Pierre F. Landry: Decentralization and Regime Transformation
18
Measuring Political Regimes
Measures of political regimes are based on the updated dataset of democratic and
authoritarian regimes around the world by
Cheibub and Gandhi
(2004).
8
While there has
been considerable debate among political scientists about how best to measure changes
among political regimes over time, the efforts of Przeworski et. al. (2000) and Cheibub &
Ghandi (2004) provide the clearest measures of authoritarianism and democracy on a
global scale. Once we accept Schumpeterian definitions of democracy of the authors of
Democracy and Development, the coding of the data—although labor intensive—is
relatively straightforward. The result is a virtually complete dataset of political regimes
around the world, including micro-states.
The key variable of interest is whether a regime is authoritarian (coded 1) or
democratic (coded 0). It is important to note that these measures capture central political
institutions, not the nature of politics at the state or local level. Thus, regimes that
introduce local elections with multiple political parties where incumbents actually lose
power are considered authoritarian. This is the case—for instance—of the KMT regime
on Taiwan before 1987 where local competitive elections between the KMT and
independent candidates were routine.
Table 2 confirms that the nature of the bias is political rather that economic.
Probit estimates where the dependent variables take the value of 1 if fiscal
decentralization measures are available clearly indicate that federal democracies are
overrepresented in the dataset, whether one uses revenue or expenditure data. Given the
8
For the details regarding these coding rules, see Przeworski et al. (2000). I am grateful to my colleague
José A. Cheibub for graciously sharing his updated dataset.
Convention