Wage determinants and school attainment among men in Peru

LSM - 38 LSII"h.S JUNE 1987 Living Standards Measurement Study Working Paper No. 38 Wage Determinants and School Attainment Among Men in Peru  LSMS WORKING PAPER SERIES No. 1. Living Standards Surveys in Developing Countries. No. 2. Poverty and Living Standards in Asia: An Overview of the Main Results and Lessons of Selected Household Surveys. No. 3. Measuring Levels of Living in Latin America: An Overview of Main Problems. No. 4. Towards More Effective Measurement of Levels of Living", and "Review of Work of the United Nations Statistical Office (UNSO) Related to Statistics of Levels of Living. No. 5. Conducting Surveys in Developing Countries: Practical Problems and Experience in Brazil, Malaysia, and the Philippines. No. 6. Household Survey Experience in Africa. No. 7. Measurement of Welfare: Theory and Practical Guidelines. No. 8. Employment Data for the Measurement of Living Standards. No. 9. Income and Expenditure Surveys in Developing Countries: Sample Design and Execution. No. 10. Reflections on the LSMS Group Meeting. No. 11. Three Essays on a Sri Lanka Household Survey. No. 12. The ECIEL Study of Household Income and Consumption in Urban Latin America: An Analytical History. No. 13. Nutrition and Health Status Indicators: Suggestions for Surveys of the Standard of Living in Developing Countries. No. 14. Child Schooling and the Measurement of Living Standards. No. 15. Measuring Health as a Component of Living Standards. No. 16. Procedures for Collecting and Analyzing Mortality Data in LSMS. No. 17. The Labor Market and Social Accounting: A Framework of Data Presentation. No. 18. Time Use Data and the Living Standards Measurement Study. (List continues on the inside of the back cover)  LSMS Working Paper Number 38 Wage Determinants and School Attainment OAmong Men in Peru Morton Stelcner Ana-Maria Arriagada Peter Moock Development Research Department The World Bank Washington, D.C. 20433, U.S.A.  - ii - This is a working document published informally by the Development Research Department of The World Bank. The World Bank does not accept responsiblity for the views expressed herein, which are those of the authors and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported by the Bank; they do not necessarily represent official policy of the Bank. The designations employed, the presentation of materials, and any maps used in this document are solely for the convenience of the reader and do not imply the expression of any opinion whatsoever on the part of the World Bank or its affiliates concerning the *legal status of any country, territory, city, area, or of its authorities, or concerning the delimitation of its boundaries, or national affiliation. The LSMS working paper series may be obtained from the Living Standards Measurement Study, Development Research Department, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A. Morton Stelcner is a Consultant in the Living Standards Unit of the Development Research Department and Associate Professor, Department of Economics, Concordia University, Montreal, Quebec. Ana-Maria Arriagada is a Researcher in the Education and Training Department. Peter Moock is a Senior Economist in the Education and Training Department. June 1987  - iii - LIVING STANDARDS MEASUREMENT STUDY The Living Standards Measurement Study (LSMS) was established by the World Bank in 1980 to explore ways of improving the type and quality of household data collected by Third World statistical offices. Its goal is to foster increased use of household data as a basis for policy decision making. Specifically, the LSMS is working to develop new methods to monitor progress in raising levels of living, to identify the consequences for households of past and proposed government policies, and to improve communications between survey statisticians, analysts, and policy makers. The LSMS Working Paper series was started to disseminate intermediate products from the LSMS. Publications in the series include critical surveys covering different aspects of the LSMS data collection program and reports on improved methodologies for using Living Standards Survey (LSS) data. More recent publications recommend specific survey, questionnaire and data processing designs, and demonstrate the breadth of policy analysis that can be carried out using LSS data.  - iv - ACKNOWLEDGENTS The helpful comments of Dennis de Tray and the competent research assistance of Martin McCambridge and Gail Spence are gratefully acknowledged. We also thank Carmen Martinez for her patience and expertise in typing various parts of the manuscript. We are particularly indebted to Jacques van der Gaag for many valuable suggestions that shaped the final version of this paper. The authors are entirely responsible for the contents of this study.  -v - ABSTRACT This paper is the first of a series that assesses the impact of education on labor market outcomes in Peru using data from the Peruvian Living Standards Survey that was conducted between June 1985 and July 1986. The present study concentrates on factors that affect wages and school attainment of male wage and salary earners. Particular attention is given to assessing the effects of formal schooling and parental education on wages, and to the effects of primary school quality and parental education on school attainment. The analysis (1) presents estimates of rates of return to schooling, (2) assesses the effects of parental education on wages and school attainment, and (3) examines regional differences in wage structures. We also explore the impact of non-market forces on pay structures by considering sector of employment (public vs private) and the effects of firm size and unionization. The main findings are as follows. Formal schooling plays an important role in explaining wage variations, and the pattern of rates of return reflects that found in most developing countries. The estimated magnitudes are similar to those found in other Latin American countries: 10% for (a year of) primary schooling, 6% for secondary schooling, and 8% for post-secondary schooling. The analysis also suggests that there are differences in wage structures among metropolitan Lima, other urban areas, and rural areas. The impact of post-secondary schooling and vocational training are strong in Lima and rural areas, but not in other urban areas. We also find that, when parental schooling effects are excluded, having attended a public school has a significant negative impact on wages in Lima and rural areas only. As regards the effects of parental education on wages, the results suggest that these are not as strong as may have been expected in a society that is often characterized as "socially stratified", and the effects are largely confined to other urban areas, that is, the point estimates of returns to schooling are reduced. In Lima parental schooling mitigates the negative effects of public education, but they remain strong in rural areas. With respect to school attainment, we find that, although parental education does have a positive and significant impact, the effects diminish as cohorts get younger and are reduced when primary school quality indicators are included among the regressors, especially in rural areas. In general, we find that the primary school quality variables contribute significantly to educational attainment and their effects are stronger as the degree of urbanization decreases.  - vi - TABLE OF CONTENTS I. Introductiou ...... ......... ............ ..................... .................. I II. Statistical Models, Data and Variables ................................. lO III. Regressions Results ................. ........... ...... . 24 Private Sector i) Country-Wide and Regional Returns to Schooling*................. 27 ii) Parental Schooling Effects on Wages and Educational Attainment ..... 29 iii) Cohort Effects . .......................................... . 34 iv) Marital Status, Migration, Union and Firm Size Effects ............. 37 v) Returns to Schooling in Peru and Other Latin American Countries .... 37 Public Sector ........................................................ .......... 38 IV. Concluding Remarks ..................................................... 41 REFERENCES ................................................................. 44 APPENDICES .................... so ... .*.*...... *es........ ......... .......... 46 Appendix A Wage Regressions With Controls for: 1) Marital Status and Migration Table Al 2) Unionization Table A2 3) Firm Size Table A3 4) Firm Size and Unionization Table A4 Appendix B Wage Regressions by Cohort Tables B1-B3 Appendix C School Attainment Regressions, All Sectors Tables C1-C2 Correlations of Son's and Parents' Schooling Tables C3-C4 Appendix D Wage Regressions by Firm Size Tables D1-D5 Appendix E Wage Regressions by Unionization Tables E1-E6  I. INTRODUCTION Education is a major consumer of government budgets in both developing and industrialized countries. Policy makers have placed considerable emphasis on human capital investments, particularly education, as a means of improving the quality of human resources for economic growth and of altering income distribution. A vast number of studies, based on the framework developed by Becker (1964) and Mincer (1958,1974), has shown that there is a positive relationship between human' capital accumulation, especially schooling, and earnings. Estimates of rates of return to schooling have often been used to advocate increased public provision of educational services. 11 In this paper we estimate wage functions on a sample of Peruvian male employees (wage and salary earners) using a recent (1985/86) micro data base. 2/ Particular attention is given to assessing the impact of education on earnings, i.e. on private rates of return to schooling. Why do we need another empirical investigation of the relationship between education and earnings in a developing country? There are several reasons. Perhaps the most important reason is that since the early 1960s the Peruvian education system has experienced dramatic growth in enrollments, and in spending on education. 3i The following brief overview of the growth of See Psacharopoulos (1985) for a comprehensive survey of rates of return to education in various countries. Conspicuous by their absence in this survey of more than 60 countries are estimates of "private" (Mincer-type) rates of return for Peru. 2 This paper is the first of a series that assesses the impact of education on labor market outcomes in Peru. Future studies will consider female workers and the self-employed. 3/ Details of the development of Peru's education system and data for the 1960s are provided by Hay (1976) and by Drysdale and Myers (1975).  - 2 - the education system distinguishes among three time-periods: -/ the 1960s, the 1970s, and 1980-85. For the period 1960-70, total school enrollments grew at an annual rate of 7.3%, twice the population growth rate. Primary school enrollments rose from 1.5 million in 1960 to 2.5 million in 1970; secondary school enrollments from 198,000 to 550,000 (674,000 if one includes evening programs); higher non-university enrollments from 4,000 to 24,000; and those for universities from 31,000 to 110,000. Although most of the expansion was met by public schools, private sector education also increased considerably. As a proportion of enrollments, the private sector comprised about 14% at the primary level in both 1960 and 1970; at the secondary level it accounted for 29% in 1960 and 16% in 1970; and at the university level for 10% and 22% for 1960 and 1970, respectively. In absolute terms, the greatest increase was at the secondary level. The expansion in enrollments was mirrored by growth in educational spending, reflecting to a large extent the policies of the Belaunde Government (1963-68). School expenditures, as a percentage of GDP, rose from 2.6% in 1960, to 5.0% in 1965, and then declined to 3.7% in 1970. The share of education in the government budget rose from 20% to 26% between 1960 and 1965 and then fell to 19% in 1970. / These periods correspond to three political eras in Peru: the first Belaunde Government, June, 1963 - October, 1968; the Revolutionary Government of the Armed Forces, October, 1968 - July, 1980, (during which there were two military presidents, General Velasco, 1968-75 and General Morales-Bermudez, 1975-80); and the second Belaunde Government, July, 1980 - July, 1985. On July 28, 1985, the Government of President Alan Gracia took office.  - 3 - In October 1968, the Revolutionary Government of the Armed Forces assumed power and remained in.office until mid 1980. One of its first policy initiatives was to reform the educational system as part of a larger process of the "structural transformation" of Peruvian society. In 1972, after three years of study and discussion, the Government embarked on a comprehensive reform of its education system. This reform, which was in large part completed by the mid-1980s, revamped the structure of the education system, l/ altered curricula, emphasized the education of school dropouts, and attempted to increase access in the poorest areas of the country. In addition to these initiatives, this Government also stressed the promotion of non-formal and out-of-school education, i.e., vocational and technical training 2/, literacy l The new structure was phased in between 1972 and 1985. The old system consisted of six years of primary schooling, five years of secondary schooling, and four to five years of higher education. Under the new system there are nine years of "basic" education, followed by 3 stages of "higher" education: (i) a first cycle of three to four years, Escuela Superior de Educacion Profesional, which stresses vocational curricula, (ii) a university cycle of four years and (iii) a final cycle of graduate studies (see Hay 1976). 2/ These non-formal education programs are offered by both the public and private sectors. As regards government-sponsored programs, by 1977, the Ministry of Education had 403 Centros Educativos de Calificacion Profesional Extraordinaria (CECAPES). The Ministry of Labor offered 6 training programs, and the Ministry of Health one training program and agricultural training was provided by CENCIRA (Centro Nacional de Cooperacion e Investigacion para la Reforma Agraria). Training programs and courses offered by the private sector or state enterprises include: industrial training by SENATI (Servicio Nacional de Adiestramiento Industrial), electricity training by CEFOCAP (Centro de Formacion y Capacitacion de Personal Electro-Peru) construction training by SENCICO (Servicio Nacional de Capacitacion para la Industria de la Construccion), transportation training by INICTEL (Instituto Nacional de Investigacion y Capacitacion de Telecomunicaciones), training in tourism by CENFOTUR (Centro de Formacion en Turismo), and handicraft training by EPPA-PERU (Empresa Publica de Promocion Artesanal).  - 4 - programs, and extension services. These policy measures have been accompanied by substantial increases in enrollments at all levels, and have continued into the 1980s. Primary school enrollments increased from 2.5 million in 1970 to 3.2 million in 1980, and to 3.6 million in 1984, resulting in a net enrollment rate for 6-11 year olds of 90%. Secondary school enrollments rose from 550,000 in 1970 to 1.2 million in 1980, and reached 1.6 million in 1984 with a net enrollment rate of 50%. Enrollments in higher non-university institutions were 24,000, 56,000, and 100,000 for 1970, 1980, and 1984, respectively, and university enrollments were 109,000, 248,000, and 334,000 for the three years.-1 As was the case during the 1960s, the private sector has played an important role. In the mid 1980s, private schools accounted for about 15% of primary/secondary enrollments, and about 25% of university enrollments. Given these dramatic increases in enrollment rates, it is important to assess whether investment in education continues to be warranted on the basis of private returns to this investment. Furthermore, we will analyze whether the returns for younger and generally better educated cohorts are as high as for older cohorts that have less school attainment. That is, we consider whether the increased supply of human capital has kept up, fell short of, or exceeded the increased demand associated with continuing economic development. Although school enrollments increased dramatically, the severe recession and rapid inflation experienced by Peru since the mid 1970s has resulted in substantial cutbacks in government spending on social services. Data for the 1970s and 1980s are provided in INE (1986) and the World Bank (1983, 1985).  - 5 - Spending on these items accounted for about one-third of the total budget in the early 1970s, falling to less than one-fifth in the mid 1980s. The education component has been the most severely reduced. As a fraction of total central government expenditures, the Ministry of Education budget fell from 19% in 1970 to 12.8% in 1980 and to 9.6% in 1985. As a proportion of GDP, it declined from 3.7% in 1970 to 3.1% in 1980, and to 2.2% in 1984. Coupled with the considerable enrollment increases over the period 1960-85, the severe reductions in education spending could lead to a serious deterioration in school quality and/or reductions in educational attainment. This is the second reason why Peru is a timely case for assessing returns to schooling. If returns to schooling are found to be at least as high as in countries at a similar stage of development, there is a strong case for arguing that continued reductions in spending on education may have serious adverse effects on long-term development. The results presented here will provide some insight into the potential effects of cutbacks in education spending on future productivity. A third reason why Peru is an interesting case for analysis is that its striking geographical and ecological differences translate into considerable regional economic diversity, This diversity is further complicated by a large variety of ethnic/linguistic categories, e.g., Misti, Campesinos, Cholos; Aymara/Quechua- and Spanish-speakers (see Nyrop, 1981 and Weeks-Vagliani, 1985). As suggested in the recent literature (Heckman and Hotz, 1986; Birdsall and Behrman, 1984), returns to. schooling can vary significantly by region, especially if regional differences are pronounced. The extent of regional differences in returns to schooling has direct  consequences for various public policies, for example, migration, spatial distribution of educational resources, and regional economic expansion. Finally, it is curious to note that, given the growth of education over the last 25 years, there are surprisingly few Mincerian descriptions of income determination for Peru. The few that do exist are based on data for the late 1960s and early 1970s and confined to Lima only (see Flores 1980, Musgrove 1982, Suarez 1986, Toledo 1984). 1' Thus, this study fills a "research gap", and will allow an up-to-date comparison of Peru with other countries. This study also responds to two criticisms that are sometimes levied against the conventional methodology used to estimate returns to schooling. Both criticisms imply that standard estimates of returns to schooling may be biased upward, but for somewhat different reasons. The first, and more familiar criticism, maintains that the upward bias is due to failure to control for variables such as innate ability, motivation, and home environment. Of particular concern in this regard is the influence of family background during childhood and adolescence both on the acquisition of schooling and on earnings outcomes. The role of family influences has been the subject of a large and growing body of research, particularly in industrialized countries. Although the importance of family effects is generally recognized, their magnitude and the manner in which they / Some results of these studies are difficult to,interpret because the estimates of returns to schooling appear to be implausible. In some instances they are negative; in others they are unusually high, for example, 60%-117%.  are transmitted are not clearly understood. (For recent literature, see Kearl and Pope, 1986). Little attention, however, has been given to this question in the context of developing countries until recently (see: for Nepal, Jamison and Moock, 1984; for Nicaragua, Behrman and Wolfe, 1984; for Panama, Heckman and Hotz, 1986; for Kenya and Tanzania, Armitage and Sabot, 1983). An examination of the role of family background in earnings determination is important for at least two reasons. First, to what extent does failure to account for family effects (and other factors, such as innate ability) bias estimates of returns to education? 11 If the bias involved is large and to the extent that these estimates are used to justify increased spending on education (as a means of achieving economic growth), the resource misallocations are apparent. Second, if the objective of public spending on education is to equalize income distribution, are the intended effects neutralized, or does such spending in fact worsen the distribution? That is to say, do children from "good" home environments benefit proportionately more (in terms of labor market outcomes) from increases in government spending on education than children from unfavorable family circumstances? Analogously, would reductions in public subsidization of schooling harm children from privileged backgrounds more than other children? / Family background variables such as parental education proxy a variety of considerations that influence educational attainment and labor market outcomes: for example, genetic endowments, home investments during childhood (time spent with the child, parental encouragement, availability of learning materials in the home), economic resources that allow investments in formal schooling, and parental "tastes". If these factors, as captured by observable proxy variables, are strongly correlated with educational attainments, then the standard regression of earnings on schooling will yield upwardly biased estimates of returns to schooling.  -8- Estimates of returns to schooling may also be overstated because of what Birdsall and Behrman (1984) call "geographical aggregation bias" effects. In countries with pronounced regional differences, country-wide estimates of returns to schooling may be biased upward for a variety of reasons including spatial differences in cost of living, failure to take account of migration patterns and costs, school quality differences, and differences in regional labor and product markets. Using the 1970 census of Brazil, Birdsall and Behrman (1984) found that geographical aggregation bias is large and that returns to education do not equilibrate across regions. Similar conclusions are reached by Behrman, Wolfe and Blau (1985) using 1937-78 survey data for Nicaragua and by Heckman and Hotz (1986) using 1983 survey data for Panama. The regional diversity of Peru suggests that "geographical aggregation bias" effects may be important. The main objectives of this study are thus : (1) to present estimates of returns to education comparable to those in other countries, (2) to assess regional differences in rates of return to schooling and (3) to examine the effects of parental education on the earnings of their offspring. Furthermore, we will explore the impact of non-market forces on pay structures by comparing wage structures in the public and private sectors and by considering the effects of unionization and firm size on wages. The analysis does not use theoretical sophistications nor "powerful" estimation methods. Rather, we opt for the standard human capital framework (with some extensions) and ordinary least-squares procedures. In the spirit of Occam's razor, we believe that it is useful to start with a straightforward approach to understanding unknown complex phenomena. A simple methodology often suffices to detect statistical patterns/regularities that are useful for  - 9 - policy formulation. The "baseline" regressions reported here will suggest directions for future research and will allow (future) comparison of rates of return to education of those not employed as wage earners. Since schooling, as expected, turns out to be an important determinant of wage variations, we augment the investigation by an analysis of school attainment in which we assess the roles of parental education and primary school quality (books, furniture, food, teachers). The paper proceeds as follows: Section II briefly reviews the statistical model used and describes the data and the variables. The estimates are presented and evaluated in Section III. Section IV contains the conclusions, a discussion of the policy implications and future research plans.  - 10 - II. STATISTICAL MODELS, DATA AND VARIABLES Our analysis of wage determination is based on the widely used model developed by Becker (1964) and Mincer (1958, 1974). The basic postulate is that variations in wages arise from differences in investments in human capital, in particular schooling and post-school work experience. The wage function yielded by this model can be augmented to include other socioeconomic factors thought to be correlated with wages. The equations estimated in this paper are of the form: (1) lnW A0 + A1S + A2X + A X2 + Z'B + u where lnW is the natural log of hourly wages (cash as well as inkind), S is years of formal schooling, X and X2 are respectively post-school work experience and experience squared, Z is a vector of socioeconomic characteristics, and u is an error term assumed to have the usual properties. We choose to estimate a wage-rate function (where the dependent variable is the natural log of the hourly wage rate) 1/ rather than an earnings function (where the dependent variable is the natural log of annual earnings). As discussed in detail by Blinder (1976) and Blomquist (1978), the 1/ In the data base workers reported: (1) the nominal value (Intis) of each of the remuneration components (cash, bonuses, housing, food, clothing, transportation, etc.) and the periodicity of payment, each of which were used to obtain real monthly earnings using region-specific consumer price indices, June 1985 =100; (2) the number of months worked in the past year; and (3) the usual weekly hours worked and the actual hours worked in the past. week. ( The Pearson correlation between usual and actual hours was 0.95). The regressions based on usual hours did not differ from those based on actual hours worked.. The real hourly wage rate (RHW) was calculated as follows: RHW = (AC)/ (AH), where: AC = annual compensation = monthly pay x months worked in the past year, AH = annual hours = weekly hours x months worked in the past year x 4.33.  - 11 - earnings function is a hybrid that may confound wage rate -differences and issues related to the amount of labor supplied (earnings = wage rates x hours worked), and may thus bias the estimates of returns to human capital traits. An implicit, but dubious, assumption in the earnings function approach is that the amount of labor supplied does not vary with wage rates. The set of regressors includes the standard human capital variables (education and experience) and is expanded to include job-specific work experience, vocational training, type of school last attended (public vs private), diplomas held, parental education, current place of residence, marital status, and migrant status. 11 Also, since wage rates are likely to be influenced by work environment and institutional factors, we examine the effects of unionization, firm size and sector of employment (government versus private sector). In order to facilitate comparability of our estimates with those of other studies for LDCs we perform the regressions using ordinary least squares. 2/ For the most part, the additional explanatory variables are standard and require little elaboration. The variable for type of school last attended is used to proxy school quality. Current place of residence (Lima, other urban areas, rural areas) is included to control for differences in labor market conditions, cost of living, and other pecuniary and nonpecuniary factors (e.g. "tastes", amenities). The marital status variable serves to capture the effects of "job commitment" in that, because of family obligations, married men have a stronger attachment to their job or are so regarded by employers. 2/ Future research will incorporate procedures to mitigate self-selection bias, e.g. wage earner versus self-employed, public sector versus private sector workers, migrants versus non-migrants.  - 12 - As regards the analysis of schooling attainment, we also use ordinary least squares. The dependent variable is total years of schooling completed and the explanatory variables are each parent's years of schooling, a dummy variable indicating whether the person lived with his parents at age 10, indicators of the quality of the primary school attended (number of teachers, provision of food, availability of books and furniture), and type of birthplace (rural, town, and city). The data used in this study are drawn from an unusually comprehensive and "clean" microdata set developed jointly by the World Bank and the Peruvian Instituto Nacional de Estadistica (INE). The Peruvian Living Standards Survey (PLSS) provides detailed socioeconomic information on over 5,000 households and 27,000 individuals as well as labor force activities information on about 22,000 persons 6 years of age and older. The PLSS was conducted between June 1985 and July 1986 (for details, see Grootaert and Arriagada, 1986). The analysis is confined to male wage and salary earners who were over 14 years of age and who reported positive remuneration and pcsitive hours worked in their main occupation during the week prior to the enumeration. Excluded from the sample are females, the self-employed, urpaid family workers, and domestics. 11 The sample consists of 2,269 men. This group accounts for about 40% of the male labor force (15 years of age and older) and about 55% of males (of the corresponding age group) who received some form of compensation from either self-employment activities or as wage earners. In / Women dre excluded from this analysis because of the well-recognized problems in properly specifying their wage function. The self-employed are excluded because of the difficulty in disentangling returns to physical capital and human capital. Future research will focus on these groups.  - 13 - the analysis public administration workers comprise civilians employed by the government and members of the military. Private sector workers are individuals employed by state enterprises, private firms, cooperatives, or in private homes (but not as domestic workers). Table A displays the definitions and measurement of the variables used in the analysis of wage determinants and Table B presents the mean characteristics of the sample of male wage earners. Table C presents the definitions of the variables used in the analysis of school attainment, Table D shows their mean values by region and Table E by cohort. A Brief Profile of Male Wage Earners The summary statistics presented in Table B reveal that there are consideraole differences between public and private sectors workers and among workers in Lima, other urban areas (OUAs) and rural areas within the private sector. 2/ We highlight these as a prelude to the regression results. The first feature which stands out is the difference between public and private sector workers. The former, comprising about 25% of all wage earners, receive higher wage rates, are older, have more years of job- specific experience and tend to be married. Differences between the two groups of workers are more pronounced if one compares education and training State enterprise employees are included in the private sector because their pay scales are not determined by the government. 2/ Lima includes the greater metropolitan area as well as the constitutional province of Callao (the main port of Lima). OUAs are towns and cities with a population of 2,000 or more. Rural areas comprise villages of less than 2,000 and agricultural communities. These definitions are based on information provided by INE.  - 14 - characteristics. Average years of schooling in the public sector is 10.6 compared to 7.4 in the private sector. More importantly, 43% of public sector workers have post-secondary education, compared to 15% of private sector workers, and almost 31% of public sector workers have diplomas versus 9% in the private sector. Over 40% of public sector workers reported that they took training courses, compared to 26% of private sector workers. We also see that there are differences in parental education. Parental schooling levels are higher among public sector workers than among the private sector workers. Finally, we note that 55% of public sector workers are unionized versus 26% in the private sector, and that 77% of public sector workers migrated to their current place of residence compared to 67% of private sector workers. The one similarity is that a high proportion of each group reported that the last school attended was public: 89% for government workers and 83% for private sector workers. Thus far we have referred to the private sector as if it were a single homogeneous national wage market. However, the statistics displayed in Table B show that this is not the case. As is readily seen, wage rates and almost all socioeconomic characteristics vary appreciably among metropolitan Lima, other urban areas (OUAs), and rural localities. The pattern of variable means indicates that as the degree of urbanization decreases, wage rates and the "stock" of schooling and training endowments also decline. Lima residents have the highest mean wage rates, followed by those who live in OUAs and rural areas. School attainment, however measured, and parental education are highest in Lima and lowest in rural areas. The average years of schooling of Lima residents is 1.8 years higher than that of residents in OUAs and 4.6 years higher than that of rural residents. The disparity in school attainment among regions increases as the level of education increases. The mean years of  - 15 - primary schooling completed is about the same for residents of, Lima and OUAs (4.8 and 4.6 years, respectively) and 3.4 years for rural residents. As regards secondary schooling, the means are 3.4 years for Lima, 2.4 years for OUAs and only one year for rural residents. At the post-secondary level the divergence is quite pronounced: Lima, 0.9 years, OUAs, 0.4 years, and rural areas, 0.2 years. The schooling advantages of Lima residents are reflected further by the distribution of educational attainment and diplomas held. Just under 20% of Lima residents have completed less than primary schooling versus 40% in OUAs and over 70% in rural areas. About 25% of Lima residents have post- secondary schooling compared to 12.5% in OUAS and 4.3% in rural areas. The proportion of Lima residents who have diplomas (13%) is about twice that in OUAs and three times that in rural areas. As regards parental schooling, the pattern is similar. The mean years of father's schooling in Lima is 5.8 years versus 4.2 in OUAs and 1.8 in rural areas. The corresponding values for mother's education are 4.0, 2.6 and 0.8 years. There are some exceptions to the pattern described above. Job- specific work experience is highest in rural areas followed by OUAs, and is lowest in Lima. The average age is about the same across the regions. In OUAs a higher percentage of the workers (89%) reported that the last school attended was public than in Lima (82%) or in rural areas (78%). The statistics in Table B provide substantial evidence that there are important differences in the characteristics of wage earners among the three regions and between the public and private sectors. In the regression analysis we explore how these attributes affect wage variations in these labor markets.  - 16 - Interesting information is also provided by the means of the school attainment variables (Table D, Table E). For private sector workers, the overall years of schooling obtained is twice as high in Lima than in rural areas (9.2 years vs 4.6 years, Table D). However, the corresponding data on parental education show a much larger gap (mother's schooling: 4.1 years vs 0.9 years; father's schooling: 5.9 years vs 1.8 years), and parental schooling is lower than son's schooling. The disparities in primary school quality are reflected mainly in the availability of math and reading books and in the availability of furniture. Taken together, these data indicate that past education policies have been reasonably successful in increasing the overall education level of the population and in narrowing the gap in scnool attainment between Lima and rural residents. Indeed, in Table E we see that, with the exception of 15-19 year olds, every cohort shows a higher level of education than the previous one. 11 This pattern is also apparent in the data corresponding to parental education and to primary school quality. / The apparent lower education attainment of the youngest cohort may reflect a "school drop-out" effect.  - 17 - Table A: Definitions of Variables Used in the Wage Regression Analysis VARIABLE DEFINITION Dependent Variable LNEWHCM7 Natural log of the real hourly wage rate in main occupation (Intis at June 1985 prices) Experience GEXPR1 Years of potential work experience estimated as: age - 6 - accredited years of schooling - years repeated GEXPR1SQ Years of potential work experience squared. XOCM7 Years of job specific experience in main occupation XOCSQM7 Years of job specific experience squared. Education and Training SPLYRSC1 Years of primary schooling SPLYRSC2 Years of secondary schooling SPLYRSC3 Years of post-secondary schooling TRAIN = 1 if did a vocational training course, 0 otherwise. DIPLOMAI = 1 if has secondary-technical diploma, 0 otherwise. DIPLOMA2 = 1 if has post-secondary non-university diploma, 0 otherwise. DIPLOMA3 = 1 if has university degree, 0 otherwise PUBSCHL = 1 if last school attended was public (state owned and operated), 0 otherwise. Parental Education FYR-SCHL Father's years of schooling. MYR-SCHL Mother's years of schooling. Marital Status MARITALO = 1 if currently married or cohabiting, 0 otherwise Migration RESDIFPL = 1 if ever lived elsewhere other than current residence, O otherwise * The regressions included missing observations designators (dummy variables), but their coefficients are not reported. (Continued)  - 18 - Table A: (Continued) VARIABLE ,DEFINITION Current Place of Residence TYPRES2 = 1 if lives in other urban area, 0 otherwise TYPRES3 = 1 if lives in rural area, 0 otherwise Work Environment LUNIONM7 = 1 if there is a union on the job, 0 otherwise FRM620 = 1 if firm employs 6- 20 workers, 0 otherwise FSIZE5M7* = 1 if firm employs 21- 50 workers, 0 otherwise FRM51200 = 1 if firm employs 51-200 workers, 0 otherwise FRM201ST = 1 if firm employs more than 200 workers or if a state enterprise, 0 otherwise  - 19 - Table B: Mean Characteristics of Wage Earners PRIVATE SECTOR PUBLIC CHARACTERISTICS LIMA OTHER URBAN RURAL ALL PERU SECTOR No. of observations 759 517 467 1743 526 Real hourly wage rate 7.686 5.361 4.021 6.015 7.691 lIntis at June 1985 prices] (15.6) (6.7) (9.1) (12.0) (7.8) Ln real hourly wage rate 1.598 1.236 0.821 1.282 1.814 (0.8) (0.9) (0.9) (0.9) (0.6) Age 33.7 33.7 34.1 33.8 37.5 (12.3) (12.6) (13.6) (12.7) (11.4) Potential work experience 18.1 20.0 23.2 20.0 20.8 (13.4) (13.9) (15.2) (14.2) (12.6) Job specific experience 7.5 8.5 11.6 8.9 10.4 (8.3) (8.9) (11.9) (9.7) (8.7) Years of Schooling - Total 9.2 7.4 4.6 7.4 10.6 (3.4) (3.5) (3.6) (3.9) (4.0) Primary 4.8 4.6 3.4 4.4 4.8 (0.6) (0.9) (1.8) (1.3) (0.7) Secondary 3.5 2.4 1.0 2.5 3.9 (2.0) (2.2) (1.8) (2.3) (1.9) Post-secondary 0.9 0.4 0.2 0.5 1.8 (1.7) (1.3) (0.9) (1.5) (2.4) Level of schooling completed Less Than Primary 0.010 0.012 0.155 0.050 0.010 Primary 0.186 0.396 0.563 0.349 0.152 Secondary Regular 0.513 0.397 0.206 0.395 0.342 Secondary Technical 0.050 0.070 0.032 0.051 0.065 Post-Secondary Non-University 0.050 0.048 0.015 0.040 0.122 University 0.191 0.077 0.028 0.114 0.309 - Accredited. NOTE: The numbers in parentheses are standard deviations. (Continued)  - 20 - Table B: (Continued) PRIVATE SECTOR PUBLIC CHARACTERISTICS LIMA OTHER URBAN RURAL ALL PERU SECTOR Diplomas Obtained Secondary technical 0.022 0.027 0.014 0.021 0.032 Post-Secondary Non-University 0.022 0.015 0.006 0.016 0.085 University 0.084 0.033 0.023 0.052 0.190 School last attended was public 0.824 0.891 0.783 0.833 0.893 Vocational Training 0.370 0.261 0.085 0.262 0.423 Father's years 5.8 4.2 1.8 4.3 5.6 of schooling (4.0) (3.6) (2.5) (3.9) (4.0) Mother's years 4.1 2.6 0.8 2.7 3.9 of schooling (3.6) (3.1) (1.9) (3.3) (3.4) Married or cohabiting 0.548 0.642 0.640 0.600 0.750 Ever lived elsewhere 0.653 0.731 0.612 0.665 0.773 Union in the firm 0.317 0.253 0.173 0.259 0.553 Firm size: 1 - 5 workers 0.248 0.377 0.498 0.353 - 6 - 20 workers 0.233 0.228 0.195 0.221 - 21 - 50 workers 0.126 0.122 0.056 0.106 - 51 - 200 workers 0.158 0.071 0.073 0.110 - 201 or more workers 0.167 0.066 0.152 0.133 - State enterprise 0.067 0.135 0.026 0.076 - Current Place of Residence Metropolitan Lima 0.435 0.483 Other Urban Areas 0.297 0.405 Rural Areas 0.268 0.112  - 21 - Table C: Definitions of Variables Used in School Attainment Regressions Variable Definitions Dependent Variable YRSCHL Son's years of schooling Family Background MYR SCHL* Mother's years of schooling FYR_SCHL* Father's years of schooling FLIVAC10 = 1 if lived with father at age 10 MLIVAC10 = 1 if lived with mother at age 10 School Quality Indicators FURNSCHL = 1 if furniture available at school FOODSCHL = 1 if food provided by school TECHCAT2 =1 if number of teachers was 4-9 TECHCAT3 = 1 if number of teachers was 10 or more BOOKCAT3 = 1 if math and reading books available at school Birthplace TBIRPL3 = 1 if born in town TBIRPL4 = 1 if born in city * The regressions included missing observations designators (4ummy variables) but the coefficients are not reported.  - 22 - Table D: Mean Characteristics, School Attainment Variables by Region Private Sector All Sectors All Peru Lima Other Urban Rural All Peru Lima Other Urban Rural YRSCHL 7.46 9.22 7.46 4.62 8.19 9.67 8.35 5.15 (4.0) (3.4) (3.5) (3.6) (4.2) (3.6) (4.0) (4.0) MYR SCHL 2.76 4.08 2.59 0.86 3.03 4.25 2.86 0.97 (3.3) (3.6) (3.1) (1.9) (3.4) (3.6) (3.2) (1.9) FYR SCHL 4.29 5.88 4.20 1.81 4.61 6.04 4.53 1.98 (3.29) (4.0) (3.6) (2.5) (4.0) (4.0) (3.7) (2.6) FLIVAGIO 0.79 0.82 0.77 0.76 0.80 0.82 0.78 0.78 MLIVAG1O 0.90 0.90 0.90 0.89 0.89 0.89 0.89 0.89 FURNSCHL 0.88 0.94 0.91 0.74 0.89 0.94 0.92 0.75 FOODSCHL 0.32 0.35 0.26 0.32 0.33 0.37 0.28 0.32 TECHCAT2 0.44 0.40 0.58 0.35 0.45 0.40 0.56 0.39 TECHCAT3 0.30 0.48 0.23 0.09 0.32 0.48 0.27 0.09 BOOKCAT3 0.71 0.82 0.72 0.54 0.73 0.83 0.72 0.55 TBIRPL3 0.27 0.30 0.28 0.21 0.28 0.31 0.29 0.23 TBIRPL4 0.42 0.58 0.49 0.10 0.44 0.58 0.49 0.11 N = 1743 759 517 467 2269 1013 730 526 Note: The numbers in parentheses are standard deviations.  - 23 - Table E: Mean Characteristics, School Attainment Variables by Cohort Private Sector All Sectors 15-19 20-on 30-39 40-49 50 + 15-19 20-29 30-39 40-49 50 + YRSCHL 6.57 8.67 8.33 6.05 5.62 6.64 9.08 9.28 7.34 6.53 (2.7) (3.2) (4.2) (4.3) (4.2) (2.7) (3.2) (4.3) (4.7) (4.8) MYR SCHL 2.52 3.23 2.99 1.92 2.40 2.58 3.48 3.30 2.37 2.68 (2.8) (3.5) (3.6) (2.9) (3.2) (2.8) (3.5) (3.6) (3.2) (3.4) FYR SCHL 4.05 4.84 4.57 3.53 3.52 4.07 5.14 5.02 3.92 3.87 (3.4) (3.8) (4.1) (3.6) (4.1) (3.4) (3.9) (4.2) (3.8) (4.2) FLIVAG10 0.83 0.83 0.79 0.69 0.77 0.84 0.84 0.80 0.73 0.77 MLIVAG10 0.94 0.93 0.88 0.83 0.87 0.94 0.93 0.88 0.82 0.86 FURNSCHL 0.90 0.94 0.93 0.78 0.76 0.90 0.95 0.93 0.83 0.80 FOODSCHL 0.50 0.37 0.36 0.19 0.11 0.52 0.38 0.37 0.24 0.12 TECHCAT2 0.47 0.41 0.47 0.43 0.46 0.46 0.41 0.47 0.46 0.47 TECHCAT3 0.38 0.40 0.29 0.20 0.15 0.40 0.42 0.32 0.22 0.17 BOOKCAT3 0.77 0.82 0.72 0.57 0.54 0.78 0.83 0.74 0.62 0.57 TBIRPL3 0.22 0.24 0.28 0.31 0.33 0.21 0.25 0.28 0.34 0.33 TBIRPL4 0.93 0.49 0.41 0.34 0.77 0.50 0.87 0.46 0.35 0.37 N = 206 571 427 287 252 215 715 582 412 345 Note: The numbers in parentheses are standard deviations.  - 24 - III. REGRESSION RESULTS Numerous variancs of the basic wage equation were estimated with OLS procedures. - In general, the estimates were not very sensitive to alternative specifications. Also, we tested for the equality of the full set of coefficients across various subsamples: across regions (Lima, OUAs, and rural areas), between the public administration and the private sector (including state enterprises), between unionized and non-unionized workers, and across firm sizes. The null hypothesis of the equality in the pay structures across these various groupings was almost always strongly rejected using the standard Chow test. The only exception was in the case of public administration workers: the null hypothesis was not rejected between unionized and non-unionized workers and across regions. The results presented in the tables below record estimates for two specifications of the wage function. The first variant (Table 1) restricts the set of explanatory variables to work experience, schooling, diplomas held, training, and type of school attended. The second variant (Table 2) augments this set of regressors by including parental schooling. Additional variants are reported in Appendix A: Table Al adds controls for marital status and migration; Table A2 controls for unionization; Table A3 incorporates firm size variables, but not unionization (private sector only), and Table A4 includes both firm size and unionization variables (private sector only). / For example, we experimented with age rather than the standard proxy for potential work experience, categorical variables for schooling, schooling sqtiared, continuous and spline variables for training hours, whether the individual lived with his parents when he was 10 years old, and various interactions of the explanatory variables.  - 25 - Table 1: Wage Regres5ions for Peruvian Male Workers, 1985/86 - Parental Schooling Excluded Public Sector Private Sector All Peru All Peru Lima Other Urban Rural VARIABLE (1) (2) (3) (4) (5) (6) INTERCEPT 1.0935 0.0144 -0.3334 0,2218 -0.4199 0.0832 (4.93) (0.14) (3.62) (1.08) (1.84) (0.52) GEXPRI 0.0114 0.0446 0.0449 0.0415 0.0596 0.0312 (1.24) (8.05) (8.00) (5.26) (5.28) (2.99) GEXPRISQ -0.0001 -0.0006 -0.0005 -0.0005 -0.0008 -0.0004 (0.22) (5.54) (5.13) (3.34) (4.07) (2.14) XOCM7 0.0176 0.0290 0.0291 0.0293 0.0469 0.0089 (1.60) (4.95) (4.90) (2.92) (3.49) (0.90) XOCSQM7 -0.0005 -0.0007 -0.0007 -0.0006 -0.0012 -0.0003 (1.43) (4.49) (4.73) (1.86) (2.94) (1.23) SPLYRSC1 0.0064 0.1267 0.1606 0.0874 0.1347 0.0885 (0.15) (6.68) (8.79) (2.11) (3.00) (2.96) SPLYRSC2 0.0791 0.0759 0.0944 0.0658 0.0985 0.0673 (4.07) (6.78) (8.60) (4.20) (4.67) (2.75) SPLYRSC3 0.0355 0.1087 0.1130 0.1296 0.0741 0.3113 (1.82) (5.44) (5.59) (5.74) (1.79) (7.42) TRAIN 0.0474 0.2093 0.2395 0.2317 0.0270 0.5641 (0,87) (4.77) (5.42) (4.37) (0.32) (4.23) DIPLOMAI -0.0649 0.0443 -0.0062 -0.0543 0.0689 _ (0.44) (0.36) (0.05) (0.32) (0.30) DIPLOMA2 0.2943 0.2265 0.2030 0.1480 0.3119 _ (2.61) (1.50) (1.32) (0.85) (0.99) DIPLOMA3 0.3325 0.5183 0.4963 0.4064 0.0446 - (2.90) (4.27) (4.05) (3.05) (0.16) PUBSCHL -.0213 -0.2050 -0.2273 -0.1870 -0.1780 -0.2296 (0.25) (3.80) (4.18) (2.73) (1.43) (2.04) TYPRES2 -0.1238 -0.1798 - _ _ _ (2.23) (4.16) TYPRES3 -0.2609 -0.3365 - - _ _ (2.95) (6.53) N 526 1743 1743 759 517 467 ADJ R2 0.25 0.38 0.37 0.40 0.26 0.31 F-STAT 12.02 77.37 84.44 40.86 14.48 22.63 Note: t-values in parentheses.  - 26 - Table 2: Wage Regressions for Peruvian Male Workers, 1985/86 - Parental Schooling IncIuded Public Sector Private Sector All Peru All Peru Lima Other Urban Rural VARIABLE (1) (2) (3) (4) (5) (6) INTERCEPT 1.0329 -0.0980 -0.3790 0.1249 -0.4923 0.0664 (4.63) (0.93) (4.17) (0.61) (2.20) (0.41) GEXPRI 0.0134 0.0466 0.0472 0.0442 0.0611 0.0315 (1.46) (8.50) (8.54) (5.65) (5.50) (3.00) GEXPR1SQ -0.0001 -0.0006 -0.0006 -0.0006 -0.0009 -0.0004 (0.33) (5.90) (5.67) (3.66) (4.32) (2.12) XOCM7 0.0168 0.0286 0.0286 0.0281 0.0507 0.0090 (1.54) (4.92) (4.89) (2.84) (3.84) (0.90) XOCSQM7 -0.0005 -0.0007 -0.0007 -0.0006 -0.0013 -0.0003 (1.42) (4.54) (4.74) (1.89) (3.22) (1.20) SPLYRSCI 0.0019 0.1047 0.1261 0.0609 0.0929 0.0821 (0.04) (5.48) (6.78) (1.47) (2.07) (2.67) SPLYRSC2 0.0718 0.0591 0.0701 0.0479 0.0767 0.0626 (3.63) (5.17) (6.20) (3.01) (3.61) (2.47) SPLYRSC3 0.0332 0.0805 0.0788 0.1112 0.0177 0.2940 (1.70) (3.98) (3.86) (4.93) (0.42) (6.27) TRAIN 0.0528 0.2056 0.2275 0.2339 0.0055 0.5648 (0.97) (4.73) (5.22) (4.47) (0.07) (4.19) DIPLOMAI -0.0306 0.0708 0.0386 -0.0195 0.1321 _ (0.21) (0.57) (0.030) (0.12) (0.59) DIPLOMA2 0.3084 0.3132 0.3119 0.2185 0.4842 _ (2.73) (2.08) (2.06) (1.27) (1.54) DIPLOMA3 0.3074 0.5192 0.5038 0.3810 0.1903 - (2.67) (4.33) (4.17) (2.89) (0.69) PUBSCHL -0.0149 -0.1423 -0.1486 -0.1088 -0.0961 -0.2180 (0.17) (2.63) (2.72) (1.57) (0.78) (1.92) FYR SCHL 0.0033 0.0255 0.0289 0.0290 0.0316 0.0153 (0.37) (3.77) (4.26) (3.57) (2.36) (0.75) MYR SCHL 0.0158 0.0212 0.0260 0.0125 0.0398 0.0042 (0.37) (3.77) (4.26) (1.42) (2.64) (0.16) TYPRES2 -0.1093 -0.1495 - _ _ _ (1.95) (3.48) TYPRES3 -0.2232 -0.2660 _ _ _ (2.46) (5.10) N 526 1743 1743 759 517 467 ADJ R2 0.26 0.40 0.39 0.42 0.29 0.31 F-STAT 9.74 64.06 69.28 33.21 12.97 15.76 Note: t-values in parentheses.  - 27 - PRIVATE SECTOR i) Country-Wide and Regional Returns to Schooling First, we consider the sensitivity of the estimates to regional aggregation. The country-wide results are displayed in Table 1 Columns (2) and (3); those by degree of urbanization in Columns (4)- (6). We note that most coefficients on the human capital variables are estimated with statistical precision (low standard errors) and the pattern of estimates display no major surprises. A glance at the coefficients on the schooling variables reveals their impact on wages is not linear. The rate of return to an additional year of schooling decreases between primary and secondary schooling, and then increases between secondary and post-secondary levels. This pattern of rates of return by level of schooling conforms to that found in most developing countries. 1- The regressions for Peru as a whole, Table 1, Column (2), shows that the rate of return to a year of primary schooling is 0.13, to secondary schooling 0.08 and to post-secondary schooling 0.11. If the control variables for region are omitted (Column 3), the rate of return to primary schooling rises considerably to 0.16 and that to secondary schooling increases somewhat, to 0.09, while that to post-secondary schooling remains unchanged. Although optimistic conclusions about the effects of education on earnings may be drawn from these results, a comparison of Column (1) or (2) and Columns (4) - (6) reveals that nation-wide estimates may be somewhat overstated. Also, the pooled wage equation masks important differences in the - However, this pattern differs from that found in a recent study of the Ivory Coast (see van der Gaag and Vijverberg, 1986).  - 28 - wage structures among the regions and may be misleading because it imposes common slope coefficients among the regions, allowing only for differences in the intercept terms. Although the Chow test showed that each region-specific wage equation differs from the other, this test does not reveal explicitly whether the differences are due to differences in the intercepts, the slope coefficients, or combinations of the two. Hence, we tested for joint effects of region and the other explanatory variables by estimating a model in which the region dummy variables were interacted with each of the explanatory variables (not reported here). We found that only the coefficients of "region by training" and "region by years of post-secondary schooling" were statistically significant. For OUAs, both coefficients were negative (-0.19 for training and -0.097 for post-secondary schooling), while for rural areas both were positive (0.34 and 0.13). These findings suggest that the regional differences in pay structures are mainly due to intercept effects and differences in the effects of vocational training and post-secondary education. The region specific wage equations (columns 4-6) confirm this impression. The rate of return to a year of primary or secondary schooling is roughly the same across regions: Lima (0.09 and 0.07), OUAs (0.13 and 0.10), and rural areas (0.09 and 0.07). Although the point estimates differ, there is no significant statistical difference among these coefficients across regions. However, the return to a year of post-secondary schooling is highest in rural areas (0.31) followed by Lima (0.13) and OUAs (0.07), and these coefficients are statistically different. Similarly, the coefficients on the training variable also differ among regions. It is notable that in OUAs training does  - 29 - not have a significant impact on wages, but there is a strong positive effect of training on wages in both Lima <0.23) and rural areas (0.56). As regards diploma effects, these are not significant in OUAs, while in rural areas the number of individuals with diplomas is too small to make any statistical inferences. In'Lima, having a university degree has a strong impact on wages (0.41); the effects of other diplomas are not statistically significant. The university diploma effect in Lima probably reflects the primacy of Lima as an industrial and financial centre and the concentration of higher education facilities. The country-wide wage regression also shows that the coefficient on type of school last attended (public) is negative (-0.21) and statistically significant. However, in the region-specific wage equations this coefficient is statistically significant only in Lima and rural areas (-0.19 and -0.23). In Lima, 83% of private sector employees last attended public schools; 78% in rural areas, and about 90% in OUAs. These results indicate that the wage-determination process does differ with the degree of urbanization and that there is some evidence of geographical aggregation bias on the estimates of returns to schooling. It would appear that country-wide estimates mask important differences among regions. To avoid tedious repetition the remainder of the discussion is limited to the regional wage functions. ii) Parental Schooling Effects on Wages and Educational Attainment Now we consider the question of whether estimates of returns to schooling are biased upward because of failure to control for family background. The wage regressions are reported in Table 2 and-a summary of the i.  - 30 - Effects of Parental Education On Schooling Coefficients, Private Sector LIMA OUAS RURAL Parental Schooling Excluded Included Excluded Included Excluded Included Years of Primary Schooling .087 .061 .135 .093 .089 .082 Years of Secondary Schooling .066 .048 .100 .077 .067 .063 Years of Post-sec Schooling .130 .111 .074* .018* . .311 .294 'A- Not significant at 10% level. In Table 2 we see that in rural areas parents' education does not have a statistically significant impact on wages. In Lima, only the father's education is statistically significant; an increase of one year of the father's schooling raises the son's earnings by about 3%. In OUAS, both the mother's and the father's education have an impact on the son's earnings, and the effects are about the same (0.040 and 0.032). As seen in the summary table above, the inclusion of parental education variables among the regressors reduces the point estimates of the returns to schooling, especially in OUAs. It is also of some interest that in Lima the negative effect of the public school variable is weakened considerably (-0.11) and becomes insignificant when parental education variables are included in the equation. This suggests that, at least for residents of Lima, parental education has a strong effect on the choice of public or private schools, and these choices affect earnings outcomes. In  - 31 - OUAs, the more pronounced impact of parental schooling may simply reflect home quality effects on earnings since about 90% of OUA wage earners attended public schools. The conclusion which may be reached regarding the family effects on estimates of returns to schooling and on wages is that they are significant, but perhaps not as strong as might have been expected for a Latin American country that is often characterized as being stratified along sociocultural and class lines. Estimates of the returns to schooling after including parental schooling variables remain high. The strong impact of education on wages makes it worthwhile to consider the determinants of schooling attainment. We regressed the son's years of schooling against: 1) each parent's years of schooling 1/ and whether the son lived with his parents at age 10; 2) quality indicators of the primary school attended by the son (availability of books and furniture, provision of food at school, number of teachers); and 3) type of birthplace (rural, town, and city). Two specifications of the equation were estimated: the first restricts the explanatory variables to "parental effects" only. The second adds school quality and type of birthplace variables. This exercise was done across regions (Table 3) and then across age groups (Table 4). Further estimation results and correlations are reported in Appendix C. The correlations between the father's and son's schooling are 0.50 for Lima and OUAs and 0.60 for rural areas. The correlations between mother's and son's schooling are 0.48 for Lima, 0.44 for OUAs, and 0.53 for rural areas. The correlations between father's and mother's years of schooling are: all Peru, 0.69; Lima, 0.62; OUAs, 0.64; and rural areas, 0.68. The correlations for the entire sample, private and public sector workers, are almost identical.  - 32 - Table 3: School Attainment Regressions for Peruvian Male Workers, Private Sector, by Region VARIABLE ALL PERU LIMA OTHER URBAN RURAL INTERCEPT 3.87 0.87 5.74 3.58 4.80 1.90 2.06 0.02 (15.0) (3.0) (14.5) (5.8) (10.3) (3.0) (4.7) (0.0) MYR SCHL 0.32 0.23 0.25 0.20 0.25 0.22 0.47 0.42 (10.6) (8.4) (6.7) (5.4) (4.4) (4.0) (5.1) (5.4) FYR SCHL 0.43 0.28 0.30 0.25 0.34 0.26 0.62 0.33 (16.4) (11.3) (8.9) (7.4) (7.0) (5.4) (8.7) (5.1) FLIVAGIO 0.64 0.29 0.58 0.38 0.59 0.20 0.56 0.16 (3.3) (1.7) (2.0) (1.3) (1.7) (0.6) (1.7) (0.5) MLIVAG10 0.38 0.39 0.35 0.33 0.14 0.12 0.64 0.56 (1.5) (1.6) (0.9) (1.0) (0.3) (0.3) (1.4) (1.5) FURNSCHL 2.03 1.11 1.7 2.28 (8.7) (2.4) (3.6) (7.6) FOODSCHL 0.27 0.26 -0.18 0.63 (1.8) (1.2) (0.6) (2.5) TECHCAT2 1.37 0.79 1.29 1.14 (7.5) (2.3) (3.6) (4.3) TECHCAT3 2.00 1.49 1.46 1.93 (9.1) (4.1) (3.3) (4.4) BOOKCAT3 0.76 0.53 0.72 0.80 (4.6) (1.9) (2.3) (3.1) TBIRPL3 0.59 -0.11 0.43 0.35 (3.2) (0.3) (1.2) (1.3) TBIRPL4 0.90 0.42 0.68 0.26 (4.7) (1.2) (1.9) (0.6) N 1743 1743 759 759 517 517 467 467 ADJ R 2 0.40 0.51 0.30 0.34 0.25 0.32 0.38 0.57 F-STAT 192.2 143.1 54.2 30.9 29.7 19.3 48.6 49.3 Note: t-values in parentheses.  - 33 - Table 4: School Attainment Regressions for Peruvian Male Workers, Private Sector, by Cohort VARIABLE 15-19 20-29 30-39 40-49 50 + - INTERCEPT 4.12 0.89 5.37 2.64 4.08 0.82 3.51 0.84 1.61 -0.27 (5.3) (1.0) (10.8) (4.4) (8.5) (1.3) (6.4) (1.5) (2.7) (0.4) MYR SCHL 0.21 0.16 0.25 0.21 0.34 0.26 0.40 0.27 0.40 0.22 (2.6) (2.2) (5.9) (5.2) (5.8) (4,5) (4.4) (3,4) (4.3) (2.5) FYR SCHL 0.22 0.15 0.30 0.19 0.45 0.34 0.59 0.37 0.50 0.33 (3.5) (2.4) (7.6) (4.9) (8.8) (6.8) (7.8) (5.2) (6.9) (4.8) FLIVAG10 -0.41 -0.51 0.31 0.09 0.82 0.58 0.54 0.25 0.86 0.48 (0.8) (1.1) (1.0) (0.3) (2.1) (1.6) (1.2) (0.6) (1.8) (1.1) MLIVAG10 1.45 1.79 0.80 0.46 0.58 0.62 -0.79 -0.39 0.72 0.94 (1.8) (2.5) (1.6) (1.0) (1.2) (1.4) (1.4) (0.8) (1.1) (1.6) FURNSCHL 1.14 1.44 2.27 1.67 1.50 (2.1) (3.1) (3.8) (3.7) (2.8) FOODSCHL -0.50 0.09 0.38 0.65 -0.30 (1.6) (0.4) (1.3) (1.5) (0.5) TECHCAT2 2.37 1.20 0.94 1.45 1.30 (4.8) (3.9) (2.4) (3.5) (2.9) TECHCAT3 1.54 1.72 1.93 2.85 2.46 (2.6) (5.1) (4.2) (5.0) (3.8) BOOKCAT3 0.58 1.08 0.39 0.59 0.38 (1.4) (3.6) (1.1) (1.6) (0.9) TBIRPL3 0.37 0.39 0.95 0.72 0.70 (0.8) (1.3) (2.5) (1.6) (1.5) TBIRPL4 0.84 0.79 0.88 0.96 1.51 (1.8) (2.7) (2.1) (1.8) (2.9) N 206 206 571 571 427 427 287 287 252 252 ADJ R2 0.18 0.33 0.32 0.41 0.45 0.53 0.45 0.58 0.49 0.59 F-STAT 8.6 8.7 45.0 31.8 59.3 37.5 39.5 31.8 41.9 29.1 Note: t-values in parentheses.  - 34 - The estimates of the school attainment equations for private sector workers in Table 3 show that parental schooling does have a positive and significant impact on educational attainment. For residents of Lima and OUAs the effect of father's education is slightly higher than mother's education. In rural areas, father's schooling has a significantly stronger impact than mother's schooling. The addition of primary school quality indicators and birthplace variables reduces the impact of parental schooling on educational attainment, especially for rural residents where the inclusion of these variables renders the impact of father's schooling to be less strong than that of the mother. Most of the primary school quality variables contribute significantly to school attainment and their effects are stronger as the degree of urbanization decreases. Type of birthplace seems to have little effect on school attainment. -/ These findings suggest the conclusion that, although there is a positive relationship between parental education and school attainment, the provision of good quality primary schools weakens this relationship, especially in rural areas. iii) Cohort Effects Regressions were also estimated for separate age groups to assess differential effects of schooling on wages and parental education on wages and / We also estimated the school attainment regressions including public sector workers. The estimates were essentially the same, except for the effects of birthplace. Having been born in a town became positively significant for residents of OUAs and having been born in a city became significant for Lima residents. See Appendix C.  - 35 - school attainment. The following age categories were used: -15-19, 20-29, 30- 39, 40-49, and 50 and over. 11 As before, the wage equations were estimated with and without controls for parental education but we did not disaggregate across regions and excluded diploma variables because of sample size problems. The results are reported in Appendix B, Tables Bi - B3. We first briefly discuss the wage regressions without controls for parental education. Returns to primary schooling are not significant for the oldest and youngest groups, are the same for those 20-29 and 30-39 (0.13), and are slightly higher (0.16) for the 40-49 group. Returns to secondary schooling do not vary for those under 40 (about 0.07), but are higher for the 40-49 group (0.10) and are not significant for those 50 and over. With the exception of the youngest age group, there are significant returns to post- secondary education : 20-29, 0.14; 30-39, 0.16; 40-49, 0.23; and 50 and over, 0.18. The training variable has a significant effect only for three age groups, with a return of 0.15 for those 20-29, 0.23 and 0.28 for the 30-39 and 40-49 age groups, respectively. About one third of each of these age groups reported that they did vocational training. The coefficient on the public school variable is significantly negative for only two age groups: 20-29, -0.38 and -0.20 for the 30-39 year olds. Regional effects are generally the same, as for the entire sample. The coefficient on the rural area variable is / The correlations between father's and son's schooling by age group are: 0.40, 15-19, 0.53, 20-29, 0.67, 30-39, 0.67, 40-49, and 0.70 for 50 and over. The corresponding correlations between mother's and son's schooling are: 0.39, 0.50, 0.59, 0.59, and 0.65. The correlations between father's and mother's years of schooling for each age group are: 0.60, 0.65, 0.70, 0.71 and 0.79.  36 - strongly negative for all age groups. The effect of residing in OUAs is statistically significant (and negative) only for three age groups: 15-19, 30- 39, and 50 and over. Now we summarize the effects of parental education on wages. For the youngest age group and the 40-49 year olds, parental education is not significant; for the 20-29 age group, only the education of the father is statistically significant (0.04), and for the 30-39 and 50 and over age categories only the mother's education is significant (0.03 and 0.08, respectively). As regards the influence of parental education on the schooling coefficients, its effect is only apparent in the 40-49 year age group. For this group, the inclusion of parental schooling variables lowers the returns to schooling by 4 to 5 percentage points at each schooling level. Overall, the effects of parental education are neither substantial nor widespread. The school attainment regressions (Table 4) show that the effects of parental schooling diminish as the cohorts get younger. The influence of primary school quality indicators are generally positive and significant across all age groups, but no discernible pattern emerges. The influence of being born in a city is positive and strongly significant for those 20-29, 30- 39, and 50 and over, but only marginally significant for the two remaining age groups. Being born in' a town is significant only for the 30-39 age group. Again, the addition of primary school characteristics and birthplace variables lowers the effects of parental schooling, especially that of the father, and particularly for those 40 and over. 1! The inclusion of public sector workers affected the birthplace coefficients as well. Being born in a town became significant for all age groups except those 15-19.  - 37 - iv) Marital Status, Migration, Union and Firm Size Effects We now briefly summarize the effects of marital status and migration (Appendix A, Table Al) and of unions and firm size (Appendix A, Tables A2 and A3). The coefficient on the marriage variable is significant only for Lima residents (0.20), while the migration coefficient is significant only in rural areas (0.29). In Table A2, the coefficient on the union variable has the expected sign, but its magnitude varies across regions--Lima (0.16), OUAs (0.57) and rural (0.46). Table A3 shows that firm size effects on wages are uneven across regions. Briefly, the patterns are as follows (firms that employ less than 6 workers is the excluded category): in rural areas, a wage premium is associated only with "very large firms" (201+ workers and state enterprises) and with firms employing 6-20 workers; in OUAs, with firms employing 6-20, 21- 50 workers, and with very large firms; and in Lima, with firm sizes 21-50, 51- 100, and very large firms. 1- v) Returns to Schooling In Peru and Other Latin American Countries We compare our estimates of (private) rates of return to schooling to those obtained for other Latin American countries. First we refer to estimates based on data for the late 1960s and early 1970s as compiled by Psacharopoulos (1985). Then we compare our estimates to more recent studies. We also estimated a model which included both firm size and unionization. In Lima and in rural, areas, the unionization effects were "washed out" because of collinearity between firm size and unionization. (See Appendix A, Table A4). Appendix D contains the regression results when the sample is stratified by firm size; Appendix E reports the results for unionized and nonunionized workers in the private sector.  - 38 - Psacharopoulos (1985) reports that average (private) returns to primary, secondary, and post-secondary schooling in Latin America are 0.32, 0.23, and 0.23, respectively. Our estimates for Peru are below these averages even when region effects are excluded (0.16, 0.09, and 0.11) and lower than for any other Latin American country cited. However, many of the estimates reported by Psacharopolous are somewhat dated in that they use data for the 1960s and early 1970s.!1 Estimates based on more recent data for wage earners suggest that the returns to schooling in Peru are not out of line. Corbo and Stelcner (1983) using 1978 data for Chile report a rate of return to a year of schooling of about 14%, while Heckman and Hotz (1986) using 1983 Panamanian data estimate a rate of 12%. Steir (1987) reports a rate of 10% for workers in Caracas, Venezuela using 1984 data. A rate of return of 11% is estimated by Psacharopolous, Arriagada and Velez (1987) using 1984 Colombian data, and using 1978 data for BogotA, Colombia Mohan (1985) reports rates of 7% for primary schooling, 9% for secondary schooling and about 13% for post-secondary schooling. These results are similar to our estimates for Peru. Public Sector Finally, we comment on the results for government workers. The interpretation of these results is somewhat hazardous because our estimation method does not address the question of why workers are "sorted" into the Included in the averages for Latin America are 1959 estimates for Puerto Rico which were exceptionally high (0.68, 0.52 and 0.29). If Puerto Rico is excluded the average returns are 0.23, 0.18 and 0.23.  - 39 - government sector. 11 Nevertheless, our OLS regressions do provide some indication of the pay structure for government workers. The first noteworthy feature is that the magnitude of the intercept term (about unity) is much higher than that in any of the private sector regressions. This suggests that government workers enjoy a high overall average base pay, irrespective of other wage determining attributes. Second, with the exception of post-primary schooling, post-secondary diplomas, place of residence and migration variables, none of the remaining variables is statistically significant. Also, as compared to the private sector, the wage- schooling relationship is somewhat weaker.2/ These findings are suggestive of two hypotheses. The first is that government workers receive a quasi-rent or overpayment as a consequence of being employed in the public sector. That is, they are paid a substantial wage premium or "markup" (regardless of productive attributes) because of political pressures or as a result of a collective bargaining process involving a highly unionized (55%) work force. An alternative explanation is that the wage premium is paid in order to attract highly qualified workers. That is, the In general, the estimation of separate private/public sector wage equations hinges on the assumption that the distribution of workers across the sectors, given observed personal characteristics, is random. Future analysis will consider the issue of sectoral attachment. See van der Gaag and Vijverberg (1986). 2/ These impressions were confirmed when we estimated a fully interactive wage equation for the pooled sample of private and public sector workers. The interactive terms were the public sector variable x each of the explanatory variables. The statistically significant interactive terms were those on : potential work experience (-.033), potential work experience squared (0.0005), years of primary schooling (-0.12), years of post-secondary schooling (-0.073), training (-0.16), and the dummy variable for public sector (1.079).  - 40 - government acts as a cost-minimizer and sets wages no higher than necessary to attract the required work force. Appropriate testing of these competing hypotheses requires estimation methods that incorporate the endogeneity of sector choice.  - 41 - IV. CONCLUDING REMARKS This study has analyzed the determinants of wages among Peruvian male employees, particular attention being given, to the roles of education and parental schooling. The empirical results do not reveal any major surprises and they generally conform to the expectations of economic theory. The study confirms that schooling is a sizable contributor to variation in wages among Peruvian male employees. As regards the effect of family background, the -results suggest that these a,re not as strong as may have been expected in a society that is often characterized as "socially stratified". This may reflect the expansion of the educational system that started in the 1960s. The productivity-raising effects of formal schooling remain strong after incorporating influences of parental education. It is important to note that the impacts of vocational training and post-secondary schooling are strong in Lima and rural areas but not in OUAs. However, the effects of formal primary and secondary schooling are not statistically different across regions. Of particular interest then are the regional variations in wage structures indicated by this study. The estimates suggest that there exist labor market disequilibria for which there are several explanations: regional variations in the demand for labor; migration patterns and costs; regional differences in the quality of schools; and finally selectivity bias may also be a reason since the estimates are based only on a sample of workers whose employment choices (wage earners vs self-employed), which may also be region specific, are not taken into account.I/ / Birdsall and Behrman (1984) and Heckman and Hotz (1986) discuss in detail the possible explanations for labor market disequilibria. For a discussion of spatial differences in poverty in Peru see Thomas (1980).  - 42 - Of what importance then are the results of this study? First, they lend support, once again, to the -human capital framework in explaining wage variations. As in numerous other studies, we find strong evidence that schooling plays an important role in explaining wage variations, particularly in the private sector. The pattern of returns to schooling in Peru reflects that found in most other developing countries: the rates of return to primary education are higher than those to secondary schooling and, with the exception of OUAs, they increase for post-secondary education. It is also of some interest to note that these estimates are bracketed by recent ones for other Latin American countries and are respectable given the economic conditions of Peru in the 1980s. In brief, the results presented here show that there is an economic "payoff" to schooling. The research reported here has some policy implications. As the Peruvian government struggles with its external debt problem and the overall weakness of the economy, there is little doubt that public spending on education will not increase above the present levels in the foreseeable future. It is more likely that it will continue to decrease. The results of this study suggest that (at least) two considerations should be taken into account in squeezing the education budget. The regional differences displayed by the regressions indicate that closer attention should be paid to regional effects in formulating education policies. Of importance in this regard are the relatively high rates of return to post-secondary schooling and training in Lima and in rural areas, but not in the collection of cities and towns (which we call OUAs), each of which has a population of well under one million inhabitants.  - 43 - The strong negative impact of public school education on wage outcomes in Lima and rural areas (Table 1) is also conveying a message. This result suggests that, although public school enrollments have increased substantially during the past quarter century, the performance of its graduates has not been commensurate. The reasons for this may well be related to internal efficiency aspects of the public school system. A basic policy (and research) issue at stake is whether reductions in public provision of education in Peru coupled with modest increases in public financing of privately-operated schools is a desirable alternative. The "descriptive regressions" of this study suggest a research agenda containing a variety of interrelated topics. What determines whether an individual is a wage earner and, if so, in which sector (public/private, union/nonunion)? What role do migration decisions play in determining employment and earning outcomes, and in understanding regional differences? Since training has an important effect on wage variations (in rural areas and in Lima), what are the determinants of training choices, and what is the relationship between formal schooling and training? How does wage determination among women compare to that of men? Finally, what are the comparative rates of return to education and training of those engaged in non- agricultural self-employment activities? The authors are in the process of investigating these topics.  - 44 - REFERENCES Armitage, J. and R. Sabot, (1983), "Social Economic Background and the Returns to Schooling in Two Low Income Economies", Washington: World Bank, Mimeo. Becker, C., (1964), Human Capital, New York: Columbia University Press. Behrman, J. and B. Wolfe, (1984), "The Socioeconomic Impact of Schooling in a Developing Country" Review of Economic and Statistics, Vol. 46. Behrman, J., B. Wolfe and D. Blau, (1985), "Human Capital and Earnings Distribution in a Developing Country. The Case of Prerevolutionary Nicaragua", Economic Development and Cultural Change, Vol. 34, No. 1. Birdsall, N. and J. Behrman, (1984), "Does Geographical Aggregation Cause Overestimates of the Returns to Schooling? Oxford Bulletin of Economics and Statistics, Vol. 46. Blinder, A., (1976), "On Dogmatism in Human Capital Theory", Journal of Human Resources, Vol. 11. Blomquist, S., (1978), Wage Rates and Personal Characteristics, Vancouver: Department of Economics, University of British Columbia, Discussion Paper Corbo, V. and M. Stelcner, (1983), "Earnings Determination and Labour Markets: Gran Santiago, Chile - 1978", Journal of Development Economics, Vol. 12 Drysdale, R. S. and R.G. Myers, (1975), "Continuity and Change: Peruvian Education", in A.F. Lowenthal (ed). The Peruvian Experiment--Continuity aid Change under Military Rule, Princeton: Princeton University Press. Flores, R., (1980), "La Segmentacion del Mercado Laboral y la determinacion de los ingresos: El caso de Lima Metropolitana", Washington: The American University, Seminar In Applied Research, mimeo. Grootaert, C. and A-M. Arriagada, (1986), The Peruvian Living Standards Survey: An Annotated Questionnaire, Washington: The World Bank, July. Hay, G., (1976), Educational Resources and Educational Reform in Peru, Paris: International Institute for Educational Planning. Heckman, J. and J. Hotz, (1986), "An Investigation of The Labor Market Earnings of Panamanian Males: Evaluating Sources of Inequality", Journal of Human Resources, Vol. 21, No. 4. Instituto Nacional de Estadistica, (1986), Peru: Compendio Estadistico 1985, Lima: May. Jamison, D. and P. Moock, (1984), "Farmer Education and Farm Efficiency in Nepal: The Role of Schooling, Extension Services and Cognitive Skills", World Development, Vol. 12, No. 1.  - 45 - Kearl, J.R. and C.L. Pope, eds. (1986), "The Family and the Distribution of Economic Rewards", Proceedings of the Sunmark Conference on Economics, Utah, September, 1984, Journal of Labor Economics, Vol. 4, No. 3, Part 2. Mohan, R., (1986), Work, Wages, and Welfare in a Developing Metropolis, Consequences of Growth in Bogota, Colombia, New York: Oxford University Press for the World Bank. Mincer, J., (1958), "Investments in Human Capital and Personal Income Distribution", Journal of Political Economy, Vol. 56. --------., (1974), Schooling, Experience, and Earnings, New York: Columbia University Press. Nyrop, R.F. (ed.), (1981), Peru: A Country Study, Washington, D.C.: American University, Foreign Area Studies. Musgrove, P. (ed.), (1982), Ingreso, Desigualdad y Pobreza en America Latina, Rio de Janeiro: ECIEL-BID Publication. Psacharopoulos, C., (1985), "Returns to Education: A Further International Update and Implications", Journal of Human Resources, Vol. 20. ---, A-M Arriagada and E. Velez, (1987), " Earnings and Education Among The Self-Employed in Colombia", Washington: Education and Training Department, World Bank, Mimeo. Steir, F., (1987), "Schooling, Experience, and Earnings: Issues in Venezuelan Development, 1975-84", New York: Ph.D. Dissertation, Columbia University. Suarez, R., (1987), "Peru Informal Sector, Labor Markets, and Returns to Education", Washington, D.C: World Bank, Development Research Department, Living Standards Measurement Study Working Paper No. 32. Toledo, A.C., (1984), "Labor Market Segmentation Test: Changes in Schooling, Employment, and Earnings Inequality in the Urban Male Labor Force in Peru", Stanford: Graduate School of Education, Stanford University, mimeo. Thomas, V., (1980), "Spatial Differences in Poverty: The Case of Peru", Journal of Development Economics, Vol. 7. Van der Caag J., and W.P.M. Vijverberg, (1986), " A Switching Regression Model for Wage Determinants in the Public and Private Sectors of a Developing Country", Washington: Development Research Department, World Bank, Mimeo. Weeks-Vagliani W., (1985), Actors and Institutions in the Food Chain: The Case of Peru, Paris: OECD. World Bank, (1983i, Peru: Education Sector Strategy Paper, Washington: April. World Bank, (1985), Peru: Country Economic Memorandum, Washington: December.  - 46- APPENDIX A Wage Regressions with Controls for: Table Al. Marital Status and Migration Table A2. Unionization Table A3. Firm Size Table A4. Firm Size and Unionization  - 47 - Table Al: Wage Regressions for Peruvian Male Workers, 1985/86 Public Sector Private Sector All Peru All Peru Lima Other Urban Rural VARIABLE (1) (2) (3) (4) (5) (6) INTERCEPT 0.9905 -0.0985 -0.3873 0.1742 -0.5449 0.0728 (4.44) (0.92) (4.23) (0.85) (2.36) (0.46) CEXPR1 0.0106 0.0367 0.0384 0.0325 0.0569 0.0138 (1.08) (5.98) (6.23) (3.72) (4.64) (1.15) GEXPR1SQ -0.0001 -0.0005 -0.0005 -0.0004 -0.0008 -0.0002 (0.14) (4.33) (4.28) (2.35) (3.84) (0.88) XOCM7 0.0140 0.0284 0.0287 0.0268 0.0503 0.0130 (1.27) (4.89) (4.90) (2.71) (3.78) (1.30) XOCSQM7 -0.0004 -0.0007 -0.0007 -0.0006 -0.0013 -0.0004 (1.15) (4.54) (4.77) (1.85) (3.19) (1.53) SPLYRSC1 -0.0049 0.1025 0.1231 0.0609 0.0916 0.0761 (0.12) (5.38) (6.64) (1.48) (2.04) (2.52) SPLYRSC2 0.0714 0.0573 0.0690 0.0471 0.0769 0.0458 (3.61) (5.02) (6.11) (2.97) (3.60) (1.80) SPLYRSC3 0.0308 0.0731 0.0719 0.1059 0.0131 0.2724 (1.58) (3.61) (3.53) (4.67) (0.31) (5.90) TRAIN 0.0465 0.1895 0.2125 0.2158 -0.0162 0.5383 (0.85) (4.36) (4.88) (4.12) (0.19) (4.07) DIPLOMAl -0.0004 0.0542 0.0222 -0.0456 0.1117 _ (0.01) (0.44) (0.18) (0.28) (0.50) DIPLOMA2 0.3229 0.3262 0.3257 0.2134 0.5158 _ (2.86) (2.18) (2.16) (1.25) (1.64) DIPLOMA3 0.3066 0.5010 0.4888 0.3606 0.1966 - (2.67) (4.19) (4.06) (2.75) (0.72) PUBSCHL -0.0129 -0.1573 -0.1646 -0.1169 -0.0927 -0.2412 (0.15) (2.91) (3.02) (1.68) (0.75) (2.17) FYR_SCHL 0.0035 0.0261 0.0293 0.0304 0.0323 0.0153 (0.39) (3.87) (4.34) (3.76) (2.41) (0.77) Note: t-values in parentheses. (Continued)  - 48 - Table Al: (Continued) Public Sector Private Sector All Peru All Peru Lima Other Urban Rural VARIABLE (1) (2) (3) (4) (5) (6) MYR SCHL 0.0170 0.0235 0.0283 0.0128 0.0404 0.0145 (1.68) (3.10) (3.74) (1.45) (2.67) (0.55) MARITALO 0.0312 0.1210 0.0967 0.2028 0.0505 0.1055 (0.44) (2.54) (2.02) (3.11) (0.52) (1.14) RESIDFPL 0.1365 0.1225 0.1300 -0.0214 0.1114 0.3200 (2.19) (3.13) (3.31) (0.39) (1.36) (4.31) TYPRES2 -0.1140 -0.1666 - _ _ (2.04) (3.87) TYPRES3 -0.2233 -0.2660 - - - _ (2.47) (5.10) N 526 1743 1743 759 517 467 ADJ R2 0.26 0.41 0.40 0.42 0.30 0.34 F STAT 9.07 59.08 63.02 30.36 11.66 15.62 Note: t-values in parentheses.  - 49 - Table A2: Wage Regressions for Peruvian Male Workers, 1985/86 Public Sector Private Sector All Peru All Peru Lima Other Urban Rural VARIABLE (1) (2) (3) (4) (5) (6) INTERCEPT 0.9887 -0.0519 -0.3084 0.1640 -0.3955 0.1547 (4.40) (0.50) (3.43) (0.81) (1.79) (0.98) CEXPR1 0.0105 0.0345 0.0359 0.0320 0.0503 0.0112 (1.07) (5.73) (5.95) (3.68) (4.28) (0.96) GEXPR1SQ -0.0001 -0.0004 -0.0004 -0.0004 -0.0007 -0.0002 (0.13) (4.21) (4.16) (2.31) '3.38) (0.94) XOCM7 0.0139 0.0228 0.0229 0.0238 0.0425 0.0107 (1.24) (3.99) (3.97) (2.41) (3.33) (1.09) XOCSQM7 -0.0004 -0.0006 -0.0006 -0.0006 -0.0012 -0.0003 (1.13) (3.89) (4.08) (1.79) (3.11) (1.11) SPLYRSC1 -0.0049 0.0939 0.1119 0.0610 0.0768 0.0726 (0.11) (5.03) (6.16) (1.49) (1.78) (2.46) SPLYRSC2 0.0713 0.0567 0.0671 0.0468 0.0767 0.0477 (3.61) (5.07) (6.07) (2.96) (3.76) (1.92) SPLYRSC3 0.0306 0.0629 0.0615 0.0996 0.0167 0.2192 (1.56) (3.17) (3.08) (4.40) (0.41) (4.70) TRAIN 0.0465 0.1622 0.1818 0.2066 -0.0525 0.4766 (0.85) (3.81) (4.26) (3.96) (0.64) (3.67) DIPLOMAl -0.0012 0.0266 -0.0030 -0.0416 0.0081 _ (0.01) (0.22) (0.03) (0.25) (0.04) DIPLOMA2 0.3224 0.0265 0.2597 0.1777 0.4974 - (2.85) (1.79) (1.76) (1.04) (1.65) DIPLOMA3 0.3061 0.4588 0.4465 0.3529 0.1047 - (2.66) (3.92) (3.79) (2.70) (0.40) PUBSCHL -0.0123 -0.1647 -0.1715 -0.1226 -0.0858 -0.2860 (0.14) (3.11) (3.22) (1.77) (0.73) (2.61) FYR_SCHL 0.0035 0.0259 0.0287 0.0307 0.0248 0.0201 (0.38) (3.92) (4.35) (3.81) (1.92) (1.03) Note: t-values in parentheses. (Continued)  - 50 - Table A2: (Continued) Public Sector Private Sector All Peru All Peru Lima Other Urban Rural VARIABLE (1) (2) (3) (4) (5) (6) MYR_SCHL 0.0170 0.0218 0.0261 0.0125 0.0419 0.0013 (1.68) (2.95) (3.53) (1.42) (2.90) (0.05) MARITALO 0.0317 0.0948 0.0720 0.1952 0.0166 0.0726 (0.44) (2.03) (1.54) (3.01) (0.18) (0.80) RESIDFPL 0.1371 0.1133 0.1196 -0.0126 0.0589 0.2954 (2.18) (2.96) (3.12) (0.23) (0.75) (4.06) LUNIONM7 0.0049 0.3647 0.3779 0.1617 0.5710 0.4547 (0.09) (8.74) (9.02) (2.97) (6.87) (4.62) TYPRES2 -0.1148 -0.1507 - _ - _ (2.02) (3.58) TYPRES3 -0.2242 -0.2386 - - - - (2.46) (4.66) N 526 1743 1743 759 517 467 ADJ R2 0.26 0.43 0.42 0.43 0.36 0.37 F STAT 8.62 62.37 66.76 29.53 14.55 16.63 Note: t-values in parentheses.  - 51 - Table Xi: Wage Regressions for Peruvian Male Workers, 1985/86 Private Sector All Peru Lima Other Urban Rural VARIABLE (1) (2) (3) (4) (5) INTERCEPT -0.0710 -0.3459 0.1403 -0.4672 0.2181 (0.68) (3.91) (0.69) (2.14) (1.40) GEXPR1 0.0333 0.0345 0.0314 0.0516 0.0038 (5.63) (5.79) (3.59) (4.49) (0.34) GEXPR1SQ -0.0004 -0.0004 -0.0004 -0.0007 -0.0001 (4.04) (3.94) (2.26) (3.66) (0.25) XOCM7 0.0210 0.0214 0.0221 0.0395 0.0086 (3.74) (3.78) (2.26) (3.15) (0.90) XOCSQM7 -0.0005 -0.0006 -0.0005 -0.0011 -0.0002 (3.68) (3.91) (1.77) (2.86) (1.01) SPLYRSC1 0.0844 0.1031 0.0541 0.0725 0.0620 (4.59) (5.75) (1.34) (1.72) (2.12) SPLYRSC2 0.0510 0.0618 0.0392 0.0714 0.0433 (4.64) (5.69) (2.50) (3.56) (1.77) SPLYRSC3 0.0579 0.0560 0.0881 0.0160 0.2171 (2.96) (2.85) (3.92) (0.40) (4.85) TRAIN 0.1387 0.1581 0.1912 -0.0634 0.4148 (3.30) (3.75) (3.70) (0.91) (3.26) DIPLOMAI 0.0156 -0.0113 -0.0357 0.0187 (0.13) (0.10) (0.09) (0.22) DIPLOMA2 0.3067 0.3066 0.2163 0.4746 (2.12) (2.11) (1.29) (1.60) DIPLOMA3 0.4461 0.4379 0.3646 0.0948 _ (3.86) (3.77) (2.81) (0.37) PUBSCHL -0.1449 -0.1493 -0.1128 -0.0883 -0.2539 (2.78) (2.85) (1.65) (0.76) (2.37) FYR_SCHL 0.0244 0.0272 0.0312 0.0208 0.0148 (3.76) (4.19) (3.92) (1.65) (0.78) Note: t-values in parentheses. (Continued)  - 52 - Table A3: (Continued) Private Sector All Peru Lima Other Urban Rural VARIABLE (1) (2) (3) (4) (5) MYR SCHL 0.0202 0.0245 0.0095 0.0376 0.0165 (2.77) (3.37) (1.09) (2.65) (0.66) MARITALO 0.0447 0.0207 0.1563 -0.0634 0.0402 (0.96) (0.45) (2.43) (0.69) (0.44) RESDIFPL 0.1040 0.1134 -0.0143 0.0919 0.2485 (2.75) (2.99) (0.27) (1.20) (3.45) FRM620 0.1660 0.1896 0.0993 0.2453 0.1835 (3.58) (4.09) (1.45) (2.81) (2.07) FSIZE5M7 0.3319 0.3543 0.3248 0.4382 -0.0023 (5.45) (5.79) (3.92) (3.92) (0.02) FRM51200 0.2409 0.2844 0.2219 0.2674 0.1266 (4.01) (4.75) (2.84) (1.96) (0.98) FRM201ST 0.6091 0.6201 0.3912 0.8307 0.6975 (11.89) (12.04) (5.24) (8.49) (6.94) TYPRES2 -0.1436 - _ (3.45) TYPRES3 -0.2491 - - _ _ (4.91) N 1743 1743 759 517 467 ADJ R2 0.45 0.45 0.45 0.39 0.41 F-STAT 59.42 62.76 27.42 14.25 16.17 Note: t-values in parentheses.  - 53 - Table A4: Wage Regressions for Peruvian Male Workers, 1985/86 Private Sector All Peru Lima Other Urban Rural VARIABLE (1) (2) (3) (4) (5) INTERCEPT -0.0634 -0.3266 0.139Y -0.4360 0.2279 (0.61) (3.69) (0.68) (2.01) (1.46) GEXPR1 0.0330 0.0342 0.0313 0.0497 0.0040 (5.59) (5.75) (3.59) (4.34) (0.35) GEXPR1SQ -0.0004 -0.0004 -0.0004 -0.0007 -0.0001 (4.04) (3.95) (2.26) (3.45) (0.32) XOCM7 0.0202 0.0205 0.0220: 0.0390 0.0084 (3.59) (3.62) (2.24) (3.13) (0.88) XOCSQM7 -0.0005 -0.0006 -0.0005 -0.0011 -0.0002 (3.57) (3.78) (1.77) (2.92) (0.94) SPLYRSC1 0.0833 0.1009 0.0542 0.0701 0.0631 (4.53) (5.64) (1.34) (1.67) (2.15) SPLYRSC2 0.0520 0.0625 0.0393 0.0713 0.0442 (4.73) (5.77) (2.50) (3.58) (1.80) SPLYRSC3 0.0555 0.0535 0.0879 0.0140 0.2076 (2.84) (2.73) (3.90) (0.35) (4.56) TRAIN 0.1370 0.1555 0.1911 -0.0736 0.4095 (3.27) (3.70) (3.69) (0.92) (3.21) DIPLOMAl 0.0148 -0.0114 -0.0352 0.0131 _ (0.13) (0.10) (0.22) (0.06) DIPLOMA2 0.2935 0.2919 0.2143 0.5085 _ (2.03) (2.01) (1.27) (1.72) DIPLOMA3 0.4430 0.4354 0.3645 0.1000 - (3.84) (3.76) (2.81) (0.39) PUBSCHL -0.1486 -0.1535 -0.1131 -0.0838 -0.2668 (2.86) (2.94) (1.66) (0.73) (2.48) FYR_SCHL 0.0245 0.0273 0.0312 0.0195 0.0162 (3.79) (4.21) (3.92) (1.56) (0.85) Note: t-values-in parentheses. (Continued)  - 54 - Table A4: (Continued) Private Sector All Peru Lima Other Urban Rural VARIABLE (2) (3) (4) (5) MYR SCHL 0.0201 0.0243 0.0095 0.0397 0.0122 (2.77) (3.34) (1.09) (2.80) (0.48) MARITALO 0.0447 0.0213 0.1566 -0.0761 0.0406 (0.96) (0.46) (2.43) (0.83) (0.45) RESDIFPL 0.1037 0.1123 -0.0138 0.0743 0.2474 (2.75) (2.97) (0.26) (0.97) (3.44) LUNIONM7 0.1465 0.1605 0.0128 0.2780 0.1300 (2.86) (3.12) (0.20) (2.51) (1.11) FRM620 0.1607 0.1830 0.0990 0.2335 0.1733 (3.48) (3.95) (1.44) (2.69) (1.95) FSIZE5M7 0.3131 0.3327 0.3224 0.4255 -0.0237 (5.12) (5.42) (3.85) (3.82) (0.16) FRM51200 0.1822 0.2189 0.2159 0.1741 0.0926 (2.88) (3.46) (2.57) (1.24) (0.69) FRM201ST 0.5092 0.5104 0.3826 0.6198 0.6209 (8.22) (8.20) (4.41) (4.82) (5.09) TYPRES2 -0.1425 - _ _ (3.43) TYPRES3 -0.2398 - - - _ (4.73) N 1743 1743 759 517 467 ADJ R2 0.46 0.45 0.45 0.40 0.41 F-STAT 57.61 60.76 26.19 14.05 15.43 Note: t-values in parentheses.  - 55 - APPENDIX B Wage Regressions by Cohort, Private Sector Table B1. Mean Characteristics Table B2. Wage Regressions - Parental Schooling Excluded Table B3. Wage Regressions - Parental Schooling Included  - 56 - Table Bi: Mean Characteristics of Wage Earners, Private Sector, by Cohort CHARACTERISTICS 15-19 20-29 30-39 40-49 50+ No. of observations 206 571 427 287 252 Real hourly wage rate 2.253 4.769 7.060 7.354 8.618 [Intis at June 1985 prices] (1.7) (6.6) (10.0) (9.1) (24.8) Ln real hourly wage rate 0.599 1.185 1.503 1.483 1.461 (0.6) (0.8) (0.9) (1.0) (1.1) Potential work experience 4.3 9.4 19.4 32.0 44.5 (2.6) (4.0) (5.5) (5.5) (6.7) Job specific experience 2.1 3.6 8.3 14.96 20.8 (2.5) (3.7) (6.6) (9.5) (12.6) Years of Schooling 6.6 8.7 8.3 6.0 5.6 Total (2.7) (3.2) (4.2) (4.3) (4.2) Primary 4.5 4.8 4.5 3.9 3.7 (1.1) (0.8) (1.2) (1.7) (1.8) Secondary 2.0 3.3 2.9 1.7 1.5 (2.0) (2.1) (2.3) (2.2) (2.2) Post-secondary 0.0 0.6 0.9 0.5 0.4 (0.2) (1.4) (1.9) (1.5) (1.4) Level of schooling completed Less Than Primary 0.027 0.022 0.034 0.082 0.117 Primary 0.359 0.011 0.307 0.522 0.556 Secondary Regular 0.570 0.200 0.361 0.195 0.210 Secondary Technical 0.029 0.541 0.066 0.070 0.012 Post-Secondary Non-University 0.010 0.056 0.049 0.021 0.024 University 0.005 0.061 0.183 0.108 0.079 Diplomas Obtained Secondary technical 0.000 0.021 0.030 0.038 0.008 Post-Secondary Non-University 0.005 0.021 0.023 0.010 0.008 University 0.000 0.025 0.094 0.084 0.056 -/ Accredited. NOTE: The numbers in parentheses are standard deviations. (Continued)  - 57 - Table Bi: (Continued) CHARACTERISTICS 15-19 20-29 30-39 40-49 50+ School last attended was public 0.927 0.858 0.785 0.850 0.766 Vocational Training 0.044 0.303 0.344 0.286 0.179 Father's years 4.0 4.8 4.6 3.5 3.5 of schooling (3.4) (3.8) (4.1) (3.6) (4.1) Mother's years 2.5 3.2 3.0 1.9 2.4 of schooling (2.8) (3.5) (3.6) (2.9) (3.2) Married or cohabiting 0.019 0.394 0.794 0.923 0.849 Ever lived elsewhere 0.403 0.580 0.740 0.798 0.798 Union in the firm 0.049 0.168 0.359 0.356 0.365 Firm size: 1 - 5 workers 0.557 0.295 0.115 0.218 0.229 6 - 20 workers 0.204 0.237 0.250 0.188 0.186 21 - 50 workers 0.049 0.156 0.082 0.094 0.095 51 - 200 workers 0.063 0.097 0.118 0.131 0.130 201 or more workers 0.015 0.107 0.164 0.192 0.171 State enterprise 0.005 0.035 0.126 0.111 0.103 Current Place of Residence Metropolitan Lima 0.359 0.478 0.438 0.408 0.429 Other Urban Areas 0.296 0.287 0.328 0.275 0.290 Rural Areas 0.345 0.235 0.234 0.317 0.282 .. I l . I~~~~~~~~~~~~~~~~~~~~~~~~~~~~  - 58 - Table B2: Wage Regressions for Peruvian Male Workers, Private Sector, by Cohort - Parental Schooling Excluded AGE GROUP IN YEARS VARIABLE 15-19 20-29 30-39 40-49 50+ INTERCEPT 0.5018 0.1454 0.0390 -0.6390 1.9038 (1.33) (0.52) (0.10) (1.00) (3.01) GEXPR1 0.0243 0.0357 0.0223 0.0334 -0.0192 (0.82) (3.07) (1.78) (2.16) (1.68) XOCM7 0.0465 0.0519 0.0799 0.0444 0.0210 (0.95) (2.43) (4.82) (2.78) (1.53) XOCSQM7 -0.0056 -0.0027 -0.0036 -0.0010 -0.0005 (1.13) (1.83) (4.78) (2.24) (1.63) SPLYRSC1 0.0318 0.1327 0.1330 0.1555 0.0735 (0.53) (2.99) (3.01) (3.31) (1.52) SPLYRSC2 0.0723 0.0745 0.0754 0.0963 0.0558 (2.47) (3.55) (2.93) (2.97) (1.59) SPLYRSC3 0.2163 0.1381 0.1560 0.2312 0.1793 (1.16) (5.50) (6.14) (5.71) (3.87) TRAIN 0.1488 0.1494 0.2309 0.2795 0.2396 (0.68) (2.12) (2.78) (2.49) (1.62) PUBSCHL -0.1272 -0.3827 -0.2030 -0.1220 0.0165 (0.69) (3.94) (2.13) (0.79) (0.11) TYPRES2 -0.3798 -0.1117 -0.1848 -0.0873 -0.3029 (3.42) (1.52) (2.14) (0.77) (2.32) TYPRES3 -0.3327 -0.2462 -0.2356 -0.4011 -0.4883 (2.61) (2.71) (2.15) (2.93) (3.05) N 206 571 427 287 252 ADJ R2 0.13 0.22 0.33 0.42 0.38 F-STAT 4.16 16.82 21.92 21.76 16.35 Note: t-values in parentheses.  - 59 - Table B3: Wage Regressions for Peruvian Male Workers, Private Sector, by Cohort - Parental Schooling Included AGE GROUP IN YEARS VARIABLE 15-19 20-29 30-39 40-49 50+ INTERCEPT 0.4951 0.0403 -0.0588 -0.4347 2.0014 (1.25) (0.15) (0.15) (0.67) (3.23) GEXPR1 0.0195 0.0374 0.0228 0.0265 -0.0218 (0.64) (3.25) (1.81) (1.68) (1.93) XOCM7 0.0358 0.0447 0.0823 0.0351 0.0145 (0.71) (2.11) (4.99) (2.19) (1.07) XOCSQM7 -0.0042 0.0023 -0.0036 -0.0008 0.0004 (0.81) (1.60) (4.84) (1.73) (1.38) SPLYRSC1 0.0257 0.1107 0.1178 0.1124 0.0359 (0.42) (2.50) (2.67) (2.31) (0.73) SPLYRSC2 0.0654 0.0647 0.0577 0.0607 -0.0002 (2.18) (3.08) (2.17) (1.77) (0.01) SPLYRSC3 0.2131 0.1183 0.1296 0.1899 0.1479 (1.13) (4.55) (4.87) (4.48) (3.17) TRAIN 0.1702 0.1436 0.2103 0.3138 0.2719 (0.77) (2.06) (2.54) (2.81) (1.89) PUBSCHL -0.0997 -0.3178 -0.1461 -0.0393 0.0875 (0.52) (3.22) (1.52) (0.26) (0.57) TYPRES2 -0.3404 -0.1004 -0.1649 -0.0242 -0.2830 (2.91) (1.37) (1.91) (0.21) (-2.21) TYPRES3 -0.2982 -0.1685 -0.1963 -0.3457 -0.3921 (2.24) (1.80) (1.78) (2.54) (2.48) FYR_SCHL 0.0034 0.0445 0.0122 0.0357 0.0129 (0.21) (3.95) (0.90) (i.80) (0.60) MYR_SCHL 0.0097 -0.0093 0.0338 0.0345 0.0795 (0.46) (0.78) (2.29) (1.57) (3.03) N 206 571 427 287 252 ADJ R2 0.12 0.24 0.34 0.44 0.41 F-STAT 3.08 13.75 16.77 16.87 13.45 Note: t-values in parentheses.  - 60 - APPENDIX C School Attainment Regressions, All Sectors Correlations of Son's and Parents' Schooling Table Cl. School Attainment Regressions, By Region Table C2. School Attainment Regressions, by Cohort Table C3. Correlations of Son's and Parents' Schooling, by Region Table C4. Correlations of Son's and Parents' Schooling, by Cohort  - 61 - Table Cl: School Attainment Regressions for Peruvian Male Workers, All Sectors, by Region VARIABLE ALL PERU LIMA OTHER URBAN RURAL INTERCEPT 4.37 0.99 5.93 3.81 5.18 1.89 2.47 0.15 (18.0) (3.5) (16.9) (6.7) (11.9) (3.2) (5.3) (0.3) MYR SCHL 0.32 0.23 0.23 0.19 0.30 0.27 0.46 0.39 (11.1) (8.5) (6.7) (5.3) (5.6) (5.2) (5.0) (4.8) FYR SCHL 0.43 0.31 0.31 0.28 0.34 0.26 0.69 0.40 (17.4) (12.9) (10.1) (8.9) (7.5) (5.8) (1.9) (6.1) FLIVAGIO 0.73 0.45 0.55 0.46 0.80 0.45 0.70 0.36 (3.9) (2.5) (2.0) (1.7) (2.4) (1.4) (1.9) (1.2) MLIVAG10 0.36 0.44 0.59 0.59 0.15 0.08 0.36 0.36 (1.5) (2.0) (1.8) (1.8) (0.3) (0.2) (0.7) (0.9) FURNSCHL 2.04 0.98 1.98 2.25 (8.7) (2.2) (4.1) (6.9) FOODSCHL 0.33 0.11 0.47 0.61 (2.4) (0.6) (1.7) (2.3) TECHCAT2 1.49 0.48 1.44 1.65 (8.1) (1.5) (4.1) (5.7) TECHCAT3 1.89 0.82 1.92 2.15 (8.9) (2.5) (4.7) (4.5) BOOKCAT3 0.60 0.49 0.53 0.47 (3.7) (1.8) (1.8) (1.7) TBIRPL3 1.00 0.49 0.71 0.73 (5.5) (1.5) (2.0) (2.5) TBIRPL4 0.99 0.82 0.41 0.38 (5.3) (2.5) (1.2) (0.9) N 2269 2269 1013 1013 730 730 526 526 ADJ R2 0.36 0.46 0.27 0.30 0.26 0.32 0.37 0.54 F-STAT 217.4 151.6 64.9 33.65 42.7 27.6 51.7 47.8 Note: t-values in parentheses.