Article Figures & Data
Figures
- Figure 1
Conceptual Framework
Notes: This diagram summarizes the conceptual framework for our analysis. See the text for more details on this framework.
- Figure 2
Main Regression Results
Notes: Each subpanel of the figure displays the point estimates (dots) and 95 percent confidence intervals (lines) on the treatment group terms from the regressions presented in Table 4 for the respective country (panel title) and outcome (subpanel title). Outcome measures for each country are constructed and standardized as described in Online Appendix Table 3. Regressions are performed as described in the notes to Table 4. The p-values listed next to each treatment group test the null hypothesis of no program impact (versus control) for that intervention and outcome and are drawn from the regressions presented in Table 4.
- Figure 3
K-Densities for Mental Health Index by Country
Notes: Each subpanel of this figure displays, separately by country, the kernel densities for the mental health index by treatment group. The kernel densities are weighted appropriately according to each country’s study design. The mental health measure is constructed and standardized as described in the text. In Panel B for Bangladesh, the data was trimmed below −4 (n = 3). GM and GR are abbreviations for Growth Mindset and Girl Rising, respectively. In Panel C for Ethiopia, the control group data was trimmed below −5 (n = 3). AWH is an abbreviation for Act With Her.
Tables
Tanzania Bangladesh Ethiopia Panel A: Interventions Goal Setting, implemented in 2017, encourages girls to set a goal to remain healthy and stay STI/HIV free; Empowerment and Livelihood for Adolescents (ELA), an ongoing intervention, provides a safe social space, life-skills training and community support for girls. Growth Mindset is an intervention aimed at developing “growth mindset,” the belief that ability depends on effort and is not constrained by fixed intelligence; Girl Rising uses the power of storytelling to change the way the world values girls and their education. Interventions were implemented in 2021. Act With Her is a multilevel intervention that includes a safe social space, life skills training for girls and boys (separately), awareness raising for their caregivers, and (in a random half of communities) community-level work aimed at improved adolescent-focused services; discussion of gender norms and attitudes is interwoven at each level, and goal setting for girls is a focus. Act With Her + Transfers additionally includes in-kind school supply and sexual and reproductive health–related Transfers for girls. Interventions were implemented during 2019–2020. Panel B: Sample Location Dodoma and Iringa Regions Chittagong & Sylhet Divisions South Gonder Zone (Amhara), East Hararghe Zone (Oromia) Sublocation/cluster 79 communities 100 schools 126 communities Urban/rural Rural/Peri-urban Urban/Rural Rural Baseline age range 10–19-year-olds 10–18-year-oldsa 10–12-year-olds Baseline sample size 1,449 1,200 1,863 Baseline survey year(s) 2016 2020 2017/18 Follow-up survey year(s) 2018 2021 2019/20 Panel C: Study Design Cross-cutting design Multi-arm cluster RCT Multi-arm cluster RCT Goal setting: Individually randomized (369 treated, 1,080 control)
ELA: Cluster randomized (32 treatment communities, 47 control)(36 control schools, 36 Growth Mindset, 37 Growth Mindset + Girl Rising) (39 control communities, 58 Act With Her, 29 Act With Her + Transfers) ↵a 93 percent of adolescents are aged 10–14; 61 adolescents are 15 years old, 16 are 16 years old, three are 17 years old, and one is 18 years old.
Tanzania Bangladesh Ethiopia Panel A: Adolescent Characteristics Age at baseline 14.7 12.7 10.9 (2.5) (1.1) (0.8) Age at follow-up 16.1 14.8 13.2 (3.0) (1.1) (1.1) = 1 if adolescent is enrolled in school 0.721 1.000 0.836 = 1 if adolescent aspires to higher than secondary schooling 0.807 0.852 0.581 Index of perceptions of gender stereotypical roles (0–2) 1.009 0.890 1.645 (0.439) (0.824) (0.601) = 1 if adolescent has savings 0.138 0.217 0.023 Panel B: Household Characteristics Number of household members 3.3 5.8 6.4 (1.3) (2.3) (1.8) = 1 for mother living in household 0.764 0.947 0.913 = 1 for father living in household 0.580 0.862 0.829 Household head highest grade attained 8.7 7.4 2.4 (3.0) (5.3) (3.8) = 1 if improved flooring (wood/tiles/ceramic/carpet/cement/bamboo) 0.894 0.656 0.017 = 1 for primary female caregiver in the household engages in paid work 0.580 0.151 0.089 Number of observations 1,449 1,220 1,863 Notes: This table presents summary statistics for the sample used in each country. All statistics are measured for the entire sample (by country) at baseline, with the exception of “age at follow-up.” Means are presented for all measures, and standard deviations are presented (in parentheses) for continuous measures only.
Tanzania Bangladesh Ethiopia Construct #1: Depression PHQ Depression Score Patient Health Questionnaire-2 (PHQ- 2) score—a measure of depression, with scores ranging from 0–6. Higher scores indicate higher levels of distress; measure used in regression is standardized within the sample control group. Patient Health Questionnaire-9 (PHQ 9) score—a measure of depression (Kroenke, Spitzer and Williams 2001), with scores ranging from 0–27. Higher scores indicate higher levels of distress; measure used in regression is standardized within the sample control group. Patient Health Questionnaire-9 (PHQ 9) score—a measure of depression (Kroenke, Spitzer and Williams 2001), with scores ranging from 0–27. Higher scores indicate higher levels of distress; measure used in regression is standardized within the sample control group. Summary statistics (mean, standard deviation within control group at follow-up) 1.01
(1.26)2.43
(2.50)0.59
(2.09)Indicator for moderate/severe depression Indicator of depression—binary measure taking on a value of 1 if raw PHQ-2 score is ≥3. Indicator of moderate/severe depression—binary measure taking on a value of 1 if raw PHQ-9 score ≥10. Indicator of moderate/severe depression—binary measure taking on a value of 1 if raw PHQ-9 score ≥10. Summary statistics (mean, standard deviation within control group at follow-up) 0.135
(0.342)0.019
(0.138)0.01
(0.08)Construct #2: Socio-Emotional Development Confidence/grit/resilience Confidence—a measure that takes on a value of 1 if the adolescent agrees she is confident she can complete any task set before her. Measure used in regression is standardized within the sample control group. Grit Scale—a measure of adolescent grit generated from a 7-item scale following Alan, Boneva, and Ertac (2019), with scores ranging from 0 to 4, where higher scores indicate more grit. Measure used in regression is standardized within the sample control group. Child and Youth Resilience Measure- 12 (CYRM-12)—a measure of youth resilience (Liebenberg, Ungar, and LeBlanc 2013), with scores ranging from 12 to 36, where higher scores indicate more resilience. Measure used in regression is standardized within the sample control group. Summary statistics (mean, standard deviation within control group at follow-up) 0.529
(0.499)2.97
(0.457)31.31
(4.41)Construct #3: Locus of Control Locus of control Pearlin Mastery Scale items, which is the extent to which the respondent feels in control of the events that influence her life. The Pearlin Mastery Scale (Pearlin and Schooler 1978) is a 7-item scale of seven statements to which respondents indicate if they strongly disagree, disagree, feel neutral, agree, or strongly agree. The responses to each item are coded so that all responses go in the same direction (toward feelings of more control) and summed to an index that takes on values of 12 to 40. We generate a binary indicator = 1 if the score is >28 (median score at baseline). Measure used in regression standardizes this indicator within the sample control group. Self-reported feeling of control over own life, measured on a scale from 1 to 10, where higher values indicate more control. Adapted from the World Values Survey (Haerpfer et al. 2020). Question asks: “Some people feel that they have a great deal of control over their own lives. Others feel that what they do has very little effect on what happens to them. On a scale from 1 to 10, with 1 being very little and 10 being complete control, how would you classify yourself?” Measure used in regression is standardized within the sample control group. Self-reported feeling of control over own life, measured on a scale from 1 to 10, where higher values indicate more control. Adapted from the World Values Survey (Haerpfer et al. 2020). Question asks: “Some people feel that they have a great deal of control over their own lives. Others feel that what they do has very little effect on what happens to them. On a scale from 1 to 10, with 1 being very little and 10 being complete control, how would you classify yourself?” Measure used in regression is standardized within the sample control group. Summary statistics (mean, standard deviation within control group at follow-up) 28.7
(4.66)6.789
(2.32)7.05
(3.01)PHQ Depression Score (Standardized) Indicator for Moderate/Severe Depression Socio-Emotional Development (Standardized) Locus of Control (Standardized) Mental Health Index (Standardized) (1) (2) (3) (4) (5) Panel A: Tanzania Goal Setting 0.150 0.086** −0.057 −0.112 −0.165 (0.102) (0.033) (0.111) (0.089) (0.103) [0.237] [0.355] [0.155] Goal Setting × ELA −0.208 −0.111** 0.058 0.096 0.199 (0.131) (0.043) (0.137) (0.137) (0.129) [0.746] [0.345] [0.121] ELA only 0.136 0.054 −0.135 −0.079 −0.174 (0.102) (0.037) (0.105) (0.118) (0.123) [0.384] [0.310] [0.104] Number of observations 1,167 1,167 1,176 1,176 1,167 p-value on K–S (GS + ELA vs. ELA) [0.736] [0.470] [0.271] Panel B: Bangladesh Growth Mindset −0.018 −0.002 0.317*** 0.175 0.286** (0.090) (0.009) (0.111) (0.110) (0.125) [0.220] [0.128] [0.469] [0.054] Growth Mindset + Girl Rising −0.001 −0.003 0.140 0.027 0.090 (0.107) (0.011) (0.103) (0.101) (0.095) [0.276] [0.258] [0.534] [0.275] Number of observations 1,096 1,096 1,100 1,104 1,092 p-value on (1st row – 2nd row) [0.877] [0.960] [0.126] [0.272] [0.137] p-value on K–S (GM vs. GM + GR) [0.471] [0.579] [0.429] [0.130] Panel C: Ethiopia Act With Her −0.113 −0.006 0.158** 0.104 0.172** (0.071) (0.007) (0.067) (0.072) (0.071) [0.767] [0.000] [0.301] [0.007] Act With Her + Transfers 0.011 0.007 0.087 0.056 0.063 (0.091) (0.010) (0.072) (0.075) (0.076) [0.897] [0.041] [0.483] [0.125] Number of observations 1,518 1,518 1,405 1,617 1,288 p-value on (1st row – 2nd row) [0.091] [0.130] [0.329] [0.490] [0.116] p-value on K–S (AWH vs. AWH + Transfers) [0.780] [0.086] [0.412] [0.031] Notes: Each column and panel presents results from a separate regression. The column title indicates the outcome measure (construct) of interest. Regression controls include only those necessary to account for the study sampling design by country (basic controls); more specific information on the control sets for each country can be found in Online Appendix Table 3. For regressions with other control specifications, see Online Appendix Tables 10–12. Each regression is properly weighted and clustered according to the country’s sample design. Results presented for each treatment indicator include the coefficient estimates, standard errors in parentheses, and in brackets the p-values from Kolmogorov–Smirnov (K–S) tests of equality of distributions between the treatment group and control group. Significance: *p < 0.1, **p < 0.05, ***p < 0.001. The results presented in brackets in the final two rows of each panel are (i) p-values on the test of equality of the treatment indicator coefficients and (ii) p-values from Kolmogorov–Smirnov (K–S) tests of equality of distributions between the two treatment groups.
× Indicator for Assets >Median × Indicator for Enrolled in Schoola × Indicator for Perceptions of Gendered Norms in Community >Median × Indicator for PHQ Score > Medianb × Indicator for Mother Lives in Household (1) (2) (3) (4) (5) Panel A: Tanzania Goal Setting −0.243* −0.209 −0.212* −0.157 −0.227 (0.131) (0.161) (0.119) (0.131) (0.193) Goal Setting × ELA 0.254 0.078 0.158 0.112 0.229 (0.166) (0.269) (0.144) (0.175) (0.287) GS × Heterogeneity 0.168 0.056 0.125 −0.024 0.076 (0.142) (0.204) (0.227) (0.164) (0.221) (GS × ELA) × Heterogeneity −0.114 0.162 0.012 0.157 −0.036 (0.193) (0.345) (0.277) (0.293) (0.335) Number of observations 1,167 1,167 1,167 1,139 1,167 Panel B: Bangladesh Growth Mindset 0.010 0.076 0.388** 0.337** 0.201 (0.129) (0.242) (0.155) (0.135) (0.417) Growth Mindset + Girl Rising 0.137 0.227 0.054 0.005 0.042 (0.116) (0.212) (0.118) (0.127) (0.272) GM × Heterogeneity 0.456*** 0.242 −0.318 −0.110 0.089 (0.167) (0.244) (0.224) (0.216) (0.421) (GM + GR) × Heterogeneity −0.104 −0.164 0.055 0.173 0.051 (0.133) (0.233) (0.195) (0.226) (0.275) Number of observations 1,092 1,092 1,092 1,092 1,092 Panel C: Ethiopia Act with her 0.029 0.402* 0.132** 0.227** −0.015 (0.079) (0.224) (0.066) (0.091) (0.126) Act with her + Transfers −0.013 0.403 0.017 0.118 −0.117 (0.086) (0.246) (0.076) (0.096) (0.164) AWH × Heterogeneity 0.342*** −0.261 0.127 −0.143 0.209 (0.115) (0.222) (0.187) (0.132) (0.153) (AWH + T) × Heterogeneity 0.166 −0.373 0.147 −0.133 0.198 (0.138) (0.252) (0.191) (0.154) (0.194) Number of observations 1,288 1,288 1,288 1,270 1,288 Notes: Each column and panel presents results from a separate regression. The column title indicates the heterogeneity metric of interest, and the panel title indicates the country sample. The outcome measure is the mental health index. Regression controls include only those necessary to account for the study sampling design (basic controls); more specific information on the control sets can be found in Online Appendix Table 3. Each regression is properly weighted and clustered according to the country’s sample design. Results presented include the coefficient (standard error) on a treatment indicator or the interaction between the heterogeneity metric and a treatment indicator. Significance: *p < 0.1, **p < 0.05, ***p < 0.001.
↵a For Bangladesh, the educational heterogeneity term is a measure of education aspirations, as school enrollment was a requirement for study eligibility.
↵b For Ethiopia, the depression heterogeneity term is the General Health Questionnaire (GHQ-12) measure, another validated measure of mental health (as the PHQ was not collected at baseline in Ethiopia).
Mental Health Index PHQ Depression Score Indicator for Moderate/Severe Depression Socio-Emotional Development Locus of Control Above-Median CATE Below-Median CATE Difference Above-Median CATEa Below-Median CATE Difference Above-Median CATEa Below-Median CATE Difference Above-Median CATE Below-Median CATE Difference Above-Median CATE Below-Median CATE Difference (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) Panel A: Tanzania Goal Setting −0.037 −0.142 0.105 0.006 0.123 −0.117 0.022 0.055 −0.033 0.006 −0.091 0.097 0.033 −0.115 0.148 (0.101) (0.099) [0.458] (0.108) (0.107) [0.442] (0.035) (0.039) [0.529] (0.097) (0.094) [0.472] (0.095) (0.094) [0.268] Goal Setting × ELA −0.042 −0.059 0.017 −0.092 0.097 −0.189 −0.02 0.005 −0.025 −0.004 −0.116 0.112 0.106 −0.261 0.367** (0.135) (0.145) [0.931] (0.135) (0.144) [0.338] (0.052) (0.046) [0.719] (0.134) (0.129) [0.547] (0.131) (0.129) [0.046] Panel B: Bangladesh Growth Mindset 0.106 0.08 0.025 −0.166 0.129 −0.295** −0.02 0.026 −0.046** 0.196 0.083 0.113 0.142 0.021 0.121 (0.106) (0.101) [0.864] (0.080) (0.100) [0.021] (0.010) (0.018) [0.025] (0.097) (0.083) [0.376] (0.096) (0.103) [0.390] Growth Mindset + Girl Rising 0.177 −0.091 0.268** 0.022 0.070 −0.048 −0.017 0.016 −0.033 0.16 0.118 0.042 0.105 −0.15 0.255* (0.096) (0.093) [0.045] (0.111) (0.094) [0.741] (0.016) (0.013) [0.109] (0.083) (0.089) [0.730] (0.097) (0.093) [0.058] Panel C: Ethiopia Act With Her only 0.159 0.146 0.013 −0.088 −0.01 −0.078 −0.016 0.003 −0.019*** 0.167 0.13 0.037 0.132 −0.008 0.14 (0.075) (0.070) [0.899] (0.077) (0.056) [0.413] (0.007) (0.002) [0.009] (0.066) (0.084) [0.729] (0.072) (0.068) [0.157] Act With Her + Transfers 0.074 −0.169 0.243** 0.023 0.038 −0.015 0.006 0.015 −0.009 0.061 −0.092 0.153 0.056 −0.074 0.13 (0.082) (0.088) [0.043] (0.085) (0.100) [0.909] (0.008) (0.011) [0.508] (0.083) (0.098) [0.234] (0.081) (0.083) [0.262] Notes: This table presents results of average treatment effects (ATEs) for adolescents based on groups defined by adolescents with high and low conditional average treatment effects (CATEs, estimated using grf model in R) for individuals in the three study settings. Above- (below-) median groups represent adolescents whose CATEs are greater than (less than) the median estimated CATEs. Standard errors are in parentheses. Each panel of the table shows the treatment effects for each CATE subgroup derived from country-specific treatment arms and the respective set of outcomes. Significance: *p < 0.1, **p < 0.05, ***p < 0.001.
↵a For PHQ Depression Score and Indicator for Moderate/Severe Depression, the “above-median CATE” indicates a more negative impact on the measure.
Tanzania Bangladesh Bangladesh Bangladesh Bangladesh Ethiopia Ethiopia Locus of Control Goal Setting × ELA PHQ Depression Score Growth Mindset = 1 Moderate/Severe Depression Growth Mindset Locus of Control Growth Mindset + Girl Rising Mental Health Index Growth Mindset + Girl Rising = 1 Moderate/Severe Depression Act with Her Mental Health Index Act With Her + Assets Above-Median CATE Below-Median CATE Difference Above-Median CATEa Below-Median CATE Difference Above-Median CATEa Below-Median CATE Difference Above-Median CATE Below-Median CATE Difference Above-Median CATE Below-Median CATE Difference Above-Median CATEa Below-Median CATE Difference Above-Median CATE Below-Median CATE Difference (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (21) Age 14.027 15.391 −1.364*** 14.844 14.853 −0.009 14.889 14.816 0.073 14.712 14.987 −0.275* 14.94 14.758 0.182*** 10.85 10.968 −0.118*** 10.873 10.849 0.024 (0.092) (0.105) [0.000] (0.043) (0.047) [0.888] (0.045) (0.045) [0.251] (0.042) (0.046) [0.062] (0.043) (0.047) [0.004] (0.028) (0.029) [0.003] (0.032) (0.031) [0.590] Indicator for enrolled in school 0.886 0.561 0.325*** 0.005 0.002 0.003 0.481 0.682 −0.201 0.002 0.004 −0.002 0.571 0.421 0.150*** 0.879 0.816 0.063*** 0.86 0.885 −0.025 (0.013) (0.020) [0.000] (0.003) (0.001) [0.956] (0.149) (0.153) [0.347] (0.002) (0.003) [0.579] (0.021) (0.021) [0.000] (0.012) (0.014) [0.000] (0.014) (0.013) [0.191] Household size 3.422 3.054 0.368*** 5.942 5.631 0.311** 5.819 5.754 0.065 5.895 5.69 0.205 5.799 5.769 0.03 6.642 6.124 0.518*** 6.627 6.006 0.621*** (0.056) (0.049) [0.000] (0.043) (0.094) [0.034] (0.106) (0.103) [0.660] (0.095) (0.113) [0.165] (0.109) (0.100) [0.839] (0.060) (0.068) [0.000] (0.070) (0.067) [0.000] Indicator for mother lives in HH 0.736 0.782 −0.046 0.94 0.944 −0.004 0.947 0.936 0.011 0.944 0.94 0.004 0.934 0.95 −0.016 0.974 0.871 0.103*** 0.935 0.91 0.025* (0.018) (0.017) [0.632] (0.010) (0.010) [0.777] (0.010) (0.010) [0.437] (0.010) (0.010) [0.777] (0.011) (0.009) [0.260] (0.006) (0.012) [0.000] (0.010) (0.011) [0.093] Indicator for father lives in HH 0.612 0.524 0.088*** 0.849 0.862 −0.013 0.862 0.848 0.014 0.848 0.863 −0.015 0.849 0.862 −0.013 0.904 0.771 0.133*** 0.862 0.815 0.047** (0.127) (0.021) [0.002] (0.015) (0.015) [0.540] (0.015) (0.015) [0.509] (0.015) (0.015) [0.800] (0.015) (0.015) [0.540] (0.011) (0.015) [0.000] (0.014) (0.015) [0.022] HH head years of education 8.586 8.836 −0.25 7.528 6.48 1.048*** 6.054 7.934 −1.88*** 9.571 4.425 5.146*** 7.362 6.648 0.714** 2.397 2.473 −0.076 2.315 2.695 −0.38* (0.092) (0.125) [0.161] (0.223) (0.207) [0.000] (0.204) (0.221) [0.000] (0.199) (0.172) [0.000] (0.228) (0.202) [0.019] (0.147) (0.139) [0.707] (0.143) (0.169) [0.086] Indicator for parent participation in paid work 0.526 0.658 −0.132*** 0.162 0.118 0.044** 0.127 0.153 −0.026 0.132 0.147 −0.015 0.138 0.139 −0.001 0.024 0.157 −0.133*** 0.095 0.09 0.005 (0.021) (0.020) [0.000] (0.016) (0.014) [0.039] (0.014) (0.015) [0.205] (0.014) (0.015) [0.465] (0.015) (0.015) [0.962] (0.006) (0.013) [0.000] (0.012) (0.011) [0.759] Asset count 5.556 4.578 0.978*** 6.027 4.871 1.156*** 4.912 5.974 −1.062*** 6.824 4.066 2.758*** 5.505 5.378 0.127 5.241 5.881 −0.64*** 5.602 5.507 0.095 (0.125) (0.110) [0.000] (0.137) (0.099) [0.000] (0.127) (0.114) [0.000] (0.111) (0.103) [0.000] (0.128) (0.117) [0.464] (0.108) (0.103) [0.000] (0.113) (0.119) [0.563] Index of perceptions of gender stereotypical roles −0.002 −0.01 0.008 0.106 0.038 0.068 0.094 0.051 0.043 −0.026 0.172 −0.198*** 0.009 0.13 −0.121** 0.011 −0.019 0.03 0.102 −0.125 0.227*** (0.044) (0.043) [0.897] (0.040) (0.042) [0.241] (0.041) (0.042) [0.463] (0.043) (0.039) [0.001] (0.041) (0.042) [0.039] (0.038) (0.036) [0.567] (0.037) (0.043) [0.000] Indicator for aspires to earn bachelor's degree 0.838 0.791 0.047** 0.837 0.807 0.03 0.813 0.831 −0.018 0.9 0.744 0.156*** 0.836 0.809 0.027 0.696 0.523 0.173*** 0.475 0.731 −0.256*** (0.015) (0.017) [0.038] (0.016) (0.017) [0.199] (0.017) (0.016) [0.441] (0.013) (0.019) [0.000] (0.016) (0.017) [0.247] (0.017) (0.018) [0.000] (0.020) (0.017) [0.000] Notes: This table presents country-level covariate means estimates for conditional average treatment effect (CATE) subgroups (above-median and below-median CATE) for selected outcomes and treatments. In particular, the outcome–treatment combinations shown are those that displayed statistically significant heterogeneous treatment effects across CATE subgroup in Table 6. Significance: *p < 0.1, **p < 0.05, ***p < 0.001.
↵a For PHQ Depression Score and Indicator for Moderate/Severe Depression, the “above-median CATE” indicates a more negative impact on the measure.
Free alternate access to The Journal of Human Resources supplementary materials is available at https://uwpress.wisc.edu/journals/journals/jhr-supplementary.html
