We attempt to resolve a recent controversy in the study of cabinet terminat
ions pertaining to the shape of hazard rates. On the one hand, Warwick (199
2b) provides evidence that cabinets are more likely to terminate the longer
they are in office. Alt and King's (1994) analysis, on the other hand, sug
gests that hazard rates are constant over the life-time of a cabinet. This
issue is of particular theoretical importance, since a constant hazard rate
would add support to the nonstrategic model of cabinet termination due to
Browne et al. (1986) while an increasing hazard rate would seem to favor Lu
pia and Strom's (1995) strategic approach. By applying a semi-parametric co
mpeting risk approach to data on cabinet durations, we are able to show tha
t through its use of theory-based censoring the previous literature in effe
ct analyzed only one mode in which cabinets terminate: the case where one c
abinet is replaced by another without a new election. Once cabinet terminat
ions that lead to chamber dissolutions with subsequent elections are analyz
ed directly, we can show that they are governed by a very different stochas
tic process. Hazard rates are not flat as in the case of replacements, but
increase over the life of the government. Further the covariates governing
replacement terminations fail to explain dissolution terminations. These fi
ndings add support to the strategic approach suggested by Lupia and Strom.