Education and the Intergenerational Transmission of Violence (Appendix)

Appendix A List of Variables Outcome Variables: • Years of schooling: Number of years of school that the respondent completed. • Completed junior high school: A dummy variable equal to one if the respondent completed junior high school or above (i.e., completed at least 8 years of schooling). • Completed high school: A dummy variable equal to one if the respondent completed high school or above (i.e., completed at least 11 years of schooling). • Completed primary school: A dummy variable equal to one if the respondent completed primary school or above (i.e., completed at least 5 years of schooling). • Child abuse: A dummy variable equal to one if the respondent has ever hit or physically abused her children. • Frequent child abuse: A dummy variable equal to one if the respondent has hit or physically abused her children frequently, e.g., a number of times, or many times. • Men can beat their partners in certain situations: A dummy variable equal to one if the respondent agrees with the statement that men can beat their partners in certain situations. • It may be necessary to beat children for discipline: A dummy variable equal to one if the respondent agrees with the statement that it may be necessary to beat children for discipline. • Age at first pregnancy: The age of the respondent during her first pregnancy. • Number of children: The number of children that the respondent has. • Employed: A dummy variable equal to one if the respondent was employed last week. • Employed in services: A dummy variable equal to one if the respondent was employed in services last week. • Social security: A dummy variable equal to one if the respondent had social security benefits from her job last week. • Personal income index: A z-score constructed by averaging the z-scores of the income dummy variables, which are calculated by using the mean and standard deviation of the variable. These dummy variables take the value of one if the respondent earns a personal income from the following six sources: rent from owning land, rent from owning a house, income from owning a company or workplace, income from owning a vehicle, having money in the bank, and income from other asset ownership. • Asset ownership index: A z-score constructed by averaging the z-scores of the asset ownership dummy variables, which are calculated by using the mean and standard deviation of the variable. These dummy variables take the value of one if the respondent’s household owns the asset. The following assets are included: refrigerator, deep freezer, gas/electric oven, microwave oven, dishwasher, garbage dispenser, washing machine, drying machine, iron, vacuum cleaner, plasma TV (LCD), home theater, television, satellite TV, paid TV service, DVD/VCD player, cellphone, non-mobile telephone, laptop/tablet computer, desktop computer, internet, air conditioner, car, taxi/mini-bus/bus or other commercial vehicles, and tractor. • Partner’s years of schooling: Number of years of school completed by the respondent’s partner. • Partner’s age: The age of the respondent’s partner. • Partner’s religiosity index: A z-score calculated as an average of z-scores of partners’ characteristics, including a dummy variable that takes the value of one if the partner never drinks alcoholic beverages, a dummy variable that takes the value of one if the partner never gambles, a dummy variable that takes the value of one if the partner never uses narcotic drugs, and a dummy variable that takes the value of one of the partner never had an affair. • Marriage age: The age of the respondent at the time of her first marriage. • Marriage decision: A dummy variable equal to one if the respondent decided on marriage together with her husband instead of the decision being made by her or his family. • Divorced: A dummy variable equal to one if the respondent has ever divorced. • Physical violence index: A z-score constructed by averaging the z-scores from each of the 6 physical violence indicators, including dummy variables that equal one if the respondent reports that she experienced intimate partner violence acts of (i) slapping or throwing an object that would hurt; (ii) pushing, shoving, or pulling hair; (iii) hitting with his fist or in a way that hurts; (iv) kicking, pushing on the ground, or beating; and (v) choking or burning. • Psychological violence index: A z-score constructed by averaging the z-scores from each of the following indicators, including dummy variables that equal one if the respondent reports that she experienced intimate partner violence acts of (i) insulting, (ii) humiliating, (iii) scaring or threatening, (iv) attempting to isolate her from her friends, (v) attempting to prevent contact with her family, (vi) insisting on knowing her location, (vii) ignoring her, (viii) becoming angry if she speaks to other men, (ix) suspecting that she is cheating on him, (x) wanting his permission before she seeks healthcare, and (xi) intervening in her clothing choices. • Financial control index: A z-score constructed by averaging the z-scores from two of the financial control behaviors, including dummy variables that equal one if the respondent reports that she experienced the following behaviors from her intimate partner: (i) taking income from her despite her disapproval and (ii) refusing to give her money for household spending. • Somatic depression index: A z-score calculated by averaging the z-scores from each of the 4 somatic depression indicators, including dummy variables equal to one if the respondent reports that she experienced the following within the last four weeks: (i) frequent headaches, (ii) trembling hands, (iii) digestion problems, and (iv) heartburn or other stomach problems. • Nonsomatic depression index: A z-score calculated by averaging the z-scores from each of the 16 nonsomatic depression indicators, including dummy variables equal to one if the respondent reports that she experienced the following within the last four weeks: (i) appetite loss, (ii) trouble sleeping, (iii) felt easily frightened from several things, (iv) felt anxious or nervous, (v) had trouble in thinking clearly, (vi) felt unhappy, (vii) cried more often, (viii) did not enjoy daily activities, (ix) had difficulty making decisions, (x) delayed daily activities, (xi) felt useless, (xii) lost interest in activities that she previously enjoyed, (xiii) felt worthless, (xiv) thought about suicide, (xv) felt tired all the time, and (xvi) got tired easily. • Overall depression index: A z-score calculated by averaging the z-scores from 20 depression indicators, including 4 somatic and 16 nonsomatic depression indicators, as listed above. • Childhood region, rural: A dummy variable equal to one if the respondent lived in a rural village or district until she was 12 years old. • Childhood region, urban: A dummy variable equal to one if the respondent lived in an urban area until she was 12 years old. Covariates: • Non-Turkish Speaker: A dummy variable equal to one if the respondent speaks a non-Turkish language as her primary language. • Region dummies: Dummy variables for each of the 12 regions where the respondents lived until they were 12 years old. • Childhood violence: A dummy variable equal to one if the respondent experienced physical or sexual violence from her own family after age of 15. Outcome Variables in Appendix B: • Childhood violence intensity: A dummy variable equal to one if the respondent experienced violence from her own family often during childhood. • Childhood violence (overall): A dummy variable equal to one if the respondent experienced violence from her own family or others such as teachers, strangers, etc. during childhood. • Childhood violence intensity (overall): A dummy variable equal to one if the respondent experienced violence from her own family or others such as teachers, strangers, etc. often during childhood. • Home violence: A dummy variable equal to one if the respondent witnessed her mother experiencing domestic violence from her husband. Appendix B Additional Figures and Tables – For Online Publication Figure A1: Percentage of Children Aged 2 to 14 Years Who Experienced Any Violent Discipline (Psychological Aggression and/or Physical Punishment) in the Past Month Note: Data are from UNICEF global databases, 2016, based on Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS) (2005-2015), accessed from https://data.unicef.org/topic/child-protection/violence/violent-discipline/ on June 20, 2017. Table A1: Effects of the Reform on Education by Childhood Violence Years of schooling Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations Completed junior high school Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations Overall sample Rural sample (1) RF (2) RF 0.662* (0.336)† 0.608 (0.544) -0.544 (0.371) 8.48 89 2,377 0.979** (0.467)†† 0.904 (0.683) -0.560 (0.445) 7.43 89 1,382 0.180*** (0.033)††† 0.081 (0.054) -0.037 (0.040) 0.60 118 3,162 0.248*** (0.050)††† 0.113 (0.070) -0.042 (0.054) 0.49 118 1,832 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes all women. The optimal bandwidth is estimated by using the Imbens and Kalyanaraman (2009) algorithm. The RD treatment effects of the reform are reported by exposure to childhood violence, i.e., whether the respondent experienced violence from her own family members during her childhood. Columns 1 and 2 report reduced-form RD treatment effects of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity for the overall sample and the subsample of women whose childhood region is rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level. (Based on p-values unadjusted for multiple-hypothesis testing.) ††† Significant at the 1 percent level. †† Significant at the 5 percent level. † Significant at the 10 percent level. (Based on p-values adjusted for multiple-hypothesis testing using Simes adjustment.) Table A2: Effects of the Reform on Childhood Violence and Having Children Overall sample Childhood violence Schooling Mean Bandwidth Observations Childhood violence intensity Schooling Mean Bandwidth Observations Childhood violence (overall) Schooling Mean Bandwidth Observations Childhood violence intensity (overall) Schooling Mean Bandwidth Observations Number of children Schooling Mean Bandwidth Observations Has children Schooling Mean Bandwidth Observations Rural sample (1) OLS (2) RF (3) IV (4) OLS (5) RF (6) IV -0.002 (0.002) 0.15 95 2,526 -0.004 (0.032) 0.15 95 2,526 -0.006 (0.046) 0.15 95 2,526 -0.001 (0.003) 0.14 95 1,471 -0.034 (0.039) 0.14 95 1,471 -0.036 (0.044) 0.14 95 1,471 -0.004** (0.002)†† 0.08 91 2,423 -0.001 (0.021) 0.08 91 2,423 -0.002 (0.03) 0.08 91 2,423 -0.005** (0.002)† 0.09 91 1,408 -0.006 (0.028) 0.09 91 1,408 -0.005 (0.025) 0.09 91 1,408 -0.001 (0.003) 0.18 91 2573 -0.026 (0.032) 0.18 91 2573 -0.041 (0.052) 0.18 91 2573 -0.001 (0.003) 0.17 91 1494 -0.061 (0.04) 0.17 91 1494 -0.061 (0.047) 0.17 91 1494 -0.001 (0.003) 0 91 2573 -0.026 (0.032) 0 91 2573 -0.041 (0.052) 0 91 2573 -0.001 (0.003) 0 91 1494 -0.061 (0.04) 0 91 1494 -0.061 (0.047) 0 91 1494 -0.088*** (0.006)††† 1.99 86 1806 -0.007 (0.073) 1.99 86 1806 -0.017 (0.181) 1.99 86 1806 -0.094*** -0.019 (0.009)††† (0.103) 2.08 2.08 86 86 1112 1112 -0.018 (0.092) 2.08 86 1112 -0.047*** (0.003)††† 0.76 83 2,332 -0.06 (0.044) 0.76 83 2,332 -0.072 (0.052) 0.76 83 2,332 -0.039*** -0.057 (0.004)††† (0.047) 0.81 0.81 83 83 1,357 1,357 -0.049 (0.04) 0.81 83 1,357 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes all women. The optimal bandwidth is estimated by using the Imbens and Kalyanaraman (2009) algorithm. Column 1 reports OLS results using years of schooling as the independent variable for an optimal bandwidth ĥ estimated by the Imbens and Kalyanaraman algorithm. Columns 2 – 3 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity. Columns 1 – 3 report these results for the overall sample, and columns 4 – 6 report results from the same specifications for the subsample of respondents whose childhood region of residence was rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level. (Based on p-values unadjusted for multiple-hypothesis testing.) ††† Significant at the 1 percent level. †† Significant at the 5 percent level. † Significant at the 10 percent level. (Based on p-values adjusted for multiple-hypothesis testing using Simes adjustment.) Table A3: RD Treatment Effects on Schooling Outcomes (Static Bandwidth) Outcome Years of schooling Completed education: Junior high school High school Primary school (1) (2) (3) (4) (5) (6) Linear RD ĥ bandwidth Linear RD 0.75ĥ bandwidth Linear RD 1.5ĥ bandwidth Bandwidth N Mean 0.825** (0.335)†† 0.674* (0.369)† 0.981*** (0.274)††† 85 2,386 8.48 0.201*** (0.037)††† 0.092** (0.041)†† -0.018 (0.026) 0.172*** (0.04)††† 0.116*** (0.043)†† -0.032 (0.029) 0.186*** (0.03)††† 0.160*** (0.038)††† -0.017 (0.021) 85 2,386 0.59 85 2,386 0.40 85 2,386 0.91 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. All columns use a static bandwidth of 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood. Columns 1 – 3 report local RD regressions with linear polynomials in the month-year of birth using the static bandwidth ĥ, 0.75 ĥ and 1.5 ĥ, respectively. Column 5 reports the number of observations used in estimations, and column 6 reports the outcome mean within the static bandwidth. All results are reported for the full sample of women. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level. (Based on p-values unadjusted for multiple-hypothesis testing.) ††† Significant at the 1 percent level. †† Significant at the 5 percent level. † Significant at the 10 percent level. (Based on p-values adjusted for multiple-hypothesis testing using Simes adjustment.) Table A4: RD Treatment Effects on Schooling Outcomes Using a Quadratic Polynomial in the Forcing Variable Outcome Years of schooling Mean Bandwidth Observations Completed education: Junior high school Mean Bandwidth Observations High school Mean Bandwidth Observations Primary school Mean Bandwidth Observations (1) (2) Quadratic RD optimal bandwidth Quadratic RD static bandwidth 0.694** 0.303†† 8.48 89 2,492 0.808** (0.339)†† 8.48 85 2,386 1.191*** (0.031)††† 0.60 118 3,308 0.201*** (0.037)††† 0.59 85 2,386 0.123*** (0.045)†† 0.40 65 1,837 0.091** (0.042)†† 0.40 85 2,386 -0.022 (0.025) 0.91 93 2,630 -0.021 (0.026) 0.91 85 2,386 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. All results are reported for the full sample of women. Columns 1 and 2 report local RD regressions with quadratic polynomials in the month-year of birth using the optimal bandwidth estimated by the Imbens and Kalyanaraman (2009) algorithm, and the static bandwidth of 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood, respectively. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. *** Significant at the 1 percent level. ** Significant at the 5 percent level. * Significant at the 10 percent level. (Based on p-values unadjusted for multiple-hypothesis testing.) ††† Significant at the 1 percent level. †† Significant at the 5 percent level. † Significant at the 10 percent level. (Based on p-values adjusted for multiple-hypothesis testing using Simes adjustment.) Table A5: RD Treatment Effects on Schooling Outcomes by Childhood Region (Static Bandwidth) Rural childhood region Bandwidth: (1) ĥ (2) 0.75 ĥ 1.160** (0.456) 7.42 85 1,385 1.112** (0.518) 7.40 64 1,036 (3) 1.5 ĥ Urban childhood region (4) ĥ (5) 0.75 ĥ (6) 1.5 ĥ 1.307*** 0.523 (0.367) (0.468) 7.47 9.68 128 85 2,027 1,001 0.439 (0.507) 9.68 64 747 0.526 (0.399) 9.64 128 1,508 1.184*** -0.328 (0.452) (0.509) 6.70 8.55 128 85 1,504 684 -0.235 (0.478) 8.44 64 521 -0.671 (0.459) 8.52 128 933 Panel A: Sample of All Women Years of schooling Mean Bandwidth Observations Panel B: Sample of Women Who Have Children Years of schooling Mean Bandwidth Observations 1.151** (0.517) 6.73 85 1,100 1.103* (0.578) 6.81 64 847 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. Columns 1 – 3, and 4 – 6 report local RD regressions with linear polynomials in the month-year of birth using the static bandwidth ĥ, 0.75 ĥ, and 1.5 ĥ, respectively. The static bandwidth is 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood. The outcome mean, bandwidth, and observation numbers are reported in the rows under the dependent variables. Columns 1 – 3 report the results for the sample of women who grew up in a rural region, and columns 4 – 6 report them for the sample of women who grew up in an urban region. Panel A reports the results for the sample of all women, and Panel B reports them for the sample of women who have children. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively. Table A6: Effects of Education on Violence Against Children (Static Bandwidth) Overall sample Rural sample (1) OLS (2) RF (3) IV (4) OLS (5) RF (6) IV -0.020*** (0.003)††† 0.48 85 1,776 0.007 (0.047) 0.48 85 1,776 0.017 (0.109) 0.48 85 1,776 -0.026*** (0.005)††† 0.51 85 1,095 0.030 (0.069) 0.51 85 1,095 0.027 (0.062) 0.51 85 1,095 -0.020*** (0.004)††† 0.41 85 1,776 0.040 (0.048) 0.41 85 1,776 0.096 (0.134) 0.41 85 1,776 -0.021*** (0.005)††† 0.44 85 1,095 0.049 (0.075) 0.44 85 1,095 0.043 (0.069) 0.44 85 1,095 Panel A: RD Treatment Effects Child abuse Schooling Mean Bandwidth Observations Frequent child abuse Schooling Mean Bandwidth Observations Panel B: RD Treatment Effects by Exposure to Childhood Violence Child abuse Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations Frequent child abuse Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations -0.018*** (0.004)††† 0.003 (0.011) 0.186** (0.093)†† 0.48 85 1,711 0.039 (0.048) -0.081 (0.085) 0.243*** (0.049)††† 0.48 85 1,711 0.247 -0.025*** (0.508) (0.005)††† -0.257 0.002 (0.431) (0.018) 2.180 0.233* (3.311) (0.125)† 0.48 0.51 85 85 1,711 1,051 0.093 (0.071) -0.229** (0.106)†† 0.341*** (0.058)††† 0.51 85 1,051 0.099 (0.089) -0.154** (0.077)†† 1.347** (0.548)†† 0.51 85 1,051 -0.018*** (0.004)††† -0.007 (0.012) 0.249** (0.106)†† 0.41 85 1,711 0.072 (0.049) -0.104 (0.083) 0.241*** (0.049)††† 0.41 85 1,711 0.421 -0.020*** (0.812) (0.005)††† -0.407 -0.001 (0.680) (0.018) 3.322 0.263* (5.236) (0.135)† 0.41 0.44 85 85 1,711 1,051 0.114 (0.077) -0.292*** (0.106)†† 0.363*** (0.063)††† 0.44 85 1,051 0.121 (0.100) -0.194** (0.082)†† 1.629*** (0.589)†† 0.44 85 1,051 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes women who have children. All columns use a static bandwidth of 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood. Panel A reports the RD treatment effects of the reform, and Panel B reports them by exposure to childhood violence, i.e., whether the respondent experienced violence from her own family members during her childhood. Columns 1 reports OLS results using years of schooling as the independent variable. Columns 2 – 3 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity. Columns 1 – 3 report these results for the overall sample, and columns 4 – 6 report results from same specifications for the subsample of respondents whose childhood region of residence was rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values unadjusted for multiple-hypothesis testing). †††, ††, and † denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values adjusted for multiple-hypothesis testing using Simes adjustment). Table A7: Effects of Education on Violence Against Children (Overall and Home Violence) Overall sample (1) OLS (2) RF Rural sample (3) IV (4) OLS (5) RF (6) IV Panel A: RD Treatment Effects by Exposure to Childhood Violence (Overall) Child abuse Schooling Schooling × Childhood violence (overall) Childhood violence (overall) Frequent child abuse Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations -0.019*** (0.004)††† -0.010 (0.009) 0.267*** (0.073)††† 0.49 94 1,930 -0.018*** (0.003)††† -0.016* (0.009) 0.278*** (0.078)††† 0.42 106 2,129 0.026 (0.047) -0.071 (0.078) 0.220*** (0.043)††† 0.49 94 1,930 0.072 (0.044) -0.049 (0.075) 0.181*** (0.040)††† 0.42 106 2,129 0.819 -0.023*** (7.802) (0.005)††† -0.547 -0.006 (4.718) (0.014) 4.428 0.292*** (36.698) (0.096)††† 0.49 0.51 94 89 1,930 1,139 0.774 -0.020*** (2.834) (0.005)††† -0.530 -0.008 (1.902) (0.014) 4.246 0.285*** (14.655) (0.103)††† 0.42 0.44 106 92 2,129 1,163 0.080 (0.068) -0.195** (0.096)†† 0.312*** (0.048)††† 0.51 89 1,139 0.106 (0.073) -0.219** (0.096)†† 0.296*** (0.051)††† 0.44 92 1,163 0.094 (0.099) -0.134* (0.070)† 1.181** (0.492)†† 0.51 89 1,139 0.133 (0.125) -0.159** (0.075)† 1.334** (0.538)†† 0.44 92 1,163 -0.005 (0.047) 0.031 (0.059) 0.176*** (0.033)††† 0.49 94 1,878 0.038 (0.045) 0.052 (0.060) 0.143*** (0.034)††† 0.42 106 2,067 -0.020 (0.201) 0.059 (0.124) -0.264 (0.936) 0.49 94 1,878 0.104 (0.206) 0.103 (0.200) -0.627 (1.514) 0.42 106 2,067 0.039 (0.073) -0.019 (0.074) 0.247*** (0.039)††† 0.51 89 1,102 0.053 (0.076) 0.008 (0.078) 0.204*** (0.043)††† 0.44 92 1,126 0.044 (0.088) -0.023 (0.064) 0.392 (0.439) 0.51 89 1,102 0.065 (0.106) -0.013 (0.071) 0.287 (0.486) 0.44 92 1,126 Panel B: RD Treatment Effects by Exposure to Home Violence Child abuse Schooling Schooling × Home violence Home violence Frequent child abuse Mean Bandwidth Observations Schooling Schooling × Home violence Home violence Mean Bandwidth Observations -0.018*** (0.004)††† -0.011 (0.007) 0.266*** (0.058)††† 0.49 94 1,878 -0.019*** (0.004)††† -0.008 (0.007) 0.216*** (0.061)††† 0.42 106 2,067 -0.021*** (0.005)††† -0.017** (0.009)† 0.369*** (0.066)††† 0.51 89 1,102 -0.020*** (0.006) -0.010 (0.009)††† 0.280*** (0.078)††† 0.44 92 1,126 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes women who have children. The optimal bandwidth is estimated by using the Imbens and Kalyanaraman (2009) algorithm. Panel A reports the RD treatment effects of the reform by exposure to overall childhood violence, i.e., whether the respondent experienced violence from her own family members or others (teachers, strangers, etc.) during her childhood; and Panel B reports them by exposure to home violence, i.e., whether she witnessed violence against her mother during her childhood. Columns 1 reports OLS results using years of schooling as the independent variable for an optimal bandwidth ĥ estimated by the Imbens and Kalyanaraman algorithm. Columns 2 – 3 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity. Columns 1 – 3 report these results for the overall sample, and columns 4 – 6 report results from the same specifications for the subsample of respondents whose childhood region of residence was rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values unadjusted for multiple-hypothesis testing). †††, ††, and † denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values adjusted for multiple-hypothesis testing using Simes adjustment). Table A8: Effects of Education on Attitudes toward Violence Overall sample Rural sample (1) OLS (2) RF (3) IV (4) OLS (5) RF (6) IV -0.024*** (0.004)††† 0.38 83 1,651 0.024 (0.057) 0.38 83 1,651 0.040 (0.101) 0.38 83 1,651 -0.024*** (0.005)††† 0.41 88 1,080 0.023 (0.069) 0.41 88 1,080 0.021 (0.063) 0.41 88 1,080 -0.019*** (0.003)††† 0.28 107 2,131 0.031 (0.041) 0.28 107 2,131 0.104 (0.185) 0.28 107 2,131 -0.025*** (0.004)††† 0.29 99 1,236 0.022 (0.051) 0.29 99 1,236 0.022 (0.051) 0.29 99 1,236 Panel A: RD Treatment Effects Men can beat their partners in certain situations. Schooling Mean Bandwidth Observations It may be necessary to beat children for discipline. Schooling Mean Bandwidth Observations Panel B: RD Treatment Effects by Exposure to Childhood Violence (Static Bandwidth) Men can beat their partners in certain situations. Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations It may be necessary to beat children for discipline. Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations -0.024*** (0.004)††† -0.001 (0.011) 0.110 (0.088) 0.38 85 1,625 0.051 (0.054) -0.080 (0.091) 0.129*** (0.047)†† 0.38 85 1,625 -0.018*** (0.004)††† 0.001 (0.011) 0.042 (0.094) 0.29 85 1,712 0.045 (0.046) 0.027 (0.070) 0.048 (0.045) 0.29 85 1,712 0.161 -0.023*** (0.225) (0.005)††† -0.188 -0.012 (0.213) (0.013) 1.532 0.195* (1.622) (0.104) 0.38 0.41 85 85 1,625 998 0.175 (0.381) -0.110 (0.339) 0.918 (2.602) 0.29 85 1,712 -0.024*** (0.004)††† -0.005 (0.014) 0.105 (0.116) 0.30 85 1,052 0.060 0.056 (0.069) (0.068) -0.084 -0.067 (0.113) (0.068) 0.141** 0.587 (0.057)†† (0.464) 0.41 0.41 85 85 998 998 0.054 (0.051) 0.011 (0.091) 0.084 (0.060) 0.30 85 1,052 0.051 (0.057) -0.017 (0.056) 0.219 (0.400) 0.30 85 1,052 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes women who have children. Panel A reports the RD treatment effects of the reform, and Panel B reports them by exposure to childhood violence, i.e., whether the respondent experienced violence from her own family members during her childhood. Panel A uses an optimal bandwidth ĥ estimated by the Imbens and Kalyanaraman algorithm, and Panel B uses a static bandwidth of 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood. Columns 1 reports OLS results using years of schooling as the independent variable. Columns 2 – 3 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity. Columns 1 – 3 report these results for the overall sample, and columns 4 – 6 report results from same specifications for the subsample of respondents whose childhood region of residence was rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values unadjusted for multiple-hypothesis testing). †††, ††, and † denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values adjusted for multiple-hypothesis testing using Simes adjustment). Table A9: Effects of Education on Mental Health Outcomes Overall sample (1) OLS (2) RF -0.024*** (0.005)††† 0.02 77 1,644 -0.035*** (0.005)††† 0.00 106 2,122 -0.023*** (0.005)††† 0.03 75 1,574 -0.037 (0.046) 0.02 77 1,644 -0.019 (0.055) 0.00 106 2,122 -0.037 (0.050) 0.03 75 1,574 Rural sample (3) IV (4) OLS (5) RF (6) IV -0.060 (0.065) 0.03 115 1,412 -0.028 (0.071) 0.03 96 1,212 -0.032 (0.076) 0.04 95 1,201 -0.062 (0.071) 0.03 115 1,412 -0.031 (0.075) 0.03 96 1,212 -0.036 (0.085) 0.04 95 1,201 0.033 (0.075) -0.285** (0.123)† 0.379*** (0.066)††† 0.04 85 1,056 0.056 (0.079) -0.292* (0.151)† 0.303*** (0.088)††† 0.02 85 1,056 0.027 (0.080) -0.284** (0.134)† 0.398*** (0.070)††† 0.04 85 1,056 0.043 (0.080) -0.159** (0.080)† 1.382** (0.574)† 0.04 85 1,056 0.065 (0.091) -0.172* (0.104)† 1.400* (0.751)†† 0.02 85 1,056 0.038 (0.083) -0.155* (0.083)† 1.379** (0.594)†† 0.04 85 1,056 Panel A: RD Treatment Effects Overall depression index Schooling Somatic depression index Mean Bandwidth Observations Schooling Nonsomatic depression index Mean Bandwidth Observations Schooling Mean Bandwidth Observations -0.073 -0.023*** (0.106) (0.005)††† 0.02 0.03 77 115 1,644 1,412 -0.056 -0.033*** (0.162) (0.006)††† 0.00 0.03 106 96 2,122 1,212 -0.063 -0.020*** (0.096) (0.005)††† 0.03 0.04 75 95 1,574 1,201 Panel B: RD Treatment Effects by Exposure to Childhood Violence (Static Bandwidth) Overall depression index Schooling Schooling × Childhood violence Childhood violence Somatic depression index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Nonsomatic depression index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations -0.023*** (0.004)††† 0.006 (0.011) 0.282*** (0.094)††† 0.02 85 1,718 -0.033*** (0.006)††† 0.021 (0.015) 0.127 (0.117) 0.00 85 1,718 -0.021*** (0.005)††† 0.002 (0.012) 0.321*** (0.103)††† 0.02 85 1,718 0.015 0.165 -0.021*** (0.047) (0.351) (0.005)††† -0.140 -0.254 -0.007 (0.093) (0.335) (0.017) 0.375*** 2.242 0.323** (0.045)††† (2.558) (0.125)†† 0.02 0.02 0.04 85 85 85 1,718 1,718 1,056 0.036 0.353 -0.030*** (0.061) (0.679) (0.007)††† -0.275** -0.521 0.006 (0.114)† (0.622) (0.026) 0.380*** 4.216 0.142 (0.061)††† (4.775) (0.177) 0.00 0.00 0.02 85 85 85 1,718 1,718 1,056 0.010 0.119 -0.019*** (0.050) (0.299) (0.006)††† -0.106 -0.188 -0.011 (0.099) (0.295) (0.018) 0.374*** 1.752 0.368*** (0.049)††† (2.247) (0.132)†† 0.02 0.02 0.04 85 85 85 1,718 1,718 1,056 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes all women. Panel A reports the RD treatment effects of the reform, and Panel B reports them by exposure to childhood violence, i.e., whether the respondent experienced violence from her own family members during her childhood. Panel A uses an optimal bandwidth ĥ estimated by the Imbens and Kalyanaraman algorithm, and Panel B uses a static bandwidth of 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood. Columns 1 reports OLS results using years of schooling as the independent variable. Columns 2 – 3 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity. Columns 1 – 3 report these results for the overall sample, and columns 4 – 6 report results from same specifications for the subsample of respondents whose childhood region of residence was rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values unadjusted for multiple-hypothesis testing). †††, ††, and † denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values adjusted for multiple-hypothesis testing using Simes adjustment). Table A10: Effects of Education on Fertility Outcomes Overall sample Rural sample (1) OLS (2) RF (3) IV (4) OLS 0.441*** (0.025)††† 21.48 121 2,481 0.183 (0.253) 21.48 121 2,481 0.442 (0.600) 21.48 121 2,481 0.357*** (0.035)††† 21.14 106 1,388 0.723** 0.703* (0.335)† (0.373) 21.14 21.14 106 106 1,388 1,388 -0.144*** (0.007) 1.51 73 2,056 -0.113 (0.100)††† 1.51 73 2,056 -0.137 (0.112) 1.51 73 2,056 -0.147*** (0.009)††† 1.69 88 1,445 -0.149 (0.108) 1.69 88 1,445 (5) RF (6) IV Panel A: RD Treatment Effects Age at first pregnancy Schooling Mean Bandwidth Observations Number of children Schooling Mean Bandwidth Observations -0.137 (0.099) 1.69 88 1,445 Panel B: RD Treatment Effects by Exposure to Childhood Violence (Static Bandwidth) Age at first pregnancy Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations Number of children Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations 0.418*** (0.028)††† 0.029 (0.077) -0.496 (0.582) 21.34 85 1,801 0.364 (0.290) 0.911* (0.499) -0.519 (0.392) 21.34 85 1,801 0.443 0.359*** (1.235) (0.038)††† 0.606 -0.019 (1.498) (0.086) -4.789 0.086 (11.657) (0.613) 21.34 21.08 85 85 1,801 1,096 0.855** 0.770* (0.398)† (0.437) 1.107* 0.177 (0.621) (0.435) -0.179 -0.900 (0.476) (3.209) 21.08 21.08 85 85 1,096 1,096 -0.141*** (0.007)††† -0.029* (0.015) 0.137 (0.160) 1.50 85 2,274 -0.183* (0.096) -0.141 (0.136) 0.022 (0.098) 1.50 85 2,274 -0.229* (0.131) -0.007 (0.072) -0.032 (0.635) 1.50 85 2,274 -0.171 (0.119) -0.104 (0.207) -0.025 (0.129) 1.70 85 1,322 -0.142*** (0.010)††† -0.030 (0.024) 0.096 (0.219) 1.70 85 1,322 -0.169 (0.121) 0.025 (0.078) -0.286 (0.623) 1.70 85 1,322 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes all women. Panel A reports the RD treatment effects of the reform, and Panel B reports them by exposure to childhood violence, i.e., whether the respondent experienced violence from her own family members during her childhood. Panel A uses an optimal bandwidth ĥ estimated by the Imbens and Kalyanaraman algorithm, and Panel B uses a static bandwidth of 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood. Columns 1 reports OLS results using years of schooling as the independent variable. Columns 2 – 3 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity. Columns 1 – 3 report these results for the overall sample, and columns 4 – 6 report results from same specifications for the subsample of respondents whose childhood region of residence was rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values unadjusted for multiple-hypothesis testing). †††, ††, and † denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values adjusted for multiple-hypothesis testing using Simes adjustment). Table A11: Effects of Education on Labor Market Outcomes Overall sample (1) OLS (2) RF Rural sample (3) IV (4) OLS (5) RF (6) IV Panel A: RD Treatment Effects Employed Schooling Employed in services Mean Bandwidth Observations Schooling Social security Mean Bandwidth Observations Schooling Personal income index Mean Bandwidth Observations Schooling Asset ownership index Mean Bandwidth Observations Schooling Mean Bandwidth Observations 0.018*** (0.003)††† 0.19 96 1,960 0.023*** (0.003)††† 0.14 94 1,940 0.023*** (0.003)††† 0.11 81 1,723 0.023*** (0.003)††† -0.05 143 2,633 0.029*** (0.002) 0.03 84 1,762 0.002 0.009 (0.031) (0.141) 0.19 0.19 96 96 1,960 1,960 -0.015 -0.063 (0.028) (0.161) 0.14 0.14 94 94 1,940 1,940 0.008 0.018 (0.024) (0.052) 0.11 0.11 81 81 1,723 1,723 -0.050 -0.198 (0.038) (0.308) -0.05 -0.05 143 143 2,633 2,633 -0.004 -0.009 (0.021) (0.053) 0.03 0.03 84 84 1,762 1,762 0.012*** (0.004)††† 0.18 93 1,184 0.020*** (0.004)††† 0.11 88 1,141 0.020*** (0.004)††† 0.10 97 1,232 0.016*** (0.005)††† -0.10 89 1,145 0.030*** (0.002) 0.00 87 1,120 -0.010 -0.010 (0.036) (0.038) 0.18 0.18 93 93 1,184 1,184 -0.024 -0.023 (0.030) (0.032) 0.11 0.11 88 88 1,141 1,141 -0.011 -0.010 (0.027) (0.026) 0.10 0.10 97 97 1,232 1,232 -0.061 -0.061 (0.054) (0.065) -0.10 -0.10 89 89 1,145 1,145 0.028 0.024 (0.027) (0.021) 0.00 0.00 87 87 1,120 1,120 Panel B: RD Treatment Effects by Exposure to Childhood Violence (Static Bandwidth) Employed Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations 0.022*** (0.004)††† -0.002 (0.009) 0.055 (0.067) 0.19 85 1,718 0.025 0.127 (0.034) (0.234) -0.033 -0.127 (0.067) (0.219) 0.048 1.007 (0.043) (1.684) 0.19 0.19 85 85 1,718 1,718 0.015*** 0.021 0.022 (0.004)††† (0.039) (0.038) -0.006 -0.041 -0.030 (0.013) (0.081) (0.049) 0.093 0.073 0.270 (0.097) (0.056) (0.357) 0.18 0.18 0.18 85 85 85 1,056 1,056 1,056 Table A11: Effects of Education on Labor Market Outcomes, Cont’d Overall sample Employed in services Schooling Schooling × Childhood violence Childhood violence Social security Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Personal income index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Asset ownership index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations (1) OLS (2) RF 0.026*** (0.004)††† -0.003 (0.009) 0.050 (0.059) 0.14 85 1,718 0.025*** (0.003)††† -0.006 (0.008) 0.047 (0.053) 0.11 85 1,718 0.021*** (0.004)††† 0.009 (0.017) -0.091 (0.112) -0.08 85 1,718 0.030*** (0.002)††† -0.013** (0.006)†† 0.039 (0.039) 0.03 85 1,718 0.015 (0.028) -0.092* (0.055) 0.055 (0.041) 0.14 85 1,718 0.020 (0.026) -0.045 (0.043) 0.015 (0.035) 0.11 85 1,718 -0.060 (0.044) -0.061 (0.065) -0.008 (0.052) -0.08 85 1,718 -0.010 (0.021) 0.059 (0.038) -0.077*** (0.022) ††† 0.03 85 1,718 Rural sample (3) IV (4) OLS 0.132 0.022*** (0.236) (0.004)††† -0.184 -0.007 (0.215) (0.014) 1.417 0.077 (1.658) (0.094) 0.14 0.11 85 85 1,718 1,056 0.117 0.019*** (0.222) (0.004)††† -0.130 -0.004 (0.202) (0.012) 0.988 0.029 (1.559) (0.071) 0.11 0.09 85 85 1,718 1,056 -0.198 0.014*** (0.426) (0.005)††† 0.094 0.026 (0.376) (0.033) -0.778 -0.210 (2.895) (0.192) -0.08 -0.10 85 85 1,718 1,056 -0.086 0.030*** (0.203) (0.002)††† 0.120 -0.005 (0.183) (0.008) -0.966 -0.018 (1.407) (0.054) 0.03 -0.01 85 85 1,718 1,056 (5) RF (6) IV -0.001 (0.032) -0.071 (0.061) 0.056 (0.048) 0.11 85 1,056 -0.009 (0.030) -0.049 (0.054) 0.021 (0.048) 0.09 85 1,056 -0.059 (0.061) -0.082 (0.095) -0.011 (0.079) -0.10 85 1,056 0.016 (0.028) 0.050 (0.048) -0.069** (0.031) -0.01 85 1,056 0.002 (0.031) -0.036 (0.040) 0.277 (0.299) 0.11 85 1,056 -0.006 (0.030) -0.021 (0.036) 0.148 (0.281) 0.09 85 1,056 -0.053 (0.070) -0.017 (0.073) 0.064 (0.556) -0.10 85 1,056 0.013 (0.025) 0.019 (0.027) -0.177 (0.195) -0.01 85 1,056 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes all women. Panel A reports the RD treatment effects of the reform, and Panel B reports them by exposure to childhood violence, i.e., whether the respondent experienced violence from her own family members during her childhood. Panel A uses an optimal bandwidth ĥ estimated by the Imbens and Kalyanaraman algorithm, and Panel B uses a static bandwidth of 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood. Columns 1 reports OLS results using years of schooling as the independent variable. Columns 2 – 3 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity. Columns 1 – 3 report these results for the overall sample, and columns 4 – 6 report results from same specifications for the subsample of respondents whose childhood region of residence was rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values unadjusted for multiple-hypothesis testing). †††, ††, and † denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values adjusted for multiple-hypothesis testing using Simes adjustment). Table A12: Effects of Education on Partner Characteristics and Marriage Market Outcomes Overall sample Rural sample (1) OLS (2) RF (3) IV (4) OLS (5) RF (6) IV 0.517*** (0.027)††† 8.80 87 1,816 0.288*** (0.035)††† 24.89 106 2,136 0.006 (0.006) 0.01 84 1,784 0.307*** (0.032)††† 20.94 37 849 0.041*** (0.004)††† 0.58 66 1,427 -0.001 (0.002) 0.05 71 1,496 0.571* (0.338) 8.80 87 1,816 0.227 (0.400) 24.89 106 2,136 0.005 (0.078) 0.01 84 1,784 0.048 (0.405) 20.94 37 849 0.150*** (0.050)†† 0.58 66 1,427 -0.018 (0.021) 0.05 71 1,496 1.613 (1.572) 8.80 87 1,816 0.737 (1.423) 24.89 106 2,136 0.010 (0.169) 0.01 84 1,784 0.271 (2.058) 20.94 37 849 0.299 (0.228) 0.58 66 1,427 -0.031 (0.039) 0.05 71 1,496 0.496*** (0.031)††† 8.47 79 1,020 0.287*** (0.039)††† 24.71 111 1,374 -0.004 (0.005) 0.06 82 1,066 0.270*** (0.038)††† 20.74 40 557 0.043*** (0.004)††† 0.53 87 1,120 -0.001 (0.002) 0.05 122 1,467 0.994** (0.446)† 8.47 79 1,020 0.211 (0.530) 24.71 111 1,374 0.007 (0.065) 0.06 82 1,066 -0.245 (0.532) 20.74 40 557 0.190*** (0.055)††† 0.53 87 1,120 0.006 (0.023) 0.05 122 1,467 0.966* (0.508) 8.47 79 1,020 0.213 (0.518) 24.71 111 1,374 0.006 (0.061) 0.06 82 1,066 -0.866 (3.044) 20.74 40 557 0.167** (0.082) 0.53 87 1,120 0.006 (0.020) 0.05 122 1,467 1.135*** (0.429)†† 0.806 (0.731) -0.756* (0.427) 8.43 85 1,043 1.190* (0.628) -0.284 (0.533) 1.871 (3.802) 8.43 85 1,043 Panel A: RD Treatment Effects Partner’s years of schooling Schooling Partner’s age Mean Bandwidth Observations Schooling Partner’s religiosity index Mean Bandwidth Observations Schooling Marriage age Mean Bandwidth Observations Schooling Marriage decision Mean Bandwidth Observations Schooling Divorced Mean Bandwidth Observations Schooling Mean Bandwidth Observations Panel B: RD Treatment Effects by Exposure to Childhood Violence (Static Bandwidth) Partner’s years of schooling Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations 0.513*** (0.029)††† -0.047 (0.067) -0.016 (0.548) 8.81 85 1,703 0.588* 2.242 0.504*** (0.354) (4.346) (0.035)††† 1.052* -0.964 -0.094 (0.540) (3.915) (0.104) -0.848*** 7.334 0.299 (0.325)†† (30.142) (0.835) 8.81 8.81 8.43 85 85 85 1,703 1,703 1,043 Table A12: Effects of Education on Partner Characteristics and Marriage Market Outcomes, Cont’d Overall sample (1) OLS Partner’s age Schooling Schooling × Childhood violence Childhood violence Partner’s religiosity index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Marriage age Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Marriage decision Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Divorced Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations 0.299*** (0.036)††† -0.193 (0.155) 2.013 (1.282) 24.81 85 1,715 0.004 (0.005) 0.023 (0.032) -0.457* (0.265) 0.01 85 1,718 0.354*** (0.025)††† -0.081 (0.083) 0.311 (0.665) 21.28 85 1,715 0.036*** (0.004)††† 0.029*** (0.010)†† -0.247*** (0.087)†† 0.57 85 1,718 -0.002 (0.002) 0.004 (0.008) 0.010 (0.060) 0.06 85 1,718 (2) RF Rural sample (3) IV (4) OLS 0.170 -0.366 0.283*** (0.439) (2.244) (0.048)††† 1.412 1.607 0.084 (0.886) (2.428) (0.134) 0.140 -11.335 -0.219 (0.408) (18.336) (0.948) 24.81 24.81 24.60 85 85 85 1,715 1,715 1,054 0.001 0.023 0.004 (0.068) (0.292) (0.006) -0.028 -0.043 -0.023 (0.169) (0.303) (0.030) -0.282*** 0.033 -0.038 (0.106)†† (2.323) (0.228) 0.01 0.01 0.06 85 85 85 1,718 1,718 1,056 0.313 0.580 0.282*** (0.310) (1.394) (0.033)††† 0.888** 0.376 0.008 (0.439) (1.381) (0.113) -0.563* -2.866 -0.079 (0.330) (10.561) (0.841) 21.28 21.28 21.09 85 85 85 1,715 1,715 1,054 0.118*** 0.409 0.037*** (0.044)† (0.682) (0.005)††† 0.094 -0.225 0.042*** (0.075) (0.585) (0.011)††† -0.075* 1.726 -0.299*** (0.044) (4.508) (0.092)††† 0.57 0.57 0.52 85 85 85 1,718 1,718 1,056 -0.008 -0.031 -0.002 (0.021) (0.110) (0.002) -0.003 0.021 -0.004 (0.042) (0.110) (0.007) 0.043 -0.120 0.042 (0.031) (0.844) (0.062) 0.06 0.06 0.05 85 85 85 1,718 1,718 1,056 (5) RF (6) IV 0.427 (0.598) 1.350 (0.876) 0.003 (0.541) 24.60 85 1,054 -0.024 (0.068) 0.187 (0.160) -0.265** (0.124) 0.06 85 1,056 0.489 (0.416) 1.001* (0.587) -0.160 (0.436) 21.09 85 1,054 0.163*** (0.059)†† 0.271*** (0.086)†† -0.106* (0.061) 0.52 85 1,056 0.010 (0.029) -0.010 (0.044) 0.019 (0.036) 0.05 85 1,056 0.353 (0.555) 0.512 (0.561) -2.984 (3.966) 24.60 85 1,054 -0.031 (0.067) 0.105 (0.097) -0.930 (0.723) 0.06 85 1,056 0.421 (0.388) 0.303 (0.366) -1.801 (2.652) 21.09 85 1,054 0.144 (0.088) 0.069 (0.076) -0.450 (0.540) 0.52 85 1,056 0.010 (0.028) -0.009 (0.028) 0.080 (0.211) 0.05 85 1,056 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes all women. Panel A reports the RD treatment effects of the reform, and Panel B reports them by exposure to childhood violence, i.e., whether the respondent experienced violence from her own family members during her childhood. Panel A uses an optimal bandwidth ĥ estimated by the Imbens and Kalyanaraman algorithm, and Panel B uses a static bandwidth of 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood. Columns 1 reports OLS results using years of schooling as the independent variable. Columns 2 – 3 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity. Columns 1 – 3 report these results for the overall sample, and columns 4 – 6 report results from same specifications for the subsample of respondents whose childhood region of residence was rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values unadjusted for multiple-hypothesis testing). †††, ††, and † denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values adjusted for multiple-hypothesis testing using Simes adjustment). Table A13: Effects of Education on Spousal Violence Overall sample (1) OLS (2) RF -0.026*** (0.006)††† 0.00 83 1,747 -0.022*** (0.004)††† 0.04 115 2,260 -0.015*** (0.005)††† -0.04 83 1,740 -0.163** (0.081) 0.00 83 1,747 -0.014 (0.058) 0.04 115 2,260 0.031 (0.074) -0.04 83 1,740 Rural sample (3) IV (4) OLS (5) RF (6) IV -0.119 (0.089) 0.01 108 1,362 -0.045 (0.082) 0.06 79 1,020 0.062 (0.103) -0.04 74 963 -0.121 (0.101) 0.01 108 1,362 -0.039 (0.070) 0.06 79 1,020 0.051 (0.087) -0.04 74 963 Panel A: RD Treatment Effects Physical violence index Schooling Psychological violence index Mean Bandwidth Observations Schooling Financial control index Mean Bandwidth Observations Schooling Mean Bandwidth Observations -0.382 -0.026*** (0.365) (0.007)††† 0.00 0.01 83 108 1,747 1,362 -0.049 -0.027*** (0.209) (0.006)††† 0.04 0.06 115 79 2,260 1,020 0.069 -0.016** (0.172) (0.007)†† -0.04 -0.04 83 74 1,740 963 Panel B: RD Treatment Effects by Exposure to Childhood Violence (Static Bandwidth) Physical violence index Schooling Schooling × Childhood violence Childhood violence Psychological violence index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Financial control index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations -0.016*** (0.006)††† -0.052** (0.024)† 0.951*** (0.196)††† 0.00 85 1,718 -0.018*** (0.004)††† 0.002 (0.018) 0.349*** (0.130)††† 0.05 85 1,718 -0.009* (0.005)† -0.023 (0.029) 0.439* (0.231)† -0.04 85 1,711 -0.121 (0.079) -0.153 (0.185) 0.618*** (0.111)††† 0.00 85 1,718 0.027 (0.065) -0.033 (0.116) 0.381*** (0.057)††† 0.05 85 1,718 0.068 (0.071) -0.120 (0.163) 0.316*** (0.095)††† -0.04 85 1,711 -0.376 -0.019** -0.049 -0.031 (0.737) (0.008)†† (0.098) (0.096) 0.145 -0.045 -0.373* -0.168 (0.699) (0.029) (0.217) (0.136) -0.608 0.794*** 0.613*** 1.635* (5.347) (0.259)††† (0.156)††† (0.988) 0.00 0.00 0.00 0.00 85 85 85 85 1,718 1,056 1,056 1,056 0.135 -0.024*** 0.042 0.045 (0.363) (0.006)††† (0.079) (0.082) -0.132 0.001 -0.124 -0.080 (0.356) (0.025) (0.127) (0.079) 1.381 0.264 0.323*** 0.845 (2.708) (0.171) (0.073)††† (0.564) 0.05 0.05 0.05 0.05 85 85 85 85 1,718 1,056 1,056 1,056 0.341 -0.013* 0.083 0.086 (0.570) (0.007)† (0.099) (0.104) -0.368 -0.062** -0.167 -0.120 (0.525) (0.028)† (0.161) (0.102) 3.077 0.530** 0.186 0.976 (4.028) (0.265)† (0.132) (0.769) -0.04 -0.05 -0.05 -0.05 85 85 85 85 1,711 1,051 1,051 1,051 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes all women. Panel A reports the RD treatment effects of the reform, and Panel B reports them by exposure to childhood violence, i.e., whether the respondent experienced violence from her own family members during her childhood. Panel A uses an optimal bandwidth ĥ estimated by the Imbens and Kalyanaraman algorithm, and Panel B uses a static bandwidth of 85 months, which is the optimal bandwidth estimated for the years of schooling in rural regions of childhood. Columns 1 reports OLS results using years of schooling as the independent variable. Columns 2 – 3 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity. Columns 1 – 3 report these results for the overall sample, and columns 4 – 6 report results from same specifications for the subsample of respondents whose childhood region of residence was rural. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values unadjusted for multiple-hypothesis testing). †††, ††, and † denote significance at the 1, 5, and 10 percent levels, respectively (based on p-values adjusted for multiple-hypothesis testing using Simes adjustment). Table A14: RD Treatment Effects in Rural Childhood Regions with Different Optimal Bandwidth Selection Methods IK CCT (1) OLS Years of schooling Schooling Child abuse Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Frequent child abuse Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Men can beat their partners in certain situations. Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence It may be necessary to beat children for discipline. Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Age at first pregnancy Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations -0.021*** (0.005) 0.007 (0.014) 0.165* (0.095) 0.51 158 1,703 -0.019*** (0.007) 0.034 (0.021) 0.113 (0.163) 0.45 51 640 -0.012 (0.008) 0.019 (0.032) -0.033 (0.277) 0.43 24 284 -0.024*** (0.004) -0.004 (0.014) 0.098 (0.114) 0.30 90 1,097 0.338*** (0.041) -0.067 (0.092) 0.798 (0.588) 20.76 55 712 (2) RF (3) IV 1.172** (0.464) 7.45 81 1,316 0.069 0.073 (0.061) (0.068) -0.157* -0.113* (0.095) (0.066) 0.267*** 1.008** (0.045) (0.454) 0.51 0.51 158 158 1,703 1,703 0.111 0.164 (0.087) (0.220) -0.225* -0.174 (0.120) (0.147) 0.427*** 1.590 (0.065) (1.066) 0.45 0.45 51 51 640 640 0.049 -1.958 (0.152) (76.347) 0.024 0.980 (0.184) (37.465) 0.094 -6.871 (0.133) (266.660) 0.43 0.43 24 24 284 284 0.039 0.041 (0.051) (0.061) 0.019 -0.008 (0.088) (0.053) 0.076 0.147 (0.058) (0.383) 0.30 0.30 90 90 1,097 1,097 0.448 0.535 (0.473) (0.586) 1.128 0.100 (0.680) (0.586) 0.076 -0.314 (0.439) (4.216) 20.76 20.76 55 55 712 712 (4) OLS -0.023*** (0.005) 0.002 (0.018) 0.238* (0.123) 0.51 89 1,096 -0.020*** (0.005) -0.001 (0.018) 0.250* (0.134) 0.44 92 1,119 -0.023*** (0.005) -0.010 (0.013) 0.169 (0.103) 0.41 88 1,039 -0.024*** (0.004) -0.008 (0.013) 0.119 (0.104) 0.29 99 1,189 0.360*** (0.037) -0.095 (0.083) 0.767 (0.599) 21.14 106 1,336 (5) RF (6) IV 1.134** (0.451) 7.42 85 1,396 0.083 0.097 (0.068) (0.096) -0.224** -0.148** (0.103) (0.074) 0.341*** 1.308** (0.056) (0.531) 0.51 0.51 89 89 1,096 1,096 0.108 0.134 (0.073) (0.122) -0.268** -0.185** (0.103) (0.084) 0.344*** 1.557** (0.062) (0.613) 0.44 0.44 92 92 1,119 1,119 0.053 0.052 (0.068) (0.070) -0.067 -0.056 (0.110) (0.065) 0.120** 0.490 (0.056) (0.446) 0.41 0.41 88 88 1,039 1,039 0.025 0.027 (0.049) (0.055) 0.009 -0.009 (0.084) (0.048) 0.075 0.147 (0.052) (0.344) 0.29 0.29 99 99 1,189 1,189 0.650* 0.672 (0.359) (0.437) 0.871 0.161 (0.547) (0.401) 0.036 -0.778 (0.411) (2.922) 21.14 21.14 106 106 1,336 1,336 Table A14: RD Treatment Effects in Rural Childhood Regions with Different Optimal Bandwidth Selection Methods, Cont’d CCT (1) OLS Number of children Schooling Schooling × Childhood violence Childhood violence Employed Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Personal income index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Partner’s years of schooling Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Marriage decision Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations -0.135*** (0.011) -0.039 (0.026) 0.168 (0.240) 1.71 68 1,055 0.013** (0.005) -0.014 (0.013) 0.184** (0.091) 0.12 36 482 0.015** (0.006) 0.034 (0.039) -0.269 (0.226) -0.08 69 877 0.500*** (0.043) -0.164 (0.121) 0.627 (0.935) 8.59 46 600 0.035*** (0.006) 0.038** (0.015) -0.265** (0.123) 0.55 53 664 (2) RF IK (3) IV -0.095 -0.077 (0.129) (0.113) -0.184 -0.026 (0.224) (0.099) -0.025 0.104 (0.145) (0.792) 1.71 1.71 68 68 1,055 1,055 -0.017 1.384 (0.059) (171.338) -0.164 -0.532 (0.114) (55.249) 0.135* 4.192 (0.071) (436.585) 0.12 0.12 36 36 482 482 -0.033 -0.027 (0.068) (0.073) -0.070 -0.018 (0.107) (0.087) -0.017 0.078 (0.090) (0.666) -0.08 -0.08 69 69 877 877 0.948 2.886 (0.603) (5.023) 1.803* -1.151 (0.924) (3.280) -1.164** 8.299 (0.558) (24.158) 8.59 8.59 46 46 600 600 0.116 0.148 (0.070) (0.144) 0.213** -0.014 (0.103) (0.117) -0.071 0.105 (0.080) (0.836) 0.55 0.55 53 53 664 664 (4) OLS (5) RF -0.146*** -0.135 (0.010) (0.113) -0.014 -0.109 (0.026) (0.198) -0.059 -0.045 (0.242) (0.125) 1.69 1.69 88 88 1,382 1,382 0.014*** 0.027 (0.004) (0.037) -0.003 -0.061 (0.014) (0.079) 0.087 0.088 (0.100) (0.055) 0.18 0.18 93 93 1,139 1,139 0.013*** -0.052 (0.005) (0.058) 0.026 -0.069 (0.033) (0.091) -0.218 -0.020 (0.190) (0.077) -0.10 -0.10 89 89 1,101 1,101 0.502*** 1.043** (0.034) (0.440) -0.150 0.647 (0.095) (0.726) 0.632 -0.730* (0.765) (0.420) 8.47 8.47 79 79 979 979 0.038*** 0.162*** (0.005) (0.057) 0.041*** 0.265*** (0.011) (0.085) -0.292*** -0.103* (0.092) (0.060) 0.53 0.53 87 87 1,076 1,076 (6) IV -0.139 (0.120) 0.006 (0.077) -0.164 (0.616) 1.69 88 1,382 0.034 (0.046) -0.044 (0.049) 0.378 (0.360) 0.18 93 1,139 -0.053 (0.074) -0.011 (0.069) 0.019 (0.528) -0.10 89 1,101 1.149* (0.669) -0.301 (0.566) 1.901 (4.052) 8.47 79 979 0.146* (0.086) 0.064 (0.072) -0.409 (0.518) 0.53 87 1,076 Table A14: RD Treatment Effects in Rural Childhood Regions with Different Optimal Bandwidth Selection Methods, Cont’d IK CCT (1) OLS Physical violence index Schooling Schooling × Childhood violence Childhood violence Psychological violence index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Overall depression index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Somatic depression index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Nonsomatic depression index Mean Bandwidth Observations Schooling Schooling × Childhood violence Childhood violence Mean Bandwidth Observations (2) RF (3) IV -0.014 -0.147 0.182 (0.012) (0.228) (0.519) -0.024 -0.471 -0.317 (0.066) (0.337) (0.300) 0.591 0.561** 2.619 (0.433) (0.271) (2.170) 0.02 0.02 0.02 24 24 24 325 325 325 -0.025*** 0.021 0.026 (0.006) (0.088) (0.088) -0.003 -0.107 -0.068 (0.029) (0.136) (0.096) 0.289 0.317*** 0.758 (0.196) (0.084) (0.697) 0.06 0.06 0.06 70 70 70 877 877 877 -0.018** -0.062 -0.294 (0.008) (0.096) (0.850) -0.033 -0.552*** -0.100 (0.030) (0.143) (0.395) 0.509*** 0.449*** 0.858 (0.188) (0.087) (2.916) 0.05 0.05 0.05 44 44 44 567 567 567 -0.027*** -0.003 0.004 (0.008) (0.085) (0.091) -0.005 -0.403*** -0.178* (0.031) (0.137) (0.106) 0.162 0.276*** 1.348* (0.210) (0.091) (0.774) 0.01 0.01 0.01 64 64 64 817 817 817 -0.016* -0.170 -1.379 (0.009) (0.115) (9.691) -0.039 -0.546*** 0.226 (0.033) (0.190) (3.675) 0.584*** 0.476*** -1.758 (0.207) (0.111) (28.426) 0.06 0.06 0.06 37 37 37 496 496 496 (4) OLS (5) RF (6) IV -0.021*** -0.040 -0.031 (0.007) (0.089) (0.104) -0.044* -0.303 -0.141 (0.026) (0.197) (0.129) 0.754*** 0.547*** 1.419 (0.233) (0.130) (0.919) 0.01 0.01 0.01 108 108 108 1,311 1,311 1,311 -0.025*** 0.036 0.041 (0.005) (0.084) (0.090) 0.004 -0.097 -0.071 (0.027) (0.133) (0.090) 0.252 0.315*** 0.785 (0.189) (0.076) (0.648) 0.06 0.06 0.06 79 79 79 980 980 980 -0.022*** 0.017 0.039 (0.005) (0.067) (0.088) 0.004 -0.286** -0.169** (0.015) (0.115) (0.081) 0.257** 0.368*** 1.449** (0.111) (0.056) (0.575) 0.03 0.03 0.03 115 115 115 1,360 1,360 1,360 -0.033*** 0.054 0.071 (0.007) (0.075) (0.112) 0.013 -0.340** -0.192* (0.025) (0.143) (0.103) 0.122 0.333*** 1.543** (0.170) (0.081) (0.736) 0.03 0.03 0.03 96 96 96 1,167 1,167 1,167 -0.019*** 0.040 0.054 (0.005) (0.077) (0.106) -0.011 -0.290** -0.160** (0.017) (0.127) (0.081) 0.378*** 0.407*** 1.414** (0.120) (0.063) (0.575) 0.04 0.04 0.04 95 95 95 1,156 1,156 1,156 Notes: Data are from the 2014 National Survey on Domestic Violence against Women in Turkey. The sample includes women who have children and whose childhood region is rural. The optimal bandwidth is estimated by using the Calonico et al. (2014) (CCT) algorithm in columns 1 - 3, and the Imbens and Kalyanaraman (2009) (IK) algorithm in columns 4 - 6. The RD treatment effects of the reform are reported by exposure to childhood violence, i.e., whether the respondent experienced violence from her own family members during her childhood. Columns 1 and 2 report OLS results using years of schooling as the independent variable. Columns 2 and 5, and columns 3 and 6 report reduced-form RD treatment effects and two-stage least-squares RD treatment effects (by using treatment as an instrument for years of schooling) of being born after January 1987 with a linear control function in the month-year of birth on each side of the discontinuity, respectively. The variables are described in Appendix A. All specifications control for a dummy variable for whether the respondent grew up in a rural location, a dummy variable for whether the respondent’s mother tongue is not Turkish, month-of-birth fixed effects, region fixed effects, and interactions of region fixed effects with an indicator of rural regions. Standard errors are clustered at the month-year cohort level. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively.