useful as a rough guide to assess model fit, but should probably not
be used for model selection (see Long, 1997). Nevertheless, the
measures do indicate a relatively good fit for themodel. Almost 80%
of the cases are correctly predicted with this specification (Count
R2 = 0.798).
A few variables stand out in Table 3. First, notice that having a
woman as a head of household plays a significantly positive role in
determining whether a man or women has the predominant
responsibility to attend meetings. Households headed by a woman
are about 13 percentage points more likely to have women attend
meetings than households that are headed by men (significant at
the 0.01 level) holding all other variables at their means. On the
other hand, households headed by a woman are about 15
percentage points less likely to have men responsible for meeting
attendance. This result seems obvious. If household self-identify as
women they will be more likely to have women attend FUG
meetings and less likely to have men attend FUG meetings.
An important result is seen in the restrictions to FUG
membership. These restrictions are in terms of fees or time
commitments to join the FUG. Notice that this variable disproportionately
disadvantages women’s participation. A marginal change
in the proportion of households that are excluded by entrance
requirements is predicted to decrease the probability of nobody
with responsibility to attend meetings by almost 26 percentage
points (significant at the 0.05 level), holding all other variables at
their means. However, these gains are almost completely realized
by an increased probability of the husband’s responsibility to
attend meetings. The predicted probability of husband’s participation
increases by 25 percentage points (significant at the 0.01
level), holding all other variables at their means. The predicted
probability of equal participation increases by about 10 percentage