Attitudes to Crime and Punishment in England and Wales, 1964–2023: A Reinterpretation of the 1980s and a Model of Interactions Between Concern, Punitiveness and Prioritization

Abstract

This paper assembles the largest set of British survey questions about criminal justice to date (1,190 question-year pairs) and uses it to measure crime concern, punitiveness, support for the death penalty and the prioritization of crime as a social issue from the 1960s to today. Results lend some support to existing narratives of public opinion, showing that concern and prioritization grew steadily through the 1970s before declining from the mid-2000s, and that support for the death penalty has been falling since at least the 1960s. But they contradict orthodox accounts of the 1980s as a period of rising punitiveness, showing instead that support for tougher policing and sentencing was highly volatile and subject to significant demographic variation until the late 1990s. I also show that crime concern is particularly responsive to the true rate of crime and propose a model for the interaction between these different strands of public opinion.

INTRODUCTION

What the general public thinks about crime and punishment is a vexed question. On the one hand, politicians often claim to be following ‘the people’ in their support for tougher policing and sentencing (Miller 2016). But on the other hand, there is little consensus on what ‘the people’ really want. Amongst scholars who focus on the contemporary period, there are disagreements about who thinks what, how much attitudes have changed and why (see summaries in Gray et al. 2008; Farrall et al. 2009; Hough et al. 2013; Enns 2016; Gramlich 2020). And there are also disagreements among those who look across the longue durée, with some writers finding evidence of widespread moderation and others a more ambiguous transformation (Durkheim 1900 [1969]; Foucault 1975 [1991]; Pratt 2002).

A recent intervention in The British Journal of Criminology, entitled ‘Public Responsiveness to Declining Crime Rates in the United States and England and Wales’ (2022), represents a major advance in these debates. By compiling the aggregate results of a large number of different surveys, Peter Enns, Jacob Harris, John Kenny, Andra Roescu and Will Jennings provide the most comprehensive estimates yet for two key dimensions of public opinion: crime concern, and punitiveness. They then go on to make four claims: (1) In England and Wales, crime concern tracks the real rate of crime, with both rising in the 1980s, peaking in the early 1990s and falling steadily thereafter; (2) In the United States, the correspondence between crime concern and crime itself is much weaker, but both are now lower than they were in the early 1990s; (3) In both countries, punitiveness is highly responsive to crime rates, increasing during the 1980s and falling significantly in recent decades and (4) Where the real level of crime diverges from police records, public opinion responds to the former and not the latter.

My purpose in this paper is to conduct a direct replication of Enns et al.’s study of attitudes to crime and punishment in England and Wales, drawing on a much larger sample of survey questions (1,190 question-year pairs) to test the robustness of those claims and to extend their analysis in several directions. On the positive side, I confirm that concern about crime did rise steadily in the 1980s before falling back after 1995, that punitiveness has fallen over the last 15 years and that neither was affected by the recent (artificial) increase in police recorded violent crime. Extending their analysis, I then produce a new estimate for the prioritization of crime as a social issue, which seems to follow a similar trend to that of crime concern: increasing from the early 1970s until the mid-2000s, before falling back significantly. Replicating earlier work by some of the paper’s co-authors (Jennings et al. 2017), I also show that people have been steadily turning away from the death penalty over the last 60 years, a trend which seems to be totally unresponsive to changing levels of crime. Finally, I show that these trends in public opinion are, for the most part, shared across different demographic groups (the young and the old, men and women and white and non-white communities) and that all these findings satisfy a series of robustness checks (including Bayesian Item Response Theory [IRT] methods, jackknife procedures and comparisons with the underlying question series).

There are, however, several areas where my findings diverge from those of Enns et al. Most importantly, although their paper focusses on the last three decades, Enns et al.’s estimates begin in the 1980s and at numerous points in the paper they repeat the claim that ‘during much of the second half of the twentieth century, public opinion in the United States and Britain became more punitive as crime rates rose’ (Enns et al. 2022: 1093). However, my estimates reveal a much more complicated picture. In England and Wales, punitiveness does not seem to have increased substantially during the 1980s and 1990s. Instead, the aggregate trend was high but volatile in the 1980s and early 1990s, before stabilizing under New Labour and falling substantially thereafter. There was also significant demographic variation: with the over-50s becoming steadily less punitive since 1981, while the trend for younger people is more n-shaped, reaching a peak in the late 2000s.

I then conclude by examining how much the three main dimensions of public opinion (concern, punitiveness and prioritization) respond to the true level of crime and how they interact with one another. Using a structural equation modelling framework, I show that crime concern is much more responsive than the other two dimensions and that it works as the ‘transmission mechanism’ through which the underlying crime rate affects punitiveness and prioritization. To put it another way, it is only when people are worried about crime that they start to prioritize the issue and become more punitive. These are only preliminary examinations and further work will need to be done to confirm and extend these findings, but I hope that this analysis proves the value of thinking about attitudes to crime and punishment in a disaggregated and interactive way.

In the spirit of Open Science (UNESCO 2022), I have also made all the aggregated data publicly available for the first time, allowing other scholars to put these materials to use in their own work (Tiratelli 2024).

DATA AND METHODOLOGY

Over the last 20 years, there has been a growing awareness of the importance of reproducing and replicating scientific findings (Ioannidis 2005; Pridmore et al. 2018; Moody et al. 2022). This paper therefore sets out to conduct a direct replication of Enns et al. (2022): mirroring their procedures as closely as possible and comparing the two sets of results (Schmidt 2009; Nosek and Errington 2020; Calin-Jageman 2023). In brief, this involves collecting British survey data from a wide range of sources, harmonizing question responses across those different surveys and then using the dyad ratios algorithm (Stimson 1991; 2018) to estimate latent trends from the aggregated survey results.

Following Enns et al. (pp. 1104, 1107–9), I therefore begin by searching through seven survey series looking for questions relevant to crime and punishment: the British Election Study (BES), which includes 15 post-election cross-sectional studies and nine separate panel studies (all available through the BES website), covering the period from 1963 to 2023;¹ the British Social Attitude survey (BSA), conducted annually since 1983 and available through the UK Data Service (NatCen Social Research 2023); the Crime Survey for England and Wales (CSEW), previously known as the British Crime Survey and conducted at irregular intervals between 1982 and 2001 before becoming an annual survey, available through the UK Data Service (ONS 2021); YouGov have produced two relevant collections, their Crime Trackers collection (2019–23) and their Most Important Issue Tracker (2011–23), both available through the YouGov website²; the Ipsos Issues Tracker is also available through their website and covers the period 1974–2023³; finally, Gallup polls from 1938 to 2000 were compiled from three sources (Gallup 1976; King et al. 2001; Jennings et al. 2022). For reasons that will become apparent below, only questions that were asked on at least two occasions were included in the analysis.

Following Enns et al., I then distinguished between crime concern (‘perceptions of the crime rate and the degree to which the public is fearful or concerned about crime’ p. 1103) and punitiveness (‘the degree to which public opinion supports being tougher on crime or supports less punitive and more rehabilitative policies’, p. 1103).⁴ I also follow them in excluding ‘questions about confidence in policing and criminal justice’ (p. 1109). Table 1 shows the results of this data-gathering exercise for the period covered in Enns et al.’s paper. As we can see, there are far more questions available than were included in their original analysis.

The fact that I uncovered more than twice as many questions for crime concern and 60 per cent more for punitiveness would be noteworthy on its own. However, this replication study is also broader in several ways (see Table 2). First, I include new data on the prioritization of crime as a social issue (questions like ‘What would you say is the most important issue facing Britain today?’ [Gallup]) and on support for the death penalty (questions like ‘The death penalty, even for very serious crimes, is never justified’ [BES]). This separation of the death penalty from other forms of punitiveness deserves some comment. At the level of theory, one might think that someone who was in favour of harsher punishment, in general, would support the death penalty as well as longer prison sentences and tougher conditions inside. But, as with most other elements of ideology, there is no necessary logical reason for why particular sets of opinions hang together (Author Self Cite). In fact, the meaning of execution, incarceration and other forms of state-sanctioned punishment has been a site of political and social struggle for several centuries and capital punishment today has become a distinctive political issue with its own historical trajectory (at least in Britain—Jennings et al. 2017).

Second, I extend the temporal coverage of the data, with questions measuring crime concerns going back to 1965, those for the death penalty going back to 1964,⁵ and those for the prioritization of crime starting in 1973. Third, wherever possible I retain separate scores for different demographic groups, splitting up the original samples by gender, age and ethnicity. (Due to small sample sizes, particularly in earlier surveys, I focus the analysis below on fairly crude categories: women/men, over-50/under-50 and white/non-white. But it would be possible to disaggregate further if we restricted the analysis to the period covered by large surveys like the most recent editions of the CSEW.) This allows me to present a much more detailed and expansive account of how attitudes to crime and punishment in England and Wales have shifted over the last 50 years. All data and a complete list of the original questions is available from the UK Data Service (Tiratelli 2024) (see also Supplementary Appendix A1).

I harmonized these different survey questions by calculating the average binary response for each question-year pair, using survey weights where possible and taking the weighted annual average if the same question was asked multiple times in 1 year. In all cases, higher numbers indicate a greater level of concern/punitiveness/etc. Following Enns et al., if the same question is asked in a different poll/survey, I treat it as a new question series, thus controlling for house effects (constant biases associated with a survey or polling company, for example, different recruitment methods or interview techniques). Similarly, if the same question is posed to different respondents in different sections of the same survey, I treat it as a different question series, thus controlling for question order effects. The various question series for each dimension of public opinion are plotted over time in Supplementary Appendix A2.

In order to estimate the latent trends behind this mass of aggregated survey results, I follow Enns et al. in relying on James Stimson’s (1991; 2018) dyad ratios algorithm. This technique has been described in detail elsewhere (including in Enns et al., p. 1102–3) and so I will concentrate here on the general intuition. The key move is to focus not on the absolute score for each question but rather on the ratio between pairs of questions. For example, if in 1991 60 per cent of people replied ‘Yes’ to the question ‘Do you think that crimes have increased in your area during the last 6 months?’, but only 50 per cent said ‘Yes’ in 1995, then the ratio for 1995 would be 0.5/0.6. The algorithm then calculates the average of all available ratios for each point in time and, starting from an arbitrary initial value, produces a time series that represents the trend in the underlying latent variable. Three other features of the algorithm are worth noting: (1) as the procedure only uses information from previous periods in calculating each ratio, the algorithm is applied forward and backward in time before taking the average of the two; (2) each question series is weighted according to its correlation with the latent trend, then the trend is recalculated and new weights applied in an iterative process until they converge and (3) the estimated trend is also exponentially smoothed.

The dyad ratios algorithm is widely used in social and political science, but it is not without its critics. In particular, Anthony McGann has queried its lack of clear micro-foundations and proposed an alternative framework grounded in Bayesian IRT (see McGann 2014; Bartel et al. 2019; McGann et al. 2019; and the response from Stimson 2018). In other cases, the trends derived from these two approaches have proven to be highly correlated (see McGann 2014; Stimson 2018; Alvarez 2023), and the same is true here, with the IRT estimates confirming the broad patterns described below (see Supplementary Appendix A3).⁶ This should give readers confidence that these results are an adequate (if not perfect) representation of the real policy mood in England and Wales.

ATTITUDES TO CRIME AND PUNISHMENT, 1964–2023

Starting with crime concern, my analysis suggests that crime concern rose steadily through the 1970s and 1980s, increasing by more than 40 per cent over a 20-year period (Figure 1). It then remained fairly high throughout the 1990s, before declining steadily after 2003, eventually reaching levels not seen since the early 1970s. As I will argue in more detail below, this adds further support to the argument that people are aware of the real level of crime in society and update their beliefs in response to it (contra Gramlich 2020; Koerth and Thomson-DeVeaux 2020).

This confirms Enns et al.’s (2022) findings but extends it into the 1960s and 1970s. The earlier data for this period comes from a set of 11 Gallup polls which asked whether respondents considered a series of issues to be ‘raising very serious social problems in Britain today’. Fortunately for my purposes, the list includes several types of crime (crimes of violence, drug taking, large-scale organized crime and rape) alongside prostitution, gang warfare and juvenile delinquency, all of which are included as separate question series in the analysis. I also subject these findings to several robustness checks, showing that they mirror the trends in the key underlying question series and are robust to a jackknife procedure (iteratively recalculating the latent trend after dropping each of the question series in turn) (Supplementary Appendices A4 and A5).

I will now turn to the second dimension of public opinion. Enns et al. claim that public desire for tougher policing and harsher punishments (excluding the death penalty) increased throughout the 1980s, before plateauing in the 1990s and then falling in the 2010s (pp. 1093, 1102). They also repeatedly cite earlier work by some of the paper’s co-authors which argues that ‘penal populism’ (a broader measure that combines crime concern, punitiveness and dissatisfaction with the police and courts) increased steadily from the mid-1980s and that this shift in public opinion drove the increase in incarceration in that period (Jennings et al. 2017). This is not the place to tackle the broader question of what caused the massive growth of Britain’s prison population after the 1980s (although putting it down to public opinion is made more difficult by the fact that Britain’s prisons are still overflowing despite a decade of falling punitiveness and concern). Instead, I want to focus more narrowly on Enns et al.’s account of public opinion. There are very few relevant survey questions for the late 1970s and early 1980s and so the estimated trend for punitiveness is highly sensitive to which year is taken as the starting point (see the three trend lines in Figure 2, starting in 1978, 1979 and 1981).⁷ Following Enns et al., my preferred series starts in 1981 with three usable question series (a number that quickly increases) and which closely tracks the trends in the underlying variables and is robust to the jackknife procedure described above (Supplementary Appendices A6 and A7). From that starting point, there is some evidence of a slight increase in punitiveness: the decadal average is 2 per cent higher for the 1990s than in the 1980s, and a further 3 per cent higher in the 2000s. But, of the 16-question series covering the period from the late 1970s to 1990, less than one-third show an upward trend and the unweighted average actually shows punitiveness falling over that period. There is also some indication of a downward trend in punitiveness in the IRT estimates (Supplementary Appendix A3). All of this makes it very difficult to sustain the claim that public opinion became noticeably more punitive in this period.

This creates space for an alternative interpretation of the 1980s. Rather than punitiveness following crime in an escalating cycle of violence and retribution, I would argue that the 1980s was characterized by a highly unstable public attitude towards punishment and imprisonment. In particular, I want to draw attention to two phenomena: the deeper historical roots of Britain’s ‘disciplinary common sense’, and the significant volatility of the 1980s and the early 1990s. Starting with those historical roots, Stuart Hall’s pioneering account of the rise of Britain’s ‘law-and-order society’ focusses on three key moments: the imperial crises of the late nineteenth century, which generated popular anxiety about British decline and the massive centralization of the British state; the backlash against the social liberalism of the 1960s; and the economic slowdown and legitimacy crisis of the 1970s (Hall et al. 1978; Hall 1980a). By the mid-1970s, these processes had led the state to be increasingly dependent on coercion and policing to manage social unrest and committed to building popular consent for this use of force. This is what Hall famously described as ‘authoritarian populism’ (Hall 1979; 1985). And it meant that there was already a deep well of public support for punitive law-and-order politics well before the 1980s.

The other important feature in the history of punitiveness is the substantial reduction in volatility between the 1980s and the late 1990s (Figure 2). Knowing how profound Thatcher’s transformation of the British political and economic system would prove to be, it is easy now to forget how contested, contradictory and cautious her reign really was. In particular, while Thatcher did win, in Hall’s telling phrase, a ‘measure of electoral support’ (Hall 1980b: 26), attitudes towards crime and punishment were far from settled in this period. 1981 saw an outbreak of anti-police riots across Britain’s metropolitan centres, beginning in April in Brixton after an aggressive stop and search operation targeting the black community there, but quickly growing to incorporate other sectors of the British working class (Hall 1981). That same year, the H-Block hunger strikes in Northern Ireland became international news, leading to the death of 10 Republican prisoners. In 1982, the Gay London Police Monitoring Group emerged out of a decade of gay and lesbian militancy. In 1984, footage of police officers breaking their batons on the backs of striking miners was (despite the best efforts of the BBC) broadcast on the evening news (BBC 2014; Milne 2014; Conn 2015). In 1985, there was a series of anti-police uprisings across London after bungled police raids left one black mother paralyzed and another dead. This was also the time of football ‘hooliganism’ (Bufford 1992), the Battle of the Beanfield (Worthington 2005), of the heyday of the National Council for Civil Liberties (Jessop et al. 1984: 53). I am not claiming that any of these political moments commanded majority support from the British public. But they do represent a resilient current of anti-police sentiment and suggest that, despite Thatcher’s attempts to remake British common sense, her ‘tough on crime’ agenda remained a contested issue throughout her time in office. In fact, as Figure 2 shows, it was not until the 1990s and, in particular the New Labour period, that a settled law-and-order consensus finally emerged.⁸

Turning to the death penalty, here my findings again reinforce earlier work, showing a decline of nearly 25 per cent in support for capital punishment since the last execution in Britain 60 years ago (Jennings et al. 2017; Hoyle 2023). This pattern is robust to different starting dates (1964 and 1981 are shown below), the jackknife procedure and closely follows the trends in the underlying variables (Supplementary Appendices A8 and A9). It also seems to mirror the decline in support for corporal punishment, which Gallup polls show fell in the early 1980s.⁹ This does not mean that support for the death penalty is not still fairly widespread (it is also polarized—Tiratelli 2023: 53–4), but simply that it is trending down and has been doing so steadily for several decades. There are many different explanations for this shift in attitudes, but the fact that it is such an impressively secular decline, seemingly unresponsive to changes in criminality, politics or the economy, does suggest that something as grand as Norbert Elias’s ‘civilising process’ might be at work here (Elias 1939 [2000]). It also reinforces the need for a distinction between support for capital punishment and support for longer, tougher imprisonment. It is important to remember however that there is nothing logically necessary about that distinction. Instead, it is a result of a history of social and political contestation which has marked out the death penalty as a distinctive political issue, a more barbarous predecessor to our modern, humane systems of punishment (Pratt 2002, chapter 1).

The fourth dimension of public opinion I track is the prioritization of crime as a social issue. The question of ‘what is the most important issue or problem facing the country’ has been a core feature of British polling since the 1940s (Jennings and Wlezien 2011). Gallup has data on the ‘most urgent problems facing the government’ from the late 1940s, and those facing ‘the country’ from the late 1950s (Gallup 1976). But ‘Law and Order’ only started to appear as a category in these series from 1973, which suggests that it was a rather insignificant issue before that point. Ipsos then began its own series on the ‘most important issue facing Britain today’ in 1974. The BSA began to ask whether people ‘would like to see more or less government spending on…’ in 1982. YouGov’s ‘most important issue’ series then started in 2010. And the BES Internet Panel study included the same question from 2018. These five question series all track each other extremely closely (Supplementary Appendix A10) and the latent trend shows a general increase in the prioritization of law-and-order from the mid-1970s until the late 2000s, after which it fell substantially (Figure 4). But it is also worth noting that this series is extremely volatile. In fact, the coefficient of variation (σ/μ) is 0.6 for the prioritization series, compared with only 0.14 for concern, 0.07 for punitiveness and 0.07 for the death penalty. This volatility should not be surprising given that attention and prioritization are normally considered to be far more changeable than beliefs and values (Evans et al. 1996; Tiratelli 2023: 51–2). Moreover, the total number of people choosing crime as the most important social issue of the day has always been fairly small (far lower than the number choosing the economy, health or international affairs for example) and this therefore makes the change ratios much larger.

In sum, the data assembled here provides robust estimates for four dimensions of public opinion: (1) Concern about crime seems to have risen from the mid-1960s, reaching a peak in the 1990s, before declining through the 2010s. (2) People’s prioritization of crime has followed a similar, albeit more volatile and pattern. (3) There were already deep wells of support for tougher penal punishment in the 1980s, but it only became a stable consensus in the late 1990s and only started to fall in the last 10 years. (4) Support for the death penalty has been falling steadily for more than half a century, a decline which shows no sign of abating. I will now turn to the question of how this data can inform our understanding of opinion formation in general.

ANALYSIS

Variation across demographic groups

Enns et al. ends their paper by calling for research ‘exploring how public responsiveness differs between different subgroups’ (p. 1112). Therefore, wherever possible, I have estimated separate latent trends for men/women, those over 50/under 50 and for white/non-white groups. These are fairly crude categories but, as mentioned above, more fine-grained analysis quickly runs into the problem of very small samples. Even using a category as vague as ‘non-white’, I had to start that series in 1991 to avoid unreliable, small-n samples distorting the results. Unfortunately, it also proved impossible to find consistent measures of class across time and across surveys and so, as with the more detailed analysis of ethnic subgroups, this will have to wait for a more focussed analysis using a subset of larger surveys. It is also not possible to disaggregate the trends in prioritization for different subgroups as only one of the sources (the YouGov Most Important Issue Tracker) contains a demographic breakdown.

Despite these limitations, the bivariate correlations shown in Table 3 reveal a remarkable level of consistency across these different demographic groups. It is important to note that this does not mean that these different subgroups have the same level of concern, punitiveness or support for the death penalty, but merely that the different trends move in parallel with one another. This suggests that different parts of the population are responding fairly evenly to shifting levels of crime and to political and economic events.

The one exception to this rule is the trend in punitiveness across age groups. Indeed, as Figure 5 shows, punitiveness for the over-50s has declined steadily from the mid-1980s, while the trend amongst younger people increased from the early 1980s until 2009 and fell sharply thereafter. This echoes recent work showing that those who came of age under Margaret Thatcher and Tony Blair have distinctive (and more right-authoritarian) political attitudes (Grasso et al. 2019; Gray et al. 2019). It also adds a further wrinkle to the discussion above about the level of punitiveness in the 1980s, reinforcing my argument that a simple story of increasing penal populism cannot do justice to a more complex pattern of contestation, volatility and demographic variation.

Does public opinion respond to changes in crime?

A key question for scholars of public opinion is whether people’s beliefs track objective conditions. In this case, there is a substantial debate about whether concern, punitiveness and prioritization reflect real rates of crime (Gramlich 2020; Koerth and Thomson-DeVeaux 2020; Enns et al. 2022). Measuring crime is notoriously difficult, but there are five commonly used metrics for England and Wales: CSEW estimates of violent crime and property crime, and police records for property crime, violent crime and homicide (Figure 6).¹⁰ These patterns will be familiar to many readers from the global north: crime increased in the 1970s and 1980s, before declining steadily and significantly thereafter. (It is worth pointing out that murder rates reach their peak somewhat later than other crime categories, only starting to decline in the early 2000s.) The only apparent exception to this is police-recorded violent crime, which has been increasing relentlessly since the late 1990s. It is now widely accepted that this series is extremely unreliable and that the recent increases have been driven by changes in police recording practices (Stripe 2023). Indeed, since 2014, it has no longer been used as an official ‘National Statistic’. Nevertheless, as Enns et al. points out (p. 1098), this discrepancy does allow us to test whether public opinion responds to the real level of crime or to police records.

As expected, the bivariate correlation coefficients show that the different dimensions of public opinion respond to the four more accurate measures of crime but not to police-recorded violent crime (Table 4, support for the death penalty has been dropped because its linear trend is an obvious outlier). But comparing the graphs in Figure 6 with those of the latent trends (Figures 1–4) also reveals some interesting patterns. First, only crime concern responds immediately to the decline in crime from the mid-1990s, with prioritization and punitiveness staying higher for longer. Second, punitiveness does not seem particularly responsive to the clear increase in crime in the 1970s and 1980s. Taken together, this suggests that crime concern might be more responsive to crime than the other dimensions of public opinion. To further investigate this possibility, I use maximum-likelihood factor analysis to create a latent variable combining both CSEW series and police records of property crime and homicides (Table 5). I then use that latent variable as the predictor in three bivariate regressions with the different dimensions of public opinion (all standardized to enable comparison). The results are shown in Table 6 and confirm that different dimensions of public opinion are differentially responsive to crime itself. The effect on prioritization is not significant at all, and the effect on punitiveness is nearly a third smaller than that on concern. These results therefore lend further support to the idea that crime concern is particularly responsive to the real level of crime.

A model for the interaction of concern, prioritization and punitiveness

A related theoretical question is how different dimensions of public opinion interact with one another. In this case, concern, punitiveness and prioritization are strongly correlated, with correlation coefficients that vary between 0.72 and 0.76.¹¹ Building on the finding of differential responsiveness, this suggests a simple model where concern acts as the transmission mechanism between real levels of crime and other attitudinal variables (Dowler 2003 and Armborst 2017 suggest a similar role for the related but narrower concept of ‘fear of crime’). I also hypothesize that prioritization leads to punitiveness, on the grounds that a tougher system of policing and punishment would only be needed if crime was a serious issue. In sum, the model I am proposing implies that worrying about crime makes people more punitive and more likely to prioritize crime as a social issue, which in turn then also feeds into their desire for harsher punishment. This hypothesized model is depicted graphically in the path diagrams in Figure 7.¹²

Turning to a structural equation modelling framework, I test this transmission mechanism model against two different individual-level datasets, using survey data from the BSA (n = 5,143 individuals) and BES (n = 40,290 individuals). (These represent the longest runs of surveys which contained a consistent question for each dimension of public opinion.) I first pooled BSA surveys from 1990, 1994 and 1996, and measured crime concerns with the question ‘How common are incidences of burglary in this area’, punitiveness with ‘Criminals should be given stiffer sentences’ and prioritization with ‘Please show whether you would like to see more or less government spending in each area [Law and Order]’. Next, I combined three waves of the BES Internet Panel (May 2018, June 2019 and May 2023). The concern is measured using the question ‘Do you think that the level of crime is getting higher, getting lower or staying about the same?’, punitiveness by ‘People who break the law should be given stiffer sentences’, and prioritization by ‘As far as you’re concerned, what is the single most important issue facing the country at the present time?’. Only 32 per cent of respondents answered these questions on more than one occasion and so, rather than treating the BES data as a full panel study, I took the first of those repeated responses, transforming it into a pooled cross-section.

I then control for the level of crime by de-meaning each of the variables by year (this is known as the ‘within transformation’ and is equivalent to time-period fixed effects). I also include controls for gender, age and ethnicity (white/non-white). For ease of interpretation, these control variables are omitted from the path diagrams shown below. All models were estimated using lavaan (Rosseel 2012, vers. 0.6-17) with standardized variables and coefficients, and bootstrapped standard errors with 50,000 iterations. Full regression tables can be found in Supplementary Appendices A11 and A12.

The results of these two tests are displayed as path diagrams in Figure 7. Taken together, they provide strong preliminary evidence in support of the transmission mechanism hypothesis, showing that crime concern has an independent effect on the other two dimensions of public opinion. This suggests that it is primarily concern that drives people to focus on the issue of law and order and to prefer more punitive policy solutions—a finding with profound implications for those who are committed to reforming the current system of policing and incarceration in Britain. However, there is significant variation in the relative weights of the paths across the two models and these tests should not be seen as the final word on how attitudes to crime and punishment interact. I will therefore leave it to others to explore and enhance this model. In particular, it would be useful to try to incorporate measures of elite opinion, the effects of the media, and the various moral panics that play such an important role in the development of penal policy.

DISCUSSION AND CONCLUSION

Enns et al.’s ‘Public Responsiveness to Declining Crime Rates in the United States and England and Wales’ (2022) marks a significant advance in our knowledge about the evolution of public attitudes to crime and punishment in the Anglophone world. But as is now well-documented, empirical studies need to be subjected to careful replication before being accepted as definitive (Ioannidis 2005; Pridmore et al. 2018; Moody et al. 2022). In this article, I have therefore attempted a direct replication of their analysis for England and Wales. I began by uncovering a much larger set of relevant survey questions (1,190 question-year pairs), allowing for more confident estimations of their chosen trends in public opinion. I then extended their analysis in several different directions: examining a longer time period, different demographic groups, new dimensions of public opinion and also complicating their analysis of responsiveness to crime. Finally, I show that all these results pass a series of robustness checks: Bayesian IRT estimation of the latent trends, jackknife procedures and comparisons with the underlying question series.

What then do these findings tell us about public attitudes to crime and punishment in England and Wales? Starting in the 1970s and 1980s, the British public seems to have become more concerned about crime as, for the first time since World War II, rates of violence and property crime started to rise. They also gradually started to prioritize law and order as one of the major public issues of the day (although it never got close to the levels of prioritization given to the economy, health or international affairs). But if people were more worried about crime, they weren’t necessarily becoming more punitive in how they wanted the state to respond. In fact, support for the death penalty and for corporal punishment seems to have been falling steadily for many years and, while there was already a deep reservoir of support for tougher policing and longer prison sentences, the 1980s and early 1990s were marked by significant contestation, volatility and demographic variation. But by the time Tony Blair came to power in 1997, these different trends had converged: concern about crime was still high, opposition to punitiveness had fallen away and law-and-order had become a priority issue for many people. But in the last decade, as crime rates continued to fall, so too did the public’s concern about it, their prioritization of it and their punitiveness. To be blunt, we now live in a comparatively low-crime era, and people’s attitudes to some extent reflect that. The fact that this hasn’t translated into a liberalization of law-and-order politics remains a significant problem for historians and social scientists to grapple with.

Aside from this historical narrative, my analysis can also inform our wider understanding of opinion formation. First, I show that crime concern seems to be more directly responsive to changing crime rates than the other dimensions of public opinion (punitiveness, prioritization and support for the death penalty). Second, I show that the different dimensions of public opinion are connected, with concern acting as the transmission mechanism through which prioritization and punitiveness respond to the underlying level of crime. Finally, although my analysis does not directly address the question of whether shifts in public opinion were driven by politicians and the media, I can make a few suggestions on the basis of the evidence presented above. On the one hand, concern and prioritization seem to follow underlying trends in crime fairly closely, indicating that elites played only a small role in that sphere. On the other hand, my results show that punitiveness stayed high until 2010, despite crime falling steadily from the early 1990s. This discrepancy suggests that competition between the Conservative and Labour parties to prove their ‘tough on crime’ credentials may have been effective in solidifying a punitive turn in the wider British public, a hypothesis which is also supported by the divergent trajectories of different age groups (we would expect younger people growing up under Margaret Thatcher and Tony Blair to be particularly susceptible to their anti-crime agenda—Grasso et al. 2019; Gray et al. 2019). This may not be a definitive account of opinion formation, but I hope that it provides scholars with a useful model for thinking through the evolution of public opinion and that it can feed into a discussion about where those interested in changing Britain’s police and prisons should focus their energies.

SUPPLEMENTARY MATERIAL

Supplementary material is available at British Journal of Criminology online.

ACKNOWLEDGEMENTS

I would like to thank Soraya Odubeko and Wenbo Saigo for excellent research assistance, Peter Enns and Will Jennings for their patient explanation of their data collection process and the anonymous reviewers at BJC for their careful critiques of earlier drafts.

FUNDING

No funding was received for this publication.

REFERENCES

Footnotes