Oil and Female Labor Force Participation Mahdi Majbouri Babson College Preliminary - Please do not cite without permission. Abstract: Despite the rapid rise of women’s education in the Middle East and the simultaneous fall in their fertility rates, female labor force participation (FLFP) is at the same low levels of two decades ago. Although many hypotheses have been proposed, this still remains a puzzle. In a recent study, Ross (2008) argued that it is not Islamic culture but oil and gas rents that contribute to low FLFP. Controlling country fixed effects, I show that oil and gas rents reduce FLFP in Muslim countries three times more than in non-Muslim countries. I argue that Islam and oil and gas rents jointly matter, and traditional institutions actually may be strengthened in the presence of oil and gas rents. JEL Classification Code: J71, J21, P48, Q39 Key Words: Female Labor Force Participation, Islam, Oil and Gas Rents 1. Introduction Women in the Middle East have achieved extraordinary accomplishments over the past three decades. As shown in Table 1, fertility rates have fallen in the region. For example, in Iran, they have fallen universally in one of the largest and fastest transitions in modern human history, from about seven in 1984 to almost two children per woman in 2002. At the same time, education for girls has increased continuously in such a way that in some countries of the region more than half of the students in college are now women. Table 2 depicts the change in average years of schooling for women aged 20 to 30 between 1970 and 2010. In some countries the increase is more than 8 years in that period. For example, in Saudi Arabia, one of the most conservative societies in the region, they increased by 9.1 years. Expansion of infrastructure, increased access to electricity and clean water, and a rise in incomes allowed households to acquire home appliances, which reduce the time needed to do household chores. For example, Table A.1 shows such developments for the case of Iran. As a result of these changes, household production technologies have continuously improved, giving women more free time and making them more productive at home. The economic theory and evidence show that this should increase FLFP (Cavalcanti and Tavares 2008; Coen-Pirani et al 2010). Despite these achievements, female labor force participation (FLFP) rates remain at their low levels of two decades ago. Figure 1 shows FLFP rates of MENA countries in comparison to OECD countries. These rates are lower not only compared to OECD countries but also to the rest of the world. For example, in Iran, 15% of urban women aged 21 through 65 and 25% of rural women in the same age group participate in the labor force. Global Entrepreneurship Monitor (GEM) which is the largest study of entrepreneurship across countries reports that entrepreneurship rates in GEM countries in the Middle East and North Africa (MENA) region compared to the other GEM 2 countries “are generally lower than might be expected for countries at their level of development” (GEM 2009, page xv). Moreover, as depicted in Figure 2, “Men are much more likely than women to be involved in early-stage entrepreneurial activity: an average of about 19% of adult men and 9% of adult women” (GEM 2009, page xvi). This interesting phenomenon has raised curiosity in social scientists, and led them to argue that FLFP is inelastic to many economic and non-economic forces. Unable to explain it, they called it a puzzle and it still remains one (see Majbouri (2011) for a detailed description of the puzzle)1. There are many competing hypotheses that try to explain this phenomenon (for a survey of them, please see Chapter 1 of Majbouri (2010)). Discrimination on the supply as well as the demand side are among the prominent ones. Traditional institutions including Islam were considered to play a major role (for example see World Bank 2004), though it has been difficult to provide strong scientific evidence for this. Some argue that Islam cannot be blamed for this difference as there are countries which are predominantly Muslim but have high FLFP rates, such as Bangladesh and Indonesia. On the other hand, there are many cultures which have discriminatory norms against women but have higher FLFP than the countries in the Middle East and North Africa. So it seems that in addition to the traditional discriminatory norms, there may be other factors, unique to the Middle East, that affect labor force participation. The first factor that may come to mind is oil and natural resource rent. Using cross-country regressions, Ross (2008) argued that in fact it is oil and gas rent that lowered FLFP, and not Islam. Since Islam does not change over time, it is difficult to rule out unobservable characteristics of the countries. This paper has two contributions to the literature. First, it challenges the argument by Ross (2008) that Islamic norms and culture does not affect female labor force participation (FLFP). I show that this argument depends on the model 1 Labor Force Participation of Women has been discussed in Bahramitash and Salehi-Esfahani (2011), Moghaddam (2009), Moghaddam (1999), Moghaddam (2005), Robinson (2005), Salehi-Esfahani and Shajari (2010), Salehi-Isfahani (2002), Salehi-Isfahani (2005), Spierings and Smith (2007) 3 specification and is not robust. More importantly, I show that although oil and gas rents reduce FLFP, they reduce it three times more in Muslim countries than non-Muslim ones. In other words, traditional Islamic institutions could be strengthened in the presence of oil and gas rents and as a result they reduce FLFP more. We knew that rents reduce FLFP through several mechanisms, discussed in the literature and explained by Ross (2008). The second contribution of this paper is to add a new mechanism that was overlooked: oil and gas rents may reduce FLFP through strengthening the traditional institutions. In Section 2, I explain this hypothesis when I discuss the impact of oil and gas rents on FLFP. Then, I will briefly talk about the data I use in Section 3. Section 4 describes the econometric models to test this hypothesis. First, I will use cross-country regressions to show whether such an argument stands, and then I employ first-difference regressions with country fixed-effects to be able to control for unobserved fixed characteristics of countries. The results are robust to model changes. I will conclude the paper in Section 5. 2. The Impact of Oil and Gas Rents Countries whose oil, gas, or other natural resources make up a large portion of their exports are affected by both positive and negative consequences of these valuable commodities. Since the cost of production is considerably smaller than the world price for such commodities, they produce a large windfall of rent for the economy as a whole, and most of the time for the government in charge. Although this large influx of income can benefit the citizens on many fronts, it has significant negative consequences for them in the short and long-run, unless the rent is managed and used wisely on the macro level. The most common negative impact of natural resource rent is termed the “Dutch disease,” which happens when the influx of foreign currency into the economy, earned by selling a natural resource, leads to a rise in real exchange rates and makes the tradable 4 sector uncompetitive. It also directs investments and resources towards the non-traded sector and makes it larger than normal. 2 Moreover, the literature on the social and political consequences of natural resource rents is extensive. Here, I want to briefly discuss the impact of natural resource rent on FLFP. For a detailed discussion, please refer to Ross (2008). The decision to participate in the labor force for female (and male) members of the household depends on three factors: 1) female members’ wages, 2) male members’ wages, and 3) household non-labor income. These factors can be shown in a canonical model of labor force participation. Suppose that the representative household is a single entity maximizing its inter-temporal utility subject to budget and time constraints as follows: { } ∑( ) Subject to in which, (1) ∑ ( ) ( ) is the vector of goods household consumes at time , and , are leisure times for female and male members of the household at time . In the budget constraint, household’s initial wealth, and is the price vector at time . and male members of the household, and , , are hours worked by female are their respective wages. government transfers to household at time and denotes the represents shows the interest rate at time . Assuming that the household has perfect foresight, the first order condition for leisure can be written as 2 For a detailed discussion of natural resource economic impact see van der Pleog (2011). 5 ( in which, ) ( ) is the Lagrangian multiplier of budget constraint, when time constraints are substituted into the utility function and the budget constraint. The Lagrangian multiplier is a function of initial wealth, past and current wages, and prices. I assume that household has perfect foresight therefore is constant over time. This inequality becomes an equality if the individual works. Hence hours worked can be written as a function of wages, prices, interest rates, and transfers as follows: ( ) (2) When rent from oil and gas is one of the major sources of government income, governments may be tempted to transfer part of the rent to households. The comparative statics of equation (2) reveals that non-labor income, such as government transfers, reduce hours worked, i.e. Since , one can conclude that larger rents would reduce labor force participation for both male and female members of the household. Hence in countries where rent is dominant, government transfers to households are larger, and that could be one reason why labor force participation is smaller. For example, in Iran, recently the government transformed the large subsidies of energy and basic food into lump-sum monthly transfers to households. In addition to larger government transfers, the influx of oil and gas rent into government coffers increases government tendencies to make large investments in social welfare by subsidizing some goods and services, from free education and health care programs to highly subsidized energy and 6 food staples. These investments reduce prices households face (i.e. in Equations (1) and (2)). The economic model predicts that such reductions in prices have two impacts on hours worked: one is the substitution effect and the other is the income effect. If the income effect dominates the substitution effect, then people work less. Another mechanism by which rent may reduce female labor force participation is through distortions in the labor market. In developing countries, two of the major sectors for female employment in the early and middle stages of development are agriculture and export-oriented manufacturing. Jobs in the export-oriented manufacturing sector, such as in textile factories and electronic device manufacturing plants, do not need physical strength, but at the same time they require precision and smaller hands. In addition, many manufacturing jobs require little education or skills. These factors qualify many women in poor countries for such jobs. There is evidence that export-oriented firms employ women at a higher rate (Başlevent and Onaran 2004; Ozler 2000). One reason is that export-oriented industries sell to the global market and grow very quickly if they are successful, and hence they cannot just rely on male workers for labor. Another reason is that firms in such industries are usually owned by foreign companies that tend to discriminate against women (in hiring) far less than do domestic firms, because of legal or cultural issues. Moreover, these firms want to minimize costs as much as possible to be able to compete in global markets. Therefore, they prefer to hire women, whose wages are usually lower then men’s. Manufacturing is a tradable sector and because of the Dutch disease, usually it becomes weak in economies that are dominated by natural resource rent. Hence the demand for female labor in such economies could be weak if the non-traded sector could not compensate for loss of jobs. On the other hand, the non-traded sector in a poor economy mostly consists of construction and services such as retail. Construction is a male-dominated sector as it requires physical strength. Moreover, 7 services such as retail in poor countries are usually small-size businesses that may not require more than one employee who is usually the (male) owner of the business. Hence, a weak manufacturing sector in natural resource exporting economies means fewer opportunities and hence lower wages for women (smaller ). The economic theory explains that lower wages would induce women to participate less only if the substitution effect is larger than the income effect of lower wages. The general evidence points to the fact the substitution effect for women is larger than the income effect. So this can be considered another reason for low female labor force participation in economies dominated by rent. The fact that the non-traded sector is becoming larger also means that there will be more opportunities for men. This in turn increases their wages. Theoretically, higher wages for male members of household (larger force (smaller ), reduce female members’ tendency to participate in the labor ) if female and male labor are substitutes.3 From this discussion one can say that in an economy dominated by oil and gas rent, the wages, prices, interest rates, and more importantly government transfers all are affected by that rent. We can formally write this as ( ) (3) (4) (5) (6) in which, is the size of rent in the economy at time that affects prices in all markets and government transfers. 3 and are the observed and unobserved characteristics of individual in For a detailed discussion of income and substitution effects of wages, and non-labor income see Killingsworth and Heckman (1986). 8 household at time , and is the time trend. is the country (economy) specific unobservable characteristics that are constant over time (such as institutions). , , and are random disturbances in each time period. Substituting Equations (3) through (6) in Equation (2), linearizing the function and averaging over all individuals of the same gender in a country, one finds a reduced form equation for each gender. For females the equation would be (7) in which, , , are respectively the labor force participation of women, rent from oil and gas, and observed characteristics of the female population in country fixed characteristics of country over time, and is the country at time . is unobserved specific time trend coefficient. In this paper, I argue that the impact of oil and gas rents on FLFP can be larger in Islamic countries, as they can strengthen the legal as well as cultural forces that tend to limit FLFP. Since men’s wages are larger in natural resource exporting countries, they have the power to limit women’s participation if they want. In other words, if women’s participation in the labor force is disliked by men, and if norms give male members of the household the privilege to decide for (or overrule decisions by) female members of the household, then they are more likely to exercise such a privilege. In fact, it is not only the norm but the legal responsibility of women to seek sanction of the male head of the household for their decision to participate in the labor force. Hence, oil and gas rent may reinforce traditional Islamic roles, norms, and rules in the society. 9 3. Data I use data produced by John Ross at the University of California at Los Angeles, and used in Ross (2008).4 The Islam variable in this data set, which shows the percentage of the population that are Muslim in every country around the world, is standardized. Since for this study I needed the actual values of this variable in order to produce the interaction of this variable with oil and gas rents variable (simply Rents from now on), I used the 2009 Pew Report to obtain values of this variable for 20095. 4. Econometric Models and Results The parsimonious econometric model to identify the joint role of oil and gas rents and Islam or other cultural norms and FLFP can be written as a simple OLS regression of country-level variables. In fact, this model is simply the average of Equation (7) over time, and can be written as Equation (8). FLFP, i.e. , is the dependent variable and oil and gas rents, Islam, and the interaction of the two are among the explanatory variables. (8) includes variables such as log of GDP per capita and squared log of GDP per capita for country , represents the oil and gas rents per capita for country , and of Muslims out of the total population in country . regression, except is the share is the error term. All variables used in the , contain their standardized average values between 1993 and 20026. For further information about data please see Ross (2008). The data is available for download at: http://dvn.iq.harvard.edu/dvn/dv/mlross/faces/study/StudyPage.xhtml?globalId=hdl:1902.1/14307 5 Miller, Tracy, ed. October 2009. Mapping the Global Muslim Population: A Report on the Size and Distribution of the World’s Muslim Population, Pew Research Center, retrieved 2009-12-17. 6 I used the same sample as Ross (2008) so that the results would be comparable. The variables were averaged over a ten-year period, 1993 through 2003, to reduce the variation. Then the result was standardized. 4 10 has its country-level value in 2009. Since this variable hardly changes in a span of one or two decades, I do not expect that the results would be different if I use the average values of Islam between 1993 and 2002. In order to get comparable coefficients, all variables were standardized. To standardize the interaction variable, first rent was multiplied by Islam, and then the result was standardized. Columns 1 and 2 of Table 3 depict this regression. In column 1 the interaction of Islam and Rents is not included. It shows that both oil and gas Rents and Islam are statistically significant predictors of FLFP. They both have negative and large correlations with FLFP. With the introduction of the interaction between Islam and Rents, the coefficient of Rents loses its significance while Islam and its interaction with Rents have statistically significant relationships with FLFP. This may imply that it is not the rent per se that lowers FLFP, but when it is combined by cultural norms set in Muslim societies, FLFP decreases. In columns 3 and 4, I add MENA, which is a dummy variable equal to one if the country belongs to the Middle East and North Africa Region, to regressions specified in columns 1 and 2, respectively. Ross (2008) includes this dummy variable in the regression, but provides little theoretical reasoning to explain why this variable should be included. This variable captures any common characteristic among countries of the region except the large share of Muslim population. Adding this variable to the regression changes the coefficients and their significance levels and overrides the results in columns 1 and 2. For example, coefficient of Islam is no longer significant in any of these settings. Although Rents is negatively correlated with FLFP in column 3, when the interaction of Islam and Rents is included, it loses its significance. This time the interaction of Islam is also insignificant. It is not clear whether the inclusion of MENA would rule out the hypothesis that Islam and its interaction with Rents reduces FLFP. For example, MENA may show the length of time countries 11 were Muslim. One may argue that Muslim countries outside of the MENA region were converted to Islam in recent centuries, while those in MENA have a long history of Islam and Islamic culture. Therefore, the MENA dummy would capture such differences which are not controlled by the share of Muslim population in the country (Islam variable). Even if this MENA dummy does not capture anything related to Islam, still it includes social and cultural norms common among these countries. So its significance shows that cultural norms matter even when we control for oil and gas rents. If we assume that it is the cultural norms shared in the MENA region (among all other common factors) that affect FLFP, it may be interesting to interact this dummy variable with Rents. The hypothesis would be whether oil and gas rents enforce these cultural norms the same way that they may enforce Islamic traditions. Column 5 of Table 3 shows the result of such regression. The coefficient of MENA is strongly significant and negative while Rents is also negatively correlated with FLFP. The coefficient of the interaction between MENA and Rents is not significant at 10% but it is significant at the 14% level. In other words, there is a 14% chance that this coefficient was zero and we found it not to be zero because of our sample. Again, since we do not know what MENA represents conceptually7, we cannot provide a reliable interpretation of its coefficient or the coefficient of its interaction with Rents, theoretically. Since unobserved heterogeneity among the countries is large and the problem of omitted variables may become a major issue in interpreting the coefficients and their significance levels, cross-country regression is not generally a reliable method for identifying causal effects. This may explain why our results are sensitive to model specification and choice of variables. 7 MENA can represent any common social, political, and economic characteristic. 12 One way to mitigate the problem of omitted variables is to use data over (a long) time and control for any unobserved heterogeneity that is fixed over time in each country by first differencing Equation 1. The result is Equation (9). (9) As you can see in this first difference, Islam and MENA are cancelled out. Moreover, any omitted variable that is time-invariant is cancelled out. Therefore, this new regression is more reliable in identifying the causal relationship between the explanatory variables and FLFP. Table 4 depicts this regression. Like the regressions in Table 3, all variables are standardized. Column (1) does not correct for heteroskedasticity in the error term while Columns (2) and (3) do. In Column (3), Saudi Arabia, the country that have a large impact on the regression estimates was omitted from the sample. Interestingly, the coefficients of Rents and its interaction with Islam are negative and statistically significant. This shows that not only do all countries that have larger rents have lower FLFP, but also the negative impact of oil and gas rents on FLFP is larger in Muslim countries. The results are robust to the specification. Interestingly, the impact of oil and gas rents in Muslim countries is three times larger than its impact in other countries. These results may be evidence for the hypothesis that oil and gas rents strengthen traditional Islamic norms that limit women’s participation in the labor market. This enforcement could be through various mechanisms. For instance, oil and gas rents make the public sector larger than normal and also the primary employer for women. But in the eyes of Islamic states only some professions may be considered “suitable” for women, such as teaching or nursing. On the other hand, oil and gas rents increase household purchasing power in a way that makes it possible for male 13 members of the household to provide a decent level of living without any need for female income. Hence they may not allow female members of the household to work. Instead of the interaction of Islam and Rents, one may use the interaction of MENA and rent in Equation 2. Table 5 depicts such a regression. Coefficients of Rents and its interaction with MENA are both significant. But the impact of oil and gas rents in MENA countries is twice as large as its impact in other countries. Similar to Table 4, Column (1) does not correct for heteroskedasticity in the error term while Columns (2) and (3) do. In Column (3), Saudi Arabia, the country that have a large impact on the regression estimates was omitted from the sample. The results are similar across columns. 5. Conclusion Ross (2008) showed that when one controls for the oil and gas rents in countries, the negative correlation between Islam and a number of outcomes for women including FLFP disappears. The first contribution of this paper is that Islam may have a negative correlation with FLFP, but this relationship is not robust to the econometric model specifications. Moreover, these cross-country regressions may imply that almost all of the correlation between FLFP and oil and gas rents is for Islamic countries. This means that in Islamic countries, oil and gas rent may reinforce Islamic norms that in turn reduce FLFP. This challenges the Ross (2008) argument. I used country-level data over time and employed first difference regressions with country fixedeffects to control for any time-invariant omitted variable, since the results from cross-country regressions are not robust. I showed that oil and gas rents produce a larger impact in Muslim countries than in other countries. It could be because oil and gas rents help to enforce traditional Islamic norms. We knew that rents reduce FLFP through several mechanisms discussed in the 14 literature and explained in Ross (2008) but the second contribution of this paper is to add a new mechanism that was overlooked: reducing FLFP through strengthening the traditional institutions. This has an important implication as it explains that in the absence of oil and gas rents, Islamic norms may not be as strong in the society. Further research is needed to identify the mechanisms by which oil and gas rents coupled with Islamic norms affect the supply and demand for female labor. 15 References Abbasi-Shavazi, M. J., P. McDonald, and M. Hosseini-Chavoshi (2009). The Fertility Transition in Iran: Revolution and Reproduction. Springer. Bahramitash, Roksana, and Hadi Salehi Esfahani. 2011. “The Transformation of Female Labor Market,” in Roksana Bahramitash and Hadi Salehi Esfahani (eds.), Veiled Employment: Islamism and a Political Economy of Women's employment in Iran, Syracuse University Press. Barro, Robert J., and Jong-Wha Lee. 2012. “A New Data Set of Educational Attainment in the World, 1950-2010.” Working paper. Başlevent Cem, and Ozlem Onaran. 2004. “The Effect of Export-Oriented Growth on Female Labor Market Outcomes in Turkey.” World Development 32 (August): 1375–93. Cavalcanti, Tiago V. de V., and Jose Tavares. 2008. “Assessing the ‘Engines of Liberation’: Home Appliances and Female Labor Force Participation.” Review of Economics and Statistics, 90.1: 81-88. Coen-Pirani, Daniele, Alexis León, and Steven Lugauer. 2010. “The Effect of Household Appliances on Female Labor Force Participation: Evidence from Microdata.” Labour Economics, 17.3: 503-513. Global Entrepreneurship Monitor. 2009. Global Entrepreneurship Monitor Report of the MENA Region 2009. Killingsworth, Mark R., and James Heckman. 1986. “Female Labor Supply: A Survey,” in Handbook of Labor Economics, Volume I, Edited by O. Ashenfelter and R. Layard. Elsevier Science Publishers B V. 16 Majbouri, Mahdi. 2011. “Against the Wind: Labor Force Participation of Women in Iran.” Dissertation. Miller, Tracy, ed. 2009. “Mapping the Global Muslim Population: A Report on the Size and Distribution of the World’s Muslim Population.”. Pew Research Center. Moghadam, Fatemeh E. 2009. “Undercounting Women’s Work in Iran.” Iranian Studies, 42.1: 81-95. Moghadam, Valentine. 1999. “Gender and Globalization: Female Labor and Women’s Movements.” Journal of World-Systems Research 5 (Summer): 367–88. Moghadam, Valentine M. 2005. “Gender And Social Policy: Family Law and Women’s Economic Citizenship in the Middle East,” International Review of Public Administration, 10.1: 23-44. Ozler, Sule. 2000. “Export Orientation and Female Share of Employment: Evidence from Turkey.” World Development 28 (July): 1239–48. Robinson, Julia. 2005. “Female Labor Force Participation in the Middle East and North Africa.” Wharton Research Scholars Journal, University of Pennsylvania, http://repository.upenn.edu/wharton research scholars/28. Ross, Michael L. 2008. “Oil, Islam, and Women.” American Political Science Review 102(1): 107-123. Salehi-Esfahani, Hadi, and Parastoo Shajari. 2010. “Gender, Education, Family Structure, and the Allocation of Labor in Iran.” Working paper. Salehi-Isfahani, Djavad, ed. 2002. “Labor and Human Capital in the Middle East: Studies of Markets and Household Behavior.” Reading, UK: Ithaca. Salehi-Isfahani, Djavad. 2005. “Human resources in Iran: potentials and challenges.” Iranian Studies 38 (1): 117–147. 17 Spierings, Neils, and Jeroen Smith. 2007. “Women’s labour market participation in Egypt, Jordan, Morocco, Syria & Tunisia: A three-level analysis.” Working paper. van der Ploeg, Frederick 2011. “Natural Resources: Curse or Blessing?” Journal of Economic Literature, 49(2): 366–420. World Bank. 2004. Gender and Development in the Middle East and North Africa. MENA Development Report, Washington DC: World Bank. 18 Figures Figure 1 – Female Labor Force Participation and Unemployment Rate in MENA and OECD Countries Source: World Bank (2004). 19 Figure 2 – Female and Male Total Entrepreneurship Activity in MENA Countries in 2009 Note: The country abbreviations are as follows: YE: Yemen, DZ: Algeria, MA: Morocco, LB: Lebanon, EG: Egypt, TN: Tunisia, JO: Jordan, PS: West Bank and Gaza, SY: Syria. Source: Global Entrepreneurship Monitor: GEM-MENA Regional Report 2009 (GEM 2009). 20 Tables Table 1 – Fertility Transition in MENA Region Total Fertility Rates Algeria Bahrain Egypt Iran Jordan Kuwait Lebanon Morocco Qatar Sudan Syrian Arab Republic Tunisia United Arab Emirates Change in 1980-85 1995-2000 Fertility Rates 6.4 4.6 5.1 6.9 6.8 4.9 3.8 5.4 5.5 6.0 7.4 4.9 5.2 3.3 2.6 3.4 2.3 4.7 2.9 2.3 3.4 3.7 4.9 4.0 2.3 3.2 3.1 2.0 1.7 4.6 2.1 2.0 1.5 2.0 1.8 1.1 3.4 2.6 2.0 Source: United Nations Population Division Estimates. Table 2 – Average Years of Schooling for Women Aged 20 to 30 Algeria Bahrain Cyprus Egypt Iran Iraq Jordan Kuwait Libya Morocco Qatar Saudi Arabia Syrian Arab Republic Tunisia United Arab Emirates Yemen 1970 2010 Change 1.3 2.9 7.8 1.3 1.7 0.9 2.4 3.5 0.5 0.9 4.3 1.5 1.5 1.7 3.9 0.0 10.2 11.8 13.0 8.8 10.0 5.4 10.8 8.4 11.8 5.4 9.8 10.6 5.4 10.2 10.5 3.2 8.9 8.9 5.2 7.5 8.3 4.5 8.4 4.9 11.3 4.5 5.5 9.1 3.9 8.5 6.6 3.2 Data Source: Barro and Lee (2012) 21 Table 3 – Cross-country Regressions of FLFP (1) (2) (3) (4) (5) ln(GDP per capita) -3.92*** (0.60) -3.97*** (0.59) -3.07*** (0.56) -3.12*** (0.57) -3.19*** (0.57) [ln(GDP per capita)]2 3.88*** (0.57) 3.89*** (0.55) 3.11*** (0.54) 3.14*** (0.54) 3.21*** (0.55) -0.11 (0.13) -0.08 (0.13) -0.12 (0.11) -0.11 (0.12) -0.10 (0.11) Communist 0.49*** (0.07) 0.48*** (0.07) 0.44*** (0.06) 0.44*** (0.06) 0.43*** (0.06) Oil and Gas Rents per capita -0.37** (0.14) 0.27 (0.36) -0.24** (0.09) 0.02 (0.25) -0.09+ (0.05) Islam -0.27*** (0.07) -0.24** (0.07) -0.04 (0.06) -0.03 (0.06) -0.05 (0.06) % of population aged 15 to 64 -0.65+ (0.37) Rents per capita MENA -0.27 (0.29) -0.43*** (0.07) -0.41*** (0.07) -0.39*** (0.08) -0.22^ (0.15) Rents per capita Constant -0.02 (0.06) -0.02 (0.05) -0.02 (0.05) -0.02 (0.05) -0.02 (0.05) Observations Adjusted R-squared 161 0.526 161 0.538 161 0.620 161 0.620 161 0.624 Note: FLFP is the dependent variable. Communist is a dummy variable for countries equal to one if the country had a communist system of government for some time during the last 5 decades. Islam is the share of Muslims out of the total population. MENA is a dummy variable equal to one if the country is in the MENA region. All variables used in these regressions, except Islam, contain their average values between 1993 and 2002. Islam contains its country-level value in 2009. In addition, all variables are standardized. Robust-heteroskedastic standard errors are in parentheses. *** p<0.001, ** p<0.01, * p<0.05, + p<0.10, ^ p 0.14 22 Table 4 – First-differenced with Country Fixed-effects Regressions (1) (2) (3) ln(GDP per capita) -0.031 (0.047) -0.031 (0.043) -0.033 (0.044) [ln(GDP per capita)]2 0.006 (0.048) 0.006 (0.049) 0.009 (0.050) % of population aged 15 to 64 0.144*** (0.012) 0.144*** (0.037) 0.144*** (0.037) Oil and Gas Rents per capita -0.035** (0.011) -0.035** (0.011) -0.042*** (0.007) -0.073*** (0.019) -0.073** (0.028) -0.061+ (0.033) 0.048*** (0.009) 0.048*** (0.002) 0.040*** (0.002) No Yes Yes Observations 5,360 5,360 5,326 Adjusted R-squared 0.540 0.033 0.032 Islam Constant Robust-heteroskedastic standard errors Note: FLFP is the dependent variable. Islam is the share of Muslims out of the total population. MENA is a dummy variable equal to one if the country is in the MENA region. All variables used in these regressions, except Islam, are from an annual sample that covers 1960 to 2002. Islam contains its country-level value in 2009. In addition, all variables are standardized. All columns report first-difference with fixed effects estimates. Column (1) has standard errors in parentheses. Columns (2) and (3) have robust-heteroskedastic standard errors in parentheses. Column (3) is estimates for a sample without Saudi Arabia. *** p<0.001, ** p<0.01, * p<0.05, + p<0.10 23 Table 5 – First-differenced with Country Fixed-effects Regressions (1) (2) (3) ln(GDP per capita) -0.046 (0.047) -0.046 (0.046) -0.044 (0.047) [ln(GDP per capita)]2 0.024 (0.048) 0.024 (0.052) 0.023 (0.054) % of population aged 15 to 64 0.140*** (0.012) 0.140*** (0.037) 0.140*** (0.037) Oil and Gas Rents per capita -0.055*** (0.011) -0.055*** (0.014) -0.062*** (0.011) MENA -0.050*** (0.012) -0.050*** (0.011) -0.059*** (0.007) 0.047*** (0.009) 0.047*** (0.001) 0.040*** (0.001) No Yes Yes Observations 5,395 5,395 5,361 Adjusted R-squared 0.539 0.033 0.033 Constant Robust-heteroskedastic standard errors Note: FLFP is the dependent variable. Islam is the share of Muslims out of the total population. MENA is a dummy variable equal to one if the country is in the MENA region. All variables used in these regressions, except Islam, are from an annual sample that covers 1960 to 2002. Islam contains its country-level value in 2009. In addition, all variables are standardized. All columns report first-difference with fixed effects estimates. Column (1) has standard errors in parentheses. Columns (2) and (3) have robust-heteroskedastic standard errors in parentheses. Column (3) is estimates for a sample without Saudi Arabia. *** p<0.001, ** p<0.01, * p<0.05, + p<0.10 24 Appendix Table A.1 – Share of Households Owning Various Home Production Technologies in Iran (in %) Rural Electricity Piped Water 1990 73 62 1995 87 74 2000 97 83 2006 99 90 1990 99 96 53 58 7 3 71 73 12 6 85 82 16 15 95 92 28 40 92 88 40 27 Refrigerator Stove Washer Vacuum Cleaner Source: Majbouri (2011) 25 Urban 1995 2000 100 100 98 100 96 94 48 42 97 96 54 58 2006 100 100 99 98 69 81