empirical effect. (See Table 1.) We were optimistic that many
of the village variables would help to explain differences between
villages in contraceptive choice; in the end, however,
we found that only four had any effect.
We model contraceptive choice as an outcome of individual,
household, and community variables, along with selected
interactions:
Pr(My- = k)
log =akC.+f3kX;+8kCXr+J..l· (I)
Pr( M u =I) 1 ~ 1 ~ '
The dependent variable for the analysis of temporary
method choice is the log odds that couple i in village j