to rule this out, but it would need to be generated by quite unusual decision making.

4.2

# Results on Camps and Gender Inequality

The central question in this section is how the gender imbalance in vaccination is related to the number of health camps. The basic result can be seen in Figure 3, which reports the gender difference in average number of vaccinations for children six months to two years, graphed against the number of vaccination camps in the previous year, as well as the average number of vaccinations for each gender.^{6 }The graph shows two measures of the gender difference. The first line (the black squares) just shows the raw differences in number of vaccines. This points to a non-monotonic relationship. In particular, moving from zero or one camp to two camps causes a large increase in the gender difference. Above two camps girls begin to gain. The adjusted line (the black triangles) shows these differences once the measure of vaccines is adjusted for a variety of covariances – child

age, mother’s education and income and village characteristics.^{7 }made, the non-monotonicity remains.

Even once these adjustments are

In terms of levels, the figure suggests that moving from zero camps to five camps increases the number of vaccinations on average, although vaccinations are slightly lower in the group with the most camps.^{8 }This decline may be related to camp placement. Perhaps areas with very low levels of vaccinations were targeted to receive a larger number of camps. Even if this is the case, however, and placement is related to pre-existing levels of vaccinations, it may not be a problem for the work here. I am focusing here on the interaction between gender and number of camps, so the relationship in levels is not a central issue.

Table 4 explores the relationship between camps and gender imbalance in vaccination in a regression context with controls. I control for standard demographics and family characteristics. In addition, all regressions include state fixed effects and interactions between state and gender dummies, as well as interactions between gender and quadratics in village population, village area and distance to the nearest PHC or CHC.^{9 }All standard errors are clustered at the village level.

^{6}I restrict to children in this age group since they are the ones who would have needed vaccinations in the previous year. Consistent with this, the results are less strong for older children.

^{7}To do this adjustment the measure of vaccines is regressed on these controls, and the predicted residuals for boys and girls are subtracted to get the gender di erence.

^{8}For girls, moving from one to two camps actually is slightly negative in terms of the average number of vaccinations. This is likely due to the relatively small sample of villages; this di erence is not statistically significant.

^{9}The regressions in Table 4 use as the dependent variable the number of vaccination camps, rather than (for example) number of camps per area or per capita. However, I note that by controlling for area and village population flexibly among the independent variables, we are implicitly considering camps per capita or per area, but with a more flexible functional form.

17