tems, majoritarian, mixed and proportional. Each observation corresponds to a
particular legislature in a particular country in the 1960-98 panel. The statistics
in the table are computed from pooling together these observations. Most ob-
servations are either classified as majoritarian or proportional. Comparing these
two more extreme electoral systems, we immediately find large differences in line
with all the predictions of the theory. Majoritarian electoral systems have a less
fragmented party system, a smaller incidence of coalition governments, a larger in-
cidence of single party government, and smaller government spending, compared
to proportional systems. The few mixed electoral systems display outcomes in
between the polar types. Interestingly, 63% of the observations from majoritarian
countries exhibit single-party government, whereas the incidence for proportional
countries is only 17%. Taking our model literary, the residual presence of coalition
governments under majoritarian elections might reflect either the multiplicity of
equilibria or heterogeneity of districts in a subset of countries. Note also that
the standard deviation of the type of government is large within each class of
electoral rules, suggesting that there may also be independent shocks to coalition
formation, a feature not present in our model (see futher below).
As discussed above, however, the theory has additional and precise predictions
that go beyond such simple cross tabulations. Moreover, in Table 1 we are not
controlling for other differences across countries or time periods. In the remainder
of this section we show that more sophisticated statistical analysis confirms the
simple message of Table 1 and provides support for the precise predictions of the
theory.
6.2. Cross sectional estimates
We first confront the predictions of the theory with the cross country-variation in
the data. We use both our cross sections: the one for the 1990s includes a larger
number of countries for a single decade; and the one for 1960-98 includes a smaller
number of countries for (at most) four decades.
First, we study the predictions of the theory concerning the political variables,
asking how party structure and type of government vary with electoral rules.
Second, we study the economic predictions, asking how the electoral rules influence
government spending. Since we have relatively few observations, we keep the
specification with regard to other variables as parsimonious as possible.
37