 11.3% 0.5% Not working 13.2% 0.5% 6.6% 0.6% 11.2% 0.4% Employee 42.9% 0.8% 19.0% 1.0% 35.4% 0.7% Table A2. Means/Percentages and Standard Errors, Sierra Region Urban Rural Sierra Region Std. Mean / Std. Err. Mean / Std. Err. Mean / Err. of Percentag of Mean Percentag of Mean Percentag Mean / e / Percent e / Percent e Variable Percent Demographics Number of babies 0.2 0.0 0.3 0.0 0.2 0.0 Number of babies squared 0.3 0.0 0.4 0.0 0.3 0.0 Number of children 0.7 0.0 1.0 0.0 0.8 0.0 Number of children squared 1.4 0.1 2.5 0.1 1.8 0.1 Number of teenagers 0.5 0.0 0.6 0.0 0.5 0.0 Number of teenagers squared 0.8 0.0 1.1 0.0 0.9 0.0 Number of adults 2.4 0.0 2.4 0.0 2.4 0.0 Number of adults squared 5.6 0.1 5.1 0.1 5.4 0.1 Female household head 22.6% 1.0% 20.1% 0.9% 21.7% 0.7% Male household head 77.4% 1.0% 79.9% 0.9% 78.3% 0.7% No spouse, single 9.2% 0.6% 7.3% 0.5% 8.5% 0.4% No spouse, separate/divorced/widow 18.9% 0.8% 17.3% 0.8% 18.3% 0.6% No spouse, married 2.0% 0.3% 3.1% 0.4% 2.4% 0.2% Spouse, married 69.9% 1.0% 72.3% 1.0% 70.7% 0.7% Age of household head 46.0 0.4 49.2 0.5 47.1 0.3 Age of household head squared 2358.9 35.9 2700.1 47.5 2480.2 29.3 Mixed race (mestizo) household head 83.7% 1.0% 68.9% 2.2% 78.5% 1.0% White household head 7.3% 0.7% 4.9% 0.5% 6.4% 0.5% Black household head 1.7% 0.3% 1.7% 0.6% 1.7% 0.3% Mixed race (mulato) household head 1.6% 0.3% 1.0% 0.3% 1.4% 0.2% Other ethnicity household head 0.0% 0.0% 0.0% 0.0% 0.00% 0.00% Indigenous 5.7% 0.8% 23.5% 2.2% 12.0% 1.0% Education of Household 21 Head 1 - 4 years 7.0% 0.6% 21.1% 1.0% 12.0% 0.6% 5 - 7 years 28.6% 1.3% 44.1% 1.3% 34.1% 1.0% 8 - 10 years 12.9% 0.7% 6.2% 0.6% 10.5% 0.5% 11 - 13 years 21.0% 0.9% 6.8% 0.6% 16.0% 0.6% 14 - + years 27.5% 1.8% 4.9% 1.0% 19.5% 1.3% None 3.0% 0.4% 16.9% 1.0% 7.9% 0.5% Education of Spouse 1 - 4 years 5.0% 0.5% 14.1% 0.7% 8.3% 0.4% 5 - 7 years 17.0% 0.9% 29.7% 1.1% 21.5% 0.7% 8 - 10 years 11.3% 0.7% 5.8% 0.6% 9.3% 0.5% 11 - 13 years 16.6% 0.8% 4.6% 0.6% 12.3% 0.6% 14 - + years 17.0% 1.1% 3.0% 0.6% 12.0% 0.8% None / No spouse 33.1% 1.0% 42.8% 1.4% 36.5% 0.8% House Ownership Rent 31.6% 1.1% 4.2% 0.8% 21.8% 0.8% Own and paying 4.3% 0.7% 1.1% 0.2% 3.2% 0.5% Other 15.9% 0.8% 20.1% 1.2% 17.4% 0.7% Own and paid 48.2% 1.2% 74.6% 1.4% 57.6% 0.9% Position in Occupation Employer 8.9% 0.6% 2.6% 0.3% 6.6% 0.4% Self-employed 24.4% 0.8% 11.5% 1.0% 19.8% 0.7% Employee without pay 1.3% 0.2% 1.8% 0.2% 1.5% 0.1% Farm labourer 1.6% 0.3% 14.1% 1.2% 6.0% 0.5% Owner farmer 0.9% 0.2% 4.2% 0.6% 2.1% 0.2% Self-employed farm labourer 2.2% 0.3% 35.3% 1.8% 14.0% 0.8% Not working 13.3% 0.8% 6.6% 0.9% 10.9% 0.6% Employee 47.4% 1.1% 24.0% 1.5% 39.1% 0.9% Source: Own calculations based on ECV Survey. 22 Table A3. Means/Percentanges and Standard Errors, Costa Region Urban Rural Costa Region Std. Std. Mean / Mean / Mean / Std. Err. Err. of Err. of Percentag Percentag Percentag of Mean Mean / Mean / e e e / Percent Variable Percent Percent Demographics Number of babies 0.3 0.0 0.3 0.0 0.3 0.0 Number of babies squared 0.3 0.0 0.4 0.0 0.3 0.0 Number of children 0.9 0.0 1.0 0.0 0.9 0.0 Number of children squared 1.8 0.1 2.4 0.1 2.0 0.1 Number of teenagers 0.5 0.0 0.5 0.0 0.5 0.0 Number of teenagers squared 0.9 0.0 1.1 0.1 0.9 0.0 Number of adults 2.5 0.0 2.5 0.0 2.5 0.0 Number of adults squared 6.0 0.1 5.8 0.2 6.0 0.1 Female household head 24.1% 0.8% 10.3% 0.9% 20.7% 0.7% Male household head 75.9% 0.8% 89.7% 0.9% 79.3% 0.7% No spouse, single 5.5% 0.4% 4.1% 0.6% 5.2% 0.4% No spouse, separate/divorced/widow 26.3% 0.9% 17.7% 1.1% 24.1% 0.7% No spouse, married 1.2% 0.2% 0.8% 0.3% 1.1% 0.2% Spouse, married 67.0% 0.9% 77.4% 1.3% 69.6% 0.8% Age of household head 46.1 0.3 47.4 0.6 46.4 0.3 Age of household head squared 2355.0 33.7 2518.6 59.1 2395.9 29.4 Mixed race (mestizo) household head 80.7% 0.9% 85.0% 1.6% 81.8% 0.8% White household head 9.9% 0.7% 6.2% 0.9% 9.0% 0.5% Black household head 4.0% 0.5% 5.6% 1.0% 4.4% 0.4% Mixed race (mulato) household head 3.9% 0.5% 2.5% 0.5% 3.6% 0.4% Other ethnicity household head 0.1% 0.1% 0.0% 0.0% 0.10% 0.05% Indigenous 1.3% 0.3% 0.8% 0.3% 1.1% 0.2% Education of Household Head 1 - 4 years 10.0% 0.6% 25.0% 1.4% 13.8% 0.6% 5 - 7 years 29.1% 1.1% 40.9% 1.5% 32.1% 0.9% 8 - 10 years 15.1% 0.7% 7.7% 0.7% 13.3% 0.6% 11 - 13 years 21.3% 0.8% 7.7% 0.8% 17.9% 0.7% 23 14 - + years 20.2% 1.3% 3.4% 0.6% 16.0% 1.0% None 4.3% 0.4% 15.4% 1.2% 7.0% 0.5% Education of Spouse 1 - 4 years 5.1% 0.4% 13.2% 1.1% 7.2% 0.4% 5 - 7 years 18.6% 0.9% 35.3% 1.7% 22.8% 0.9% 8 - 10 years 10.5% 0.6% 9.7% 1.1% 10.3% 0.5% 11 - 13 years 16.6% 0.8% 6.4% 0.8% 14.1% 0.6% 14 - + years 13.0% 0.9% 1.3% 0.3% 10.1% 0.7% None / No spouse 36.1% 1.0% 34.2% 1.4% 35.6% 0.8% House Ownership Rent 19.2% 1.1% 1.4% 0.4% 14.7% 0.9% Own and paying 3.1% 0.5% 0.4% 0.2% 2.5% 0.4% Other 13.4% 0.7% 20.0% 1.9% 15.1% 0.7% Own and paid 64.3% 1.2% 78.1% 2.0% 67.8% 1.1% Position in Occupation Employer 8.2% 0.6% 1.5% 0.4% 6.5% 0.5% Self-employed 31.1% 1.0% 9.4% 1.4% 25.7% 0.9% Employee without pay 1.6% 0.2% 1.3% 0.3% 1.6% 0.2% Farm labourer 4.7% 0.6% 34.4% 2.1% 12.1% 0.9% Owner farmer 1.2% 0.2% 9.8% 1.2% 3.3% 0.4% Self-employed farm labourer 0.9% 0.2% 24.6% 1.8% 6.9% 0.7% Not working 13.5% 0.7% 7.2% 0.9% 11.9% 0.6% Employee 38.7% 1.1% 11.7% 1.3% 32.0% 1.0% 24 Table A4. Means/Percentanges and Standard Errors, Oriente Region Urban Rural Oriente Region Std. Std. Mean / Mean / Mean / Std. Err. Err. of Err. of Percentag Percentag Percentag of Mean / Mean / Mean / e e e Percent Variable Percent Percent Demographics Number of babies 0.3 0.0 0.5 0.0 0.4 0.0 Number of babies squared 0.3 0.0 0.7 0.1 0.5 0.1 Number of children 0.9 0.1 1.5 0.1 1.3 0.1 Number of children squared 1.8 0.2 4.7 0.5 3.6 0.3 Number of teenagers 0.6 0.1 0.8 0.1 0.7 0.1 Number of teenagers squared 1.0 0.1 1.8 0.2 1.5 0.2 Number of adults 2.1 0.1 2.4 0.1 2.3 0.1 Number of adults squared 4.7 0.3 5.9 0.4 5.4 0.3 Female household head 18.7% 2.1% 12.6% 2.0% 15.0% 1.5% Male household head 81.3% 2.1% 87.4% 2.0% 85.0% 1.5% No spouse, single 9.9% 1.7% 3.7% 0.9% 6.2% 0.9% No spouse, separate/divorced/widow 16.5% 2.0% 13.1% 1.8% 14.4% 1.4% No spouse, married 1.4% 0.5% 1.7% 0.7% 1.5% 0.5% Spouse, married 72.2% 2.5% 81.5% 2.3% 77.9% 1.8% Age of household head 41.6 1.0 44.0 0.9 43.1 0.7 Age of household head squared 1933.6 99.3 2157.1 91.6 2069.5 69.1 Mixed race (mestizo) household head 77.0% 2.1% 50.3% 6.8% 60.8% 4.6% White household head 8.7% 1.8% 6.6% 1.5% 7.4% 1.2% Black household head 2.2% 0.7% 0.2% 0.2% 1.0% 0.3% Mixed race (mulato) household head 2.6% 1.6% 0.8% 0.5% 1.5% 0.7% Other ethnicity household head 0.0% 0.0% 0.0% 0.0% 0.00% 0.00% Indigenous 9.5% 1.7% 42.0% 7.4% 29.3% 5.2% Education of Household Head 1 - 4 years 6.4% 1.2% 14.2% 2.1% 11.1% 1.4% 5 - 7 years 30.3% 2.4% 49.0% 3.4% 41.7% 2.6% 8 - 10 years 15.4% 1.4% 12.1% 1.7% 13.4% 1.2% 11 - 13 years 23.6% 1.7% 9.5% 1.7% 15.0% 1.5% 14 - + years 23.2% 3.0% 5.4% 1.3% 12.4% 1.7% None 1.0% 0.4% 9.9% 1.6% 6.4% 1.1% 25 Education of Spouse 1 - 4 years 4.6% 0.9% 15.1% 2.0% 11.0% 1.5% 5 - 7 years 20.7% 1.9% 35.9% 2.6% 29.9% 2.0% 8 - 10 years 12.9% 2.0% 10.1% 1.7% 11.2% 1.3% 11 - 13 years 20.9% 1.9% 8.2% 1.5% 13.2% 1.4% 14 - + years 12.1% 1.8% 3.5% 1.0% 6.9% 1.0% None / No spouse 28.9% 2.6% 27.1% 2.2% 27.8% 1.7% House Ownership Rent 33.3% 3.4% 5.3% 1.8% 16.2% 2.3% Own and paying 2.4% 0.9% 0.8% 0.4% 1.4% 0.4% Other 14.8% 1.9% 14.0% 2.3% 14.3% 1.6% Own and paid 49.5% 3.3% 80.0% 3.1% 68.0% 2.9% Position in Occupation Employer 9.4% 1.8% 2.3% 0.9% 5.1% 0.9% Self-employed 18.0% 1.5% 6.9% 1.6% 11.2% 1.3% Employee without pay 1.2% 0.7% 1.9% 0.6% 1.7% 0.5% Farm labourer 4.0% 0.9% 10.8% 2.1% 8.2% 1.3% Owner farmer 0.7% 0.4% 2.0% 0.6% 1.5% 0.4% Self-employed farm labourer 5.3% 1.3% 50.8% 5.4% 33.0% 4.5% Not working 6.8% 1.3% 3.8% 1.0% 5.0% 0.8% Employee 54.5% 2.7% 21.4% 3.4% 34.4% 3.2% 26 Table A5. Regressions for Log Welfare Ratio, Sierra Region Urban Sierra Rural Sierra Standard Standard Explanatory variables Coefficient Error Coefficient Error Constant -0.543 * 0.150 -0.453 * 0.130 Demographics Number of babies -0.290 * 0.072 -0.369 * 0.047 Number of babies squared 0.013 0.050 0.060 * 0.025 Number of children -0.286 * 0.024 -0.267 * 0.024 Number of children squared 0.033 * 0.007 0.026 * 0.006 Number of teenagers -0.223 * 0.041 -0.126 * 0.033 Number of teenagers squared 0.019 0.020 0.008 0.012 Number of adults -0.138 * 0.026 -0.058 * 0.026 Number of adults squared 0.006 0.004 0.008 * 0.004 Sex of the head -0.065 0.044 -0.068 0.051 No spouse, single 0.144 0.091 0.092 0.064 No spouse, separate/divorced/widow 0.211 * 0.087 0.251 * 0.057 No spouse, married 0.333 * 0.098 0.479 * 0.082 Age of household head 0.030 * 0.005 0.020 * 0.004 Age of household head squared 0.000 * 0.000 0.000 * 0.000 Mixed race (mestizo) household head 0.103 * 0.049 0.132 * 0.042 White household head 0.212 * 0.069 0.212 * 0.071 Black household head 0.102 0.103 0.026 0.141 Mixed race (mulato) household head 0.183 * 0.078 0.276 ** 0.143 Other ethnicity household head Education of Household Head 1 - 4 years 0.423 * 0.090 0.180 * 0.035 5 - 7 years 0.705 * 0.086 0.352 * 0.041 8 - 10 years 0.843 * 0.084 0.499 * 0.056 11 - 13 years 1.064 * 0.094 0.701 * 0.074 14 - + years 1.479 * 0.100 1.233 * 0.141 Education of Spouse 1 - 4 years -0.014 0.081 0.029 0.038 5 - 7 years 0.060 0.082 0.096 * 0.043 8 - 10 years 0.078 0.085 0.124 * 0.059 11 - 13 years 0.167 * 0.084 0.264 * 0.064 14 - + years 0.364 * 0.089 0.408 * 0.088 House Ownership Rent -0.203 * 0.023 0.050 0.078 Own and paying -0.006 0.051 0.395 * 0.125 Other -0.214 * 0.030 -0.081 * 0.033 Position in Occupation Employer 0.259 * 0.039 0.330 * 0.063 Self-employed -0.010 0.026 -0.011 0.041 Employee without pay 0.075 0.063 -0.207 * 0.071 Farm labourer -0.148 ** 0.083 -0.258 * 0.043 Owner farmer 0.242 * 0.090 0.281 * 0.052 Self-employed farm labourer -0.227 * 0.081 -0.203 * 0.039 Not working -0.032 0.043 -0.255 * 0.058 Source: Own calculations based on ECV Survey. Note: Number of observations: 3780 (urban) and 3436 (rural). R2=0.59 (urban) and 0.48 (rural), * indicates that the coefficient is significant at 5 percent level, and ** at 10 percent level. The base categories are: male household head, married with spouse, indigenous household head, no education of household head, no education of spouse or no spouse, own house and employee. 27 Table A6. Regressions for Log Welfare Ratio, Costa Region Urban Costa Rural Costa Standard Standard Explanatory variables Coefficient Error Coefficient Error Constant -0.308 * 0.133 -0.233 0.151 Demographics Number of babies -0.275 * 0.042 -0.348 * 0.049 Number of babies squared 0.028 0.023 0.069 * 0.025 Number of children -0.271 * 0.017 -0.278 * 0.020 Number of children squared 0.022 * 0.004 0.030 * 0.005 Number of teenagers -0.198 * 0.025 -0.133 * 0.034 Number of teenagers squared 0.010 0.010 0.011 0.010 Number of adults -0.089 * 0.019 -0.079 * 0.028 Number of adults squared 0.004 0.003 0.002 0.004 Sex of the head 0.024 0.036 -0.022 0.062 No spouse, single 0.297 * 0.063 0.150 ** 0.083 No spouse, separate/divorced/widow 0.201 * 0.055 0.171 * 0.054 No spouse, married 0.550 * 0.091 0.647 * 0.183 Age of household head 0.027 * 0.004 0.020 * 0.004 Age of household head squared 0.000 * 0.000 0.000 * 0.000 Mixed race (mestizo) household head -0.153 * 0.074 0.033 0.134 White household head -0.077 0.083 -0.027 0.142 Black household head -0.188 * 0.080 0.021 0.143 Mixed race (mulato) household head -0.199 * 0.084 0.014 0.146 Other ethnicity household head -0.212 0.377 Education of Household Head 1 - 4 years 0.180 * 0.049 0.162 * 0.047 5 - 7 years 0.330 * 0.047 0.218 * 0.043 8 - 10 years 0.447 * 0.052 0.316 * 0.055 11 - 13 years 0.629 * 0.055 0.354 * 0.065 14 - + years 0.939 * 0.061 0.650 * 0.101 Education of Spouse 1 - 4 years 0.002 0.046 0.063 0.053 5 - 7 years 0.140 * 0.047 0.100 * 0.046 8 - 10 years 0.188 * 0.049 0.190 * 0.054 11 - 13 years 0.309 * 0.049 0.188 * 0.082 14 - + years 0.504 * 0.055 0.596 * 0.144 House Ownership Rent -0.044 0.028 0.025 0.102 Own and paying -0.031 0.055 -0.251 0.281 Other -0.097 * 0.027 0.011 0.039 Position in Occupation Employer 0.345 * 0.038 0.250 * 0.113 Self-employed -0.081 * 0.024 -0.123 * 0.059 Employee without pay 0.020 0.071 -0.078 0.113 Farm labourer -0.178 * 0.032 -0.077 0.049 Owner farmer 0.178 * 0.074 0.180 * 0.062 Self-employed farm labourer -0.206 * 0.097 -0.156 * 0.051 Not working -0.071 * 0.036 -0.326 * 0.070 Source: Own calculations based on ECV Survey. Note: Number of observations: 3697 (urban) and 1520 (rural). R2=0.55 (urban) and 0.50 (rural), * indicates that the coefficient is significant at 5 percent level, and ** at 10 percent level. The base cathegories are: male household head, married with spouse, indigenous household head, no education of household head, no education of spouse or no spouse, own house and employee. 28 Table A7. Regressions for Log Welfare Ratio, Oriente Region Urban Oriente Rural Oriente Standard Standard Explanatory variables Coefficient Error Coefficient Error Constant -0.390 0.388 -1.111 * 0.311 Demographics Number of babies -0.540 * 0.159 -0.413 * 0.091 Number of babies squared 0.221 ** 0.128 0.123 * 0.048 Number of children -0.377 * 0.058 -0.289 * 0.042 Number of children squared 0.038 * 0.019 0.020 * 0.008 Number of teenagers -0.225 * 0.081 -0.193 * 0.066 Number of teenagers squared 0.044 0.034 0.016 0.019 Number of adults -0.038 0.050 -0.118 * 0.048 Number of adults squared -0.005 0.007 0.014 * 0.006 Sex of the head -0.124 0.089 -0.089 0.096 No spouse, single 0.931 * 0.307 0.164 0.172 No spouse, separate/divorced/widow 0.820 * 0.293 0.205 0.151 No spouse, married 1.161 * 0.393 0.741 * 0.227 Age of household head 0.020 ** 0.011 0.046 * 0.011 Age of household head squared 0.000 0.000 0.000 * 0.000 Mixed race (mestizo) household head 0.151 ** 0.084 0.569 * 0.072 White household head 0.127 0.113 0.548 * 0.121 Black household head 0.053 0.185 0.646 * 0.146 Mixed race (mulato) household head -0.091 0.127 0.571 * 0.140 Other ethnicity household head Education of Household Head 1 - 4 years 0.037 0.269 0.088 0.109 5 - 7 years 0.108 0.257 0.215 * 0.098 8 - 10 years 0.263 0.258 0.318 * 0.097 11 - 13 years 0.273 0.265 0.509 * 0.124 14 - + years 0.555 * 0.258 0.307 * 0.144 Education of Spouse 1 - 4 years 0.427 ** 0.249 0.074 0.148 5 - 7 years 0.651 * 0.305 0.049 0.123 8 - 10 years 0.701 * 0.279 0.268 * 0.126 11 - 13 years 0.859 * 0.288 0.244 ** 0.132 14 - + years 1.005 * 0.310 0.660 * 0.181 House Ownership Rent -0.138 * 0.054 -0.069 0.094 Own and paying 0.169 0.180 0.000 0.111 Other -0.079 0.082 -0.053 0.083 Position in Occupation Employer 0.245 * 0.069 0.147 0.191 Self-employed -0.126 ** 0.069 -0.024 0.081 Employee without pay -0.011 0.243 -0.464 * 0.170 Farm labourer -0.367 * 0.087 -0.372 * 0.091 Owner farmer 0.412 0.424 0.190 ** 0.098 Self-employed farm labourer -0.421 * 0.134 -0.435 * 0.063 Not working -0.059 0.104 -0.412 * 0.117 Source: Own calculations based on ECV Survey. Note: Number of observations: 473 (urban) and 525 (rural). R2=0.58 (urban) and 0.70 (rural), * indicates that the coefficient is significant at 5 percent level, and ** at 10 percent level. The base cathegories are: male household head, married with spouse, indigenous household head, no education of household head, no education of spouse or no spouse, own house and employee. 29 WB148263 C:\Users\wb148263\Documents\Documents in PREM Computer\Ecuador\Sources of Welfare Disparities in Ecuador Sep 09 2009.docx 9/8/2009 10:04:00 PM 30