Earnings, occupational choice, and mobility in segmented labor markets of India

WDP-154 1154 N~ World Bank Discussion Papers Earnilngs, Occupati.onal Choi'ce, and Mobi''ity in Segmented Labor Markets of India Shahidur R. Khandker FiLE COPYD  Recent World Bank Discussion Papers No. 96 Household Food Security and the Role of Women. J. Price Gittinger and others No. 97 Problems of Developing Countries in the 1990s. Volume I: General Topics. F. Desmond McCarthy, editor No. 98 Problems of Developing Countries in the 1990s. Volume 11: Country Studies. F. Desmond McCarthy, editor No. 99 Public Sector Management Issues in Structural Adjustment Lending. Barbara Nunberg No. 100 The European Communities' Single Market: The Challenge of 1992for Sub-Saharan Africa. Alfred Tovias No. 101 International Migration and Development in Sub-Saharan Africa. Volume 1: Overview. Sharon Stanton Russell, Karen Jacobsen, and William Deane Stanley No. 102 International Migration and Development in Sub-Saharan Africa. Volume II: Country Analyses. Sharon Stanton Russell, Karen Jacobsen, and William Deane Stanley No. 103 Agricultural Extensionfor Women Farmers in Africa. Katrine Saito and C. Jean Weidemann No. 104 Enterprise Reform and Privitization in Socialist Economies. Barbara Lee andJohn Nellis No. 105 Redefining the Role of Govemment in Agriculturefor the 1990s. Odin Knudsen,John Nash, and others No. 106 Social Spending in Latin America: The Story of the 1980s. Margaret E. Grosh No. 107 Kenya at the Demographic Turning Point? Hypotheses and a Proposed Research Agenda. Allen C. Kelley and Charles E. Nobbe No. 108 Debt Management Systems. Debt and International Finance Division No. 109 Indian Women: Their Health and Economic Productivity. Meera Chatterjee No. 110 Social Security in Latin America: Issues and Optionsfor the World Bank. William McGreevey No. 111 Household Consequences of High Fertility in Pakistan. Susan Cochrane, Valerie Kozel, and Harold Alderman No. 112 Strengthening Protection of Intellectual Property in Developing Countries: A Survey of the Literature. Wolfgang Siebeck, editor, with Robert E. Evenson, William Lesser, and Carlos A. Primo Braga No. 113 World Bank Lendingfor Small and Medium Enterprises. Leila Webster No. 114 Using Knowledgefrom Social Science in Development Projects. Michael M. Cernea No. 115 Designing Major Policy Reform: Lessonsfrom the Transport Sector. Ian G. Heggie No. 116 Women's Work, Education, and Family Wefare in Peru. Barbara K. Herz and Shahidur R. Khandker, editors No. 117 Developing Financial Institutionsfor the Poor and Reducing Barriers to Accessfor Women. Sharon L. Holt and Helena Ribe No. 118 Improving the Performance of Soviet Enterprises. John Nellis No. 119 Public Enterprise Reform: Lessonsfrom the Past and Issuesfor the Future. Ahmed Galal No. 120 The Information Technology Revolution and Economic Development. Nagy K. Hanna No. 121 Promoting Rural Cooperatives in Developing Countries: The Case of Sub-Saharan Africa. Avishay Braverman, J. Luis Guasch, Monika Huppi, and Lorenz Pohlmeier No. 122 Pe!formance Evaluationfor Public Enterprises. Leroy P. Jones No. 123 Urban Housing Reform in China: An Economic Analysis. George S. Tolley No. 124 The New Fiscal Federalism in Brazil. Anwar Shah No. 125 Housing Reform in Socialist Economies. Bertrand Renaud (Continued on the inside back cover.)  154 ~ 1z1World Bank Discussion Papers Earnings, Occupational Choice, and Mobility in Segmented Labor Markets of India Shahidur R. Khandker The World Bank Washiington, D.C.  Copyright e 1992 The International Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing February 1992 Discussion Papers present results of country analysis or research that is circulated to encourage discussion and comment within the development community. To present these results with the least possible delay, the typescript of this paper has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data induded in this publication and accepts no responsibility whatsoever for any consequence of their use. Any maps that accompany the text have been prepared solely for the convenience of readers; the designations and presentation of material in them do not imply the expression of any opinion whatsoever on the part of the World Bank, its affliates, or its Board or member countries concerning the legal status of any country, territory, city, or area or of the authorities thereof or concerning the delimitation of its boundaries or its national affiliation. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to photocopy portions for classroom use is not required, though notification of such use having been made will be appreciated. The complete back]ist of publications from the World Bank is shown in the annual Index of Publications, which contains an alphabetical title list (with full ordering information) and indexes of subjects, authors, and countries and regions. The latest edition is available free of charge from the Distribution Unit, Office of the Publisher, Department F, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'Iena, 75116 Paris, France. ISSN: 0259-210X Shahidur R. Khandker is a research economist in the Women in Development Division of the World Bank's Population and Human Resources Department. Library of Congress Cataloging-in-Publication Data Khandker, Shahidur R. Earnings, occupational choice, and mobility in segmented labor markets of India / Shahidur R. Khandker. p. cm. - (World Bank discussion papers; 154) Includes bibliographical references. ISBN 0-8213-2062-9 1. Labor market-India-Bombay. 2. Poor-Employment-India- -Bombay. 3. Occupational mobility-India-Bombay. 4. Wages-India- -Bombay. I. Title. II. Series. HD5820.B6K46 1992 331.12'0954'7923-dc2O 92-3419 CIP  Hii ABSTRACT This paper, using labor market survey data from Bombay, attempts to identify factors that determine men's and women's earnings, occupational choices, and mobility in segmented labor markets of India. The paper develops a model that considers three categories of labor-protected wage, unprotected wage, and self-employment-representing three different forms of labor market segmentation according to the type of labor contract and job vulnerability. The results indicate the presence of labor market segmentation; however, human capital variables such as education and training have important influence on both sectoral job allocation and worker's income and occupational mobility. Thus, policies directed to raise the productive capacity and employment levels of the poor may help alleviate poverty. The labor market outcomes, however, vary by gender. Women are less paid, less mobile, and more occupied in the unprotected wage sector than men. Men have higher education and so can more easily move to the protected wage employment or better remunerated self-employment than women endowed with lower levels of education. Men have better access to credit than women; thus, self- employed women are more constrained by lack of capital than self-employed men in raising their productivity. Policies such as structural adjustment programs which, inter alia, improve competition, reduce regulatory barriers to employment creation in the protected wage sector, and strengthen fiscal measures to accelerate growth are perhaps necessary but not sufficient for poverty alleviation. This is because, these measures cannot improve the conditions of the poor, especially women, who are trapped in the unprotected wage sector because of low-levels of education and other forms of human capital. More targeted approach is perhaps required. It is therefore necessary to complement structural adjustment by human capital development programs that improve productive capacity of the poor as well as institution strengthening measures that remove the legal and other regulatory constraints to employment expansion or job mobility. These measures are expected to benefit the poor to gain access to protected wage jobs and also provide an efficient allocation of employment.  iv ACKNOWLEDGEMENTS The author wishes to thank Barbara Herz, Dipak Mazumdar, Paul Schultz, Steve Talbot, and seminar participants at World Bank and Harvard University for helpful comments on earlier drafts of this Paper. John de New provided excellent computer programming and research assistance. Stella David prepared the text of this Paper. The author appreciates partial financial support for the Government of Norway, Ministry of Development Corporation. He is also thankful to the Asian Regional Team for Employment Promotion (ARTEP) for releasing the Bombay labor market survey data.  v FOREWORD Unlike their counterparts in developed nations, women and men in developing countries work more in the informal sector of the economy where job continuity is uncertain and returns to labor are low. Women, however, are disproportionately represented in this vulnerable sector. Because jobs in the protected sector (where job continuity is guaranted by laws) are in short supply relative to demand, personal contacts and "inside" information are important influences on entry to the sector. The result is that inter-sectoral mobility is restricted for many of the poor, especially women. Research is needed why women are so numerous in the unregulated sector, what influences women's inter-sectoral mobility (if any), and the scope for policy interventions to improve the plight of the poor, including women. This paper makes an attempt to answer some of these questions using household survey data from Bombay, India. India provides an interesting case study as workers' job vulnerability is not only affected by economic conditions but also by non-economic influences such as caste and religion. The paper suggests that, although non-economic factors influence market outcomes, policy makers can improve income and labor mobility of workers, especially women, in the informal sector by investing in their human capital. Poor workers may also gain if the degree of protection granted to the formal sector is reduced. This paper -- a product of the Women in Development Division, Population and Human Resources Department -- is part of a larger effort in the Bank to determine how women's productivity can be improved by enhancing their access to education, training, credit, health care, and other public services. Ann 0. Hamilton Director Population and Human Resources Department   vii TABLE OF CONTENTS 1. INTRODUCTION ............................... 1 II. OCCUPATIONAL CHOICES, EARNINGS AND MOBILITY: SEGMENTATION THEORIES REVISITED. 3 III. JOB CHARACTERISTICS AS MEASURE OF MARKET SEGMENTATION AND WORKERS' EARNINGS AND LABOR MOBILITY: AN ECONOMIC FRAMEWORK. 5 IV. STUDY AREA AND WORKERS CHARACTERISTICS. 8 V. DETERMINANTS OF EARNINGS IN SEGMENTED LABOR MARKETS OF BOMBAY .11 Impact of human capital variables .12 Impact of worker's background variables .13 Impact of job characteristics .13 VI. SECTOR SELECTION AND ITS IMPACT ON WAGES AND LABOR SUPPLY BEHAVIOR . ......................................... 14 Determinants of sectoral job distribution ............................... 17 Sector allocation bias: Test of labor market segmentation ............................................... 19 Determinants of labor supply ........... ........................... 19 VII. DETERMINANTS OF OCCUPATIONAL AND INCOME MOBILITY .... ....... 20 Occupational mobility . ......................................... 21 Income mobility .............................................. 22 VIII. CONCLUSIONS .23 Bibliography 26  viii Tables Table 1. Descriptive Statistics .................................... 29 Table 2. Determinants of Protected Wage ................................ 31 Table 3. Determinants of Unprotected Wage ............. ................ 32 Table 4. Determinants of Hourly Earnings from Self-employment .33 Table 5. Participation Function Estimated by Multinomial Logit for Male Subsample .34 Table 6. Participation Function Estimated by Multinomial Logit for Female Subsample .35 Table 7. Predicted Male Participation by Sample Characteristics .36 Table 8. Predicted Female Participation by Sample Characteristics .................................... 37 Table 9. Employment Function (Hours Worked) Estimation ...... .............. 38 Table 10. Censored Regression (Tobit) Estimates of Occupational Mobility .39 Table 11. MNL Estimates of the Determinants of Occupational Mobility Among the Unprotected Wage Workers 40 Table 12. Random Effects Estimates of Income Mobility ........ .............. 41 Appendices Table AI. OLS Estimates of Wages/Earnings with Sample Selection Bias Correction .42 Table A2. Employment Function (Hours Worked) Estimation With Sample Selection Bias Correction .43 Table A3. Probit Estimates of the Determinants of Occupational Mobility Among the Self-employed .44 Table A4. Probit Estimates of the Determinants of Occupational Mobility Among the Wage Workers .45  1 I. INTRODUCTION 1.01 Labor market segmentation is an important concern for policymakers in many developing countries. It produces institutional barriers to the smooth operation of markets, causing inefficiencies that can seriously frustrate development. In particular, when the labor market is highly segmented and inter- sectoral labor mobility is restricted, the market process can thwart policies, especially those relying on market forces such as structural adjustment programs, that are aimed at alleviating poverty. 1.02 In India, about 25 percent of the population live in urban areas, where 40 percent live below poverty level (World Bank 1989). Evidence indicates that many Indian urban workers are trapped into the unprotected segment of the labor market where the incidence of poverty is high (Harris 1989).' The evidence also suggests that the disparity between the earnings of protected and unprotected workers is widening, and the earnings of unprotected workers are stagnating. 1.03 The growing difference in the wages of protected and unprotected workers reflects the influence of a large labor supply into the unprotected segment of the market as well as the inability of the unprotected workers to move into the protected sector. Because of continuing in-migration from rural areas and because the protected wage sector cannot readily absorb these migrant workers, the workforce in the unprotected sector is expanding and the inter-sectoral wage differentials persist as institutional barriers restrict workers' inter-sectoral mobility. Thus, efforts to eliminate urban poverty must reduce market segmentation to promote labor mobility and thereby ease poverty. This requires a better understanding of the extent and influence of market segmentation on occupational choices, earnings, and mobility of poor urban workers. 1.04 The purpose of this paper is to estimate the occupational choice, earnings, mobility, and labor supply behavior of men and women among low-income households in Bombay. A recent study suggests that about half of Bombay's 10 million people live in slums and shanty towns. Most of these people are poor and some 60 percent of them work in the unprotected sector of the labor market (Acharya and Jose 1990). Who are these unprotected workers and what is the extent of their poverty? Would better education and other forms of human capital investment help them break out of poverty? How do women fare in the segmented labor market process? These questions matter not only for policy purposes, but also for understanding the dynamics of urban poverty. 1.05 How does labor market segmentation occur in developing countries such as India? Evidence suggests that caste, religion, gender, location, labor laws, and union are some of the influences which create and perpetuate a segmented labor market (Harris 1989; Bardhan 1989; Rogers 1989). The segmented labor market (SLM) view argues that these influences make jobs heterogeneous and affect the incidence of urban poverty via their influence on differential access to jobs based on a worker's gender, caste or education. In this view, skill differentials among individuals cannot overcome the institutional barriers. Thus raising the productive capacity of the poor is not enough to reduce the impact of segmentation on income and labor mobility. In contrast, the neo-classical labor market (NLM) view 'A job is protected if it is secured from market forces either through restrictions on entry or by contracts and legal constraints. In contrast, a job is unprotected if it is characterized by insecurity and/or irregularity and hence its continuity is uncertain (Rogers 1989). The casual or contract jobs belong to the unprotected category, while the regular job belongs to the protected category.  2 argues that the socio-institutional factors reflect labor rather than job heterogeneity and thus a way to remove poverty is to provide individuals with better education, skills, and training (Taubman and Wachter 1986). 1.06 Although it is empirically difficult to make a clear distinction between competing models of labor market segmentation (Heckman and Hotz 1986), this paper attempts to identify the relative influence of the human and non-human capital influences on the pattern of occupation, earnings and labor supply of women and men. Identification of non-human capital influences may provide information for policy changes to improve earnings, occupations, and hence the poverty levels of urban population. Similarly, identifying the returns to education and experience in different types of work will help show whether human capital investment can induce income and labor mobility in segmented labor markets. 1.07 This paper focuses only on people in the low-income strata who live in urban slums of Bombay to study the impact of labor market segmentation on urban poverty, occupational choice and mobility, and labor supply behavior. Traditionally, urban poverty has been explained as an outcome of rural poverty. Rural poor attempt to escape rural unemployment or under-employment by migrating to towns and mostly settle in slums.2 This potential pool of unskilled labor increases competition in the urban labor market and helps the urban employers maintain an unprotected workforce (Rogers 1989). The persistence of rural poverty thus puts pressure on urban labor processes and contributes to the growth of a segmented labor market. Understanding urban poverty, therefore, means understanding the labor market choices of these low-income households and the factors that influence their earnings, occupational choices, and mobility. An understanding of how the urban labor markets operate and interact with the occupational choice and earnings of low-income households may help understand better the role of labor market segmentation and its impact on urban poverty. 1.08 The results based on Bombay labor market survey indicate the presence of labor market segmentation. Nevertheless, human capital variables have important influence on workers' earnings, occupational choices, and mobility. Thus, labor market outcomes and urban poverty are not preordained by market and institutional structures. Although employers play an upperhand in the distribution of employment and wages, the results support the view that policies directed to raise the productive capacity of the poor may help alleviate poverty. 1.09 The paper is organized as follows. Section II reviews both the SLM and NLM views on labor market segmentation and concludes that individual's job characteristics are better indicators of labor market segmentation than segmentation usually identified on the basis of wage or production system. Section III outlines a model for analysing occupational choices, earnings, and mobility of workers in segmented labor markets. Section IV discusses the survey data from Bombay and their characteristics. Section V presents the results of the determinants of earnings of different types of workers. Section VI discusses the estimates of occupational choice and its influence on earnings and labor supply. Section VII analyses the income and occupational mobility of the poor urban workers of India. Section VIII concludes the paper with policy implications. 'This, of course, does not mean that rural out-migration takes place only for those who wish to escape the wrath of rural poverty. Self-selected migration from rural to urban areas may occur for educated workers and landed elite who move into cities for better employment and living conditions.  3 II. OCCUPATIONAL CHOICES, EARNINGS AND MOBILITY: SEGMENTATION THEORIES REVISITED 2.01 How to model labor market segmentation? There are two models that try to explain labor market segmentation and its impact on labor market outcomes such as distribution of wages and employment, and income and occupational mobility of workers. The NLM model emphasizes that the labor market outcomes reflect the interactions of profit-maximizing behavior of firms and the utility-maximizing behavior of workers. In this view, supply and demand forces clear the market, and thus earnings, occupational choice and mobility reflect on-job training and other investment in human capital. The model thus tends to emphasize differences among people, rather than among jobs, as a determinant of the distribution of jobs and income (Dickens and Lang 1985). 2.02 In contrast, the SLM view suggests that labor market segmentation results more from differences in jobs than from differences in people. Differences in jobs arise because of tastes and institutional rigidities that prevent the market from operating as the NLM view suggests (e.g., Doeringer and Piore, 1971; Piore, 1983). Various categories of jobs emerge because job access depends more on the implicit job-specific labor contract than on the skill of the worker? Although the SLM view does not offer any concrete model, it emphasizes that demand and supply forces cannot compete away the wage differentials across sectors; nor skill differentials among individuals can overcome the institutional barriers. It also argues that human capital investment is not enough to promote income and employment of the poor; it is thus also necessary to understand the nature and causes of labor market segmentation and its impact on the poor. 2.03 What is important from the SLM view is that since good quality jobs are short in supply relative to demand, they are rationed and thus contact and influence rather than market forces interact with the market outcomes. Furthermore, because institutional barriers and preferences restrict entry into good jobs, labor market segmentation produces labor immobility across sectors, especially among the poor. A large literature has been developed along the lines of SLM view which casts doubts on the efficacy of the NLM view of the labor market (see Taubman and Wachter 1986). 2.04 In response to SLM criticisms, the NLM model allows for some labor market segmentation. It allows geographical and biological factors (e.g., age of the worker) to make labor imperfect which then creates market segmentation. It also allows institutional factors such as labor unions and government laws or minimum wages to cause market segmentation which arises because of market's response to externalities created by these institutional influences (Williamson, Wachter, and Harris, 1975). However, the presence of such non-human capital influences distorting the market process and thus affecting market outcomes does not contradict the basic concept of the maximizing behavior of the firms or the individuals. 2.05 Although the controversy continues over the sources of labor market segmentation and their theoretical underpinnings, important empirical questions still remain: a) what causes labor market 3Broadly speaking, jobs can be classified as being in the primary sector when they are better paid, with good working conditions and better opportunities for advancement. On the other hand, jobs can be grouped in the secondary sector when they are low paid, with bad working conditions and less opportunity for advancement. In the similar fashion, one can identify other dual forms of labor market segmentation, such as high versus low wage markets, formal versus informal sectors, and public versus private jobs.  4 segmentation; b) how much do the causes of market segmentation affect the distribution of wages and employment; c) how strong are these effects on occupational and income mobility of workers; d) how many people are trapped in low paying jobs, and e) whether there is any room for policy interventions to improve the plight of the poor. Empirical research addressing these questions may help understand what determines the distribution of income and employment and the policy options for eliminating poverty. 2.06 Empirical verification of such issues is not an easy task. It depends on how the labor market segmentation is viewed. The dual labor market segmentation based on either production system (e.g., primary vs. secondary, public vs. private, and formal vs. informal) or wage (e.g., high vs. low wage) is not mutually exclusive and so remain unsatisfactory (Rogers 1989). For example, in the formal segment of labor market, informal arrangements such as contractual employment may emerge which are not different from the casual work of the informal sector. Similarly, in the informal sector where informal arrangements dominate, some formal labor arrangements may exist for certain categories of labor. 2.07 Therefore, a satisfactory way of identifying labor market segmentation should not be based on the characteristics of the production system or wages but on the job characteristics of an individual worker. Vijverberg and Van der Gaag (1990) conduct a test of labor market segmentation in the private wage sector of Ghana where they use job characteristics to predict the formality index of a worker's job and then find its impact on her or his productivity.4 They find labor market segmentation even in the private wage sector. To the extent that job characteristics may involve a worker's choice and hence the formality index may be jointly determined with wages, their study may suffer from potential self-selection bias in characterizing labor market segmentation and its impact on wages. Furthermore, identifying the formality index and its impact on wages has very little policy relevance. 2.08 However, as this study suggests, individual-specific job characteristics are important sources of labor market segmentation. Rogers (1989) has identified five categories of labor market segmentation based on individual's "labor vulnerability, protection, and control over work."5 The job vulnerability approach (based on worker's job characteristics) to sort out labor market segmentation identifies not only different forms of labor market segmentation but the level of poverty among various groups of urban workers as well. This approach also has important policy implications. A number of ILO-sponsored labor market studies document how urban.poverty can be identified with worker's labor vulnerability (Rogers 1989). 'They treat a host of job characteristics such as whether or not minimum wage law applies, the number of workers at the place of employment, whether a worker is unionized, whether she or he receive social security benefits, housing allowance, clothing allowance, bonus, and retirement pension, etc. 5Rogers' (1989) five categories of work are protected wage work (where jobs are protected from market forces by restrictions on entry), competitive regular wage work (where competition exists but yet jobs are secured through experience), unprotected wage work (which includes both casual and contract jobs which are not secured), self-employment and family labor in productive small-scale production, and employment in marginal activities such as hawking, shoe-shining, etc.  5 2.09 Identifying market segmentation based on an individual's job characteristics has also a number of attractive properties: a) it is less ambiguous than the approach based on type of wage/industry/production system; b) it provides a framework to identify the relative influence on a worker's sectoral job allocation and wage of individual characteristics and the characteristics of the industry where she or he is employed; and c) more importantly, it facilitates the use of a sectoral selection bias test of market segmentation that sorts out who plays an upperhand in the distribution of employment and wages, workers or employers.6 III. JOB CHARACTERISTICS AS MEASURE OF MARKET SEGMENTATION AND WORKERS' EARNINGS AND LABOR MOBILITY: AN ECONOMIC FRAMEWORK 3.01 A worker's job characteristics are important sources of labor market segmentation. It is, however, important to distinguish two types of job characteristics: individual-specific job characteristics (indexed J) and industry-specific job characteristics (indexed K). Individual-specific job characteristics include variables such as whether or not a worker receives social security benefits or retirement pension etc. which produce labor heterogeneity and hence labor market segmentation.' In contrast, the industry- specific job characteristics such as whether the production system of an industry is formal, whether labor laws apply, and the number of workers work in an industry may produce both job and labor heterogeneity and hence labor market segmentation. The individual-specific job characteristics may reflect a worker's preference, which the industry-specific job characteristics proxy the preference of an employer. Although both types of job characteristics influence wages, the difference is that J-job characteristics may involve a worker's decision, while K characteristics may be predetermined.8 3.02 More formally, an industry with given K characteristics provides an array of jobs and their characteristics that every worker does not need to choose the same bundle of J-job characteristics. Based on own charateristics and tastes, a worker draws a bundle of job characteristics that maximize her or his utility which is different from another worker's chosen bundle. These job characteristics are then sources of a worker's non-pecuniary returns and capable of producing implicit labor markets within an industry with K characteristics. This idea should then be the driving force for modeling labor market segmentation. The idea can be traced to the work by Atrostic(1982) and Rosen(1974). Atrostic argues that workers derive non-pecuniary returns (i.e., job satisfaction) from individual-specific job characteristics which they choose from an array of jobs and their characteristics provided by an industry. Rosen, on the other hand, argues that the chosen job characteristics vary across workers and may create implicit job markets within a job market. 6For more discussion on this sectoral selection bias test, see Section III below. 'lndividual-specific job characteristics may also create job heterogeneity as job and labor heterogeneity are really indistinguishable. SK characteristics may also involve a worker's choice if she or he decides which industry to work for where each industry has an array of K-job characteristics. In this case, a dual rather than a single decision-making process is involved. However, for simplicity, this case is ignored here.  6 3.03 For simplicity, we assume that every individual participates in the labor market. Assume that each individual worker derives utility from composite bundle goods (X), leisure (L), and job characteristics (J): (1) U = U(X, L, J; A) where A is an array of her or his background characteristics such as age, gender, caste or ethnic race, religion, etc which determine the curvature of the utility function. 3.04 Also aasume that the worker faces the following constraints in her or his attempt to maximize the utility function (1): (2) H =T - L); (3) W = W(J;K, M, A); (4) PX =WH + I; where H is hours worked per week, T is total weekly hours available, M is a vector of human capital endowments, P is the price of X goods, W is wage, and I is unearned income. Equation (2) is a time constraint. Equation (3) -- a wage equation -- states that an individual's wage is not given but depends on job characteristics (J) chosen and predetermined human capital endowments (M), industry characteristics (K) and background characteristics (A). The predetermined M, K, and A variables determine the level of the wage-job characteristics frontier from which the specific bundle(J) is chosen.' 3.05 The maximization of utility function (1) subject to the constraints (2)-(4) yield the following optimum conditions: (5a) U,, = TP (5b) UL = TW (5c) Uj = T(L-T)Wj (5d) I - PX + W(J; K,M,A)(T-L) = 0 where T is the marginal utility of income (positive), U; (i=X,L,J) is the marginal utility of ith good and Wj is the marginal return of J-job characteristics to individual's productivity. 3.06 Thus, a worker when selects J-job characteristics faces a wage which is determined by her or his human capital and other background characteristics as well as the industry characteristics. In maximizing her or his utility, a worker equates the marginal utility of leisure with the imputed cost of her or his labor evaluated at the wage rate given by these job characteristics. The worker also equates the marginal utility of J-job characteristics with the imputed value of her or his labor supplied and evaluated at the marginal return of J-job charateristics to productivity. For an optimum, the worker will choose J-job characteristics until they yield negative return to her or his productivity provided that the worker derives 9The endogeneity of wage due to J-job characteristics makes the budget constraint non-linear which contrasts with the standard NLM assumption of the linear budget constraint. The endogeneity of wage can also occur if wage depends on the quantity of labor supply. Moffit(1984) has shown that when the budget constraint becomes nonlinear because of endogeneity of wage, the true wage elasticities are lower than those derived under the assumption of linear budget constraint for given market wage.  7 job satisfaction (i.e., positive marginal utility) from J-job characteristics. 3.07 Under suitable conditions, the above model yields a system of reduced-form equations for labor supply (H), the chosen bundle of job characteristics (J), and consequently wages (W): (6a) H = H(P,I,T,K,M,A) (6b) J = J(P,I,T,K,M,A) (6C) W = W(P,I,T,K,M,A) 3.08 The system of equations (6) shows that job status (sources of market segmentation), labor supply and wages are jointly determined by a number of exogeneous individual, household, and industry characteristics that affect an individual's production and consumption of income. 3.09 Note that the impact of wage determining variables such as human capital (M) has two effects on wages - one is the direct effect for given J and the other is the indirect effect via changing J. In other words, 6W/8M = 6W/6Mlj + (6W/6J).(6J/6M) Similarly, the impact of M on hours of work can be decomposed in two different ways: SHlM = (bH/6M j + (6H/6J). (6J/6M) or 6H/6M = (6H/6W).(6W/6M) + (6HL6J.(6J/OM) 3.10 The above model (6) can be estimated in two ways: either we estimate the reduced-form by ordinary least squares (OLS) or we impose suitable exclusion restrictions and estimate (6) by instrumental variable method. The instrumental variable method is used to estimate the choice of J-job characteristics and their impact on wage and labor supply. This involves the following: a) for given J, estimate the wage function (3), assuming that J-job characteristics are given (section V below); b) estimate the distribution of J-job characteristics among workers and its impact on their wages (section VI); and c) estimate the impact of wages on labor supply (section VI). This paper examines the income and labor mobility of workers by looking at the changes in J and W over time. This paper however, estimates mobility in wages and J-job characteristics using the reduced-form approach (section VII). 3.11 To operationalize the model, the J vector needs to be characterized. The approach used here to characterize the J vector is Roger's (1989) job vulnerability approach. J-job characteristics are classified into three major occupational groups implying three forms of labor market segmentation which are protected wage, unprotected wage, and self-employment."0 In other words, J vector approximates three '5he protected wage sector consists of regular and competitive wage work which is secured by the contract of the job where entry is difficult. The unprotected wage sector comprises both casual and contract jobs where entry is easier but the job is not secured and well paid. Self-employment consists of activities where entry is not restricted but entry depends on the availability of investable funds. Unlike the protected wage employment, self-employment is not secured, but its continuity is not uncertain either like a job in the unprotected wage sector. This classification of market segmentation is available from the Bombay labor market survey data. When such a classification is not available, an alternative procedure may be used. Following Atrostic  8 occupational choices of workers signalling a worker's level of job security. Note that J vector then becomes discrete rather than continuous as proposed earlier. This approximation of J-characteristics is simple but capable of identifying the role of market segmentation in the determination of workers' earnings and labor mobility. Furthermore, such characterization has a clear-cut policy implication for poverty alleviation. 3.12 The paper includes both human capital (M) and non-human capital variables(e.g., vectors A and K) to explain sectoral choices, earnings, and mobility among low-income workers of Bombay. The non- human capital A vector measures factors such as religion, caste, gender, and place of birth. The K vector implies industry characteristics such as scale, and whether labor laws apply. The human capital (M) variables include age (proxy for experience), education and training. The difference between NLM and SLM models is that the NLM model emphasizes on the powerful role of human capital variables to explain wages, sectoral choices and mobility of workers, while the SLM model considers them largely irrelevant for these outcomes. Thus, according to NLM, workers can influence the market outcomes, while the SLM view emphasizes that employers determine workers' sectoral allocation and wages. To determine whether labor market segmentation exists, we thus need a test that sorts out who determines the market outcomes: workers or employers. 3.13 Two tests are conducted to distinguish who -- workers or employers -- determine the market outcomes. First is the sectoral difference test where appropriate tests are conducted to examine whether the distribution of J-job characteristics (i.e., in this paper three types of employment) and the accompanying wages are inflexibly given. The purpose of this test is to determine a) whether a single wage of job-status equation characterizes the entire labor market and b) whether workers' charateristics (e.g., A and M) and/or industry characteristics (K) influence the market outcomes. Second, if the sectoral difference test suggests that sectoral wage and job allocation are non-random, then a sector selection bias test is applied to determine whether non-random allocation of employment influences a worker's wage in a given sector (Gindling 1991). The purpose of this test is to determine whether workers are able to self-select employment in a sector where their expected wages are high. Thus, if the test shows that sectoral selection bias exists in wage determination, this supports the hypothesis of no market segmentation. Market segmentation, however, exists when non-random job allocation is determined by employers and hence, sector selection bias does not exist in the determination of wages. IV. STUDY AREA AND WORKERS CHARACTERISTICS 4.01 Bombay is the largest urban center in India with about 10 million people. More than half of its population live in slums and shanty towns. The present study is based on a random sample survey of 2,192 workers who live in the "recognized' slum areas of Bombay.'" The survey was conducted from January to June of 1989. Among the 2,192 workers, 1,469 are males and 723 are females. Table 1 (1982), one can use the principal component analysis to aggregate individual-specific J-job characteristics into a single, composite, measure of job characteristics. Atrostic's approach is not different from Vijverberg's and Van der Gaag's approach of estimating the formality index. "There are 615 recognized slums in Bombay city. The recognized slums are those which have been in existence for at least 10 years and their residents pay taxes to the municipal authority for and have legal entitlement to space. The sample survey covers 30 such slums, i.e., about 5 percent of the recognized slums in Bombay city (Acharya and Jose, 1990)  9 presents the descriptive statistics of some of the salient features of the dataset. 4.02 According to individual occupational status, among the male workers 33 percent work as self- employed, 20 percent are unprotectd wage (i.e., casual or contract) workers, and 47 percent are protected wage (i.e., regular wage) employees. Similarly, among women, 27 percent work as self-employed, 59 percent are unprotected wage workers, and only 14 percent are protected wage workers. Thus, more women work in the unprotected wage sector than men. 4.03 The distribution of these workers by their first job status is noteworthy. Among currently self- employed men, originally only 31 percent were self-employed, 50 percent worked as casual-contract workers, and some 19 percent were employed in the protected wage sector. In contrast, among self- employed women, initially some 65 percent were self-employed, while 33 percent were casual or contract workers, and only 2 percent were regular wage workers. 4.04 Among men who are at present casual or contract workers, some 79 percent were initially employed in this unprotected wage sector, while some 15 percent were employed in the protected wage sector, and 6 percent were self-employed. Among women who currently work in the unprotected wage sector, some 90 percent were initially employed in this sector and some 5 percent each switched from the self-employed and protected wage sectors. 4.05 Among men who are currently employed in the protected wage sector, only 35 percent were originally employed in this sector, while 61 percent moved from the unprotected wage sector and some 4 percent came from self-employed sector. In contrast, among women regular wage workers, some 57 percent were employed in their first job in this sector and some 42 percent came from the unprotected wage sector and only 1 percent came from the self-employed sector. These statistics do suggest that inter-sectoral mobility is not restricted but also that men are more mobile than women. 4.06 In terms of job mobility measured by the number of times a worker changes job (after adjusting for the amount of time she or he works in the labor market), we find that job mobility among women and men who work in the unprotected wage sector is higher than that for their counterparts employed in other sectors. For example, men who are employed in the unprotected wage sector change their job twice in every ten years, while their counterparts employed in other sectors change job only once during the same time. However, men are on average more mobile than women. Men workers of each type work longer than their counterpart women workers. In terms of hourly wage rate, men earn more than women in each sector even though men and women are of similar age. Women receive 69 percent of men's earnings in the protected wage sector, 61 percent in unprotected wage sector, and 70 percent in self-employment. 4.07 If per capita income (total household income divided by family size) is a proxy for poverty level, the data clearly supports the claim that workers -- both men and women -- who work in the unprotected wage sector are the poorest among the poor in urban India (Harris 1989). 4.08 Both male and female workers vary in education, training, and other background variables as well as in their job characteristics. There are also noticeable intra-sectoral differences in these characteristics of men and women. Men are, on average, more educated than women. For example, among the self- employed, 51 percent of women have less than primary schooling compared with only 17 percent for men. Among the unprotected wage workers, 32 percent of women have secondary level schooling compared to 38 percent for men. Some 32 percent of the protected wage workers of both genders had completed secondary schooling. So the educational differential between men and women falls as the job  10 security improves. 4.09 Men are more trained than women but this differential again falls with job security. Among the self-employed, 57 percent men are trained compared to 20 percent for women, while among the unprotected wage workers, 51 percent men are trained compared with 16 percent for women. in contrast, some 44 percent men are trained compared to 34 percent for women in the protected wage sector. Occupational choices of men and women also vary by caste: women of scheduled castes are more self- employed than men of the same caste background. 4.10 Among the self-employed, only 17 percent of men were born in Bombay compared with some 25 percent for women. Among the unprotected wage workers, 25 percent of men are from Bombay compared with 37 percent of women. Among the protected wage workers, some 48 percent of women and only 20 percent of men were born in Bombay. Among those who have migrated to Bombay from Maharashtra, men are largely (56 percent) employed in the protected wage sector, while women (56 percent) are largely self-employed. Among the migrants from outside of Maharashtra, men account for 52 percent of the self-employed compared with only 20 percent of women. 4.11 In sum, migrant men account for 83 percent of the self-employed, 75 percent of the unprotected wage workers, and 80 percent of the protected wage workers. In contrast, migrant women account for 76 percent of the self-employed, 63 percent of the unprotected wage workers, and 53 percent of the protected wage employees. However, among the migrants who have migrated recently(i.e., within 10 years), more men and women are employed in the unprotected wage sector than in other sectors. In other words, both men and women first work in the unprotected wage sector before they move to the protected wage or self-employed sector. 4.12 Among the protected wage workers, 55 percent of men and only 11 percent of women report that some labor laws apply to their job."2 In contrast, among the casual-contract wage workers, only 9 percent men and some 1 percent women report that labor laws apply to their job. Both men and women are employed in the large-scale enterprises (in terms of number of workers) if they work in the protected wage sector. When self-employed, men operate businesses with more capital as reflected by the size of sales turnover than women. Among the protected wage workers, 43 percent men and some 22 percent women have joined the trade union. Women are less unionized than men. 4.13 The employed men are mostly married. The percentage of unmarried women is higher in the protected wage sector (40 percent) than in other two sectors (20 percent each). If the worker's father or guardian were self-employed, both women (61 percent) and men (65 percent) are more likely to be self- employed. "Labor laws include govemment legislation on matters such as security of employment, payment of wages linked to cost of living, leave facilities and social security schemes. For details of these protective labor legislations, see The Indian Factories Act, 1948, Govemment of India, New Delhi.  C, V. DETERMINANTS OF EARNINGS IN SEGMENTED LABOR MARKETS OF BOMBAY 5.01 In this section we estimate and present the wage equation (3) for three occupational groups, assuming that these choices are given. We will relax this assumption in section VI. A semilogarithmic wage function, which is now a standard practice in the literature, is fitted which relates logarithmic wage to workers' education, experience, religion, caste, location of the job, and some industry-specific job characteristics such as whether or not labor laws apply and scale or sales turnover of an enterprise. Such a hedonic price function can be estimated separately for men and women as well as jointly to assess whether these functions differ beyond the intercept between men and women. 5.02 Furthermore, the wage function can be fitted for each segment of the labor market by gender to assess how they differ among these three labor market segments."3 More formally. (7) InWj = a + B1jAJ + B2JMJ + 82JKj + ej where lnWj is the natural logarithmic wage of jth type of worker, M is a vector of human capital variables such as age, age squared, education, and training; A is vector of individual's background variables such as caste, religion, and area dummy where the worker was born; K is a vector of industry- specific job-related characteristics such as the location of the job, whether any labor law applies to job and scale of an enterprise; and c's and B's are the parameter estimates. The error terms, ej, is assumed to be independent and normally distributed. 5.03 A few additional variables, even though they are endogenous, can be introduced to the basic model (equation 7) one at a time in order to assess how they are associated with earnings of men and women across different categories of work. One such variable is whether or not the worker is a trade union member. This variable reflects a worker's decision and hence endogenous, but measures the impact of unionization on workers' productivity. Although the impact of trade union membership is therefore biased, the differences in coefficients may be suggestive of whether unions provide similar or greater benefits for men and women with different occupation. 5.04 The second variable is migration status of workers. This variable measures the timing of migration and hence implies a worker's decision to migrate which jointly determines her or his productivity. Even though its impact is hence biased, the differences in coefficients may indicate whether different timing of migration has provided similar or larger market opportunities for men and women to improve their productivity. 5.05 The migration impact may vary by the gender of the worker, as migration for women in mostly due to marriage and not for seeking job opportunities in the urban center. Thus, new migration status may appear irrelevant for women's wages, but for men's wages it would signal a worker's commitment to labor market advance. In Latin America, migration status exhibits an initial negative impact on wages, but that diminishes and becomes positive ten years after migration to the urban center (Ribe 1979). We may find a similar pattern for the impact of migration status on wages. "3The assumption of structural differences among these three sectors is tested with appropriate F-tests that test whether a single wage structure can explain wage determination in all three sectors. The Chow-test rejects the null hypothesis. This suggests that a single wage structure that explains wages in all three sectors does not exist. This provides support for the presence of segmentation.  12 5.06 Capital stock is an important factor determining productivity, especially for the self-employed. But information on this important variable is not available from the survey. We include, therefore, a proxy variable - the sales turnover of an enterprise -- to measure the impact of capital on productivity of the self-employed. However, sales turnover is clearly endogenous, because it is jointly determined with a worker's productivity, especially if she or he is self-employed. Thus, although its impact is also biased, the differences in coefficients would indicate whether capital provides similar or larger opportunities for men and women to improve productivity. 5.07 The OLS method is applied to estimate the wage equation for each category of worker by men and women and the results are reported in Tables 2 to 4. Model I(i.e., wage equation 7) does not include any controversial variable, while Model II includes sales or union, and Model III includes new migration status as additional variables to the basic model. Model I explains 29 percent of wage variations for men and 44 percent for women who are protected wage workers, some 11 percent each for men and women who work in the unprotected wage sector, and some 20 and 7 percent, respectively for men and women who are self-employed. The model's explanatory power dramatically increases for the self-employed men and women if sales turnover enters into the wage regression. 5.08 The OLS estimates provide interesting comparisons between men and women across different categories of jobs. They exhibit differences in the response patterns of men and women and the relative importance of human capital and non-human capital factors in the determination of wages. Impact of human capital variables 5.09 The results support the human capital model of NLM view that wage variations are partly explained by variations in workers' human capital such as experience and education (Becker 1964; Mincer 1974). According to Model I estimates, experience has an important infuence on earnings in the protected wage sector, at least, for men. Education beyond primary level affects men's and women's productivity in the protected wage sector and women's productivity in the unprotected wage sector. Education influences earnings for the self-employed, but only for men. Training improves both men's and women's productivity in the unprotected wage sector, but not in the self-employed sector. The results indicate that, even after controlling for the effects of job-related and other characteristics, human capital significantly affects both men's and women's productivity. 5.10 The importance of human capital varies by sector of employment, however. Schooling has higher returns for both men and women in the protected wage sector than in the unprotected wage. Schooling has the highest return for men but lowest return for women in the self-employment sector. The returns to education at the secondary level for men and women in the protected sector are, respectively 2 and 5 percent, while they are 2 percent each for both men and women employed in the unprotected wage sector.'4 Also, the returns to schooling at the secondary level is 2 percent for self-employed men. These rates are lower than those reported by Tilak (1987; p.85).'5 The return estimates, however, tell '4The education dummy variable approach is not a good measure to estimate retums to schooling, especially for the primary school level (Psacharopoulos 1981). "Tilak reports about 20 percent rates of private return for secondary schooling. These figures when adjusted for growth and unemployment are only some 4 percent and thus are consistent with the rates reported here for the low-income workers in Bombay.  13 a remarkable consistent story that is observed in other developing countries. That is, the returns to schooling are higher for women than for men at the secondary school level (Schultz, 1989; Khandker 1990). This is true if the return estimates are calculated based on the wages received from the protected and unprotected wage sectors. In the case of the self-employed sector, the returns to education are much higher for men than for women, however."6 5.11 The returns to training are also higher for men (27 percent) than for women (20 percent) employed in the unprotected wage sector. Training has no effect on wages in the protected wage sector. Moreover, training does not affect productivity of the self-employed (according to Model I), but when sales turnover enter into the wage regression, the coefficient of training becomes positive and significant. Thus, according to Model n of Table 4, the returns to training are higher for self-employed women (20 percent) than for self-employed men (11 percent). Impact of worker's background variables 5.12 Although caste is a serious concern for policymakers in India, it does not affect productivity of low-income workers.'7 Religion, however, influences workers' productivity and its impact varies by gender. Religion does not influence women's productivity in the protected wage or self-employed sector, but muslim women earn less in the unprotected wage sector. In contrast, muslim men get less in the protected wage sector. Hindu, Muslim and Buddhist men alike earn more in the unprotected wage sector and less as self-employed than men of other religions. Women's place of birth affects wages in the protected wage sector. Thus, a non-Maharashtrian woman earns more than a woman who was born in Bombay. 5.13 The new migration status measures the impact of the length of stay in Bombay on a worker's productivity. Women's length of stay in Bombay does not affect their productivity, a finding consistent with a priori expectation that women often migrate for family reasons, not for better job opportunities (Acharya and Jose 1990). In contrast, men who have migrated in the last 10 or less year earn less compared to those who have migrated more than ten years ago. This finding appears consistent with the finding from Brazil and Columbia (Yap 1976; Ribe 1979). Impact of job characteristics 5.14 The location of a job has a significant effect on earnings in the protected wage and self-employed sectors. The protected job located closer to home has a positive effect on a worker's productivity. This is true for both men and women, which may reflect a worker's efficiency because of less time involved in commuting. The reverse is the case for the self-employed: both men and women who are self- employed earn more if they work away from home and are also mobile. Although the results are sensitive to varying wage function specification, they confirm that self-employed men earn more than self- employed women if both are mobile. The finding that a mobile business pays off more than a non-mobile '6Note that the estimates of rates of returns to schooling are largely insensitive to varying wage function specification. '7This does not mean that caste is not a detrimental factor for improving productivity in India. On the contrary, the low- caste workers of high-income groups may earn less compared to the high-caste workers of the same income groups.  14 business is consistent with findings from countries such as Peru (Smith and Stelcner 1990). 5.15 The industry-specific characteristics also influence a worker's productivity. As expected, a job pays off more if it is covered by any labor law that protects the interest of the workers. This is equally true for both men and women who are employed in the protected wage sector. This result holds irrespective of whether controversial job characteristic such as trade union membership is included in the wage regression. 5.16 An industry's scale (in terms of the number of employees) increases men's productivity in the protected wage or self-employed sector. This is not the case for women, however. Male-female differences in productivity may suggest that men benefit more than women from large-scale enterprises.'8 5.17 Trade union membership improves the productivity of men in both the protected and unprotected wage sectors. Improved productivity may result from the positive feedback of a union, as suggested by SLM model. Alternatively, higher wages may follow from union pressure, as contemplated by NLM model. If NLM view is right, then the role of education in improving a worker's productivity is minimum when trade union membership status is included along with the education variables in the wage function (see Model II in Tables 2 and 3). The results, however, suggest that including union membership status does not reduce the role of education or training variables in improving a workers' productivity. This perhaps indicates that a trade union may have a positive feedback on improving a workers' productivity, a case that supports the SLM view. 5.18 As expected, sales turnover increases the productivity of both men and women who are self- employed. This variable captures the impact of capital on the productivity of self-employed workers. Not surprisingly, inclusion of sales as an additional regressor improves substantially the explanatory power of the model. The marginal impact of sales turnover is higher for women than for men. This suggests that self-employed women are more constrained than men in raising productivity due to lack of capital. This may also indicate a higher labor intensity of women's self-employed output than men's, i.e., women;s value-added in output (sales) is greater than men's. VI. SECTOR SELECTION AND ITS IMPACT ON WAGES AND LABOR SUPPLY BEHAVIOR 6.01 Section V above discusses the results of the effects of various types of human capital and socio- economic-institutional factors on a worker's productivity in three types of labor markets. The results indicate that there are structural differences in wages across three sectors of employment and both individual-and industry-specific characteristics influence a worker's wage. These results are, however, conditional on the assumption that a worker's sectoral job is given. The results are not tenable if the observed wage involves a worker's choice of being in a particular sector, as we hypothesized in Section III. More specifically, if the unobserved characteristics that influence a worker's choice of sectoral job "8Note that including these two non-controversial job characteristics - labor laws and scale of operations - in the wage regression does not any way influence the impact of human capital variables. The ftndings may suggest that one can treat these job-characteristics as exogenous in the wage regression.  15 also affects her or his productivity, the wage estimates reported in Section V are then biased. The extent of bias, however, depends on how important is the sample selection problem, i.e., how representative is the sample of workers in each market segment from the population. In other words, sector selectivity bias exists when the workers in a given sector do not constitute a random subset of the population. 6.02 The sector selectivity bias in turn provides a test of labor market segmentation as it provides an evidence of whether sectoral selectivity decision is made by workers or employers (Gindling 1991). Thus, if the sectoral allocation of workers is non-random and yet this non-random allocation does not affect a worker's wage, then sector allocation does not involve a worker's but an employer's choice. That means, employers determine which workers are to be allocated to which sector and this is clearly an evidence of labor market segmentation. Alternatively, if the sectoral allocation is non-random and this non-random allocation affects a worker's wage in a particular sector, then a sector selectivity bias exists. The market segmentation, thus, cannot occur as workers choose a job based on their comparative advantages and wages."9 Sample selection bias correction critically depends on identification, i.e., how valid are the instruments for identifying the wage model from the sectoral job allocation equation. In other words, the sectoral allocation test of market segmentation depends on the availability of appropriate instruments for identifying sectoral allocation decision from the wage regression. 6.03 Assume that each individual works for living and selects one among the three mutually exclusive job alternatives: (1) Working as self-employed (indexed s), (ii) working in the protected wage sector (indexed f) and (iii) working in the unprotected wage sector (indexed c). Workers' maximize utility as modelled in Section III. Let Vj; be the maximum utility attainable for individual i if she or he chooses participation status j = f,c,s. Following McFadden (1974), we assume that utility is random and the indirect utility function can be decomposed into a nonstochastic component (R) and a stochastic component(O): (8) Vi; = Rj; + e;. where Rj; is a function of observed variables and %, is a function of unobserved variables. The utility maximization principle also ensures the optimal choice of optimal hours of work in each given occupational status (Hji* for j =f,c,s). 6.04 Define the probability of ith individual participating in any of three alternatives as Dji such that Dr, is j = f= 1 for f workers and 0 elsewhere, D, is j = c =1 for c workers and 0 elsewhere, and D,i is j = s = I for s workers and 0 elsewhere. Then the probability that the ith individual selects the jth sector job is: (9) Pjj = Prob.(Dji = 1) = Prob.(Vji > Vid for k=j, k,j=f,c,s); "See more about this argument in Gindling (1991). Labor market segmentation can be defined as a situation where a worker, say, in the unprotected wage sector has less than full access to a job in the protected wage sector held by an observationally identical worker. The NLM model argues that workers choose their sector of employment on the basis of their comparative advantage. However, if segmentation exists, assignment of workers to the protected sector does not reflect the workers' decisions but reflect employers' decisions to hire workers from the pool of workers waiting to obtain a job in the protected wage sector.  16 Substituting (8) in (9) we get, (10) Pii = Prob.(R; - RK > O ;- for k