Analysis of Data

What do the data from the post-communist region show about the possible effects of these structural variables on regime outcomes? Table 1.1 presents the results of ordinary least squares regressions of political regime scores on the four independent variables of interest, with robust standard errors and independent variables lagged by 2 years.

Many more specifications than the seven presented here are of course possible. With statistical analyses of this type, one must always be cautious about the dangers of presenting those models that conform to some preestablished notions held by the analyst. That is to say, we have to be wary of presenting findings that we “expect” or “want” and ignoring others. The findings presented in the seven models here are representative of the findings that we could present showing all specifications produced by alternative combinations of two or three independent variables per model (e.g. a model including only fuel income per capita and distance to Vienna); I strive to engage in no “cherry-picking.”

Each of the independent variables may exert an effect. The signs are all in the expected directions; there are no counterintuitive findings. Wealth and fuels production are both statistically significant at a demanding level in all specifications and the signs are in the expected directions. Higher levels of wealth are associated with higher democracy scores; more fuels production with lower scores. Distance from Vienna is statistically significant in three of the five specifications in which it is included, including the fully specified model (model 7). The sign is negative, indicating that greater distance from the West is associated with lower democracy scores. Religious tradition appears to exert effects as well. The variables for Orthodox Christianity and Islam are statistically significant at a demanding level. The signs on the coefficients are negative, indicating that higher proportions of Orthodox Christians and higher proportions of Muslims are both correlated with lower democracy scores. While Western Christian has an expected positive coefficient and is significant in model 5, it fails to reach significance in the fully specified model.

We may be reasonably confident in the effects of level of economic development, fuels production, proportion Orthodox Christian, and proportion Muslim. Proximity to the West may exert effects, but since it is not

Table 1.1 Regressions of extent of regime outcomes on hypothesized determinants

^<0.1, ^<0.05, ^<0.01

robust to all specifications, we must be more guarded about reporting its influence on regime type. Because Western Christian does not retain significance in the fully specified model, we have evidence that any influence it has on regime type is superseded by the other included variables.

The results suggest that conjecture based on conventional assumptions about the influence of economic development, oil dependence, geographical location, and predominant religious tradition would not have misled the observer. Wealthier countries have fared better than poorer ones and less fuel-dependent economies better than more fuel-dependent ones. Those that are closer to the West appear to have done better than those that are farther away. And, even according to the fully specified model, Catholic countries have done better than Orthodox countries, which have done better than Muslim countries.