includes more cases than other datasets, e.g. the World
Bank indicators of good governance which are available
only from 1996). The analysis focuses thus on 160 parliamentary
elections between 1984 and 2011 in 32 countries.5
I had to exclude the elections when the prime minister's
party was in a pre-electoral alliance or the party split
during the electoral cycle, because of the level of error
involved in disentangling its vote share from the vote share
of the alliance.
Each case is an election in country c at time t. Because
the elections when the prime-minister’s party was in a preelectoral
alliance or the party split during the electoral term
disrupt the country time series, I cannot use panel data
analysis, but still have to control for the dependence between
multiple elections in the same country. To do that, I
use the Huber-White standard error clustered according to
country.6 The dependent variable is change in vote share of
the prime minister's party since the last elections. Thus, I
run OLS regressions with clustered standard error. Similar
results are obtained when support for the incumbent is
estimated as the change in overall vote share for the coalition
parties in the cabinet. For elections when several
cabinet changes took place during the electoral term, I
considered the change in vote share for the prime minister
of the last cabinet formed before the elections.
The key independent variable is point change in good
governance in the election year, with high values indicating
an improvement in the quality of governance. It is
measured using the International Country Risk Guide
(ICRG, 2012) indicators of political risks in a country. Good
governance is calculated as the average of corruption (i.e.
excessive patronage, nepotism, job reservations, favor-forfavors,
secret party funding, and suspiciously close ties
between politics and business), bureaucratic efficiency (i.e.
whether the bureaucracy is autonomous from political
pressure, has an established mechanism for recruitment
and training, and can govern without drastic changes in
policy and interruption in government services), business
regulation (i.e. an assessment of factors affecting the risk to
investment: contract viability/expropriation, profits repatriation
and payment delays), and judicial effectiveness (an
assessment of the strength and impartiality of the legal
system and of popular observance of the law).7 The ICRG
indicators are based on the evaluations of the Political Risk
Group (PRG) editors using pre-set questions about the
political risk in each country. These are calculated using
identical rules for all countries, which makes the crossnational
and cross-time comparison possible, as opposed
to other indicators of governance based on public opinion
or expert surveys (e.g. World Bank Worldwide Governanc