University of Southern California Dissertations and Theses

Against the wind: labor force participation of women in Iran

(USC Thesis Other)

Against the wind: labor force participation of women in Iran

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content AGAINST THE WIND: LABOR FORCE PARTICIPATION OF WOMEN IN IRAN by Mehdi Majbouri A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulllment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (ECONOMICS) August 2010 Copyright 2010 Mehdi Majbouri Dedication to Azadeh, and to the memory of my mother. ii Acknowledgements I am deeply indebted to Professor John Strauss for his continuous encouragement, valu- able guidance, and stimulating suggestions, and above all, for showing me the way, with- out which this project would not have been possible. I am especially grateful to Professor Jeery B. Nugent for oering me his valuable time, his protable advice, warm support, and patience with the diculties I made. I am pleased to thank Professor Djavad Salehi-Isfahani for providing me with his expertise, wisdom, and precious help which jump-started my research on Iran. I am thankful to Professor Geert Ridder, for whenever I needed his advice, he patiently listened to my erroneous ideas and oered me with his valuable time and comments. I am grateful to Professor Tatiana Melguizo for her warm encouragements. I sincerely wish to thank Azadeh, my wife, for her grace, love, sincerity, and patience throughout the last six years when I needed them most. I am deeply grateful to my father and late mother because of everything. I would like to express my gratitude to the Statistical Center of Iran and my pro- fessors in Iran for providing me with the data and oering me necessary assistance in understanding them. Financial support was provided by the Wallis Annenberg fellowship iii for research on women in families and College of Letters, Arts, and Sciences, University of Southern California. Special thanks goes to Young Miller, Morgan Ponder, and the Economics department sta and graduate students who made the last six years a pleasant journey for me. All the remaining errors are mine. iv Table of Contents Dedication ii Acknowledgements iii List Of Tables vii List Of Figures xi Abstract xiii Chapter 1 Introduction 1 1.1 Revolutions in Post-1979 Iran . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Female Labor Force Participation in Iran . . . . . . . . . . . . . . . . . . 9 1.3 The FLFP Puzzle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Chapter 2 FemaleLaborForceParticipationinIraninaStaticSetting 23 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.2 Labor Force Participation in Theory . . . . . . . . . . . . . . . . . . . . . 25 2.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 2.3.1 Household Expenditure and Income Surveys (HEIS) . . . . . . . . 29 2.3.2 Socio-Economic Characteristics of Households Surveys (SECH) . . 31 2.4 The Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 2.4.1 Estimating the Structural Model for LFP . . . . . . . . . . . . . . 45 2.4.2 Estimating the Structural Model for Hours . . . . . . . . . . . . . 57 2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Chapter 3 Female Labor Force Participation in Iran in a Dynamic Set- ting 65 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.2 Economic Crisis of 1994-95 . . . . . . . . . . . . . . . . . . . . . . . . . . 68 3.3 Work and Macro-destabilization . . . . . . . . . . . . . . . . . . . . . . . 72 3.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 3.4.1 Caveats of Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 3.5 The Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 3.5.1 Labor Force Participation and Economic Crisis . . . . . . . . . . . 84 3.5.2 Hours and Economic Crisis . . . . . . . . . . . . . . . . . . . . . . 97 v 3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 Chapter 4 Conclusion 105 4.1 Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 Appendix A Structural Model Estimations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 Bibliography 125 vi List Of Tables 1.1 Trends in Literacy Rates for Men and Women in Urban an Rural Areas . 2 1.2 Change in Fertility rates in Iran and Other Middle Eastern Countries . . 8 1.3 Share of Households Owning Various Home Production Technologies in Urban an Rural Areas (in per cents) . . . . . . . . . . . . . . . . . . . . . 10 1.4 Distribution of Types of Employments for Rural and Urban Women and Men in 1992 (in %) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.5 Distribution of Employed Women and Men Across Industries for Rural and Urban Areas in 1992 (in %) . . . . . . . . . . . . . . . . . . . . . . . . . . 17 2.1 Number of Households in Each Household Expenditure and Income Survey 30 2.2 Distribution of Number of Households in Each Cluster . . . . . . . . . . . 32 2.3 Distribution of the Households Ever Interviewed Based on the Wave They Were Added to the Panel . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 2.4 Weights of Various Assets Used to Construct an Asset Index for HEIS Data Series following Sahn and Stifel (2000) . . . . . . . . . . . . . . . . . . . . 36 2.5 Linear Probability Model of Labor Force Participation for Women Aged 21 through 65 with Province Fixed Eects, 1990-1996 . . . . . . . . . . . 38 2.6 Linear Probability Model of Labor Force Participation for Women Aged 21 through 65 with Province Fixed Eects, 1997-2004 . . . . . . . . . . . 40 2.7 First Step of the Two-Step Estimation of Heckman Selection Model on Observed Wages across People Aged 21 through 65, Pooled Data 1992-95 48 vii 2.8 Estimation of Log of Real Wages for People Aged 21 through 65 Controlling for Heckman Selection on Wages (Two-Step Method; First Step in Table 2.7), Pooled data 1992-95 . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 2.9 Linear Probability Model of Labor Force Participation on Predicted Log of Real Wages from Regression in Table 2.8 for People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors), Pooled data 1992-95 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 2.10 Linear Probability Model of Labor Force Participation on Predicted Log of Real Wages from Regression in Table A.3 for Never-Married People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors), Pooled data 1992-95 . . . . . . . . . . . . . . . . . . . . . . . . . 54 2.11 Linear Probability Model of Labor Force Participation on Predicted Log of Real Wages from Regression in Table A.5 for Married People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors), Pooled data 1992-95 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 2.12 Linear Estimation of Log of Hours Worked on Predicted Log of Real Wages from Regression in Table 2.8 for People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors) Controlling for Heckman Selection on Hours (Two-Step Method; First Step in Table A.6), Pooled data 1992-95 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 2.13 Linear Estimation of Log of Hours Worked on Predicted Log of Real Wages from Regression in Table A.3 for Never-Married People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors) Controlling for Heckman Selection on Hours (Two-Step Method; First Step in Table A.7), Pooled data 1992-95 . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 2.14 Linear Estimation of Log of Hours Worked on Predicted Log of Real Wages from Regression in Table A.5 for Married People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors) Controlling for Heckman Selection on Hours (Two-Step Method; First Step in Table A.8), Pooled data 1992-95 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 3.1 Change in Real per capita Expenditure During the Crisis Years, Control- ling for HH Head's Fixed Eects (in 1994 Rials) . . . . . . . . . . . . . . 72 3.2 Summary Statistics of the 1992 Values of Some Variables of Interest for Men and Women Aged 21 through 65 Who Were in the Panel for More Than One Year . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 3.3 Years in Panel for households who were interviewed in the rst wave (1992) 81 viii 3.4 Waves in the Panel for households who were interviewed in the rst wave (1992) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 3.5 Percentage of People with at Least 2 Jobs Among Working People in 1992 83 3.6 Linear Probability Model with Individual Fixed Eects Estimates of Labor Force Participation of People Aged 21 through 65 . . . . . . . . . . . . . 86 3.7 Linear Probability Model with Individual Fixed Eects Estimates of Labor Force Participation of People Aged 21 through 65 in Rural Areas . . . . . 87 3.8 Linear Probability Model with Individual Fixed Eects Estimates of Labor Force Participation of People Aged 21 through 65 in Urban Areas . . . . 88 3.9 Estimated Unemployment Rates in Each Year in the Panel (in %) . . . . 89 3.10 Linear Probability Model with Individual Fixed Eects Estimates of Labor Force Participation of People Aged 21 through 65 in Rural Areas, Control- ling for Rainfall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 3.11 Coecient of Instability () in the Linear Probability Model with Indi- vidual Fixed Eect Estimates of Labor Force Participation for All People Aged 21 through 65 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 3.12 Distribution of Types of Employment for Rural Women who Entered the Labor Market during Instability Relative to Rural Women in 1992 (in %) 97 3.13 Least Squares with Individual Fixed Eect Estimates of Hours Worked per Week in the Main Job for All People Aged 21 through 65 . . . . . . . . . 99 3.14 Least Squares with Individual Fixed Eect Estimates of Hours Worked per Week in the Main Job for People Aged 21 through 65 in Rural Areas . . 100 3.15 Least Squares with Individual Fixed Eect Estimates of Hours Worked per Week in the Main Job for People Aged 21 through 65 in Urban Areas . . 101 3.16 Coecient of Instability () in the regression of Hours Worked in the Main Job for All People Aged 21 through 65 . . . . . . . . . . . . . . . . . . . 102 A.1 Weights of Various Assets Used to Construct an Asset Index for SECH 1992-95 Panel Data following Sahn and Stifel (2000) . . . . . . . . . . . . 112 A.2 First Step of the Two-Step Estimation of a Heckman Selection Model on Observed Wages across Never-Married People Aged 21 through 65, Pooled Data 1992-95 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 ix A.3 Estimation of Log of Real Wages for Never-Married People Aged 21 through 65 Controlling for Heckman Selection on Wages (Two-Step Method; First Step in Table A.2), Pooled data 1992-95 . . . . . . . . . . . . . . . . . . . 115 A.4 First Step of the Two-Step Heckman Selection Model on Observed Wages across Married People Aged 21 through 65, Pooled Data 1992-95 . . . . . 116 A.5 Estimation of Log of Real Wages for Married People Aged 21 through 65 Controlling for Heckman Selection on Wages (Two-Step Method; First Step in Table A.4), Pooled data 1992-95 . . . . . . . . . . . . . . . . . . . 118 A.6 First Step of the Two-Step Heckman Selection Model on Non-Zero Hours Worked across People Aged 21 through 65, Pooled Data 1992-95 . . . . . 119 A.7 First Step of the Two-Step Heckman Selection Model on Non-Zero Hours Worked across Never-Married People Aged 21 through 65, Pooled Data 1992-95 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 A.8 First Step of the Two-Step Heckman Selection Model on Non-Zero Hours Worked across Married People Aged 21 through 65, Pooled Data 1992-95 123 x List Of Figures 1.1 Average Years of Education Across Birth Cohorts . . . . . . . . . . . . . . 4 1.2 Total Number of Students and Female to Male Student Ratio in Public and Private Colleges, 1985-2006 . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3 Number of Public Primary, Middle, and High Schools, 1963-2003 . . . . . 6 1.4 Own-Children Estimates of Total Fertility Rates for Iran by Urban and Rural Areas, 1976-2006 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.5 Labor Force Participation of Women Aged 21 through 65 in Rural and Urban Areas, 1992-2003 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.6 Labor Force Participation of Men Aged 21 through 65 in Rural and Urban Areas, 1992-2003 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.7 Labor Force Participation of Married and Never-married Women Aged 21 through 65 in Rural Areas, 1992-2003 . . . . . . . . . . . . . . . . . . . . 13 1.8 Labor Force Participation of Married and Never-married Women Aged 21 through 65 in Urban Areas, 1992-2003 . . . . . . . . . . . . . . . . . . . . 13 3.1 Trends in Import per Capita and In ation (CPI) . . . . . . . . . . . . . . 70 3.2 GDP and Private Consumption per Capita Growth Rates During the Crisis 70 3.3 Log Real Wages For Women Aged 21 through 65 in Rural and Urban Areas, 1992-1995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 3.4 Log Real Wages For Men Aged 21 through 65 in Rural and Urban Areas, 1992-1995 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 3.5 Average Province Level Rainfall over time . . . . . . . . . . . . . . . . . . 91 xi 3.6 Histogram of Change in Province Level Rainfall Between 92-93 and 94-95 periods (in%) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 4.1 Comparison of per capita energy subsidies and FLFP rates for women aged 25 to 55 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 xii Abstract Women in Iran have garnered extraordinary achievements in the last two and a half decades. Fertility rate has fallen in one of the largest and fastest transitions in modern human history. Meanwhile, education levels have consistently increased, to the extent that currently, women who are in their late 20s are more educated than their male counterparts. But still, female labor force participation (FLFP) rates remain at the low levels of two decades ago (FLFP puzzle), a fact which has led many researchers to suggest that FLFP is inelastic to economic forces. In this manuscript, I analyze some economic characteristics of the women's labor supply in Iran and show that female labor force participation is elastic, at least for some women and with respect to some economic forces. In the second chapter, I provide the reduced form as well as structural estimations of women's \observed" labor force participation and hours worked in Iran. In both models, I show that the results are compatible with the basic predictions of economic theory and the empirical evidence from other countries. For instance, women with higher education are more likely to participate; age prole of participation is concave and has its peak between the ages of 30 to 50; more assets leads to less participation; and presence of more adult men decreases the likelihood of participation while more adult women increases it. Moreover, similar xiii to the results for the developed countries, the structural estimations show that Iranian women's participation in urban areas, regardless of marital status, is highly elastic with respect to wages, while their hours worked is in-elastic. In Chapter 3, I argue that despite its rigidities, \observed" FLFP did, for some women, respond to economic pressures induced by macroeconomic instabilities. Looking at the Iranian minor economic crisis of 1994-95, I show that, controlling for individual xed eects, married rural women and unmarried urban women, between the ages of 21 to 65, joined the labor market at an increase of about 8.4 and 8.9 percentage points respectively (30% and 36% rise). During the same period, married urban women did not change their participation rates. These results are compatible with the hypothesis that marriage considerably limits urban women's responsiveness to economic forces (\marriage lock" theory). On the other hand, the high wage elasticities of participation for urban married women, estimated in Chapter 2, may contradict this hypothesis. No change in hours worked, during the crisis, was found for any group of women. I discuss my future research in Chapter 4. xiv Chapter 1 Introduction I am happy that my wife works and contributes to the family with her income, as I cannot earn a decent living with my income. But I am concerned that I may lose my authority. { A worker whose wife had a small tailoring business at home in Tehran, Iran 1.1 Revolutions in Post-1979 Iran Iranian women have prevailed in two remarkable revolutions during the past two and a half decades. The rst one is the continuous dramatic rise in education levels of women relative to men during the three decades after the Islamic revolution. Thanks to the implementation of a national literacy program, the literacy rates, as depicted in Table 1.1, improved continuously after the revolution for both genders, but faster for women in urban as well as rural areas. While only 36.3% of rural women were literate in 1985, almost 70% could read and write in 2005. The change is even larger in younger cohorts. In 1976, merely 10% of women aged 21 to 24 in rural areas were literate. Today, this number is 91%. Promoting universal education, the government invested considerably, especially in impoverished urban and rural areas, even in dicult times such as the Iran-Iraq war. 1 Table 1.1: Trends in Literacy Rates for Men and Women in Urban an Rural Areas Rural Urban Women Men Women Men 1985 36.3 60.0 65.4 80.4 1990 54.2 72.6 76.8 86.7 1995 62.4 76.7 81.7 89.6 2005 69.0 81.2 85.6 92.2 Data Source: Statistical Center of Iran, Census data. As Figure 1.1 depicts, although both men and women's average years of education rose across the board, the increase was larger and faster for women. Not only was the gender gap in education of youth in rural areas eliminated, but young urban women's education surpassed young urban men's. 1 Those same young girls who went to primary school during the 1980s demanded a college education in the mid 90s. In Iran, in order to go to college, one needs to take a national examination. (Exams are separate for public and private colleges.) Supply of college education was always behind the demand and during 90s and early 2000s, on average, only the top 30% could enter college. The public colleges, which oer free education, are considered to be of higher quality. Therefore, only the top 10% of the participants in the national exams are competent enough to enroll in a public university. 1 Among the plausible reasons are the gender segregation of primary, mid and high schools, higher quality of teachers in girls' schools, and school construction in poor urban neighborhoods and rural areas (Keddie, 2006; Salehi-Isfahani, 2005a, p. 286). Everything equal, the more traditional families, presumably a large portion of the population, were more eager to send their girls to a segregated school than a non-segregated one and hence many of them, who would have not sent their female ospring to school under the previous regime, willingly enrolled them after the revolution. New school constructions and teacher training program were other factors that promoted education. After revolution, all private schools became public and education was oered free of charge. Figure 1.3 shows the change in the number of public schools over time. A jump in the number of public primary schools in 1980 is pronounced and related to nationalize private schools after the revolution. The last factor is the hypothesis that female teachers were of higher quality than their male counterparts, since women who teach are among the selected few who participate in the labor force and hence of higher ability on average. The causes of the rise in women's education after revolution still remains a question of interest requiring further research. 2 Despite this intense competition for free high quality higher education, women who were born and educated after the revolution began to surpass men and enter (mostly public) colleges in large numbers. In less than 10 years, women who had made up only 30% of public colleges' student population in the early 1990s became the majority, constituting 66% of the enrollments by 2003. Becker et al. (2009) explain that rise in the women's education is a global phenomenon and discuss the reasons behind it, nevertheless this phenomenon is quite pronounced in Iran. 2 Figure 1.2 depicts this sharp change. The second revolution came with the fastest and largest drop in fertility rates in modern human history. As depicted in Figure 1.4, fertility rates fell dramatically in less than two decades from about seven to almost two births per woman (Abbasi-Shavazi et al., 2009; Aghjanian, 1991; Aghajanian, 1995). This revolution started in both urban and rural areas a few years after the 1979 Islamic revolution. It was only in 1989, nearly ve years after the decline in fertility rates began, that the government launched its family planning program with a particular focus on rural areas. 3 Countries that are considered fertility miracles, like Malaysia, experienced the same transitions over longer time spans { three to four decades. 2 These changes may turn on a cycle, in which each generation invest more in the next generations. Mothers with more education and less children may tend to invest more in the health and education of the future generations, as Thomas and Strauss (1992) and Thomas et al. (1991) show. Although recent studies such as Strauss (1990), Behrman and Wolfe (1987), Behrman and Rosenzweig (2002), and Du o and Bereierova (2002) oer some evidence that there is no dierence between mother's and father's education on human capital investments. Meanwhile this new evidence does not refute that mothers with more education invest more in the health and education of their children. 3 The reasons behind this fall in fertility rates, especially compared to other countries in the region, are not quite well known. Salehi-Isfahani et al. (2009) nd that only 7 to 18 percents of this large fall can be explained by the government family planning program, a nding consistent with those from other countries, such as Bangladesh (Joshi and Schultz, 2007), Columbia (Miller, 2005), and Peru(Angeles et al., 2005). 3 Figure 1.1: Average Years of Education Across Birth Cohorts Data Source: Household Expenditure and Income Survey 2006. 4 Figure 1.2: Total Number of Students and Female to Male Student Ratio in Public and Private Colleges, 1985-2006 Data Source: Extracted from 21 yearbooks of Statistical Center of Iran. 5 Figure 1.3: Number of Public Primary, Middle, and High Schools, 1963-2003 Data Source: Extracted from 40 Statistical Yearbooks of Statistical Center of Iran. These transitions are remarkable, particularly since on many fronts life under the Is- lamic Republic was more dicult for women than it had been under the previous regime. 4 Throughout these decades of radical change, traditional social norms have faded. For instance, Rajabzadeh (1999) showed that while 70% of names given to newborn babies in 1979 were religious, by 1994 the rate was only 35%. In another study, he explains that none of the women, surveyed in a college in Tehran, was willing to choose a religious name for her children. As name giving is an indicator of what parents perceive to be benecial 4 Within a few months after the revolution, the 1967 family protection laws, which had given equal rights to women in divorce and custody, were abrogated, legal age of marriage for women were lowered to 13, Islamic dress code were imposed on women in the workplace (and eventually all public places), women were barred from becoming judges, and some public and private venues such as primary and secondary educational institutions, beaches, and sport facilities were segregated. These changes were strictly enforced (Gheytanchi, 2000). 6 Figure 1.4: Own-Children Estimates of Total Fertility Rates for Iran by Urban and Rural Areas, 1976-2006 Source: Abbasi-Shavazi et al. (2009) for the future of their children, it could be an index of people's expectations about future trends in religiosity. Moreover, a longitudinal survey between 1995 and 2000 suggests a decline in the number of decisions made by fathers for their daughters (Moheseni, 2000). In addition, as newly educated young women have become more involved in society, the ocial traditional dress code has been losing its rigor. (For more discussion on fading traditional norms and institutions in Iranian society, see Majbouri, 2007). Similar trends of an increase in women's education and a decrease in fertility rates have been observed in other Middle Eastern countries, but none had such a fast transition in education and fertility rates, despite the fact that they are smaller, more urbanized countries. For example, measures of tertiary education for several oil producing countries 7 in the Middle East show that female students dominate domestic universities by a ratio of more than two to one. 5 Comparing these measures with Iran's may be a bit misleading, since it is widely known that in those countries boys are usually sent abroad to study in European and American colleges, while girls do not have such opportunities and therefore enroll in domestic colleges. 6 Table 1.2: Change in Fertility rates in Iran and Other Middle Eastern Countries Total Fertility Rates Change in 1980-85 1995-2000 Fertility Rates Iran 6.9 2.3 4.6 Algeria 6.4 3.3 3.1 Bahrain 4.6 2.6 2.0 Egypt 5.1 3.4 1.7 Jordan 6.8 4.7 2.1 Kuwait 4.9 2.9 2.0 Lebanon 3.8 2.3 1.5 Morocco 5.4 3.4 2.0 Qatar 5.5 3.7 1.8 Sudan 6.0 4.9 1.1 Syrian Arab Republic 7.4 4.0 3.4 Tunisia 4.9 2.3 2.6 United Arab Emirates 5.2 3.2 2.0 Source: United Nations Population Division Estimates. Education in universities abroad is also subsidized by the governments of these coun- tries. Hence, measures of enrollment in domestic colleges in ates the perception of women's progress in these countries. In Iran, on the other hand, the largest number of students studying abroad, including graduate students, is reported to be about 100,000 5 For Kuwait, for example, this is 3 to 1. 6 This is mainly because girls are rarely allowed to live far from their families. One reason that these countries bring top American and European universities to build campuses on their soil is to provide similar opportunities of quality education for girls as well. 8 people, which is not even close to 8% of the total student population. Moreover, this number includes graduate students as well as girls who have been sent abroad. Since, in Iran, studying abroad is not subsidized by the government, and Iran's GDP per capita is less than half and sometimes one third of the other oil producing countries in the region, there is less incentive to send children to study abroad. Fertility rates, on the other hand, are easier to compare. As shown in Table 1.2, while other countries in the Middle East had lower fertility rates in 1980-85, their rates at the end of 20th century were higher than Iran's. 1.2 Female Labor Force Participation in Iran Considering the phenomenal rise in women's education, and the dramatic fall in fertility, one would expect that female labor force participation (FLFP) should have increased in response. One should add improved home production technologies to these forces that aect fLFP. As Table 1.3 shows, Iran has one of the highest rates of access to safe drinking water in the Middle East { especially considering the diculties arising from size and geographical challenges { with an estimated 96% of its people enjoying clean piped water: nearly 100% in urban areas and about 90% in rural areas as of 2006 (see Table 1.3). Electricity coverage is also universal and use of home appliances has increased continuously during the last three decades. Figure 1.5 shows the trends in labor force participation of women and men aged between the ages of 21 to 65 in urban and rural areas. In the data, one is considered a participant in the labor force if she is looking for work, or worked for at least two days 9 Table 1.3: Share of Households Owning Various Home Production Technologies in Urban an Rural Areas (in per cents) Rural Urban 1990 1995 2000 2006 1990 1995 2000 2006 Electricity 73 87 96 99 99 100 100 100 Piped Water 62 74 83 90 96 98 99 100 Refrigerator 53 71 85 95 92 96 97 99 Stove 58 73 82 92 88 94 96 98 Washer 7 12 16 28 40 48 54 69 Vacuum Cleaner 3 6 15 40 27 42 58 81 Data Source: Household Expenditure and Income Surveys of 1990, 1995, 2000, and 2006. in the week preceding the survey. 7 As shown, the FLFP rates in both urban and rural areas are about 15% and 25% respectively. These rates are quite low relative to other developing countries. FLFP rate in rural areas had an initial rise and then uctuated at around 25% between 1993 and 2003, when these economic forces were becoming stronger. In urban areas, on the other hand, FLFP remained at around 15% during this period. One can easily conclude that there is little uctuation as well as change in urban areas relative to rural ones. For a report on trends in FLFP before 1990, please refer to Esfahani and Bahramitash (2010) who provide the trend of FLFP in urban and rural areas since 1956 based on census data. This phenomenon { that FLFP is low relative to other developing countries 8 , and all these economic forces did not raise it as much as expected { is prevalent across the Middle East. Confounded by this, social scientists have called it a puzzle and indeed it 7 Since 2005, the denition of employment has changed to cover anybody who worked for at least an hour during the week prior to the interview. 8 Goldin (1994) shows that married women's LFP rate rst falls and then rises as a country develops. But FLFP rates in the middle east are signicantly lower than countries with similar income levels. 10 Figure 1.5: Labor Force Participation of Women Aged 21 through 65 in Rural and Urban Areas, 1992-2003 Note: An individual is a participant in the labor force if she is considered \employed" or \unemployed" in the surveys. One is \employed" if she worked for at least two days during the week prior to responding to the questionnaire. \Unemployed" in the survey is dened as someone with no job who is looking for it in the week prior to the interview. Data Source: Household Expenditure and Income Surveys, 1990 through 2004. still remains a puzzle. In Iran, the puzzle is more pronounced in urban areas, as FLFP in rural areas has an upward trend as well as more uctuations. Before discussing this puzzle, I will explain other important characteristics of partic- ipation of women in the labor force. One major characteristic of FLFP in Iran is that never-married women have signicantly higher \observed" labor force participation than married women. This phenomenon, which is also true in some other Middle Eastern countries, is depicted in Figures 1.7 and 1.8. 9 9 Goldin (1990) explains that in the rst decades of the 20th century in the United States, a large dierence between married and unmarried women's LFP rates existed, albeit for dierent reasons. At that time, businesses were ring women who were getting married. In Iran, the employee-friendly labor laws prohibits ring anyone except in special circumstances, such as medical reasons or termination of employment by the employee. 11 Figure 1.6: Labor Force Participation of Men Aged 21 through 65 in Rural and Urban Areas, 1992-2003 Note: Please see the notes for Figure 1.5. Figure 1.7 shows that the trend in rural FLFP which we saw in Figure 1.5 is largely because of the trend in married women's participation rate. As shown, FLFP rates for married women resembles those for all rural women. It had an initial upward movement and then uctuated around 23%. On the other hand, FLFP rate for never-married women in rural areas did not have any trend and was uctuating at around 33%. In urban areas, as shown in Figure 1.8, these are never-married women who increased their participation rate and show more uctuation in it. They also have the largest participation rates among women. Moreover, the FLFP rates for married women in urban areas is quite sluggish and did not change over time. Overall, the groups that had almost no trend in their participation where unmarried women in rural areas and married ones in urban areas. In addition, the only group that had almost no uctuation in his participation was urban married women. These are important patterns to which we will come back 12 Figure 1.7: Labor Force Participation of Married and Never-married Women Aged 21 through 65 in Rural Areas, 1992-2003 Note: Please see the notes for Figure 1.5. Figure 1.8: Labor Force Participation of Married and Never-married Women Aged 21 through 65 in Urban Areas, 1992-2003 Note: Please see the notes for Figure 1.5. 13 later in Chapter 3 when we analyze the responses of dierent groups of women to a macro-economic shock. Interestingly, the dierence between married and never-married women's LFP in ur- ban areas is almost three times that of rural areas and it remains somewhat constant over time. Note that part of this could be the product of the fact that one is \observed" to be working only when she is employed for at least 2 days { that is, if women decrease their hours worked to less than two days per week after marriage, in the data they are con- sidered out of the job market. Keeping this in mind, the other implication of this result may be that marriage locks women out of the \observed" labor market (\marriage lock theory"). This eect of marriage may or may not be due to discrimination. For instance, women after marriage may have a stronger preference for home production, like raising children, rather than market work. In the case of discrimination, a husband may not \allow" her wife to work outside home, for various reasons such as: 1) wife's work outside home is a breach in traditional norms, 2) husband wants to limit social interactions of his wife with other men, and 3) husband loses part of his authority and bargaining power at home since his wife has a source of income of her own. There is large anecdotal evidence that discrimination is a major reason for marriage eect. In any case, this is a factor that should be considered in any analysis of labor force participation of women in Iran and in the Middle East in general, and I account for it in this study as well. Despite similarities in the characteristics of the labor market for women between Iran and some other Middle Eastern countries, dierences do exist. For example, unlike other Middle Eastern countries such as Saudi Arabia and countries in the southern part of Per- sian Gulf, the workplace is not segregated based on gender. Except for primary, middle 14 and high schools which are gender-segregated 10 , women and men are free to work along- side each other. Although women are under scrutiny to abide by the ocial dress code and codes of conduct in governmental organizations, we nd that in private workplaces there is less emphasis on these rules. In addition, there had been restrictions on the professions women could choose at the beginning of the revolution. For instance, women could not become agricultural engineers in 80s and early 90s (Keddie, 2000). But these restrictions were abrogated. In today's Iran, one can nd women bus and taxi drivers, re ghters, engineers in dicult working conditions (such as oil wells and deserts), pilots, and soldiers { a diversity which is rare in the Middle East. The type of employment that men and women in urban and rural areas choose is also important. Table 1.4 describes the jobs rural and urban men and women accepted in 1992, at the beginning of this study. As one might expect, the majority of women in rural areas (63%) were working as unpaid family members, while only 20% of urban women had the same kind of job. The fact that rural women have the option of working as an unpaid family member in rural areas may be the reason behind higher FLFP rates in rural areas and also smaller dierences between married and unmarried FLFP rates in rural areas. Urban women, not surprisingly, were more likely to work in the public sector (43% vs. 2% in rural areas). Interestingly, there is no dierence amongst urban and rural women between those who are employers and those who are self-employed. Men were more likely to be employers or self-employed (about 40% and 50% in rural and urban areas, respectively). Similar to women, there were more men working in the public sector in urban rather than rural areas (35% vs. 16%), but the dierence is smaller for men. 10 One should add to this list a university in Tehran which is for women only. There are a few sections in some governmental organizations which are gender segregated as well. 15 Table 1.4: Distribution of Types of Employments for Rural and Urban Women and Men in 1992 (in %) Women Men Rural Urban Rural Urban Employer 2.1 0.9 16.2 7.2 (0.5) (0.4) (0.7) (0.5) Self Employed 19.9 17.0 33.2 29.7 (1.4) (1.8) (1.0) (0.8) Employee in public sector 2.9 43.4 16.5 36.0 (0.6) (2.3) (0.8) (0.9) Employee in private sector 12.0 17.9 21.9 25.0 (1.2) (1.8) (0.8) (0.8) Unpaid family worker 63.1 20.9 12.3 2.2 (1.7) (1.9) (0.7) (0.3) Total 100 100 100 100 Observations 765 459 2141 2831 Note: Standard errors are in parentheses. Interestingly, men were slightly more likely to participate as employees in the private sector. Men who were working as unpaid family members were a small minority. Rural men were about 5 times less likely to work as unpaid family labor relative to rural women, and in urban areas they were 10 times less likely to work as unpaid family labor. Table 1.5 looks more closely into industries in which women and men worked in 1992. In rural areas, women were almost equally divided into two industries: agriculture and manufacturing. Manufacturing is dened here as any production of goods that involves using machines, tools and labor, and includes arts and crafts, a major industry in rural areas that includes carpet production. The number of rural women working in manu- facturing was slightly higher than in agriculture (51% vs. 44%). In urban areas, on the other hand, the majority of women were working in manufacturing (39%) and education 16 Table 1.5: Distribution of Employed Women and Men Across Industries for Rural and Urban Areas in 1992 (in %) Women Men Rural Urban Rural Urban Agriculture 44.2 5.7 53.1 5.4 Manufacturing 51.6 39.4 9.0 22.4 Construction 0.0 0.0 10.8 9.5 Wholesale & Retail 0.3 3.3 5.3 17.2 Hotels & Restaurants 0.0 0.2 0.5 1.5 Transportation 0.0 0.4 5.4 9.1 Financial 0.0 1.1 0.2 1.4 Real Estate 0.0 1.5 0.1 1.0 Public Service 0.4 5.0 9.3 16.2 Education 2.0 30.9 2.5 5.4 Health 1.2 6.3 0.4 1.9 Other 0.4 6.1 3.4 9.0 Observations 765 459 2141 2831 Note: \Manufacturing" includes \Arts and Crafts" production, which is a major industry in rural areas containing carpet production. Estimated using all observations in year 1992 of SECH 92-95. (31%), while healthcare (6%) ranked third. This demonstrates that most urban work- ing women were teachers, factory workers and nurses. The majority of rural working men (53%) worked in agriculture, while only 9% (compared with 51% of rural women) worked in manufacturing. The rest were distributed in a variety of industries including construction, trade, and transportation. Urban men spread across all industries. While, like urban women, the largest share of them (22%) were in manufacturing, contrary to urban women only 5.4% were in education. Sales and public service at 17% and 16%, respectively, were the second and third largest groups. 17 1.3 The FLFP Puzzle The Iranian version of FLFP puzzle is a mixture of three implicit questions: the rst is why labor force participation was initially low relative to other developing countries, the second is why it remained low, especially in urban areas, despite all these changes, and the third is why women sought education and decreased their fertility rates. A few non- tested hypotheses are provided. I categorize these hypotheses in three groups: 1) those attributing the problem to the shortcomings of data, 2) those assigning it to the supply side of the labor market, 3) and those that approach the question from the demand side. I provide two (complementary) hypotheses which focus on the shortcomings of the data. One limitation is the denition of \employment," which is too narrow. Since we do not observe those who work part time for less than two days per week, we may miss any changes in FLFP rates that occurred due to changes in the share of women (out of total women) who were working part-time for less than two hours during this period. While this claim is dicult to verify, one may still expect that based on the dramatic rise in economic forces explained above, the observed FLFP rates should rise. 11 The second hypothesis proposes that urban women in Iran who own their own busi- nesses, such as tailoring, hairdressing, and tutoring, usually hide them by working at home, in order to escape the burdens of business licenses, rent and taxes. These women are reluctant to report that they work. Hence, working women are always under-reported in governmental surveys. To the extent that this under-reporting changes over time, we do not observe true changes in FLFP. This hypothesis is also hard to verify, although 11 Recent surveys have changed the denition of \employed" to include all people above age 10 who are working for at least an hour during the week preceding the survey. They found an increase of about 50% in LFP rates of women aged 10 and above, in both urban and rural areas, relative to past surveys. 18 there is anecdotal evidence that this is a common practice among urban women. For example, a female graduate of a top university in Iran explained that she wants to set up a small professional kitchen at her home to provide quality home-made catering for small VIP meetings. But even if this hypothesis is true, it should explain why we should not expect similar changes for reported labor force participation. In other words, while this hypothesis may explain the rst question, it should be modied to be able to explain the second question as well. On the forefront of hypotheses approaching the puzzle from both demand and supply is one that claims that traditional norms, stemming from religion, are quite strong and impede women from seeking work in the market. The traditional norms on the supply side make households reluctant to have their female family members working (especially married ones). On the demand side, businesses also may be reluctant to hire women because of these traditional norms. Discrimination, both on the supply and demand side, is a very important and relevant hypothesis. Extensive rm and household surveys, creative research designs may be required to identify these discrimination eects. Perhaps we may only need to wait more in order that exogenous factors, such as changes in policy or law, happen and help nding evidence for discrimination. On the other hand, there are countries which are predominantly Muslim but have higher female participation rates in the labor force than the Middle East, such as In- donesia (58.5%), Malaysia (50.2%), Brunei (58.4%), and Bangladesh (60%) 12 . Moreover, countries with similar traditional norms, such as India, have higher FLFP rates. Mean- while, if the norms are proven to behave dierently in Iran (and the Middle East), then 12 These FLFP rates are for women aged 25 through 54 in late 1990s and were taken from Oce (2001). 19 we may be closer to the answer. But still the question that follows is why such norms in the labor market still prevail while there is evidence of them fading in other aspects of contemporary Iran. (For more discussion on fading traditional norms and institutions in Iranian society, see Moheseni, 2000; Majbouri, 2007; Rajabzadeh, 1999). These observations suggest that research should focus on economic and non-economic forces which impact traditional institutions. For instance, a husband may lose part of his authority and bargaining power at home if his wife has a source of income of her own. Hence, he has more incentive to adhere to those traditions. 13 Assuming that traditional norms and culture have similar eects in all Muslim countries, one may expect dier- ences that would lead to dierent outcomes. Ross (2008) argues that \oil, not Islam, is at fault." Employing cross-country analysis, he nds that low FLFP is associated with oil rents rather than Islam. He explains that the mechanism through which oil revenues can hinder female participation is related to Dutch disease and the fall in traded goods and rise in non-traded goods such as construction. If non-traded goods sectors such as con- struction have more masculine employment and traded goods sectors have more feminine employment, then Dutch disease will increase male wages and also reduced female wages. Moreover more oil revenues bring more government transfers which increases non-labor income. Both these forces reduces female labor force participation. One of the major mechanisms that may relate oil to low FLFP is the way Oil revenues are used to provide government transfers. For example, the Iranian government provides large subsidies for various consumer goods and services such as food, utilities, energy, transportation, health care, and education. Iran spent $55 billion in 2008 on energy 13 The dierential between married and unmarried women's LFP rates may be partly explained by this. 20 subsidies (ten times what was spent in 1990), ranking rst among all countries in the world. This is translated to $786 per capita, second only to Saudi Arabia with $1036 per capita in fuel subsidies, and is signicantly lower than Indonesia, India, and China with $77, $20, and $29 in energy subsidies per capita, respectively (The Economist, 2009). Primary and secondary education is almost fully subsidized, while supply meets the demand. Public tertiary education is also free, although the government only covered about 10% of the demand in the 1990s and 20% in the 2000s. Although over 85 percent of the Iranian population use an insurance system to reim- burse their drug expenses, the government heavily subsidizes pharmaceutical production and importation. In order to provide cheap basic food for low-income households, agri- cultural products, such as cereals, milk, and dairy, are subsidized. In the early 90s, food subsidies amounted to between one and three percent of the GDP (Pesaran, 2000). These subsidies may provide support for traditional (discriminatory) institutions to exist. That is, since discriminating members of the household can maintain a standard of living, and do not need income gained by their wives, they can enforce such institutions. The subsidy explanation for the puzzle is one of the issues that I would like to work in near future. In my dissertation, I do not intend to solve this puzzle; rather I would like to explain some of the characteristics of FLFP and show that economic forces may aect labor force participation of some women, even under rigid institutions. Chapter 2 adapts a simple model of labor force participation to Iran to show how some of the major characteristics of household aects decision for FLFP. I show that the model gives a very consistent story with what has been found for many developed and developing countries and the Iranian household behavior is quite similar qualitatively. 21 In Chapter 3, I look at the response of FLFP to macroeconomic forces, in particular the economic downturn of 1994-95. I show that FLFP is not as rigid as has been perceived and responds to economic forces at least for some women. Never-married women in urban areas and married women in rural areas are the groups that show more exibility in terms of their participation. This is the rst but an important step to show that economic factors play a role in the decision of FLFP in Iran and possibly caused the puzzle itself. 22 Chapter 2 Female Labor Force Participation in Iran in a Static Setting 2.1 Introduction Understanding the economics of labor force participation is one of the rst steps in ex- plaining the puzzle of FLFP in Iran. The economic theory has extensively modelled the factors that aect the decision to participate in the labor market (for a review of this liter- ature see Killingsworth and Heckman, 1986). A set of these models approach the question as if there is a single lifetime period (\Static" models). In these models, the individual has already been endowed with a set of characteristics and her decision to participate is usually aected by these predetermined characteristics and will not change. For example, her level of education and family background will not change. These models, which are usually applied to cross-sectional data are useful for understanding long-term decisions. The other set of models are those in which the individual makes many long and short- term decisions over time and adjusts the decisions based on changes in the environment and her status. For instance, her decision of whether to go to college, to marry, to 23 have certain number of children, and to work are continuously revised in response to the socio-economic forces, such as prices and returns. These elaborate and complex models are \dynamic" models in which people are forward looking and consider the long-term consequences of each of their decisions. Although static models tend to simplify the reality, their predictions match the intuition derived from dynamic models and therefore a relevant approach to the analysis of labor force participation. In this chapter, I apply the standard static model of labor force participation to the available data to see how much FLFP in Iran follows the predictions of a simple economic theory. In Section 2.2, I review the theory of Labor force participation and discuss the estimation methods used to apply the theory to the data. Section 2.3 reviews the two major data series, that are used in this and next chapter, in detail. Following that, in Section 2.4, I provide the empirical evidence for the static model both in terms of reduced form and structural estimation. The results are quite consistent with the predictions of the static model and the evidence for other countries, showing that FLFP in Iran is qualitatively similar to other countries. In section 2.5, I explain the implications of the results. 24 2.2 Labor Force Participation in Theory In this section, I describe the canonical static model of labor force participation for a household. Consider a household as a single entity that maximizes her utility function as follows max U i (L i f ;L m i ;X i ;E i ) (2.1) S.t. pX i +W f i L f i +W m i L m i W f i T +W m i T +Y i 0L f i ;L m i T in whichL f i andL m i are the leisure time of the female and male members of the household i, and X is a composite good. The exogenous elements of this model are Y i , non-labor income, p, the price of the composite good X, and W f i and W m i , the hourly wage rates of female and male members. E i is a measure of all household characteristics that shape the household i's utility function. The Kuhn-Tucker conditions for leisure are simply @U i @L f i i W f i 0 (2.2) @U i @L m i i W m i 0 (2.3) 25 in which i is the Lagrangian multiplier of the combined time and budget constraints. The individual's rst order condition binds, if she works, i.e. when optimal leisure is less than T . In this case, the optimal solution for Leisure is L j i =L i (W f i ;W m i ;Y i ;p;E i ) (2.4) in which E i contains household variables that shapes the utility function especially with respect to labor force participation. It has an observed component V i as well as an unobserved part e i . In Iran, as most women do not work, their optimal leisure is at the corner solution, i.e. L f i =T . Hence, the female leisure rst-order condition, Equation (2.2), is an inequality. Note that, one can normalize wages by the price of the composite good, p, in the utility maximization problem and obtain L j i which does not include p. If the utility function is quadratic, it can be shown that L j i is a linear function and can be written as L j i = +' 0 V i +' 0 f W f i +' 0 m W m i +r 0 Y i +e i +v j i j =m;f (2.5) Moreover, normalized hourly wage rates (from now on, wages) are functions of pre- determined (exogenous) characteristics such as schooling level and age. Wages can be specied as W f i =w f i (Z f i ;u f i ) (2.6) W m i =w m i (Z m i ;u m i ) (2.7) 26 in which Z f i , and Z m i are the vector of observable characteristics of female and male members, respectively. They include variables, such as schooling and age. u f i , and u m i are mean zero constant variance disturbances. If the disturbances for wages and tastes in Equations (2.5), (2.6), and (2.7), i.e. v j i and u j i (j = f;m), are correlated, the OLS estimates of ' 0 f and/or ' 0 m are biased and inconsistent. To solve this problem, I employ instrumental variables to estimate wages rst and then use the predicted wages for all individuals to estimate labor force participation (Equation (2.5)). This procedure is similar but not identical to the two-stage least square method (2SLS). Here, after estimating the rst stage, I predict the dependent variable, i.e. wages, for everybody, that is all observations regardless of working status. But in the 2SLS, we only predict wages for those observations used to estimate the rst stage, i.e. those who had reported wages. In this \pseudo" two-stage least square, the instrument is individual education. It can be argued that education may only explain labor force participation through wages. In other words, we assume that education does not aect the decision to work by itself, but only through changing wages. More education increases wages and increased wages in turn increase the incentives to work. 1 In Section 2.4.1, I utilize this method to estimate a simpler form of Equation (2.5). One can substitute normalized wage functions (Equations (2.6) and (2.7)) in Equation (2.4) and obtain the following: L j i =L i (V i ;Z f i ;Z m i ;Y i ;e i ;u f i ;u m i ) j =f;m (2.8) 1 This assumption may not be true. Becker argued that we should take preferences as given. 27 which can in turn be linearized into L j i = + 0 V j i + 0 Z f i + 0 Z m i +r 0 Y i +e i +w j i j =m;f (2.9) in whichw j i is a mean zero constant variance disturbance that includes u f i andu m i . This reduced form gives an understanding of how household and individual characteristics aect labor force participation and in particular FLFP. In the beginning of Section 2.4, I report estimates of a simplied form of this equation. 2.3 Data Statistical Center of Iran (SCI) is the main organization in charge of gathering micro datasets in Iran. As this body of government is one of the two major agencies which are responsible for measuring macro and micro socio-economic indicators 2 , these relatively large datasets are gathered with substantial eort and resources. With about ve decades of experience, this center provides quality datasets and statistics for policy makers as well as the general public. 3 In this study, I use two major household datasets, 1) the household expenditure and Income Surveys (HEIS), a series of cross-sectional datasets conducted annually, and 2) the socio-economic characteristics of households (SECH), which are household panel datasets. 2 The other agency is the Central Bank of Iran. 3 Until recently these raw datasets were unavailable to researchers, but thanks to new rules, most of them are currently accessible. Compiled statistics are available on their website, http://amar.sci.org. ir/ 28 2.3.1 Household Expenditure and Income Surveys (HEIS) This annual survey has been gathered since 1963 in rural areas and 1968 in urban areas but only 1990 through 2004 was made available to me. Each year, new samples are drawn from the population. The samples are nationally representative and are stratied at the urban and rural areas of each province. Sample selection follows a two-stage sampling method which has remained the same over the years. In the rst stage, the total number of primary sampling units (PSUs) in each geographical block (rural or urban areas in each province) are determined based on the most recent census. This number is equal to the population in the block divided by ve. Each PSU corresponds to a census track and consists of ve households. In the second stage, a number of PSUs in each block are randomly chosen to be surveyed. This number is related to the population of the block and variance of some of the variables of interest such as food expenditure in that block. Hence, households have dierent probability of selection. For instance, more rural households have been selected. Probability of selection is known for each household and is taken into account whenever national average of a variable, like lfp, is estimated in this study. Data gathering process is done uniformly throughout the year so that 1 12 of all house- holds are surveyed each month. The number of households for years 1990 through 2006 is reported in Table 2.1 and varies from 12,763 in 1993 to 36,579 in 1998. The questionnaire consists of about 600 elds in eight sections. The rst section contains basic demographic information, including gender, age, education, relationship to the household head, and marital and employment status, for each member of the 29 Table 2.1: Number of Households in Each Household Expenditure and Income Survey Year No. of Households 1990 18,439 1991 18,672 1992 18,653 1993 12,763 1994 19,904 1995 36,572 1996 21,963 1997 21,949 1998 17,477 1999 27,464 2000 26,941 2001 26,961 2002 32,152 2003 23,134 2004 24,534 2005 26,895 2006 30,910 household. The second section has data on home, appliances, car, motorcycle, and bike ownerships as well as access to electricity, piped water, gas, telephone, and in recent years, internet. The ownership questions are yes/no questions, similar to demographic and health surveys (DHS). Section 3 has extensive detailed information on food expenditure and section 4 has expenditure on non-food non-durable or semi-durable goods such as clothing, rent, and utilities. Section 5 is on durable goods expenses including appliances, furniture, and vehicles, as well as services such as vacation, and schooling tuition, in the last year. Sections 6, 7, and 8 respectively contain information on self employment, salary, and other types of income, including retirement pensions, rent, transfers and interest. Over the years, these surveys became richer as more questions were added. 30 Until 2005, the denition of employment in these surveys was limited to a person who works for at least two days during the week prior to the survey. Since the adoption of ILO standards in 2005, everyone who is working for at least one hour during the week prior to the interview is considered employed. As the purpose of these surveys was mainly to calculate indices such as in ation, unfortunately, they were week on other sections, such as assets and labor market participation. For instance, hours worked was not asked until 2005. Hence, it is not possible to calculate hourly wages, a major component of labor market analysis with these surveys. 2.3.2 Socio-Economic Characteristics of Households Surveys (SECH) Socio-Economic Characteristics of Household (SECH), which are panel data sets similar to HEIS but with more questions on demographics, individual characteristics, and labor market participation. For example, this is one of the few surveys that include variables such as age at rst marriage, at rst pregnancy and at rst birth, children ever born, disability, and hours worked to name a few. Since the 1979 revolution, three separate SECH datasets have been gathered, each time with a dierent set of households: 1987-89, 1992-95, and 2001-03. Waves of each panel are gathered during the months of November in the years covered by the panel. SECH 1992-95 is chosen to be the primary dataset for this and the next chapter since it contains hours worked in addition to income, providing the opportunity to compute wages, an important labor market participation variable. HEIS does not have such information. First wave of SECH 1992-95 is a nationally random sample of 5090 households, en- compassing 172 sampling clusters (62 rural and 109 urban), each having an average of 31 about 30 households. Almost 73% of rural clusters and 49% of urban clusters have exactly 30 households. Table 2.2 shows the distribution of number of households in clusters. In each cluster, selected households are neighbors living side-by-side. Table 2.2: Distribution of Number of Households in Each Cluster Number of Frequency (in %) Households Rural Urban Less than 30 12 23 30 73 49 More than 30 15 28 Total 100 100 In successive rounds, some households were not part of the survey any more. This attrition, which I will extensively discuss in Chapter 3, may make the successive waves of SECH 92-95 not representative of the population. To have a more representative sample in each wave, by design, a few randomly selected households were added to the sample. Table 2.3 reports on the number of households added in each wave. Only 2.3% of the households ever interviewed were not part of the survey in the rst wave and were added in later waves. Table 2.3: Distribution of the Households Ever Interviewed Based on the Wave They Were Added to the Panel Number Percent Cumulative 1 5,090 97.60 97.60 2 58 1.11 98.71 3 16 0.31 99.02 4 51 0.98 100.00 32 The questionnaire is the same for all waves of this panel data and has 8 sections. The rst section has extensive demographic as well as labor market information, including age at rst marriage, pregnancy, and birth, exact birth and death dates of children ever born, family planning, rst and second jobs, hours worked, job locations, leisure activities, and unemployment duration. Second section contains data on adult activities in home production, deaths occurred in the household, and Yes/No questions on home, appliances, and amenities ownership. Section 3 has questions on the total income and expenditure, and sections 4 and 5 asked about expenditure on food and expenditure on services and semi-durables respectively. Section 6 contains data on investments during the prior year and the last two have information on salary, self-employment, and other types of income, including retirement pensions, transfers, interest, and income from making handicraft products such as carpets. The denition of employment in SECH 1992-95 is the same as the denition used prior to 2004 for HEIS. This makes these surveys comparable in terms of employment. Since the denition of employment has changed in 2005 onward, it is dicult to compare the results of years prior to 2005 with the years after. Therefore, I only use HEIS data before 2005. On the other hand, SECH 92-95 allows for the estimation of a structural labor force participation model as it permits calculation of hourly wages. Hence, for the years 1992 through 1995, I employ SECH 92-95 instead of HEIS. Moreover, as the analysis in Chapter 3 is based on SECH 92-95, for the purpose of consistency throughout this study, I adopt SECH 92-95 in this chapter as well. Therefore, here, I only use HEIS data for the years 1990 through 2004 and employ SECH 1992-95 for estimating the structural model and also the dynamic model in Chapter 3. 33 2.4 The Evidence To begin with, let us estimate the reduced form, i.e. Equation (2.9), with labor force participation dummy as the dependent variable instead of leisure hours. Tables 2.5 and 2.6 depict the coecients of the linear probability model of labor force participation for women aged 21 through 65 for each year between 1990 and 2004. Labor force participation is a dummy variable equal to one if the individual participates and zero otherwise. In this case,e i , andw j i would be part of the error term and are assumed to be independent of the covariates. Ignoring Z m i for females, the covariates include education, age, urban-rural location, female head dummy, number of female and male household members between age 15 and 18 and above age 18, an index for assets, home value, and province xed eects. Education is a categorical variable which reports the completed or partially completed level of education. The levels of education are \Illiterate", \Primary", \Mid School", \High School", and \College and higher". Dummy variables for each of these categories were dened and employed to estimate the linear model. Since the dependent variable is a dummy variable, some continuous variables such as age were transformed to form categorical variables. In other words, age was divided into ten-year categories, 20 to 30, 30 to 40, etc., and dummy variables were introduced for each age group. For instance, I[20<Age30] represents a dummy variable which is equal to one if age is between 20 and 30 (including 30 but excluding 20) and zero otherwise. Women aged 61 through 65 are the control group. 34 Location is identied by a simple dummy, \Urban", which is equal to one if the household lives in an urban area and zero otherwise. Number of teenage female and male members (between age 15 and 18) is included in the regression as this is the age in which many households, especially in urban areas, allocate so many resources to their children's education so that they would have a higher chance of entering college. Numbers of adult female and male members (above age 18) are also included in this regression as a proxy for the household available resources that can be used to generate income. Wealth and assets are the other variables that should be included in the analysis. As noted in Section 2.3, assets are simply appliances (fridge, stove, TV, radio, cassette player, computer, etc.) , and transportation vehicles (car, motorcycle, and bike) as well as house amenities such as access to electricity, piped water, gas, telephone, and in recent years internet. Assets are not reported in terms of their value but rather as dichotomous ownership dummies, i.e. one if household owns the specic asset and zero otherwise. One way of accounting for these variables is to dene a new variable that is the sum of these dummy variables. In this method, technically, there is no dierence between owning a car or owning a radio as both increase the new dened variable by one unit. But since households who have cars are very likely to own simpler and cheaper assets such as radio as well, number of assets owned (sum of the ownership dummies) is possibly a good measure of wealth. However, instead of giving the same weights to all assets, one can give dierent weights to various assets based on how much the ownership of an asset signals the wealth of the household. Sahn and Stifel (2000) use factor analysis to nd weights for each type of asset. Their main argument is that weights should be determined by the data that is 35 being used. As a second approach, I followed their method and constructed the asset index using weights computed for households in 1992. Interestingly, the results are qualitatively similar using both of these approaches. Therefore, I only report the results using the asset index computed by Sahn and Stifel (2000) factor analysis method. In this method, assets owned by many households get small or negative weights. The fewer people owning an asset, the larger the weights will be, as it shows that the asset is more valuable. I report the weights for each asset in Table 2.4. Table 2.4: Weights of Various Assets Used to Construct an Asset Index for HEIS Data Series following Sahn and Stifel (2000) Weights Car 0.3328 Motorcycle 0.0057 Bicycle 0.1795 Black & White TV 0.2176 Color TV 0.4856 Radio -0.2459 Cassette Player 0.2585 Refrigerator 0.4236 Freezer 0.4505 Gas stove -0.0160 Washing Machine 0.4612 Vacuum Cleaner 0.4638 Sewing Machine -0.0037 Telephone 0.3712 Electricity -0.8682 Piped Gas 0.2938 Air Conditioner 0.3101 Central Heating 0.4665 Bath 0.3142 Note: Each of the assets is a dummy variable equal to one if the households owns it and zero otherwise. \Owned home value", the other covariate in the reduced form regression, is the answer to the question, \If you wanted to rent this house, how much would be the rent?" It is a 36 proxy for the value of the house (or wealth in general) owned by the household. Province xed eects were also used to account for any xed unobservable factor in the province, especially province labor market. Since these are xed eects, the robust-heteroskedastic standard errors are corrected for correlation within province. 37 Table 2.5: Linear Probability Model of Labor Force Participation for Women Aged 21 through 65 with Province Fixed Eects, 1990-1996 1990 1991 1992 1993 1994 1995 1996 Primary 0.018 0.023 0.010 0.015 0.016* 0.008 -0.002 (0.016) (0.016) (0.013) (0.019) (0.008) (0.014) (0.011) Mid School 0.009 0.029* -0.019 0.019 0.025 y 0.002 -0.007 (0.013) (0.012) (0.017) (0.019) (0.014) (0.014) (0.014) High School 0.244*** 0.255*** 0.216*** 0.246*** 0.227*** 0.183*** 0.166*** (0.016) (0.016) (0.022) (0.024) (0.018) (0.017) (0.020) College & higher 0.531*** 0.551*** 0.553*** 0.570*** 0.538*** 0.514*** 0.503*** (0.035) (0.041) (0.028) (0.035) (0.027) (0.025) (0.031) I[20<Age30] 0.048* 0.028 0.048* 0.076* 0.053* 0.053* 0.053* (0.022) (0.020) (0.022) (0.029) (0.025) (0.022) (0.022) I[30<Age40] 0.083** 0.065*** 0.087*** 0.130*** 0.092*** 0.100*** 0.097*** (0.023) (0.015) (0.017) (0.026) (0.019) (0.019) (0.018) I[40<Age50] 0.073** 0.065*** 0.090*** 0.116*** 0.104*** 0.087*** 0.095*** (0.022) (0.014) (0.013) (0.017) (0.014) (0.014) (0.014) I[50<Age60] 0.042** 0.023* 0.019* 0.063*** 0.056*** 0.036*** 0.054*** (0.013) (0.010) (0.009) (0.012) (0.010) (0.009) (0.012) Urban -0.092** -0.099** -0.097*** -0.148*** -0.108** -0.122*** -0.151*** (0.026) (0.026) (0.023) (0.033) (0.031) (0.029) (0.037) No. of females b/w 15 to 18 0.018*** 0.025*** 0.011 y 0.020* 0.012* 0.019** 0.009 (0.005) (0.006) (0.006) (0.008) (0.006) (0.006) (0.006) No. of males b/w 15 to 18 0.002 0.005 -0.002 -0.000 -0.003 0.003 -0.003 (0.007) (0.005) (0.006) (0.008) (0.006) (0.005) (0.005) Continued on next page 38 Table 2.5 { continued from previous page 1990 1991 1992 1993 1994 1995 1996 No. of females above age 18 0.028*** 0.023*** 0.028*** 0.020** 0.026*** 0.024*** 0.023*** (0.005) (0.005) (0.005) (0.006) (0.005) (0.004) (0.004) No. of males above age 18 -0.037*** -0.042*** -0.038*** -0.036*** -0.034*** -0.025*** -0.026*** (0.008) (0.006) (0.005) (0.007) (0.004) (0.005) (0.006) Asset Index -0.019** -0.016* -0.014* -0.026*** -0.022*** -0.021*** -0.018* (0.007) (0.007) (0.006) (0.007) (0.004) (0.004) (0.007) Owned Home Value10 6 -0.030 0.011 -0.095 -0.033 -0.113* -0.158* -0.096* (0.061) (0.016) (0.110) (0.090) (0.051) (0.058) (0.042) Constant 0.122*** 0.143*** 0.150*** 0.170*** 0.135*** 0.174*** 0.185*** (0.015) (0.016) (0.019) (0.018) (0.027) (0.020) (0.019) Province FE Yes Yes Yes Yes Yes Yes Yes Observations 19705 20126 19714 13566 21628 39639 23780 Average FLFP 0.141 0.133 0.157 0.181 0.168 0.199 0.187 (0.003) (0.003) (0.003) (0.004) (0.003) (0.002) (0.003) Note: Dependent variable is a dummy equal to one if the individual was looking for a job or was working for at least two days in the week prior to the survey, and zero otherwise. \Primary", \Mid School", \High School", and \College & higher" are education dummy variables. Illiterates are the control group. I[a<Ageb] is a dummy variable equal to one if Age is between a and b (b included) and zero otherwise. Women aged 61 through 65 are the control group. \Asset" index is computed following Sahn and Stifel (2000). \Owned home value" is the answer to the questions \If you wanted to rent this house, how much would be the rent?" Robust heteroskedastic standard errors are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 39 Table 2.6: Linear Probability Model of Labor Force Participation for Women Aged 21 through 65 with Province Fixed Eects, 1997-2004 1997 1998 1999 2000 2001 2002 2003 2004 Primary 0.006 -0.007 0.016 y 0.009 -0.001 0.013 0.025* -0.016 (0.013) (0.011) (0.009) (0.010) (0.012) (0.009) (0.011) (0.012) Mid School 0.001 -0.027 y 0.015 0.015 -0.021 0.023 0.007 -0.016 (0.015) (0.015) (0.012) (0.009) (0.014) (0.014) (0.013) (0.013) High School 0.140*** 0.119*** 0.142*** 0.141*** 0.112*** 0.151*** 0.144*** 0.089*** (0.017) (0.024) (0.017) (0.018) (0.020) (0.021) (0.019) (0.020) College & higher 0.501*** 0.456*** 0.505*** 0.472*** 0.430*** 0.477*** 0.475*** 0.394*** (0.031) (0.035) (0.022) (0.023) (0.026) (0.020) (0.025) (0.028) I[20<Age30] 0.066** 0.047* 0.039 y 0.043* 0.045* 0.030 0.041* 0.048** (0.021) (0.022) (0.021) (0.016) (0.018) (0.020) (0.015) (0.017) I[30<Age40] 0.113*** 0.100*** 0.085*** 0.090*** 0.079*** 0.064*** 0.089*** 0.090*** (0.019) (0.019) (0.017) (0.015) (0.015) (0.017) (0.014) (0.017) I[40<Age50] 0.122*** 0.110*** 0.088*** 0.105*** 0.093*** 0.070*** 0.087*** 0.102*** (0.019) (0.015) (0.017) (0.014) (0.015) (0.016) (0.014) (0.017) I[50<Age60] 0.076*** 0.053*** 0.058*** 0.056*** 0.041** 0.036** 0.056*** 0.062*** (0.016) (0.012) (0.013) (0.013) (0.012) (0.011) (0.013) (0.015) Urban -0.139*** -0.106*** -0.106*** -0.105*** -0.110*** -0.095*** -0.076** -0.111*** (0.033) (0.026) (0.022) (0.021) (0.023) (0.022) (0.022) (0.023) No. of females b/w 15 to 18 0.012** 0.006 0.013* 0.012* 0.019*** 0.017*** 0.015*** 0.014** (0.004) (0.007) (0.006) (0.005) (0.005) (0.004) (0.004) (0.005) No. of males b/w 15 to 18 -0.004 -0.010 -0.001 -0.003 0.004 0.006 0.012* 0.008 (0.003) (0.007) (0.004) (0.004) (0.004) (0.004) (0.005) (0.005) Continued on next page 40 Table 2.6 { continued from previous page 1997 1998 1999 2000 2001 2002 2003 2004 No. of females above age 18 0.030*** 0.025*** 0.021*** 0.025*** 0.021*** 0.022*** 0.034*** 0.035*** (0.005) (0.006) (0.004) (0.004) (0.005) (0.004) (0.007) (0.005) No. of males above age 18 -0.024*** -0.023*** -0.025*** -0.027*** -0.023*** -0.015*** -0.014** -0.024*** (0.004) (0.005) (0.005) (0.004) (0.003) (0.003) (0.005) (0.003) Asset Index -0.030*** -0.024** -0.026*** -0.028*** -0.028*** -0.028*** -0.036*** -0.031*** (0.007) (0.007) (0.007) (0.006) (0.007) (0.005) (0.006) (0.006) Owned Home Value10 6 0.071 y -0.051 -0.054 -0.016 -0.082 y -0.011 -0.066 -2.756 (0.035) (0.048) (0.077) (0.073) (0.045) (0.033) (0.052) (4.862) Constant 0.156*** 0.154*** 0.160*** 0.151*** 0.173*** 0.133*** 0.110*** 0.171*** (0.013) (0.020) (0.018) (0.015) (0.019) (0.017) (0.020) (0.020) Province FE Yes Yes Yes Yes Yes Yes Yes Yes Observations 23960 19468 30611 29926 30318 36661 26268 28087 Average FLFP 0.191 0.175 0.187 0.181 0.177 0.179 0.194 0.202 (0.003) (0.003) (0.003) (0.004) (0.003) (0.002) (0.003) (0.003) Note: Dependent variable is a dummy equal to one if the individual was looking for a job or was working for at least two days in the week prior to the survey, and zero otherwise. \Primary", \Mid School", \High School", and \College & higher" are education dummy variables. Illiterates are the control group. I[a<Ageb] is a dummy variable equal to one if Age is between a and b (b included) and zero otherwise. Women aged 61 through 65 are the control group. \Asset" index is computed following Sahn and Stifel (2000). \Owned home Value" \Owned home value" is the answer to the questions \If you wanted to rent this house, how much would be the rent?" Robust heteroskedastic standard errors are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 41 As Tables 2.5 and 2.6 show, generally, women with primary and middle school edu- cation are not likely to work more than illiterate women (the control group). However, there are some discrepancies. For instance, in 1994, 1999, and 2003, women with primary education were slightly more likely to work than illiterate women. The same was true for women with middle school education in 1991, and 1994. In 1998 the opposite was the case. But, interestingly, the coecients are small and it can be safely inferred that primary and middle school education may not change the likelihood of a woman working. As education increases to high school and higher, participation increases signicantly. The coecient of high school is around 0.2 before 1997 and decreases to about 0.14 afterwards. College education coecient is almost two times that of high school before 1997 and becomes three times that afterwards. Generally, it uctuates around 0.5. 4 These coecients are particularly large, especially compared to any other coecient reported. For instance, they show that women with tertiary education are about 50% more likely to work than illiterate women. The education prole is increasing and convex. There are ve age groups: 21 to 30, 31 to 40, 41 to 50, and 51 to 60, and 61 to 65 (the control group). As the coecients show, women aged 61 to 65, i.e. the control group, are less likely to participate in the labor force than other age groups. Moreover, consistent with the evidence from other countries and the economic theory, they show that the age prole is generally concave. In other words, women aged 31 to 50 are more likely to participate than any other group. Being in an Urban area decreases the probability of participation. This is certainly consistent with Figure 1.5. Having more teenage females in the households (females aged 4 There is a decline in the coecients of high school and college in 2004. 42 between 15 and 18) increases the likelihood of participation for women slightly, while more teenage males does not change this likelihood. This is particularly interesting. As households, especially in urban areas, invest more on their children's education in this age group, we expect that more resources, both money and time, are spent on these children. On one hand, households need more income to be able to pay for the tuition of quality high school education as well as future college tuitions, and on the other hand, they would like to spend more time with their kids improving the quality of their education through home schooling. The former will entice mothers to participate in the labor force while the latter persuades them to stay at home. Moreover, households are saving more for their daughters' future dowry when they are at this age. This motivates women to work more. In addition, teenage females in the household may contribute to home production and make more free time for their mothers available. These two reasons entice women to work. The empirical results show that from the these forces, those that motivate women to work are stronger. On the other hand, we do not see any correlation between the number of teenage males in the household and FLFP. Presence of one more adult female in the household increases the likelihood of women working by about three percent (more than twice the coecient of teenage females). This is not large nor surprising, as more adult females in the household would increase the number of people who can potentially contribute to home production and hence provides more free time to more female members to contribute to labor market. In addition, since these adult females are at the age of marriage, saving for their dowries is a strong motivation to work for all members of the household. Interestingly, more adult male members have negative correlation with FLFP. The coecient is almost as large as 43 the coecient for adult females but in opposite direction. Adult males in the household usually contribute to household income by working in the labor market. Since households with more adult males may have more sources of income, they are less in need of income brought by adult females of the household and hence fewer women would work in such households. Not surprisingly, the asset index computed following Sahn and Stifel (2000) has a negative coecient, implying that owning more assets decreases the likelihood of partici- pation. Using equal weights for each type of asset does not aect the results qualitatively. The results are consistent with the theory and the evidence from other countries. Marginal eects of owned home value are negative but quite small and insignicant. 5 One expects these coecients to be negative and signicant. This dierence might be because of selection. Households with large amount of wealth are a selected group with dierent unobservable factors than the average households. Women in such households are likely to work more because of those unobservable factors that are correlated with wealth. So the coecient of wealth may capture the eect of these unobservable factors as well. Overall, the results show that FLFP characteristics in Iran are quite similar to those of other countries. Moreover, although none of the marginal eects represents a causal relationship, they follow the predictions of a standard simple economic theory. 5 The coecients become signicant when I do not employ province xed eects. 44 2.4.1 Estimating the Structural Model for LFP In addition to estimating the reduced form, Equation (2.9), one may be interested in the direct economic factors that aect FLFP. In other words, one may want to understand the relationship that is predicted by the model in Equation (2.4) or its linear form Equation (2.5). One of the main predictors of participation according to this model is wages. Estimating the eect of wages on participation is dicult as we do not observe wages for those who do not participate. Moreover, the error terms in Equations (2.5), (2.6), and (2.7), v j i and u j i (j =f;m), may be correlated with each other making OLS estimates of linear equation biased. To overcome these issues, as explained in Section 2.2, I employ education as an in- strumental variable to predict wages and use those predicted wages instead of the actual wages to estimate Equation (2.5). This is not exactly the same as a 2SLS method. The reason is that while I can only use observations that reported wages to estimate the wage equation (rst stage), I predict wages, using these estimates, for all observations, i.e. everyone regardless of whether they reported wages or whether they worked at all. Then, I use these predictions in the second stage instead of wages. If it was a 2SLS method, I should only have predicted wages for those observations used in the rst stage, i.e. those who reported wages. Then I would have used only those predictions in the second stage. In other words, I would have run the labor force participation regression only for those who reported wages, i.e. those who worked. This is not possible, as the dependent variable in the second stage would be constant and equal to one. 45 In order to implement this \pseudo" two-stage least square method, I need to correct for selection on wage variable in the rst stage, as wages are not observed for everyone. So, in the rst stage, I estimate a two-step Heckman selection model of whether wages are observed for an individual. Estimating the structural model has the following steps: 1. In the rst step of this Heckman selection model, inverse Mills ratios can be com- puted using a probit model in which household characteristics are the selection identifying variables. 2. In the second step, I estimate the log of hourly wages using education as explanatory variables and correcting for selection by inverse Mills ratios. 3. Using this regression, I then predict the log of wages for all women regardless of participation status, and substitute these predictions instead of actual wages in the second stage, i.e. Equation (2.5). In order to satisfy the exclusion restriction in the second stage regression, I assume that education does not predict labor force participation directly, but only through wages. This three-step procedure would give us incorrect standard errors in the second stage. Therefore, I use bootstrapping for the whole procedure, to obtain correct standard errors in the second stage. First, a bootstrapped sample is drawn from the population. Then, a Heckman two-step selection model of log of wages is estimated on this sample. Using the estimates, log of wages are predicted for all observations. Then, employing the same bootstrapped sample, the linear regression of labor force participation on predicted log of wages and other covariates are estimated. This procedure is repeated a thousand times, 46 each time with new bootstrapped samples to be able to compute the correct standard errors for each regression. Resampling is implemented based on the clusters. Since we can only calculate hourly wages from the SECH 92-95, I am using the pooled data for all four years to estimate this structural model. Tables 2.7 and 2.8 report the Heckman two-step selection model on wages. Specically, Table 2.7 reports the Probit regression of whether wages are observed and Table 2.8 represents the rst stage which I use to predict wages. Table 2.9 reports the second stage in which I estimate the partial elasticities of wages on labor force participation. In all these regressions, standard errors are corrected for correlation inside clusters. As depicted in Table 2.7, for the Probit regression of the Heckman selection model of wages, I use the same covariates that are used in the reduced form estimates in Tables 2.5 and 2.6 as well as a dummy variable called \Crisis Years". This dummy is equal to one for the years 1994 and 95 when a minor economic crisis happened and zero for the other two years, 1992 and 93. \Asset Index" variable is recalculated with new weights that are computed for SECH 92-95 Panel Data. These weights are reported in Appendix Table A.1. In this regression, the selection on log of wages is identied using household composition and asset variables. 6 Table 2.7 shows that the coecients for women partly resembles those in the reduced form for labor force participation. Since there were few women in rural areas with college and higher level degrees, I combined them with women with High School education. The education variables are especially signicant for females in urban areas. Age prole of reporting wages is close to concave for both men and women. Number of teenage males 6 These are No. of females and males aged between 15 and 18, and above 18, asset index, and owned home value. 47 and females as well as adult females is not correlated with observed wage, but No. of adult males is. Interestingly, a larger asset index increases the likelihood of an observed wage for men in urban and rural areas but decreases it for urban women. The 2 statistics for all four regressions show that the coecients of selection identifying variables, i.e. household composition and asset variables, are jointly signicant. Table 2.7: First Step of the Two-Step Estimation of Heckman Selection Model on Ob- served Wages across People Aged 21 through 65, Pooled Data 1992-95 Rural Urban Women Men Women Men Primary 0.293* 0.039 0.285** 0.194* (0.138) (0.091) (0.106) (0.082) Mid School -0.109 -0.152 0.401** 0.185* (0.195) (0.112) (0.124) (0.091) High School 0.718*** -0.294** 1.012*** -0.082 (0.202) (0.100) (0.126) (0.089) College & higher -0.443** 1.880*** -0.441*** (0.164) (0.128) (0.100) Crisis Years 0.184 y 0.029 -0.007 -0.020 (0.103) (0.099) (0.078) (0.059) Trend -0.016 -0.010 -0.012 -0.045 (0.042) (0.045) (0.036) (0.028) I[20<Age30] 0.194 -0.582*** -0.077 0.079 (0.158) (0.125) (0.171) (0.093) I[30<Age40] 0.216 y -0.037 0.243 0.812*** (0.126) (0.130) (0.167) (0.094) I[40<Age50] 0.208 y 0.130 0.227 0.707*** (0.118) (0.122) (0.154) (0.080) No. of females b/w 15 to 18 0.009 -0.005 -0.050 -0.002 (0.059) (0.047) (0.048) (0.038) No. of males b/w 15 to 18 0.061 -0.068 -0.145*** -0.058 (0.059) (0.045) (0.043) (0.037) No. of females above age 18 -0.016 -0.064 0.062 -0.034 (0.055) (0.040) (0.044) (0.023) Continued on next page 48 Table 2.7 { continued from previous page Rural Urban Women Men Women Men No. of males above age 18 -0.338*** -0.362*** -0.138** -0.276*** (0.064) (0.039) (0.044) (0.026) Asset Index 0.172 0.160*** -0.091* 0.088** (0.107) (0.048) (0.046) (0.030) Owned home value10 6 -2.168 -1.827** -0.388* -0.420*** (2.071) (0.591) (0.153) (0.108) Constant -0.942*** 2.253*** -1.647*** 1.160*** (0.212) (0.161) (0.165) (0.108) Observations 6345 6401 10009 10630 2 test for identifying variables z 40.43*** 189.86*** 27.81*** 191.44*** Note: Dependent variable is a dummy equal to one if the individual's wage is observed in the data and zero otherwise. \Crisis Years" is a time dummy equal to one for 1994 and 95, and zero otherwise. For a description of other covariates, please see Table 2.5. SECH 1992-95 is the data used. As there were few women in rural areas with \college & higher" level of education, they are combined with rural women with high school education. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z This 2 statistic tests whether the coecients of selection identifying variables, i.e. No. of females and males b/w 15 to 18, No. of females and males above age 18, Asset Index, and owned home value, are jointly signicant. Table 2.8 reports the regression of log of real wages on education, crisis dummy variable, trend, and age dummies, using the inverse Mills ratio to correct for selection. 7 This is like a rst stage of a two-stage least square. Education dummies are instruments for wages as they are assumed to be independent of the decision to participate in the labor force and hence excluded from the second stage, i.e. LFP regression in Table 2.9. Education coecients are increasing as the level of education increases. Age prole is concave with its maximum usually at ages above 50. The 2 statistics for testing the joint signicance of instruments are signicant in all regressions except for rural women. 7 This regression is estimated jointly with the Probit Regression using Stata heckman command to get correct standard errors. 49 When signicant, the statistics are larger than 10, the threshold below which instruments are generally considered weak. Table 2.8: Estimation of Log of Real Wages for People Aged 21 through 65 Controlling for Heckman Selection on Wages (Two-Step Method; First Step in Table 2.7), Pooled data 1992-95 Rural Urban Women Men Women Men Primary 0.279 0.147 0.259 0.180** (0.179) (0.092) (0.223) (0.063) Mid School 0.242 0.280* 0.514** 0.230** (0.247) (0.110) (0.174) (0.071) High School 0.527 y -0.134 1.096*** 0.239** (0.274) (0.162) (0.256) (0.073) College & higher 0.907*** 1.931*** 0.644*** (0.169) (0.420) (0.099) Crisis Years 0.087 0.004 0.222 0.105* (0.213) (0.094) (0.140) (0.052) Trend -0.167 y -0.011 -0.245*** -0.160*** (0.101) (0.050) (0.056) (0.027) I[20<Age30] -0.454 y -0.317** -0.270 -0.832*** (0.268) (0.110) (0.194) (0.074) I[30<Age40] -0.244 0.025 0.222 -0.529*** (0.272) (0.084) (0.212) (0.097) I[40<Age50] -0.028 -0.014 0.398* -0.317*** (0.311) (0.068) (0.200) (0.084) Constant 10.777*** 10.490*** 9.011*** 11.647*** (0.691) (0.119) (0.690) (0.123) 2 test for education z 4.47 39.77*** 25.32*** 45.83*** Note: Dependent variable is the individual's log of real wage. For a description of covariates, please see Tables 2.7 and 2.5. SECH 1992-95 is the data used. Education dummies are the instruments for log of real wage in the second stage in Table 2.9. As there were few women in rural areas with \college & higher" level of education, they are combined with rural women with high school education. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z This 2 statistic tests whether education dummies which are the instruments for log of wage in the regression in Table 2.9, are jointly signicant. 50 Using these estimates, I predict wages for all individuals between the ages of 21 and 65, even if they were not participating in the labor force, i.e. wages were not observed for them. These predicted wages are used in the structural model, i.e. Equation (2.5), instead of log of wages. Other covariates are crisis dummy, trend, age dummies, number of teenage and adult females and males in the household, asset index, and owned home size. A linear probability model with cluster random eect is used to control for correlation inside clusters. Table 2.9 depicts these regressions. Table 2.9: Linear Probability Model of Labor Force Participation on Predicted Log of Real Wages from Regression in Table 2.8 for People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors), Pooled data 1992-95 Rural Urban Women Men Women Men ln(Wage) -0.018 -0.080 0.259*** -0.303*** (0.151) (0.066) (0.061) (0.062) Crisis Years 0.082 y -0.010 -0.055 0.026 (0.046) (0.012) (0.043) (0.018) Trend -0.018 0.001 0.059** -0.056*** (0.037) (0.006) (0.023) (0.013) I[20<Age30] 0.107 0.016 0.024 -0.034 (0.081) (0.022) (0.041) (0.049) I[30<Age40] 0.106 y 0.065*** -0.031 0.109** (0.057) (0.018) (0.050) (0.038) I[40<Age50] 0.083 0.047** -0.062 0.136*** (0.051) (0.014) (0.048) (0.030) No. of females b/w 15 to 18 0.016 0.018** -0.007 0.006 (0.012) (0.006) (0.009) (0.008) No. of males b/w 15 to 18 0.013 -0.005 -0.017* -0.005 (0.012) (0.006) (0.008) (0.008) No. of females above age 18 0.024* 0.009 y 0.030** 0.003 (0.010) (0.006) (0.009) (0.006) No. of males above age 18 -0.030*** -0.026*** -0.018** -0.043*** Continued on next page 51 Table 2.9 { continued from previous page Rural Urban Women Men Women Men (0.008) (0.006) (0.006) (0.006) Asset Index 0.013 0.023** -0.030** 0.018** (0.016) (0.007) (0.010) (0.007) Owned home value10 6 -0.028 -0.142 -0.048 y -0.052* (0.164) (0.088) (0.028) (0.026) Constant 0.380 1.799** -2.293** 4.363*** (1.716) (0.690) (0.726) (0.724) Observations 6345 6401 10009 10630 Average LFP 0.253 0.951 0.149 0.891 (0.005) (0.003) (0.004) (0.003) Elasticities -0.071 -0.084 1.738 -0.340 Note: Dependent variable is a dummy equal to one if the individual is participating in the labor force and zero otherwise. For a description of other covariates, please see Tables 2.5 and 2.7. SECH 1992-95 is the data used. ln(Wage) is the predicted log of real wages from the rst stage regression reported in Table 2.8. Education dummies are the instruments for log of real wage in the second stage in Table 2.9. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 The main coecient of interest, coecient of log of real wages, is dierent across women and men of urban or rural areas. Interestingly, women's and men's wage rates in rural areas are not a predictor of their participation status. However, for women in the urban areas, the hourly wage has a large signicant coecient. For every one percent rise in wages, the likelihood to participate increases by 0.003. Average LFP rates and elasticities computed at the average LFP are reported at the bottom of the table. They are particularly quite large for urban women as one percent increase in wages increases LFP by 1.7%. Urban men with higher wages, on the other hand, are likely to work less. Looking at elasticities, one observes that a one percent increase in wages, decreases urban 52 men's participation by 0.3%. Interestingly, These results are quite similar to the evidence found for the United States. Note that men's wages in rural areas may not be predicted accurately, since household members working as unpaid family labor contribute to household income, but this income is reported as the income of only one individual in the household, i.e. household head. Thus, the hourly wages for heads in such households are overestimated. Age prole is concave for men in both areas but is almost at for women. No. of adult females and males in the household increases and decreases participation respectively, similar to the reduced form equations. More teenage females and males does not change participation. Assets are negatively correlated with urban women participation but positively correlated with men's. Home value has generally negative correlation with participation for almost everyone. These results, for the most part, are consistent with the expectations and comparable to evidence presented in Tables 2.5 and 2.6. Overall, we observe that not only does the reduced form follows the economic theory, but also the structural labor force participation model gives similar results that have been found for the developed countries, including the United States (Killingsworth and Heckman, 1986; Heckman, 1980; Dooley, 1982; Stelcner and Brestaw, 51; Franz, 1981; Franz and Kawasaki, 1981; Renaud and Siegers, 1984). After all, leaving the low labor force participation aside, there is little dierence among Iran and other countries, even the developed ones. As explained in Chapter 1, there is a signicant dierence between the married and unmarried women's participation rates. The eect of marriage may be the result of discriminatory and/or non-discriminatory factors. But whatever caused it, one may be 53 inclined to redo the structural estimation for these two groups separately. Here, I divide the pooled panel data into two groups: currently married and never married people. I then re-estimate the whole model, i.e. the Heckman selection, rst stage, and second stage, for each sub group. The second stage regressions for never-married and currently married people are reported in Tables 2.10 and 2.11 respectively. The Heckman selection model and the rst stage regressions are reported in the Appendix Tables A.2, A.3, A.4, and A.5. Table 2.10: Linear Probability Model of Labor Force Participation on Predicted Log of Real Wages from Regression in Table A.3 for Never-Married People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors), Pooled data 1992-95 Rural Urban Women Men Women Men ln(Wage) 0.015 0.061 0.286** -0.446 y (0.170) (0.050) (0.110) (0.253) Crisis Years -0.038 -0.013 -0.014 0.040 (0.071) (0.048) (0.066) (0.072) Trend 0.018 0.009 0.018 -0.104 y (0.044) (0.022) (0.032) (0.056) I[20<Age30] 0.054 0.322** 0.059 -0.412 (0.119) (0.106) (0.062) (0.271) No. of females b/w 15 to 18 -0.001 0.014 -0.056 -0.033 (0.034) (0.020) (0.035) (0.023) No. of males b/w 15 to 18 0.016 -0.042 y -0.042 0.005 (0.039) (0.025) (0.033) (0.020) No. of females above age 18 0.063** 0.001 0.047* -0.012 (0.023) (0.012) (0.019) (0.014) No. of males above age 18 -0.077** 0.011 -0.020 -0.006 (0.028) (0.014) (0.015) (0.013) Asset Index 0.014 -0.004 -0.063* -0.021 (0.043) (0.023) (0.029) (0.019) Owned home value10 6 0.315 0.077 -0.093 -0.107 Continued on next page 54 Table 2.10 { continued from previous page Rural Urban Women Men Women Men (0.566) (0.286) (0.061) (0.088) Constant 0.061 0.093 -2.522* 6.316* (2.050) (0.479) (1.175) (2.954) Observations 649 1135 1202 2377 Average LFP 0.354 0.867 0.353 0.785 (0.019) (0.010) (0.014) (0.008) Elasticities 0.042 0.070 0.810 -0.568 Note: Dependent variable is a dummy equal to one if the individual is participating in the labor force and zero otherwise. For a description of other covariates, please see Tables 2.5 and 2.7. SECH 1992-95 is the data used. ln(Wage) is the predicted log of real wages from the rst stage regression reported in Table A.3. Education dummies are the instruments for log of real wage in the second stage in Table 2.10. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 Interestingly, the results for currently married and never-married women are quite similar to the results for the whole sample. The coecient of log of wages is insignicant for both genders in rural areas regardless of marital status, but they are positive and large for urban women whether the individual has never-married or is currently married. However, the elasticities are smaller for never married urban women, showing that change in wage rate has a smaller impact on participation for them. That is never-married women act more like men than married women do. The high wage elasticity of FLFP for married women is contradictory to the \marriage lock" theory which argues that marriage limits the responsiveness of women to economic forces. This is a new and interesting result in the literature of FLFP in Iran. In Chapter 3, on the other hand, I will present a dierent result that is compatible with \marriage lock" theory. 55 This coecient for urban men who are never married is negative and large while it is insignicant for married men. However, the absolute value of elasticity for never married urban men is lower than that for never married urban women. The fact that the results are quite comparable to those reported in Table 2.9 is interesting as we might have expected that there would be a dierence between never married and currently married women. Table 2.11: Linear Probability Model of Labor Force Participation on Predicted Log of Real Wages from Regression in Table A.5 for Married People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors), Pooled data 1992-95 Rural Urban Women Men Women Men ln(Wage) -0.055 -0.059 0.235* -0.057 (0.211) (0.070) (0.093) (0.040) Crisis Years 0.113 -0.009 -0.053 -0.006 (0.070) (0.012) (0.049) (0.010) Trend -0.036 0.001 0.055* -0.012 y (0.054) (0.007) (0.027) (0.007) I[20<Age30] 0.047 0.053* -0.083 0.236*** (0.208) (0.021) (0.073) (0.025) I[30<Age40] 0.088 0.074*** -0.094 0.257*** (0.178) (0.019) (0.083) (0.025) I[40<Age50] 0.061 0.053*** -0.106 0.214*** (0.143) (0.014) (0.082) (0.024) No. of females b/w 15 to 18 -0.001 0.015** -0.002 0.019** (0.011) (0.005) (0.010) (0.007) No. of males b/w 15 to 18 -0.006 0.004 -0.011 0.002 (0.011) (0.005) (0.009) (0.008) No. of females above age 18 0.001 0.001 -0.005 -0.002 (0.011) (0.005) (0.007) (0.006) No. of males above age 18 -0.011 -0.010 -0.008 -0.017* (0.008) (0.007) (0.007) (0.007) Asset Index 0.034 y 0.019** -0.013 0.012* Continued on next page 56 Table 2.11 { continued from previous page Rural Urban Women Men Women Men (0.018) (0.007) (0.012) (0.006) Owned home value10 6 -0.207 -0.154 -0.030 -0.021 (0.198) (0.129) (0.029) (0.026) Constant 0.837 1.557* -1.947 y 1.385** (2.542) (0.722) (1.090) (0.452) Observations 5197 5153 7873 8091 Average LFP 0.234 0.971 0.120 0.926 (0.006) (0.002) (0.004) (0.003) Elasticities -0.235 -0.061 1.958 -0.062 Note: Dependent variable is a dummy equal to one if the individual is participating in the labor force and zero otherwise. For a description of other covariates, please see Tables 2.5 and 2.7. SECH 1992-95 is the data used. ln(Wage) is the predicted log of real wages from the rst stage regression reported in Table A.5. Education dummies are the instruments for log of real wage in the second stage in Table 2.11. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 2.4.2 Estimating the Structural Model for Hours After estimating the elasticity of labor force participation with respect to wages, one may think that the natural next step is to estimate the elasticity of hours worked with respect to wages. The econometric framework is similar to what I explained in Section 2.4.1. Here, I regress log of hours worked on predicted log of wages. Predicted log of wages are obtained from the same regressions explained in Section 2.4.1 and depicted in Tables 2.8, A.3, and A.5 for whole sample, never-married, and currently married respectively. Here, I use log of non-zero hours worked as the dependent variable in the second stage. Therefore, I correct for selection for hours worked using a two-step Heckman Selection Model. The selection identication variables for hours worked are education 57 dummies. The rst step of this Heckman selection model for the whole sample, i.e. the Probit estimation of a dummy for whether hours worked is non-zero, is reported in Table A.6. This regression is very similar to the rst step of two-step Heckman selection model on wages. Employing the Mills ratios predicted from this regression, we can correctly estimate the structural model for hours and estimate the elasticity of hours worked with respect to wages. Table 2.12 reports this structural model. The coecient of log of wage represents the elasticity since the dependent variable is in terms of log as well. As shown, this elasticity is insignicant for all groups except urban men. This is consistent with the empirical evidence for the United States, i.e. although female participation is elastic with respect to wages (extensive margin), their hours worked is not (intensive margin). Almost none of the other coecients are signicant for rural and urban women as well as rural men. Table 2.12: Linear Estimation of Log of Hours Worked on Predicted Log of Real Wages from Regression in Table 2.8 for People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors) Controlling for Heckman Selection on Hours (Two-Step Method; First Step in Table A.6), Pooled data 1992-95 Rural Urban Women Men Women Men ln(Wage) 0.049 -0.105 0.656 -0.255*** (0.633) (0.073) (0.672) (0.068) Crisis Years 0.045 0.023 -0.245 -0.004 (0.301) (0.045) (0.202) (0.021) Trend 0.031 -0.048* 0.221 -0.040** (0.149) (0.021) (0.173) (0.013) I[20<Age30] 0.089 0.039 -0.215 -0.155** (0.508) (0.033) (0.234) (0.051) I[30<Age40] 0.011 0.044 -0.297* -0.138*** (0.422) (0.039) (0.147) (0.039) Continued on next page 58 Table 2.12 { continued from previous page Rural Urban Women Men Women Men I[40<Age50] 0.016 0.068* -0.294 y -0.101** (0.334) (0.028) (0.160) (0.034) No. of females b/w 15 to 18 -0.023 -0.012 -0.074 -0.004 (0.074) (0.013) (0.056) (0.012) No. of males b/w 15 to 18 -0.017 -0.007 -0.028 0.010 (0.080) (0.014) (0.082) (0.009) No. of females above age 18 0.077 -0.013 0.118* -0.006 (0.057) (0.020) (0.058) (0.008) No. of males above age 18 0.005 0.027 -0.089 0.021 y (0.122) (0.019) (0.089) (0.011) Asset Index 0.084 0.032 -0.080 0.015 (0.119) (0.034) (0.092) (0.010) Owned home value10 6 -2.022 0.113 -0.227 0.059 (3.710) (0.359) (0.333) (0.049) Constant 2.425 5.060*** -4.019 6.974*** (7.793) (0.759) (7.951) (0.781) Note: Dependent variable is the positive (non-zero) hours worked by an individual in the week preceeding the survey. Heckman two-step selection model is used to correct for selection on Hours. The selection identifying variables for hours worked are education dummies and the rst step Heckman selection model is reported in Table A.6. For adescription of other covariates, please see Tables 2.5 and 2.7. SECH 1992-95 is the data used. ln(Wage) is the predicted log of real wages from the rst stage regression reported in Table 2.8. Education dummies are the instruments for log of real wage in this regression. Bootstrapped standard errors, resampled at clusterlevel and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 Elasticity of hours worked with respect to wages for urban men is negative and signif- icant showing that if wages increase by one percent hours worked will decrease by 0.26%. This result is not surprising as men with lower wages particularly work over-time and usually have multiple jobs to be able to maintain a decent living standard. As wage rates increase there is less need to work more and hours worked decrease. 59 Table 2.13: Linear Estimation of Log of Hours Worked on Predicted Log of Real Wages from Regression in Table A.3 for Never-Married People Aged 21 through 65 with Cluster Random Eects (Bootstrapped Standard Errors) Controlling for Heckman Selection on Hours (Two-Step Method; First Step in Table A.7), Pooled data 1992-95 Rural Urban Women Men Women Men ln(Wage) 0.105 -0.077* -0.108 -0.097 y (0.451) (0.033) (0.139) (0.052) Crisis Years -0.094 0.033 -0.190 y -0.047 (0.257) (0.063) (0.111) (0.035) Trend 0.103 -0.051 y 0.079 -0.014 (0.140) (0.031) (0.049) (0.018) I[20<Age30] 0.128 -0.054 -0.077 -0.045 (0.641) (0.094) (0.103) (0.062) No. of females b/w 15 to 18 -0.214 -0.030 -0.116 0.011 (0.155) (0.025) (0.097) (0.019) No. of males b/w 15 to 18 -0.165 -0.005 0.053 -0.018 (0.191) (0.037) (0.079) (0.014) No. of females above age 18 0.074 0.008 -0.029 -0.007 (0.100) (0.020) (0.034) (0.012) No. of males above age 18 -0.073 -0.004 -0.005 -0.003 (0.131) (0.024) (0.034) (0.010) Asset Index 0.040 0.028 0.026 0.002 (0.141) (0.034) (0.048) (0.015) Owned home value10 6 -0.562 0.724 0.174 0.037 (4.598) (0.533) (0.154) (0.106) Constant 1.713 4.780*** 4.813** 5.175*** (5.803) (0.372) (1.597) (0.613) Note: Dependent variable is the positive (non-zero) hours worked by an individual in the week preceeding the survey. Heckman two-step selection model is used to correct for selection on Hours. The selection identifying variables for hours worked are education dummies and the rst step Heckman selection model is reported in Table A.7. For adescription of other covariates, please see Tables 2.5 and 2.7. SECH 1992-95 is the data used. ln(Wage) is the predicted log of real wages from the rst stage regression reported in Table A.3. Education dummies are the instruments for log of real wage in this regression. Bootstrapped standard errors, resampled at clusterlevel and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 60 Following the same concern explained in Section 2.4.1, one can divide the sample into never-married and currently married people, and redo the same analysis for hours. This analysis is reported in Tables 2.13 and 2.14 for the never-married and currently married respectively. The results in these tables are generally similar to the evidence for the whole sample presented in Table 2.12. In other words, hours worked for women, both in urban and rural areas and across marital status, is inelastic with respect to wages. However, never-married men in both urban and rural areas have dierent elasticities relative to currently married men. Never-married men in rural and urban areas have similar elasticities of about 0.08 to 0.09 percent. On the other hand, hours worked for currently married men in rural areas is not elastic, while for urban married men the elasticity is negative, large and signicant at about -0.36%. Table 2.14: Linear Estimation of Log of Hours Worked on Predicted Log of Real Wages from Regression in Table A.5 for Married People Aged 21 through 65 with Cluster Ran- dom Eects (Bootstrapped Standard Errors) Controlling for Heckman Selection on Hours (Two-Step Method; First Step in Table A.8), Pooled data 1992-95 Rural Urban Women Men Women Men ln(Wage) -0.061 -0.137 0.355 -0.335*** (0.565) (0.159) (0.321) (0.078) Crisis Years -0.026 0.010 -0.094 0.025 (0.264) (0.054) (0.124) (0.025) Trend 0.025 -0.043 y 0.115 -0.053*** (0.176) (0.024) (0.092) (0.014) I[20<Age30] -0.176 -0.045 -0.374* -0.298* (0.562) (0.050) (0.154) (0.133) I[30<Age40] -0.255 -0.018 -0.278 y -0.245 (0.452) (0.093) (0.161) (0.150) I[40<Age50] -0.201 0.024 -0.168 -0.156 (0.376) (0.070) (0.172) (0.120) Continued on next page 61 Table 2.14 { continued from previous page Rural Urban Women Men Women Men No. of females b/w 15 to 18 0.017 -0.026 -0.043 -0.024 (0.050) (0.020) (0.055) (0.018) No. of males b/w 15 to 18 0.042 -0.029 -0.026 0.017 (0.058) (0.019) (0.049) (0.014) No. of females above age 18 0.065 -0.015 0.055 -0.003 (0.053) (0.024) (0.038) (0.011) No. of males above age 18 0.017 0.024 -0.020 0.026 y (0.084) (0.024) (0.052) (0.015) Asset Index 0.033 0.029 -0.020 0.011 (0.145) (0.041) (0.054) (0.014) Owned home value10 6 -1.341 0.180 -0.223 0.096 (3.456) (0.505) (0.218) (0.068) Constant 4.301 5.482*** -0.382 7.959*** (6.735) (1.599) (3.653) (0.843) Note: Dependent variable is the positive (non-zero) hours worked by an individual in the week preceeding the survey. Heckman two-step selection model is used to correct for selection on Hours. The selection identifying variables for hours worked are education dummies and the rst step Heckman selection model is reported in Table A.8. For adescription of other covariates, please see Tables 2.5 and 2.7. SECH 1992-95 is the data used. ln(Wage) is the predicted log of real wages from the rst stage regression reported in Table A.5. Education dummies are the instruments for log of real wage in this regression. Bootstrapped standard errors, resampled at clusterlevel and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 The results for married men in urban areas are more reliable than those in rural, as it is very likely that wages are over-estimated for some men in rural areas. Overall, one can observe that the results are quite consistent with the economic intuition, and the evidence from other countries such as the United States. 62 2.5 Conclusion This chapter provides a simple way to estimate the static model of labor force participa- tion for women and men in Iran in a reduced form as well as a structural model. From the reduced form estimations, we observed that female and male labor force participation follow the standard theory of labor force participation in Iran. Women increase their participation dramatically if their education increases above a threshold (the threshold for Iran is high school and above). They have a concave age prole and their maximum participation rate is at around age 40. Households who have more assets are likely to have fewer women participating in the labor market (this seems not to be true for owning a house). Interestingly, women in households that have more teenage females between ages of 15 and 18 are more likely to participate. In addition, more adult women and fewer adult male are more likely to have more women participating. Estimating the eect of wages on labor force participation using education as instru- ments, I nd that urban women have an upward sloping supply curve, i.e. they are more likely to participate if their wage rates increase. Moreover, their response is quite strong as the elasticity is about 1.7. Urban men, on the other hand, have a backward bending supply curve as their elasticity is about -0.34. These results are quite strong and robust to specication. In addition, they are consistent with the evidence for the United States showing that despite low labor force participation rate in Iran, women's behavior in the labor market is quite similar to women's behavior in a well-developed less-discriminatory country such as the United States. 63 Moreover, the high wage elasticity of FLFP for married women is contradictory to the \marriage lock" theory which argues that marriage limits the responsiveness of women to economic forces. This is a new and interesting result in the literature of FLFP in Iran. In Chapter 3, on the other hand, I will present a dierent result that is compatible with \marriage lock" theory. Even structural analysis of hours worked conrms these ndings. Similar to women in the United States, Iranian women's hours worked is inelastic with respect to wage rates. Interestingly, there is almost no dierence between currently married and never-married women for any of these structural models. This is a bit surprising as one could have expected that never-married women may have been less constrained and the decision to participate may be less complex for them. This chapter provides the rst structural analysis of labor force participation of women (and men) in Iran. I showed that despite low FLFP rates in Iran, women's decision to participate or not has the same reactions to wage rates, the most important economic factor in the labor market, as women's decision in the United States. A surprising result with quite important implications for the economic theory and policy. 64 Chapter 3 Female Labor Force Participation in Iran in a Dynamic Setting When economic forces are strong enough, culture may not dominate behavior. { Gary S. Becker 3.1 Introduction In this Chapter, I would like to argue that despite low rates of female labor force partici- pation and rigid institutions that limit women's market work, some economic forces may aect participation of some women. To show this, I look at Iran's economic crisis of 1994 and 95, and its eect on labor force participation when it produced a large exogenous rise in prices. This economic crisis originated with the Iranian government's inability to repay its large short-term foreign debts. This could have been avoided, but mismanage- ment aggravated the problem and made it into a crisis that doubled in ation from 20% to about 50% in a short time, unprecedented in the last six decades. The large abrupt jump in prices led to a negative growth rate for GDP per capita and private consumption per capita. I discuss the characteristics of this instability in Section 3.2. 65 There are only a handful of studies on the eect of economic crises on labor force participation of women in developing countries. This is likely because historically few good data sets have been gathered during economic crises in the developing world. In addition, disentangling and identifying the eects of any economic crisis is dicult. The Indonesian economic crisis of 1998 provides one of those rare opportunities for research, leading to studies on the eects of economic crisis on education (Thomas et al., 2004), household expenditure (Thomas et al., 2005), wealth and welfare (Frankenberg and Smith, 2003), health and family planning (Frankenberg et al., 2003, 1999), and also labor market outcomes (Smith et al., 2002). Thomas et al. (2003), in particular, studied the labor force participation of women during the Indonesian economic crisis of 1998 and found that unemployment did not change during this period, although real wages fell by about 40%. Interestingly, women's employment rates increased during this crisis in almost all types of employment, especially \unpaid family work." McKenzie (2003) studied the eect of the Mexican Peso crisis as an aggregate shock on household behavior, including labor force participation. Mu (2006) studied the eect of income shocks on labor force participation decision of men and women in Russia. 1 This chapter can stand among the few contributing to this literature. Moreover, studying the eect of economic instability and FLFP in Iran is more benecial, since it may shed light on the long held mysterious hypotheses that labor force participation of women in the Middle East is not responsive to economic forces (the FLFP puzzle). Economic theory predicts that under such shocks to real wages and exogenous income, labor force participation (LFP) of men and women can increase or decrease, depending 1 One may add to this list the literature on the eects of transition to market economy on labor force participation of women. 66 on the preferences of people. Hence, the question of what would happen to LFP is largely an empirical one. Understanding the eect of economic instability of 1994-95 on female LFP in the context of Iran is fruitful, since it sheds light on whether economic forces can aect female LFP in this rigid environment. I explain the economic argument behind this research design in detail in Section 3.3. Using panel data which covers 1992 through 95 and is discussed in Section 3.4, I show that during the economic instability of 1994-95 labor force participation of women partic- ularly in rural areas, increased by about 7.1 percentage points (Section 3.5). Considering that labor force participation in rural areas was about 25% before the instability, a 7 percentage point increase is equal to 28% rise in labor force participation, a large change in Iran's context. For men on the other hand, there was little dierence in LFP rates. Analyzing the results for currently married and never-married people, one can see that married rural women on average were \observed" to participate by 8.4 percentage points (34%) more in the labor market during the crisis, while never-married rural women and all rural men did not change their participation rates. Moreover, never-married urban women were \observed" to join the labor force by 8.9 percentage points (36%) more. Married urban women (like married urban men) did not change their labor force partici- pation. The results in urban areas are compatible with the \marriage lock" theory which argues that in the Middle East, women opt out of the labor market after marriage and they rarely come back to the labor force, unless they become unmarried again. Interest- ingly, the results in the rural areas seem not to t with this theory. This is a new result in the literature of labor force participation in the Middle East. The results are subject to the caveats explained in this chapter; nevertheless, they provide insights into the puzzle 67 of female labor force participation in Iran. I discuss these results and their consequences in more detail in Section 3.5.1. The hours worked at main jobs generally did not change for women. But interestingly, it increased for never-married rural women. Although these women did not increase their participation rates like married rural women, they raised their hours worked. Moreover, the hours decreased for all urban men regardless of marital status. Section 3.5.2 depicts that the fall in hours on average is about 2.7 for working married urban men and 3.5 hours for working never-married urban men. These results are quite similar to what Thomas et al. (2003) found for Indonesia during the Asian foreign exchange crisis. 3.2 Economic Crisis of 1994-95 At the end of the Iran-Iraq war in 1989, the rst ve-year economic plan after the Islamic republic was devised. It was aimed to reconstruct the war-ravaged regions of the country, utilize and extend capacities in agriculture and industries, promote non-oil exports, and implement a liberalization program especially in trade and foreign exchange. The rst half of the plan achieved respectable results such as an annual growth rate in real output of about 8.6 percent, which was largely due to utilization of unused capacity in the economy, in addition to trade and foreign exchange liberalization (Pesaran, 2000). This growth of output was accompanied by large private consumption growth rates which demanded increased imports of nal consumer goods. The imports increased from $13.5 billion in 1989 to about $25 billion in 1991. While non-oil exports increased from 68 $1 billion to 3.5 $billion, oil exports remained relatively stable. Therefore, exports could not meet the large increase in imports and this led to a substantial decrease in external current balance, creating major problem for the government to pay its foreign debts which was around $23.2 billion in the beginning of 1994. This amount of foreign debt should not usually make a signicant problem for an oil-producing economy as large as Iran, but the fact that 76.1% of this debt was in short term, compared to Thailand (65.7%), Malaysia (56.4%), Philippines (58.8%), Korea (67.9%), and Indonesia (59%) at the end of June 1997 in Asian currency crises, transformed this into a debt payment crisis for the government and forced it to stop trade liberalization policies and promote a \closed economy" (Pesaran, 2000). Because of the new strict controls on trade, imports fell by 68% to $13.5 billion, where it was in the beginning of the ve-year plan. The large fall in imports and the uncertainties on the future exchange rate increased rate of price in ation dramatically from about 23% in 1993 to 35.2% in 1994 and then to 49.4% in 1995, rates far from the average in ation rates in the last 6 decades. Only the 60% in ation rate in 1974 could rival with it. These changes are shown in Figure 3.1. Large unexpected rise in in ation led to a fall in aggregate real private consumption in addition to real output. Figure 3.2 depicts these falls. The large rise in in ation in 1994 and 95 particularly hit the households hard and as Figures 3.3 and 3.4 show, real wages fell by about 42% and 35% for women and men from 1992 to 1995, respectively. The fall in GDP per capita was not as large as major economic crises such as the Asian currency crisis, but the average changes in household level economic well-being was large enough to consider this as at least a minor crisis. 69 Figure 3.1: Trends in Import per Capita and In ation (CPI) Data Source: Central Bank of Iran. Figure 3.2: GDP and Private Consumption per Capita Growth Rates During the Crisis Data Source: Central Bank of Iran. 70 Figure 3.3: Log Real Wages For Women Aged 21 through 65 in Rural and Urban Areas, 1992-1995 Note: Wages are calculated by dividing income from the main job by the number of hours worked for the main job during last week. Real wages are computed using rural and urban CPIs (1994 were used as the base year.) Although people who were unpaid family workers reported hours worked for their main job, for obvious reasons have missing values for their income. Therefore, wages are missing for about 63% of working rural women, 20% of working urban women, 12% of working rural men and 2% of working urban men. Figure 3.4: Log Real Wages For Men Aged 21 through 65 in Rural and Urban Areas, 1992-1995 Note: Please see the notes for Figure 3.3. 71 A within estimator of real per capita expenditure, on a sample of household heads only, controlling for household head's xed eects, shows that real per capita expenditure (Real PCE) fell during these crisis years. Table 3.1 reports the results of this exercise and depicts that the fall in rural areas was 8.5% of the standard deviation of rural real PCE in 1992 and in urban areas, it was 10% of the standard deviation of 1992 urban real PCE. Considering that economy was growing at an average rate of about 8% in the three years prior to the instability, this change is considerable although small. Table 3.1: Change in Real per capita Expenditure During the Crisis Years, Controlling for HH Head's Fixed Eects (in 1994 Rials) HH Real per capita Expenditure Rural Urban Crisis Years -6462.9* -14141.4** (3242.0) (6235.9) Standard Deviation of Real PCE in 1992 54996.5 141235.5 Change relative to Standard Deviation -8.5% -10.0% Note: Fixed-eect estimator of real per capita expenditure on Crisis Years dummy (a time dummy equal to one for 1994 and 95 and zero otherwise) using the sample of household heads only and controlling for household heads' xed eects. Robust-heteroskedastic Standard errors are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 3.3 Work and Macro-destabilization Consider householdi as a single decision making entity which maximizes an inter-temporal utility function, U , as T X t=1 1 1 + t U it (L f it ;L m it ;X it ;E it ) (3.1) 72 in which L f it and L m it are the leisure time at time t for the main female and male family members, respectively. X it is the household consumption of a composite good at time t andE it contains all household characteristics that shape the householdi's utility function at time t. The maximization problem is subject to the household's lifetime budget and time constraints as, Y i0 = T X t=1 1 1 +r t t (P t X it W f it H f it W m it H m it ) H f it +L f it = 1 t = 1; ;T H m it +L m it = 1 t = 1; ;T in which, Y i0 is the household's initial wealth (at time 0), H f it , W f it are hours worked and wages by female family member respectively (similarly, H m it , W m it for male family member) andr t is the interest rate at timet. The rst order condition for an individual's hours worked at time t is 1 1 + t @U it @L j it i 1 1 +r t t W j it 0 j =f;m in which, i is the Lagrangian multiplier of the combined budget and time constraints. i is a function of initial wealth as well as past, current, and future wages and prices, and hence allows us to identify current hours worked as a function of current wages and prices. Moreover, since I assume perfect foresight for the household, i is constant over 73 time. The inequality becomes an equality if the individual works. Therefore, her hours worked can be written as, H j it =H j it (W f it ;W m it ; i ;P t ;E it ) (3.2) Note that hours worked for each family member could be a corner solution and end up being zero; something which is quite prevalent for female family members in Iran. Nominal hourly wage rates are functions of predetermined (exogenous) characteristics such as schooling level and age. They can be specied as, E it =E it (V it ; i ;e it ) (3.3) W f it =w f it (Z f it ;C t ;t; f i ;u f it ) (3.4) W m it =w m i (Z m it ;C t ;t; m i ;u m it ) (3.5) r t =r t (C t ;t;v r t ) (3.6) P t =P t (C t ;t;v p t ) (3.7) in which V it , Z f it , and Z m it are the vectors of time-varying observable characteristics of household, female, and male members. They include variables, such as schooling and age. C t is an indicator of whether the economy is in crisis. i is the household xed eect, and f i and m i are female and male individual xed eects. t captures the time trend in wages, prices and interest rate. e it , u f it , u m it , v r t , and v p t are mean zero constant variance serially uncorrelated disturbances which may be contemporaneously correlated. As shown, an economic crisis can aect wages as well as prices and interest rates. 74 Substituting Equations (3.3), (3.4), (3.5), (3.6), and (3.7) into Equation (3.2) and linearizing it, one can nd the reduced form equation as, H j it = +C t + t + 0 V it + 0 f Z f it + 0 m Z m it + j i +w j it j =m;f (3.8) Individual xed eect, j i , captures i , household xed eect, i , as well as other member's xed eects 2 , k i k6=j, since they are all xed for individualj. In this setting, age which is part of Z f it and Z m it acts as a time trend variable, t, and captures any trend in wages, prices, interest rate, and other macroeconomic factors that aect them. A crisis can aect wages, interest rate (returns on assets), and prices each of which will in turn change hours worked (see Equations (3.4), (3.5), (3.6), (3.7), and (3.2)) or LFP. At the time of an economic crisis, when prices rise, and returns on non-labor incomes and wages fall, what happens to the labor force participation is an empirical question, since it depends on the sum of the substitution and income eects from all these changes. In other words, whether individuals, during a crisis, on average, tend to work more to maintain their living standards or set back and consume less cannot be answered by theory and varies from population to population. The coecient of C t , , is the sum of the eect of crisis on hours through wages, interest rate, and prices. In this reduced form specication, each of these eects are not separately identied but we can estimate the total eect. This is what we are interested in as well. Having data for the same individuals over time, one can use least squares estimation with individual xed eects to estimate Equation (3.8) and nd marginal eect of C t on 2 As long as they do not leave the household for reasons such as divorce or death during the survey period. The divorce rate during the survey period, 1992 to 1995, was close to one percent. 75 hours or labor force participation 3 ,. An individual xed eect estimator also allows for dividing the sample based on possibly endogenous but xed characteristics, like region of residence and marital status. I use this property later to estimate the model for urban and rural areas, as well as currently married and never-married people separately. For the reasons that I explained in Section 1.2, marriage has important implications for any analysis of labor force participation in Iran. An individual xed eect estimator also controls for any selection based on time- constant unobservable factors. That is it permits estimating the model for a dependent variable which is only observed on a selected sample, such as hours worked or log of wages, as long as selection is based on time-constant (observable or unobservable) factors. The concern with the reduced form specication is that the crisis dummy is a time dummy and hence is a time eect. In other words, the eect is the sum of the eects of all aggregate forces that happened during 1994-95 period, aected everyone, were dierent from aggregate forces in 1992-93, and could not be captured by a trend variable. But, since I include Age in the regression, it acts like a trend variable and captures most of these aggregate forces which have a trend. Therefore, what remains and is identied by the coecient of the crisis dummy can be categorized as the eect of economic crisis of 1994-95. In this chapter, I estimate a simple form of Equation (3.8) for both hours and labor force participation as dependent variables. This equation can be modied to account for dierential marginal eect of crisis across age and education. In this case, the coecient of C t , , can be written as a linear function of a vector of individual characteristics, Z j it , 3 If the dependent variable is LFP dummy. 76 such as age and education, i.e. the marginal eect of crisis can be dierent for old vs. young and for educated vs. uneducated. Hence, can be written as 0 +Z j it z . As I will explain later, I estimate , 0 , and z for dierent sub-samples. 3.4 Data SECH 1992-95 is chosen to be the primary dataset for the purpose of this chapter for four major reasons. First, unlike HEIS, it is a panel data that provides more exibility in overcoming omitted variables bias. Second, SECH contains hours worked by individuals in addition to income, providing the opportunity to compute wages, while HEIS data prior to 2004 do not. Third, it is the only panel data which include an unexpected economic crisis with large rise in prices, 4 necessary for addressing the main question of this chapter. Fourth, the panel covers a period in which women aged 21 through 65, the sample of interest in this study, did not experience a systematic rise in their educational attainment 5 . As shown in Figure 1.2, the rise in college enrollment for women happened after 1996 and for generations whose oldest were seventeen in 1995, the last wave of the panel. The distribution of educational attainment for this group of women is similar across these four years. Summary statistics of some of the variables of interest are reported in Table 3.2. 4 The other panels, i.e. SECH 87-89 and SECH 2001-03, only cover boom cycles. 5 But women younger than 20 years old in 1995 had a dramatic rise in their tertiary education. 77 Table 3.2: Summary Statistics of the 1992 Values of Some Variables of Interest for Men and Women Aged 21 through 65 Who Were in the Panel for More Than One Year Women No. of Observations Mean St. Dev. Min Max Urban 4527 0.62 0.49 0 1 Age 4527 37.73 12.14 21 65 Education 4028 3.72 4.82 0 16 Illiterate 4527 0.45 0.50 0 1 Read/Write 4527 0.21 0.41 0 1 Primary School 4527 0.35 0.48 0 1 Middle School 4527 0.07 0.26 0 1 High School 4527 0.09 0.29 0 1 University 4527 0.04 0.19 0 1 Married 4527 0.83 0.38 0 1 Widowed 4527 0.07 0.25 0 1 Divorced 4527 0.01 0.10 0 1 Never-Married 4527 0.09 0.29 0 1 Labor Force Participation 4527 0.21 0.40 0 1 Hours Worked per Day 906 5.69 2.64 1 24 Days Worked per Week 906 5.64 1.39 1 7 Hours Worked per Week 906 33.36 18.15 1 112 ln(Real Wage) 494 10.48 1.12 5.91 15.00 Men No. of Observations Mean St. Dev. Min Max Urban 4603 0.62 0.49 0 1 Age 4603 37.99 12.69 21 65 Education 4294 5.80 5.03 0 16 Illiterate 4603 0.23 0.42 0 1 Read/Write 4603 0.20 0.40 0 1 Primary School 4603 0.44 0.50 0 1 Middle School 4603 0.13 0.33 0 1 High School 4603 0.13 0.34 0 1 University 4603 0.07 0.26 0 1 Married 4603 0.81 0.40 0 1 Widowed 4603 0.01 0.10 0 1 Divorced 4603 0.00 0.06 0 1 Never-Married 4603 0.18 0.38 0 1 Labor Force Participation 4603 0.92 0.27 0 1 Hours Worked per Day 4047 8.77 2.52 1 24 Days Worked per Week 4046 6.05 0.86 1 7 Hours Worked per Week 4046 53.59 18.20 2 168 ln(Real Wage) 3790 10.32 1.16 4.13 14.94 Continued on next page 78 Table 3.2 { continued from previous page Note: \Urban" is a dummy variable equal to one if the individual lived in urban areas and zero if in rural areas. \Age" and \Education" are measured in years. \Illiterate", \Read/ Write", \Primary School", \Middle School", \High School", and \University" are dummy variables equal to one if the individuals' attained level of education is the same as the name of variable. \Married", \Widowed", \Divorced", and \Never-Married" are dummy variables equal to one according to marital status of the individual. Hours worked per day, Days worked per week, and Hours worked per week are for main job only. ln(Real Wage) is calculated by dividing income at 1994 price levels by Hours worked per week and taking natural log. One of the main issues with this panel data is that if an individual left the household, her identication number was given to a younger member of the household. This way, the younger member has dierent identication numbers over time and the same identication number was given to dierent members across survey waves. To overcome this problem, I wrote a program that matched individuals ID based on gender, and age over time. Since the surveys are conducted in the same time each year (month of November), age should increase by one year, year by year, for an individual, if it is measured accurately over time. Using this, the program takes an individual in the household in the rst year (1992) and tries to nd an individual with the same gender in the next year which is one year older. If the search was successful, these individuals would get the same identication number. Running this program on a sample of 16 to 70 year olds, I was able to match almost 80% of the individuals over time. Since age and gender are not measured accurately, I could not infer that the other 20% were unmatched individuals who left the household. So I matched those unmatched observations (7848 observations) manually one by one. 6 I tried to identify measurement 6 I manually matched 7848 observations in the sample of 16 to 70 year old individuals. 79 errors by looking at education, household composition over time, and size of the error in age. It is certainly not a perfect match but very close to the actual observations. 7 In addition to SECH, rainfall data for each province was taken from the Iran Meteo- rological Organization website 8 . Since GPS locations of the clusters or their names were not available to me, I obtained rainfall data at province level. For each province and year, monthly rainfall data were obtained separately and then merged and aggregated to get the average annual rainfall. 3.4.1 Caveats of Data Attrition is a common issue with panel data and SECH 92-95 is not an exception. In order to understand attrition, the focus should be on households who were surveyed in the rst wave, omitting those who were added in successive rounds. Among the households who were surveyed in the rst year, share of those who participated in one, two, three, or all four rounds across regions are depicted in Table 3.3. Like most longitudinal surveys, attrition rates are higher in urban areas where geographic mobility is higher. Also, it mainly occurred in the second round of the survey, i.e. 1993, when 14.3% of households left the survey. Those who participated in the rst two waves are likely to stay in the survey. Table 3.4 has a more detailed account of attrition, reporting the pattern of participation. 7 During this process, I realized that there are a few households who completely change in the later rounds and it was unlikely that they would be the same households but were given the same household identication number over time. Interestingly, they were all in the same location and presumably inter- viewed by one interviewer. Supposedly the original household who was given that id number moved out and a new household came in. The interviewer did not compare the new household to the old one and gave the same id number to them. If the data is sorted based on household id number, these households can be identied with id numbers between 11505803 and 11505821. 8 http://www.irimet.net/ 80 Table 3.3: Years in Panel for households who were interviewed in the rst wave (1992) Years Region in Panel Rural Urban Total 1 7.3 18.3 14.3 2 5.4 10.4 8.6 3 7.7 12.9 11.0 4 79.7 58.4 66.1 Total 100.0 100.0 100.0 Note: The Pearson 2 test rejects the null hypothesis that columns labeled urban and rural are the same. The test statistic is 244.28 and the P-value is less than 0.1 percent. Salehi-Isfahani and Majbouri (2008) show that attrition is strongly correlated with whether a household owns or rents its residence or lives in urban or rural areas. Assuming that attrition only happened based on these observable factors, which is a strong assump- tion, one can use inverse response probabilities to control for attrition (Wooldridge, 2002). The results using this attrition correction method, are quite similar to the results reported in this chapter which do not account for attrition. The second issue with the dataset is with the denition of employment. The ques- tion which asks about economic activity, is a multiple choice question with the follow- ing choices: 1) \employed", 2) \unemployed", 3) \having income without working", 4) \student", 5) \homemaker", and 6) \other". Several choices might be relevant for an individual, but only \the most relevant" is chosen in the data. The \most relevant" could vary depending on the circumstances. When it is dicult to choose the \most relevant" activity, the one that takes the most amount of time, is chosen. In other circumstances, 81 Table 3.4: Waves in the Panel for households who were interviewed in the rst wave (1992) Waves Region in Panel Urban Rural Total 1st 7.3 18.3 14.3 1st & 2nd 4.7 9.1 7.5 1st & 3rd 0.4 1.0 0.8 1st & 4th 0.3 0.3 0.3 1st, 2nd, 3rd 5.3 9.2 7.8 1st, 2nd, 4th 1.0 1.1 1.0 1st, 3rd, 4th 1.4 2.6 2.2 All waves 79.7 58.4 66.1 Total 100.0 100.0 100.0 Note: The Pearson 2 test rejects the null hypothesis that columns labeled urban and rural are the same. The test statistic is 247.77 and the P-value is less than 0.1%. the choice is chosen based on its denition. \Employed" is referred to a person \who works for more than 2 days in the week prior to the interview." This denition does not count those who work part-time for less than 2 days per week, which could disproportion- ately be women. In addition, some women, especially in urban areas, who have their own businesses, such as tailoring, hairdressing, and tutoring, usually hide their businesses and work at home in order to escape the burden of getting many business licenses, decrease some xed business costs such as rent, and evade paying taxes. These women may be reluctant to report that they work in the government surveys. 9 Hence, working women may always be under-reported in the surveys. \Unemployed" is checked if someone was 9 Household surveys are collected by Statistical Center of Iran or Central Bank of Iran, both of which are governmental agencies. 82 not employed but was looking for a job in the week preceding the survey. \Student" is the status of someone whose majority of time is spent studying in the school. \Other" is given to any status that does not match with any other choice. Table 3.5: Percentage of People with at Least 2 Jobs Among Working People in 1992 Women Men Rural Urban Rural Urban 11.4 2.2 39.4 9.0 Observations 765 459 2141 2831 Note: Estimated using all observations in year 1992 of SECH 92-95. Another issue with the dataset is that while hours worked is reported for everyone who is considered \working" in this data set, it is only reported for the main job. Table 3.5 reports that in 1992, 11.4% of women in rural areas have at least two jobs, while only 2.2% of urban women have more than one job. On the other hand, 39.4% of working men in rural areas reported to have two jobs or more, while this number is 9% in urban areas. Since hours are only reported for the main job, wages can only be calculated for that job. These may show that one should be cautious in interpreting any results obtained for hours worked and wages by rural men. The last issue for consideration is that income gained, and hence wages, are missing for those who are working as unpaid family workers, that is about 60% of working women in rural and 20% in urban areas. Therefore, one should be cautious in interpreting the results on wages. 83 3.5 The Evidence The economic instability aects prices, nominal wages, and non-labor income. Hence, its marginal eect on LFP is the sum of three forces: marginal eect of prices, marginal eect of own and other's wages, and marginal eect of returns on assets. Theory explains that marginal eects of prices and wages could be positive or negative depending on preferences. Hence, it is not informative about the eect of economic instability on labor force participation. As this question is inherently an empirical one, this Section is of main interest. Evidence is presented in two Sections: the eect of economic instability on LFP (Section 3.5.1) and on hours worked on the main job (Section 3.5.2). Estimations of LFP and hours worked are within estimators with individual xed eects. I report the results in Tables each of which consists of two panels. The top panel is for women and the bottom one for men. Note that within estimator with individual xed eects samples individuals who were in the panel data for at least two years (two waves). 3.5.1 Labor Force Participation and Economic Crisis An individual is considered \employed" if she worked for at least two days during the week prior to the interview. No hour limit is mentioned in the questionnaire. As it is clear from the summary statistics (Table 3.2), there are people who worked for less than two days in their main job, but they account for less than 0.7% of the sample. \Unemployed" in the survey is dened as someone with no job who is looking for it in the week prior to the interview. Labor force participation is dened as a dummy variable equal to one if an 84 individual is \employed" or \unemployed". In all the regressions in this chapter, labor force participation is the dependent variable, and \age" and \education" are measured in years. I dene \Crisis Years" as a dummy equal to one if the observation belongs to 1994 or 95, i.e. the years economy was in crisis, and zero otherwise, i.e. for 1992 and 93. The summary statistics of these variables in the rst wave are in Table 3.2. In all the regressions, standard errors are corrected for correlation inside clusters. Tables 3.6 through 3.8 report results for within estimator with individual xed eect. Table 3.6 depicts regressions for all women and men in the sample respectively. Interest- ingly, coecients of \Crisis Years" for women are almost always positive and signicant. Looking at column 2 of top panel in Table 3.6, labor force participation of women in- creased by about 2.9 percentage points on average, controlling for individual xed eects (signicant at 5% level). This is about 15% rise in women's LFP rates. Although it seems that the eect of economic crisis is not dierent for dierent age groups but women with less education are aected more (columns 3 through 5), that is they increased their participation slightly more (0.5% for each additional years of education). Meanwhile, for men (bottom panel of Table 3.6, column 2), economic crisis on average did not change labor force participation. The economic crisis eect is not dierent across education levels nor age, for men. Comparing Tables 3.7 and 3.8, one realizes that this phenomenon is largely a rural one. Female labor force participation increased by about 7.1 percentage points (28% rise) in rural areas, a large change in the context of Iran, but it did not change in urban areas. Interestingly, the eect is not signicantly dierent across education and age groups, neither in rural (Table 3.7) nor urban areas (Table 3.8). On the other hand, male LFP 85 Table 3.6: Linear Probability Model with Individual Fixed Eects Estimates of Labor Force Participation of People Aged 21 through 65 Women (1) (2) (3) (4) (5) Crisis Years 0.023* 0.029* 0.023 0.031** 0.058 y (0.009) (0.012) (0.030) (0.012) (0.034) Age 0.005 0.008 0.008 (0.010) (0.015) (0.015) Age 2 10 3 -0.109 -0.144 -0.141 (0.104) (0.165) (0.164) Crisis Age10 3 0.173 -0.402 (0.647) (0.679) Crisis Education10 3 -3.349 y -4.916* (1.779) (2.077) Constant 0.188*** 0.165 0.123 0.188*** 0.121 (0.004) (0.262) (0.333) (0.004) (0.334) Observations 17405 17405 17405 17405 17405 Men Crisis Years -0.005 -0.005 0.027 -0.025*** 0.017 (0.004) (0.007) (0.024) (0.006) (0.027) Age 0.049*** 0.036** 0.036** (0.007) (0.011) (0.011) Age 2 10 3 -0.610*** -0.445** -0.445** (0.083) (0.142) (0.142) Crisis Age10 3 -0.831 -0.710 (0.592) (0.623) Crisis Education10 3 4.881*** 1.147 (1.297) (1.363) Constant 0.916*** 0.051 0.254 0.916*** 0.256 (0.002) (0.155) (0.218) (0.002) (0.218) Observations 17260 17260 17260 17260 17260 Note: The sample includes all people Aged 21 through 65 who were at least in two waves of the panel. Labor force participation is a dummy equal to 1 if the individual is looking for a job or working for more than 2 days in the week preceding the survey. \Crisis Years" is a dummy variable equal to one if year is 1994, and 95, and zero otherwise (1992, and 93). \Age" and \Education" are measured in years. Robust-heteroskedastic standard errors, corrected for correlation inside clusters, are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 86 Table 3.7: Linear Probability Model with Individual Fixed Eects Estimates of Labor Force Participation of People Aged 21 through 65 in Rural Areas Women (1) (2) (3) (4) (5) Crisis Years 0.054* 0.071* 0.027 0.049* 0.012 (0.021) (0.027) (0.060) (0.020) (0.060) Age 0.007 0.024 0.024 (0.019) (0.029) (0.029) Age 2 10 3 -0.206 -0.432 -0.432 (0.193) (0.301) (0.301) Crisis Age10 3 1.139 1.414 (1.201) (1.240) Crisis Education10 3 4.305 3.587 (5.413) (5.981) Constant 0.246*** 0.303 0.025 0.246*** 0.028 (0.010) (0.556) (0.692) (0.010) (0.692) Observations 7055 7055 7055 7055 7055 Men Crisis Years 0.005 -0.003 0.017 -0.010 y -0.011 (0.003) (0.009) (0.029) (0.006) (0.036) Age 0.025** 0.017 0.017 (0.008) (0.014) (0.014) Age 2 10 3 -0.262* -0.163 -0.164 (0.103) (0.198) (0.198) Crisis Age10 3 -0.500 -0.071 (0.725) (0.838) Crisis Education10 3 5.485* 3.795 y (2.163) (2.221) Constant 0.950*** 0.444** 0.566* 0.950*** 0.572* (0.002) (0.162) (0.248) (0.002) (0.247) Observations 6825 6825 6825 6825 6825 Note: The sample includes all people Aged 21 through 65 who were at least in two waves of the panel. Labor force participation is a dummy equal to 1 if the individual is looking for a job or working for more than 2 days in the week preceding the survey. \Crisis Years" is a dummy variable equal to one if year is 1994, and 95, and zero otherwise (1992, and 93). \Age" and \Education" are measured in years. Robust-heteroskedastic standard errors, corrected for correlation inside clusters, are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 87 Table 3.8: Linear Probability Model with Individual Fixed Eects Estimates of Labor Force Participation of People Aged 21 through 65 in Urban Areas Women (1) (2) (3) (4) (5) Crisis Years 0.002 0.001 0.026 0.006 0.048 (0.005) (0.008) (0.028) (0.008) (0.031) Age 0.003 -0.006 -0.007 (0.010) (0.014) (0.014) Age 2 10 3 -0.036 0.093 0.095 (0.108) (0.167) (0.167) Crisis Age10 3 -0.644 -0.993 (0.623) (0.650) Crisis Education10 3 -1.288 -2.366 (1.600) (1.769) Constant 0.147*** 0.088 0.247 0.147*** 0.246 (0.002) (0.226) (0.293) (0.002) (0.293) Observations 10350 10350 10350 10350 10350 Men Crisis Years -0.011 y -0.007 0.035 -0.047*** 0.013 (0.005) (0.010) (0.037) (0.011) (0.040) Age 0.068*** 0.052** 0.052** (0.010) (0.016) (0.016) Age 2 10 3 -0.885*** -0.672*** -0.672*** (0.118) (0.194) (0.194) Crisis Age10 3 -1.073 -0.841 (0.887) (0.904) Crisis Education10 3 7.252*** 2.344 (1.886) (1.938) Constant 0.893*** -0.273 -0.009 0.893*** -0.006 (0.003) (0.233) (0.325) (0.002) (0.325) Observations 10435 10435 10435 10435 10435 Note: The sample includes all people Aged 21 through 65 who were at least in two waves of the panel. Labor force participation is a dummy equal to 1 if the individual is looking for a job or working for more than 2 days in the week preceding the survey. \Crisis Years" is a dummy variable equal to one if year is 1994, and 95, and zero otherwise (1992, and 93). \Age" and \Education" are measured in years. Robust-heteroskedastic standard errors, corrected for correlation inside clusters, are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 88 on average did not change in rural nor urban areas. These results depict that the crisis was largely aecting women in rural areas and increasing their participation. One may suspect that this rise in labor force participation is due to more women coming into the labor market but not being able to nd any jobs, and hence it is really a rise in unemployment levels rather than employment. In this chapter, I study the decision of women and men to join the labor force and whether economic instability aects that. Therefore, we are interested in any change in the total number of employment and unemployment. Nevertheless, as Table 3.9 shows, interestingly, unemployment did not change during the years of instability. Unemployment only increased for urban men, by a mere 1 percentage point (signicant at 5% level). There was virtually no change for rural or urban women as well as rural men. Tests strongly reject that unemployment rates changed for any of them. 10 Table 3.9: Estimated Unemployment Rates in Each Year in the Panel (in %) Rural Urban Women Men Women Men 1992 0.6 3.9 4.9 4.8 (0.3) (0.5) (1.0) (0.4) 1993 0.1 4.1 2.1 4.4 (0.1) (0.5) (0.7) (0.4) 1994 0.6 3.5 3.2 5.6 (0.4) (0.5) (0.9) (0.5) 1995 0.7 4.1 3.7 5.7 (0.3) (0.5) (1.0) (0.5) Note: Estimated using all observations in SECH 92-95. Standard errors are in parentheses. 10 P-values are 32%, 91%, and 67% for rural women, urban women and rural men respectively. 89 One may suggest that the reason behind the increase in labor force participation of women is not necessarily the crisis, and other aggregate factors could have aected that, especially that it is a rural phenomenon. For instance, an increase in rainfall might have led to such response in FLFP. Unfortunately, it was not possible to match rainfall data to the clusters in this panel data set, since the location of clusters were condential. However, since the provinces were known, it was easy to match province level rainfall data to this dataset. 11 Using data on rainfall during 1992-95 period for all 25 provinces, Figure 3.5 shows the trend in average precipitation during this period. Not only did not the average rainfall increase during the crisis period, but rather it fell sharply relative to 1992 and 93. Looking at medians, one observes similar patterns. considering the percentage change in rainfall during 1994-95 relative to 1992-93 for each province, one observes that only 5 out of 25 provinces experienced an increase in rainfall, while the rest had a decline in precipitation levels, sometimes up to 50%. Figure 3.6 reports this by depicting a histogram of percentage change in province level rainfall, between the two periods. In Table 3.10, I control for province level rainfall in the labor force participation regressions of rural areas. I nd similar results to what was reported before in Table 3.7 without controlling for rainfall. That is the coecient of \Crisis Years" is still positive and signicant for women but not for men. Interestingly, none of the coecients of Rainfall is signicant for both men and women. Moreover, since rainfall is measured in meters 12 , the reported coecients are quite small and negligible. 11 Note that Iran is approximately 17.2% of the United States and 17.7% of China in terms of land area. Most provinces, even some of the large ones, are close in size to the State of Vermont in the U.S. 12 Each meter is about three feet. 90 Figure 3.5: Average Province Level Rainfall over time Data Source: Iran Meterological Organization Figure 3.6: Histogram of Change in Province Level Rainfall Between 92-93 and 94-95 periods (in%) Data Source: Iran Meterological Organization 91 Table 3.10: Linear Probability Model with Individual Fixed Eects Estimates of Labor Force Participation of People Aged 21 through 65 in Rural Areas, Controlling for Rainfall Women (1) (2) (3) (4) (5) Crisis Years 0.040* 0.067* 0.025 0.037 y 0.016 (0.019) (0.027) (0.060) (0.018) (0.060) Rainfall (in meters) -0.093 -0.103 -0.103 -0.092 -0.102 (0.088) (0.089) (0.089) (0.088) (0.089) Age 0.002 0.018 0.018 (0.019) (0.028) (0.028) Age 2 10 3 -0.210 -0.430 -0.430 (0.187) (0.299) (0.299) Crisis Age10 3 1.104 1.261 (1.210) (1.243) Crisis Education10 3 3.298 2.039 (5.262) (5.880) Constant 0.298*** 0.561 0.291 0.297*** 0.290 (0.046) (0.559) (0.699) (0.046) (0.699) Observations 7055 7055 7055 7055 7055 Men Crisis Years 0.006 -0.003 0.017 -0.009 -0.011 (0.004) (0.009) (0.029) (0.006) (0.036) Rainfall (in meters) 0.008 0.010 0.010 0.011 0.012 (0.014) (0.013) (0.013) (0.014) (0.014) Age 0.025** 0.018 0.018 (0.008) (0.014) (0.014) Age 2 10 3 -0.261* -0.163 -0.163 (0.103) (0.198) (0.198) Crisis Age10 3 -0.496 -0.052 (0.726) (0.840) Crisis Education10 3 5.543* 3.918 y (2.164) (2.211) Constant 0.946*** 0.420* 0.541* 0.944*** 0.541* (0.008) (0.167) (0.255) (0.008) (0.254) Observations 6825 6825 6825 6825 6825 Continued on next page 92 Table 3.10 { continued from previous page Note: The sample includes all people Aged 21 through 65 who were at least in two waves of the panel. Labor force participation is a dummy equal to 1 if the individual is looking for a job or working for more than 2 days in the week preceding the survey. \Crisis Years" is a dummy variable equal to one if year is 1994, and 95, and zero otherwise, i.e. 1992, and 93. \Rainfall" is measured yearly and at province level. Most provinces, even some of the large ones, are close in size to the State of Vermont in the U.S. \Age" and \Education" are measured in years. Robust-heteroskedastic standard errors, corrected for correlation inside clusters, are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 As depicted in Figures 1.7 and 1.8, there is a dierence in LFP rates of currently married and never-married women, especially in urban areas. Although the reasons be- hind this dierence are not well-known, it would be necessary to see how the response diers across married and never-married women. There are two major ways to categorize married and never-married women in the panel data: 1) following those who were married or never-married in all four years of the panel, and 2) following those who were married or never-married in the rst wave, i.e. 1992. Similar results were found by using any of these two samples. I report the results, using the rst group, in Table 3.11. Column 1 shows that married women did increase their LFP rates during the crisis years, while married men on average decreased their participation rates. Looking at the within area dierences in columns 2 and 3 of Table 3.11, one observes that similar to the results for all women (Tables 3.7 and 3.8) the rise in married women LFP was also a rural phenomenon. It is not surprising that married women responded the same as all women, since they account for the majority of women, but it is interesting that even married women's LFP rates increased during the instability time by 8.4 percentage points (more than 30% rise). On the other hand, bottom panel of column 3 shows that married urban men decreased their participation rates slightly, while married urban 93 Table 3.11: Coecient of Instability () in the Linear Probability Model with Individual Fixed Eect Estimates of Labor Force Participation for All People Aged 21 through 65 Women Married Never-married All Rural Urban All Rural Urban Crisis Years 0.030* 0.084** -0.008 0.055 0.001 0.089* (0.013) (0.029) (0.008) (0.036) (0.061) (0.043) Observations 14143 5828 8315 2004 785 2031 Men Crisis Years -0.011 y -0.008 -0.012 0.015 0.019 0.013 (0.005) (0.006) (0.008) (0.022) (0.030) (0.030) Observations 13247 5406 7841 3753 1319 2434 Note: The sample includes all people Aged 21 through 65 who were at least in two waves of the panel. Labor force participation is a dummy equal to 1 if the individual is looking for a job or working for more than 2 days in the week preceding the survey. \Crisis Years" is a dummy variable equal to one if year is 1994, and 95, and zero otherwise (1992, and 93). \Age" and \Education" are measured in years. Robust-heteroskedastic standard errors, corrected for correlation inside clusters, are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 women did not experience a signicant change in their LFP rates. This is also in accord with the results for the average population. Interestingly, looking at all never-married women and men in column 4 of Table 3.11, one can see that contrary to the results for married people, never-married women and men did not respond during the crisis. Within area estimates in columns 5 and 6 show that never-married women in urban areas increased their participation rates by about 8.9 percentage points (about 36%) during the crisis years. This is the largest change in participation rates among all groups studied here. In summary married women in rural areas as well as never-married urban women increased their participation levels. These results match the conclusions from Figures 1.7 94 and 1.8 explained in Section 1.2. They are consistent with the hypothesis that considers marriage in the Middle East as an institution which locks women out of the labor market, although they are not the evidence for it. This eect of marriage may or may not be due to discrimination. For instance, women after marriage may have a stronger preference for home production, like raising children, rather than market work. In the case of discrimination, a husband may not \allow" his wife to work outside the home, for various reasons such as: 1) wife's work outside home is a breach in traditional norms, 2) husband wants to limit social interactions of his wife with other men, and 3) husband loses part of his authority and bargaining power at home since his wife has a source of income of her own. The results for urban areas are compatible with this theory as they show that never-married women are more responsive to the instability while they already had higher participation rates. Looking at the results for rural areas, one nds the opposite pattern, that is the FLFP rate increased for married women while it did not change for never- married ones. This may suggest that the \marriage lock" theory may not be relevant for rural areas, and only urban women may be constrained not to work after marriage. Figures 1.7 and 1.8 are also in harmony with this suggestion. Since the dierence in LFP between married and never-married women in urban areas is almost twice that of rural areas (15 to 20 percents for urban vs. 7-10 percents in rural), it may provide further (suggestive) evidence that, compared to rural married women, urban married women are more likely to behave dierently relative to never-married ones. Therefore, it is likely that marriage has more impact on female labor force participation decisions in urban areas than it does in rural areas. 95 But if one looks closer into the types of jobs women do in rural areas, as depicted in Table 1.4, she nds that the majority of employed women (63%) work as unpaid family worker there. Since they work for the family, they essentially have to work on the farm or in the family business and it would not be a breach of social norms. Moreover, social interactions of female family members and other men are controlled when the wife works in the family business under her husband's supervision. In addition, there is no wage being paid, therefore women may not necessarily gain bargaining power by working as unpaid workers for the family. Hence, husbands (and other male relatives) are less concerned when women work as unpaid family workers, which is the majority of labor force participation in rural areas. In addition, there is more exibility in terms of hours and load of work when women work in the family business, therefore they can manage the housework better and more eciently. This advantage gives them more motivation to work in rural areas. In urban areas, on the other hand, since the possibility of women working as unpaid family labor is limited, fewer women are \allowed" to work or would \like" to work (according to the \marriage lock" theory), and hence labor force participation is lower, the dierence in labor force participation of married and never-married women is higher, and married women do not seem to change their participation rates over time. Thomas et al. (2003) found quite similar results for Indonesian crisis, that is women increased their participation and they predominantly contributed to the family by working as unpaid family labor. It is also interesting to see what types of employment were chosen by those who entered the labor market during the crisis in rural areas. Table 3.12 compares the distribution 96 Table 3.12: Distribution of Types of Employment for Rural Women who Entered the Labor Market during Instability Relative to Rural Women in 1992 (in %) Rural Women New Entrants 1992 in 1994-95 Dierence Employer 2.8 2.9 0.1 (0.7) (1.1) (1.3) Self employed 21.0 25.0 4.0 (1.8) (2.8) (3.3) Employee in public sector 3.6 1.7 -1.9 (0.8) (0.8) (1.3) Employee in private sector 10.0 11.0 1.0 (1.3) (2.0) (2.4) Unpaid family worker 62.6 60.0 -2.6 (2.2) (3.2) (3.8) Total 100 100 Observations 501 240 Note: Standard errors are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 of new jobs with the distribution of jobs in 1992, for rural women. As depicted, like 1992, the majority of women who newly joined the labor force during the crisis chose unpaid family work. There is also almost no dierence between these two distributions. Moreover, I nd that there is no dierence in the types of industries chosen by the new entrants compared to those in 1992. These imply that there was no structural change in the rural labor market during this crisis. 3.5.2 Hours and Economic Crisis Observing these changes in participation rates, one is naturally interested to see how hours worked responded during the crisis. Tables 3.13 through 3.15 report the least square with 97 individual xed eects estimation of hours worked per week in the main job conditional on being \employed", that is working for at least two days per week (no minimum hour is mentioned). Therefore, all the results here are subject to considering these caveats, 1) looking at the hours worked in the main job not all hours worked, and 2) for only those who worked more than 2 days a week. Nevertheless, the evidence is interesting. Tables 3.13, 3.14, and 3.15 show that, interestingly, hours worked on the main job by women did not change during the crisis, neither in rural nor urban areas. Looking at column 2 in the bottom panel of Table 3.13, one can see that for all men this fall was about 1.5 hours on average (signicant at 10% level). However, for men in rural areas, there was no signicant change in hours worked (column 2 of the bottom panel in Table 3.14). In urban areas, the fall was about 2.9 hours (Table 3.15). The eect does not vary by age or education. It is hard to interpret the dierence between men and women, but a few explanations exist. One may attribute it to the dierent types of jobs done by women. While most urban women's main job had a xed-hour schedule in public sector, like teachers, nurses, and factory workers, men worked in a variety of jobs with a variety of hours worked and some of them might have more freedom to change their hours worked. But this is not a good explanation for rural areas. Indeed, it is still a question of interest requiring further research. Looking at married and never-married people separately in Table 3.16, one generally nds that hours worked on the main job increased for married and never-married men in urban areas. These changes within each group were not dierent across education or age levels. Therefore, only , the coecient of crisis years, is reported for each group in this table. Never-married men in rural areas had no change in hours similar to that 98 Table 3.13: Least Squares with Individual Fixed Eect Estimates of Hours Worked per Week in the Main Job for All People Aged 21 through 65 Women (1) (2) (3) (4) (5) Crisis Years -0.514 0.090 8.653 y -1.600 7.347 (0.732) (1.526) (4.588) (1.109) (5.277) Age 1.681 -1.467 -1.450 (1.242) (2.010) (2.012) Age 2 -27.166 y 18.582 18.409 (15.601) (27.897) (27.945) Crisis Age -0.240* -0.220 y (0.117) (0.126) Crisis Education 0.297 y 0.148 (0.169) (0.202) Constant 34.646*** 12.235 60.899 34.700*** 60.564 (0.374) (28.618) (37.306) (0.387) (37.294) Observations 3391 3391 3391 3391 3391 Men Crisis Years -3.175*** -1.539 y -0.918 -3.690*** -2.252 (0.488) (0.849) (1.830) (0.736) (2.256) Age -1.167 -1.406 -1.412 (0.742) (0.900) (0.899) Age 2 4.079 7.251 7.263 (7.291) (11.279) (11.271) Crisis Age -0.016 0.001 (0.046) (0.050) Crisis Education 0.130 0.166 (0.105) (0.115) Constant 53.765*** 91.306*** 95.343*** 53.758*** 95.535*** (0.228) (19.717) (20.119) (0.225) (20.107) Observations 15039 15039 15039 15039 15039 Note: The sample includes all people Aged 21 through 65 who were at least in two waves of the panel. Dependent variable is positive (non-zero) hours worked in the week preceding the survey. For the description of other covariates, please see Table 3.6. Robust-heteroskedastic standard errors, corrected for correlation inside clusters, are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 99 Table 3.14: Least Squares with Individual Fixed Eect Estimates of Hours Worked per Week in the Main Job for People Aged 21 through 65 in Rural Areas Women (1) (2) (3) (4) (5) Crisis Years -0.894 1.600 12.383 y -1.196 13.859 y (1.151) (2.584) (6.863) (1.352) (7.523) Age 0.164 -3.729 -3.770 (1.397) (2.745) (2.764) Age 2 -18.401 37.131 37.768 (13.345) (35.859) (36.114) Crisis Age -0.297 y -0.326 y (0.165) (0.176) Crisis Education 0.201 -0.259 (0.272) (0.351) Constant 34.045*** 53.982 114.368* 34.033*** 114.942* (0.625) (41.857) (55.595) (0.618) (55.892) Observations 1913 1913 1913 1913 1913 Men Crisis Years -4.044*** 0.407 2.169 -3.960*** 2.112 (1.019) (1.616) (2.833) (1.066) (3.521) Age -2.970* -3.647* -3.648* (1.300) (1.409) (1.405) Age 2 8.672 17.589 17.591 (12.648) (18.409) (18.387) Crisis Age -0.045 -0.044 (0.072) (0.083) Crisis Education -0.033 0.008 (0.212) (0.206) Constant 53.799*** 152.230*** 163.537*** 53.802*** 163.550*** (0.495) (35.735) (33.070) (0.499) (33.008) Observations 6244 6244 6244 6244 6244 Note: The sample includes all people Aged 21 through 65 who were at least in two waves of the panel. Dependent variable is positive (non-zero) hours worked in the week preceding the survey. For the description of other covariates, please see Table 3.6. Robust-heteroskedastic standard errors, corrected for correlation inside clusters, are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 100 Table 3.15: Least Squares with Individual Fixed Eect Estimates of Hours Worked per Week in the Main Job for People Aged 21 through 65 in Urban Areas Women (1) (2) (3) (4) (5) Crisis Years -0.070 -1.642 4.265 -2.909 y 0.505 (0.866) (1.510) (5.433) (1.472) (5.717) Age 3.605 1.344 1.396 (2.748) (3.257) (3.241) Age 2 -39.563 -6.005 -7.277 (39.645) (47.447) (47.056) Crisis Age -0.169 -0.120 (0.146) (0.146) Crisis Education 0.459* 0.334 (0.196) (0.222) Constant 35.482*** -37.538 -2.548 35.558*** -2.650 (0.407) (47.357) (54.421) (0.407) (54.346) Observations 1478 1478 1478 1478 1478 Men Crisis Years -2.540*** -2.907** -3.187 -2.985*** -4.107 (0.386) (0.872) (2.341) (0.706) (2.621) Age 0.149 0.257 0.253 (0.784) (1.076) (1.074) Age 2 0.528 -0.911 -0.920 (7.714) (13.057) (13.050) Crisis Age 0.007 0.017 (0.058) (0.059) Crisis Education 0.089 0.106 (0.118) (0.118) Constant 53.754*** 47.372* 45.534 y 53.749*** 45.687* (0.175) (20.058) (23.033) (0.175) (22.992) Observations 8795 8795 8795 8795 8795 Note: The sample includes all people Aged 21 through 65 who were at least in two waves of the panel. Dependent variable is positive (non-zero) hours worked in the week preceding the survey. For the description of other covariates, please see Table 3.6. Robust-heteroskedastic standard errors, corrected for correlation inside clusters, are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 101 Table 3.16: Coecient of Instability () in the regression of Hours Worked in the Main Job for All People Aged 21 through 65 Women Married Never-married All Rural Urban All Rural Urban Crisis Years -0.170 -0.200 -0.337 2.702 10.452 z -3.295 (1.668) (2.819) (1.517) (3.893) (6.840) (3.803) Observations 2465 1469 996 676 298 378 Men Crisis Years -1.414 0.380 -2.699** -1.615 1.302 -3.518* (0.900) (1.736) (0.895) (1.476) (2.808) (1.706) Observations 12211 5144 7067 2643 1020 1623 Note: The sample includes all people Aged 21 through 65 who were at least in two waves of the panel. Dependent variable is positive (non-zero) hours worked in the week preceding the survey. For the description of other covariates, please see Table 3.6. Robust-heteroskedastic standard errors, corrected for correlation inside clusters, are in the parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z P-value for this coecient is 0.13. of rural married men. But both never-married and married urban men decreased their hours worked. The reduction was larger for never-married urban men (-3.5 hours vs. -2.7 hours). Interestingly, never-married women in rural areas have increased their hours worked by the large amount of 10.5 hours, while married rural women did not. The coecient is signicant at 13% but note that the magnitude of the coecient is large and the number of observations is small. This is quite interesting as it shows that although never-married rural women did not change their participation rates (see Section 3.5.1), they increased their hours worked. In fact, they are the only group who increased their hours worked. 102 3.6 Conclusion This chapter is an attempt to understand aspects of the FLFP puzzle and has two con- tributions to this literature. Firstly, contrary to what has been perceived but compatible with stylized facts explained in Section 1.2, FLFP was (and probably is) responsive to economic forces at least for some women. Secondly, I nd some evidence compatible with \marriage lock" theory which says women lock out of the labor market after marriage. Never-married urban women \observed" to join the labor force by 8.9 percentage points (36%) more. Meanwhile, married urban women like married urban men did not change their labor force participation. The results in urban areas are quite compatible with the \marriage lock" theory that says in the Middle East women opt out of the labor market after their marriage and they rarely come back to the labor force, unless they become unmarried. On the other hand, one can see that married rural women on average \observed" to participate by 8.4 percentage points (30%) more in the labor market during the crisis. The change is similar to never-married urban women. This is while never-married rural women and all men did not change their participation rates. In summary, the dierence between married and never-married women in urban areas is the opposite of those in rural areas. The results in urban areas t with the \marriage lock" theory. As discussed in Section 3.5.1, although the evidence in rural areas is the opposite, it does not necessarily reject this theory. Since in urban areas unpaid family work is less possible, the majority of married women cannot work and do not change their decision based on small economic forces. Hence, even the results in rural areas may be compatible with the \marriage lock" theory. 103 Moreover, generally, hours worked did not change for women except for never-married ones in rural areas, who worked for 10.5 hours more. Urban men, regardless of marital status, decreased their hours worked as well, while rural men were not responsive during the crisis. These results show that the crisis may have aected rural women, both married and never-married, but in dierent ways, as well as never-married urban women. Married rural women increased their participation while never-married rural ones increased their hours worked. Similar to married rural women, the crisis increased participation for urban never-married women. These new results are subject to the caveats explained in this chapter; nevertheless, it provides insights to the puzzle of female labor force participation in Iran. 104 Chapter 4 Conclusion Women's employment is an important development issue as not only does it lead to eco- nomic independence and a greater say in family decision-making, but it also increases competition and productivity of labor markets and the economy as a whole, while trans- forming the traditional discriminatory norms and institutions against women, such as the preference for having sons. This study is the continuation of eorts by many scholars during the past decade on understanding and explaining the puzzle of female labor force participation in the Middle East. In the rst chapter, I explained the stylized facts about women's achievements dur- ing the last three decades. Surprisingly, these remarkable achievements did not translate to more participation of women in the markets (the FLFP puzzle). I made three main contributions to this literature. Firstly, in Chapter 2, I applied a canonical economic labor force participation model to data on Iranian women and estimated a reduced form as well a structural model. Interestingly, the results are quite compatible with the standard predictions of this economic model. For instance, women with higher education are more likely to work; age prole of participation is concave and 105 has its peak between the ages of 30 to 50; more assets leads to less participation; presence of more adult men decreases the likelihood of participation while more adult women increases it. Moreover, similar to the results for other countries such as the United States, Iranian women's participation is quite elastic with respect to wages in urban areas and the elasticity is positive, while their hours worked is not elastic at all. These interesting results conrm that despite low labor force participation, Iranian women's decision to participate has features that have been predicted by economic theory and were found for other countries. In Chapter 3, I showed that FLFP was (and probably is) responsive to economic forces at least for some women and it is not as rigid as was perceived. Secondly, contrary to the results from the structural estimations in Chapter 2, I nd some evidence compatible with the \marriage lock" theory which says that for many reasons, women lock out of the labor market after marriage. In order to argue these, I looked at the economic crisis of 1994-95 and its eect on FLFP. The evidence shows that rural married women increased their participation during the crisis and worked predominantly as unpaid family labor on family business. Moreover, rural never-married women increased their hours worked during the same period, although they did not change their participation. Interestingly, in urban areas, the only change during crisis was that never-married women increased their participation. The result in urban areas is compatible with \marriage lock" theory, while there are explanations for why the evidence in rural areas does not reject it. Surprisingly, the high wage elasticity for married women's LFP in urban areas is not consistent with the \marriage lock" theory which says that for many reasons, women lock out of the labor market after marriage. 106 Similar to regressions in Chapter 3, structural estimations in Chapter 2 contain \Crisis Years" dummy as an independent variable. Therefore, one may be interested in comparing the estimates of the coecients of this variable in these two chapters. For some of the regressions in Chapter 2, such as the rst step of Heckman selection model on wages (Tables 2.7, A.2, and A.4), and on hours (Tables A.6, A.7, and A.8), the coecient of \crisis Years" has similar sign and signicance level to the results depicted in Chapter 3. This is not surprising as wages and hours are observed for a subset of people who participate in the labor force. But controlling for log of wages in the structural model of labor force participation and hours, the coecients of \Crisis Years" for rural women are not signicant (see Tables 2.10, 2.11, 2.12, 2.13, and 2.14). 1 This is interesting as it shows that, controlling for wages, crisis may not aect participation. Interesting dynamics are happening for women in the Middle East in general and in Iran, in particular, while we know very little about their causes and consequences. Despite these dynamics, the puzzle of FLFP is still a major question in economic literature of the region, which make future research necessary, valuable, and exciting. 4.1 Future Research With many unanswered questions, tackling the FLFP puzzle is a promising line of research and one to which I would like to dedicate the next few years. In Section 1.3 of Chapter 1, I organized and discussed possible hypotheses in three major categories: 1. Those attributing the problem to the shortcomings of measuring FLFP, 1 This is not true for the structural estimation of LFP for all rural women, i.e. Table 2.9, in which the coecient is positive and signicant, similar to the results in Chapter 3. 107 2. Those assigning the problem to the supply side of the labor market, and 3. Those that approach the question from the demand side. Not surprisingly, a combination of these hypotheses may nally solve the puzzle. Discrimination, both on the supply and demand side, is a very important and relevant hypothesis among these explanations. It is also one of the dicult hypotheses to identify empirically. Extensive rm and household surveys, and/or creative research designs may be required to identify these discrimination eects. Perhaps we may only need to wait more in order that exogenous factors, such as changes in policy or law, happen and help nding evidence for discrimination. Two of the major explanations for the FLFP puzzle on the supply side, that are related to discrimination, are the \marriage lock" theory and the \subsidy" hypothesis. \Marriage lock" theory explains that for many reasons women are locked out of the labor market after their marriage. The \subsidy" hypothesis argues that the Iranian government, like some other governments in the Middle East, provides large subsidies on many consumer goods such as energy (the second largest in the world), education (close to 100% subsidized), healthcare (85% subsidized), and basic food (over 70% subsidized). These subsidies may have a large income eect that decreases labor force participation of household members while providing support for traditional (discriminatory) institutions to exist. In other words, household heads, who are predominantly male, can aord a decent living and do not need income gained by female household members and hence, can continue to discriminate. This is the main hypothesis which I would like to investigate in the near future. 108 Figure 4.1: Comparison of per capita energy subsidies and FLFP rates for women aged 25 to 55 Source: Energy subsidies per capita from The Economist (2009), and FLFP rates from ILO statistics on http://www.ilo.org/global/What_we_do/Statistics/lang--en/index.htm (accessed Jan 21, 2010) The \subsidy" hypothesis may not be limited to Iran. Figure 4.1 compares the amount of subsidies on fossil fuels as well as FLFP in a few Middle Eastern and other developing countries. As shown, there is some negative correlation between the amount of subsidies and the rate of female labor force participation. This hypothesis may explain part of the puzzle. In the short run, in addition to \marriage lock" theory, my main focus would be specically on the \subsidy" hypothesis. One approach is to analyze the eect of gov- ernmental subsidies on female labor force participation across countries, controlling for country xed eects. 109 Having a rich cross-country dataset and using dynamic panel data methods to control for all country-level unobservable time-constant characteristics, I hope to identify the eect of subsidies on female labor force participation. Recently, the Iranian parliament passed a law which eliminates all subsidies, but gives the President full control (without oversight) of spending the savings from this subsidy- elimination policy. A recent study by a think tank that advises Irans parliament predicts that such a policy will quadruple gasoline prices. Subsequent eects of increase in gasoline prices may include increases in the costs of basic goods and services (Worth, 2009). In a second project, I would like to exploit this policy change as an exogenous factor and compare the labor force participation of women before and after the policy imple- mentation, using program evaluation methods. Currently, no one knows how the surplus from eliminating subsidies will be spent, but there is talk in policy circles that some of the savings may be spent as cash transfers to low income households. The government needs to nalize the implementation plan before spring 2010. Choosing the right identication methods, to some extent, depends on how the savings from the subsidy elimination are (or are not) going to be distributed among poor households. If there is no cash transfer, rst dierencing over time or dierence-in-dierence across various socio-economic groups and over time are suitable methods. Depending on the type of transfer, propensity score matching with rst dierencing, or with dierence-in-dierence, may help to identify the eects of the new policy. Another method is to compare those households that are on the margin, i.e. those who are close to the threshold and on either side of it, over time. Depending on the amount of money transferred, households that are close to the threshold may be better or (most 110 likely) worse o. In the case that they are worse o, the magnitude of the estimates will be lower-bound for the eect of subsidies on FLFP. Regression discontinuity is the applicable method, although careful consideration should be invested on its implementation. Women in the Middle East and Iran are experiencing dynamic changes, while re- searchers know very little about their causes and consequences. This research project is thus necessary, valuable and most intriguing. 111 Appendix A Structural Model Estimations Table A.1: Weights of Various Assets Used to Construct an Asset Index for SECH 1992-95 Panel Data following Sahn and Stifel (2000) Weights Car 0.2248 Motorcycle 0.1111 Bicycle 0.1435 TV -0.0477 Radio -0.3231 Cassette Player 0.3067 Refrigerator -0.0349 Freezer 0.3000 Gas Stove -0.2799 Washing Machine 0.4281 Vacuum Cleaner 0.4204 Telephone 0.2629 Piped Water -0.3694 Gas Pipe 0.2626 Air conditioner 0.3292 Central Heating 0.3205 Bath 0.4038 Note: Each of the assets is a dummy variable equal to one if the households owns it and zero otherwise. 112 Table A.2: First Step of the Two-Step Estimation of a Heckman Selection Model on Observed Wages across Never-Married People Aged 21 through 65, Pooled Data 1992-95 Rural Urban Women Men Women Men Primary 0.197 0.442** 0.912* 0.650* (0.283) (0.164) (0.360) (0.269) Mid School 0.433 0.298 0.848** 0.531 y (0.418) (0.185) (0.298) (0.281) High School 0.601 0.262 1.586*** 0.267 (0.376) (0.181) (0.275) (0.277) College & higher -0.035 1.755*** -0.270 (0.248) (0.324) (0.286) Crisis Years 0.056 -0.273* 0.090 -0.045 (0.197) (0.134) (0.159) (0.099) Trend -0.038 0.136 y -0.051 0.040 (0.107) (0.070) (0.073) (0.051) I[20<Age30] 0.427 0.688** -0.768*** -0.316* (0.435) (0.226) (0.166) (0.131) No. of females b/w 15 to 18 -0.035 -0.023 -0.151 -0.079 (0.130) (0.078) (0.113) (0.062) No. of males b/w 15 to 18 0.186 0.004 -0.327* 0.073 (0.147) (0.079) (0.140) (0.058) No. of females above age 18 0.155 y -0.061 0.025 -0.044 (0.093) (0.066) (0.071) (0.042) No. of males above age 18 -0.077 -0.019 -0.091 -0.003 (0.105) (0.063) (0.057) (0.049) Asset Index 0.093 -0.042 -0.086 0.104* (0.185) (0.080) (0.095) (0.048) Owned home value10 6 -0.324 -0.697 -0.163 -0.345 (3.778) (1.374) (0.276) (0.262) Constant -1.947*** -0.971** -1.066*** 0.027 (0.495) (0.318) (0.316) (0.297) Observations 649 1135 1202 2377 2 test for identifying variables z 9.04 2.24 9.55 11.31 y Continued on next page 113 Table A.2 { continued from previous page Note: Dependent variable is a dummy equal to one if the individual's wage is observed in the data and zero otherwise. \Crisis Years" is a time dummy equal to one for 1994 and 95, and zero otherwise. For a description of other covariates, please see Table 2.5. SECH 1992-95 is the data used. As there were few women in rural areas with \college & higher" level of education, they are combined with rural women with high school education. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z This 2 statistic tests whether the coecients of selection identifying variables, i.e. No. of females and males b/w 15 to 18, No. of females and males above age 18, Asset Index, and owned home value, are jointly signicant. 114 Table A.3: Estimation of Log of Real Wages for Never-Married People Aged 21 through 65 Controlling for Heckman Selection on Wages (Two-Step Method; First Step in Table A.2), Pooled data 1992-95 Rural Urban Women Men Women Men Primary 0.161 1.009* -0.128 -0.430 (0.499) (0.471) (0.558) (0.639) Mid School -0.570 0.680 0.205 -0.566 (0.630) (0.415) (0.556) (0.606) High School -0.173 -0.491 0.626 -0.757 y (0.702) (0.449) (0.712) (0.453) College & higher 0.756 y 1.222 0.025 (0.455) (0.789) (0.419) Crisis Years 0.053 -0.016 0.231 0.079 (0.353) (0.513) (0.240) (0.211) Trend -0.112 0.102 -0.206* -0.232* (0.179) (0.230) (0.094) (0.116) I[20<Age30] -0.583 0.439 -0.657* -0.898** (0.517) (0.832) (0.324) (0.285) Constant 13.438*** 5.954** 10.059*** 12.703*** (1.772) (2.279) (1.015) (1.168) 2 test for education z 1.74 28.23*** 29.12*** 12.46** Note: Dependent variable is the individual's log of real wage. For a description of covariates, please see Tables A.2 and 2.5. SECH 1992-95 is the data used. Education dummies are the instruments for log of real wage in the second stage in Table 2.10. As there were few women in rural areas with \college & higher" level of education, they are combined with rural women with high school education. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z This 2 statistic tests whether education dummies which are the instruments for log of wage in the regression in Table 2.9, are jointly signicant. 115 Table A.4: First Step of the Two-Step Heckman Selection Model on Observed Wages across Married People Aged 21 through 65, Pooled Data 1992-95 Rural Urban Women Men Women Men Primary 0.277 y -0.013 0.270** 0.076 (0.147) (0.089) (0.104) (0.084) Mid School -0.553** -0.149 0.329* 0.074 (0.204) (0.131) (0.132) (0.114) High School 0.709** -0.172 0.842*** 0.038 (0.229) (0.132) (0.144) (0.106) College & higher -0.177 2.183*** -0.174 (0.322) (0.147) (0.119) Crisis Years 0.286* 0.208 y -0.028 -0.033 (0.125) (0.118) (0.092) (0.072) Trend -0.049 -0.076 -0.007 -0.039 (0.054) (0.052) (0.041) (0.034) I[20<Age30] 0.397* -0.379* 0.047 0.573*** (0.195) (0.155) (0.171) (0.102) I[30<Age40] 0.492** -0.003 0.289 y 0.966*** (0.182) (0.155) (0.158) (0.102) I[40<Age50] 0.341* 0.198 0.282 0.695*** (0.174) (0.123) (0.175) (0.080) No. of females b/w 15 to 18 -0.095 -0.013 -0.030 0.029 (0.062) (0.056) (0.060) (0.054) No. of males b/w 15 to 18 -0.055 -0.059 -0.091 y -0.072 (0.068) (0.053) (0.053) (0.050) No. of females above age 18 -0.115 -0.121** -0.055 -0.073* (0.077) (0.044) (0.052) (0.029) No. of males above age 18 -0.184** -0.345*** -0.111 y -0.193*** (0.064) (0.037) (0.058) (0.031) Asset Index 0.278* 0.158** -0.032 0.057 (0.119) (0.053) (0.056) (0.038) Owned home value10 6 -3.126 -2.063* -0.548* -0.284 y (2.504) (0.803) (0.237) (0.157) Constant -1.116*** 2.471*** -1.666*** 1.109*** (0.245) (0.197) (0.167) (0.121) Continued on next page 116 Table A.4 { continued from previous page Rural Urban Women Men Women Men Observations 5197 5153 7873 8091 2 test for identifying variables z 23.77*** 131.40*** 18.13** 71.81*** Note: Dependent variable is a dummy equal to one if the individual's wage is observed in the data and zero otherwise. \Crisis Years" is a time dummy equal to one for 1994 and 95, and zero otherwise. For a description of other covariates, please see Table 2.5. SECH 1992-95 is the data used. As there were few women in rural areas with \college & higher" level of education, they are combined with rural women with high school education. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z This 2 statistic tests whether the coecients of selection identifying variables, i.e. No. of females and males b/w 15 to 18, No. of females and males above age 18, Asset Index, and owned home value, are jointly signicant. 117 Table A.5: Estimation of Log of Real Wages for Married People Aged 21 through 65 Controlling for Heckman Selection on Wages (Two-Step Method; First Step in Table A.4), Pooled data 1992-95 Rural Urban Women Men Women Men Primary 0.357 0.162 y 0.264 0.230*** (0.238) (0.091) (0.281) (0.054) Mid School 0.197 0.246* 0.482* 0.297*** (0.423) (0.114) (0.214) (0.069) High School 0.573 0.290* 1.145*** 0.363*** (0.491) (0.137) (0.319) (0.066) College & higher 0.864*** 2.224*** 0.607*** (0.194) (0.624) (0.091) Crisis Years 0.105 -0.025 0.200 0.089 y (0.276) (0.091) (0.169) (0.047) Trend -0.169 0.003 -0.239*** -0.149*** (0.127) (0.052) (0.072) (0.022) I[20<Age30] -0.801* -0.203 y 0.101 -0.355* (0.374) (0.105) (0.339) (0.167) I[30<Age40] -0.572 0.065 0.437 -0.060 (0.401) (0.091) (0.361) (0.215) I[40<Age50] -0.415 0.048 0.582 y 0.084 (0.324) (0.071) (0.348) (0.168) Constant 11.005*** 10.251*** 8.541*** 10.932*** (1.355) (0.119) (0.999) (0.258) 2 test for education z 3.12 19.78*** 15.44** 46.50*** Note: Dependent variable is the individual's log of real wage. For a description of covariates, please see Tables A.4 and 2.5. SECH 1992-95 is the data used. Education dummies are like instruments for log of real wage in the regression in Table 2.11. As there were few women in rural areas with \college & higher" level of education, they are combined with rural women with high school education. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z This 2 statistic tests whether education dummies which are the instruments for log of wage in the regression in Table 2.9, are jointly signicant. 118 Table A.6: First Step of the Two-Step Heckman Selection Model on Non-Zero Hours Worked across People Aged 21 through 65, Pooled Data 1992-95 Rural Urban Women Men Women Men Primary 0.173 0.096 0.178 y 0.196* (0.111) (0.100) (0.094) (0.087) Mid School -0.214 -0.071 0.192 0.229* (0.162) (0.128) (0.122) (0.101) High School 0.233 -0.386*** 0.787*** -0.063 (0.197) (0.110) (0.134) (0.099) College&higher -0.815*** 1.705*** -0.525*** (0.201) (0.148) (0.115) Crisis Years 0.303** 0.018 -0.022 -0.078 (0.095) (0.095) (0.050) (0.062) Trend -0.085* -0.015 0.010 -0.028 (0.043) (0.047) (0.026) (0.029) I[20<Age30] 0.206 -0.024 0.106 0.380*** (0.142) (0.135) (0.150) (0.090) I[30<Age40] 0.246** 0.420** 0.356* 1.023*** (0.089) (0.136) (0.143) (0.103) I[40<Age50] 0.208** 0.262* 0.332* 0.767*** (0.075) (0.122) (0.130) (0.081) No. of females b/w 15 to 18 0.039 0.094 y -0.045 0.032 (0.049) (0.050) (0.044) (0.041) No. of males b/w 15 to 18 0.046 0.030 -0.101* -0.037 (0.051) (0.050) (0.044) (0.041) No. of females above age 18 0.039 0.038 0.064 -0.001 (0.046) (0.045) (0.041) (0.025) No. of males above age 18 -0.160*** -0.176*** -0.125** -0.231*** (0.044) (0.034) (0.039) (0.025) Asset Index 0.151 0.167** -0.101* 0.095** (0.098) (0.059) (0.048) (0.034) Owned home value10 6 -4.482* -2.029* -0.455** -0.290 y (2.158) (1.012) (0.174) (0.151) Constant -0.441* 1.717*** -1.534*** 0.983*** (0.185) (0.146) (0.144) (0.100) Continued on next page 119 Table A.6 { continued from previous page Rural Urban Women Men Women Men Observations 6345 6401 10009 10630 2 test for identifying variables z 11.53** 39.29*** 162.73*** 101.73*** Note: Dependent variable is a dummy equal to one if the individual's hours worked is observed in the data and zero otherwise. \Crisis Years" is a time dummy equal to one for 1994 and 95, and zero otherwise. For a description of other covariates, please see Table 2.5. SECH 1992-95 is the data used. As there were few women in rural areas with \college & higher" level of education, they are combined with rural women with high school education. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z This 2 statistic tests whether the coecients of selection identifying variables, i.e. education dummies, are jointly signicant. 120 Table A.7: First Step of the Two-Step Heckman Selection Model on Non-Zero Hours Worked across Never-Married People Aged 21 through 65, Pooled Data 1992-95 Rural Urban Women Men Women Men Primary 0.458 y 0.604* 0.796** 0.654* (0.270) (0.246) (0.298) (0.318) Mid School 0.380 0.312 0.478 y 0.565 y (0.358) (0.253) (0.261) (0.341) High School 0.147 0.026 1.275*** 0.244 (0.338) (0.253) (0.230) (0.323) College & higher -0.454 1.427*** -0.461 (0.352) (0.262) (0.334) Crisis Years -0.013 -0.094 0.055 -0.074 (0.178) (0.140) (0.137) (0.102) Trend 0.023 0.042 -0.064 0.039 (0.081) (0.072) (0.065) (0.049) I[20<Age30] 0.449 0.698** -0.609*** -0.239 (0.335) (0.268) (0.160) (0.146) No. of females b/w 15 to 18 0.040 0.026 -0.119 -0.069 (0.119) (0.080) (0.109) (0.061) No. of males b/w 15 to 18 0.210 -0.032 -0.235 y 0.053 (0.140) (0.092) (0.122) (0.062) No. of females above age 18 0.067 -0.054 0.061 -0.044 (0.072) (0.058) (0.063) (0.042) No. of males above age 18 -0.186 y 0.088 -0.097 y -0.014 (0.095) (0.060) (0.055) (0.049) Asset Index -0.076 0.028 -0.128 0.074 (0.177) (0.092) (0.086) (0.054) Owned home value10 6 -1.638 -1.022 -0.280 -0.026 (3.394) (1.676) (0.283) (0.286) Constant -1.058* -0.332 -0.744** 0.378 (0.435) (0.377) (0.283) (0.344) Observations 649 1135 1202 2377 2 test for identifying variables z 3.39 31.6*** 45.48*** 83.64*** Continued on next page 121 Table A.7 { continued from previous page Note: Dependent variable is a dummy equal to one if the individual's hours worked is observed in the data and zero otherwise. \Crisis Years" is a time dummy equal to one for 1994 and 95, and zero otherwise. For a description of other covariates, please see Table 2.5. SECH 1992-95 is the data used. As there were few women in rural areas with \college & higher" level of education, they are combined with rural women with high school education. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z This 2 statistic tests whether the coecients of selection identifying variables, i.e. education dummies, are jointly signicant. 122 Table A.8: First Step of the Two-Step Heckman Selection Model on Non-Zero Hours Worked across Married People Aged 21 through 65, Pooled Data 1992-95 Rural Urban Women Men Women Men Primary 0.043 0.010 0.107 0.072 (0.120) (0.099) (0.090) (0.090) Mid School -0.561*** 0.043 0.071 0.078 (0.158) (0.181) (0.131) (0.128) High School 0.242 0.002 0.553*** 0.062 (0.214) (0.173) (0.156) (0.120) College & higher -0.593 2.008*** -0.168 (0.450) (0.173) (0.141) Crisis Years 0.403*** 0.104 -0.054 -0.117 (0.104) (0.132) (0.058) (0.073) Trend -0.124** -0.039 0.030 -0.015 (0.048) (0.064) (0.028) (0.034) I[20<Age30] 0.130 0.243 0.160 0.955*** (0.156) (0.174) (0.165) (0.103) I[30<Age40] 0.276** 0.486** 0.354* 1.206*** (0.104) (0.155) (0.150) (0.110) I[40<Age50] 0.185* 0.362** 0.372* 0.761*** (0.084) (0.114) (0.158) (0.082) No. of females b/w 15 to 18 -0.023 0.134* -0.029 0.091 (0.053) (0.067) (0.055) (0.059) No. of males b/w 15 to 18 -0.042 0.119 y -0.072 -0.031 (0.050) (0.063) (0.049) (0.055) No. of females above age 18 -0.009 0.018 -0.059 -0.017 (0.059) (0.060) (0.047) (0.033) No. of males above age 18 -0.116* -0.123* -0.122* -0.120*** (0.046) (0.053) (0.048) (0.034) Asset Index 0.249* 0.129* -0.033 0.086* (0.099) (0.064) (0.055) (0.041) Owned home value10 6 -6.003* -2.295 y -0.622* -0.281 (2.382) (1.226) (0.247) (0.178) Constant -0.235 1.663*** -1.451*** 0.825*** Continued on next page 123 Table A.8 { continued from previous page Rural Urban Women Men Women Men (0.201) (0.204) (0.169) (0.123) Observations 5197 5153 7873 8091 2 test for identifying variables z 21.22*** 2.06 172.70*** 5.04 Note: Dependent variable is a dummy equal to one if the individual's hours worked is observed in the data and zero otherwise. \Crisis Years" is a time dummy equal to one for 1994 and 95, and zero otherwise. For a description of other covariates, please see Table 2.5. SECH 1992-95 is the data used. As there were few women in rural areas with \college & higher" level of education, they are combined with rural women with high school education. Bootstrapped standard errors, resampled at cluster level and computed using 1000 replications are in parentheses. y p<0.10, * p<0.05, ** p<0.01, *** p<0.001 z This 2 statistic tests whether the coecients of selection identifying variables, i.e. education dummies, are jointly signicant. 124 Bibliography Abbasi-Shavazi, M. J., P. McDonald, and M. Hosseini-Chavoshi (2009). The Fertility Transition in Iran: Revolution and Reproduction. Springer. Aghajanian, A. (1995). A new direction in population policy and family planning in the islamic republic of Iran. Asia-Pacic Population Journal 10 (1). Aghjanian, A. (1991). Population change in Iran, 1966-86: A stalled demographic tran- sition. Population and Development Review 17 (4), 703{715. Angeles, G., D. K. Guilkey, and T. A. Mroz (2005). The determinants of fertility in rural Peru: Program eects in the early years of the national family planning program. Journal of Population Economics 18 (367-389). Bahramitash, R. (2007). Iranian women during the reform era (1994-2004): A focus on employment. Journal of Middle East Women's Studies 3 (2), 86{109. Becker, G. S., W. H. J. Hubbard, and K. Murphy (2009). Explaining the worldwide boom in higher education of women. Working Paper. Behrman, J. R. and M. R. Rosenzweig (2002). Does increasing women's schooling raise the schooling of the next generation? American Economic Review 92 (1), 323{334. Behrman, J. R. and B. L. Wolfe (1987). How does mother's schooling aect family health, nutrition, medical care usage, and household sanitation? Journal of Economet- rics 36(1-2), 185{204. Deolalikar, A. B. (1993). Gender dierences in the returns to schooling and in school enrollment rates in Indonesia. The Journal of Human Resources 28 (4), 899{932. Dooley, M. D. (1982). Labor supply and fertility of married women: an analysis with grouped and individual data from the 1970 u.s. census. Journal of Human Resources 17, 499{532. Du o, E. and L. Bereierova (2002). The impact of education on feritlity and child mor- tality: Does fathers really matter less than mothers? Esfahani, H. S. and R. Bahramitash (2010). Veiled Employment: Islamism and a Political Economy of Women's employment in Iran, Chapter 4: The Transformation of Female Labor Market. Syracuse University Press. Forthcoming. 125 Frankenberg, E., K. Beegle, B. Sikoki, and D. Thomas (1999). Health, family planning and well-being in indonesia during an economic crisis: Early results from the Indonesian family life survey. RAND Corporation Labor and Population Program Working Paper Series No. 99-06. Frankenberg, E., B. Sikoki, and W. Suriastini (2003). Contraceptive use in a changing service environment: Evidence from Indonesia during the economic crisis. Studies in Family Planning 34 (2), 103{116. Frankenberg, E. and T. D. Smith, James P. (2003). Economic shocks, wealth, and welfare: Evidence from the Indonesia family life survey. Journal of Human Resources 38 (2), 280{321. Franz, W. (1981). Schatzung regionaler arbeitsangebotsfunktionen mit hilfe der tobit- methode und des probit-verfahrens unter berucksichtigung des sog. 'sample selection bias'. Discussion Paper No. 171-81, Institut fur Volkswirtschaftslehre und Statistik, University of Mannheim, Mannheim, Federal Republic of Germany. Franz, W. and S. Kawasaki (1981). Labor supply of married women in the federal republic of germany: theory and empirical results from a new estimation procedure. Empirical Economics 6, 129{143. Gheytanchi, E. (2000). Appendix: Chronology of events regarding women in Iran since the revolution of 1979. Social Reserach Journal 67 (2). Goldin, C. (1990). Understanding the Gender Gap: An Economic History of American Women. Oxford Universiry Press. Goldin, C. (1994). The u-shaped female labor force function in economic development and economic history. Heckman, J. (1980). Female Labor Supply, Chapter Sample Selection Bias as a Speci- cation Error, pp. 206{248. Princeton, NJ: Princeton University Press. Joshi, S. and T. P. Schultz (2007). Family planning as an investment in development: Evaluation of a program's consequences in matlab, bangladesh. Keddie, N. R. (2000). Women in iran since 1979. Social Research Journal 67 (2). Keddie, N. R. (2006). Modern Iran: Roots and Results of the Revolution. Yale University Press. Updated. Killingsworth, M. R. and J. J. Heckman (1986). Female labor supply: A survey. In O. C. Ashenfelter and R. Layard (Eds.), Handbook of Labor Economics (First ed.), Volume 1 of Handbooks in Economics, pp. 103{204. Amsterdam: North-Holland. Majbouri, M. (2007). On the dynamics of women's rights movement in Iran, 1996-2006. McKenzie, D. J. (2003). How households cope with aggreagte shocks? evidence from the Mexican Peso crisis. Wolrd Development (7), 1179{1199. 126 Miller, G. (2005). Contraception as development? new evidence from family planning in Colombia. NBER working paper w11704. Moheseni, M. (2000). Study of awareness, attitudes, and socio-cultural behaviors in Iran (in Persian). Technical report, Secretariat of Public Culture of Iran. Mu, R. (2006). Multiple job holdings as a way to smooth consumption: Labor supply response to wage arrears among Russian couples. Oce, I. L. (2001). Key indicators of labor market 2001-2002. Technical report, Interna- tional Labour Oce, Geneva, Switzerland. Pesaran, M. H. (2000). Economic trends and macroeconomic policies in post-revolutionay Iran. In P. Alizadeh (Ed.), The Economy of Iran: The Dilemmas of an Islamic State, pp. 63{100. London: I.B. Tauris. Rajabzadeh, A. (1999). Tahlil-e Ijtima'i-ye Namgozari (The Social Analysis of Name- giving). Raza, F. A. (2007). Reasons for lack of women's participation in Pakistan's workforce. Journal of Middle East Women's Studies 3 (3), 99{102. Renaud, P. S. A. and J. J. Siegers (1984). Income and substitution eects in family labour supply. De Economist 132, 350{366. Ross, M. L. (2008). Oil, islam, and women. American Political Science Review 102 (1), 107{123. Sahn, D. E. and D. C. Stifel (2000). Poverty comparisons over time and across countries in Africa. World Development 28 (12), 2123 { 2155. Salehi-Isfahani, D. (2000). Demographic aspects of economic development in Iran. Social Research 67(2). Salehi-Isfahani, D. (2003). Labor supply behavior of urban women in Iran. Working paper, Department of Economics, Virginia Tech. Salehi-Isfahani, D. (2005a). Human resources in Iran: potentials and challenges. Iranian Studies 38 (1), 117{147. Salehi-Isfahani, D. (2005b). Labor force participation of women in Iran. Available from http://lebox/users/salehi/research/htm. Salehi-Isfahani, D., M. J. Abbasi-Shavazi, and M. Hosseini-Chavoshi (2009). Family planning and rural fertility decline in Iran: A study in program evaluation. Working paper. Salehi-Isfahani, D. and M. Majbouri (2008). Mobility and poverty dynamics in Iran: Evidence from 1992-95 panel survey. 127 Salehi-Isfahani, D. and M. Marku (2006). Women and market work in Iran: A cohort approach. Working paper, Department of Economics, Virginia Tech. Smith, J. P., D. Thomas, E. Frankenberg, K. Beegle, and G. Teruel (2002). Wages, employment, and economic shocks: Evidence from Indonesia. Journal of Population Economics 15, 161{193. Stelcner, M. and J. Brestaw (51). Income taxes and the labor supply of married women in Quebec. Southern Economic Journal 1985, 1053{1072. Strauss, J. (1990). Households, communities, and preschool children's nutrition outcomes: Evidence from rural Cote d'Ivoire. Economic Development and Cultural Change 38 (2), 231{61. The Economist (2009, October). Fossilised policy: Fuel subsidies. The Economist. Thomas, D., K. Beegle, and E. Frankenberg (2003). Women in the Labour Market in Changing Economies: Demographic issues, Chapter Labor Market Transitions of Men and Women During an Economic Crisis: Evidence from Indonesia. Oxford University Press. Thomas, D., K. Beegle, E. Frankenberg, J. Strauss, B. Sikoki, and G. Teruel (2004). Education in a crisis. Journal of Development Economics 74, 53{85. Thomas, D., E. Frankenberg, K. Beegle, and G. Teruel (2005). Household budgets, household composition and the crisis in Indonesia: Evidence from longitudinal survey data. Califronia Center for Population Research On-line Working Paper Series. Thomas, D. and J. Strauss (1992). Prices, infrastructure, household characteristics and child height. Journal of Development Economics 39 (2), 301{331. Thomas, D., J. Strauss, and M. H. Henriques (1991). How does mother's education aect child height? Journal of Human Resources 26 (2). Wooldridge, J. (2002). Econometric Analysis of Cross Section and Panel Data. Cam- brdige, MA: MIT Press. Worth, R. F. (2009, December 2). Iran plans to phase out subsidies brings frenzied debate. New York Times, A11. 128

Abstract (if available)

Abstract Women in Iran have garnered extraordinary achievements in the last two and a half decades. Fertility rate has fallen in one of the largest and fastest transitions in modern human history. Meanwhile, education levels have consistently increased, to the extent that currently, women who are in their late 20s are more educated than their male counterparts. But still, female labor force participation (FLFP) rates remain at the low levels of two decades ago (FLFP puzzle), a fact which has led many researchers to suggest that FLFP is inelastic to economic forces.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays on political economy of privatization

PDF

Essays on education: from Indonesia to Los Angeles, CA

PDF

Community forest management in Nepal: saving the forest at the expense of the poorest

PDF

Essays on health economics

PDF

Essays on child labor and poverty in the context of a conditional cash transfer program in Nicaragua

PDF

Essays in political economy and mechanism design

Asset Metadata

Creator Majbouri, Mehdi (author)

Core Title Against the wind: labor force participation of women in Iran

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Publication Date 08/05/2010

Defense Date 04/26/2010

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Economic development,Iran,labor force participation,OAI-PMH Harvest,Women

Place Name Iran (countries)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Strauss, John A. (committee chair), Melguizo, Tatiana (committee member), Moon, Hyungsik Roger (committee member), Nugent, Jeffrey B. (committee member), Salehi-Isfahani, Djavad (committee member)

Creator Email mahdi.majbouri@gmail.com,majbouri@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m3304

Unique identifier UC1324925

Identifier etd-Majbouri-3875 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-366397 (legacy record id),usctheses-m3304 (legacy record id)

Legacy Identifier etd-Majbouri-3875.pdf

Dmrecord 366397

Document Type Dissertation

Rights Majbouri, Mehdi

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu