-0.086*** 0.813*** -0.097*** 2.768*** -0.106*** 1.131** (0.030) (0.300) (0.030) (0.635) (0.029) (0.484) Female -0.035** -0.043*** -0.037** -0.060*** -0.030* -0.041** (0.016) (0.016) (0.016) (0.017) (0.016) (0.016) Age 0.614*** 0.362*** 0.640*** -0.243 0.647*** 0.265* (0.020) (0.088) (0.022) (0.199) (0.022) (0.152) Age2 -0.027*** -0.016*** -0.028*** 0.010 -0.028*** -0.012* (0.001) (0.004) (0.001) (0.009) (0.001) (0.007) Urban -0.015 -0.005 -0.011 0.011 -0.148*** -0.129*** (0.018) (0.018) (0.018) (0.018) (0.024) (0.026)
Father Attended School 0.009 0.023 0.013 0.054 -0.030 -0.004
(0.050) (0.050) (0.050) (0.049) (0.046) (0.047)
Mother Attended School 0.077 0.084 0.080 0.084 0.033 0.043
(0.070) (0.068) (0.066) (0.065) (0.059) (0.059) Migrated -0.005 -0.165** -0.013 -0.525*** -0.052 -0.262*** (0.040) (0.066) (0.037) (0.119) (0.037) (0.084) First Child 0.013 0.016 0.014 0.013 0.003 0.005 (0.019) (0.019) (0.019) (0.019) (0.019) (0.019) Adopted -0.030 0.010 -0.023 0.096 -0.030 0.025 (0.067) (0.066) (0.065) (0.067) (0.060) (0.063) Niece/Nephew 0.054 0.032 0.048 -0.034 0.026 -0.006 (0.039) (0.039) (0.037) (0.042) (0.037) (0.040) Ln (Household Size) 0.022 0.033 0.013 0.040* 0.123*** 0.118*** (0.024) (0.024) (0.024) (0.024) (0.029) (0.030) Share of boys 0-1 0.106 0.110 0.118 0.190 0.328** 0.323** (0.145) (0.146) (0.143) (0.143) (0.143) (0.143) Share of boys 2-4 0.221** 0.212** 0.260** 0.234** 0.423*** 0.376*** (0.106) (0.106) (0.103) (0.103) (0.102) (0.105) Share of boys 5-9 0.059 -0.014 0.078 -0.172* 0.154* 0.031 (0.082) (0.085) (0.081) (0.094) (0.084) (0.096) Share of boys 10-14 -0.081 -0.112 -0.052 -0.158* 0.069 0.004 (0.084) (0.085) (0.082) (0.084) (0.084) (0.090) Share of men above 65 -0.189 -0.214 -0.181 -0.240 -0.039 -0.086 (0.186) (0.189) (0.181) (0.185) (0.181) (0.185) Share of girls 0-1 0.056 -0.125 0.080 -0.453** 0.257* -0.006 (0.148) (0.160) (0.145) (0.187) (0.148) (0.186) Share of girls 2-4 0.088 0.114 0.104 0.223** 0.164 0.200* (0.111) (0.114) (0.108) (0.113) (0.108) (0.110) Share of girls 5-9 0.226*** 0.187** 0.222*** 0.077 0.295*** 0.216** (0.083) (0.084) (0.081) (0.087) (0.081) (0.089) Share of girls 10-14 -0.048 -0.080 -0.028 -0.166* 0.011 -0.057 (0.085) (0.085) (0.082) (0.086) (0.082) (0.089)
Share of women above 65 0.177 0.175 0.181 0.185 0.157 0.155
(0.182) (0.180) (0.172) (0.170) (0.165) (0.165)
Completed Years of Education
-0.002 0.027*** -0.017*** -0.002
Table 30: Primary Education Demand – Marginal Effects - Standard Models with Conflict 2: Extreme Killings Shock
(1) (2) (3) (4) (5) (6)
Individual, Household & Local
+ Education as Investment Standard Model Probit IV Probit Probit IV Probit Probit IV Probit
b/se b/se b/se b/se b/se b/se
Explained variable = Probability of child, between ages 6 and 14, to have attended school in the 2004/05 school year
E(Returns to Education) -4.275** 11.238*** -4.989*** 1.724
(1.878) (3.948) (1.797) (3.280)
Average local wage 0.031 0.019 -0.050 -0.040
(0.043) (0.042) (0.042) (0.043)
Number of cows owned 0.016*** 0.018*** 0.010*** 0.012***
(0.003) (0.003) (0.003) (0.004)
Household had a loan denied -0.142*** -0.073 -0.074* -0.053
(0.044) (0.045) (0.043) (0.043)
E (Per capita Monthly Household Income)
0.015*** 0.012*** (0.002) (0.003)
Average local education costs -0.002** -0.002***
(0.001) (0.001)
Average local subsidies to education
0.000 0.001 (0.002) (0.002) Average time to primary
school -0.001 -0.001 (0.001) (0.001) Insufficient Access -1.767*** -1.631*** (0.400) (0.419) Insufficient Quality -9.603*** -10.758*** (2.270) (2.300) Insufficient Security -3.182 -3.856 (2.932) (3.000)
District fixed effects Yes Yes Yes Yes Yes Yes
Ethno-linguistic fixed effects Yes Yes Yes Yes Yes Yes
N 6031 6022 6031 6022 6022 6022
* p<0.10, ** p<0.05, *** p<0.01 Source: author’s calculations on data from TLSLS (2007) and CAVR (2006)
Although the indication of a shock would suggest reading the marginal effect result directly, the values would be misleading. Again, although counter-intuitively, it is more meaningful to translate the impact of a one standard deviation equivalent change in the (let us call it for the sake of the interpretation) probability of having experienced the violent 1999 post-referendum shock. In this case, such an increase (of 40%) translates into a probability of such a child having attended primary education in 2004/05 lower, on average, in 4% if considering the (likely to be overestimated) Probit estimate, or higher in 45% if considering the IV Probit estimate. Again, such a high potential correction of the coefficient, together with much higher standard deviations, suggests the need to control for its robustness. A further robustness test was added in this case. The analysis was to test the effects of extreme shocks of violence, defined as the experience of at least a year when the number of killings and disappearances reported in the district of birth exceeded the yearly average for all districts throughout the 25 years of Indonesian occupation in two standard deviations. In Table 31, below, the values of
the coefficients result from a complete empirical model with the threshold defined, with alternatives of one and three standard deviations in excess of the average. The signs and statistical significance of the coefficients, both in the Probit and IV Probit estimations hold, despite, again, of higher standard deviations, indicators of lower efficiency of the latter estimations.
Table 31: Primary Education Demand - robustness threshold of extreme violence
(1) (2) (3) (4) (5) (6)
Threshold = +1 Standard Dev. Baseline Threshold = +3 Standard Dev. Probit IV Probit Probit IV Probit Probit IV Probit
b/se b/se b/se b/se b/se b/se
Conflict 2: Shock of extreme violence during life
-0.568*** 1.649** -0.402*** 4.260** -0.402*** 4.260** (0.097) (0.709) (0.112) (1.831) (0.112) (1.831)
N 6022 6022 6022 6022 6022 6022
Log-Likelihood -120627.692 -121698.393 -121308.848 -121698.394 -121308.848 -121698.394 * p<0.10, ** p<0.05, *** p<0.01 Note: Coefficients of the conflict variables were estimated with the remaining covariates of the study’s standard models. Source: author’s calculations on data from TLSLS (2007) and CAVR (2006)
Following a robustness test strategy similar to the one explored above, the lateness of entry into primary school can be explored. Table 56, in the appendix, shows the estimates of medium run effects of conflict considering different age cohorts to account for later entry. When the age of entrance is assumed to be higher than 6, neither the Probit nor the IV Probit estimates are significant, reinforcing, therefore, the indication of a lack of robustness of the estimates of medium run impact of conflict.
The same can be read from the estimates in Table 32 and Table 33 and the first stage tests for the IV regressions in Table 57 and Table 58. As it can be seen, the IV Probit estimates do not show themselves to be robust to the decomposition. In fact, although distance to the border is shown to be a strong regressor of the conflict indicator, exogeneity cannot be rejected and the IV estimates for girls and boys cannot be used in inference. Therefore, an equally negative and significant IV Probit coefficient of the conflict indicator, found in the regression on the subsample of girls, fails to be valid for inference. The IV estimates for Rural and Urban subsamples are also not valid for inference for, as it can be seen in Table 58, the Wooldridge test of significance of the
residual does not allow us to reject the exogeneity of the experience of conflict to unknown drivers of demand for education, using distance to the border as an instrument. A robustness test performed in the location where violence was experienced, as presented in Table 3470, gives the same indication.
On the other hand, the empirical results regarding structural drivers of demand for primary education in the empirical models that incorporate the experience of the 1999 post-referendum violent shock hold with the same signs and significance as reported in the case of the use of average killings during the life of the child as a proxy of conflict.
Table 32: Primary Education Demand– Marginal Effects - Conflict 2: Shock of extreme violence during life of child - Decomposition by Gender
(1) (2) (3) (4) (5) (6)
All Sample Girls Boys
Probit IV Probit Probit IV Probit Probit IV Probit
b/se b/se b/se b/se b/se b/se
Explained variable = Probability of child, between ages 6 and 14, to have attended school in the 2004/05 school year
Conflict 2: Shock of