Nicole M. Fortin Department of Economics and CIFAR University of British Columbia University of California at Davis May 2010 Source: Goldin (2006) 2 .9 .8 .7 .6 .5 1975 1980 1985 All Men Predicted by Quadratic Canadian Women 1990 Year 1995 2000 2005 All Women College-Educated Women Source: U.S. BLS, March CPS and Canada, Cansim 4 Given that the FLP of college-educated women had almost reach parity with men's in the mid-1990s, this stabilization or slight retreat was disappointing for the women's movement. It has been characterized as “opting out” in the popular press (Belkin, 2003; Wallis, 2004; Story, 2005) and among sociologists (Cotter et al., 2007; Stone, 2007). The “opting-out” phenomena has raised more skepticism among economists (Boushey, 2005; Goldin and Katz, 2007) Women’s educational attainment has continued to rise, as well as their relative wages. Their husband’s income has remained relatively unchanged, with an elasticity of income quite low (Blau and Kahn, 2005) Demand-side factors (e.g. technological change, sectoral shifts; Black and Spitz-Oener, 2007) favor women’s work. Some evidence in male-dominated jobs (Antecol, 2010). 5 Placed in the recent literature that has emphasized the role of social norms and beliefs in modulating the impact of economic fundamentals on labor market outcomes, this paper appeals to changing gender role attitudes, whose progression halted in the mid-1990s, as a source of explanation for the retreat in FLP. I will argue that gender role attitudes are the missing genderspecific factors that explain the differences in the concavity of time trends in male and female labor force participation, which remain after accounting for the usual factors, which include years of schooling, number of children, dummies for white, married, ever divorced, preschooler present, mother ever worked, living in an intact family, and dummies (9) for religion at age 16, and region dummies (8). 6 Recognizing that traditional gender role attitudes, which capture the notion the husband should be the main “breadwinner" and the wife the main “homemaker", are not necessarily antagonistic to egalitarian attitudes, which capture the notion that women are as capable as men in the workforce (more an equal opportunity view), the impact of both type of attitudes is studied. Accommodating both views or identities (“homemaker” vs. “career women”) in the Work-Life Balance (WLB) has arguably become the new face of feminism. Gender role attitudes are thought to impact not only the work decisions of married women, but also those of lone-mothers who have the option of relying on government assistance, or indeed the career and lifestyle choices of single women. 7 Gender Role Attitudes Birth Years Cohort Name Survey: Labor Force Participation CPS GSS Age at HIV/AIDS First Subjective Traditional Egalitarian Marriage Risk GSS GSS GSS NHIS Years: 1983-1985 1966-75 Generation X All 59.0 63.1 — 61.7 — 45.0 — 66.3 — 20.5 — — 65.7 67.3 56.8 67.8 36.2 41.4 75.2 70.6 18.7 20.8 6.4 5.3 70.5 69.8 68.3 71.4 32.2 38.0 78.7 75.9 19.9 21.6 20.2 16.3 77.0 71.9 75.4 72.9 36.8 40.3 73.2 72.5 — — 13.2 11.5 73.8 70.1 72.9 70.2 41.0 39.9 69.6 72.2 23.2 22.2 10.6 9.5 Years: 1988-1990 1966-75 Generation X All Years: 1993-1994 1966-75 Generation X All Years: 1998-2000 1966-75 Generation X All Years: 2004-2006 1966-75 Generation X All 8 Economic Fundamentals: Education, Wages, Non-wage Income Mother’s Labor Force Participation Woman’s Decision to Participate in the Labor Market Who am I ? Career Woman or Stay at Home Mom Other influences 9 When gender role attitudes observed in the past are used to explain current labor force participation (as with the NLS72), the issue of reverse causality does not arise. But when gender role attitudes and FLP are observed contemporaneously (as with the GSS), potential endogeneity arises as an important challenge in the analysis of the impact of these attitudes. The analysis uses a double prong instrumental variable strategy appealing to extraneous attitudes found in the GSS, and to an exogenous shock to attitudes, namely the AIDS scare, which may have acted as a counter-current to the “Pill Revolution”. The AIDS scare is thought to be an exogenous shock to attitudes, but perhaps only one factor among others affecting these attitudes. 10 Economic Fundamentals: Education, Wages, Non-wage Income Mother’s Labor Force Participation Woman’s Decision to Participate in the Labor Market Who am I ? Career Woman or Stay at Home Mom Other influences 11 Economic Fundamentals: Education, Wages, Non-wage Income Mother’s Labor Force Participation Woman’s Decision to Participate in the Labor Market Who am I ? Career Woman or Stay at Home Mom Other influences AIDS Scare 12 This paper contributes to the study of the impact of gender role attitudes on FLP in several novel ways, by 1) Accounting for non-linear time-period, life-cycle and cohort effects, as well as a host of background variables, using data from the 1977-2006 General Social Surveys (GSS) 2) Using longitudinal data from the single cohort National Longitudinal Survey of the High School Class of 1972 (NLS72) to address concerns about reverse causality and corroborate the age-period-cohort specification. 3) Using a double prong instrumental variables strategy based on extraneous attitudes about sexual morality and political views found in the GSS, and on an exogenous shock to attitudes, namely the mid-1990s AIDS scare using repeated cross-sectional data from the 19882006 National Health Interview Surveys (NHIS) in the context of a variant of two-sample two-stage least squares (TS2LS) to address concerns about endogeneity. 13 Gender role attitudes, whose secular trends reversed in the mid1990s when the AIDS crisis peaked, are found to explain at least a third of the recent leveling-off in FLP, that is, as much as all the usual variables combined while general cultural trends towards more conservative social, religious and political views do not. More precisely, the estimated coefficients of gender role attitudes imply that the 2 points rise in average traditional attitudes and 4 points decline in average egalitarian attitudes from 1993-94 to 200406 would account for one half to a full percentage point decline in FLP. 14 (1) Explanatory variables: Econometric Specification All Men OLS (life-cycle & cohort controls) All Women OLS (life-cycle & cohort controls) OLS (additional usual controls) LPM 2SLS 2SLS/TS2SLSa Time (2) (3) Traditional 2 Time /100 attitudes (4) Egalitarian attitudes (5) (6) Time Time /100 Years: 1977-2006 2 0.019*** -0.040*** (0.008) (0.016) 0.015*** (0.003) 0.010*** (0.001) 0.006 (0.004) 0.007* (0.004) 0.006* (0.003) 0.039*** (0.011) 0.031*** (0.010) 0.029*** (0.011) 0.012*** (0.004) 0.010*** (0.004) -0.219*** (0.024) -0.231*** (0.070) (8) Egalitarian attitudes Years: 1988-2006 0.006*** -0.012** (0.002) (0.006) -0.025*** (0.008) -0.019*** (0.007) -0.010 (0.008) -0.011 (0.007) -0.010 (0.008) (7) Traditional attitudes 0.051** (0.021) 0.257*** (0.088) -0.067*** (0.020) -0.056*** (0.020) -0.052*** (0.019) -0.054*** (0.017) -0.046*** (0.018) -0.196*** (0.027) -0.239*** (0.083) 0.048* (0.025) 0.313** (0.133) Note: Robust standard errors are in parentheses. Significance at 1% level denoted by ***, at 5% level denoted by **, at 10% level denoted by *. Also included in the regressions are quadratic in age, cohort dummies (7), years of schooling, number of children, dummies for white, married, ever divorced, preschooler present, mother ever worked, living in an intact family, and dummies (9) for religion at age 16, and region dummies. 15 0 .2 .4 .6 .8 A. Attitude Sample - Table 4 1980 Actual Trend Col.1 1985 1990 Year Pred. Col.1 Trend Col.2 1995 Pred. Col.2 Trend Col.5 2000 2005 Pred. Col.5 Male Trend .2 .4 .6 .8 B. Post-1988 Sample - Table 7 1990 1995 2000 2005 Year Actual Trend Col. 2 Pred. Col. 2 Trend Col. 3 Pred. Col. 3 Trend Col. 9 Pred. Col. 9 Male Trend 17 The results are strongest among women with less than a college degree, for these women the AIDS scare as an instrument is actually more significant. Among married women, where the analysis include husband’s income, gender role attitudes account for about one third of the time trends. Among college-educated married women (about 800 obs), the results are less precise and not as strong, leaving room for explanations based on negative feedback from the labor market, such as glass ceiling effects. Among African-American women however, self-rated health is found to be as important as gender role attitudes in accounting for the evolution of FLP. 18 Relevant Literature Theoretical Underpinnings: Economic Identity Theory and the AIDS Scare Data and Descriptive Evidence Time-period, life-cycle and cohorts effects in FLP Measurement of gender role attitudes Subjective risk of HIV/AIDS Econometric Specification and variant of TS2SLS Regression Results Main results Longitudinal analysis with the NLS-72 Instrumental variables strategies Alternative hypotheses (divorce, religious, social and political conservatism, ethnic and health factors) Sub-groups (African-American women, College-educated women vs. women with less than college, men) Conclusion 19 Impact of gender role attitudes on FLP and other labor market outcomes Levine (1993), Vella (1994), Fortin (2005), Fernandez and Fogli (2005), Charles, Guryan, and Pan (2009) Intergeneration transmission of gender role attitudes Fernandez, Fogli and Olivetti (2004), Farré-Olalla and Vella (2007) Dynamic macro-models with gender role attitudes Fernandez (2007) and Fogli and Veldkamp(2007) Cohort-effects – “Pill Revolution” Goldin and Katz (2002), Goldin (2004) and Bailey (2006) Identity and Gender Akerlof and Kranton (2000, 2002), Goldin and Shim (2004), Goldin (2006) • Impact of AIDS Crisis, Mad Cow Scare • Ahituv, Hotz and Philipson (1996), Adda (2007), Johnson and Raphael (2009) 20 • Akerlof and Kranton (2000, 2002, 2005) have proposed to incorporate one's sense of self as an important element of the utility function. Identity translates cultural values and social norms into motivational factors: agents act as they “ should” given their chosen social category. Benabou and Tirole (2006) have introduced competing identities that are competing for time or resources, such as a traditional identity and a modern identity, where investing in the identitycapital of one can damage the other. Here I retain some of the basic elements of their framework: retain the basic elements of the framework: identity-endowment, identity-asset and saliency of identities. 21 The “Women’s Liberation Movement” can be seen as having proposed the new identity of “career women” equal to men in the workplace, and assuming their own identity by keeping their birth name (Goldin and Shim, 2004). The “Pill Revolution” enabled the new career woman to be a “liberated” woman in control of her sexuality and fertility, that is, created a disassociation between the “marriage market” and the “sex market”. In that context, the AIDS scare may have acted as a countercurrent to the Pill Revolution making the lifestyle of the single, but condom-able, career woman less comfortable. Other health scares, such as the 2003 Mad Cow and SARS scare, the 2009 Swine Flu scare, also had profound impacts on attitudes, identities and even trade. 22 Data from the National Health Interview Surveys which have been conducted yearly, for over fifty years, as one of the major data collection programs of the National Center for Health Statistics (NCHS) combine information on health characteristics and many demographic and socioeconomic characteristics. Designed to monitor the health of the U.S. population, these large repeated cross-sectional surveys comprise from 20,000 to 30,000 observations a year. In the late 1980s, the NCHS added they an “AIDS Knowledge and Attitudes" supplement; in 1997 a subset of that module was incorporated into the core components “Sample Adult” . The information one's own chances of getting HIV/AIDS is from the question: “What are your chances of getting HIV/AIDS? High/Already have HIV/AIDS (1), Medium (2/3), Low (1/3), or None (0).“ 23 0 .05 .1 .15 .2 .25 A. Changes over Time 1990 1995 2000 2005 Year <=1920 1946-52 1966-75 1921-39 1953-58 1976-88 1940-45 1959-65 All Note: Average of “What are your chances of getting HIV/AIDS? High /Already have HIV/AIDS (1), Medium (2/3), Low (1/3), or None (0).“ 25 The main data are drawn the 1977 to 2006 General Social Surveys (GSS) conducted yearly (or bi-yearly) by National Opinion Research Center. Each cross-section comprises 1372 to 2992 observations per year with a total of total of 20,000 females and 19,194 males between the ages of 18 and 65. But the sample for which consistent gender role attitudes are measured comprises a subset of about 9000 women, because the same questions are not asked in each survey. Data from National Longitudinal Survey of the High School Class of 1972 (NLS72) which follows the first post-Pill cohort (1954-55 birth cohorts) are also used. 29 .2 .3 .4 .5 .6 .7 .8 .9 1 B. Changes over the Life-Cycle 20 25 <=1920 1946-52 1966-75 30 35 40 Age 1921-39 1953-58 1976-88 45 50 55 60 65 1940-45 1959-65 All 30 .2 .3 .4 .5 .6 .7 .8 .9 1 A. Changes over Time 1975 1980 <=1920 1946-52 1966-75 1985 1990 Year 1921-39 1953-58 1976-88 1995 2000 2005 1940-45 1959-65 All 31 Various flexible A-P-C specifications show the following parsimonious specification captures well the time trend in FLP Yit = α0+ α1T + α2T2 + α3 A + γ4A2 + Σj δj Bj + βgGit + βxXit + εit where Yit is the outcome of interest where T is time , A is age and Bj are the 8 birth cohort categories where Git are gender role attitudes that capture the saliency of traditional or egalitarian identities where Xiti individual characteristics that capture the identityendowment (living in intact family, mother ever worked, religion at 16, etc.), identity-asset (education, children, married, divorced, etc.) variables For simplicity, the model is estimated with a Linear Probability Model, but corroborated by a Probit model 36 (1) (2) (3) Mean LFP Time (1977=1) 0.012*** (0.002) -0.025*** (0.005) 0.017*** (0.002) 2 Time /100 -0.034*** (0.005) Age 0.044*** 0.043*** (0.002) (0.002) 2 Age /100 -0.051*** -0.057*** (0.003) (0.002) Birth cohort (1953-1958 omitted) <=1920 -0.354*** -0.290*** (0.027) (0.030) 1921-39 -0.175*** -0.162*** (0.013) (0.016) 1939-45 -0.072 -0.079*** (0.015) (0.015) 1946-52 0.005 -0.017 (0.012) (0.012) 1959-65 -0.011 0.025** (0.012) (0.012) 1966-75 -0.051*** 0.033** (0.013) (0.014) 1975-88 -0.101*** 0.069*** (0.019) (0.030) Observations 20000 20000 20000 R-squared 0.04 0.059 0.061 (4) (5) All Women: 0.672 (6) (7) (8) (9) (10) (11) All Men: 0.879 0.018*** (0.002) -0.033*** (0.005) Full set of age dummies (65 omit.) Full set 0.018*** of time (0.002) -0.033*** dummies -0.031*** (0.005) (1998 omit.)(0.005) Full set Full set 0.036*** of age of age (0.003) dummies dummies -0.050*** (65 omit.) (65 omit.) (0.003)*** 0.012*** (0.002) -0.023*** (0.005) 0.044*** (0.003) -0.057*** (0.003)*** 0.002 (0.002) -0.004 (0.003) 0.052*** (0.002) -0.067*** (0.003) 0.002 (0.002) -0.007* (0.004) Full set of age dummies (65 omit.) 0.027 (0.054) -0.013 (0.033) 0.001 (0.022) 0.024 (0.014) -0.011 (0.014) -0.054** (0.022) -0.091*** (0.035) 20000 0.073 Full set of birth cohort dummies (1988 birth omitted) -0.001 (0.052) 0.019 (0.032) 0.006 (0.021) 0.013 (0.014) -0.008 (0.014) -0.036* (0.021) -0.054 (0.034) 19919 0.132 0.016 (0.050) 0.019 (0.026) 0.007 (0.017) 0.017* (0.010) 0.010 (0.010) -0.007 (0.018) -0.013 (0.029) 16193 0.123 0.078 (0.049) 0.031 (0.026) 0.020 (0.017) 0.021** (0.010) 0.007 (0.010) -0.025 (0.018) -0.040 (0.029) 16193 0.140 20000 0.076 0.023 (0.054) -0.011 (0.033) 0.002 (0.022) 0.025 (0.014) -0.012 (0.014) -0.057** (0.022) -0.095*** (0.035) 20000 0.075 0.004 (0.053) 0.007 (0.032) 0.010 (0.021) 0.026 (0.014) -0.020* (0.014) -0.069*** (0.021) -0.090*** (0.035) 20000 0.063 -0.006 (0.004) Full set of age dummies (65 omit.) Full set of birth cohort dummies (1988 birth omitted) 16193 0.146 In the NLS72, ten questions were asked in 1979: 5 on traditional views and 5 on egalitarian attitudes Table 1. Average Agreement with Gender Role Attitudes by Labor Force Participation Variable Name FT196A FT196B FT196C FT196D FT196E FT196F FT196G FT196H FT196I FT196J EGAL TRAD A: National Longitudinal Survey of 1972 How do you feel about the following statements? Strongly agree, agree, disagree, or strongly disagree. (asked in 1979) a. A working mother of pre-school children can be just as good a mother as the woman who doesn’t work b. It is usually better for everyone involved if the man is the acheiver outside the home and the woman takes care of the home and family c. Young men should be encouraged to take jobs that are usuall filled by women (nursing, secretarial, work. etc.) d. Most women are just not interested in having big and important jobs e. Many qualified women can’t get good jobs; men with the same skills have much less trouble f. Most women are happiest when they are making a home and caring for children g. High schools counselors should urge young women to train for jobs which are now held mainly by men h. It is more important for a wife to help her husband than to have a career herself i. Schools teach women to want the less important jobs (reverse) j. Men should be given first chance at most jobs because they have the primary responsibility for providing for a family Composite of (FT196A+FT196C+FT196E+FT196G+FT196I) Composite of (FT196B+FT196D+FT196F+FT196H+FT196J) NonLFP in LFP in 1986 1986 ∆ 57.5 68.8 -11.3 46.7 35.0 11.7 46.0 48.0 -2.0 32.0 58.2 30.6 60.5 1.4 -2.3 44.6 37.9 6.7 54.3 56.9 -2.6 45.0 36.3 8.7 41.3 37.0 43.1 30.0 -1.8 7.0 51.5 40.4 55.5 20.5 -4.0 19.9 38 In the GSS, out of a total of eight questions on gender role attitudes, only four are asked in the 2000s Table 1. Average Agreement with Gender Role Attitudes by Labor Force Participation Variable Name FEPOL B: General Social Survey 1977-2006 Do you agree or disagree with this statement? Tell me if you agree or disagree with this statement: Most men are better suited emotionally for politics than are most women. (reverse) Now I'm going to read several more statements. As I read each one, please tell me whether you strongly agree, agree, disagree, or strongly disagree with it. For example, here is the statement: FECHLD A. A working mother can establish just as warm and secure a relationship with her children as a mother who does not work. FEPRESCH C. A preschool child is likely to suffer if his or her mother works FEFAM D. It is much better for everyone involved if the man is the achiever outside the home and the woman takes care of the home and family. EGAL Composite of (FEPOL+FECHLD) TRAD Composite of (FEFAM+FEPRESCH) Non-LFP LFP ∆ 67 77.3 -10.3 58.4 69.2 -10.8 50.3 47.9 41.1 35.7 9.2 12.2 62.3 49.1 73.0 38.5 -10.7 10.6 39 .3 .4 .5 .6 .7 .8 A. Changes over Time 1975 1980 <=1920 1946-52 1966-75 1985 1990 Year 1921-39 1953-58 1976-88 1995 2000 2005 1940-45 1959-65 All 41 .2 .3 .4 .5 .6 .7 .8 .9 1 A. Changes over Time 1975 1980 <=1920 1946-52 1966-75 1985 1990 Year 1921-39 1953-58 1976-88 1995 2000 2005 1940-45 1959-65 All 43 Mean LFP: 0.692 Time (1977=1) Time2/100 (1) 0.015*** (0.003) -0.025*** (0.008) Traditional attitudes (2) 0.010*** (0.001) -0.019*** (0.007) (3) 0.007* (0.004) -0.011 (0.008) -0.246*** (0.022) Egalitarian attitudes (4) (5) 0.008** 0.006 (0.004) (0.004) -0.014* -0.010 (0.008) (0.008) -0.219*** (0.024) 0.135*** 0.051** (0.019) (0.021) FEFAMa (6) (7) Full set 0.007** of time (0.003) dummies -0.013* (1998 omit.)(0.008) -0.223*** (0.024) 0.047** (0.021) -0.174*** (0.020) FEPOLa FEPRESCHa FECHLDa Mother ever worked 0.027** 0.018* (0.011) (0.011) 0.027** 0.018 (0.011) (0.011) 0.020 (0.011) 0.020* (0.011) (8) (9) (10) 0.008** 0.008** 0.008*** (0.004) (0.004) (0.003) -0.014* -0.015* -0.014* (0.008) (0.008) (0.008) (11) 0.005* (0.003) -0.009 (0.008) -0.089*** (0.023) 0.024* -0.011 (0.013) (0.013) -0.185*** -0.113*** (0.019) (0.022) 0.189*** 0.125*** (0.018) (0.021) 0.024** 0.020* 0.022** 0.018* (0.011) (0.011) (0.011) (0.011) Note: Parameter estimates from Linear Probability Model. Robust standard errors are in parentheses. Significance at 1% level denoted by ***, at 5% level denoted by **, at 10% level denoted by *. Columns (2)-(11) also includes years of schooling, number of children, everworking, living in an intact family and dummies (9) for religion at age 16, and region dummies (8). 45 At -0.246 (0.022), the impact of traditional attitudes implies that the slight rise [0.028 (0.011)] in average traditional attitudes from 0.373 in 1994 to 0.401 in 2006 would account for a [(0.401-0.373)*-0.246*100] 0.7 percentage point decline in FLP. By comparison, the increase in years of schooling from 13.04 in 1994 to 13.12 in 2006 would have lead to an increase 0.18 percentage point. Importantly, the introduction of traditional attitudes reduces the magnitude of quadratic term of the time trend from 0.018 to -0.011, rendering it insignificant and comparable in point estimate to that of men. 46 0 .2 .4 .6 .8 A. Attitude Sample - Table 4 1980 Actual Trend Col.1 1985 1990 Year Pred. Col.1 Trend Col.2 1995 Pred. Col.2 Trend Col.5 2000 2005 Pred. Col.5 Male Trend Mean LFP in 1986 at age 32: 0.728 Traditional attitudes in 1979a Egalitarian attitudes in 1979a LFP in 1979: 0.734 (1) (2) -0.294*** (0.034) (3) (4) (5) (6) -0.278*** -0.232*** (0.037) (0.040) 0.173*** 0.049 0.133*** (0.045) (0.044) (0.044) 0.180*** 0.189*** (0.016) (0.016) Note: Parameter estimates from Linear Probability Model. Robust standard errors are in parentheses. Significance at 1% level denoted by ***, 5% level denoted by **, at 10% level denoted by *. Included in the regressions are number of children, dummies for education levels (6), married, ever divorced, mother ever worked, presence of pre-school children, white, living in an intact family and dummies (6) for religion while growing up and region dummies (3). 48 Sample Mean LFP Time (1977=1) Time2/100 (1) GSS 0.589 0.020*** (0.003) -0.038*** (0.007) Traditional attitudes Egalitarian attitudes Ever divorced Presence of Pre-school children Mother ever worked Log of other income Observations R-squared 10612 0.071 (2) (3) (4) GSS-Attitude Sample 0.655 0.016*** 0.014*** 0.009* (0.004) (0.004) (0.005) -0.036*** -0.033*** -0.022** (0.011) (0.011) (0.010) -0.297*** (0.032) 0.048 (0.030) (0.003) (0.003) (0.003) 0.014 0.014 0.006 (0.018) (0.018) (0.018) -0.173*** -0.173*** -0.174*** (0.019) (0.019) (0.019) 0.018 0.018 0.009 (0.016) (0.016) (0.011) -0.080*** -0.078*** (0.009) (0.009) 3967 3967 3967 0.140 0.158 0.183 (5) (6) NLS72 0.695 -0.260*** (0.058) 0.088 (0.062) 0.054*** (0.021) -0.102*** (0.020) 0.025 (0.018) -0.089*** (0.013) 3469 0.113 0.040* (0.021) -0.105*** (0.019) 0.022 (0.018) -0.091*** (0.013) 3469 0.112 Note: Parameter estimates from Linear Probability Model. Robust standard errors are in parentheses. Significance at 1% level denoted by ***, 5% level denoted by **, at 10% level denoted by *. All columns also include number of children, dummies for white, living in an intact family and dummies for religion while growing up and region dummies as in Appendix Table A1. 49 In the GSS, questions about gender role attitudes and labor market decisions are asked contemporaneously, this raises the issue of possible biases associated with the avoidance of cognitive dissonance (Akerlof and Dickens, 1982). Letting A ( A, A ), P (1) can be rewritten as r 2 r (t , t 2 ), Cr (C1 ,..., C8 ) , equation Yi G i β g A ir β a Pir β p Cir β c X i β x i , Cognitive dissonance generates an errors-in-variables, G G * where G * denote the true attitudes. In the classical case, E ( G * ) 0 , this would lead to an attenuation bias in β . ( 2) g 50 This issue is addressed using a double prong instrumental variables strategy with some instruments coming from the GSS and another instrument from a TS2SLS. The instruments from the GSS are answers to questions about the respondents’ political views and attitudes towards sexual relations. These variables are correlated with gender role attitudes and are thought to impact labor market decisions only through attitudes toward whether women should work outside the home or not. 51 The exact questions are: There's been a lot of discussion about the way morals and attitudes about sex are changing in this country. If a man and woman have sex relations before marriage, do you think it is always wrong, almost always wrong, wrong only sometimes, or not wrong at all? [VAR:PREMARSX] “We hear a lot of talk these days about liberals and conservatives. I'm going to show you a seven point scale on which the political views that people might hold are arranged from extremely liberal point 1 to extremely conservative point 7? Where would you place yourself on this scale?” [VAR:POLVIEWS] 52 Risky .2 .3 .4 .5 .6 .7 A. Changes over Time 1975 1980 <=1920 1946-52 1966-75 1985 1990 Year 1921-39 1953-58 1976-88 1995 2000 2005 1940-45 1959-65 All 53 .4 .5 .6 C. Changes over Time 1975 1980 <=1920 1946-52 1966-75 1985 1990 Year 1921-39 1953-58 1976-88 1995 2000 2005 1940-45 1959-65 All 55 Econometric Specification Dependent variable: Time Time2 /10 (1) (2) (3) (4) First-Stage Traditional attitudes -0.017*** (0.002) 0.030*** (0.003) 2SLS First-Stage Egalitarian attitudes 0.027*** (0.004) -0.045*** (0.008) 2SLS Traditional attitudes Egalitarian attitudes Instruments: Premarital sex 0.131*** wrong (0.006) Liberal political -0.137*** views (0.011) R-squared 0.17 F-Test on instruments/ Anderson canon corr. 386.6 Sargan/ Anderson-Rudin Overid : p-value FLP 0.007* (0.004) -0.011 (0.007) -0.231*** (0.070) FLP 0.006* (0.003) -0.010 (0.008) 0.257*** (0.088) -0.107*** (0.009) 0.147*** (0.015) 0.14 744.6 0.001 0.9795 0.13 111.4 220.7 0.074 0.7854 Note: Robust standard errors are in parentheses. Significance at 1% level denoted by ***, at 5% level denoted by **, at 10% level denoted by *. Quadratic in age and cohorts dummies included , as wells years of schooling, number of children, dummies for white, married, ever divorced, preschooler present, mother everworking, living in an intact family, and dummies (9) for religion at age 16, and region dummies (8). 57 In the usual case (Angrist and Krueger, 1991; Inoue and Solon, 2005) , both samples contain on the instruments. Here the instrument is available only in sample 2. So an estimate of the instrument in sample 1 has to be constructed. Let Z be the excluded instrument not available in sample 1 1 and let X1 ( A r , P r , Cr ) From each cross-section of sample 2 (the NHIS), I estimate Z 2it A i Γ 2t 2it , t 1,...T where A i is a M-vector of age dummies. Stacking the estimates results in a M Χ T matrix Γ̂ 2 . 58 An estimate of the subjective risk of HIV/AIDS is constructed as ˆ P Zˆ1it Ai Γ 2 where P is a T-vector of time dummies, assuming as in Inoue and Solon that Γ1 Γ2 . Writing the linear projection of included instruments Zˆ1i : Zˆ1*i X1 ˆ Zˆ Zˆ * then net out the included , the residuals R i 1i 1i onto X1 instruments, and can be used as excluded instruments to identify β p and β g . In other words, the instrumentation relies on the fact that women of different ages at different time periods evaluated their chances of getting HIV/AIDS differently. 59 (1) (2) (3) (4) (5) Econometric Specification LPM LPM LPM LPM LPM Dependent Variable: MLP FLP FLP FLP FLP 0.019*** (0.008) -0.040*** (0.016) 0.039*** (0.011) -0.067*** (0.020) 0.031*** (0.010) -0.056*** (0.020) 0.029*** (0.011) -0.052*** (0.019) -0.196*** (0.027) 0.048* (0.025) 0.029*** (0.010) -0.052*** (0.019) Time Time2/10 Traditional attitudes Egalitarian attitudes Instruments: Liberal political views Estimated HIV/AIDS risk residual R-squared 0.14 F-Test on instruments Sargan/ Anderson-Rudin Overid : p-value 0.05 0.12 0.12 (6) (7) (8) (9) First-Stage TS2SLS First-Stage TS2SLS Egalitarian Egalitarian attitudes FLP attitudes FLP 0.007*** (0.002) -0.031*** (0.010) 0.122*** (0.022) 0.010 (0.007) -0.042 (0.031) 0.009*** (0.006) -0.035*** (0.010) 0.415 (0.829) 0.313** (0.133) -0.469** 0.172*** (0.014) -0.469** (0.234) (0.231) 0.13 0.1 4.02 0.010*** (0.004) -0.046*** (0.018) 78.56 0.12 156.3 0.016 0.8996 Note: Standard errors are in parentheses. Significance at 1% level denoted by ***, at 5% level denoted by **, at 10% level denoted by *. Quadratic in age and cohorts dummies included , as well as years of schooling, number of children, dummies for white, married, ever divorced, preschooler present, mother everworking, living in an intact family, and dummies (9) for religion at age 16, and region dummies (8). 60 .2 .4 .6 .8 B. Post-1988 Sample - Table 7 1990 1995 2000 2005 Year Actual Trend Col. 2 Pred. Col. 2 Trend Col. 3 Pred. Col. 3 Trend Col. 9 Pred. Col. 9 Male Trend 61 I explored various alternative hypotheses 1) Increase in divorce rates and attitudes toward divorce 2) Social conservatism (attitudes toward premarital sex) 3) Political conservatism 4) Increased religiosity (church attendance, e.g. Glaeser and Sacerdote, 2007; bible inerrancy also tested, e.g. Sherkat, 2000) 5) Cultural background (42 dummies on ethnic ancestry, e.g. Fernandez and Fogli (2005), Zaiceva and Zimmerman (2007)) 6) Increasing rates of ill-health (morbid obesity, e.g. Cawley, 2004) 62 .4 .5 .6 .7 C. Changes over Time 1975 1980 <=1920 1946-52 1966-75 1985 1990 Year 1921-39 1953-58 1976-88 1995 2000 2005 1940-45 1959-65 All 65 Time (1977=1) Time2/100 (1) (2) (3) (4) 0.011*** (0.004) -0.018** (0.008) 0.006 (0.004) -0.010 (0.008) -0.220*** (0.024) 0.052** (0.022) 0.018 (0.012) -0.007 (0.024) 0.010*** (0.003) -0.018** (0.008) 0.006 (0.004) -0.010 (0.008) -0.221*** (0.025) 0.052** (0.022) 0.019 (0.012) Traditional attitudes Egalitarian attitudes Ever divorced Liberal political views Premarital sex wrong Divorce should be easier Church attendance Self-rated health Current religion (9 dummies) 0.025** (0.012) 0.046* (0.023) 0.022* (0.012) (5) (6) 0.009** 0.005 (0.003) (0.004) -0.016** -0.007 (0.008) (0.008) -0.226*** (0.025) 0.058** (0.022) 0.034*** 0.028** (0.012) (0.012) (7) (8) 0.010*** (0.003) -0.019** (0.008) 0.026** (0.012) 0.006 (0.004) -0.011 (0.008) -0.220*** (0.025) 0.053** (0.022) 0.021* (0.012) 0.020 (0.017) 0.048*** (0.017) (9) 0.011** (0.004) -0.018* (0.009) 0.032** (0.015) (10) 0.007 (0.004) -0.011 (0.009) -0.234*** (0.030) 0.026 (0.027) 0.025* (0.015) -0.035*** 0.005 (0.013) (0.014) -0.026* (0.013) -0.044*** (0.013) 0.187*** 0.187*** (0.026) (0.026) No No No No No No Yes Yes No No Note: Parameter estimates from Linear Probability Model. Robust standard errors are in parentheses. Significance at 1% level denoted by ***, at 5% level denoted by **, at 10% level denoted by *. All columns also include years of schooling, number of children, dummies for white, married, preschooler present, mother everworking, living in an intact family and dummies (9) for religion at age 16, and region dummies (8). 67 (1) GSS 1977-2006 Mean LFP: Time 2 Time /10 (2) All Women 0.657 0.009* 0.007 0.005 (0.005) (0.006) (0.006) -0.023 -0.014 -0.008 (0.013) (0.016) (0.016) Traditional attitudes Egalitarian attitudes Self-rated health Married (3) 0.031 (0.022) 0.289*** (0.039) 0.020 (0.022) (4) 0.003 (0.008) -0.015 (0.019) 0.023 (0.029) (5) (6) Attitude Sample 0.693 0.003 0.000 (0.008) (0.010) -0.010 -0.003 (0.019) (0.023) -0.175*** (0.065) 0.056 (0.054) 0.275*** (0.065) 0.030 -0.008 (0.029) (0.036) (7) -0.003 (0.010) 0.004 (0.023) -0.303*** (0.079) -0.077 (0.065) 0.273*** (0.064) 0.000 (0.025) Note: Parameter estimates from Linear Probability Model. Robust standard errors are in parentheses. Significance at 1% level denoted by ***, at 5% level denoted by **, at 10% level denoted by *. Other demographics include number of children, dummies for living in an intact family and dummies (9) for religion at age 16, and region dummies (8). 68 The possible cascade of effects from the HIV/AIDS crisis, to more negative attitudes towards premarital sex and less favorable egalitarian attitudes, to lower level of FLP would appear to have several weaker links among African-American women. Importantly, they are less likely to be married, Data from the 2006 March CPS reveals that 60 percent of white women are currently married, whereas the percentage among black women is 30 percent. In the GSS, the number are 62 percent for white women and 34 percent for black women in 2006, There appears to be a strong link between health and the evolution of FLP among black women, although it is not clear which health factor is at play here given the self-reported nature of the variable. 69 Risky .4 .5 .6 .7 .8 D. Changes over the Life-Cycle 20 25 <=1920 1946-52 1966-75 30 35 40 Age 1921-39 1953-58 1976-88 45 50 55 60 65 1940-45 1959-65 All 70 Risky .4 .5 .6 .7 .8 B. Changes over the Life-Cycle 20 25 <=1920 1946-52 1966-75 30 35 40 Age 1921-39 1953-58 1976-88 45 50 55 60 65 1940-45 1959-65 All 72 Explanatory variables: Econometric Specification A: All Women LPM 2SLS 2SLS/TS2SLSa (1) (2) Time Time /100 2 (3) Traditional attitudes (4) Egalitarian attitudes (5) (6) Time Time /100 Years: 1977-2006 0.006 (0.004) 0.007* (0.004) 0.006* (0.003) -0.010 (0.008) -0.011 (0.007) -0.010 (0.008) B: Less than College Women LPM 0.004 -0.005 (0.003) (0.009) 2SLS 0.004 -0.005 (0.003) (0.008) a 2SLS/TS2SLS 0.003 -0.004 (0.003) (0.009) -0.219*** (0.024) -0.231*** (0.070) (8) Egalitarian attitudes Years: 1988-2006 0.051** (0.021) 0.257*** (0.088) -0.223*** (0.029) -0.266*** (0.087) 2 (7) Traditional attitudes 0.043* (0.024) 0.301*** (0.099) 0.029*** (0.011) 0.012*** (0.004) 0.010*** (0.004) -0.052*** (0.019) -0.054*** (0.017) -0.046*** (0.018) -0.196*** (0.027) -0.239*** (0.083) 0.019* (0.010) 0.020* (0.010) 0.007 (0.006) -0.044* (0.023) -0.045* (0.009) -0.033 (0.025) -0.189*** (0.033) -0.283** (0.112) 0.048* (0.025) 0.313** (0.133) 0.044 (0.029) 0.451** (0.208) 75 (1) (2) (3) Traditional 2 Time /100 attitudes Explanatory Time variables: Econometric Years: 1977-2006 Specification C: All Married Women LPM 0.009* -0.022** -0.297*** (0.005) (0.010) (0.032) 2SLS 0.010*** -0.023** -0.306*** (0.005) (0.011) (0.103) 2SLS/TS2SLSa 0.009* -0.022** (0.005) (0.011) D: College-Educated Married Women LPM 0.010 -0.019 -0.350*** (0.009) (0.020) (0.065) 2SLS 0.010 -0.019 -0.444** (0.009) (0.020) (0.204) 2SLS/TS2SLSa 0.009 -0.019 (0.009) (0.020) (4) Egalitarian attitudes (5) (6) Time Time /100 2 (7) Traditional attitudes (8) Egalitarian attitudes Years: 1988-2006 0.048 (0.030) 0.297 *** (0.100) 0.074 (0.062) 0.390** (0.194) 0.012* (0.007) 0.013* (0.007) 0.013* (0.006) -0.061** (0.030) -0.063** (0.030) -0.054** (0.026) -0.263*** (0.041) -0.253** (0.121) 0.030** (0.013) 0.031** (0.013) 0.027** (0.014) -0.118** (0.053) -0.120** (0.052) -0.101* (0.055) -0.382*** (0.070) -0.527** (0.231) 0.057 (0.036) 0.265* (0.158) 0.061 (0.068) 0.499* (0.281) 76 Explanatory variables: Econometric Specification E: All Men LPM (1) (2) Time Time /100 2 (3) Traditional attitudes (4) Egalitarian attitudes (5) (6) Time Time /100 Years: 1977-2006 0.007*** -0.013** (0.003) (0.006) LPM 0.007*** -0.013** (0.003) (0.006) LPM 0.007*** -0.013** (0.003) (0.006) F: College-Educated Single Women LPM 0.004 -0.006 (0.010) (0.024) 2SLS 0.005 -0.009 (0.010) (0.024) 2SLS/TS2SLSa 0.003 -0.003 (0.010) (0.023) 0.028 (0.021) 0.039** (0.018) (8) Egalitarian attitudes Years: 1988-2006 -0.019 (0.016) -0.028** (0.014) 0.019 (0.056) -0.213 (0.137) 2 (7) Traditional attitudes 0.067 (0.066) 0.210 (0.168) 0.021*** (0.007) 0.021*** (0.007) 0.021*** (0.007) -0.043*** (0.016) -0.043*** (0.016) -0.043*** (0.016) 0.024 (0.023) 0.037* (0.020) 0.003 (0.009) 0.003 (0.009) 0.002 (0.009) -0.010 (0.040) -0.008 (0.040) -0.000 (0.040) -0.019 (0.054) -0.211 (0.141) -0.023 (0.019) -0.032* (0.016) -0.026 (0.068) 0.183 (0.174) 77 This paper provides compelling evidence that beliefs about gender roles are an essential element of the analysis of the evolution of FLP over the latter part of the twentieth century. Gender role attitudes, whose secular trends reversed in the mid1990s when the AIDS crisis peaked, are found to explain the recent leveling-off in FLP, while general cultural trends towards more conservative social, religious and political views do not. While this paper solves the puzzle of the gender differences of the evolution of LP, it seemingly seems to open another one. How are gender role attitudes formed? How can their evolution be explained? 78 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 1972 1974 1976 1978 1982 1984 Women Very happy Women Not too happy 1986 1988 1990 1993 1996 2000 2004 Men Very happy Men Not too happy 79 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 1972 1974 1976 1978 1982 1984 1986 1988 1990 1993 1996 LFP Very happy LFP Not too happy NLFP Very happy NLFP Not too happy 2000 2004 80 81 2000 1800 1600 1400 1200 1000 Total 800 Females 600 Males 400 200 0 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 Year of Diagnostic of AIDS Source: Public Health Agency Canada 82 Source: Mosher et al. (2004) “Use of Contraception and Use of Family Planning Services in the United States: 1982-2002. Advance Data No. 350 83 √ 0 .05 .1 .15 .2 C. Changes over Time 1975 1980 <=1920 1946-52 1966-75 1985 1990 Year 1921-39 1953-58 1976-88 1995 2000 2005 1940-45 1959-65 84 √ 0 .05 .1 .15 .2 D. Changes over the Life-Cycle 20 25 <=1920 1946-52 1966-75 30 35 40 Age 1921-39 1953-58 1976-88 45 50 55 60 65 1940-45 1959-65 All 85 0 .1 .2 .3 .4 D. Changes over the Life-Cycle 20 25 <=1920 1946-52 1966-75 30 35 40 Age 1921-35 1953-58 1976-86 45 50 55 60 65 1936-45 1959-65 All 86 0 .1 .2 .3 .4 B. Changes over Time 1975 <=1920 1946-52 1966-75 1980 1985 1990 Year 1921-35 1953-58 1976-86 1995 2000 2005 1936-45 1959-65 All 87 .3 .4 .5 .6 .7 .8 A. Changes over Time 1975 1980 <=1920 1946-52 1966-75 1985 1990 Year 1921-39 1953-58 1976-88 1995 2000 2005 1940-45 1959-65 All 88 .4 .5 .6 .7 .8 C. Changes over Time 1975 1980 <=1920 1946-52 1966-75 1985 1990 Year 1921-39 1953-58 1976-88 1995 2000 2005 1940-45 1959-65 All 89 .2 .3 .4 .5 .6 .7 GSS cohorts vs. NLSY79 cohort 1980 1985 1990 1995 2000 2005 Year <1935 1952-56 1976-86 1936-45 1957-65 All GSS 1946-51 1966-75 NLSY79 Note: Averages of strong agreement (1), agreement (2/3), disagreement (1/3), strong disagreement (0) of “It is much better (for everyone involved) if the man is the achiever outside the home and the woman takes care of the home and family” 90