Fertility and Life Satisfaction in Rural Ethiopia

Online Resource 1

Fertility and Life Satisfaction in Rural Ethiopia

Pierluigi Conzo, Giulia Fuochi, and Letizia Mencarini

Robustness check for the IV estimation: the weak-instrument problem

In order to ascertain the extent to which weak instrument problems can affect our estimates, we perform the following statistical tests in the specifications in columns 3-4 and 7-8 in Table 4: i) we run the Kleibergen and Paap (2006) rk LM test to determine whether the minimal correlation between endogenous variables and the instrument is statistically different from zero; as argued by Bazzi and Clemens (2013), “the LM test for identification provides a lower hurdle than the tests for weak instruments”; ii) we implement the Kleibergen-Paap (2006) Wald test (robust to heteroskedasticity and clustering at village level) and compare the results with the critical values which were originally tabulated by Stock and Yogo (2005) for the Cragg-Donald statistic (Baum et al. 2007; Bazzi and Clemens 2013).

As far as the first check is concerned, the Kleibergen and Paaprk LM significantly rejects the null of under-identification: the rk statistic is 5.41 (p-value = 0.02) and 5.73 (p-value = 0.02) in the specifications in columns 3-4 and 7-8 in Table 4 respectively. Secondly, in these specifications the Kleibergen-Paap Wald test reports a rk statistic equal to 4.90 (p-value = 0.03) and 6.61 (p-value = 0.01) respectively; the last figure is larger than (close to) the Stock and Yogo critical value of 5.53 (6.66) when restricting the bias of the IV estimator to 25 (20) percent of the OLS bias. Results from all these diagnostics suggest that the common problems associated with instrument relevance (i.e. weak instruments) are likely to be mitigated in our identification strategy.

Attrition bias: IPW estimation

In general, attrition due to the data-cleaning steps described in section 3.1 is accounted for in our IV regression since the use of the instrumental variable reduces the effects of attrition bias in linear regression models (e.g. Leigh et al, 1993). In this section we further deal with panel attrition with respect to cases that are present in Wave 6 (2004) but not in Wave 7 (2009), i.e. the waves we use for our main analyses. Panel attrition can lead to biased estimates when non-response is not at random. Even though panel attrition in our data is not remarkable (14%) given the difficulties in collecting data in a rural area of a developing country, we however check whether our estimates change dramatically when weighing them by the inverse of the estimated probability of attrition (IWP estimation, e.g., among many others, Raab et al. 2005; Nicoletti and Peracchi 2005; Vandecasteele and Debels 2007). This procedure gives less weight to individuals who have a low probability of non-response, as they maybe overrepresented – on the basis on their personal characteristics - in the observed sample.

The attrition model is estimated as follows:

where is a dummy variable = 1 for respondents interviewed at t-1 (wave 6, 2004) but not at t (wave 7, 2009) and = 0 for respondents who were in the panel in both at t and t-1; SES is a set of K controls capturing respondents’ socio-economic status at t-1 (2004); Regioni are G dummies for respondents’ region of residence in 2004. Since we estimate the attrition probability conditionally on the respondent’s characteristics observed at time t-1, we select controls that are fairly stable over time and that minimize the number of missing responses. The probability of contacting a respondent at t is inversely related to its degree of geographical mobility and to the probability of finding someone at home. Individuals’ socio-economic status and the characteristics of the region of residence may account for these factors. For these reasons, controls in the first model (MODEL 1) include religion, gender, land size and the absence of toilet in the house. In an alternative model (MODEL 2) we add per capita food expenditure, even though this variable has some missing values. Table S5 reports descriptive statistics of personal characteristics by attrition. On average respondents dropping in 2009 are mainly Christian, own a slightly larger land and enjoy higher per capita food expenditure than those remaining in both waves. Continuing respondents, however, happen to have more often a toilet in their house than those who drop in 2009.

When we estimate the baseline attrition model (MODEL 1), the only significant variable is the presence of a toilet, our proxy for wealth (Table S6, col. 1). The inverse of this estimated probability is then used in our main models so to give more weight to respondents that - on the basis of their personal characteristics - are more likely to attrite (Inverse Probability Weighting approach, IPW). Results are reported in Tables S7a-c and are consistent with our main findings (unweighted). Similar results are obtained when using MODEL 2 (i.e. including per capita food expenditure in the attrition model, Table S6, col. 2) and are reported in Tables S8a-c. This robustness check, jointly with the IV estimation, document that attrition is not leading to a severe bias in our main estimates and, therefore, attrition bias can be excluded as main interpretation of our findings.

Selection on age cut-off: IPW estimation

We apply the IPW approach described above to account for a potential source of selection deriving from the sample restriction to participants who are aged 60 or less. The descriptive statistics in Table S9 show that on average respondents aged 60+ are more likely to be male and Christian, to own a larger piece of land and to have more often a toilet in their house than those aged ≤ 60 (these differences are significant at 5% level in all comparisons except for total land size). To control for this potential source of bias we follow the IPW approach described above and weight the baseline models for older participants (columns 5-6, Tables 2a-b) with the inverse of the estimated probability of being selected (i.e. being 60 years old or less). Results from the first-step selection model (Table S10) are in line with the descriptive statistics, with selected respondents less likely to be Christian and female and less likely to have a toilet in their house. In the second step, we re-estimate the main models by weighing each observation with the inverse of the estimated selection probability. Results are reported in Table S11 and confirm the baseline results. This robustness check suggests that selection on age cut-off is not leading to a severe bias in our main estimates and, therefore, that is not the main interpretation of our findings.

References

Baum, C. F., Schaffer, M. E., & Stillman, S. (2007). Enhanced routines for instrumental variables/GMM estimation and testing.Stata Journal,7, 465–506.

Bazzi, S., & Clemens, M. A. (2013). Blunt instruments: Avoiding common pitfalls in identifying the causes of economic growth. American Economic Journal: Macroeconomics, 5(2), 152–186.

Kleibergen, F., & Paap, R. (2006). Generalized reduced rank tests using the singular value decomposition. Journal of Econometrics, 133, 97–126.

Leigh, J. P., Ward, M. M., & Fries, J. F. (1993). Reducing attrition bias with an instrumental variable in a regression model: Results from a panel of rheumatoid arthritis patients. Statistics in Medicine, 12, 1005–1018.

Nicoletti, C.,Peracchi, F. (2005). Survey response and survey characteristics: Microlevel evidence from the European Community Household Panel. Journal of the Royal Statistical Society:Series A, 168, 763–781.

Raab, G., Purdon, S., Buckner, K. (2005).Practical exemplars and survey analysis. Retrieved from

Stock, J. H., & Yogo, M. (2005). Testing for weak instruments in linear IV regression. In D. W. K. Andrews & J. H. Stock (Eds.), Identification and inference for econometric models: Essays in honor of Thomas Rothenberg (pp. 80–108). New York, NY: Cambridge University Press.

Table S1a. OLS regression of women’s life satisfaction on fertility, ERHS 2009

(1) / (2) / (3)
Age<50 / Age<50 / Age 50-60
N. children ever born / 0.021 / 0.028 / 0.058
(0.041) / (0.040) / (0.047)
Birth event in the last 5 years / -0.465*
(0.199)
Controls / YES / YES / YES
Observations / 284 / 284 / 195
R-squared / 0.063 / 0.069 / 0.233
Robust standard errors clustered at village level in parentheses. Controls include all the regressors in column 2 and 5 of Table 2a but Log per capita household, Food expenditure, N. open loans, Total land size, No toilet in the household and Adequacy perception index.***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S1b. OLS regression of men’s life satisfaction on fertility, ERHS 2009

(1) / (2) / (3)
Age<50 / Age<50 / Age 50-60
N. children ever born / -0.006 / -0.003 / 0.099**
(0.040) / (0.040) / (0.035)
Birth event in the last 5 years / -0.286
(0.173)
Controls / YES / YES / YES
Observations / 284 / 284 / 195
R-squared / 0.063 / 0.069 / 0.233
Robust standard errors clustered at village level in parentheses. Controls include all the regressors in column 2 and 5 of Table 2b with the exception of Log per capita household, Food expenditure, N. open loans, Total land size, No toilet in the household and Adequacy perception index. ***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S2. Fixed effects regressions of life satisfaction on fertility, ERHS 2004 and 2009

(1) / (2)
Men, age<50 / Women, age<45
N. children ever born / -0.115 / -0.170
(0.130) / (0.125)
Controls / YES / YES
Observations / 561 / 566
R-squared / 0.018 / 0.024
N. of respondents / 284 / 293
Robust standard errors clustered at village level in parentheses. Controls include all the regressors in column 2 and 4 of Table 3 with the exception of Log per capita household food expenditure, N. open loans, Total land size, No toilet in the household and Adequacy perception index. ***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S3. OLS regression of life satisfaction on fertility, ERHS 2009 with lagged controls.

(1) / (2) / (3) / (4)
Women, Age<45 / Men, Age<50 / Women, Age<45 / Men, Age<50
N. children ever born / -0.034 / 0.005 / -0.030 / -0.000
(0.036) / (0.042) / (0.036) / (0.042)
Birth event in the last 5 years / -0.607** / -0.328† / -0.581** / -0.329†
(0.208) / (0.169) / (0.215) / (0.170)
Adequacy perception index / 0.858*** / 1.272*** / 0.833*** / 1.180***
(0.215) / (0.198) / (0.212) / (0.209)
Adequacy perception index at wave 1 / 0.893*** / 0.412 / 0.680** / 0.108
(0.205) / (0.276) / (0.248) / (0.285)
Life satisfaction at wave 1 / 0.135* / 0.188**
(0.065) / (0.065)
Controls / YES / YES / YES / YES
Controls at wave 1 / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES
Dummies for socio-political/household shocks at wave 1 / YES / YES / YES / YES
Region dummies / YES / YES / YES / YES
Observations / 248 / 260 / 248 / 260
R-squared / 0.416 / 0.269 / 0.427 / 0.296

Robust standard errors clustered at village level in parentheses. Controls include: Has a co-resident partner; Went to school; Physical limitations; Non-Protestant Christian (Ref. cat.: Muslim household) ;Protestant; Other religion; Log per capita household; food expenditure; N. open loans; Total land size; No toilet in the household. Controls at wave 1 include the following regressors measured at wave 1 (2004): Log per capita household, food expenditure; N. open loans; Total land size; No toilet in the household. ***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S4a. OLS regression of women’s life satisfaction on fertility, ERHS 2009

(1) / (2) / (3) / (4) / (5) / (6)
VARIABLES / Age<45 / Age<45 / Age<45 / Age<45 / Age 45-60 / Age 45-60
N. children ever born / -0.004 / 0.011 / 0.005 / 0.005 / 0.055 / 0.055
(0.040) / (0.038) / (0.035) / (0.035) / (0.041) / (0.040)
Birth event in the last 5 years / -0.539** / -0.523** / -0.523**
(0.199) / (0.182) / (0.182)
Adequacy perception index / 0.826*** / 0.826*** / 0.772*
(0.177) / (0.177) / (0.360)
Life satisfaction at wave 1 / 0.289*** / 0.291*** / 0.253*** / 0.253*** / 0.187* / 0.159†
(0.052) / (0.051) / (0.050) / (0.050) / (0.075) / (0.081)
Controls / YES / YES / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES / YES / YES
Region dummies / YES / YES / YES / YES / YES / YES
Observations / 269 / 269 / 269 / 269 / 229 / 229
R-squared / 0.304 / 0.325 / 0.367 / 0.367 / 0.308 / 0.332

Table S4b. OLS regression of men’s life satisfaction on fertility, ERHS 2009

(1) / (2) / (3) / (4) / (5) / (6)
VARIABLES / Age<50 / Age<50 / Age<50 / Age<50 / Age 50-60 / Age 50-60
N. children ever born / -0.009 / -0.007 / -0.013 / -0.013 / 0.092* / 0.092*
(0.044) / (0.044) / (0.040) / (0.040) / (0.036) / (0.036)
Birth event in the last 5 years / -0.249 / -0.232 / -0.232
(0.166) / (0.157) / (0.157)
Adequacy perception index / 0.936*** / 0.936*** / 1.116***
(0.176) / (0.176) / (0.255)
Life satisfaction at wave 1 / 0.314*** / 0.312*** / 0.280*** / 0.280*** / 0.245*** / 0.203**
(0.057) / (0.058) / (0.057) / (0.057) / (0.067) / (0.061)
Controls / YES / YES / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES / YES / YES
Region dummies / YES / YES / YES / YES / YES / YES
Observations / 258 / 258 / 258 / 258 / 184 / 184
R-squared / 0.190 / 0.194 / 0.249 / 0.249 / 0.371 / 0.441

Table S5. Socio-economic controls by attrition, ERHS 2004 and 2009

Both in 2004 and 2009 (A=0) / Only in 2004 (A=1)
Variable / Obs / Mean / Std. Dev. / Obs / Mean / Std. Dev.
Christian / 1585 / 0.495 / 0.500 / 265 / 0.796 / 0.404
Other / 1585 / 0.026 / 0.159 / 265 / 0.023 / 0.149
Protestant / 1585 / 0.223 / 0.417 / 265 / 0.083 / 0.276
Muslim / 1585 / 0.256 / 0.436 / 265 / 0.098 / 0.298
Female / 1585 / 0.508 / 0.500 / 265 / 0.426 / 0.495
Total land size / 1577 / 6.923 / 47.917 / 262 / 7.399 / 27.782
No toilet / 1585 / 0.632 / 0.483 / 265 / 0.891 / 0.313
Per capita Food exp. (ln) / 1522 / 1.779 / 1.128 / 250 / 2.146 / 1.094

Table S6.Attrition probability models, ERHS 2004 & 2009.

Depvar: Pr(A=1) / (1) / (2)
MODEL 1 / MODEL 2
Per capita food exp (ln) / 0.018**
(0.007)
Christian / 0.033 / 0.021
(0.087) / (0.085)
Other / 0.125 / 0.107
(0.200) / (0.187)
Protestant / 0.123 / 0.078
(0.136) / (0.118)
Female / -0.009 / -0.012
(0.024) / (0.025)
Total land size / 0.000 / 0.000
(0.000) / (0.000)
No toilet / 0.072† / 0.066†
(0.038) / (0.038)
Region dummies / YES / YES
Observations / 1,839 / 1.764
Marginal effects from probit estimation of attrition probability (A=0 if respondent was interviewed both in 2004 and 2009; A=1 if respondent was interviewed only in 2004). Robust standard errors clustered at village level in parentheses. ***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S7a. IPW regression of women’s life satisfaction on fertility, ERHS 2009

(1) / (2) / (3) / (4) / (5) / (6)
VARIABLES / Age<45 / Age<45 / Age<45 / Age<45 / Age 45-60 / Age 45-60
N. children ever born / -0.030 / -0.018 / -0.011 / -0.011 / 0.019 / 0.026
(0.058) / (0.053) / (0.049) / (0.049) / (0.043) / (0.045)
Birth event in the last 5 years / -0.463* / -0.516* / -0.516*
(0.213) / (0.212) / (0.212)
Adequacy perception index / 0.739** / 0.739** / 0.579
(0.228) / (0.228) / (0.433)
Life satisfaction at wave 1 / 0.283*** / 0.278*** / 0.229** / 0.229** / 0.284** / 0.266**
(0.058) / (0.056) / (0.067) / (0.067) / (0.088) / (0.092)
Controls / YES / YES / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES / YES / YES
Region dummies / YES / YES / YES / YES / YES / YES
Observations / 267 / 267 / 267 / 267 / 225 / 225
R-squared / 0.311 / 0.328 / 0.368 / 0.368 / 0.352 / 0.368

Robust standard errors clustered at village level in parentheses. Controls include: Has a co-resident partner; Went to school; Physical limitations; Non-Protestant Christian (Ref. cat.: Muslim household);Protestant; Other religion; Log per capita household; food expenditure; N. open loans; Total land size; No toilet in the household. Weights: inverse estimated probability of attrition (Model 1, Table S6). ***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S7b. IPW regression of men’s life satisfaction on fertility, ERHS 2009

(1) / (2) / (3) / (4) / (5) / (6)
VARIABLES / Age<50 / Age<50 / Age<50 / Age<50 / Age 50-60 / Age 50-60
N. children ever born / 0.054 / 0.054 / 0.049 / 0.049 / 0.109* / 0.113**
(0.068) / (0.068) / (0.062) / (0.062) / (0.045) / (0.042)
Birth event in the last 5 years / -0.264 / -0.284 / -0.284
(0.217) / (0.204) / (0.204)
Adequacy perception index / 0.851*** / 0.851*** / 1.209***
(0.206) / (0.206) / (0.329)
Life satisfaction at wave 1 / 0.323*** / 0.317*** / 0.278*** / 0.278*** / 0.253** / 0.194*
(0.058) / (0.060) / (0.064) / (0.064) / (0.090) / (0.074)
Controls / YES / YES / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES / YES / YES
Region dummies / YES / YES / YES / YES / YES / YES
Observations / 258 / 258 / 258 / 258 / 183 / 183
R-squared / 0.268 / 0.273 / 0.322 / 0.322 / 0.354 / 0.435

Robust standard errors clustered at village level in parentheses. Controls include: Has a co-resident partner; Went to school; Physical limitations; Non-Protestant Christian (Ref. cat.: Muslim household);Protestant; Other religion; Log per capita household; food expenditure; N. open loans; Total land size; No toilet in the household. Weights: inverse estimated probability of attrition (Model 1, Table S6). ***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S7c. IPW Fixed effects regressions of life satisfaction on fertility, ERHS 2004 and 2009

(1) / (2) / (3) / (4)
Women, age<45 / Men, age<50
N. children ever born / -0.318* / -0.332* / -0.045 / -0.081
(0.151) / (0.145) / (0.184) / (0.166)
Adequacy perception index / 0.641** / 0.519*
(0.207) / (0.259)
Controls / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES
Observations / 509 / 509 / 498 / 498
R-squared / 0.139 / 0.174 / 0.172 / 0.202
N. of respondents / 273 / 273 / 260 / 260

Robust standard errors clustered at village level in parentheses. Controls include: Has a co-resident partner; Went to school; Physical limitations; Log per capita household; food expenditure; N. open loans; Total land size; No toilet in the household. Weights: inverse estimated probability of attrition (Model 1, Table S6). ***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S8a. IPW regression of women’s life satisfaction on fertility, ERHS 2009

(1) / (2) / (3) / (4) / (5) / (6)
VARIABLES / Age<45 / Age<45 / Age<45 / Age<45 / Age 45-60 / Age 45-60
N. children ever born / -0.032 / -0.025 / -0.019 / -0.019 / 0.027 / 0.030
(0.052) / (0.047) / (0.045) / (0.045) / (0.044) / (0.044)
Birth event in the last 5 years / -0.360† / -0.432* / -0.432*
(0.207) / (0.213) / (0.213)
Adequacy perception index / 0.737** / 0.737** / 0.650
(0.218) / (0.218) / (0.426)
Life satisfaction at wave 1 / 0.250*** / 0.248*** / 0.204** / 0.204** / 0.291*** / 0.268**
(0.059) / (0.056) / (0.068) / (0.068) / (0.081) / (0.087)
Controls / YES / YES / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES / YES / YES
Region dummies / YES / YES / YES / YES / YES / YES
Observations / 266 / 266 / 266 / 266 / 223 / 223
R-squared / 0.298 / 0.309 / 0.350 / 0.350 / 0.364 / 0.383

Robust standard errors clustered at village level in parentheses. Controls include: Has a co-resident partner; Went to school; Physical limitations; Non-Protestant Christian (Ref. cat.: Muslim household);Protestant; Other religion; Log per capita household; food expenditure; N. open loans; Total land size; No toilet in the household. Weights: inverse estimated probability of attrition (Model 2, Table S6). ***p < .001, ** p<0.01, * p<0.05, †p<0.1

Table S8b. IPW regression of men’s life satisfaction on fertility, ERHS 2009

(1) / (2) / (3) / (4) / (5) / (6)
VARIABLES / Age<50 / Age<50 / Age<50 / Age<50 / Age 50-60 / Age 50-60
N. children ever born / 0.044 / 0.044 / 0.039 / 0.039 / 0.108** / 0.108**
(0.054) / (0.055) / (0.050) / (0.050) / (0.038) / (0.037)
Birth event in the last 5 years / -0.238 / -0.262 / -0.262
(0.209) / (0.199) / (0.199)
Adequacy perception index / 0.793*** / 0.793*** / 1.076**
(0.204) / (0.204) / (0.321)
Life satisfaction at wave 1 / 0.326*** / 0.321*** / 0.287*** / 0.287*** / 0.229* / 0.183*
(0.051) / (0.052) / (0.055) / (0.055) / (0.093) / (0.082)
Controls / YES / YES / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES / YES / YES
Region dummies / YES / YES / YES / YES / YES / YES
Observations / 256 / 256 / 256 / 256 / 182 / 182
R-squared / 0.288 / 0.292 / 0.335 / 0.335 / 0.407 / 0.468

Robust standard errors clustered at village level in parentheses. Controls include: Has a co-resident partner; Went to school; Physical limitations; Non-Protestant Christian (Ref. cat.: Muslim household);Protestant; Other religion; Log per capita household; food expenditure; N. open loans; Total land size; No toilet in the household. Weights: inverse estimated probability of attrition (Model 2, Table S6). ***p < .001, ** p<0.01, * p<0.05, †p<0.1

Table S8c. IPW Fixed effects regressions of life satisfaction on fertility, ERHS 2004 and 2009

(1) / (2) / (3) / (4)
Women, age<45 / Men, age<50
N. children ever born / -0.341* / -0.352* / -0.024 / -0.066
(0.158) / (0.149) / (0.186) / (0.163)
Adequacy perception index / 0.634** / 0.482†
(0.213) / (0.260)
Controls / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES
Observations / 507 / 507 / 495 / 495
R-squared / 0.146 / 0.179 / 0.162 / 0.189
N. of respondents / 272 / 272 / 258 / 258

Robust standard errors clustered at village level in parentheses. Controls include: Has a co-resident partner; Went to school; Physical limitations; Log per capita household; food expenditure; N. open loans; Total land size; No toilet in the household. Weights: inverse estimated probability of attrition (Model 2, Table S6). ***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S9. Socio-economic controls by age cut, ERHS 2009

Respondents aged 60+ / Respondents aged ≤ 60
Variable / Obs / Mean / Std. Dev. / Obs / Mean / Std. Dev.
Christian / 263 / 0.593 / 0.492 / 1023 / 0.463 / 0.499
Other / 263 / 0.023 / 0.150 / 1023 / 0.027 / 0.163
Protestant / 263 / 0.163 / 0.371 / 1023 / 0.243 / 0.429
Muslim / 263 / 0.221 / 0.415 / 1023 / 0.266 / 0.442
Female / 263 / 0.357 / 0.480 / 1023 / 0.532 / 0.499
Total land size / 261 / 3.984 / 37.248 / 1023 / 2.802 / 13.430
No toilet / 263 / 0.376 / 0.485 / 1023 / 0.293 / 0.455
Per capita Food exp. (ln) / 261 / 2.429 / 1.012 / 1017 / 2.375 / 0.901

Table S10. Selection probability model, ERHS 2009

VARIABLES / Pr(Age ≤ 60)
Christian / -0.073*
(0.035)
Other / -0.061
(0.060)
protestant / -0.002
(0.038)
Female / 0.138***
(0.025)
Total_land_size / -0.000
(0.000)
Per capita food exp (ln) / -0.021
(0.015)
No_toilet / -0.076**
(0.027)
Region dummies / YES
Observations / 1,277
Marginal effects from probit estimation of selection probability. Robust standard errors clustered at village level in parentheses. ***p ‹ .001, ** p<0.01, * p<0.05, †p<0.1

Table S11. IPW regression of life satisfaction (0-10 scale) on fertility, ERHS 2009

(1) / (2) / (3) / (4)
Women / Men
VARIABLES / Age 45-60 / Age 45-60 / Age 50-60 / Age 50-60
N. children ever born / 0.052 / 0.052 / 0.101** / 0.101**
(0.042) / (0.041) / (0.034) / (0.034)
Adequacy perception index / 0.816* / 1.114***
(0.374) / (0.253)
Life satisfaction at wave 1 / 0.191* / 0.162* / 0.228*** / 0.190**
(0.074) / (0.080) / (0.065) / (0.058)
Controls / YES / YES / YES / YES
Dummies for socio-political/household shocks / YES / YES / YES / YES
Region dummies / YES / YES / YES / YES
Observations / 229 / 229 / 184 / 184
R-squared / 0.311 / 0.338 / 0.380 / 0.448