In model (4), however, we allow a little longer time horizon in the assessment of regime
types. Apart from looking at the immediate impact of current regime type, this model includes
the previous regime type as a set of regressors.
This change in time perspective also changes
the results. Apart from a positive effect of having had a dominant party system in the previous
regime period as opposed to being a monarchy, no democracy enhancing effects of “multi-
partyness” remains. What matters in this time perspective is, to simplify, whether the
authoritarian regime is a monarchy or not. Recall that monarchies rarely change, and when
they do they usually oscillate back and forth between being a pure and an amalgam monarchy.
The effect of being a monarchy is thus more or less the same whatever time perspective we
apply. And this effect is clearly negative—although only significantly different from the
military, one-party and the military one-party amalgam regimes (apart from dominant regimes
and democracies). Thus, in the intermediate time perspective what matters for democracy in
terms of authoritarian regime types is mostly the fact that monarchies are highly resistant to
democratization.
Finally, we would like to say a few words about the level of explained variance achieved
when regime types are taken into account. As previously mentioned, the full control model
only accounts for some 10 percent of the variance in the yearly democratization rate (Teorell
and Hadenius 2005). As Table 3 makes clear, however, this figured is pitched to a nearly 30
percent when the regime typology is added. True, this may be an effect in part of having
added a set of explanatory variables that in the causal chain are fairly proximate to the
phenomenon we are trying to explain. Yet, it is not a reflection of the degree of democracy, as
we control for that. Hence, we would argue that our findings indicate that taking type of
21
Obviously the time perspective for this assessment varies by country and year, dependending on the amount
of time that has ellapsed since the last regime change occures. Lagging all regime variables by 5 years, however,
yields almost identical results. Also note that some countries lack ”previous regimes” in the sense that they have
had the same regime since the start of our observation period (in 1972). For these countries we have assigned the
previous regime type to be the same as the current.
29
Convention