MEASURING THE EFFECT OF GOVERNMENT INTERVENTIONS: EVIDENCE FROM RESTRUCTURING THE U.S. NUCLEAR POWER SECTOR

摘要

The predominant empirical strategy when measuring the effect of endogenousgovernment interventions is to represent the intervention by a dummy variable and toinstrument it with either exogenous shocks directly, or predicted values from a probit model.These modelling strategies assume that if the net instrumental effect changes sign after theimplementation of the intervention, then one would expect the post-intervention conditionsto revert to their original state. In practice, market conditions are sometimes subject topersistence with the probability of reversal being close to 0 in the time periods following anintervention. We evaluate government interventions in the US nuclear power sector bycreating instruments for interventions based on proportional hazard models. This procedurereduces the instrument variability to 0 following the intervention when the likelihood ofreversed conditions is low. We find that divesture, with the vertical unbundling of utilities,reduces the unavailability factor substantially more than previously found. Contrary toprevious studies, we find that wholesale market creation has no significant impact on outagelevel.