Exploring individual-level predictors of punitive attitudes in Australia

ABSTRACT

Western countries tend to display preferences for the harsh punishment of people with criminal justice involvement. Drawing on a representative survey of the Australian population, the present study explores punitive attitudes and what factors shape the development of these attitudes at an individual level. More specifically, the study considers the role of age, sex, level of education, geographic location, perceptions of crime, fear of crime, confidence in the criminal justice system, media consumption, crime causation, beliefs in redeemability, interpersonal trust, political ideology, racial essentialism, and minority threat in predicting the punitive attitudes of Australians. The results indicate that generally Australians are somewhat punitive and that the strongest predictors of these attitudes are: internal attribution of crime, perceptions of rising crime rates, a lack of belief in redeemability, geographic location (specifically rural areas), a lack of interpersonal trust and a lack of support for multicultural principles together, creating a more robust understanding of punitive attitudes in Australia, which is currently lacking.

KEYWORDS:

• Punitive attitudes
• punitiveness
• predictors
• criminal justice policy
• Australia

Introduction

Punitive attitudes have been consistently identified in Western nations including Australia, Canada, the United Kingdom, and the United States. Harsh criminal justice responses to crime are understood as being driven in part, by the publics’ desire for harsh penalties against people with criminal justice involvement (PCJI) and populist politicians willing to empower such sentiments for political gain. Flowing on from this, public opinion and attitudes are central to the implementation of policy, therefore the public is a powerful political figure when it comes to criminal justice policy. As such, it is imperative to understand public attitudes and how they are shaped.

Drawing on a representative survey of the Australian public, this study explores punitive attitudes in Australia and identifies its primary predictors. While previous research, has identified several predictors of punitive attitudes, studies have tended to focus on a small number of predictors and as such, are limited in scope. This paper examines 14 predictors of punitiveness (age, level of education, sex, geographic location, perception of crime, fear of crime, confidence in the criminal justice system, media consumption, crime causation, belief in redeemability, interpersonal trust, political ideology, racial essentialism, and minority threat) and considers the relationships between these predictors and their relative importance in predicting punitive attitudes in the Australian public.

Taken together, this approach provides a holistic assessment of what predicts individual punitiveness in Australia and develops a better understanding of public attitudes towards crime and punishment and thus informs how politicians should engage with public attitudes on matters of criminal justice. The article begins by reviewing the literature on punitive attitudes, with particular attention to the Australian context. Next, the paper considers evidence establishing 14 key predictors of punitiveness, before proceeding to the results of this study which examine the extent that these predict punitive attitudes. The article concludes by discussing the results of the study, contextualising these findings within the current literature, and considering how these results may inform engagement with public attitudes on matters of punishment.

What are punitive attitudes?

Over time, how punitive attitudes are measured has changed. Early research simply asked respondents if judicial sentences are ‘too harsh’, ‘not harsh enough’, or ‘just right’ (Gelb, 2008; Hough & Roberts, 2002), while others focussed on the goals of punishment (Doob & Roberts, 1988; Indermaur & Roberts, 2009) or support for specific forms of punishment (i.e. imprisonment or community sanctions) (Ramirez, 2013; Robbers, 2006; Unnever et al., 2005). Finally, some research has explored sentencing options by examining an individual’s support for punitive and non-punitive sentences for a particular crime (Applegate et al., 1996; Hough & Roberts, 2005; Mitchell & Roberts, 2012).

Punitive attitudes in Australia

Previous research has shown that Australians tend to hold punitive attitudes expressing a desire for harsher punishment of PCJI as well as little confidence in the court system (Roberts & Indermaur, 2007). Jones et al. (2008) found that 37% of respondents believed that sentences were too lenient (see also, Indermaur, 1987). In the same token, Mackenzie et al. (2012) explored punitive attitudes, acceptance of alternatives to imprisonment and confidence in the criminal justice system where results identified that a majority of respondents expressed high punitive attitudes, highlighting sentences were too lenient.

When comparing punitive attitudes in Australia to those of other countries, we are presented with some interesting findings. De Soto et al. (2022) compared punitive attitudes in a sample of Australians to Americans towards offences including drug sales, drug possession, robbery, rape, molestation, and burglary. The results revealed that Australians were more punitive regarding drug offences and were on par with respondents from the United States in all other offending areas. Likewise, Huang et al. (2012) explored the relationship between punitive attitudes and the concept of support for rehabilitation of offenders among a sample of Japanese (n = 1544) and Australian (n = 1967) respondents, with results identifying that both samples were supportive of rehabilitative approaches to dealing with crime; however, Japanese respondents were more supportive of viewing punishment as an active prevention measure for future crimes. When examined holistically, Australians tend to be more cynical about the usefulness of punishment as a successful means of deterrence and display high punitive attitudes towards PCJI, particularly when compared to Japanese and American samples (Ahlin et al., 2017; Huang et al., 2012).

Predictors of punitive attitudes

Some research has explored what factors influence punitive attitude formulation and found that a myriad of factors play a role. Demographic factors such as age, level of education, sex, and geographic location have been identified as predictors of punitive attitudes. However, the literature has often yielded inconsistent results. Some studies have found that punitive attitudes increase with age (Payne et al., 2004; Simmler et al., 2021) while others have identified no relationship at all (Hartnagel & Templeton, 2008; Kury & Ferdinand, 1999). Spiranovic et al. (2012) found that, on its own age was significantly correlated with punitive attitudes; however, when other predictors were considered in the model, age was non-significant. Similarly, the relationship between sex and punitive attitudes is also inconsistent. Most studies find that males are more likely to support harsher penalties than females (Applegate et al., 2002; Kury & Ferdinand, 1999; Roberts & Indermaur, 2007). However, studies have also found the opposite (Applegate et al., 2002; Dodd, 2018; Kelley & Braithwaite, 1990; Spiranovic et al., 2012). Moreover, regarding level of education, research has identified that people with higher educational attainment tend to be less punitive (Dodd, 2018; Fitzgerald et al., 2020; Spiranovic et al., 2012). Finally, those residing in rural areas hold more punitive attitudes than urban respondents (Mulrooney & Wise, 2019; Walker et al., 1987).

Crime salience variables (how ‘visible’ people believe crime is) is a key model applied in punitive attitudes research. This model includes perceptions of crime (if crime rates are rising), fear of crime (fear of victimisation) and confidence in the criminal justice system. Perceptions of crime have been identified as having a positive association between the perception of increasing crime rates and punitive attitudes, with these perceptions often being regularly overestimated (Gelb, 2006; Kornhauser, 2013; Roberts & Indermaur, 2007; Spiranovic et al., 2012). Similarly, research found a positive association between fear of crime and overall punitiveness (Brookman & Wiener, 2017; Gelb, 2011; Maruna & King, 2009; Spiranovic et al., 2012). Finally, confidence in the criminal justice system (typically measured through confidence in the police, courts, and justice system) has been linked to punitive attitudes, with people who report higher levels of distrust towards the police, courts, and criminal justice system in general, displaying higher levels of punitive attitudes (Brookman & Wiener, 2017; Kornhauser, 2013; Roberts & Indermaur, 2007).

The perceived drivers of crime (e.g. crime causation, belief in redeemability and interpersonal trust) are instrumental in punitive attitude development and have a direct impact on other predictors contributing to long-term punitive beliefs at the individual level. Crime causation can be broken into two distinct concepts: internal or external. Individuals who place the fault of crime solely on the individual (internal) are more likely to present with higher punitive attitudes in comparison to individuals who place the fault of offending on wider social circumstances such as poverty (external) (Carroll et al., 1987; Jeong, 2009; Templeton & Hartnagel, 2012). Likewise, belief in redeemability outlines the extent to which an individual believes that PCJI can desist offending and become positive contributors to society (Maruna & King, 2009; O'Sullivan et al., 2017). Individuals holding a strong belief in the potential redeemability of PCJI are less punitive and more supportive of rehabilitative practices (Burton et al., 2020; Maruna & King, 2009; O’Sullivan et al., 2017). Similarly, interpersonal trust is a crucial element of social cohesion, solidarity, and human capital. Tyler and Boeckmann (1997) found that individuals who do not trust others believe it is difficult for PCJI to rehabilitate and consequentially are more likely to be punitive and show a preference for incarceration in contrast those reporting elevated levels of interpersonal trust who show a preference for less punitive approaches to dealing with crime such as community-based punishments (see also, Barker, 2007; Lappi-Seppälä, 2010). To date, there has been no research investigating interpersonal trust and punitive attitudes in Australia but given that Australians tend to score highly on measures of interpersonal trust, one might surmise there would be an inverse relationship between interpersonal trust and punitive attitudes.

Moreover, media has been previously identified as contributing to the development of punitive attitudes. Regarding media consumption, Roberts and Indermaur (2009) found that 80% of Australians reported that television, radio, and newspapers were ‘fairly’ or ‘very’ important resources for informing the public about crime and crime-related issues, despite the misrepresentation of crimes amongst the media (Weatherburn & Indermaur, 2004). Spiranovic et al. (2012) explored several predictors of punitive attitudes, one of which was media consumption, with findings indicating that reliance on tabloid media for news and information was one of the top three predictors of punitive attitudes, along with perceptions of crime and education.

Equally, political ideology has been linked to punitive attitudes, with left-wing ideologies generally inclined towards social welfare and believing that inequities in society need to be reduced, whereas right-wing ideologies are characterised by an emphasis on authority, order, duty, tradition, reaction, and nationalism (Drakulich, 2015). Drakulich (2015) identified exactly this, with people ascribing to left-wing (liberal) political ideologies presenting with lesser punitive attitudes than their right-wing (conservative) counterparts. Similarly, Sivasubramaniam et al. (2020) explored this in jury decision-making in Australia, with results highlighting that conservative political ideology is associated with punitive attitudes which in this case translated to higher conviction rates.

Finally, racial beliefs have been identified as a predictor of punitive attitudes. Unnever and Cullen (2010) identified a ‘culturally universal’ relationship between racial prejudice and punitive attitudes. The current study will explore race and punitive attitudes using the principles of minority threat hypothesis (operationalised as support for multicultural principles and collective threat) and the lay theory of race (operationalised as racial essentialism). Firstly, the Minority Threat Hypothesis outlines that that the larger the relative size of a minority group, the more the dominant group is likely to perceive it as threatening and correspondingly exhibit harsher preferences for punishment as a means of maintaining the dominant group’s status and power which can manifest in an increased desire for harsh punishment of minorities (Blalock, 1967; Blumer, 1958; Weitzer & Tuch, 2005). People who perceive minorities as ‘threatening’ to their collective identity, express less tolerance and greater hostility towards members of the minority group and less support for multicultural principles (Berry & Kalin, 1995; Sumino, 2017). Secondly, the Lay Theory of Race highlights that the way individuals view, respond to, and interact with people of different racial backgrounds is dependent on their perceptions of how race shapes behaviour (No et al., 2008). Racial essentialists view race as a deep-seated, inalterable essence and that all members of this group share a specific set of characteristics specifically due to their membership in that group (Rothbart & Taylor, 1992). Individuals holding racially essentialist beliefs present with higher levels of prejudice towards certain racial groups as they believe that all members of a racial group will behave in the same way (Haslam et al., 2006; Morton et al., 2009) which theoretically will impact their punitive attitudes.

Thus far, some research has explored the predictors of punitive attitudes in the Australian context. Spiranovic et al. (2012) which identified that perceptions of rising crime rates, lower educational attainment, and reliance on tabloid/commercial media for news and information were that the three strongest predictors of punitive attitudes when accounting for the influence of age, sex, education, income, media usage, perceptions of crime, fear of crime, and personal experience with the criminal courts. Moreover, Brookman and Wiener (2017) explored the Crime-Distrust Model (perception of crime, fear of crime and confidence in the criminal justice system) and the Racial-Animus Model (community perceptions of First Nations Australians) as predictors of punitive attitudes. The results identified that both models could significantly predict high punitive attitudes.

The current study

This study builds on existing research by adopting a multifactorial approach exploring 14 theoretically informed and empirically established predictors of punitive attitudes in a single model, while previous research has tended to focus on one or two individual-level predictors. There are two research questions driving this study; (1) What are the punitive attitudes held by Australians? and (2) after controlling for the influence of other predictors, which predictors best predict punitive attitudes in Australia? This study has three hypotheses:

H₁: Respondents will present with punitive attitudes.

H₂: Age, sex, geographic location, high media consumption, high perceptions of rising crime rates, high fear of crime, conservative political ideology, high internal attribution of crime, high racial essentialism and high percieved collective threat will be positively correlated with punitive attitudes.

H₃: Low level of education, low belief in redeemability, low interpersonal trust, low confidence in the criminal justice system, low support for multicultural principles, and high external attribution of crime will be negatively correlated with punitive attitudes.

Method

Participants

This study surveyed 520 respondents and was representative of Australia’s population following a comparison with 2022 Australian Bureau of Statistics data. Respondent ages ranged from 19 to 84 (M = 47 years, SD = 17.08), with 51.9% (n = 270) of the sample identifying as male, 47.6% (n = 247) identifying as female and 0.6% (n = 3) identifying as non-binary. With regards to geographical location, 76.3% (n = 397) reported residing in an urban area and 22% (n = 123) in a rural area. See Table 1 for additional demographic details of respondents.

Table 1. Demographic information (N = 520).

Demographic variable	% (n)
Sex
Male	51.9 (270)
Female	47.6 (247)
Non-binary	0.6 (3)
Age
18–24	11.5 (60)
25–34	20.8 (108)
35–44	17.1 (89)
45–54	16.9 (88)
55–64	14.8 (77)
65–74	13.1 (68)
75+	5.8 (30)
Level of education
Primary school	1 (6)
Year 11 or below	7.5 (39)
Year 12 or equivalent	11.7 (61)
Certificate I/II/III/IV	18.7 (97)
Diploma or advanced diploma	11.5 (60)
Bachelor’s degree	30.2 (157)
Graduate diploma/certificate	5.2 (27)
Post graduate degree	13.8 (72)
Did not go to school	0.2 (1)
Residence by state
New South Wales	31.9 (166)
Northern Territory	0.8 (4)
Victoria	27.3 (142)
Queensland	19.8 (103)
South Australia	7.3 (38)
Western Australia	9 (47)
Tasmania	2.5 (13)
Australian Capital Territory	1.2 (6)

Measurements

Punitive attitudes

This index was developed to explore respondents’ feelings towards crime and punishment. This index consists of seven items on a 5-point Likert Scale (1 = ‘strongly disagree’, 5 = ‘strongly agree’). Respondents were asked to rate their level of agreement with each statement, an example being ‘The death penalty should be the punishment for murder’ and ‘People who break the law should be given stiffer sentences’. Higher scores on this scale indicated more punitive attitudes. The index was developed by Roberts et al. (2011) with Principal Axis Factoring supporting the unidimensional scale structure and demonstrating good internal consistency (Cronbach’s α = .84), which was higher in the current study (Cronbach’s α = .87).

Perception of crime

This index was developed to explore respondents’ beliefs surrounding crime. This index consists of three items on a 5-point Likert Scale (1 = ‘decreased a lot’ and 5 = ‘increased a lot’). Respondents are asked to rate their level of agreement with each statement, an example being ‘In your opinion, do you think that the level of overall crime in your State/Territory has increased, decreased, or remained the same over the last two years?’ Higher scores on this scale indicated a stronger belief that crime rates were rising. The index was developed by Spiranovic et al. (2012) and found to have acceptable internal reliability (Cronbach α = .71), which was higher within the current study (Cronbach’s α = .86).

Fear of crime

This index was developed to measure respondents’ concerns about becoming a victim of crime and consists of three items measured on a 4-point scale. One item is measured using (1 = never to 4 = many times) while the other two are measured using (1 = very safe to 4 = very unsafe). Respondents were asked to rate their level of agreement with each statement, an example being ‘How frequently have you worried about becoming a victim of a violent crime in your neighbourhood during the previous month?’ Higher scores indicated more fear of victimisation. The index was developed by Spiranovic et al. (2012) and found to have acceptable internal reliability (Cronbach α = .69), which was higher within the current study (Cronbach’s α = .78).

Confidence in the criminal justice system

This index was developed to explore respondents’ overall confidence in the criminal justice system. It consists of seven items arranged on a 5-point Likert Scale (1 = ‘Not confident at all’ and 5 = ‘Very Confident’), with higher scores reflecting higher confidence in the criminal justice system. Respondents are asked to rate their level of agreement with each statement, an example being ‘How confident are you that the criminal justice system is effective in bringing people to justice’. This scale was developed by Kautt and Tankebe (2011) and found to have acceptable internal reliability (Cronbach α = .80), which was higher within the current study (Cronbach’s α = .85).

Media consumption

This index explores the consumption of seven types of media including social media in general (e.g. Facebook, Instagram, Twitter, or Reddit), overall consumption of television, the Internet, social watching a local television news broadcast (e.g. 5pm news broadcasts), watching a national news broadcast (e.g. Sunrise or ABC), watching television crime shows (e.g. Criminal Minds, CSI, or Law & Order) and reading news stories on social media. Respondents are asked to rate the amount of time they spent engaging with each type of media in a 1-week period. Responses are arranged into six categories (none, 60 min or less, 61–120 min, 121–180 min, 181–240, and 241 min or more) with higher scores associated with more consumption in that category. The index was developed by Intravia et al. (2017), while the reliability was not reported in the original paper, in the current study this scale was found to have acceptable reliability (Cronbach’s α = .66).

Internal crime causation

This scale comprised of three items measuring the belief that crime is a biproduct of the individual. A sample item is ‘Most criminals commit crimes because they know they can get away with it’. Responses were arranged on a 5-point Likert Scale (1 = ‘strongly disagree’ and 5 = ‘strongly agree’), with higher scores reflecting more support for each type of crime attribution. This scale was developed by Templeton and Hartnagel (2012) and found to have acceptable internal reliability (Cronbach α = .66), which was higher in the current study (Cronbach α = .76).

External crime causation

This scale comprised of six items measuring the belief that crime is a biproduct of wider societal factors (Templeton & Hartnagel, 2012). A sample item is external attribution is ‘Most criminals were abused as children’. Responses were arranged on a 5-point Likert Scale (1 = ‘strongly disagree’ and 5 = ‘strongly agree’), with higher scores reflecting more support for each type of crime attribution. This scale was developed by Templeton and Hartnagel (2012) and found to have acceptable internal reliability (Cronbach α = .71), which was higher in the current study (Cronbach α = .78).

Belief in redeemability

This index was developed to measure respondents’ feelings that PCJI could ‘redeem’ themselves by desisting offending and becoming productive members of society. This index consists of 10 items arranged on a 5-point Likert Scale (1 = ‘strongly agree’ and 5 = ‘strongly disagree’). Respondents were asked to rate their level of agreement with each statement, an example being ‘After committing a crime, changing your life is more about personal effort than luck’. Items 8, 9 and 10 were reverse coded to ensure respondent consistency. Higher scores reflect a stronger belief in the redeemability of PCJI. This study utilised an updated version of the scale developed by Maruna and King (2009) detailed in O’Sullivan et al. (2017) and was found to have acceptable internal reliability (Cronbach α = .77), which was consistent with the current study (Cronbach’s α = .78).

Interpersonal trust

This index was developed to measure respondents’ level of trust in others. This index consists of five items arranged on a 5-point Likert Scale (1 = ‘strongly agree’ and 5 = ‘strongly disagree’), with higher scores on this scale reflecting lower level of interpersonal trust. Respondents were asked to rate their agreement with each statement, an example being ‘Most people tell a lie when they can benefit by doing so’. Item 4 was reverse coded. This scale was developed by Yamagishi (1986), while the reliability was not reported in the original paper, in the current study this scale was found to have low reliability (Cronbach’s α = .57).

Political ideology

Political ideology was measured in three ways: a self-report sliding scale of political ideology, a question where respondents self-reported which Australian Government Political Party they most closely identified with a 10-item index measuring conservative and liberal attitudes. Responses on the index were scored as +1 for a conservative and −1 for a liberal, with higher scores indicating a conservative political ideology and lower scores indicating a more liberal political ideology. Respondents were presented with two statements, one with a conservative ideology and one with a liberal ideology and asked to pick which statement resonated with them. An example of a conservative statement being ‘The Government is almost always wasteful and inefficient’. Whereas an example of a liberal statement is ‘The Government does a better job than people give it credit for’. This scale was developed by Dimock et al. (2014) and found to have acceptable internal reliability (Cronbach α = .72), which was lower within the current study (Cronbach’s α = .63).

Minority threat

This index was developed to measure minority threat and consists of two subscales, one measuring collective threat and the other measuring support for multicultural principles (eight items in total) arranged on a 5-point Likert Scale (1 = ‘strongly agree’ and 5 = ‘strongly disagree’). Four items measured attitudes towards multicultural principles, with an example being ‘Children born in Australia of parents who are not citizens should have the right to become Australian citizens’. Higher scores indicated less acceptance of multicultural principles. Four further items measured collective threat, with an example being ‘Immigrants increase crime rates’. Items 3 and 4 were reverse-coded, with higher scores indicating higher perceived collective threat. This scale was developed by Sumino (2017) and found to have borderline acceptable internal reliability on support for multicultural principles (Cronbach’s α = .55) and acceptable internal reliability for collective threat (Cronbach’s α = .83) while in the current study, both were borderline acceptable (support for multicultural principles, Cronbach’s α = .64; collective threat, Cronbach’s α = .65).

Racial essentialism

This index was developed to measure how respondents perceive race, either as a deep-seated, inalterable essence that dictates a person’s behaviour based on their membership to a racial group or as a socially constructed identity that is malleable and arbitrary. This index consists of eight items arranged on a 5-point Likert Scale (1 = ‘strongly disagree’ and 5 = ‘strongly agree’), with higher scores indicating higher racially essentialist beliefs. Respondents were asked to rate their level of agreement on each statement, an example being ‘To a large extent, a person’s race biologically determines his or her abilities and traits’. This scale was developed by No et al. (2008) and found to have acceptable internal reliability (Cronbach α = .83), which was lower in the current study (Cronbach’s α = .60).

Age

Age was self-reported by respondents and measured in years. During analysis, this factor was used as a continuous variable.

Sex

Sex was measured using the categories Male, Female, Non-binary.

Level of education

Respondents self-reported their highest level of education they had completed which was measured using categories drawn from the Australian census. For analysis, education was treated as a continuous variable (1 = Did not go to school, 2 = Primary School, 3 = Year 11 and below, 4 = Year 12 or equivalent, 5 = Certificate I/II/III/IV, 6 = Diploma or Advanced Diploma, 7 = Bachelor’s Degree, 8 = Graduate Diploma/Certificate and 9 = Post Graduate Degree).

Geographic location

Geographic location was measured using the Rural, Remote and Metropolitan Area (RRMA) Scale (Australian Institute of Health and Welfare [AIHW], 2004) which categorises Australia into seven areas: (1) a capital city, (2) an ‘other’ metropolitan urban centre (a non-capital city with a population equal to or greater than 100,000), (3) a large urban centre (population 25,000–99,999), (4) a small urban centre (population 10,000–24,999), (5) other rural area, (6) a remote Area (population between 5,000 and 9,999), and (7) Other remote area. During analysis, these options were coded as bivariate variables (rural = 1 and urban = 0). Prior to analysis, a bivariate variable was created where Urban = 0 and consisted of respondents residing in HMRA groups 1 and 2, whereas Rural = 1 and consisted of RRMA groups 3, 4 and 5. Groups 6 and 7 were considered remote and excluded from further analysis due to their small sample size (n = 16).

Funnel debrief

This consisted of two questions that were displayed at the end of the survey and acted as a control measure for social desirability bias. An example of an item is ‘While you were completing this survey, did you think of any research questions that we could/should look at in this study?’ Any potential issues were to be looked at on a case-by-case basis prior to analysis. Nil issues of social desirability were identified.

Procedure

The Online Research Unit (ORU) conducted the recruitment of respondents due to their ability to obtain a representative sample of the Australian population, using quotas as respondents complete the survey to control the sample across various demographic predictors including age, gender and geographic location and ensure these aligned with reported statistics by the Australian Bureau of Statistics. Prior to dissemination, the questionnaire was pilot tested to obtain an initial test of measure reliability and observe any unexpected response patterns, of which none were found. The ORU utilised offline measures (e.g. telephone, post, and print) to create a pool of potential respondents for the current study. Respondents were then recruited using random stratified sampling techniques.

Data collection took place between March 1 and 13, 2022 via a unique Qualtrics links distributed to each respondent via email. All respondents were informed that the survey was anonymous and that their responses would only be used for research purposes. Research and consent information was included for respondents prior to commencing the survey. The scales were presented to respondents in the following order: demographic information, media consumption, political ideology, punitive attitudes, perception of crime, fear of crime, confidence in the criminal justice, interpersonal trust, internal crime causation, external crime causation, belief in redeemability, collective threat, support for multicultural practices, racial essentialism and a funnel debrief. Throughout the survey, the language used was kept as simple as possible to minimise the potential of bias and ensure that respondents were clear on what each item was asking. The survey consisted of 122 questions and took 30–45 min to complete.

Analysis plan

The data analysis plan includes descriptive statistics, Pearson’s Correlation Analysis, and Hierarchical Multiple Regression Analyses (HMRA). Descriptive statistics will be used to describe the sample and respondents’ punitive attitudes. Pearson’s correlations will be used to explore the correlational relationship between individual predictors and punitive attitudes. Lastly, HMRA will be used to incorporate each predictor into one model and account for shared variance among the demographic and socio-cultural variables while predicting punitive attitudes, thus allowing for holistic assessment, and ranking of the predictors which most strongly predict punitive attitudes.

Results

All analyses were conducted using IBM SPSS V27 software. The descriptive analyses indicated that no issues present amongst variables. Normal distribution was found for all variables except for support for multicultural principles and racial essentialism which were above +2, indicating that the distribution is highly peaked. Moreover, some borderline acceptable Cronbach's alphas were observed for some scales but no further testing was required as none were under the acceptable level of .5. The lowest observed Cronbach’s alpha was .5 for the support for multicultural practices scale. Finally, the decision was made to not remove outliers prior to analysis as the removal would not be truly reflective of public attitudes (refer to Table 2).

Table 2. Descriptive information of measurement scales used (N = 520).

Scale	M (SD)	Skewness (SE)	Kurtosis (SE)
Punitive attitudes scale	3.60 (.91)	−.34 (.11)	−.26 (.21)
Media consumption	3.02 (.85)	.54 (.11)	.39 (.21)
Political ideology	−.27 (.452)	.56 (.10)	−.30 (.21)
Crime perception	3.50 (.84)	.08 (.11)	−.42 (.21)
Fear of crime	6.33 (2.57)	.82 (.11)	.59 (.26)
Confidence in the criminal justice system	2.80 (.79)	.12 (.11)	−.07 (.21)
Interpersonal trust	3.28 (.57)	.02 (.11)	.34 (.21)
Crime causation (internal)	3.72 (.79)	−.34 (.11)	−.02 (.21)
Crime causation (external)	3.17 (.71)	−.43 (.11)	.74 (.21)
Belief in redeemability	3.58 (.53)	−.17 (.11)	.16 (.21)
Collective threat	2.44 (.72)	−.26 (.11)	−.24 (.21)
Multicultural principles	9.24 (2.9)	.25 (.11)	−.25 (.21)
Racial essentialism	2.83 (.45)	−.34 (.11)	1.18 (.21)
Attitudes towards aboriginal Australians	3.11 (.50)	−.23 (.11)	1.14 (.21)
Age	46.90 (17.14)	.24 (.11)	−1.01 (.21)
Educational attainment	6.64 (1.9)	.041 (.11)	−.58 (.21)
Geographic location	.76 (.49)	−1.24 (.11)	−.45 (.21)

Attitudes towards crime and punishment

Our analyses indicated a relatively even distribution of punitive attitudes scores among respondents with 38.5% scoring 4 or 5 (out of 5) on the Punitive Attitudes Scale (M = 3.60, SD = 0.915). The average of 3.6 on the 1–5 scale indicates ‘somewhat agree’ with statements that indicate a harsh approach to dealing with PCJI in Australia. Further exploration indicated that respondents display the most punitive responses regarding the length of sentence given by the courts, with respondents demonstrating the highest agreement with the following statements ‘Courts are too soft on offenders’ (M = 3.9) and ‘People who break the law should be given stiffer sentences’ (M = 3.9). Furthermore, when examining the distribution of responses, very few respondents scored on the low end of the scale. In fact, 38.5% (n = 200) scored 3, 31% (n = 161) scored 4 and 7.5% (n = 39) scored 5, indicating punitive responses (see Figure 1 below for the spread of punitive attitudes among respondents).

Figure 1. Average Respondent Scores on the Punitive Attitudes Scale.

Assessing individual-level predictors of punitive attitudes

Pearson’s correlation analyses were undertaken to explore the relationship between each variable and punitive attitudes (refer to Table 3). There were significant positive relationships between punitive attitudes and age, geographic location (specifically rural areas), conservative, political ideology, perception of crime, fear of crime, interpersonal trust, internal attributions of crime, racial essentialism, and collective threat. Significant negative relationships were found between punitive attitudes and belief in redeemability, confidence in the criminal justice system and support for multicultural principles. Media consumption, external attribution of crime, level of education and sex did not correlate significantly with punitive attitudes and were therefore excluded from further analyses.

Table 3. Correlation matrix.

Variable	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17
1. Punitive Attitudes	-
2. Media Consumption	.015	-
3. Political Ideology	.279**	-.012	-
4. Crime Perception	.510**	.106*	.112*	-
5. Fear of Crime	.222**	.003	.163**	.334**	-
6. Confidence in the Criminal Justice System	.246**	.102*	.022	.312**	.066	-
7. Interpersonal Trust	.477**	-.037	.243**	3.75**	.201**	.239**	-
8. Crime Causation (Internal)	.586**	.002	.174**	.485**	.117*	.327**	.482**	-
9. Crime Causation (External)	-.031	.063	-.216**	-.058	-.044	-.062	.102*	.058	-
10. Belief in Redeemability	-.233**	-.001	-.234**	-.019	-.287**	.200**	-.219**	.018	-.093*	-
11. Collective Threat	.209**	.024	-.041	.176**	.011	.024	.232**	.221**	.225**	-.059	-
12. Support for Multicultural Principles	.243**	.044	.436**	.089*	.032	.118**	.148**	.131**	-.283**	-.182**	-.134**	-
13. Racial Essentialism	.253**	.021	.124**	.119**	.240**	-.071	.242**	.175**	.162**	-.290**	.295**	-.078	-
14. Age	.123**	.044	-.066	.198**	-.167**	.295**	-.002	.208**	-.162**	.252**	-.004	.139**	-.166**	-
15. Education	-.043	-.085	.005	-.018	-.055	-.073	-.009	-.031	.002	.010	-.032	-.015	-.023	.037	-
16. Geographic Location	-.179**	.050	-.009	.124**	-.027	-.143**	-.063	-.060	.041	.000	-.001	.019	-.002	-.141**	.063	-
17. Sex	.020	.040	-.126**	.182**	.260**	-.006	-.013	-.044	.058	-.066	-.026	-.061	.006	-.099*	.017	.013	-

Note. * p = .05; ** p = .001; Spearman’s Rho used for Sex and Geographic Location; Bivariate Coding (Rural = 1, Urban = 0); Pearson’s Correlations for all other variables.

Furthermore, a HMRA was performed to predict punitive attitudes and rank its strongest predictors, while controlling for the statistical influence of each predictor upon one another. Step 1 of the HMRA predicted punitive attitudes using the demographic variables that were significantly correlated with punitive attitudes which included age and geographic location. These predictors accounted for a significant 4.6% of the variance in punitive attitudes (R² = .046, F(2, 465) = 11.141, p < .001), representing a small effect size, with rurality exerting a small but significant effect on its own (β = −.195). In Step 2 all other predictors of punitive attitudes including political ideology, perception of crime, fear of crime, confidence in the criminal justice system, interpersonal trust, internal attribution of crime, belief in redeemability, racial essentialism, support for multicultural principles and collective threat were added to the model. Together these accounted for a significant 50.8% of the variance in punitive attitudes (ΔR² = .508, F(10, 455) = 41.152, p < .001), representing a large effect size. Further examination of the model revealed that age, political ideology, fear of crime, confidence in the criminal justice system and collective threat were not significant predictors of punitive attitudes despite the correlational analysis showing significant relationships with punitive attitudes. In combination, demographic and attitudinal predictors accounted for 52% of the variance in punitive attitudes (R² = .520, F(12, 455) = 45.044, p < .001). Further, internal attribution of crime, perceptions of rising crime rates, a lack of belief in redeemability, geographic location, a lack of interpersonal trust, lack of support for multicultural principles. (refer to Table 4).

Table 4. Hierarchical multiple regression analysis predicting punitive attitudes.

Predictor	β	t	sr^2	Sig.	R	R^2	F
Step 1				<.001	.214	.046	11.141
Age	.066	1.43	.065
Geographic location	−.195	−4.27	−.193
Step 2				<.001	.721	.520	41.152
Age	.018	.503	.016
Geographic location	−.129	−3.85	−.125
Political ideology	.237	5.67	.184
Perception of crime	−.002	−.04	−.001
Fear of crime	.021	.56	.018
Confidence in the criminal justice system	.113	2.84	.092
Interpersonal trust	.343	8.35	.271
Belief in redeemability	−.164	−4.37	−.142
Crime causation (internal)	.070	1.85	.060
Collective threat	−.105	−2.71	−.088
Support for multicultural principles	.052	1.47	.048
Racial essentialism	.089	2.32	.075

Note: β, standardised regression coefficient; bivariate coding for geographic location (Rural = 1, Urban = 0).

Discussion

The aim of this study was to examine the punitive attitudes held by Australians and to determine what individual-level factors predict such attitudes. Of note, 38.5% of respondents scored above the Punitive Attitudes scale mid-point of 3, revealing that just over one third of the sample endorsed punitive treatment of PCJI. This was consistent with previous research determining that Australians present with punitive preferences for punishment (Jones et al., 2008; Mackenzie et al., 2012; Roberts & Indermaur, 2007; Roberts et al., 2011). Further, the current study identified that sex, level of education, media consumption, and external attributions of crime demonstrated no significant relationship with punitive attitudes which was inconsistent with past research previously identifying significant relationships between these predictors and punitive attitudes. Positive correlations identified that political ideology, perceptions of crime, fear of crime, interpersonal trust, internal attribution of crime, collective threat, racial essentialism, age, and geographic location were associated with punitive attitudes. Finally, negative correlations identified that confidence in the criminal justice system, belief in redeemability, and support for multicultural principles were negatively correlated with punitive attitudes. The HMRA identified that several predictors (internal attribution of crime, perceptions of rising crime rates, belief in redeemability, geographic location, a lack of interpersonal trust, lack of support for multicultural principles) combined to explain 50.9% of variance in punitive attitudes while controlling for the influence of the other predictors.

Implications

When considering the strongest predictors and the broader context and implications of these findings, four different ‘pathways’ appear to be at work, location culture, ideological, misperception and minority threat.

First, is the ‘location culture’ pathway. The link between punitive attitudes and geographic location is not well established in the literature, with very few studies having this as a topic of focus. Our results revealed that people residing in rural locations tended to be more punitive (small effect size) than their urban counterparts. This finding is consistent with previous research conducted by Mulrooney and Wise (2019) in the Canadian context and Walker et al. (1987) in Australia. From a theoretical standpoint, this relationship could exist for several reasons such as rural communities holding more conservative ideologies (Roy et al., 2015), having lower education, or being in closer proximity to PCJI and victims and crime more generally, due to smaller residential populations. This finding suggests that more exploration is needed to understand the driving differences in punitive attitudes across rural and urban communities, particularly as approximately 28% of Australia’s total population resides in a rural or remote area (AIHW, 2022).

Second, is the ‘ideological pathway’ consisting of internal attributions of crime, belief in redeemability and interpersonal trust. These three predictors are key in understanding the drivers behind individual beliefs as to why people offend. Templeton and Hartnagel (2012) highlighted that respondents who place the fault of crime solely on the individual due to inherent negative characteristics are likely to present with higher levels of punitiveness (Burton et al., 2020; Maruna & King, 2009; O’Sullivan et al., 2017). Further, Tyler and Boeckmann (1997) found that individuals who do not trust others believe that it is difficult for PCJI to rehabilitate and as a result are more likely to endorse punitive sentiments such as prison (Lappi-Seppälä, 2010). Theoretically speaking, these predictors fit together because if a person holds the belief that crime is the sole responsibility of the PCJI and a by-product of ‘inner evil’, it is likely that the person would be less trusting of others and consequentially hold beliefs that once someone offends, they are unable to desist offending and redeem themselves.

Third, is the ‘misperception pathway’ consisting of perceptions of crime. Research suggests that the public widely holds the belief that crime rates are rising often despite factual evidence to the contrary (Indermaur & Roberts, 2005) and that people with high punitive attitudes often have little accurate knowledge of crime and the criminal justice system (Gelb, 2006; Roberts & Indermaur, 2007; Spiranovic et al., 2012). While further research is needed to establish a causal link between these variables, this finding raises the possibility that punitive sentiments may decrease if the public has accurate knowledge of crime and is better equipped to decipher the competing narrative that is presented by mainstream media.

Finally, the ‘minority threat pathway’ consisting of lack of support for multicultural principles. The relationship between racial prejudice and punitive attitudes has been longstanding, with an increased appetite for punitive responses having previously been theorised as a means of maintaining the dominant social group’s power (Brookman & Wiener, 2017; Weitzer & Tuch, 2005). This power struggle often sees out-group members labelled disorderly, drug-prone, criminal, and dangerous leading to increased support for more aggressive controls (Garland, 2001, p.148), that further reinforces and provides justification for punitive sentiments. In the case of the current study, a lack of support for multicultural principles was associated with high punitive attitudes. From this, one could infer that those with high punitive attitudes may view PCJI belonging to racial minorities as the ‘criminal other’ and as such have an unchangeable criminal essence that distinguishes them from the dominant social group thereby justifying punitive punishments for racial minorities.

When contemplating each of these pathways, they hold important for implications punitive attitudes in Australia, highlighting that punitive attitudes are complex and influenced by multiple factors at once, rather than just one factor. Moreover, these findings suggest that both circumstances of the individual (geographic location, perceptions of rising crime rates and a lack of support for multicultural practices) and perceptions of PCJI themselves (internal attributions of crime, lack of redeemability, interpersonal trust) combine to influence punitive attitudes. Working to understand these pathways is especially significant given that public attitudes often underpin the creation of criminal justice policies possibly resulting in the creation of both harmful and ineffective policies. Flowing on from this, how such attitudes can be used and exploited by cynical or motivated political actors seeking to employ penal populist tactics should be considered (Pratt, 2006; Roberts et al., 2002; Simon, 2007). However, a more positive interpretation of the current findings is that they are open to evidence, reason, and education. For example, regarding public views on sentencing, research has consistently pointed out that individuals tend to exhibit less punitive attitudes when they possess comprehensive information concerning the particulars of the case and the available alternatives, or when they are requested to provide insights based on their familiarity with the case in question (Doob, 2016). In other words, what people ‘know’ is a particularly important part of how they ‘feel’ and, optimistically, this can extend to issues such as perceptions around the salience of crime, the influence of race on crime, the redeemability of PCJI and attributions of crime, among others.

Overall, public attitudes must be engaged with. Not only should the public have a say in the operation of the criminal justice system in democratic societies but ignoring public attitudes risks ceding matters of crime and justice to political populists who may manipulate or exploit public misinformation and misperceptions for both political and ideological gain. The present study indicates some significant predictors of punitive attitudes which can be addressed directly. This includes predictors logically related to law and order, including perceptions of crime and reoffending, but also broader socio-cultural predictors such as interpersonal trust, collective threat, and racial essentialism, alongside cultural factors tied to residence in rural locations. Providing the public with information about the law and how it is applied is imperative, but as the findings in this study reveal; attention, and action in wider domains of social policy (such as improving social trust and relations between different cultural groups or perceptions of outgroups) may be valuable by serving to engage with punitive attitudes through a knock-on effect.

Limitations and future research

Although this research has yielded interesting findings, some limitations should be noted. First, some measures had borderline acceptable internal reliability, namely interpersonal trust. Given that this was were identified as significant strong punitive attitudes, the interpretation of these results should err on the side of caution as lower internal reliability can impact results by identifying significantly higher or lower results compared to what would be found by a more robust measurement (Nimon et al., 2012). Future research should consider further exploring the relationship between these predictors and punitive attitudes more in-depth, using more robust scales of measurement.

Second, whilst this study measured race-based predictors such as racial essentialism, collective threat and support for multicultural principles, this study does not account for attitudes towards Aboriginal and Torres Strait Islander Australians specifically. Such attitudes have previously been identified by Brookman and Wiener (2017) as an important predictor of punitive attitudes in the Australian context given the colonial history of Australia which may contribute to Australians cultural outlook on crime and punishment, thereby influencing baseline levels of punitivity. As such, future research should explore attitudes towards Aboriginal and Torres Strait Islander Australians and its relationship with punitive attitudes in more detail.

Conclusion

This study expanded on the knowledge of punitive attitudes and individual-level predictors of these attitudes in Australia by adopting a multivariable approach, with 77.2% of respondents scoring above the scale mid-point, and the sample as a whole averaging 3.6 on the 5-point scale. Additionally, our findings highlight that internal attribution of crime, perceptions of crime, lack of support for multicultural practices, belief in redeemability, interpersonal trust, and geographic location combine to predict high punitive attitudes. Together these findings suggest that punitive attitudes relate to four areas (1) location culture, (2) ideological perceptions of PCJI, (3) misinformation around crime, and (4) misperceptions around criminality and its relationship with race. This has important implications for future policymaking, particularly given that the creation of these is often informed by public opinion. The outcomes of this study underscore the significance of cognition and perceptions in shaping attitudes towards punishment, thereby highlighting the pragmatic implications of these predictors.

Open Scholarship

This article has earned the Center for Open Science badges for Open Data and Open Materials through Open Practices Disclosure. The data and materials are openly accessible at https://osf.io/c72b6/ and https://osf.io/c72b6/.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.