The specifications always include other time-varying determinants of govern-
ment spending, such as the output gap (to measure cyclical influences on govern-
ment spending), openness to international trade, and the proportion of the elderly
in the population. Finally, to reduce collinearity among the electoral rule vari-
ables, we omit the indicator variable for mixed electoral rules, including only the
indicator for plurality rule (maj) and the measure of district magnitude (district).
Columns 1 and 2 of Table 5 show the reduced form of spending on the electoral
rule variables. As expected, district magnitude has a positive and strongly signif-
icant coefficient (larger districts implying more spending). But the coefficient on
plurality rule is insignificant.
The remainder of the table displays the structural estimates. Columns 3 and
4 show the GMM estimates of the type of government (coalition or single party
majority) on government spending for the legislature panel, where the type of
government is endogenous and the electoral rule variables of column 1 are used
as additional instruments. Columns 5 and 6 show the fixed-effect, instrumental-
variable estimates; here the lagged type of government and the electoral-rules
variables are used to instrument for the current type of government.
The estimated coefficients all have the expected sign: coalition governments
spend more, single party majority governments spend less. The effect of coalition
governments is precisely estimated and large in value. Given the coefficients on
lagged spending, the implied long-run effect is 10% of GDP in column 3, and 6%
of GDP in column 5, thus of the same magnitude as the cross sectional estimates.
The effect of single-party government is smaller and much less precisely estimated.
However, note that the over-identifying restrictions now can be rejected for the
fixed effect estimates (time periods measured in years), while they cannot for the
GMM estimates.
All in all, the panel estimates are a bit more fragile than the cross-sectional
estimates, but that is perhaps to be expected given the paucity of real-world-
reforms in our data. On the whole, they give additional support to the main
insights from the theory: proportional electoral rules are associated with more
fragmented party systems, more coalition governments, and these spend more
than single-party majority governments.
7. Concluding remarks
This paper investigates the effects of electoral rules in parliamentary democracies.
In line with the recent paper by Bawn and Rosenbluth (2002), we study a formal
44