However, since this approach is novel, a sensitivity analysis on the aggregate national crime victimization costs was conducted. Weighting the data with series victimizations without imposing the maximum of 10 victimizations in 6 months would raise the estimated tangible crime costs by 7.5 percent, quality of life costs by 12 percent, and total costs by 11 percent. Weighting the data with the most recent victimization in a series counted as the only victimization would lower tangible costs by 4 percent, quality of life costs by 8 percent, and total costs by 7 percent. Thus, aggregate cost estimates are not overly sensitive to decisions about how to handle series victimizations.

The costs per rape are quite sensitive to choice of weighting for series victimizations since the quality of life losses for rape are computed per victim, then allocated across victimizations. The cost of a rape victimization is estimated to be $87,000, while the cost of being a rape victim is $110,000. The latter figure is probably a more useful estimate, since the quality of life losses (the largest component of rape costs) are estimated from jury awards to victims (not victimizations). Conversely, because of this study's methodology using jury awards for victims, the national quality of life losses due to rape are unaffected by the decision about how to weight series crimes.

Monetary valuation of intangible losses

Quality of life losses for fatal crimes. The largest cost element for all violent crimes is lost quality of life and related fear, pain, and suffering. It may also be the cost item with the highest degree of uncertainty. For fatalities, the study estimated lost quality of life using a $2.7 million value of saving an anonymous life derived from a synthesis of almost 50 published values (Miller, 1990), adjusted for the difference in expected lifespan of crime victims (by crime type) versus the average injury fatality. Miller (1990) finds the standard deviation of the quality of life values across studies is 30 percent of the mean value for a statistical life. Although one cannot use this information directly to estimate the standard errors of the composite samples, if one assumes that each of the 50 studies used to generate the $2.7 million estimate were actual samples from the population of values, it would imply a 95 percent confidence interval of + $1.3 million.

Viscusi (1993) reviews the same literature but does not adjust the values for obvious sources of variation (e.g., differences in the discount rates used in calculations). He favors a larger value of $3 million to $7 million. Using a $5 million value would raise the estimated quality of life lost due to fatalities and the estimated total cost of fatal crime by $93 billion (to $184 billion).

Alternatively, one could argue that the typical crime victim is not the same as the typical individual used to generate the lost quality of life estimate for fatal injuries. Although estimates are generally based on nationally representative samples of workers or consumers, it is known that murder victims are disproportionately young, nonwhite, and in the lowest income classes (National Research Council, 1993). They may also be more likely to display high tolerance for risk taking (which might reduce their implied valuation of