Introduction

Many Western countries, most notably the USA, have experienced a boom in prisoners rates, characterized as “carceral hyperinflation” or the “new punitiveness” (Pratt et al. 2005). To address a variety of issues related to increasing incarceration rates and prison overcrowding, many countries have successfully embarked on initiatives to create alternatives to imprisonment. Some of these initiatives are geared to providing judges with community-based options that will serve as appropriate sanctions for the crime. Some operate as diversions from the formal criminal justice system (e.g. restorative justice programmes for juveniles). Others, such as halfway houses, temporary release programmes, and furloughs, operate at the back end of the prison system and provide non-incarcerative options for use by correctional administrators and parole boards. Some countries, e.g. Finland, Canada and Poland, have in recent years successfully embarked on sustained national policies of stabilising or even decreasing the use of incarceration as a response to crime (Doob and Webster 2006). In the Netherlands the national prisoners’ rate has decreased by 20 percent since 2005 (CBS 2008), presumably as a result of the government’s standing policy to promote community service orders.

It is often assumed by politicians and judges that public opinion in their country will not accept less severe sentencing. This assumption results in hesitancy to further promote alternatives to imprisonment. For this reason, research into public attitudes towards sentencing, including towards non-custodial sentencing options has important implications for policymaking (Roberts and Hough 2002). Comparative international research into this topic has the advantage of allowing policymakers to understand both public attitudes towards sentencing and actual sentencing policies in an international perspective. Such comparison can alert readers to elements in the national situation that are seemingly normal but deviate from trends elsewhere.

In this article, I will analyse data on public attitudes towards sentencing across 50 or more different countries, derived from the International Crime Victims Surveys (ICVS) and the European Survey of Crime and Safety (EU ICS), a collaborative research effort which was launched in 1989 by a consortium of European criminologists on the initiative of Jan Van Dijk. Results of similar analyses of older ICVS data on public attitudes towards sentencing and its social correlates have been published before (Besserer 2001; Kury et al. 2002; Mayhew and Van Kesteren 2002; Van Dijk 2008). I will replicate this work, using the results of the fifth sweep of the survey conducted in 2004/2005 covering 32 different countries. I will also try to extend and deepen the analysis in several respects. Since for many countries data are now available from subsequent surveys conducted since 1987, I will discuss trends in public attitudes towards sentencing over time. Instead of relying on data on a preference for imprisonment as an indicator of punitivity, I have constructed a scale expressing a preference for several non-custodial sentences, a moderate prison sentence, a medium long prison sentence and a very long one. The scale can be interpreted as one on interval level. I have subsequently re-analysed the social correlates of punitiveness using this newly constructed scale. Public attitudes towards sentencing can be influenced by social characteristics of individuals as well as by characteristics of countries. In previous analyses these relationships at the micro and macro level have been analysed independently of each other (Mayhew and Van Kesteren 2002). To obtain more comprehensive results, I have carried out a multi-level analysis of the social correlates of punitivity at both levels. This analysis shows which social factors are associated with high punitivity at individual and country level and whether inter-country differences persist after the impact of both these factors has been accounted for.

In the past, ICVS-based data on attitudes towards sentencing have been compared with actual rates of prisoners per 100,000 (Van Dijk and Mayhew 1992; Van Dijk et al. 1990). Prisoners rates cannot be regarded as a reliable measure of the severity of sentencing tariffs because they are co-influenced by the extent of serious crime. In my analysis, I have used a scale expressing the prisoners rates controlling for the homicide rates of countries, constructed by Van Dijk (2008). To explore the extent of congruity between public attitudes towards sentencing and formal sentencing I will look at the association between public attitudes towards imprisonment and this new measure of sentencing severity.

Proportions Preferring Imprisonment or Community Service

Victimization surveys have primarily been designed as a source of statistical information on the volume and trends of crime collected independently from police records. From this perspective, prevalence and incidence rates of victimization are the key findings. The surveys typically also yield information on reporting to the police, victim satisfaction with the police, fear of crime and attitudes towards sentencing. In the case of the International Crime Victim Surveys, data about crime and about public attitudes is collected in a comparative, international perspective. The first round of ICVS surveys was in 1989, followed by surveys in 1992, 1996 and 2000. The last round of surveys was done in 2004 and 2005. For more information on the methodology of the ICVS and EU ICS, I refer to the latest reports with key findings (Van Dijk et al. 2007, 2008) and to the web-sites Footnote 1 of both projects. The results presented here are extracted from the integrated database of the ICVS (Van Kesteren 2007). We will firstly present descriptive cross-national data on the answers to the question what sentence respondents considered most appropriate for a recidivist burglar - a man aged 21 who is found guilty of burglary for the second time, this time having stolen a colour television. Respondents were asked to choose between fine, prison, community service, suspended sentence or another sentence. Those opting for imprisonment were asked what term of imprisonment seemed most appropriate. With these two items, the complex phenomenon of public attitudes to sentencing can only be assessed roughly. In local surveys, respondents are often asked what would be the most appropriate punishment, and for what reasons, in very specific cases (Boers and Sessar 1988; Roberts and Hough 2002). In large scale international surveys, such as the ICVS, the use of single item information is often unavoidable for costs reasons. It is also possible that more detailed questions or response categories, for example about finer sentencing options, would have reduced rather than enhanced comparability across countries.

Figure 1 shows percentages respondent per country opting for imprisonment and community service orders respectively in the ICVS survey of 2004/2005, including the EU ICS. It merits observing, first and foremost, that community service order was the preferred sentence for 39% of respondents in the national samples in 2004/2005. Imprisonment was recommended by 37% of respondents overall. Community service was the preferred sentence in all continental European countries except Bulgaria.

Fig. 1
figure 1

Percentage of the public opting for community service order and imprisonment as punishment for a recidivist burglar in 2004/05 in countries and main cities. 2004–2005 ICVS and 2005 EU ICS

Full size image

Imprisonment was the first choice in most developing countries, represented in the latest round of the ICVS mainly by samples from major cities. Fifty eight (58) percent of these respondents were in favour of imprisonment. From a global perspective, support for prison sentences was highest in South Africa, Phnom Penh, the Far East (Japan and Hong Kong), Latin America, English speaking/common law countries -with the exception of Ireland- and Bulgaria and Turkey.

Globally, there was a remarkably wide-ranging divergence across countries. Over 50% favoured imprisonment in Johannesburg (76%), Mexico, Hong Kong, Lima, Japan, Northern Ireland and Istanbul. In several European countries, less than 30% opted for imprisonment. The population of France (13%) and Austria (13%) are least in favour of this option. A community service order was seen as the most appropriate sentence overall in the 16 EU member states providing results in the 2004/05 ICVS - EU ICS: 49% of respondents recommended it. It was the first choice of sentence in half of the countries, with particularly strong support in Luxembourg, France and Portugal (69% opting for it) and Belgium (67%). There was, however, a fairly wide divergence of opinion within the European Union as well: a community sentence was seen as most appropriate by less than 30% in the UK.

Trends in Punitiveness

The ICVS has been executed more than once in many countries over the past two decades. Table 1 shows results from countries where the survey has been carried out more than once since 1989.

Table 1 Percentage of the public opting for imprisonment as punishment for a recidivist burglar in 2004/05 plus results from earlier surveys in countries and main cities. Countries that participated more than once in the 1989 – 2005 ICVS and 2005 EU ICS
Full size table

Leaving aside changes in relative levels of support for different individual sentencing options, the first four sweeps of the ICVS sweep showed a general hardening of attitudes towards punishment (Van Kesteren et al. 2000). This trend of increased support for imprisonment has not continued. In many countries, support for imprisonment has since 2000 remained stable. Lower percentages of the public are favouring imprisonment in 2004/05 than in 2000/1996 in the USA, Australia, the Netherlands, Belgium, Finland and Estonia. Poland displays a significant drop during the 1990ties but an increase thereafter. Between 2000 and 2004/05 the upward trend in support for imprisonment seems to have reached a plateau in most countries.

It could be argued that over the years the value of a colour TV has diminished and that this might have influenced the support for imprisonment. If this would indeed have been the case, the slackening public support for imprisonment would be an artefact. In my view respondents of the survey probably respond to the broader concept of a recidivist burglar rather than to the information on the item that was stolen last time. The interpretation that opinions have genuinely become less punitive is supported by the finding that since 2000, repondents in most developed countries have been less often victimised by crime, are less fearful of crime and more satisfied with their local police (Van Dijk et al. 2008). Reduced support for imprisonment can be interpreted as part of a broader new trend towards reduced public concern about crime.

In some countries, support for community service shows significant shifts over time. The percentage opting for a community service order in Finland increased markedly after 1989, when this sentencing option was formally introduced in Finland. This shift in public opinion suggests that changes in legislation and actual sentencing can increase support for alternatives to imprisonment. Support has fallen back somewhat in Finland since 1992, although it is still higher than in 1989. Between the 1996, 2000 and 2005 ICVS sweeps, though, there has been little further change.

Duration of Imprisonment

Respondents who opted for a prison sentence were asked “For how long do you think he should go to prison?”. The length of sentence recommended is correlated to the rate of respondents per country who opt for a prison sentence: at the country level a more punitive attitude also translates into longer sentences. Based on the 29 national surveys in the 2004/05 ICVS and 2005 EU ICS, the correlation is 0.51 and statistically significant. Figure 2 shows this relation at the country level.

Fig. 2
figure 2

Plot of the percentage of the population opting for a prison sentence against the duration of that prison sentence by country. 2004 – 2005 ICVS and 2005 EU ICS

Full size image

Inhabitants of Mexico are by far the most punitive on both dimensions. Other countries with high scores on both imprisonment and length of prison are Bulgaria, the USA, England & Wales, Northern Ireland, Scotland and Japan. Among the outliers in the scatter plot are Poland and Estonia. Relatively few respondents in these countries favour imprisonment but those that do prefer imprisonment opt for relatively long prison sentences. This result must probably be interpreted against the background of declining numbers of people in favour of imprisonment in these countries. If proportions of citizens favouring imprisonment go down, those that have changed their opinion against imprisonment are likely to have previously been in favour of short rather than long prison sentences. This factor may lead to an increase of the mean of recommended prison length. Such increases in the average length of recommended prison sentence have been observed in Poland, Estonia, Finland and Belgium, all countries where support for imprisonment has gone down significantly. In other countries, however, both the proportion of supporters of imprisonment and the recommended length of prison sentences has gone down. This is notably the case in the USA where support for imprisonment went down from 53% in 1989 to 47% in 2004 and the recommended length of prison sentences from 37 months in 1989 to 28 months in 2004.

Scaling Punitiveness

To facilitate more refined analyses of public attitudes towards sentencing, I have constructed an index that subdivides sentencing preferences in seven categories of increasing severity through an optimal scaling procedure. For this purpose, I applied a multivariate regression analysis with a categorical variable as dependent, that consisted of four types of non-custodial sentences (suspended sentence, fine, community sentence, other sentence) and three modalities of a prison sentence (short, medium and long termFootnote 2). As independent variables in the equation, I selected variables known to be correlates of punitiveness according to the literature. I added the assessment by victims of the seriousness of the offence committed against them as an additional variable, under the assumption that such assessments are associated with the severity of the punishments deemed appropriateFootnote 3. Data on serious rating are only available for respondents who have been victimised by any crime in the course of the past five years. This part of the analysis is restricted to victims in a period of 5 years, in most countries around 50% of the sample.

In the regression analysis an optimal scaling procedure is applied that assigns numerical values to categorical data so as to optimize the linear regression equation for the transformed variablesFootnote 4. For the construction of the index we used first the full, global dataset of the ICVS 1989–2004/2005, covering urban areas in 57 countries. The analysis was conducted with three different sets of independent variables (known correlates of punitiveness only, correlates including the serious ratings by victims and serious ratings by victims only). The severities of the sentencing options are expressed by a scale with an average of 0 and a standard deviation of 1. Figure 3 shows results from the three different analyses.

Fig. 3
figure 3

Result of the quantification of the type of sentence using a regression analysis with optimal scaling (CATREG)

Full size image

The results of the scaling of the criterion variable were consistent for analyses with or without the serious rating. As expected, medium long and long prison sentences were rated as the most severe options and community service orders as the least severe. Contrary to expectation, fines were rated as a slightly more severe option than a short prison sentence. Our interpretation of this result is that respondents in developing countries are inclined to consider a heavy fine as more severe than a short prison sentence. Information on what has been understood under the ‘other sentence’ category is not available. Considering their rating as more severe than a short prison sentence, respondents have probably thought of options such as labour camps or corporal punishment. However, some may also have had restitution of damages/ compensation in mind.

The rating of fines as more severe than a short prison sentence seems counter-intuitive in the context of developed nations. The analysis has therefore been repeated using the ICVS dataset from developed countries only from the period 1989 through 2000. This dataset includes data on 30 developed countries, some of them participating more than once. The use of these historical datasets for the scaling procedure has the advantage that results of the analysis of social correlates, to be discussed later in this article, are not “contaminated” by the optimal scaling procedure. The analysis of correlates was carried out on the dataset of the ICVS 2004/2005.

The results of the scaling of sentencing options for developed countries are given in Fig. 4. Once again the three analyses produced roughly similar results. The only result that is somewhat out of line is the very low severity of a fine in a scaling procedure based on the seriousness rating of incidents by victims only. The rating of the seven sentencing options conforms to common sense expectations in a Western context, with prison sentences rated as most and suspended sentences and community service orders as least severe. Figure 4 depicts the resulting scale and Table 5 presents the percentages of the respondents that opted for the various options and the value of each category according to the scaling procedure.

Fig. 4
figure 4

Result of the quantification of the type of sentence using a regression analysis with optimal scaling (CATREG). Data from the 1989–2000 ICVS in industrialised countries

Full size image

A similar CATREG scaling procedure has been conducted on three national subsets of the EU ICS database by Kühnrich and Kania (2005). Their analyses of the German dataset produced a scale starting with community service as the least severe of the available options, a fine and suspended sentence as medium severe options and a prison sentence as most severe. Replication of the procedure on French and British datasets resulted in a rating of suspended sentences as less severe than fines. Our results based on 30 developed countries are very similar as those found for the UK and France by Kühnrich and Kania (2005). Table 2 shows the result.

Table 2 Result of the quantification of the type of sentence using a regression analysis with optimal scaling (CATREG)
Full size table

Punitiveness Scores per Country

For each participating country scores on the punitiveness index can be calculated. Figure 5 shows the scores on the newly created index of punitiveness of the countries participating in the ICVS 2004/05. Lower scores on the scale indicate a preference for less severe sentencing options.

Fig. 5
figure 5

Country scores on index of punitiveness, data from the 2004/05 ICVS and 2005 EU ICS

Full size image

The distribution of the punitiveness scores is heavily skewed towards South Africa, Peru and Buenos Aires. Among the developed countries, the USA and the United Kingdom stand out with comparatively high scores for punitiveness.

The punitiveness scale can also be used to identify changes in the severity of public attitudes towards sentencing in a more comprehensive way than by looking at support for prison sentences alone. Table 3 shows trends in punitiveness scores since 1989 for countries that participated more than once in the ICVS.

Table 3 Trends in scores on the punitiveness scale for countries that participated more than once since 1989 in the ICVS
Full size table

Correlates of Punitivity

I will now turn to the question which types of people prefer more severe sentencing options and whether these correlates are invariant across countries (Walker and Hough 1988). The ICVS questionnaire has not been designed with a view of testing theory- based hypotheses about public attitudes toward punishment. Previous analyses have looked at links of punitiveness with available socio-demographics, such as age, gender, income and level of education. Several authors have also looked at a variety of other, possible explanatory factors, such as past victimization by burglary or other types of crime, fear of crime and satisfaction with the police. In past analyses, most attention has been devoted to correlates of punitiveness at the level of individuals. We will briefly sum up the main findings.

Education

Results of the first ICVS showed a strong relationship between level of education and support for community service. Subsequent analyses of ICVS data have confirmed the relationship of less punitive sentencing preferences and level of education (Mayhew and Van Dijk 1997). This result is in line with other studies showing that support for non-custodial sentences is stronger if people are better educated and/or have received more information (Boers and Sessar 1988; Killias 1989). However, Besserer found no such linkage in a multivariate analysis of ICVS data from nine developed countries (Besserer 2001).

Age and Gender

Analyses of relationships between age and punitiveness have not produced consistent results across countries. According to Kuhn (1993), younger people tend to be more punitive in most Western countries but not in the UK, Germany, Finland and Norway. Kühnrich and Kania (2005) confirm that in the ICVS 2004/2005 German youth are more punitive than their parents. They interpret this finding as a cohort effect. However, Mayhew and Van Dijk (1997) report that young men are generally somewhat less punitive than older men in the second sweep of the ICVS. They also report that young women, for their part, are more punitive than older women. The punitiveness of younger women has been linked to heightened levels of fear of crime among this group. Concerning gender, in several analyses, men were found to be somewhat more punitive (Besserer 2001; Kühnrich and Kania 2005).

Victimization and Fear of Crime

According to stereotypical thinking, victims of crime are supposed to opt for punitive sanctions as an expression of their anger and desire for revenge. Van Dijk et al. (1990) reported that in the ICVS dataset those who have recently been victimised by burglary victims are more likely to opt for community service than non-victims. Several other analysts have also found no, modest or even inversed relationships between previous victimizations and punitiveness (Kury et al. 1992; Sessar 1992). According to Sessar, victims of relatively minor crimes might have moderate views on the most appropriate sentence because they possess more realistic information and are less influenced by media based ideas. Similar views were brought forward by Walker and Hough (1988). In their analysis of ICVS 2000 data, Van Kesteren et al. (2000) concluded a modest increase in support for imprisonment among those who had been a victim of either a burglary over the past five years or a contact crime. A preliminary, multi-variate analysis of the ICVS 2004 data showed no relationship between previous victimization controlling for external factors (Van Kesteren and Van Dijk forthcoming, 2009). Some authors, finally, have found significant relationships between variables measuring fear of crime and lack of confidence in the police and punitiveness (Costelloe et al. 2002; Kühnrich and Kania 2005).

In the most comprehensive analysis of older ICVS data on the social correlates of punitiveness to date, Mayhew & Van Kesteren have replicated previous work (Mayhew and Van Kesteren 2002). They found mixed results across countries and world regions. They looked first at the bivariate associations between socio-demographic variables and support for imprisonment in the 58 individual countries from which data were available. The most consistent effect was in relation to gender, with men being more punitive in 46 out of 56 countries. Other relationships were rather less consistent across the 58 countries. Consistency in the social correlates of punitiveness was markedly lower among developing countries. In a second step of their analysis, Mayhew and Van Kesteren carried out a multivariate analysis using loglinear models. Over all, males were found to be more punitive everywhere except in Africa. The second most important predictor of preference of imprisonment was a low level of education, with again Africa being the exception. Young people were slightly more punitive than their elders, though not in East and Central Europe and Latin America. Past victimization by burglary was not significantly associated with a preference for imprisonment. The loglinear models showed that the known correlates could explain only a small part of the variance in punitiveness. Most of the variance between countries remained unexplained. This result is in line with the findings of the regression analyses discussed above. On average, very little of the variance in sentencing options could be explained by the independent variables mentioned.

Country-level Variables

Only a few authors have commented on factors associated with a preference for imprisonment at the country level. Van Dijk et al. (1990) report on a statistically significant relationship between national burglary rates and the proportion of people preferring imprisonment (n = 14). In publications on subsequent rounds of the ICVS, this association was reported again (Besserer 2001; Mayhew and Van Dijk 1997; Van Dijk and Mayhew 1992; Van Kesteren et al. 2000). Japan is mentioned as a country the public attitudes of which are surprisingly high considering the country’s low burglary rates.

Van Dijk and Mayhew have pointed in their reports on the ICVS 1992 and 1996 data at the exceptionally high levels of support for imprisonment in English speaking/common law countries (Mayhew and Van Dijk 1997; Van Dijk and Mayhew 1992). Mayhew and Van Kesteren (2002) have demonstrated a strong, linear relationship between the Human Development Index and support for imprisonment (Mayhew and Van Kesteren 2002). Support for imprisonment is stronger in countries lower down on the social and economic ladder. Their analysis confirmed the exceptional position of the USA and the United Kingdom as countries where the public is very punitive in spite of their high levels of human development.

Multi-level Analysis of Punitiveness: Developed Countries

In this section, I will try to bring the analysis of the social correlates of punitiveness a step further in two respects. First, I will use as dependent variable the scores on the index of punitiveness instead of the dichotomous variable expressing a choice for imprisonment or for any of the other options. As discussed, the index is based on an optimal scaling procedure applied to data from the historical ICVS 1998–2000 dataset concerning developed countries. The second advancement compared to previous work is the execution of an analysis that looks at the correlations at both the individual and country level at the same time.

I have first conducted this multi-level analysis on the dataset from the 2004/2005 ICVS concerning 23 developed, Western countries only. This decision was informed by two considerations. First, the index based on an analysis of global data rated, as discussed, the sentencing option of fines as more severe than a short imprisonment. This result suggests that respondents in (some) developing countries apply other standards to determine severity of sentencing options than people in developed countries. This finding complicates an analysis of data from developed and developing countries together. The second consideration is that results of an analysis of social correlates of punitiveness at the individual level by Mayhew and Van Kesteren (2002) showed relatively little consistency in the findings concerning developing countries. Especially results for Africa were often different from those of the West. Public attitudes towards sentencing in developing countries seem to be determined by other, as yet unexplored factors than in developed countries.

The multi-level analysis was done on data from 23 developed countries participating in the ICVS 2004/05 and EU ICS 2005. In the multi-level analysis we have entered seven variables at individual level mentioned above and an additional five at the country level. At the country level we have entered besides GDP per capita, also a measure of socio-economic inequality (the Gini coefficient). This factor is known to be linked to levels of crimes of violence (Van Dijk 2008). Several authors have interpreted the more severe penal policies in the USA and several European countries in the wider political framework of neo-liberalism and the rollback of the welfare state (Garland 2001; Pratt et al. 2005). In the context of developed countries, socio-economic inequality seemed a useful proxy indicator of (the impact of) such neo-liberal policies. Table 4 shows results of the analysis.

Table 4 Results of a multi-level analysis of social correlates of punitiveness with scores on the index of punitiveness as dependent variable (23 Western developed countries). Data from the 2004/05 ICVS and 2005 EU ICS
Full size table

The results of the analysis can be summed up as follows. At the individual level, the most important, independent correlates of punitiveness are gender (males showing higher scores), low satisfaction with the performance of the police in controlling crime, low level of education and a younger age. These results are broadly in line with those of previous analyses. They confirm the salience of the socio-demographic factors gender and age, as well as educational attainment in explaining punitiveness. A new finding is the importance of confidence in the police. Those who lack confidence in the police tend to be more fearful of crime. As part of this attitudinal package, such people appear to be somewhat more punitive as well. The findings concerning the victimization-punitiveness link are negative. They unambiguously falsify the notion that those recently victimized by crime are more punitive than others.

It is important to note that all effects, although statistically significant due to the large sample size, are very small from a substantive or explanatory perspective. The gender effect is for example 1/16th of a standard deviation and the age effect 1/50th. The other variables on individual level have been standardised and the Beta’s can therefore be interpreted as standardised regression coefficients. The variables explain very little variance on the individual level: σ2 goes down from 0.445 to 0.441, a reduction of a mere 0.004 or 1% of the variance. It is also important to note that at the level of countries, these four factors explain almost no inter-country variance. Composition effects are apparently very small. In other words, differences between countries in terms of age, gender, education and satisfaction with police play no important role in the variance in punitiveness across countries.

At the country level, the analysis confirmed first the association between high rates of crime (victimization by any crime) and punitivenessFootnote 5. The second most important explanatory factor was a measure of economic inequality (Gini coefficient). Together these two country-level variables explained almost half of the inter-country variance. The σ2 at country level drops from 0.034 to 0.017. It is also worth noting that GDP per capita, educational attainment and police performance appeared to be unrelated to punitiveness at the country level. The country-level variable of being Anglophone disappeared as a relevant factor in explaining punitiveness when the impact of Gini and crime level has been controlled for. This finding suggests that there is little reason to speculate about specific cultural factors explaining punitiveness in the Anglophone/common law countries, other than those related to economic inequality.

Replication at Global Level

The most striking findings of the multi-level analysis of the social correlates of punitiveness in developed countries is the association of economic inequality and high levels of punitiveness among the public. This result raises the question whether economic inequality is associated with high levels of punitiveness at the country level globally as well. We have looked at the relationship between scores on the punitiveness index and the Gini coefficient in a selection of developing and developed countries. To extend the number of countries, this analysis was executed on the historical ICVS dataset for urban areas 1989–2000 (latest data available). The results presented in Fig. 6 show that such association is indeed also in evidence at the global level, though less strongly so than among developed countries only. Table 5 presents the correlations per world regions and overall. The correlation coefficient over all is 0.57 (N = 56; P < 0.05). The conclusion that the public is more punitive in less egalitarian societies holds both within world regions and at the global scale.

Fig. 6
figure 6

Country scores on the punitiveness index by scores on a measure of economic inequality. Data from the ICVS/EU ICS and Human Development Report 2004

Full size image
Table 5 Correlations for Gini coefficient with punitivity scale for 56 countries together and by region (latest available ICVS data (urban) and Human Development Report 2004)
Full size table

Finally, we have repeated the multi-level analysis of punitiveness on a dataset from the ICVS 1996–2004 on capital cities or urban areas in 39 countries from different world regionsFootnote 6. Since the majority of these countries are developed countries, we used the scale of punitiveness based on developed countries only. As can be seen in Table 6, the results are very similar to those of the analysis of 23 developed countries. In this dataset, the variable gender was comparatively even more salient.

Table 6 Results of a multi-level analysis of social correlates of punitiveness with scores on the index of punitiveness as dependent variable (39 countries from different world regions)
Full size table

Variance at the country level is to some extent explained by the variables economic inequality and crime rate (victimisation by household burglary). The variable GDP per capita plays no role independently of economic inequality. The similarity of the results of analyses of two different datasets suggests that they are not based on sampling effects.

Public Attitudes and Imprisonment Rates

In several publications on the ICVS, authors have looked at the relationship between prisoners rates per 100.000 inhabitants and public support for imprisonment. In analyses of the first rounds of the ICVS, a moderately strong association was found (Van Dijk et al. 1990). This was confirmed in analyses of ICVS 2000 data on nine developed countries (Besserer 2001). However, analyses of larger numbers of countries showed such relationship to be non-existent (Mayhew and Van Kesteren 2002). Among the countries participating in the ICVS 2004/2005, a weak but statistically significant relationship between public opinion on sentencing and the actual number of prisoners per 100,000 was found (Van Dijk et al. 2008).

The association between public attitudes and prisoners rates is often interpreted as evidence that high prisoners rates are caused by severe sentencing tariffs reflecting public opinion. In my view this interpretation hinges on the unproven assumption that prisoners’ rates can be used as a measure of sentencing severity. In reality, imprisonment rates are determined by the level of serious crime in countries and national capacities to investigate and prosecute offenders, as well as by sentencing tariffs applied by the courts. Countries with high prisoners’ rates may have courts applying especially severe sentencing tariffs, but this is not necessarily the case. High rates of imprisonment may also be the result of higher levels of serious crime and/or the outcome of more effective action in solving such crimes and bringing to justice those who have committed them. Previous studies have found that in the Western world, levels of imprisonment are fairly closely linked to the level of serious, violent crime (Haen-Marshall 1998).

The comparative severity of sentencing tariffs is difficult to determine in a global perspective. Such assessment would require a detailed comparison of the processing of cases through the system and specifically of the proportion and average length of custodial sentences and actual time served for equivalent cases of crime. Such comparisons have been made for eight Western countries (Farrington et al. 2004). The results show that the United States, although not experiencing higher rates of burglaries and robberies than the other seven countries, has higher conviction rates for arrested burglars and robbers and more often imposes custodial sentences in such cases. Mean actual months served in custody of these groups of convicts are the highest in the United States as well. In terms of sentencing severity the USA obviously stands out. Of the other countries under study, Australia seemed to come in second place after the United States on this criterion.

The available global court statistics do not allow a replication of such refined analysis on a global scale. But it seems possible to take into account the seriousness of the national crime problem when interpreting prisoners’ rates. Homicide rates are a reliable indicator of the overall level of violent and organised crime in a country (Van Dijk 2008). Homicide rates can therefore be used as a proxy indicator of the overall level of crime punishable by imprisonment in countries in a comparative context. In order to put the national imprisonment rates in perspective, Van Dijk (2008) related the imprisonment rates per 100,000 population to the rates of homicides per 100,000 population. In his global data set of crime and justice statistics, the imprisonment rates and homicide rates of countries from 2002 are moderately strongly correlated with each other (r = .40, p < 0.000, n = 120). Countries where imprisonment rates are above average tend to have above-average homicide rates as well. The correlation is, however, not very strong. Many countries do not conform to the general pattern. Several countries incarcerate more people per 100,000 inhabitants than the global average, while their domestic levels of serious crime are comparatively modest. Such countries can be regarded as applying comparatively stiff sentencing tariffs. To obtain a proxy indicator of sentencing severity, Van Dijk subtracted rank numbers for imprisonment rates (from low to high) from homicide rank numbers per country (also from low to high)(Van Dijk 2008). Countries with higher positions on homicide than on imprisonment rates obtained positive net scores. Such scores reflect that from a global perspective, imprisonment rates in these countries are relatively low considering their homicide rates. Such incongruity possibly indicates relatively mild sentencing policies. Ecuador ranked highest on this index. Of the major countries, Colombia and India scored in the top 10. In these countries, too, relatively high homicides rates are not accompanied by high imprisonment rates.

Countries with higher positions on imprisonment rates than on homicide rates obtained negative values, possibly indicating severe sentencing or punitiveness. Singapore, a country with low homicide rates and a high imprisonment rate, came out as the country with the highest positive score. The United States was ranked 12 out of 108. Japan is the only country whose rank number on imprisonment is exactly the same as the one on homicide (106). From a global perspective, that country’s use of imprisonment seems to be perfectly balanced with its level of homicides. This finding suggest that Japanese courts apply tariffs that are close to the global mean. Other countries with fairly balanced rates are Australia and Denmark.

To explore whether sentencing severity reflects public attitudes towards sentencing, we have looked at the correlation between the proportion of the public opting for imprisonment and the sentencing severity index just described. To include the largest number of countries possible, I have conducted the analysis on the historical ICVS 1998–2004 dataset (latest data available). Figure 7 shows results.

Fig. 7
figure 7

Scatter plot of proportions preferring imprisonment as sentence for recidivist burglar and (ICVS 1996–2004/05) the index of sentencing severity (Van Dijk 2008)

Full size image

The results of the bivariate analysis show a reversed correlation between the proportion of the public opting for imprisonment and the measure of sentencing severity. This result falsifies the notion that sentencing policies or tariffs reflect public attitudes towards sentencing at a global scale. Although in many developing countries the public favours prison sentences, actual sentencing policies tend to be comparatively more lenient. This is notably the case in countries such as Paraguay, India, Mexico, Argentina and the Philippines. In contrast, sentencing tariffs tend to be fairly severe in several Western countries where public attitudes are comparatively lenient, e.g. Austria, Spain, Poland, Hungary and Rumania. In countries like the USA and New Zealand, sentencing tariffs are very severe while public attitudes are only moderately severe.

Explaining the Lack of Concurrence between Public Attitudes and Policies

To explore the possible correlates of sentencing severity as measured by discrepancies between ranking on homicide rates and imprisonment rates, we looked at the correlations between this “net ranking,” ranging from extremely high positive net values to the highest negative values, and the human development index. This exploration revealed that sentencing severity is positively correlated with Human Development (r = 0.33). In other words, more-developed affluent nations tend on average to be somewhat less restrained with the use of imprisonment, controlling for levels of homicide. The most plausible explanation for the link between human development and imprisonment rates seems to be that affluent countries can simply better afford the funding of large prison populations. This finding suggests that the extended use of imprisonment is in part a luxury that only rich countries can affordFootnote 7. While more authoritarian regimes in developing or emerging countries may ideologically support extended use of imprisonment, budgetary constraints seem to prevent them from actually expanding their prison systems. Democratic governments may have policy-based reservations about imprisonment, but they are less inhibited to expand prison capacity for financial reasons.

Discussion

In this article, I have looked at the social correlates of public attitudes towards punishment across countries using the results of both the older, historical datasets of the ICVS (1989–2000) and the data from its fifth round (2004/2005). Whereas previous analysis of ICVS data on public attitudes towards sentencing have used the proportion of those favouring imprisonment as dependent variable, I have constructed a scale measuring the severity of sentencing preferences of the public. The scale rates the options in the following order: suspended sentence, community service order, fine, short term prison sentence (less than a year), other sentence, medium term prisons sentence (1 to 2 years) and long term prison sentence (3 years or more). This scale was constructed using an optimal scaling procedure using the older, historical ICVS dataset of the developed countries. Through the same procedure, a scale was also constructed for data on a global scale from the ICVS 2004/05. An important finding was that the scale that emerged from the analysis of data from developed countries differed in one essential respect from the global scale. In the global scale a fine was rated as more severe than a short term prison sentence. This result casts doubt on the validity of previous analyses of global data on the proportion of people favouring a prison sentence. Some respondents from developing countries opting for short prison sentences have been counted as being more punitive, but might in fact have been less punitive than those favouring a fine. This effect may have distorted analyses of social correlates of punitiveness at a global scale. This distortion may also explain the lack of regional consistency in the results of previous analyses of global data on punitiveness (Mayhew and Van Kesteren 2002).

In my own analysis of social correlates, I have first focussed on the ICVS 2004/05 data for 23 developed countries. Unlike previous analysts, I have carried out a multi-level analysis, looking at correlates at both the individual and country level at the same time. The analysis was replicated on a set of historical ICVC-based data regarding 39 countries including a sizeable minority of developing countries. The results of both analyses were very similar. Since these analyses relate to data on all major developed countries and a broad representation of countries from other parts of the world, the results can be regarded as roughly representative for the world population.

The analytical results first confirmed that males tend to be somewhat more punitive than women, independently of other factors. There seems to be no obvious explanation for this robust finding. One possible explanation is that (older) women may consider recidivist burglars as potential breadwinners of (extended) families and may therefore be more in favour of non-custodial sentences, such as community service, than of imprisonment. The second most important socio-demographic factor predicting punitiveness is age. Older people tend to be more lenient in their choice of sentencing options. This factor seems to be more than just a cohort effect caused by especially harsh attitudes of the current generation of youngsters. The young age-punitiveness linkage has also been observed in the historical ICVS dataset. In fact, Buikhuisen and Van Dijk reported on the curious finding that people become more lenient or tolerant with age in a Dutch study from the 1970s (Buikhuisen and Van Dijk 1975). One possible explanation is that older people are more familiar with issues of crime and justice and therefore more aware of the social costs of imprisonment. Possibly, older people are also more inclined to weigh in the loss of family income as an important cost of prison sentences.

The results confirm the finding of previous analyses that educational attainment is an important independent predictor of punitiveness. To the extent that people have received more years of education they are more in favour of non-custodial sentences or shorter prison sentences. This finding supports our line of interpretation that support for (longer) prison sentences is reduced by more information about the social costs of imprisonment.

The analysis of the relationship between victimization and punitiveness showed negative findings. These results suggest that recent victims, however angry about what has been done to them, do not necessarily translate their personal experiences of crime into a preference for imprisonment. One consideration behind the support of victims for other sanctions than tough prison sentences might be their personal stakes in obtaining compensation from the offender.

Finally, the individual level effects point at the importance of lack of confidence in the local police as a predictor of punitiveness. This last result may indicate that concern about crime rates fuels more punitive attitudes towards sentencing.

At the level of individuals, the variables entered into the analyses explain only a small part of the variance. This result proves that there remains ample scope for new hypothetical explanations of punitiveness. Worth exploring seems in particular the notion that much variance in public attitudes towards imprisonment or sentencing generally is based on the extent of factual information on the costs and benefits of various sentencing options.

At the country level, two factors stand out as predictors of punitiveness, comparatively high levels of common crime, including burglary and economic inequality. The first factor points again at the salience of concerns about levels of crime in sentencing preferences. Some people seem inclined to blame existing crime problems to lenient sentencing (Costelloe et al. 2002). This mechanism can explain the positive association between actual burglary rates and punitiveness (Besserer 2001). In countries where crime is more of a concern the public seems more inclined to choose imprisonment as the preferred option.

The association of punitiveness with economic inequality emerged both from the analysis of data from developed countries and of data of a mixture of developed and developing countries. This result indicates that worldwide people in less egalitarian societies are more likely to opt for more severe sentencing options. The result suggests that attitudes among the public towards sentencing are more punitive in countries pursuing neo-liberal policies where economic inequality is greater (Pratt et al. 2005). The result could also be seen as evidence for increased sensitivity to the infliction of pain among the public in more interdependent, egalitarian societies. The linkage between economic inequality and punitiveness at any rate merits further empirical scrutiny and theoretical reflection.

The results concerning the possible association between public attitudes to sentencing and actual sentencing tariffs of national courts produced negative results. Among developed countries some positive association may exist, but at the global level sentencing tariffs were found to be inversely related to public attitudes. In countries where the public is more in favour of imprisonment, judges are more reluctant to impose such sentences. This paradoxical result can be explained in economic terms. In developing countries governments cannot afford the “penal populism” of the West for budgetary reasons. In this light, extensive use of imprisonment, such as in the USA, can be seen as a collective “luxury good” or, to put it more sharply, a rich man’s folly.