Political effects of electoral rules Table 2 reports our estimates of the polit-
ical effects, part a for the 1990s cross section, part b for the 1960-98 cross section.
Since the layout of both tables is the same, we comment on them together.
First, we ask how the electoral rule shapes the party structure, as measured
by party fragmentation (party_frag) and by the number of parties (nparties).
We use a slightly more parsimonious specification in the longer cross section in
Table 2b, where we have fewer observations, but we always control for UK colonial
origin and for country size. Ethno-linguistic fragmentation is only available in the
1990s, so we control for it in Table 2a. Columns 1 and 2 of both tables display the
results. As expected, plurality rule (as measured by maj) is associated with less
party fragmentation and fewer parties. Mixed electoral systems also seem to lead
to less fragmented party systems (compared to the proportional default), although
the effect is marginally significant only in the larger sample from the 1990s. Also as
expected, larger electoral districts (typical of proportional elections) are associated
with a more fragmented party system and a larger number of parties. Finally,
electoral thresholds influence the party structure as expected (higher thresholds,
fewer parties), but the effect is only significant for the number of parties.
Next, we estimate a reduced form for the type of government. Here, besides
the features of the electoral rule and the other regressors, we also control for
the constitutional features mentioned above: bicameralism, investiture vote, and
constructive vote of no confidence.22 The results are displayed in columns 3-5 of
Tables 2 a and b, with coalition governments, single-party majority governments,
and number of parties in government, as the dependent variable. Only some
of the electoral rule variables are statistically significant, and the results vary
according to the sample and the type of government indicators. This might be
due to some correlation between some of these indicators (in particular between
district magnitude and the indicators for the electoral formula). Nevertheless,
the overall effects go in the expected direction. Larger electoral districts raise
the incidence of coalition governments and the number of parties in government,
and tend to reduce the incidence of single party-majority government (the last
effect is only significant in the 1960-98 cross section). Plurality rule raises the
incidence of single-party majorities. Among the other constitutional variables,
22We do not include investiture and constructive in the regression where single is the de-
pendent variable, because these two variables are expected to reduce the incidence of minority
government, and thus should have no effect on single-party majority, given the party system -
see the dicussion above. We also omit bicameralism from the longer cross sections for 1960-98,
because it is not available over such a long period.
38