Incapacitation

This section presents incapacitation estimates derived from modeling exercises in which various statistical controls were used to overcome the inherent ambiguity between incarceration and crime rates described before. It has been noted that incapacitation estimates vary enormously by source. However, sometimes the differences in estimates are exaggerated because estimates calculated under different metrics are directly compared.

Historically there have been four conceptually different incapacitation measurement systems. The earlier measures were of the type provided by Greenberg and the Shinnars, previously Figure 1 mentioned. In this approach incapacitation is measured by the percent that crimes prevented represent relative to crimes committed, or some variant of this basic calculation, often referred to as the incapacitation effect. More recent measures of incapacitation focus on the number of crimes prevented per offender in prison, using the average or median offending rate of offenders in prison as the basis for that calculation. A third measure favored by economists measures incapacitation as an elasticity, that is, the percent change in the crime rate following a 1 percent Table 1 SOURCE: Marvell and Moody 1994, p. 132, Table V; Levitt 1996, p. 345, Table VIII; Zimming and Hawkins 1995, pp. 116-118, Tables 6.5 and 6.6; and Canela-Cacho et al. 1997, pp. 150-151, calculated by the author based on results of Tables 4 and 5, and model described in pages 137-142. change in the incarceration rate. Finally, a fourth useful measure attempts to measure what could be described as marginal incapacitation, and that is the number of crimes prevented by the incarceration of additional inmates in the event of an expansion in the prison population.

While these four measures are obviously interrelated, numerically they are not equivalent. For example, Levitt provides both elasticity and marginal incapacitation estimates based on the same data and model. The elasticity estimate is fairly small, -.31, suggesting that for a 10 percent increase in incarceration rates the crime rate would be about 3 percent below where it otherwise would have been. To some this would be evidence that incapacitation effects are negligible; yet, the marginal incapacitation estimate that accompanies this elasticity value is 14.9 crimes prevented for each additional inmate joining the prison population, evidence to many of large incapacitation effects.

When comparing incapacitation estimates one needs to make sure that the incapacitation measures in question belong to the same type, otherwise the comparison would be misleading. In this entry the incapacitation estimates discussed below belong to the type previously defined as marginal incapacitation measures (Table 1) or to measures of the elasticity of incapacitation.

The first two sets of estimates presented in Table 1 (Marvell and Moody; Levitt) were obtained using similar methods and data sets. They both rely on a statistical technique called regression analysis to assess changes in the crime rate as a result of changes in the incarceration rate, controlling for other factors that could influence that relationship, such as the demographic composition of the population and levels of economic well being. The data informing these two models spans the period 1971 to the early 1990s and includes all states. The Zimring and Hawkins incapacitation estimates (1995) also rely on the relationship between crime and incarceration rates but in a nonregression estimation context, and are based only on California data covering the period 1981 through 1990. These three sets of estimates are comparable in that they provide marginal incapacitation estimates by crime type and for the aggregation of all crimes, and they are all based on aggregate crime and imprisonment rates.

The last set of estimates in Table 1 are an extrapolation specifically prepared for this entry based on results reported by Canela-Cacho and colleagues. The latter estimates are based on a completely different methodology relying on individual offending rates to calculate incapacitation effects rather than aggregates of crime and incarceration rates. The estimates are based on California data and apply to the same time period considered by Zimring and Hawkins. In addition, the individual offending rates used in this approach derive from surveys of prison inmates where offenders have reported their crimes.

Some commonalities in the four sets of estimates are readily apparent. The marginal incapacitation estimates are all substantially higher for burglary than for robbery, and this applies generally to property crimes vis-à-vis violent crimes. In addition, the size of the estimates for marginal burglary incapacitation is virtually the same in the estimates by Levitt, Zimring and Hawkins, and Canela-Cacho and colleagues.

However, the differences across estimates are striking concerning marginal incapacitation for violent crimes. The Marvell and Moody estimate is 2.5 times that of Zimring and Hawkins, whereas Levitt's estimate is 4.6 times that of Marvell and Moody. For robbery, similar differences exist among the four sets of estimates. These differences raise concerns since the Levitt and Marvell and Moody estimates rely essentially on the same data and apply the same genre of statistical models. The estimates of Zimring and Hawkins and Canela-Cacho and colleagues are based on completely different methodologies but they both apply to California during the same time period, and again, with regard to robbery, the estimates are dramatically different. Perhaps the one conclusion we can reach is that substantial uncertainties remain in the size of incapacitation effects, despite some impressive methodological advances in the measurement of incapacitation.

The crime drop throughout the United States that began in 1992 has given new impetus to the study of incapacitation effects, as scholars are busily trying to explain what contribution imprisonment played, if any, to this large and for the most part unexpected decline. Not surprisingly, in a context where homicide and robbery rates have experienced declines of over 50 percent in seven years, the new rounds of incapacitation estimates tend to be larger than estimates based on data prior to 1992, even when the same methodology is applied. For example, Spelman (2000a) redid Levitt's analysis expanding the data through 1997 and reports an increase in the incapacitation elasticity of 41 percent between 1973 and 1997.

Similarly, some recent work examining the incapacitation effect for homicide reports elasticities of -1.5 to -1.9 (Marvell and Moody, cited in Rosenfeld). Rosenfeld analyzed the decline in homicides in the period 1990–1995 and concluded that the high elasticity estimates just cited are in agreement with his finding, obtained independently, that the homicide rate would have been 28 percent higher absent the incapacitation effect generated by an average annual increase of about 67,000 prison inmates between 1990 and 1995.

It should be noted that the crime types for which incapacitation estimates have been attempted are limited to violent offenses and to some property offenses. No one seriously has entertained the notion that putting one drug dealer behind bars would prevent a number of drug transactions; the incarcerated offender would be easily replaced by someone else in the streets as long as the demand for drugs continues unabated. Such replacement effect is unlikely to apply to offenses like robbery or burglary, except in the context of co-offending where two or more individuals team up to engage in criminal acts. Reiss pointed out that the incarceration of one of the members of a group of two or three offenders acting together need not have an incapacitative effect as the remaining free members of the group could recruit a new member to replace the incarcerated peer.

There are two additional situations that could further negate or at least diminish incapacitation. The first one relates to crime desistence, a phenomenon exhaustively documented in crime research. Simply put, as a result of aging, many offenders stop committing crimes, and thus past a certain point their incarceration yields no incapacitation benefits.

In the second case, incapacitation would initially be effective but would eventually become counterproductive, if as a result of an episode of incarceration, an offender upon release evolves into more serious crimes or engages in the same criminal behaviors but at substantially higher rates. In this instance prisons would have a criminogenic effect, preventing some crimes at first but at the expense of contributing to more serious or to a higher number of crimes in the future.

The incapacitative estimates presented before only indirectly attempt to control for replacement or desistence of offenders following incarceration, and none allow for the possibility that prisons are indeed criminogenic. These omissions further compound the uncertainties surrounding the available incapacitation estimates and show the need to develop yet better measuring techniques and substantially richer data sets.