Crime and destiny: patterns in serious offenders' mortality rates.

• Author: Tremblay, Pierre

Introduction

Common sense would expect habitual offenders to increase their odds of premature death. Hobbesians (and moral exhortations) delight in such anticipations: "In each human psyche, there is a primal awareness of a possible world of nightmarish violent anarchy ... sufficiently vivid to permit many to sense that to commit a crime is to invite unforeseeable threats against their own survival" (Novak 1986: 85). Current research does not dispute that offender death rates are higher but dismisses the conventional explanation. Instead of arguing that offenders' criminal activities themselves explain their premature deaths, the assumption is made that offenders' higher mortality rates are a consequence of "unhealthy proximal behavior" (e.g., alcohol abuse; see Laub and Vaillant 2000: 16) or collateral imprudence in lifestyle (e.g., accidents unrelated to crime activities). This theme is best developed in Gottfredson and Hirschi's (1990) general theory of crime.

A key proposition contained in this theory (outlined in Hirschi and Gottfredson 1994) is that individuals lacking self-control are more likely than others to engage in behaviour that is harmful both to others (crimes included) and to themselves (behaviour resulting in premature mortality included). Personality traits that define low self-control include impulsivity or short-sightedness, lack of persistence or tenacity, self-centredness, low tolerance for frustration, and a physical or a sensation-seeking disposition. Such personality traits are assumed to remove the moral constraints that prevent crime, to distort or compromise the calculus of decision making, and to increase, as a result, the probability that offences will be committed (Arneklev, Grasmick, Tittle, Bursik 1993: 29).

The basic proposition, then, is that higher mortality rates among offenders are not a consequence of the offences they commit but rather a reflection of the kinds of persons they are (individuals lacking in self-control). The first implication is that a significant proportion of excess mortality can be explained by offenders' involvement in a wide range of imprudent or reckless behaviour (not necessarily criminal), and that this, in turn, increases their chances of being involved in accidents (some of which will be lethal). Higher mortality rates among offenders would not require a subcultural or social interpretation but could, instead, be explained away as the aggregate outcome of unintended and somewhat random events triggered by individuals lacking self-regulation. Thus, the simple bivariate correlation between offending and premature death should vanish when self-control predisposition is controlled for. To the extent that excess offender mortality rates do, in fact, disappear under such conditions, the relevance of an occupational-hazard model of crime as a way of life is severely undermined.

A second implication derived by Hirschi and Gottfredson (1994) is that offenders will exhibit higher mortality rates, whatever the particular "cause of death": they will experience higher homicide rates because low self-control increases the odds that disputes will escalate into deadly conflicts; they will also exhibit higher overdose rates because low self-control increases the likelihood of addiction and the imprudent injection of drugs once addicted. Finally, they will also have higher suicide rates because low self-control undermines individuals' coping abilities. Since this low self-control predisposition is general in scope, it increases the odds of premature death, whatever the particular cause of death examined. One should, therefore, expect that the main difference between offenders and non-offenders will be found not in the kinds of death they experience, but in their relative mortality rates.

Since self-control attributes are viewed as latent continuous variables--some individuals being more impulsive, more insensitive, more self-centred, more physical than others--and since the more persistent offenders are expected to exhibit lower self-control than occasional offenders, it also follows that these more serious offenders should experience higher mortality rates. Because self-control is conceptualized as a mix of stable individual traits, the gap in mortality rates between offenders and non-offenders should remain roughly constant throughout the life course. Evidence that mortality rates increase over time should not be viewed as an outcome of cumulative strain but as the combined effect of age and self-selection. Age, here, is assumed to have an independent (positive) effect on mortality rates, as well as an independent (negative) effect on offending (Hirschi and Gottfredson 1983).

Residual increases in age-specific offender mortality rates, on the other hand, could be accounted for as artefacts of a self-selection bias: offenders located in the mid-range of the self-control continuum are likely to drop out of crime earlier than those located in its lower range. An interesting feature of this general-hazard model is that it allows for the possibility that premature death cannot only interrupt the course of a "crime career" but also occur after offenders cease or decelerate the frequency or the seriousness of their illegal activities. This is so because death outcomes are not viewed as intrinsically embedded in offending patterns but as indicative of an underlying predisposition to engage in a wide range of behaviours (only a subset of which could be qualified as crimes) whose consequences may be harmful for themselves, as well as for others.

Previous research

A systematic test of this general delinquency-hazard model has yet to be conducted. Relevant mortality-rate data sets for offenders are hard to obtain. (1) Large cross-sectional samples or extensive follow-up studies are required to analyse differential mortality rates of juvenile or adult offenders. A number of investigations, however, provide direct evidence on the probability that serious offenders will experience premature death. Yeager and Otnow-Lewis (1990) undertook a follow-up study analysis of 188 juvenile offenders (97 boys and 21 girls) released from a Connecticut detention centre. Of these, 188 juvenile offenders, 7 died before reaching their 25th birthday (6 males, 1 female). All died violently: two by overdose at age 21; two in car accidents between ages 16 and 21; another in prison, murdered at 23; and the remaining male, killed by the police at age 15. The last case involved a young woman who committed suicide at age 21. The prevalence of premature deaths among male juvenile offenders in this sample is 6.5%.

Lattimore, Linster, and MacDonald (1997) conducted a follow-up study of two cohorts of California state inmates released on parole. The first cohort of 1,997 inmates was released in 1981-1982 and followed until age 30. The proportion of offenders who died by age 30 was 5.5%. The second cohort of inmates (N = 1,998) was released in 1986-1987 and followed for 6 years. The proportion who died by age 26 was 3.5%. Adjusting for time at risk (11 and 6 years, respectively), the odds of premature death for both cohorts are similar.

The Gluecks (1968) examined mortality data for a 1930 cohort of persistent juvenile offenders committed to Massachusetts correctional schools and found that by age 32, delinquent individuals had a mortality rate twice that for the matched control group. Laub and Vaillant (2000) extended the follow-up until age 65 and report that 8% and 42% of offenders were dead by age 40 and 65 respectively, whereas this proportion was 4% and 27% for the matched control. Stattin and Romelsjo (1995) conducted a follow-up study of a random sample of 7,757 males born in 1951 in Stockholm. A detailed questionnaire was administered when these individuals were 18 years old. A follow-up of their life courses indicated that 1.8% of the cohort's subjects died between their 18th and 33rd birthdays (N = 138). Among those who had been convicted by criminal courts (N = 2,053 or 27% of the sample), the proportion of premature deaths was 3% (N = 61), whereas the mortality rate for the remaining young male adults, who had not been convicted by criminal courts (N = 5,524), was 1.3% (N = 74). The mortality rate was thus 2.3 times higher for offenders than for non-offenders.

Although limited, the available evidence therefore indicates that offenders are, indeed, likely to die prematurely. As cohort studies extend the scope of their follow-ups and systematically investigate adult life transitions (including desistance) of habitual offenders, patterns of excess mortality rates among offenders will presumably attain the status of well-established criminological fact and become a more elaborate focus of theoretical explanation. Gottfredson and Hirschi's general delinquency-hazard model does provide a coherent analytical framework for interpreting such a finding. Lattimore et al.'s (1997) results showed that, as expected, inmates are indeed more likely to die prematurely in accidents. But they also indicated that this is true whatever the particular cause of death: they are more likely to be killed by others (homicide), more likely to commit suicide, and more at risk of lethal overdose.

As predicted, the odds of premature death increase as a function of offenders' degree of involvement in crime. The more extensive their prior record, the higher their odds of dying prematurely (Stattin and Romelsjo 1995): the proportion of subjects who died by age 33 was 3% among all those who were convicted by criminal courts; this proportion became 4.7% for those who had been convicted twice or more and 7.2% for those convicted four times or more. Stattin and Romelsjo's findings also show that controlling for severe addiction, the more persistent offenders do not exhibit significantly higher mortality rates, suggesting that odds of premature death are not mainly determined by their crime activities (as...