A PROPORTIONAL HAZARD MODEL OF LEADERSHIP DURATION

We analyze a sample of 2,256 leaders from 167 countries to ascertain their durability in power and to account for their risk of losing power over time. A life table of the sample reveals that the risk or hazard of losing power decreased throughout leaders' tenure. The study then develops a proportional hazard model of leader longevity to account for sample heterogeneity and finds strong evidence that the hazard varies according to the leader's past time in power. Even after a number of leader and country characteristics are taken into account, the underlying risk of losing power is significantly lower in the second decade of rule than in the first decade. Low risks are linked to nonelectoral regimes which are found typically in the developing countries. It is important to distinguish system instability from length of leaders' tenure. We cannot directly measure leaders' skills and abilities. Thus, we cannot determine whether longevity in power is really a selection process over time. The length of time a leader has been in power predicts well how long that leader will continue to hold power.