Incarceration of criminals reduces crime through two main channels: deterrence and incapacitation. Because of the simultaneity between crime and incarceration-arrested criminals increase the prison population-it is difficult to measure these effects. This paper estimates the incapacitation effect on crime using a unique quasi-natural experiment namely, the recurrent collective pardoning between 1962 and 1995 of up to 35 percent of the Italian prison population. Since these pardons were enacted on a national level, unlike in Levitt (1996), we can control for the endogeneity of these laws that might be driven also by criminals' expectations: it is optimal to commit crimes shortly before a collective pardon gets enacted. This is the deterrence effect, which, if not properly controlled for, would bias our instrumental variables estimates toward zero. The incapacitation effect is large and precisely estimated. The elasticity of crime with respect to prison population ranges, depending on the type of crime, between 0 and 49 percent. These numbers are increasing during our sample period, which suggests that habitual criminals are now more likely to be subject to pardons than in the past. A cost-benefit analysis suggests that pardons, sometimes seen as a short-term solution to prison overcrowding, are very inefficient. The estimated marginal social cost of crime is more than two times the cost of incarceration. Our very conservative estimate of the total net social cost due to the July 2006 pardon is equal to 2 billion euro.

Stock-market crashes tend to follow run-ups in prices. These episodes look like bubbles that gradually inflate and then suddenly burst. We show that such bubbles can form in a Zeira-Rob type of model in which demand size is uncertain. Two conditions are sufficient for this to happen: A declining hazard rate in the prior distribution over market size and a positively sloped supply of capital to the industry. For the period 1971-2001 we fit the model to the Telecom sector.

Stock-market crashes tend to follow run-ups in prices. These episodes look like bubbles that gradually inflate and then suddenly burst. We show that such bubbles can form in a Zeira-Rob type of model in which demand size is uncertain. Two conditions are sufficient for this to happen: A declining hazard rate in the prior distribution over market size and a convex cost of investment. For the period 1971-2001 we fit the model to the Telecom sector.