18
the predicted values of the dependent variable of the first-stage equation to estimate the
impact of aggregate inequality on the aggregate levels of turnout in the context of an OLS
with robust standard errors estimation.
To control for the possible effects of the electoral system we include in the right
hand side of the equation a measure of the disproportionality of the electoral system.
Specifically, we utilize the Gallagher least index of electoral system disproportionality
measured by the square root of half the sum of the squares of the difference between
percent of vote and percent of seats for each party. In addition, the specification contains
a number of variables previously identified as relevant determinants of cross-national
differences of turnout levels (Gray and Caul 2000). These include the size of the voting
age population (log), the legal age at which people are allowed to vote, the level of
economic growth, the share of seats in obtained by labor oriented Social-Democratic
parties, whether elections are held on a holiday, whether there is a system of automatic
registration, and whether the system of representation is unicameral or not. The
estimation results are presented in Table 1.
All other variables in the model display the expected signs and levels of
significance. Holding elections on holidays inhibits turnout, whereas systems of
automatic registration promote it. Moreover, everything else equal, larger electorates (in
relation to the total population) are associated with higher levels of voter turnout in a
country. Finally, the coefficient for the variable measuring electoral systems is worth
causal chain includes already several steps, this criticism is well taken. In replicating the model with
alternative instruments whose direct link to electoral turnout is arguably less traceable, such as the overall
dependency ratio (pensioners + unemployed over working age population) we are able to produce very
similar results.
Convention