future quality of life), as demonstrated by higher rates of alcoholism, drug abuse, drug trafficking, and other forms of criminal behavior. On the other hand, more than half of all murders are domestic or involve children; society also might decide for equity reasons that differences in value of life estimates across individuals should not be used for policy analysis. Regardless of which lost quality of life value is determined to be appropriate, a sensitivity analysis can be conducted to determine the effect of varying this parameter. For example, if one values lost quality of life at the lower bound of the 95 percent confidence interval (about $500,000 instead of $1.9 million), the estimated total cost of fatal crime would fall from $91 billion to $54 billion.

Quality of life losses for nonfatal crimes. The nonfatal quality of life estimates come from regression analysis of jury verdicts. For nonfatal injuries, the standard error range around the log-linear regression estimates is +39 percent for assault and +29 percent for rape. Translating back to linear estimates of the quality of life yields extremely large confidence intervals.

As an alternative method of checking this study's quality of life estimates, they were compared to a completely different methodology described in Miller, Pindus et al. (1995). This alternative method is based on physician ratings of the loss of functioning that typically results from different injuries and surveys that reveal how people rate different functional losses relative to death. These ratings are then applied to the same $2.7 million statistical value of life used above to arrive at a monetary equivalent for the nonfatal injury. Using the values derived from physician judgments would lower the estimated quality of life losses by just 6 percent. This close match masks substantial disagreement over the losses by treatment status, with the jury verdict estimates lower than the physician judgment estimates for hospitalized cases and higher for other medically treated cases.

The concern that the fatal victim population might not have the same risk tolerance or income as the U.S. population as a whole is not as relevant for nonfatal injuries. This study used estimates of wage rates directly from NCVS, and there is less reason to believe that victims who sue for damages are different from the population of victims in general.

Until there is a larger data base on jury awards or more confidence in alternative methodologies, this is the best this study is able to do. Based on the size and uncertainty of the estimates, further research on the lost quality of life due to crime victimization appears to be a priority.

Lifestyle changes due to victimization. The psychological effects of victimization may result in lifestyle changes such as moving to a new neighborhood, switching jobs or occupations, buying more protective devices, changing daily routines, and simply enjoying life less. In some instances, victims may subsequently attempt suicide. Since this study already estimated pain, suffering, and lost quality of life, many of these lifestyle effects were incorporated into the estimates. Juries presumably take such effects into account when awarding damages for pain, suffering, and lost quality of life; and individuals also take them into account in making decisions on which risk-reducing actions to take.

Of course, it is impossible to know empirically whether or not juries take all of these costs into account. One sensitivity analysis was conducted to look at the maximum medical costs associated with suicide attempts precipitated by criminal victimization. The order-of-magnitude of these losses can be bounded for sexual assaults using excess suicide attempt rates from Golding and George (1993) and Kilpatrick et al. (1992) and for other child abuse using rates from the literature review in Daro (1988). With the costs per suicide and per hospitalization for suicide attempts from Miller, Pindus et al. (1995), suicide-related medical costs would be $107 per rape and $165 per child abuse case. Although this equates to $300 million annually, it is a very small percentage of estimated costs. Because the associations between crime and suicide are uncertain and these costs might already be implicitly taken into account in this study's quality of life costs, they are excluded from the estimates in this document.

Comparison with prior estimates

Estimates of the cost of crime have a long history, dating back as early as the 1931 Wickersham Commission. A review of some of these studies is contained in Gray (1979). These earlier studies tended to report broad aggregates and focus on direct costs such as victims' out-of-pocket costs, the criminal justice system, the cost of private security, and the value of illegal wagers. Most of the earlier studies did not go beyond tangible costs. Beginning with Thaler (1978), there have been several attempts to include indirect costs. Thaler (1978) and subsequent studies estimated the difference in housing prices that can be attributed to differential crime rates, thus inferring homeowners' willingness to pay for reduced crime. Although in theory these approaches