THE WORLD BANK EDT 66 Discussion Paper EDUCATION AND TRAINING SERIES Report No. EDT66 Does In-Service Training Affect Self Employed Earnings? The Colombian Case Robin Horn Emmanuel Jimenez March 1987 Education and Training Department Operations Policy Staff The views presented here are those of the author(s), and they should not be interpreted as reflecting those of the World Bank. Discussion Paper Education and Training Series Report No. DDT66 DOES IN-SERVICE TRAINING AFFECT SELF-EMPLOYED EARNINGS?: THE COLOMBIAN CASE Robin Horn and Emmanual Jimenez Research Division Education and Training Department March 1987 The World Bank does not accept responsibility for the views expressed herein, which are those of the author(s) and should not be attributed to the World Bank or its affiliated organizations. The findings, interpretations, and conclusions are the results of research or analysis supported by the Bank; they do not necessarily represent official policy of the Bank. Copyright 0D1987 The International Bank for Reconstruction and Development/ The World Bank Abstract This paper evaluates the effectiveness of Colombia's SENA -- one of the world's largest national job-training systems -- for raising the earnings of self-employed workers, who constitute as much as 30 percent of Colombia's labor force. The study also tests the sensitivity of previous measurements of the SENA effect to the screening hypothesis. The findings show that the earnings differential due to SENA is positive and significant for the self-employed, but that it is only about three-fifths of the differential received by SENA-trained workers in the salaried sectors, indicating that screening plays a part in those sectors. Also, the paper finds that SENA substitutes for schooling among the self-employed. Executive Summary National job-training systems are major providers of labor-force training throughout the Latin American region and absorb a significant portion of public spending. Because of the substantial investments in these programs, a number of studies have been undertaken to assess the cost effectiveness of national job training and to assist policy-makers in setting priorities for human capital investments. One of the largest of these systems, Colombia's SENA, has been the subject of a number of economic evaluations. However, all of the previous evaluations of SENA have focused on the impact of training for workers in the formal, salaried sectors. The present paper contributes to the literature by evaluating SENA's effectiveness for self-employed workers, who constitute as much as 30 percent or more of Colombia's labor force. This analysis also tests the sensitivity of previous measurements of the SENA effect to the screening hypothesis, which stipulates that, when it is costly to monitor workers' job performance, salaried-sector employers may rely upon credentials such as SENA training to set salary levels. The data used in this study are derived from a 1979 national household survey in Colombia, which included a number of items concerning the type and extent of SENA training. These data are analyzed to contrast characteristics of self-employed workers in Colombia with those of workers in other employment sectors. Semi-log earnings functions in a variety of specifications are then estimated to compare returns to education, experience, and SENA training by employment sector. The findings show that the SENA effect is positive and significant for the self-employed in Colombia. An individual with seven years of formal schooling--an incomplete secondary education--and 16 years of work experience (the sample means) would earn about 12 percent more with 4 months or more of SENA training than without SENA training. This differential is only about three-fifths of the earnings differential received by SENA trained workers in the public and private sectors. This gap is one indicator that the measured effect of SENA in the salaried sectors may be subject to screening, although add-itional tests would be needed to ensure the comparability of labor markets across the three employment sectors. The paper also finds that long-course SENA training is complementary to primary education and substitutes for secondary education among salaried private and public sector workers. Among self-employed workers, SENA training substitutes for education at all levels of schooling. Limited data on short course SENA participants reveal that SENA --non-SENA earnings differentials are consistently lower than long course differentials, course differentials, but they are complementary to schooling. INTRODUCTION In-service training programs, which are designed to provide training to individuals who are already part of or about to enter the labor force, constitute a significant portion of human capital investments in developing countries. In Latin America, where a national training system exists in almost every country, public spending on these programs amounts to as much as 0.5 percent of GNP. One of the largest of these programs is the Servicio Nacional de Aprendizaje (SENA) of Colombia. Because of the significant amounts invested in programs such as SENA, economic evaluations have been undertaken to assist policy makers in determining whether in-service training meets its goals and in setting priorities regarding investments in human resources. Many of the evaluations of the benefits of SENA have focused on the impact of in-service training on earnings of salaried workers (Puryear, 1977; Reyes and de Gomez, 1979; Gomez and Libreros, 1984). The most recent study completed (Jimenez, Kugler and Horn, 1986) additionally compared a similar benefit measure with unit cost to obtain an assessment of the net returns to SENA training. Although valid for assessing the private benefits to in-service training for certain groups of the population, this approach has some potential drawbacks, particularly if the results are to be extended to the social level. One problem is that, as in many developing countries, a significant portion of the Colombian labor force is seLf-employed. In fact, recent estimates indicate that one-third of Colombia's urban workforce is self-employed (Psacharopoulos, Arriagada and Ve:Lez, 1987). Consequently, if the impact of training on these workers differs from that on salaried workers, then, the evaluation of SENA is incomplete. - 2 - Another potential criticism of earlier studies is that earnings of salaried workers may not adequately reflect their productivity. There are at least two reasons why this argument may be valid in Colombia. One is that a significant portion of the sample may work in the public sector whose salary scales may be relatively insensitive to market forces. Another is that, like formal schooling, the SENA effect may be contaminated by screening. Such an outcome might result if employers, whether in the public or private sector, rely on credentials such as SENA or formal schooling to set salary levels. This study contributes to the literature by examining the effect of in-service training for self-employed males in Colombia. Aside from being able to obtain an evaluation based on a more representative sample of the Colombian labor force, this study also attempts to control for the screening effect, by comparing the SENA effect for salaried and self-employed workers. This is reasonable, since a critical assumption of the screening hypothesis is that it is very costly (and in the extreme, impossible) for an employer to monitor a worker's productivity. We would expect the self-employed sector to be relatively free of such screening effects. The next section describes the data and the sample characteristics. A considerable amount of information can be obtained by simple comparisons of the subsamples by sector of employment. This is followed by the section on the measurement of the SENA effect, separated by sector of employment. -3 - DATA AND BASIC CHARACTERISTICS The data are obtained from the March 1979 Colombian National Household Survey (Encuesta Nacional de Hogares), designed and implemented by DANE, the National Department of Statistics (Departmento Administrativo Nacional de Estadistica). The DANE surveys are collected yearly to obtain information on labor force characteristics from random household sampling. The 1979 survey was administered in seven of the largest cities in Colombia: Bogota, Baranquilla, Cali, Medellin, Bucaramanga, Manizales, and Pasto. It included a section of 14 supplementary survey items which examined SENA participation history, if any.1 For each respondent who participated in SENA, the following data were collected: the number of SENA courses taken in total, the number of months of SENA training in total, the individual's activity immediately prior to the first SENA course, the most important SENA course (in terms of duration or usefulness), the most recent SENA course (if different from the most important SENA course), and the location, year, hours-per-day, and number of months of both the most important and the most recent SENA course. In order to simplify our analysis, we limit our study to males who are not domestic workers.2 Table 1 contains a description of the basic data used in this analysis. These data are partitioned by employment sector. The largest sector is the private, salaried sector. The size of the firm was not examined in the survey so there is no way of determining whether any of the employees worked in small (ie., informal sector) enterprises. A small proportion of the sample reported that their main job was family work, but these workers were excluded from consideration in this paper.3 - 4 - The group we shall refer to as the "self-employed sector" consists of individuals who reported they were self-employed (the survey item was trabajador por cuenta propia), as well as those who reported that they employed others or were owners of businesses. Of the males, those who worked for themselves represented over 80 percent of this self-employed sector (less than 20 percent were employers). Age-, Table 1 indicates that, on average, self-employed male workers are older than male workers in other employment sectors. The highest proportion of private sector employees is 20-24 years of age; public sector employees, 30-34 years; and the self-employed group, 40-44 years. These figures basically support Bourguignon and Kugler's (1978) finding that, in Colombia in 1974, workers in the traditional sector were significantly older than the average working population. That paper pointed out that this was to be expected since "traditional" activities would more likely be performed by the older generation. However, the data do not support Bourguignon and Kugler's secondary finding of a high proportion of young people working in that sector -- no doubt because their definition of a traditional sector included wage-earners in small firms and excluded all individuals who had ever enrolled in post-secondary schooling. Instead, the distribution of the labor force by age across the employment sectors shows that members of the younger cohorts are far more likely to be found working for private firms than for themselves or in the public sector. The older an individual is, the more likely he is to be self-employed. Individuals over the age of 47 are in fact more likely to be self-employed than employed by the private wage sector. This finding corroborates the results of a recent study in - 5 - Malaysia which found strong evidence that "upward mobility evidently involves leaving wage employment in order to form one's own business" (Blau, 1986, p. 843). Education: One common claim of researchers studying the informal sector or traditional sector employment is that workers in these sectors tend to have particularly low levels of education compared to formal or modern sector employees (Hallak and Caillods, 1981, pp.56-68; PREALC, 1978, passim; King, 1980, pp. 219 & passim; Sethuraman, 1981, passim; Sethuraman, 1976, p.135; Bromley, 1979, p.1178; PREALC, 1978). Hallak and Caillods, for example, infer from the relatively low levels of education found among workers in the informal sector that there is some kind of a causal link from education to occupational sector: "It is as if, unable to find employment in the modern sector, the least-educated, the illiterates and school dropouts had no alternative but to seek employment in the traditional sector," (Hallak and Caillods, 1981, p. 56). We can use the Colombian data to see if such characteristics of the so-called informal sector hold for the self-employed. The distribution of education level within each employment sector in Colombia shows that the education level for the self-employed tends to be lower than the other sectors. However, the difference in the educational attainments of wage versus self-employment is not significant: the proportion of self-employed individuals who have ever attended secondary school is nine points below that of wage-employed individuals, and the mean education level for self-employed males is only three months below that of the wage-employed males. Moreover, while at every education level a higher proportion of men are found working in the private wage-sector than in the self-employed sector, the relative difference - 6 - between the proportions in each of these sectors is the least at the lowest and at the highest levels of education. What is particularly interesting here is the relatively large proportion of post-secondary educated persons in the self-employed sector. This runs counter to the existing literature, which emphasizes that workers with low levels of education are relatively more likely to be found outside the formal employment sector. The composition and definition of the employment sector is probably critical in this divergence with the existing literature on the informal sector. The self-employed, even in developing countries, do not necessarily share characteristics that many authors ascribe to the so-called "informal" or traditional sector. Migration: Another popular hypothesis concerning traditional employment is that migrants find their first work in that sector. The likelihood of working in the traditional sector is higher, the briefer the person's residence in the place of new residence; and the longer the person's stay in the new town, the greater is the likelihood of transferring to the formal sector (Hallak and Caillods, 1981). This hypothesis has been challenged empirically by some recent studies which show that traditional sector workers were no more likely to 4 be recent migrants than modern sector employees . Our results for the self-employed sector also support this challenge. Table 1 shows that for men, the proportion of migrants in the self-employed sector (71 percent) is higher than it is in the wage employed sector (61 percent). However, the table also indicates that, among the male self-employed who have migrated, the majority have been residing in the new area more than 20 years, whereas among the male wage-employed migrants, the majority have been here less than 20 years. Thus, the self-employed tend to be relatively well-settled. - 7- Earnings and economic activities: The issue of earnings is somewhat complicated by differences in how earnings are obtained in the different sectors. In particular, wage-employees simply receive periodic pay-checks, whereas self-employed individuals and small-businessmen receive revenues but also have to bear the cost of certain capital goods such as tools, workshops, street-vendor stands, and shops. In the survey, the main earnings question for public and private wage-sector employees and workers was "How much do you normally earn in all of your jobs, including tips, gratuities, and commissions but excluding daily expense allowances and payment in kind?." A second survey question asked the respondent to estimate the value of room, board, and clothing benefits received from his employer. We use the sum of these two categories as our earnings variable. No items addressed the issue of other job benefits characteristic of the private wage-sector, such as holiday leave or health and retirement benefits. Studies of the Colombian modern sector have found that the value of these benefits amount to approximately 25 percent of regular earnings (Bourguignon and Kugler, 1979). Nonetheless, the figures reported in the table below do not reflect any additional job benefits received by workers who said they were wage- or public-sector employed. Individuals who said that they were self-employed, owners, or employers were asked a different question: "What were your earnings from your business or profession, net of expenses?" Table 1 indicates that earnings for public sector employees exceeds wage and self-employed earnings. Interestingly, mean self-employed earnings are 40 percent higher than mean earnings for wage employed. Even if an additional 25 percent were added to wage-sector earnings to approximate the value of employee benefit packages, as discussed above, self-employed earnings would still outstrip wage-employed earnings. - 8 - The table also reports the standard deviation of the earnings distribution and the median earnings. The standard deviation of earnings for the male self-employed sector is twice that of male wage-employed earnings. It is also much larger than mean earnings (which does not include values below 0). This, along with the fact that the median income is so much lower than the mean (40 percent) and that there are no negative earnings reported in the sample, indicates that the self-employed earnings distribution is a considerably more right-skewed than the earnings distributions in the other sectors. This finding is further confirmation of our earlier supposition that self-employed workers do not necessarily belong to the "informal sector" category. Most studies comparing earnings in the formal and traditional sectors emphasize the lower earnings received by traditional sector workers. The PREALC studies of Gran Asuncion (1978, p.148) and San Salvador (1978, pp. 187-190), reveal significantly lower incomes for workers in the informal sector--even when domestic employees are excluded from the informal sector sample. However studies that focus on self-employed workers, small entrepreneurs, or skilled craftsmen working on their own account report that earnings for these individuals are higher. An analysis of a longitudinal sample of self-employed and wage-employed individuals in peninsular Malaysia demonstrated that earnings for the urban self-employed are consistently above employee earnings (Blau, 1986). A study of small scale entrepreneurs in Bangkok found that the self-employed have much better incomes than wage-sector employees (Larsson, 1980). The higher variation in earnings for the self-employed reflect the diversity in work intensity, as well as the nature of the occupations in that sector. For example, according to Table 1, there are large differences in length of the work-week within each of the employment - 9 - sectors. Although the self-employed do not work substantially more hours than the wage employed, the higher proportion of part-time workers in the self-employed sector indicates that there is also a considerable number of workers in this sector who work well above the mean. Similar findings have been found for Tanzania, Senegal (Hallak and Caillods, 1978, p.72) and earlier Colombian sample (Bourguignon and Kugler, 1978). In addition, the nature of the occupational activities in the self-employed subsectors is much more diverse than in the salaried sector. Among the males, category 3 (shopkeepers, store proprietors and vendors) makes up one-third of the self-employed sector. This is five times the size of the same category in the private wage-sector. On the other hand, there are almost no self-employed who are working in occupational category 2 (clerical, communication, and transportation workers). This group constitutes 12 percent of the private wage-sector and 21 percent of the public sector. SENA and the self-employed: Table 2 shows the proportion of individuals in each sector which has received any SENA training. The proportion of the self-employed men who have taken SENA courses is roughly half the proportion in the private wage-sector, and only about one-third the proportion in the public sector. The small proportion of SENA trainees in the self-employed sector is not surprising since the institution itself has been geared to the wage sector. Its original responsibility was to analyze and meet vocational training needs in the modern sector, for which purpose most of its training centers as well as its training methodology and instructor training were designed (SIENA, 1978, p.106). It was not until the 1970s that the SENA training (primarily the short courses) began to be directed to the informal or traditional sectors. Nevertheless, it is apparent that SENA does serve the self-employed subsector, despite the original design. - 10 - This table also provides a description of the kinds of SENA training which the individuals in our sample received. Self-employed individuals were more likely to have taken short courses, and the wage employed more likely to have taken long courses. Compared to the private wage-employed, the self-employed are more than twice as likely to take short courses of less than 6 months. In fact, about half of the self-employed group take SENA training for less than one year, and nearly two-thirds for less than one-and-one-half years. Among the private-sector wage-employed, about two-thirds take SENA training for one year or more, and over one-third of these employees take training for two years or more. These results are not surprising since many short-courses were designed for the "traditional sectors." In terms of the content of the training courses, a higher proportion of the self-employed (35.5 percent) than the private wage employed (15.4 percent) pursued electrical, electronic, artisanal, or agricultural technician training (35.5 percent). On the other hand, nearly twice as many private wage employed individuals took SENA training in administrative work or accounting. Surprisingly, a considerably higher proportion of private wage employed men took training in machine or motor mechanics (26.5 percent) compared to the self-employed men (15.3 percent). Gross comparisons of mean levels of education within the employment sectors reveals that SENA may substitute for formal education. Among wage employed individuals, less than half as many SENA-trained men and women have received any post-primary education. Smaller proportions of the SENA-trained self-employed have also received education at every level except primary. We will return to the issue of substitutability later in the paper since the full story turns out to be considerably more complicated than the one told by these gross statistics. - 11 - While most of the SENA-trained male subsample of all three employment sectors had been working prior to their enrollment in their first SENA course, about 30 percent more self-employed than wage employed individuals had been working. Conversely, roughly three times as many private wage sector than self-employed trainees had been students just before they enrolled in SENA. This high proportion of individuals who were working prior to electing SENA training courses gives some additional support to the hypothesis that the decision to obtain additional SENA-type training, unlike formal education, is made after the individual chooses his occupation and employment sector. Furthermore, since the overwhelming majority of the self-employed men were working just prior to SENA training, it is very likely that they enrolled in SENA courses to upgrade their skills in order to become more productive in their business. According to Table 3, the public sector and self-employed sector, SENA mean earnings are below non-SENA mean earnings, while in the private salaried sector alone, SENA mean earnings surpass non-SENA earnings. However, in all three sectors, SENA median earnings are above non-SENA earnings. In all the sectors, and especially in the self-employed sector, the distributions of non-SENA earnings are more right skewed than the SENA earnings distributions and are characterized by more variation. The implication is that individuals with unusually high earnings are less likely to have enrolled in SENA training courses--particularly individuals with high earnings in the self-employed sector. This is not surprising if our hypothesis concerning the training decisions following the occupation and employment sector decision is correct. Average individuals tend to gain the most from education and training investments. In the next - 12 - section, we employ multiple regression analysis to estimate the effects of SENA training on earnings, holding constant age, education, and other control variables. - 13 - MEASURING THE EARNINGS EFFECT OF SENA In order to isolate the effect on attending a SENA program on earnings, it is essential to hold constant for other variables which may affect earnings, such as schooling, experience and socio-economic background. Because of the cross-sectional nature of the data base (i.e., earnings of program participants cannot be observed before and after training), we use estimated earnings functions to control for these variables. This methodology has been used in several previous studies of SENA effects (see, for example, Puryear, 1977; Jimenez, Kugler and Horn, 1986). The basic functional form used in this study is presented as equation (1) below, where ln (Y) is the natural logarithm of the weekly wage; Z is a vector of explanatory variables, with the components, SP, Ss and Sh which correspond respectively, to years of primary, secondary and higher level of schooling, X, X2, and B, which corresponds to a vector of background variables; a, and the vector b are parameters to be estimated; and u is a random error term: ln (Yij) = aij + bij*Zij + uij (1) This equation is estimated separately for each of the "J" employment sectors, where j = 1 (public), 2 (private wage-employed), and 3 (self-employed); and for each of the "i" training categories, where i = 1 (no SENA training), 2 (SENA short-course training) and 3 (SENA long-course training). This is done in order not to impose the prior restriction that each of these groups are necessarily drawn from a single population of workers, all of whom choose one of the sectors and one of the training - 14 - categories. Instead, we test the assumption that may be drawn from different populations, and that different parameters of the earnings functions apply for each of the subgroups. The SENA effect is measured as the difference in average earnings between those who have taken each of the SENA courses (i=2,short; 3,1ong) and those who have not taken any SENA training (i=l), holding constant for the components of Z -- S, X and B. The functional form of equation (1) implies that the SENA effect would differ depending upon the employment sector. For example, for the self-employed sector, the SENA effect of a long course would be: E33 = Y33 - Y13, (2) where Yij is predicted earnings for the ith course and jth sector using the parameters of equation (1). Obviously, differences in the predicted earnings of equation (2) depends, not only on differences in the estimated parameters a and vector b of equation (1), but also upon the levels of Z (i.e., S, X and B) at which equation (1) is evaluated. For example, the estimated SENA effect could differ, depending upon whether we calculated it at the mean levels of Z, for those who are in SENA, as opposed to the mean levels for those out of SENA. In order to see the sensitivity of our results, we calculate the SENA effect at alternative values of the explanatory variables. The estimated earnings functions: Table 4 presents the results of the estimations of earnings functions for each of the subgroups mentioned earlier. Although a fuller investigation is beyond the scope of this paper, a few findings regarding the returns to schooling and the self-employed are of general interest. Let - 15 - us consider only the equations for the general population, which has not taken any SENA courses (columns 1, 3 and 7 of Table 4). The returns to education in Colombia vary by level of schooling. The returns generally rise with the level of education. For example, for the private salaried sector with no SENA training, the rates of return to another year of primary schooling is 6%; for secondary schooling is 12.5%; and for higher-level schooling is 22.4%. These findings are consistent with those of other studies in Colombia (Mohan, 1981; Fields, 1978), as well as with those returns estimated for modern and fast growing economies like Malaysia, Korea and Taiwan (Psacharopoulos, 1985). Within primary and secondary levels of schooling, the returns to education are higher for those in the self-employed sector than for those in the salaried sectors. Thus, for these levels of education, the screening effect does not seem to be important. Indeed, relying only on salaried workers may underestimate the returns to education. However, for higher education, there are important differences. The higher rate of returns in public and private sectors relative to the self-employed sector seems to confirm the screening hypothesis. Moreover, the returns to higher education in the public sector is less than that in the private salaried sector, which implies that public sector employees do not get paid more--and may even have to give up earnings to work there, although they are probably compensated by job security or other non-monetary benefits. In the absence of individual specific background measures (such as parnts education and income levels), we use three background variables, GPOORCITY, GRICHCITY, and GOTHCITY, which represent the location where sample members spent the majority of their schooling years. These variables were constructed from data on average urban housing, education, and income characteristics reported by DANE based upon a statistical survey - 16 - conducted in 1955 in the same seven cities as in our survey. GPOORCITY represents cities which in 1955 scored 55 to 60 percent lower than the average on indexes of housing quality, and education, and family income. GRICHCITY represents cities which scored from 10 to 40 percent higher than the average on these indexes. GOTHCITY represents the remaining major cities. Those who were raised in areas other than these seven cities fall into the omitted category. The negative sign which accompanies the coefficient for GPOORCITY indicates that being schooled in a poor city, which is probably a proxy for low quality schooling, lowers the average earnings in all sectors. GRICHCITY, with two exceptions, has the expected positive sign, and the coefficients of GOTHCITY do not follow any particular pattern, except in the public sector where the sign is always negative. Finally, we use another proxy variable to control for the effect of possible differences in labor market conditions across geographic locations. RICHCITY, a dummy variable equal to 1 for those whose present earnings come from Bogota, Medellin, or Bucaramanga, the cities with the highest average income currently, does not appear to have much impact on individual earnings. This gross indicator reveals, at least at this level of aggregation, that labor markets are not geographically segmented in Colombia. The SENA effect: The other columns of Table 4 show the earnings functions estimated for the SENA subsamples. These are used to compute the SENA effect in terms of percentage differentials in earnings. In order to have a reference base, results from earlier evaluations of SENA are shown in Table 5. Our results, computed from Table 4 using the methodology described above, are presented in Table 6. Unfortunately, the relatively few short - 17 - course participants who are currently in the public and self-employed sectors make the results from these two sub-groups unstable. In particular, we were nol: able to obtain estimates for the coefficients of primary years of schoo:Ling for the public sector short course participants and of years of higher education for the self-employed short course participants. Consequently, we do not estimate earnings effects for these subgroups. Previous studies point, on average, to a significant positive return to SENA in terms of earnings differentials between trainees and non-trainees (Table 5). According to Table 6, we find similar qualitative results, at least for long courses. For example, a primary school graduate (5 years of schooling) with an additional two years of secondary education, about 16 years of experience (the average for the sample), and SENA long-course training would earn 31 percent than a counterpart currently working in the public salaried sector; 32 percent more than a counterpart in the private salaried sector; and 12 percent more than a counterpart in the self-employed sector. For an individual with ten years of experience, and seven years of schooling, the SENA earnings differentials are, respectively, 37 percent, 24 percent, and 23 percent for the public, private and self-employed sectors. These results indicate that there are positive returns to SENA training and that these returns persist even for the self-employed. There has been some debate in the literature as to whether SENA and formal schooling complement or substitute for one another. An early study by Puryear of apprentices indicates that SENA-training is a substitute for formal education. This result is not confirmed for other studies using a more recent data set (Gomez and Libreros, 1984; Jimenez, Kugler and Horn, 1986). Our results indicate that the evidence on - 18 - complementarity depends upon the sector of economic activity, as well as the level of education. We consider SENA training to be complementary to formal education if, at a given level of education, another year of schooling would increase the impact of SENA on earnings. SENA is substitutable for education if another year would decrease its impact. These patterns of complementarity and substitutability are evident from Figure 1, which depict the earnings premium associated with SENA long courses by years of schooling for someone with 16 years of experience. This figure is simply a graphical representation of rows 5 to 10 of Table 6. The results in Figure 1 indicate that, for the salaried public and private sectors, SENA long-course training is complementary to formal schooling at the primary level; but it is substitutable for secondary and higher levels of schooling. In contrast, for the self-employed sector, SENA training is substitutable for formal education at all levels. Thus, contrary to the earlier literature, the substitution between SENA and formal education is a complex interaction. Long-course SENA training is also complementary to work experience if a greater amount of experience is associated with a larger SENA premium. Figure 2 depicts what happens to this earnings differential at various experience levels for someone with seven years of formal schooling. SENA training is complementary to work experience in the private sector, which is not surprising given its in-service nature. However, it is substitutable in the public sector, where seniority would tend to overcome all SENA effects at high experience levels. Surprisingly, it is also substitutable for work experience in the self-employed sector. Short course impacts are much lower than long course effects for the one sector that we have data for--the private salaried sector. In fact, as found in Jimenez et al., 1986 the effects of some of these courses - 19 - found to be negative. There appears to be a high degree of complementarity between formal schooling and short courses. However, these short courses were meant for the informal sector and there could be selection bias in these results. In comparing the SENA effect across sectors, one result that emerges is that the SENA impact is roughly equivalent in the public and private sectors at average years of experience (16 years) for the whole sample and over all years of secondary schooling. This is clearly shown in Figure 1. The implication is that the SENA effect would not be over-inflated in the public sector relative to the private sector due, say, to the possibility that the public sector was not responsive to market forces. This finding implies that criticisms of earlier studies which did not differentiate the SENA effect for public from private sectors may not be warranted. On the other hand, both public sector and private sector SENA effect exceed the SENA effect in the self-employed sector. This is a clear indication that some screening (with respect to SENA training) is taking place in the salaried sectors. Nevertheless, this screening effect does not make the impact of SENA on earnings disappear. _ 20 - CONCLUSIONS The goals of this paper were to characterize the differences between workers in the self-employed sector relative to the public and private salaried sectors in order to gain more understanding of this important segment of the labor force, to estimate the differential impact of SENA on the earnings of workers in each of these sectors, to evaluate the importance of screening for SENA training in the salaried sectors, and to gain new insights into the complicated interaction between schooling and post-school training on earnings. The paper found that self-employed workers differ systematically from public and private salaried workers (and from statistical descriptions of informal sector workers from other studies) in a variety of ways such as age, educational attainment, migration patterns, economic activities, earnings, and the kind of SENA training received. These differences motivated more controlled comparisons of the effect of SENA on earnings by through the estimation of separate earnings functions by employment sector. These estimations demonstrated that long SENA courses exceed short courses in their effectiveness in raising the earnings of workers. It was found that the SENA effect is nearly equivalent for public and private sector workers, implying that the public sector is no less competitive than the salaried sector. Although a lower SENA effect for workers in the self-employed sector suggests that screening for SENA training may occur in the salaried sectors, the persistence of a SENA effect among the self-employed implies that SENA does what it has been expected to do: it raises labor productivity in Colombia. Job training programs have an impact on the earnings of self-employed workers despite the fact that the training was designed exclusively for salaried sector employees. - 21 - The interaction between formal schooling and SENA was found to be quite complicated. For the self-employed there is no question that the benefit of SENA training on productivity is lower the more schooling the worker had received. For the salaried sectors, more schooling increases the SENA effect up until a primary school certificate. Additional years of secondary schooling lower the impact of SENA on earnings. In other words, in the salaried sectors, workers with very little schooling are probably improving their basic skills through SENA training. But by secondary level, SENA training is being taken to make up for an insufficient education--as a substitute for additional years of schooling at this level. And finally, it is useful to point out the main limitations of this study. First, the small numbers of self-employed workers who took SENA short courses--training of four months or less--did not allow for a reliable measure of the impact of this increasingly popular mode of training. Secondly, our findings may be affected by (i) the possible selection into the self-employed sector (although a recent paper by Psacharopoulos, Arriagada, and Velez, 1987, show that this selection bias is unimportant in Colombia) or (ii) the possible selection into SENA (as discussed by Jimenez and Kugler, 1987). While the present dataset is too limited for the analysis of these issues, we do plan to investigate them in the future through the use of a different dataset. - 22 - NOTES: 1. This module was designed by Bernardo Kugler on behalf of SENA. 2. We focus on males because lower labor force participation rates of females complicates the analysis of earnings functions (see Mohan, 1986 for a fuller treatment of female earnings functions in Colombia). We also ignore domestic workers in this analysis. There is strong evidence that in Latin America domestic workers differ systematically from workers in the self-employed sector with regard to sex, age, education, migration status, and duration of employment (see Bourguignon and Kugler, 1978; PREALC, 1978). Another problem with treating domestic workers as self-employed individuals is that they tend to under-report their earnings as many receive significant but unestimatable benefits as payments-in-kind (Bourguignon and Kugler, 1978). Moreover, since there are only 14 men in the survey who report that their work activity is in the domestic services sector, this omission does not pose severe problems. 3. Only 6.5 percent of the self-employed males in our sample lived in households in which there was at least one individual--male or female--who worked as an unpaid family worker, and 85 percent of these individuals were the only unpaid worker in the family. Adjusting self-employed earnings for these 6.5 percent can be shown to be non-significant (see Horn, 1987). 4. Bourguignon and Kugler found that traditional sector workers were no more likely to be recent migrants than modern sector employees. In El Salvador, a higher proportion of the informal sector is composed of migrants, and, excluding domestics, the probability of working in the informal sector was higher the longer the residence in the current location (PREALC, 1978, p.178). In urban Peru, approximately the same proportion of recent migrants (less than 6 years of residence in the new location) are self-employed--l1 percent--as formal sector employed--12 percent (Webb, 1974). 5. Data on the number of courses taken by individuals in each of the employment sectors is not shown in the table since no sector deviates too much from the overall division: About 75 percent of the individuals in all three sectors took only one SENA course, and another 12 percent took two SENA courses. ..head - 23 - Table 1: General Sample Characteristics - Means and Standard Deviations (parentheses) Wage Earners Self- Public Sector Private Sector Employed Age (years) Prop. < 20 2.1 11.5 2.8 Prop. 20-40 77.0 73.9 54.4 Prop. 40-60 20.1 13.9 36.7 Prop. 60+ 0.8 0.8 6.1 Total 100.0 100.0 100.0 Years of education 9.67 6.75 6.48 (4.62) (3.82) (4.27) Educational level: None 0.9 2.9 4.6 Primary (1-5 yrs.) 26.1 45.9 52.1 Secondary (6-11 yrs.) 40.3 40.9 32.1 Higher (more than 11 yrs.) 32.8 40.5 11.2 Total 100.0 100.0 100.0 Percent migrated 71.4 61.0 70.9 Years at present residence: 1-10 14.3 22.0 14.7 11-20 28.7 35.4 27.7 21-30 37.0 27.6 28.4 31+ 20.0 15.0 29.2 Total 100.0 100.0 100.0 Earnings 10,116 6,452 10,604 (9,188) (7,141) (14,869) Coeff. of variation .91 1.11 1.40 Hours worked last week 48.2 50.5 51.60 Percent part-time 4.0 4.1 10.6 Continued.. - 24 - (Continued) Table 1: General Sample Characteristics - Means and Standard Deviations (parentheses) Wage Earners Self- Public Sector Private Sector Employed Occupational group: Prof., sup., managers 40.1 14.5 13.0 Clerk, secretary 20.7 12.0 0.4 Shopkeepers .1 7.3 34.2 Cooks, personal services 18.5 8.0 2.9 Tailors, dressmakers .4 5.5 9.0 Mechanics, construc., produc. 17.4 48.6 35.5 Other 2.8 4.1 5.0 Total 100.0 100.0 100.0 No. of children < 6: None 52.3 49.7 61.8 1 30.8 31.0 23.6 2 14.4 15.1 11.1 3 or more 2.5 4.2 3.5 Total 100.0 100.0 100.0 Wife of head is working: Yes 22.1 16.4 17.7 No 77.9 83.6 82.3 Total 100.0 100.0 100.0 Means Values of Background Variables: GPOORCITY 2.0 1.6 2.2 GRICHCITY 1.8 4.1 3.1 GOTHCITY 33.6 43.5 33.4 GOTHERLOC 62.6 50.8 61.3 Total 100.0 100.0 100.0 RICiICITY 70.7 72.5 72.1 OTHERLOC 29.3 27.5 27.9 Total 100.0 100.0 100.0 - 25 - Table 2: Characteristics of SENA and the Self-Employed Subsaeples Wage Earners Self- Public Sector Private Sector Employed SENA participation 9.1 7.6 3.5 OE which: Short courses 48.0 37.6 59.3 Long courses 52.0 62.4 40.7 Total 100.0 100.0 100.0 Distribution of type of course: Agricultural technician 1.4 2.5 8.5 Administration or accounting 30.6 18.3 10.2 Office technician 5.6 3.6 3.4 Textile work or tailoring 0.0 0.7 3.4 Food processing, preparation, or catering 0.0 2.9 3.4 Wood or metal work 4.1 8.6 5.1 Machine or motor mechanic 19.4 26.5 15.3 Electricity or electronics 18.1 11.1 18.6 Construction 2.8 3.9 0.0 Graphic Arts 1.4 1.8 1.6 Sales technician 1.4 5.0 6.8 Artesanal technician 4.2 1.8 8.4 Other 11.0 13.3 15.3 Total 100.0 100.0 100.0 Education prior to SENA: Primary (1-5 yrs.) 12.0 16.8 41.7 Secondary (6-11 yrs.) 58.7 71.2 48.3 Tech/Voc/Norm. 6.7 3.5 6.7 Higher 22.7 8.4 3.3 Total 100.0 100.0 100.0 Activity prior to SENA: Student 17.6 28.8 11.7 Working 78.4 64.6 83.3 Looking for work 1.4 3.5 1.7 Other 2.6 3.1 3.3 Total 100.0 100.0 100.0 Years since SENA course 7.3 6.8 8.2 (4.2) (4.3) (4.9) - 26 - Table 3: Distribution of Earnings, Age, & Education by Employment Sector, Nales only Public Wage Self- inu Sector Sector EAloXe Earnings (Monthly, 1979 Pesos) Mean 8,871 6,979 9,530 Standard Dev. 5,833 5,768 9,882 Median 7,000 5,000 7,000 Age (years) 33 29 34 Education (years) 9.7 8.3 7.1 Non-SEN Earnings (Monthly, 1979 Pesos) Mean 10,241 6,409 10,643 Standard Dev. 9,453 7,241 15,017 Median 6,810 4,000 6,000 Age (years) 37 32 41 Education (years) 9.7 6.6 6.5 - 27 - Table 4: Earnings Functions with LNWAGLST as dependent variable, by sector and length of SENA program. Public Sector Private Salaried Sector Self-Employed Sector no SENA Short Long no SENA Short Long no SENA Short Long Programam PoroProgi Pr amr Program Program Program PRYR 0.068 -- 0.108 0.060 0.047 0.165 0.103 -0.010 0.058 (2.50) (1.25) (7.08) (0.23) (1.80) (6.04) (0.06) (0.23) SEYR 0.121 0.154 0.094 0.125 0.158 0.100 0.171 0.249 0.135 (11.38) (3.03) (2.55) (24.47) (3.75) (5.74) (14.59) (2.36) (2.03) HIYR 0.173 0.322 0.164 0.224 0.266 0.201 0.150 -- 0.368 (15.77) (3.58) (3.55) (25.27) (3.95) (4.63) (9.07) -- (2.24) EXPER 0.045 0.054 0.048 0.059 0.094 0.092 0.067 0.177 0.045 (7.25) (1.28) (1.85) (21.07) (3.33) (6.79) (9.73) (2.51) (0.64) EXPERSQ -0.0006 -0.0007 -0.0010 -0.0009 -0.0011 -0.0018 -0.0009 -0.0033 -0.0006 (4.80) (0.82) (1.57) (15.82) (1.77) (5.13) (7.80) (2.46) (0.41) GRICHCITY -0.159 -- 0.000 0.127 0.763 -0.288 0.210 0.273 0.414 (1.12) -- 0.00 (2.49) (2.29) (1.17) (1.62) (0.49) (0.43) GPOORCITY -0.053 -- -0.487 -0.266 -- -0.776 -0.314 -0.983 -- (0.36) (1.40) (3.51) (2.55) (2.11) (1.20) -- GOTHCITY -0.003 -0.953 -0.034 -0.010 0.319 -0.051 0.106 -0.205 -0.126 (0.07) (3.71) (0.28) (0.50) (1.56) (0.72) (2.27) (0.55) (0.42) RICHCITY -0.029 0.588 -0.091 0.003 -0.059 -0.040 0.033 0.201 0.052 (0.63) (2.13) (0.67) (0.11) (0.31) (0.52) (0.64) (0.53) (0.18) INTERCEPT 2.075 2.016 2.242 1.884 1.292 1.364 1.526 0.928 2.211 (14.10) (3.26) (4.61) (36.43) (1.24) (2.87) (12.01) (0.87) (1.50) R-SQUARED 55.2 81.2 42.7 47.2 67.3 36.7 36.7 55.3 32.8 SAMPLE SIZE 741 15 60 3435 40 243 1606 19 41 NOTE: t-ratios in parentheses below coefficients. - 28 - Table 5: Returns to SENA -- Previous Research Study Data Coverage Method Control % Gain in Income Puryear 1972 Bogota male Regression Random sam- All 61.6 SENA grads. coeff. of SENA ple of Bo- Yrs. of ed. of appren- dummy var. in gota males 5-6 113.8 ticeship semilog ear- aged 20-9 7-9 33.6 1965-67 nings func. 10-11 39.1 Reyes & 1975 Technical Regression Labor force All tech. ed. Gomez education coeff. of participants Males 21.4 graduates technical ed. who did not Females 25.6 (nationwide) dummy in semi- graduate log earnings from tech. Tech. ed. > 3 function education months duration Males 20.6 Females 4.6 Gomez & 1981 All SENA Regression w/ Non-SENA % of % gain Libreros grads. coeff. of SENA workers in grads income dummy var. in similar po- 4.4 20-33 semilog earn- sitions and 26.1 1-20 ings functions firms as 66.3 0 estimated for SENA grads. 3.2 < 0 each occupa- using SENA- tional cate- Holanda data gory Jimenez, 1986 Bogota male Semi-log earn- Non-SENA Estimates of Priv. Kugler, & SENA grads. ings functions workers in Internal Rate of Horn overall SENA estimated for similar po- Return & by long, each program: sitions and Overall 36% short, and SENA dummy and firms as Long 41% five types interaction SENA grads. Short < 0% of SENA effect terms. using SENA- Apprentices 32% programs. Estim. costs Holanda data Promotionals 61% and benefits. Complementary 54% Internal rate Qualifying 168% of return anyl. at mean educ. level. Note: For the Puryear and Reyes & Gomez figures, the percentage gain in income is obtained from the following equation: [exp(b)] - 1, where b is the regression coefficient of the SENA participation dummy variable in the semi-logarithmic earnings function. - 29 - Table 6: The SENA Effect by Course Type, Education, and Experience Earnings Differential (%) Years of Education by Level -------------------------- Type of --------------------------- Public Private Self- Program Primary Secondary Hiiher Sector Sector Emploved Long 10 4 0 0 39.84 17.18 42.73 5 0 0 45.62 30.93 33.31 5 2 0 37.21 23.74 22.95 5 6 0 22.39 11.08 5.51 5 6 2 20.13 -13.97 -2.02 16 3 0 0 33.19 24.77 27.64 5 0 0 38.71 39.19 21.52 5 2 0 30.88 31.65 12.42 5 6 0 16.95 18.26 -3.21 5 6 2 14.86 -8.31 -10.04 20 3 0 0 26.95 25.31 21.63 5 0 0 32.23 39.73 15.86 5 2 0 24.85 32.23 7.32 5 6 0 11.67 18.86 -7.48 5 6 2 9.72 -7.80 -13.99 Short* 13 3 0 0 -- -25.23 5 0 0 -- -28.36 -- 5 2 0 -- -19.06 -- 5 6 0 -- -3.29 -- 5 6 2 -- -18.09 -- 18 3 0 0 -- -7.50 -- 5 0 0 -- -10.30 -- 5 2 0 -- -2.71 -- 5 6 0 -- 10.50 -- 5 6 2 -- -2.38 -- 23 3 0 0 -- 6.09 -- 5 0 0 -- 3.59 -- 5 2 0 -- 10.02 -- 5 6 0 -- 21.39 -- 5 6 2 -- 10.05 -- Because there were few short course participants in the Public and Self- Employed Sectors, SENA differentials were not computed for these groups. - 30 - The SENA Effect by Years of Schooling At 16 Year of Work Experience 40 30- 20 - ~ ~ ~ ~ ~s Wri ic Sector> -10- -20- L a 0 I -10 * 0~ 3 4 5 6 7 8 9 10 11 12 13 Total Years of Schooling FIGURE 1 - 31 - Impact of Experience on SENA Training At 7 Years of Formal Schooling 60 . 50 - 1X 40 Public Sector Oc 30 I o 20- ii 10 , Private Sector _ a- f 0 -10 1 0 1 2 3 4 5 6 7 8 9101 1i2131415161718192021 22232425 Total Years of Expedence FIGURE 2 - 32 - BIBLIOGRAPHY Blau, D.M. "Self-Employment, Earnings, and Mobility in Peninsular Malaysia," World Development Report, July, 1986: 839-852. Bourguignon, F. "The Role of Education in the Urban Labor-Market During the Process of Development: The Case of Colombia," a paper presented at the 6th World Congress of the International Economic Association, Mexico City, August, 1980. Bourguignon, F., and Kugler, B. "Earnings Functions and Earnings Differentials between Modern and Traditional Urban Sectors: The Case of Colombia," Laboratoire D'Economie Politique (CNRS), Ecole Normale Superieure, Paris: October, 1978 Bromley, R. editor. (1979). The Urban Informal Sector: Critical Perspectives on Employment and Housing Policies. Oxford: Pergamon Press. Fields, G. (1980). How Segmented in the Colombian Labor Market, The World Bank, Staff Working Paper No. 434, October, 1980.. Gomez Buendia, H., and Libreros Davila, E. (1984). Formacion Prefesional y Mercados de Trabajo,: in Education Formacion Profesional y Empleo. Bogota: Ministerio de Trabajo y Seguridad Social, Servicio Nacional de Aprendizaje (SENA) pp. 251-268. Hallak, J. and Caillods, F. (1981). Education, Training and the Traditional Sector, Fundamentals of Educational Planning 31, Paris: Unesco International Institute for Educational Planning. Jimenez, E. and Kugler, B. (1987) "The Earnings Impact of Training Duration in a Developing Country: An Ordered Probit Selection Model of Colombia's Servicio Nacionale de Aprendizaie (SENA)," in Journal of Human Resources, Spring, 1987. Jimenez, E., Kugler, B, and Horn, R. (1986). An Economic Evaluation of a National Job Training System: Colombia's Servicio Nacional de Aprendizaje (SENA), The World Bank, Education and Training Department Discussion Paper Series, Report No. EDT24. Washington: The World Bank. King, K. (1980). "Education and Self-Employment," in Education, Work. and Employment; Vol. 2. Paris: UNESCO. Larsson, J. Skill Acquisition by the Self-Employed: A Case Study of Small-scale Entrepreneurs in Bangkok, Asian Regional Skill Development Programme, Islamabad: ILO, 1980 Mohan, R. (1984). "Poverty, Income and Employment in a Developing Metropolis: Consequences of Growth in Bogota and Cali, Colombia," an unpublished paper. Washington: The World Bank. - 33 - Merrick, T.W."Employment and Earnings in the Informal Sector in Brazil: The Case of Belo Horizonte," Journal of Developing Areas, April, 1976: pp.337-354. PREALC (1978). Sector Informal: Funcionamiento y Politicas. Santiago: Programa Regional del Empleo para America Latina y el Caribe (PREALC), ILO. Psacharopoulos, G. (1985). "Returns to Education: A Further International Update and Implications, in Journal of Human Resources, Fall, 1985, pp. 583-604. Psacharopoulos, G., Arriagada, A. and Velez, E. (forthcoming). Earnings and Education among the Self-Employed in Colombia, The World Bank, Education and Training Department Discussion Paper Series. Washington: The World Bank. Puryear, J. (1979) "Vocational Training and Earnings in Colombia: Does a SENA Effect Exist?" in Comparative Education Review, June, 1979, pp. 283-292. Sethuraman, S.V., editor (1981). The Urban Informal Sector in Developing Countries: Employment, Poverty, and Environment. Geneva: International Labor Office. Reyes Posada, A. and M. Isabel de Gomez (1979). El Papel y las Posibilidades de la Educacion Tecnica: El Caso de Colombia. Bogota: Corporacion Centro Regional de Poblacion, Area Socioeconomica.