of government spending, under coalition government. This direct effect of the
electoral rule is of ambiguous sign, however, because by Proposition 5 higher
spending targeted towards groups represented in government may be more than
offset by lower spending on opposition groups.
6. Empirical evidence
In this section, some of the model’s predictions on the causal chain from electoral
rules, via political outcomes, to economic policy outcomes, are taken to the data.
We use both cross-sectional and panel data for about 50 democracies in the post-
war period. The only economic policy variable is overall government spending.
6.1. The Data
Our samples As the model in the preceding sections deals with decision-making
in a parliamentary democracy, we limit the empirical investigation to democracies
with this form of government. Here, we follow Persson and Tabellini (2003), who
use the existence of a confidence vote for the executive as the main basis for
classifying different forms of government.
We use data from two different data sets, each of which combines information
from a variety of second-hand sources and first-hand constitutional documents.
For many of the variables, we rely on a broad cross sectional data set assembled
and presented in detail in Persson and Tabellini (2003). For some measures of
political institutions and outcomes, we use another data set resulting from a col-
Iaborative data collection effort jointly with political scientists from Âbo Akademi
(see Lundell and Karvonen, 2003).
Our first data set is a cross section of about 50 parliamentary democracies in
the 1990s, where each observation is the average of annual data over the period
1990-98. Here, democracies are countries labelled as free or semi-free according to
the surveys published by Freedom House. The so-called Gastil indexes of political
rights and civil liberties (gastil) vary on a discrete scale from 1 to 7, with low
values associated with better democratic institutions. We include a country in the
sample if the average of these two indexes in the 1990-98 period does not exceed
5.
Our second data set covers the period 1960-98 for a smaller group of about
40 parliamentary democracies. Here, we mainly rely on the Polity IV data for
the classification into democracies, as this data set goes farther back and is more
32