25
Entering all my variables in Model 2 yields somewhat disappointing results for
my institutional and partisan variables. The R-square only increases with about .036 and
my model therefore adds a mere 3.6 % of explanatory power to Model 1. Still, the fact
that the addition of the institutional and partisan variables makes a difference for the
explanatory power, is a positive result.
Bicameralism has no significant effect on reforms, and the coefficient is actually
positive, contrary to my hypothesis 1. The coefficient for federalism is negative as
predicted, but its p-value (.394) does not approach any conventional level of significance.
Model 2 therefore gives no support for hypothesis 2. Of the two variables measuring
partisan unity, presidential-legislative congruence has a significant positive effect at .002.
Thus, when the president has a higher level of support in the legislature, reforms are more
likely. This supports hypothesis 3 and contradicts the results by Remmer (1998) who
found that presidential power had a negative effect on reform. Consequently, she argued
that weak presidents are more likely to push for reforms in order to increase their power.
On the other hand, she measured the electoral vote for the president as a measure of
influence, where I measure actual partisan control. Although the vote for the president
may influence his/her popular support, it need not translate into legislative support.
On the other hand, my second measure of partisan unity, legislative cohesion, has
a significant effect on structural reform, but in the opposite direction than predicted. Thus
the statistical analysis rejects hypothesis 4. The coefficient for veto powers is positive as
expected but not significant. I therefore have no support for hypothesis 6. Similarly, the
coefficient for decree powers is positive, as expected, but its p-value is only .456.
Convention