the sample. The hypothesis is that the magnitude of the coefficients should be larger for the set of women who report wanting a greater ratio of boys to girls.^{10 }I define women as having a preference for boys if they report wanting more boys than girls; the alternative is desiring equal numbers or more girls.

These regressions are shown in Table 6. Columns 1 and 2 focus on total vaccinations as the measure of vaccines; Columns 3 and 4 focus on whether the child has any vaccines. Columns 1 and 3 include women who do not report wanting a male-biased sex ratio among children; Columns 2 and 4 include those who do. In Column 1, neither of the two interactions between gender and camps is significant. In Column 2, however, the interactions are more significant and much larger.^{11 }The contrast suggests that, as predicted, this relationship holds more consistently for families with more gender-biased preferences. Columns 3 and 4 show a similar pattern: no significant relationship for families in which the mother does not report wanting more boys, but a non-monotonic relationship in families that do.

Table 6 bolsters the conclusions in Table 4 and Table 5. The effect of vaccination camps is largest for families with stronger male-biased preferences. Another obvious test would be to explore whether the inequality-access relationship holds less strongly for families where there are two children of opposite gender close in age. Intuitively, in families where a female child needs vaccination at the same time as a male child, the relationship between access and inequality may be muted since there is a high fixed cost to travel. Unfortunately, the sample sizes are too limited (given the necessary closeness in age – a year or less) to test this hypothesis.

5

# Other Determinants of Vaccination Access

# A significant advantage of the analysis above is that vaccination camps are likely to operate in large

part as shocks to the availability of vaccination in the village. This makes them less obviously

correlated with existing conditions. In addition, their discrete nature makes the exploration of

non-monotonic effects relatively straightforward. The theoretical framework, however, is not specific

to vaccination camps (or even to vaccination). In general, we expect discrimination to be

non-monotonic in any measure of access. This section considers two alternative proxies for access to

1 Obviously, reported ideal number of male and female children is not a perfect measure of gender preferences. However, it should provide some proxy. In addition, it is certainly the case that the gender inequality in vaccination is larger for families where parents report wanting more male children (results available from the author). I have run similar regressions dividing the sample based on the average ideal gender ratio in the area the individual lives in, rather than th ir reported gender ratio, and the results are similar.

^{11}These are not wildly statistically precise, likely due to the smaller sample size.

19