banner supportlinkscontactaboutarticlesmain

Prison Works – A Review Of The Evidence

I’ve been looking at the evidence for or against the effectiveness of prison in reducing crime. There is no shortage of statistics published by governments and various criminal justice bodies, but when one looks closely at the data it is a minefield of inconsistency, moving goal posts, and definitions which differ between countries and change with time making like with like comparisons extremely difficult.

As an example consider murder. A layman would probably have little difficulty in defining murder, but once handed over to the legal systems of different countries one finds that murder has numerous different definitions. Furthermore, in the case of the UK, and I suspect many other countries, it is clear that more and more crimes that might once have been prosecuted under a charge of murder are now prosecuted under lesser charges such as manslaughter. This should come as no surprise. Whether this is done in the interest of expediency (plea bargaining and lesser charges increase conviction rates), in the interest of presenting the impression that crime is falling (fewer murder trials suggests there are fewer murders), or in the interest relieving pressures on the prison system doesn’t matter. The distortions are still there making comparisons over time or between different legal systems open to different interpretations.

For the above reasons, and because prison is only one of many variables which influence the incidence of crime, analysis of criminal data is going to be problematic at best, and any correlations found are unlikely to be strong. However, having made this caveat we can only work with the data that is available. So what does it tell us?

One comprehensive reference produced by the WODC (the Dutch abbreviation for Wetenschappelijk Onderzoek- en Documentatiecentrum, in English: Research and Documentation Centre), an international criminal justice knowledge centre, is the European Sourcebook of Crime and Criminal Justice Statistics which provides a comprehensive European data base up to the year 2000 in the 2003 edition, and 2000 – 2003 data in the 2006 edition (

In principle, it should be relatively easy to compare crime rates in different countries with their incarceration rates to see if more use of prison leads to less crime. For example, a first question might be, “Do those countries that send a higher proportion of convicted criminals to prison enjoy lower rates of crime than those countries that rely more on community punishments?”

The graph below is a plot of the percentage of convicted criminals who were sent to prison versus crime rate (number of crimes per 100,000 of population) for 28 European countries. The data, taken from the 2003 edition of The Sourcebook contains the latest published data from the year 2000.


A negative correlation is revealed, i.e. Those countries which employ the sanction of prison the most tend to experience lower crime rates than countries which rely more heavily on community sentencing. Running this data through a statistical analysis package revealed a correlation coefficient of -0.55. A correlation coefficient measures the degree that one variable (use of prison) is dependant on another variable (crime rate). A correlation coefficient of zero shows no relationship between the variables, whereas a correlation coefficient of 1 or -1 shows that change in one variable fully accounts for the change in the other variable. Our result (-0.55) is evidence of a moderate correlation, but in view of the caveats raised in the introduction, this correlation is perhaps as high as one might reasonably expect from such data.

The correlation coefficient squared times 100 tells us the percentage of the variation in crime rate that can be explained by the use of prison as a deterrent. In this case, 0.55 x 0.55 x 100 = 30%, i.e. nearly a third of the variation in European crime rates can be explained by the effects of prison. In the imprecise field of social science this degree of correlation is pretty good, but we must ask, is it statistically significant, or could this finding just as easily have resulted from chance. The statistical analysis answers this question. A “P” value, the probability of seeing such a correlation in a collection of random data in which prison rates had no effect was 0.003, i.e. there is only a 0.3% chance that this result was a random effect, or alternatively we can say that there is a 99.7% probability that the use of prison has some effect in reducing crime rates.

Attempts to confirm the results for later years by carrying out the same analysis on statistics presented in the 2006 edition of this reference source were thwarted because a change in data collection methodology meant that the percentage of convictions which resulted in prison was no longer reported. One cannot help wondering why the collection of official data is invariably tinkered with over time when such tinkering makes the study of trends so much more difficult. We all know that it happens with economics data (e.g. hedonic manipulation to suppress inflation data), and education data (e.g. the maintenance of standards debate, “O” levels versus GCSE’s). Well it certainly happens with crime statistics too, and I will leave you to decide whether the tinkering results from the imposition of change for the sake of change, or is the result of something more sinister, change for the sake of obscuration.

One final point that must be made is that the demonstration of a correlation, such as the fact that more prison is correlated with lower crime rates that we have demonstrated here, does not demonstrate cause and effect. Statistical analysis cannot say that more prison causes less crime, because there could be other variables not considered in this analysis that are the true causative factors. However, the onus should now be on the prison reformers to identify such unknown factors if they wish to continue claiming that prison either has no effect or even that prison causes more crime.

The reluctance of prison reformers to present analyses of data, either for or against prison is telling. It either demonstrates an inability to address this debate logically as a dispassionate search for the truth, or a fear of what such an impartial evaluation might mean to their beliefs. Either way, it reflects badly on the way they argue their case.

I will now look for evidence for and against the success of community punishments in reducing crime. I have to admit that my searches so far have revealed a total and complete absence of any published data that could be analysed scientifically, but I will continue looking, and if anyone reading this knows of any community punishment data, please let me know.

Post a comment