Does Prison Work?

A view from criminology - Timothy Ore and Astrid Birgden
Department of Justice Office of the Correctional Services Commissioner
Click here for PDF version

Introduction

In 'Does Prison Work?' recently published in Policy, Saunders and Billante 1 argued that 'a dramatic fall in the American crime rates over the past decade' could be explained primarily by two policy changes: more offenders being sent to prison and 'Broken Windows' policing. Most of the report is devoted to the former, drawing on the work of Charles Murray 2. The authors concluded that Australia could reduce its escalating crime rates by making greater use of incarceration as in the US.

The purpose of this paper is to detail fundamental methodological issues associated with the measures used. In addition, the paper will briefly indicate that deterrence-based programs are ineffective in reducing crime and consider alternative methods to reduce re-offending. Reaction to the article is important for at least three reasons. First, crime and punishment are among 'the most topical, urgent and contentious social questions of our times 3'. Second, given the tendency for crime to become a political issue, frequently represented in emotive terms, an empirical, evidence-based approach to criminological analysis can inform the debate on penal policy 4. Third, contemporary discourse of crime varies considerably from perceptions of offenders as rational individuals, who choose crime on the basis of a cost-benefit calculus, to offenders as individuals with deficits, who require rehabilitation programs to develop pro-social skills 5. Without an evidence-based debate on crime, it is doubtful if real answers can be found to important questions such as which interventions are more successful in reducing crime.

Rather than using simple correlation analyses to consider complex problems, meta-analysis statistically detects and consolidates trends across the findings of large numbers of outcome studies. Although meta-analysis is potentially effected by the quality of outcome studies, publication biases, threats to external and internal validity, and generalisation from clinical trials to the 'real world' 6, it allows the integration of findings in the behavioural sciences and criminology. Meta-analytic findings will be considered as a method to compare the effects of deterrence and rehabilitation.

Ineffectiveness of Deterrence

Saunders and Billante 1 consider prison to be effective for incapacitation and deterrence. The effect of the former is obvious as an individual is removed from society. However, the impact of deterrence requires a more sophisticated analysis. The authors believe that offenders commit crime if they consider the benefits will outweigh the cost and so legal sanctions will have a deterrent effect on behaviour. However, this utilitarian or consequentialist assumption is not supported by any consistent findings from official criminal statistics or research studies 7. Indeed, offenders differ from the general population in that they have deficits in problem solving abilities and so rarely consider the consequences of their actions 6,8-9. Current scientific opinion on an international basis is that punishment through imprisonment does not reduce crime rates and, in some instances, even worsens crime rates. For example, in a recent review of 29 evaluation studies of boot camps, this approach was considered ineffective in reducing crime 10. Therefore, recent research has failed to establish a link between length of prison sentence and recidivism as predicted by deterrence theory.

Rehabilitation to Reduce Offending

As a product of numerous factors, crime requires varying interventions targeting problem-specific areas. Best practice rehabilitation programs are those that target factors empirically linked to the risk for re-offending. These include: poor parental supervision or low attachment to families; difficulty with education and employment; socialising with criminal networks; anti-social attitudes; distorted thinking patterns or information processing; limited coping, problem solving and social skills; and low self-control or high impulsivity. There are now almost 2,000 separate studies and over 30 meta-analytic reviews that have considered whether recidivism rates can be reduced through rehabilitation 7. Evidence shows that the risk for re-offending are modifiable when such programs are delivered. For example, recidivism rates in serious or persistent young offenders can be reduced by 40% in the community and 30% in institutions 6.

Methodological Issues

Specifically in terms of the methodological approach used by Saunders and Billante 1, it is the view that incarceration rates should have been computed as the ratio of persons admitted to prison for a particular offence in a given year to the number of persons arrested for that offence in the same year. In this way, the likelihood of the results accurately capturing cross-national differences in the willingness to incarcerate is enhanced. The complex and multi-dimensional association between levels of crime and incarceration propensity cannot be explored through simple correlation analyses 11-13. The connections are multiple and reciprocal, requiring advanced statistical modelling.

Were the right measures of imprisonment employed?

Two sets of data are required for calculating a crime rate or imprisonment rate: a numerator (the number above the line in a fraction) and a denominator (the part of a fraction appearing below the line). Table 1 summarises the measures employed by Saunders and Billante 1 in Figures 1-3 compared to those proposed in the present paper.

Table 1: Summary of Reported and Proposed Measures for Incarceration Propensity

By using stock data, number of prisoners in custody on a given day, the trends in Figures 1-3 have confounded sentence length with imprisonment rates. Stock data often over-represent more serious offenders with longer sentences, with the potential for over-estimation of the propensity to incarcerate in those countries with higher serious crime rates. Flow data, the number of admissions to prison (new court commitments, excluding persons re-admitted to prison for conditions such as parole violation) are not affected by the accumulation of more serious offenders, thereby allowing the separation of the propensity to incarcerate from the length of sentence served. Despite their usefulness, flow data have limitations, which include changes in the charged offence and delays in court processing, both of which can affect incarceration rates. For instance, countries that process cases more slowly will appear to have an artificially low incarceration rate compared to countries that process cases more efficiently 14.

Failure to distinguish between sentenced and unsentenced prisoners can affect comparisons of stock-based incarceration rates since not all those held in a prison have been convicted of an offence 15. A fifth of the total Australian prison population are unsentenced 16. There is no evidence that Saunders and Billante adjusted for variations in size of unsentenced prisoners.

Crimes reported to police are known to involve numerous errors 17, which are not constant cross-nationally 18. One of the reasons for establishing the International Crime Victims Survey (ICVS) 19 was to provide an alternative mechanism to inaccurate police records on crime. Using convicted persons instead, is also problematic for reasons including the absence of court data of consistently high quality across countries. To minimise bias in comparative studies, police arrests, rather than crimes reported to police, seem to be the most appropriate data to use in Figures 2 and 3. Data on convicted persons offer a useful measure of the relative punitiveness of sentencing practices across countries as they cover only those who have been found guilty of a specific crime. However, determining the point at which conviction actually occurs in different criminal justice systems is difficult 14. For several reasons, police arrest data may be the most adequate way for adjusting incarceration rates for crime variations. The use of arrested persons means that the numerator and the denominator are consistent, in contrast to situations where one person can be responsible for multiple crimes but likely to be sentenced only once. Police arrests can be expected to vary with criminal activity rates and have similar meaning in many countries 14. Arrest data must be adjusted to match the age profile of the numerator by dividing the rate by the proportion of arrests involving adults for each offence category.

In summary, cross national comparisons of the ratio of prisoners (flow or stock) to the total population or adult population standardise for population size variations but not the relative propensity of the population to engage in crime. By not using offenders, who are the true population at risk for imprisonment, as the denominator, Figure 1 confounds imprisonment rates with the level of criminality. For reasons already stated, the second denominator in Figures 2 and 3, should be actual police arrests rather than reported crimes to police.

Were adjustments made to differences in crime seriousness?

Figures 1-3 have not controlled for differences in the seriousness of crime across the countries compared. Countries could have similar crime rates but the nature of the crimes committed could vary. The ICVS 19 (the fourth round, 2000) reported that 'there was a higher than average use of weapons in the USA, Spain, Scotland and Portugal'. Guns were used more often in the US and Spain. Without standardising for such variations, it is incorrect to attribute differences in incarceration rates to punitiveness. Therefore, offence-specific analyses provide a better approach.

Since countries seem to sentence offenders differently for different offence types, cross-national comparisons of sentencing practices ought to be crime-specific. Therefore, the robbery incarceration rate for the US, for instance, is likely to be lower than other countries for comparable crime seriousness, if robberies involving the use of firearms are held constant. What of cross-national differences in offenders' criminal history? In Australia, at least 59% of male prisoners and 51% of female prisoners have prior imprisonment. Approximately 80% of Indigenous prisoners have prior imprisonment 16. Spier 20 found that 37% of prisoners (n=22,430) released from New Zealand prisons during 1995-98 were re-imprisoned within two years and 51% within five years. Of 272,111 prisoners released in the US in 1994, 67.5% were re-arrested within three years post-release 21. Holding constant variations in recidivism may also change differences in the propensity to incarcerate.

More reliable results of the relative frequency with which the countries compared in Figures 1-3 by Saunders and Billante 1 use incarceration as a sentencing option can be obtained by adjusting incarceration rates for the incidence of crime, limiting comparisons to rates for similar offences, and by employing flow data. For instance, in a comparison of the use of incarceration in US, Canada, Germany and England, Lynch 14 found that, in terms of either population-based stock rates or population-based flow rates, the US was several times more likely than any of the countries to incarcerate for homicide, robbery, burglary, and larceny. For homicide, the US was incarcerating 7.5 times and 5.3 times more frequently than England and Germany, respectively. By contrast, flow rates based on police arrests revealed a different pattern, showing a broad similarity in the probability of incarceration for the offences. Similarly, Blumstein 22 has argued that it is more useful to measure incarceration rate per crime rather than per head of population. Correlating the level of prison-eligible crimes such as homicide with the level of incarceration for those crimes, he showed that the US appears less punitive than reported in international comparisons of imprisonment rates.

Do crime rates necessarily predict imprisonment rates?

Several studies have shown the influence of crime rates on imprisonment rates to be limited 11,23-25. Numerous and complex factors need to be explored. For example, differences in the organisation of the criminal justice system, including sentencing 11,14, socio-cultural values 26,27 inequality and reward structure 28,29, political regime 30, statistical and administrative processes and changes 11,31, and the level of media and political attention 32.

As noted by Wiles 4, many of the risk factors associated with offending behaviour are outside the control of the criminal justice agency, indicating that factors other than penology determine crime levels. Following a close examination of arrest rates for violent crime and other crimes in the US in 1971-96, Tonry 12 found that socio-economic conditions were more influential on crime rates than the action of the criminal justice system. In spite of having the second lowest imprisonment rate in Australia, Victoria's rates of assault and robbery are approximately half the Australian average and its sexual assault rate is 20% below the national average 11. In Canada, where the criminal law is the same across the country but administered provincially, Sprott and Doob 13 found that crime rates did not predict incarceration counts, although some measures of provincial charge rate did correlate with provincial prison counts. Thus, more detailed analyses are required to substantiate the claim by Saunders and Billante 1 that 'the rate of crime and incidence of punishment are closely associated'.

Conclusion

The observed differences reported by Saunders and Billante in the propensity to incarcerate cross-nationally have been made in terms that are too general to serve as a useful and valid basis for policy guide. Stringent requirements, focusing on more sensitive numerators, denominators, and specific crime classes, are critical. Analyses of comparable crimes minimise the effects of variations in crime seriousness cross- nationally, thereby yielding more credible results. Well-designed studies show that deterrence-based programs are ineffective in reducing crime and the focus should be on developing rehabilitation programs that do reduce the likelihood of recidivism. The case for Australia adopting the US approach to crime reduction through the use of imprisonment has not been established.

References

1Saunders P and Billante N, 'Does Prison Work?', Policy (Summer 2002-03), 18 (4), 3-8.

2 Murray C, Does Prison Work?, London: Institute of Economic Affairs, 1997.

3 Garland D and Sparks R, 'Criminology, social theory and the challenge of our times', British Journal of Criminology 40, 2000, 189-204.

4 Wiles P, 'Criminology in the 21st century: Public good or private interest? - The Sir John Barry Memorial Lecture', The Australian and New Zealand Journal of Criminology 35, 2002, 238-252.

5 Rose N, 'Government and control', British Journal of Criminology 40, 2000, 321-339.

6 McGuire, J. 'Integrating findings from research reviews', in McGuire, J (ed.), Offender rehabilitation and treatment: Effective programs and policies to reduce re-offending, Chichester: John Wiley & Sons, 2002, 3-38

7 McGuire, J. 'Maintaining change: Converging legal and psychological initiatives in a therapeutic jurisprudence framework'. The Western Criminology Review. Published by the Department of Criminal Justice, California State University, San Bernardino, 2003.

8 Gendreau, P and Ross RR. 'Effective correctional treatment: Bibliography for cynics'. Crime and Delinquency, 1979, 24 (4), 463-489.

9 Robinson, D and Fabiano EJ. 'Programming in cognitive skills: Reasoning and rehabilitation program'. In Hollin C.R. (ed.) Handbook of offender assessment and treatment. NY: John Wiley & Sons, 2000, 179-193.

10 Mackenzie DL, Wilson DB, and Kider SB, 'Effects of correctional boot camps on offending'. Annals of the American Academy of Political and Social Science, 2001, 578, 126-43.

11 Freiberg A and Ross S, Sentencing reform and penal change: The Victorian experience. Melbourne: The Federation Press, 1999.

12 Tonry M, 'Crime and punishment in America', Overcrowded Times 1 9 (2).

13 Sprott JB and Doob AN, 'Understanding provincial variation in incarceration rates'. Canadian Journal of Criminology (July), 1998, 305-322

14 Lynch JP, 'A comparison of prison use in England, Canada, West Germany and the United States: A limited test of the punitive hypothesis', Journal of Criminal Law and Criminology 79, 1988, 180-217.

15 Brown D, 'Battles around truth: A commentary on sentencing Act 1989', Current Issues in Criminal Justice, 1992, 3, 329.

16. Australian Bureau of Statistics, Prisoners in Australia, Cat. No. 4517.0, 30-31.

17 Langan PA and Farrington D, Crime and justice in the United States and England and Wales, 1981-96. US Department of Justice, Washington, D.C., 1998, 6.

18 Kesteren JN, Mayhew P and Nieuwbeerta P, Criminal victimisation in seventeen industrialised countries: Key findings from the 2000 International Crime Victim Survey, The Hague, Ministry of Justice, WODC, 2000.

19 International Crime Victims Survey, The Hague, Ministry of Justice, WODC, 2000.

20 Spier P, Reconviction and reimprisonment rates for released prisoners. New Zealand Ministry of Justice, Research Findings No. 1, May, 2002.

21 Langan PA and Levin DJ, Recidivism of prisoners released in 1994. US Bureau of Justice. Washington, D.C., 2002.

22 Blumstein A, 'Prisons populations: A system out of control?', in Tonry M. (ed.) Crime and Justice: A Review of Research, Chicago: University of Chicago Press, 1988, 10.

23 Michalowski R and Pearson MA, 'Punishment and social structure at the State level: A cross-sectional comparison of 1970 and 1980', Journal of Research in Crime and Delinquency, 1990, 27, 25.

24 Young W and Brown M, 'Cross-national comparisons of imprisonment' in Tonry M. (ed.) Crime and Justice: A Review of Research, Chicago: University of Chicago Press, 1993, 17

25 Quimet M and Tremblay P, 'A normative theory of the relationship between crime rates and imprisonment rates: An analysis of the penal behaviour of US States from 1972 to 1992', Journal of Research in Crime and Delinquency, 1996, 33, 109.

26 Zimring F and Hawkins G, The scale of imprisonment, Chicago: The University of Chicago Press, 1991.

27 Tyler TR and Boeckmann RJ, 'Three strikes and you are out, but why? The psychology of public support for punishing rule breakers', Law and Society Review, 1997, 31, 237.

28 Wilkins LT and Pease K, 'Public demand for punishment', International Journal of Sociology and Social Policy, 1987, 7, 16.

29 Hay D, 'Time, inequality, and law's violence', in Sarat A and Kearns TK (eds.) Law's Violence, Ann Arbor: University of Michigan Press, 1992.

30 Walker J, Wilson PR, Chappell D and Weatherburn D, A comparison of crime in Australia and other countries. Australian Institute of Criminology, Trends and Issues in Crime and Criminal Justice, 1990, No. 23, June.

31 McConville C, 'From 'criminal class' to 'underworld'', in Davidson G, Dunstan D, and McConville C (eds.) The outcasts of Melbourne: Essays in social history, Sydney: Allen and Unwin, 1985.

32 Beckett K, 'Political preoccupation with crime leads, not follows, public opinion, Overcrowded Times 1, 1997, 8 (5).

A view from sociology - Peter Saunders and Nicole Billante
The Centre for Independent Studies

A view from sociology Peter Saunders and Nicole Billante Timothy Ore and Astrid Birgden have written a very interesting and informative response, but it does not address the core concerns in our original article.

Following Charles MurrayŐs lead, we charted crime rates against the rate of imprisonment per recorded crime. We did this over time, and we did it across several different countries, and the results were striking. As the chances of criminals getting locked up fall, so the number of serious offences rises; when more criminals get locked up, the rate of serious crime drops.

Few people outside of academic social science would be surprised by these correlations, for they confirm what common sense tells us. If you ease up on the punishment, you get more of the crime.

But read their response carefully. Ore and Birgden never actually deny our core premise, that imprisonment rate has an effect on the crime rate. Not once in their critique do they directly address this question, still less try to refute it.

Rather, they question the reasons why the connection exists. In particular, they claim right from the start that prison is an ineffective deterrent, and that preventative measures are better.

They may be right. We noted in our original article that the probability of getting caught tends to be more of a deterrent to many people than the severity of the punishment. This suggests that money spent on policing may generate a better return than money spent on prisons, which is why we went on in subsequent work to look at the relation between police numbers and crime rates (sure enough, as with prison, so with policing, we found prima facie evidence of a strong association between them).

But this does not mean that prison is ineffective. Be in no doubtŃif we stopped locking away serious offenders, crime would go through the roof, no matter how many of them got caught. The reason is that prison does not work wholly, or even mainly, by its deterrence effect. It works most crucially by physically removing the worst offenders from society so they cannot go on committing crimes (that is, by incapacitation).

This is precisely what the Americans realised from the 1980s onwards, and the result has been bulging jails and a plummeting crime rate. Most Americans think thatŐs not a bad trade-off. In Australia, by contrast, we made it less and less likely as time went on that robbers, thieves, rapists and thugs would end up in prison, even if they were caught. Not surprisingly, our crime rates just kept on climbing.

What of the Ômethodological issuesŐ that Ore and Birgden raise? They make some good points.

It would, for example, make a lot of sense to follow their advice and break the statistics down into the different categories of serious crime to see how well the overall association between crime rates and probability of imprisonment stands up for, say, robberies as compared with rapes. But the way AustraliaŐs statistics were compiled prior to the 1990s makes it very difficult to do this over the sort of time period we are interested in analysing.

Ore and Birgden also argue that we should measure offenders rather than offences and convicted prisoners rather than all prisoners. Both probably very sensible, assuming it were possible to do so over a nearly 40 year period. But even if we recomputed our data in this way, it would make little difference to our findings because we are interested in overall trends, not precise figures.

Ore and Birgden are concentrating on the trees and failing to see the wood. What our graphs show is a clear trend over time. Even if our annual measures of crimes and prisoners are a bit crude and imprecise, it doesnŐt matter very much for the trends we have plotted because the same indicators were measured in the same way each year. Any error will therefore occur in the same way throughout the graph and will effectively cancel itself out. Maybe we could get more precise in the way each yearŐs figures are measured, but this will have little impact on the trend-lines we have plotted.

Some of Ore and BirgdenŐs suggestions are not very helpful. For example, they say we should use police arrests, rather than offences reported to the police, as our indicator of crime trends, but this makes little sense. Police arrest rates will depend as much on the effectiveness of policing as on the number of crimes being committed (if the police become less effective, then the number of criminals processed will fall, even if the actual number of crimes is still rising). They are not therefore a good indicator of changes in crime.

Similarly, they say we should have measured the flow of prisoners rather than the stock, but this argument only holds if (like them) you want to measure only the deterrent effect of penal policy. If you also want to measure the incapacitation effect, however, then it is perfectly reasonable to focus on the stock of prisoners as your key indicator.

Have Ore and Birgden refuted the association that we found between crime rates and the probability of imprisonment? No they havenŐt. They say that other factors, including Ôsocio-economic conditionsŐ, are important in influencing the rate of crime, and they may be right. But this does not mean that penal policy has no effect. Criminal behaviour, like almost every other aspect of social life, is influenced by the interaction of many different factors. One of them is the probability of imprisonment.

Have Ore and Birgden shown that the association between crime rates and imprisonment is not a causal relationship? No they havenŐt. We showed that penal policy and crime are statistically associated over time, and across countries. Our hypothesis is that this is a causal connection. If Ore and Birgden think we are wrong, it is incumbent on them to come up with a better explanation for this pattern (for as the philosopher of science, Imre Lakatos argues, no hypothesis should be discarded until a better one is found to put in its place). They have not done so. Until they do, the case for prison stands.

Policy is the quarterly review of The Centre for Independent Studies. For more information on subscribing to Policy, click HERE

If you are interested in the Centre's activities and publications, why not subscribe to e-PreCIS, our regular email update on the latest news and events.

(e-PreCIS requires html capable email facilities, such as Microsoft Outlook Express or Netscape Messenger)