Public - private sector wage differentials in Peru : 1985-86

LIIT1S LSM - 41 DEC. 1987 Living Standards Measurement Study Working Paper No. 41 Public-Private Sector Wage Differentials in Peru: 1985-86 Morton Stelcner Jacques van der Gaag Wim Vijverberg  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. No. 19. The Conceptual Basis of Measures of Household Welfare and Their Implied Survey Data Requirements. No. 20. Statistical Experimentation for Household Surveys: Two Case Studies of Hong Kong. (List continues on the inside of the back cover)  LSMS Working Paper Number 41 PUBLIC-PRIVATE SECTOR WAGE DIFFERENTIALS IN PERU: 1985-86 Morton Stelcner Jacques van der Gaag Wim Vijverberg Population and Human Resources Department The World Bank Washington, D.C. 20433, U.S.A.  - 11 - This is a working document published informally by the Population and Human Resources Department of The World Bank. The World Bank does not accept responsiblity for the views expressed herein, which are those of the author 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 Welfare and Human Resources Division, Population and Human Resources Department, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A. Jacques van der Gaag is Division Chief of the Welfare and Human Resources Division. Morton Stelcner is a long-term consultant for the Welfare and Human Resources Division and Associate Professor, Department of Economics, Concordia University, Montreal, Quebec. Wim Vijverberg is Assistant Professor at the University of Texas at Dallas and works as a consultant for the Welfare and Human Resources Division. December 1987  - iii - PREFACE 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 - ABSTRACT Public-Private Sector Wage Differentials in Peru: 1985-86 Fiscal deficits and external debts have placed public sector employment and compensation under increased scrutiny as developing countries confront the economic crises of recent years. Although much attention has been given to the overall problem of growing public expenditures, there have been few research efforts addressed to a basic question: Do public sector workers with the same productivity traits earn more than their private sector counterparts? The answer to this question has important policy implications because reducing the government wage bill, which comprises the major portion of recurrent public spending, is often viewed as an attractive means to reducing budget deficits. In many countries, this is accomplished by allowing the salaries of government employees to be eroded by inflation, while maintaining the level of employment. Surprisingly, there is little systematic empirical evidence on public-private sector pay comparability in developing countries, and most of the existing findings show mixed results in terms of both direction and magnitude of wage differentials. Moreover, most studies are based on statistical procedures, such as ordinary least squares regression, that do not take into account the process by which the wage earners with different personal characteristics are likely to be employed in the public rather than the private sector, i.e., selection bias. This study considers how wage differentials between male wage earners in the two sectors are generated using recent (1985/86) data for Peru (The Peru Living Standards Survey). Explicit attention is given to the endogeneity of sector choice. A switching regression model is used to consider the question of whether government workers enjoy a "pure" wage advantage or economic rent. We find that this is not the case if selectivity corrected estimates of wage functions are compared. In fact, in metropolitan Lima, public sector wages are well below those in the private sector, while in other urban areas there is no significant wage differential. The estimation procedure and results of this study should serve as useful resources for formulating wage and employment policies in developing countries, especially wage and employment reforms in the public sector.  v TABLE OF CONTENTS I. Introduction ...*.... 6......I......................l II. The Switching Regression Nodel ............................. 7 III. Data Set and Variable Description .......................... 11 IV. Estimation Results ..%*... ... ............................ o*18 Ve Discussion*. .......................................... *28 VI. Conclusions ........................................ #o...34 Appendix... .................................... 0...36 References .......................................... 41  I. Introduction It is well-documented that a large portion of formal (wage) employment in developing countries as well as in industrialized economies is in the public (not-for-profit) sector. (For a comprehensive collection of statistics on public employment and salaries, see Heller and Tait, 1983). This phenomenon has focused attention on various aspects of public sector labor markets, in particular public-private wage differentials. These wage differentials have been researched quite extensively in industrialized economies, especially Canada and the US, but only recently has systematic attention been given to this topic in the context of developing countries.!, An analysis of public-private wage differentials is particularly important and timely for Third World countries because fiscal deficits and external debts have placed public sector employment and compensation under increased scrutiny. Since the government wage bill forms a high proportion of current budget expenditures, cutting it is often viewed as an attractive means to reduce fiscal deficits. Furthermore, government pay scales in LDCs often serve as the prevailing model, if not as a lever for wage earners in the private sector. When public sector employment is the dominant component of the wage sector, the pay structures and work conditions ( e.g. job security, 1/ Studies on Canada include those of Abbot and Stengos (1986), Gunderson (1979a, 1979b, 1979c), Robinson and Tomes (1984), Shapiro and Stelcner (1980, 1986); For comprehensive surveys of studies on the US see Ehrenberg and Schwarz (1986) and Wise (1987). As regards developing countries, see Bennell (1981) on Ghana, Kenya and Nigeria; Corbo and Stelcner (1983) on Chile; House (1984) on Cyprus; Lee (1980) and Mazumdar (1981) on Malaysia; Lindauer and Sabot (1983) on Tanzania; Psacharopoulos (1983) on Greece, Portugal, Brazil, Colombia and Malaysia; Mohan (1986) and Psacharopoulos, Arriagada and Velez (1987) on Colombia, and van der Gaag and Vijverberg (1987) on the Cote d'Ivoire.  -2- fringe benefits) have strong influences on the private sector. To the extent that the wage-productivity nexus is weak in the public sector, allocative inefficiencies are generated, and modifications of public wage scales may be in order.2/ A natural question that arises is whether government workers are overpaid vis-a-vis their private sector counterparts. In other words, do workers with the same productivity traits receive equal total remuneration in the public and private sectors? The little systematic empirical evidence that exists on public-private pay comparability in developing countries shows mixed results in terms of both direction and magnitude of wage differentials. Moreover, most of the findings are based on statistical procedures that typically do not correct for selection bias in the assignment of workers to public sector versus private sector jobs. Clearly, more reliable empirical evidence is necessary for assessing policies that attempt to curb public sector growth, especially wages, or those that influence the functioning of the labor market in general. In this paper we analyse public-private wage differentials for males in Peru using a recent (1985-86) micro data base. The study considers the following questions: (1) is there a public sector wage advantage or economic rent? (2) what is the relationship between wage-determining factors, such as education and work experience, and wages in each of the sectors, and does this relationship differ between the two sectors? 2/ For a thorough discussion of a host of issues related to public sector employment and compensation in developing countries, see World Bank (1983) and Nunberg (1987).  -3- Peru is a very interesting case for analyzing public-private sector wage comparability. Since the mid 1970s it has been experiencing a severe economic crisis. For example, between 1975 and 1984 real GDP per capita (thousands, 1973 soles) fell from 29.1 to 24.5, a 16 percent drop. Average real household income dropped by 19 percent, but fell by almost 30 percent for the lowest income groups. At the same time, the annual inflation rate accelerated from 24 percent in 1975 to over 60 percent in 1977 and then stayed in the 60-70 percent range for the next 4 years. Between 1981 and 1985, the inflation rate ballooned from 73 percent in 1981, to 110 percent in 1984, and 163 percent in 1985. During the the first half of 1985, the inflation rate was over 200 percent. To add to this bleak picture, Peru's external debt as a percentage of GDP rose from about 2 percent in the mid 1970s to over 7 percent by the mid 1980s, and the government deficit as a percentage of GDP increased from 4 percent in 1973 to 8.6 percent in 1975, 11 percent in 1983, and then declined to 7 percent in 1984, and to 4 percent of GDP in 1985.3/ A large part of this state of affairs can be traced to the policies of the Revolutionary Government of the Armed Forces which seized power in October, 1968 and remained in office until mid 1980.4/ During most of the 1970s, the military regimes (there were two military presidents, General Velasco, 1968-75 and General Morales-Bermudez, 1975-80), pursued expansionary fiscal and monetary policies. The results were 3/ The data are drawn from the World Bank (1985, 1987), and INE (1986). 4/ For detailed analyses of the policies pursued during the period 1968-1980, see Scheetz (1986), Thorp and Bertram (1977), and Cline (1981).  -4- large budget and balance of payments deficits. The growth in the public sector is summarized in Table 1. Table 1: Public Spending as a Percentage of GDP 1970 1973 1975 1978 1980 1983 1984 Recurrent Expenditures 16.8 29.3 32.0 38.1 45.5 51.2 42.5 Capital Expenditures 5.2 5.9 8.6 8.1 7.5 9.6 8.1 Total Expenditures 22.0 35.2 40.6 46.2 53.0 60.8 50.6 Fiscal Deficit -0.7 -4.1 -8.6 -5.6 -4.2 -11.1 -6.9 Source: World Bank, 1987 The overall expansion in public spending during the 1970s, was mainly caused by increases in recurrent spending, rather than by capital expenditures. The role of local governments was minimal in this process, and the proportion of their spending to GDP has been historically less than 1 percent. Although there are sectoral variations, a large fraction of recurrent spending is devoted to wages and salaries, particularly in the social sectors. Indeed, a large part of the increase in government spending was due to an expansion of public sector employment which rose at an annual rate of about 5 percent during the 1970s. In 1970 government employment accounted for only 7 percent of total employment; by 1980 this more than doubled, to 17 percent. Much of the growth of public employment was concentrated in Lima which, in 1980, accounted for over 40 percent of public sector workers. A civilian government assumed power under the leadership of President Belaunde in 1980, and remained in office until July 1985. During the Belaunde era, public sector expansion continued to be sustained. By 1983, public spending reached a record high of 61 percent of GDP and then declined to 51 percent in 1984. An interesting development occurred during this period:  -5 - Public employees incurred a 57 percent reduction in their real wages between 1980 and 1985, compared to a 22 percent cut for private sector wage earners. (See Table 2). Government payrolls as a percent of CDP fell from over 5 percent during the 1970s to about 3.5 percent in 1985. Since the severe economic crisis continued into the 1980s, wage employment in the private sector declined during this period as reflected by marked increases in non- agricultural unemployment and under-employment rates, by substantial decreases in manufacturing jobs, and by a large increase in non-farm self-employment (See INE, 1986). However, it is highly unlikely that public sector employment fell by very much during this period, given the existence of a stringent set of labor laws, "estabilidad laboral", in Peru. Although "hard" evidence is not available, our estimates show that the public sector accounted for 14 percent of employment in 1985/86. TABLE 2: Indices of Remuneration in Real Terms (1980 = 100) Year Public Sector Private Sector 1980 100.0 100.0 1981 89.8 101.7 1982 81.6 109.7 1983 60.6 94.0 1984 53.0 86.8 1985 42.8 78.4 Source: INE (1986), pp. 141 and 151.  -6- TABLE 3: Public and Private Log-Wage Rates in Lima and Other Urban Areas Differentiated by Age Cohort and Educational Attainment (Standard Deviations in Parentheses) Educational Log of Log of Region Age Cohort Attainment Public Wage Private Wage Lima 20-29 secondary 1.764 (.72) 1.191 (.65) higher 1.828 (.77) 1.912 (.77) 30-39 secondary 1.899 (.43) 1.680 (.62) higher 2.155 (.51) 2.172 (.73) 40-49 secondary 1.737 (.66) 1.788 (.82) higher 2.180 (.49) 2.820 (.54) Other Urban Areas 20-29 secondary 1.685 (.74) 1.195 (.70) higher 1.554 (.50) 1.457 (1.13) 30-39 secondary 1.524 (.65) 1.459 (.84) higher 2.043 (.52) 1.781 (.74) 40-49 primary or less 1.564 (.68) 1.393 (.93) 50-59 primary or less 1.716 (.44) 1.357 (.94) Given these very different developments in employment and wages, how do remuneration levels in the public and private sectors compare ? Casual comparisons show that in 1985/86, government workers appear to receive more than private sector workers: e.g., in Lima the difference in average wages is over 31 percent. Such comparisons are deceptive, however, because the background characteristics differ substantially between the two groups of workers. For example, public sector employees are considerably more educated and older. But, as Table 3 shows, even if we look at comparably educated workers of the same age cohorts, differences persist. Regression analysis  -7- shows that, keeping a variety of background variables constant, the differential is still a significant 17.6 percent in other urban areas, but is reduced to 1.6 percent in Lima. These findings appear to answer the first question above in sufficiently strong terms: observed public sector wages exceed those in the private sector. However, even these findings can be misleading because they they are based on conditional comparisons, i.e. they describe the difference in observed wages between groups of workers who have chosen (or were chosen) to be employed in a particular sector. The more meaningful comparison is one between wage offers in the two sectors, i.e. between (predicted) wages in each of the sectors unconditioned upon whether or not the worker has been selected in. (See, Gyourko and Tracy, 1986 and van der Gaag and Vijverberg, 1988). In the next section we will provide the formal arguments against the use of conditional wage comparisons based on ordinary least squares estimates. We will also present the switching regressions model that provides the wage predictions that allow one to make unconditional (and thus unbiased) comparisons. Section III describes the data and the variables used. In section IV we present and discuss the estimation results. Section V provides a discussion of the results and their policy implications. The final section provides the conclusions and some suggestions for future research. II. The Switching Regression Model Virtually all studies of public-private wage differentials are based on the widely used model of earnings determination developed by Becker (1964) and Mincer (1958, 1974). The basic postulate is that variations in earnings arise from differences in human capital, as measured by formal schooling and  post-school work experience. The earnings function yielded by this model can be augmented to include other socioeconomic factors thought to be correlated with earnings. All of the wage determining factors are combined in the vector of explanatory variables X. Let us denote the public and private sectors as sector 1 and 2 respectively, and express the corresponding wage functions as: In w1 = X81 + u1 (1) 9n w2 Xs2 u2 (2) where In w. is the natural log of wages in sector i, ai is the vector of coefficients associated with wage-determining attributes X, and ui is a disturbance term, to be further discussed below. The standard procedure to test for equality of wage structures between the two sectors is to estimate equations (1) and (2) by OLS and then to test, for any given vector X, whether the predicted wages in the two sectors are the same. The predicted wages are set equal to Xbi and Xb2 respectively, where bi and b2 are the OLS estimates of 81 and 82. An important assumption is this approach is that Eu1 = Eu2 = 0. However, this implicit assumption that workers are randomly distributed between the public and private sectors is questionable. This is especially so if wage differentials exist, in which case (after controlling for productivity- determining factors), a selection process will determine who will choose, and indeed obtain, employment in the preferred sector.  -9- Let: I= Zy + e (3) and I =1 (public sector) if I 20 (4) and I = 0 (private sector), otherwise. Thus I* is a partially observed index that describes the selection process. We observe the outcome (public or private sector job) depending on whether I is positive or negative. OLS estimation of equations (1) and (2) will provide unbiased estimates of the unconditional wages (i.e. the wage offers) in the two sectors only if E is uncorrelated with u and u2. However, if, as is quite plausible, unobserved preference or taste variables, as well as unmeasured productivity- enhancing traits, influence the selection process as well as wage determination, then this assumption is violated and wage comparisons based on OLS are misleading. Formally, the model described by equations (1) - (4) can be summarized as follows: (See Heckman, 1979) E[IW II201 = Ka + l (5) E[W2j1 <0l = xo + a2e 2 (6) where aii is the standard deviation of ui, i=1,2 a. is the covariance between u.and e e ~~~~~~~~~~1 1= f(Zy) F(Zy)  - 10 - and x f~ f(ZY) and 2 1 - F(Zy) where F and f are the normal density and cumulative distribution functions, respectively. Thus only if al = a2c = 0 will OLS yield unbiased estimates of the wage equations. We can readily test for this by estimating the model by maximum likelihood techniques. The remaining task is to determine the factors that enter the selection process. For the sake of exposition, let us suppose that private sector jobs are easier to obtain than public sector jobs -- the argument is symmetrical and the truth may well lie in the middle. The selection process involves two steps: first, a worker will determine whether or not to try to obtain a public sector job, and secondly the employer determines whether the person is chosen for the job. The selection process thus contains supply and demand elements. The worker compares the expected benefits with the cost of applying, and the public employer uses the characteristics of the applicants to select employees from a queue. Thus, in the overall selection process, the difference of wage offers will matter, at least to the worker, and consequently all variables included in the vector X are potentially relevant. Furthermore, the applicant's personal characteristics will matter, at least to the employer, and many of these are also included in the vector X. Hence, the vector of variables X will enter the switching equation but the associated vector of coefficients will not necessarily be proportional to (61-32). Characteristics other than those contained in X may also be relevant. The vector Z in (3) combines all factors that enter the selection process. Finally, it should be noted that we choose to estimate a wage-rate function (where the dependent variable is the natural log of the hourly wage  - 11 - rate) 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 earnings function is a hybrid that may confound wage rate differences and issues related to the amount of labor supplied, 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. III. Data Set and Variable Description The Data The data used in this study are drawn from an unusually comprehensive and "clean" microdata set developed jointly by the World Bank, the Peruvian Instituto Nacional de Estadistica (INE) and the Central Bank of Peru. 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 urban male wage and salary earners who were over 14 years of age and who reported positive remuneration and positive hours worked in their main occupation during the week prior to the survey. The total sample consists of 1,743 men, 1,013 in Lima, and 730 in other urban areas (OUAs).5/ Public sector workers comprise civilians employed by the 5/ Metropolitan Lima includes the environs of Lima as well as the constitutional province of Callao (the main port of Lima). OUAs are other towns and cities with a population of 2,000 or more.  - 12 - government plus 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).61 Government workers comprise 25 percent of male wage earners in Lima, and 41 percent in OUAs. Table 4 displays the definitions and measurement of the variables used in the analysis. As stated earlier, the dependent variable is the hourly wage rate, corrected for the effect of high inflation that occurred in Peru between June 1985 and July 1986.71 Cash and the value of in-kind benefits (food, housing, transportation allowances) are included in the wage rate. The set of regressors include the standard human capital variables: education which is further distinguished by level, and (potential) experience defined as (age - years of education - 6). The set is expanded to include job-specific work experience, i.e., experience in the current main occupation that was obtained in the current job or in previous (at most) two jobs. We also include, vocational training, type of school last attended (public or private), diplomas held, parental education, and marital status.8/ 6/ State enterprise employees are included in the private sector because their pay scales are not determined by the government. Z/ In the data base, workers reported: (1) the nominal value of each of the remuneration components (cash, bonuses, housing, food, transportation, clothing, 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 (RW) was calculated as follows: RW = (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 (weeks per month). 8/ For a more complete description of these variables and the reasons for their inclusion, see Stelcner, et al (1987).  - 13 - Table 4 also shows how the variables enter the equations of the switching regression model. Since the switching equation describes sectoral choice, it would be improper to include job-specific experience in this equation. Instead, age is included as a proxy for general experience. Total years of schooling, rather than years of schooling by level, are entered into the switching equation since test results indicated that the more general model did not perform better than the more restricted one. Parental education is omitted from the switching equation for the same reason.91 Marital status is added as a preference indicator of the employer to the extent that married men are perceived as being more "stable" workers who have a stronger "job commitment". 2 9/ Preliminary tests shows insignificant x -values of .98 for the Lima sample and 1.84 for OUAs, with 4 degrees of freedom.  - 14 - Table 4 Definitions of Variables Used in the Regression Analysis Entered VARIABLE DEFINITION Wage Switching equation equation Dependent Variable LNEWHCM7 Natural log of the real hourly wage rate In main occupation (Intis at June 1985 prices) Experience AGEYR Age in years X AGEYRSQ Age squared x GEXPR Years of potential work experience estimated as: X age - 6 - accredited years or schooling - years repeated GEXPRSQ Years of potential work experience squared X XOCM7 Years of job specific experience In main occupation X XOCSQM7 Years of job specific experience squared X Education and Training YRSCHL Total years of schooling X SPLYRSC1 Years of primary schooling X SPLYRSC2 Years of secondary schooling X SPLYRSC3 Yearps of post-secondary schooling X TRAIN = 1 if did a vocational training course, 0 otherwise X X DIPLOMAl = 1 if has secondary-technical diploma, 0 otherwise X X DIPLOMA2 = 1 if has post-secondary non-university diploma, 0 otherwise X X DIPLOMA3 = 1 if has university degree, 0 otherwise x x PUBSCHL = 1 if last school attended was public (state owned and X X and operated), 0 otherwise Parental Education FYR-SCHL Father's years of schooling X MYR-SCHL Mother's years of schooling X Marital Status MARITALO = 1 if currently married or cohabiting 0 otherwise X  - 15 - A Brief Profile of Male Wage Earners The summary statistics presented in Table 5 reveal that there are considerable differences between government and private sector workers, and between workers in Lima and OUAs. We highlight these as a prelude to the estimation results. First, we consider public-private sector differences within Lima. Government workers receive on average 31 percent higher wage rates. We also see that government workers are 3.9 years older, have 3.1 years more job- specific experience, and tend to be married. The difference in education is especially pronounced at the secondary and post-secondary levels: 41 percent of the government workers have some post-secondary education compared to 24 percent of the private sector workers. Also, more of them have obtained a diploma (30 percent compared to 13 percent), and they took more vocational training courses (48 percent compared to 37 percent). There are small differences in the level of parental education, and in the percentage that attended publitc school. In OUAs, the differences between public and private sector workers are more pronounced than in Lima. The average wage gap is even wider, 44 perceint, and educational differences are larger. In particular, 47 percent of government employees obtained post-secondary schooling, as opposed to only 14 percent of workers in the private sector. There are also important differences between Lima and OUAs. In both the public and the private sector, wages are higher in Lima. As ages are the same in each sector and educational attainment is less, potential work experience in OUAs is greater. The greatest differences are found in  - 16 - educational background. Public sector workers in Lima have 1.1 years more of secondary and 0.5 years more post-secondary education. Twenty-five percent of Lima's private sector workers have post-secondary education, compared to 12.5 percent in OUAs. Among government workers, this comparison is more favorable for OUAs (41 percent in Lima and 47 percent in OUAs), but not if we compare the percentage who only completed secondary education (50 percent in Lima and 33 percent in OUAs). Finally, Lima workers have better educated parents and are less frequently married. The statistics in Table 5 provide substantial evidence that there are important differences in the characteristics of wage earners between Lima and OUAs and between the public and private sectors within each of the urban areas. In the next section we explore how these attributes affect wage variations in these different labor markets.  - 17 - TABLE 5: Mean Characteristics of Wage Earners PRIVATE SECTOR PUBLIC SECTOR CHARACTERISTICS LIMA OTHER URBAN LIMA OTHER URBAN No. of observations 759 517 254 213 Real hourly wage rate 7.686 5.361 8.711 7.007 (Intis at June 1985 prices) (15.6) (6.7) (10.2) (5.5) in wage rate 1.598 1.236 1.910 1.771 (0.85) (0.92) (0.66) (0.62) Age 33.7 33.7 37.6 37.6 (12.3) (12.6) (12.4) (10.6) (Age)2 1286.4 1297.3 1564.6 1526.4 (945.3) (960.2) (1025.) (843.6) Potential work experience 18.1 20.0 20.4 21.0 (13.4) (13.9) (13.1) (12.3) (Potential work experience)2 508.8 592.8 585.6 593.5 (655.7) (714.5) (682.5) (647.5) Job specific experience 7.5 8.5 10.6 10.2 2 (8.3) (8.9) (9.2) (8.3) (Job specific experience) 126.2 153.6 195.2 172.8 (236.7) (276.5) (268.0) (231.4) Years of Schooling Total 9.2 7.4 11.0 10.5 (3.4) (3.5) (3.7) (4.2) Primary 4.8 4.6 4.9 4.8 (0.6) (0.9) (0.4) (0.9) Secondary 3.5 2.4 4.2 3.8 (2.0) (2.2) (1.6) (2.0) Post-secondary 0.9 0.4 1.9 1.9 (1.7) (1.3) (2.6) (2.3) Level of schooling completed Less Than Primary 0.010 0.012 - - Primary 0.186 0.396 0.094 0.188 Secondary Regular 0.513 0.397 0.409 0.277 Secondary Technical 0.050 0.070 0.091 0.042 Post-Secondary Non-University 0.050 0.048 0.071 0.174 University 0.191 0.077 0.334 0.300 * Accredited. The numbers In parentheses are standard deviations. Continued  - 18 - Table 5 (Continued) PRIVATE SECTOR PUBLIC SECTOR CHARACTERISTICS LIMA OTHER URBAN LIMA OTHER URBAN Diplomas Obtained Secondary technical 0.022 0.027 0.047 0.019 Post-Secondary Non-University 0.022 0.015 0.051 0.122 University 0.084 0.032 0.220 0.174 School last attended was public 0.824 0.891 0.874 0.897 Vocational Training 0.370 0.261 0.476 0.423 Father's years 5.8 4.2 6.5 5.3 of schooling (4.0) (3.6) (4.1) (3.9) Mother's years 4.0 2.6 4.8 3.5 of schooling (3.6) (3.1) (3.5) (3.4) Married or cohabiting 0.548 0.642 0.685 0.793 IV. Estimation Results The switching regression model contains two components: two wage equations and a switching equation. Although the entire model is estimated by Full Information Maximum Likelihood (FIML), the estimates of each of the components will be discussed separately. The Wage Equations The purpose of the study is to determine to what extent wage structures differ between the public and private sectors. In the Appendix we report the estimates of the switching regression model that allows all wage equation coefficients to differ. The question is whether part or all of these coefficients are equal. Before answering this question we need to know whether it is valid to combine the sample of Lima workers with that of workers residing in other urban areas. A likelihood ratio test on the unrestricted  - 19 - model, where we allowed for a regional dummy variable in the pooled estimation, rejected the pooling soundly.i