Exploring the links between punitiveness and political economy: Are criminal justice regimes a key dimension of varieties of capitalism in OECD countries?

Abstract

The varieties of capitalism thesis suggest high-income democracies cluster into distinct types, with complementarities between political, economic, and social institutions shaping long-run cross-national differences. Although featuring strongly in the comparative political economy and social policy literatures, criminal justice has received relatively little attention in this debate, but some have suggested penal policy variations map against varieties of capitalism ideal types. Using fuzzy-set qualitative comparative analysis – and data for 22 Organisation for Economic Co-operation and Development (OECD) countries over a 25-year period – we explore connections between cross-national differences in penal punitiveness and key institutional features identified in these debates. We find that the degree of punitiveness strongly overlaps with varieties of capitalism ideal types, arguing criminal justice regimes constitute a dimension of institutional complementarity in varieties of capitalism.

Introduction

After the welfare regimes debate reshaped the field of social policy in the 1990s, macro-level cross-national comparisons of a wide range of policy sectors proliferated (Hudson and Lunt, 2022), but aside from some notable exceptions (e.g., Cavadino and Dignan, 2006, 2013), criminal justice policy has been largely immune to this trend. Lacey (2008: 29) suggested “macro-level comparative research on criminal justice is relatively thin on the ground”; more recently, Newburn (2020: 103) noted “a growth of interest in comparative research within criminology” but still concluded “such work continues to remain the exception rather than the norm.”

We contribute to this growing body of work and connect it to wider comparative political economy debates, including welfare regime analysis, by exploring the role of multiple theorised political economy drivers of penality in criminal justice systems. We build on Lacey et al.’s (2018) suggestion that varieties of capitalism (VoC) and penal policy are related because the degree of punitiveness of the criminal justice system is – as with factors such as the strength of social protection or labour market protection – one of a number of institutional complementarities found in each VoC. We extend this framework through a systematic empirical investigation of their arguments using cross-national datasets commonly employed in macro-level comparative social policy analysis. We employ fuzzy-set qualitative comparative analysis (QCA), a method rarely deployed in comparative criminal justice (Parker, 2017), and data spanning a 25-year period for 22 high-income democracies to provide an innovative exploration of comparative data.

Comparative political economy and penal policy

Despite some notable exceptions (e.g., Sozzo, 2021; Lacey et al., 2018; Lappi-Seppälä, 2011), quantitatively informed macro-comparative analysis of penal policy remains relatively rare. Comparative criminology more commonly uses case-driven small-n comparisons (e.g., Crewe et al., 2022) or theoretical generalisations based on trends in a particular national context (e.g., Duxbury, 2021). Such studies offer rich insights into individual systems, but Nelken and Hamilton (2022) argue conclusions from these approaches may not generalise across a wider sample. Historically, limited availability of robust and meaningful cross-national data hampered macro-level comparisons of criminal justice policy, but advances in data and methods have now removed some of these barriers (Nelken, 2010; Tubex, 2013).

Despite facing similar limitations, scholars in other fields of applied policy analysis have persisted with broad quantitative macro-comparative analyses. They have often hedged the risks arising from limited datasets by rooting their analysis in established comparative political economy frameworks that aid contextualisation and interpretation of data. Nunn and Tepe (2022) note growing interest in political economy among criminologists; several high-profile theories of penal policy have highlighted aspects of political economy, including Cavadino and Dignan’s (2006) work linking criminal justice with welfare regimes and Wacquant’s (2009) account of punitiveness as a central feature of neoliberal statecraft.

Recent macro-comparative analyses of criminal justice policy (e.g., Cavadino and Dignan, 2006, 2012, 2013; Crawford, 2011; Díez-Ripollés, 2013; Lacey, 2008, 2010, 2012; McAra, 2011; Nelken, 2010; Tubex, 2013) have produced a great deal of fruitful discussion, with key studies underlining the value of a political economy approach. Within this body of work, Lacey et al. (2018) offer arguably the most sophisticated attempt to apply comparative political economy theory. They identify multiple “explanatory paradigms of penal policy” and conclude the “comparative political economy model, with its close attention to institutional particularities and their concrete shaping of incentives, provides the most promising framework in terms of bringing these insights together” (Lacey et al., 2018: 212). More specifically, they strongly suggest that different VoCs are likely to be key in explaining cross-national variations in punitiveness.

The VoC perspective occupies a prominent place in comparative political economy (e.g., Hall and Soskice’s (2001) foundational text). At its core is the distinction between liberal market economies (LMEs), where economic coordination takes place primarily through market mechanisms, and coordinated market economies (CMEs), where non-market modes of coordination are more prominent. Hall and Soskice (2001) argued that different modes of coordination in LMEs and CMEs¹ reflect different institutional arrangements that shape the strategic interactions of firms with other political and economic actors, particularly with respect to industrial relations, employee relations, vocational training and education, inter-firm relations, and governance. As such, VoC represents an actor-centred political economy approach in which the strategic interaction of firms within their institutional context is seen as key in understanding the development of distinctive models of capitalism.

Although VoC’s relevance to criminal justice may seem limited at first sight, it is expansive in orientation, considering how variations in economic, social, and political institutions shape different VoC. Crucially, it postulates that institutional complementarities exist in different spheres of political economy, including social policies that constitute differences between LMEs and CMEs. Indeed, in Hall and Soskice’s (2001) foundational text, Estevez-Abe et al. (2001) argued different welfare production regimes, featuring varying levels of employment, unemployment, and wage protections, map on to each VoC. This raises the question of whether and how variations in punitiveness might be understood as an additional dimension of institutional complementarity. Lacey et al. (2018) argue that CME systems are more likely to encourage less punitive and more rehabilitative penal approaches compared to LMEs. However, while they illustrate their theory with a detailed case study of the United States, they stop short of developing and operationalising a systematic cross-national comparison to test their argument. Instead, prompting our own investigation, they lay down a future research agenda, arguing “The institutional structures and modus operandi of political-economic systems – labo[u]r markets, political systems, welfare states – are obvious places to look” (Lacey et al., 2018: 205) and conclude “There remains much detailed empirical work to be done to test out the hypotheses emerging from this work” (Lacey et al., 2018: 212).

Operationalisation challenges: distilling theory

Lacey et al. (2018) provide numerous illustrations supporting their thesis that VoC is a missing link in debates about cross-national variations in punitiveness. However, undertaking the detailed empirical work they prompt is not a straightforward task because the broad themes they articulate are not easily operationalised in an empirically rooted cross-national analysis. Moreover, even if they had defined a specific set of institutional factors, data limitations present a challenge for anyone seeking cross-national measures suitable for empirical study. As a first step towards operationalising Lacey et al.’s (2018) thesis, we begin by drawing out five key themes from their paper, focusing in particular on those that also feature prominently in other key comparative political economy and punitiveness studies.

Income inequality

Lacey et al. (2018: 205) argue “countries with lower levels of inequality, more generous welfare states, and higher levels of unionisation and social trust – all factors likely to have been shaped by the political economy as broadly understood – show consistently lower rates of imprisonment,” wrapping a focus on inequality with other themes we explore below. Their focus on income inequality has sharpened further recently (Lacey and Soskice, 2021), and inequality is a common theme in other key studies too. For instance, Cavadino and Dignan’s (2006: 451) classic review of penal policy and political economy concluded “as a general rule, economic inequality is related to penal severity” and Karstedt (2021a, 2021b) presents wide-ranging comparative analyses of the relationship between inequality and punishment but points to complex non-linear relationships.

Welfare regimes

As noted above, Lacey et al. (2018: 205) suggest more generous welfare states are linked to lower rates of imprisonment. The link between welfare state arrangements and criminal justice is the core focus of widely cited studies by Cavadino and Dignan (2006, 2012, 2013); they mapped incarceration rates against different welfare regimes, noting “a significant association between these different types of political economy and penality” (2006: 446). Lappi-Seppälä includes indicators capturing welfare state effort in analyses of incarceration rates in Europe (Lappi-Seppälä, 2011) and the United States (Lappi-Seppälä, 2018).

Culture/societal values

Lacey et al. (2018) argued that countries with lower incarceration rates tend to have higher levels of social trust. Other key studies have made similar observations, sometimes with a wider focus than “social trust.” Utilising a wide range of indicators, particularly around different dimensions of trust, Karstedt (2015) mapped overlaps between “penal cultures” and “cultures of control” in Europe, while Lappi-Seppälä (2011, 2018) identified social trust as a key variable in explaining variations in punitiveness. More generally, Tonry’s (2007) classic study pointed to the interaction of broad cultural differences – for example, between “Francophonic” and “Anglo-Saxon” cultures – with political institutions as a potential driver of variations in penality.

Political institutions

Lacey et al. (2018) highlight the specific political institutions they see as influential in shaping punitiveness and place particular emphasis on Lijphart’s (1999) classic distinction between consensual and majoritarian democracies. Here, they stress the role of electoral systems in influencing the model of democracy, consensual democracies typically featuring proportional voting systems and majoritarian democracies first-past-the-post voting systems. Lappi-Seppälä’s (2011, 2018) cross-national explorations of punitiveness also draw on Lijphart’s (1999) classic distinction between consensual and majoritarian democracies. A common claim in the literature is that first-past-the-post systems are more likely to create adversarial “winner-takes-all” two-party competition in which the median voter plays a decisive electoral role, creating an institutional context in which the incentives for parties to compete on being the toughest on crime are amplified (Lacey et al., 2018).

Economic coordination

Lacey et al. (2018) argue that CMEs display lower levels of punitiveness than LMEs because of the institutional complementarities between modes of economic coordination, political institutions, and welfare regimes.

Operationalisation challenges: complex causality

The list above is a broad, but not exhaustive, summary of themes in key studies, hinting at a challenge any analysis of political economy and penal policy needs to confront: the expansive nature of relevant theoretical contributions. This challenge is exacerbated by the fact that studies commonly argue that causal pathways are complex, with outcomes contingent on interactions between multiple institutions. Quantitative comparative methods can soon reach their limits when exploring theoretical propositions that emphasise multi-directional and non-linear relationships, particularly when datasets have limited depth. Such challenges may explain the modest number of quantitative comparative macro-level studies of political economy and penal policy to date.

Key contributors have instead used predominantly theoretical and/or qualitative case study analysis to unpack connections between penal policy and political economy (e.g., Cavadino and Dignan, 2012; Lacey et al., 2018; Garland, 2020). This mirrors the wider institutionalist comparative political economy field, not least because the “thick description” that case studies allow offers “one way to try to pin down the complicated relationships between institutions, actors, and interests” (Immergut, 2006: 245). However, this brings the risk of a small number of cases dominating theoretical discussion. Arguably, this has occurred in debates about penal policy, where attention often focuses on a small number of “exceptional” states, such as the highly punitive United States (Garland, 2020; Lacey and Soskice, 2021) or the welfare-driven approach of Nordic countries (Lappi-Seppälä, 2012; Pratt, 2007; Crewe et al., 2022). Indeed, some of the best examples of broad-based quantitative comparative analysis have also focused on these cases (e.g., Lappi-Seppälä, 2012, 2018).

Karstedt (2015: 377) argues that comparative penology can utilise configurational approaches to address these challenges because they capture the influence of combinations of multiple conditions on outcomes, providing a “middle ground between macro-level models and “exceptional” cases.” QCA is the most prominent configurational approach (Ragin, 2008; Rihoux and Ragin, 2009; Oana et al., 2021) and has been widely deployed in comparative social policy analysis. However, very few papers in criminology have used QCA as a means of unpacking complex causality (key exceptions include: Stevens, 2016; Cruz, 2021), and it has not yet been employed in a study of political economy and punitiveness.

Methods

We utilise fuzzy-set QCA (fsQCA), a set-theoretic or “configurational,” approach that centres on the analysis of relations between “sets” capturing key concepts of interest and the varying membership of cases in each set (see Ragin, 2008; Oana et al., 2021). Rooted in Boolean algebra, fsQCA allows researchers to explore relationships between intersections and combinations of conditions captured in multiple sets and the outcome(s) of interest. Established thresholds are used to determine when specific combinations of conditions are deemed “necessary” or “sufficient” for that outcome. fsQCA facilitates “case-oriented” (Rihoux and Ragin, 2009) rather than variable-centred analysis of causal relationships, providing a systematic structured technique for facilitating dialogue between case-based empirical evidence and relevant theoretical perspectives. In so doing, the approach embraces causal complexity (Ragin, 2008; Oana et al., 2021), including three key principles: conjunctural causation (some conditions may only become influential on an outcome when combined with another condition); equifinality (there can be different routes to the same outcome); and asymmetric causation (where conditions lead to an outcome we cannot assume the opposite outcome occurs when we see a mirror image of those conditions).

The existing literature on penal policy and political economy has primarily focused on explaining trends in established high-income democracies, and the same is true for the VoC debate. Reflecting this, alongside pragmatic considerations around data availability, our sample comprises 22 cases that represent long-term members of the Organisation for Economic Co-operation and Development (OECD): Australia; Austria; Belgium; Canada; Denmark; Finland; France; Germany; Greece; Ireland; Italy; Japan; South Korea (Korea); Netherlands; New Zealand; Norway; Portugal; Spain; Sweden; Switzerland; United Kingdom; and the United States of America. As studies in comparative political economy are frequently concerned with capturing long-run national differences, we base our analysis on averages that pool data over 25 years spanning 1997–2021. This allows us to capture differences between nations that reflect broadly stable elements of their institutional landscape. The choice is also pragmatic again, as pre-2000s data are patchy; indeed, even for years after 2000, annual data are sporadic or unavailable for many indicators².

We base our analysis on each country’s membership of six sets: punitiveness as our outcome plus the five institutional dimensions – income inequality; the welfare state; social trust; economic coordination; and the electoral system – identified in our theoretical review. Raw data used to operationalise sets can be quantitative and/or qualitative, but here we draw primarily on the former (see Note 2).

High punitiveness (PR) is our main outcome measure, based on the incarceration rate indicator “prison population per 100,000 people,” the most common measure of punitiveness in the literature, with data drawn from the World Prison Brief.

High inequality (IN) captures debates around the impact of income inequality on punitiveness. It is based on the indicator “Gini coefficient pre-tax national income data” from the World Inequality Database. We use the pre-tax measure in order to minimise overlaps with our welfare state set (WS), allowing the IN set to capture underlying income inequalities and WS to capture the extent to which social policy intervenes to mitigate these income inequalities. Our theoretical expectations are that high inequality is associated with high punitiveness, whereas low inequality is associated with low punitiveness.

Weak social protection (WS) captures the debates around the impact of welfare state regimes on punitiveness. The set is based on a combination of two indicators encapsulating key aspects of social protection featured in welfare production regime ideal types. First, we capture income protection for unemployed workers by using OECD data on the “net replacement rate of cash benefits (including housing benefits and social assistance safety net) after 12 months of unemployment for a single person formerly employed at the average wage and with no children.” Second, we capture employment protection for workers (e.g., safeguards against dismal, redundancy and so on) by using version 1 of the OECD’s employment protection legislation index that captures protections against the dismissal of regular workers. Countries need to be weak in both components to enter the set. Our theoretical expectations are that weak social protection is associated with high punitiveness, whereas strong social protection is associated with low punitiveness.

High level of trust (TR) captures debates around the impact of culture/societal values on punitiveness. We focus on the specific societal value most commonly cited in the studies reviewed above – social trust – using an indicator from the World Values Survey/European Values Survey designed to capture generalised trust (the proportion agreeing that “most people can be trusted”). Our theoretical expectations are that high levels of social trust are associated with lower levels of punitiveness and vice versa.

High economic coordination (EC) captures debates around the impact of VoC on punitiveness. We use Witt and Jackson’s (2016) economic coordination index as a contemporary measure of Hall and Soskice’s (2001) VOC framework. Their index captures economic coordination in five key institutional domains – corporate governance, inter-firm relations, employment relations, firm hierarchy, and occupational training – summarising them in a single indicator. Our theoretical expectations are that CMEs will display lower levels of punitiveness and that LMEs will display higher levels of punitiveness.

Majoritarian voting system (MV) captures debates around the impact of electoral systems on punitiveness, particularly the suggestion that majoritarian voting systems/democracies are more likely to produce punitive criminal justice systems. We utilise the OECD Government at a Glance electoral system classification data to construct this set. Our theoretical assumptions are that countries with majoritarian voting systems will display higher levels of punitiveness and countries with proportional voting systems will display lower levels of punitiveness.

Note that although we have theoretical expectations about the influence of each set on the outcome, the configurational approach is rooted in acknowledging causal complexity. In the context of our study, this means our goal is not to isolate the impact of each set on punitiveness, but instead understand how it aligns with commonly theorised institutional complementarities found across different types of VoC.

In standard QCA, set memberships are binary (1 = in, 0 = out); however, fsQCA assigns membership in increments between the maximum (1.0, fully in) and minimum (0.0, fully out) anchor points, reflecting that membership of sets can be “fuzzy” rather than a binary. In fsQCA, the crossover threshold (0.5) represents the point of maximum ambiguity between the membership and non-membership of a set. The calibration of sets in fsQCA should be theoretically informed, using field- or case-specific knowledge to generate conceptually driven “anchor points” or “calibration thresholds” that guide the assignation of degrees of set membership to individual cases (Oana et al., 2021).

We used a method of direct calibration for the majority of indicators, assigning theoretically or empirically informed anchor points. Given most of our sets capture relative positioning (e.g., more or less punitive, more or less unequal), this entailed looking at the distribution of scores across nations – in particular inspection of median and mean values that might inform the location of the 0.5 crossover threshold – and reference to key cases commonly cited in the literature (e.g., those commonly referred to as being punitive or as exemplars of egalitarianism) or fuzzy set calibration points deployed in previous studies. The exceptions are for MV, where qualitative descriptors of voting systems were mapped against set scores capturing the degree to which a system is majoritarian, and EC, where Witt–Jackson index scores were utilised without alteration as they were already calibrated as an fsQCA set. Table 1 summarises the scores for each country and set (see Note 2).

Table 1. Raw set data and fuzzy set scores by country

Findings

Pathways to high punitiveness

Necessity

In simple set relations, necessary conditions are always present in cases with the outcome, that is, there should be no instance in which the outcome is present but the condition is absent. In fsQCA, this logic is adapted to account for the fact that cases have fuzzy membership of sets (rather than binary “in” or “out” membership) and fsQCA protocols permit small deviations from perfect set relations as they are usually “less than perfect” in reality (Oana et al., 2021). Applying a consistency value of at least 0.9 and coverage and relevance of necessity values at a minimum of 0.6 (as suggested by Oana et al., 2021; Mello, 2022), one condition emerges as necessary for high-punitiveness (PR): non-membership of the strong economic coordination set (~EC). In other words, being an LME is a necessary condition for membership of the high punitiveness set. Therefore, our QCA findings resonate with the suggestion that VoC and punitiveness are in some way related.

Sufficiency

In simple set relations, sufficient conditions are those that always lead to the outcome of interest. Once again, in the proposed fsQCA, this logic is adapted to account for fuzzy set membership and small deviations from the perfect set relations. Sufficiency analysis, which accommodates the possibilities of conjunctural causality and equifinality detailed above, entails production of “truth tables” detailing all logically possible combinations of sets and the outcome they display, to identify all possible “solutions” (combinations of conditions that sit in a sufficiency relationship with the outcome). These are then logically minimised to the fewest possible unique solutions. Different approaches to logical minimisation exist; here, we present “intermediate” overall solutions, meaning the logical minimisation is guided by theoretical expectations, and use an enhanced version of the standard analysis (see Note 2) to remove untenable assumptions prior to logical minimisation of the solution (Oana et al., 2021).

Overall solution

Table 2 summarises the two pathways to high punitiveness. Pathway 1a (IN*WS*MV* ~ EC) – featuring Australia, Canada, the United Kingdom, and the United States – which we label “punitive Anglophone LMEs,” combines high inequality (IN), weak social protection (WS), a majoritarian voting system (MV), and weak economic coordination (~EC, i.e., an LME). This maps strongly against theoretical expectations and key cases in the literature. The second pathway to high punitiveness, 1b (IN* ~ WS* ~ TR* ~ EC) – Korea, Portugal, and Spain, countries less often invoked in comparative discussions of punitiveness – “neglected punitive LMEs,” is a little more nuanced. Here, we find countries that do not fall into the weak social protection set (~WS) but have high underlying inequality (IN) and weak economic coordination (~EC, i.e., they are also LMEs), combined with low levels of social trust (~TR). These countries’ position as LMEs with high underlying market inequality does fit claims about interconnections between punitiveness and political economy, but the stronger social protections they offer goes against the grain of theoretical expectations.

Table 2. Enhanced intermediate solution for high punitiveness

Note: ☑ indicates the presence of a condition.

☒ indicated the absence of a condition.

– indicates the condition does not matter.

Based on standard QCA metrics, the overall solution – that is, the two pathways together (IN*WS*MV* ~ EC + IN* ~ WS* ~ TR* ~ EC) – is a strong one. It is highly consistent (a measure of how well subset relations in the solution fit the outcome: 0.869 here) and has a high level of coverage (a measure of how much of the outcome can is explained by the solution: 0.762 here). It is good practice to visually inspect a sufficiency plot (Figure 1) alongside these metrics, not least because this helps maintain a focus on cases. These plots should not be read as linear regressions; the different quadrants signify different logical relationships between the cases and the solution formula. Deviant cases “consistency in kind” – holding membership of the solution but not the outcome – would undermine claims that the solution is sufficient for the outcome. These would appear in the bottom right quadrant, but there are no cases here in Figure 1. Deviant cases “consistency in coverage” – cases that hold membership of the outcome but not the solution – are shown in the top left quadrant. Here we find one case not covered by the solution formula, New Zealand. These cases are viewed as less critical for the robustness of the overall solution because they do not contradict claims that the pathway is sufficient for the outcome. Nonetheless, they are cases of interest because they display the outcome under investigation, but a consistent pathway explaining their membership cannot be identified, perhaps pointing to conditions missing from the analysis.

Figure 1. Sufficiency plot (outcome = PR).

Robustness checks

QCA good practice increasingly stresses the value of applying computer-aided processes for assessing the robustness of the solution formula (Oana et al., 2021: 123). Such checks assess how far the solution would alter with multiple alternative, but plausible, analytic choices (e.g., different thresholds for set membership or alternative thresholds for assessing the consistency of a solution). Some suggested tests are of limited applicability for our study, but we applied Oana et al.’s (2021) procedures using alternative test solutions that varied the calibration of the punitiveness set and the consistency thresholds applied to sufficiency tests (see Note 2). The solution remains highly consistent (RF_cons 0.992) when subjected to these plausible changes in key analytic decisions but coverage less so (RF_cov 0.513). The solution is therefore graded with a robust cases rank (RCR) of three out of four on the basis that some cases it covers are “shaky”: Korea, Portugal, and Spain.

The robustness checks add useful information to the overall picture. The punitive Anglophone LMEs pathway appears highly robust and, as such, a strongly plausible explanation for high punitiveness in these countries, although the absence of New Zealand from this grouping requires some explanation. By contrast, the neglected punitive LMEs pathway highlights strongly punitive cases less often discussed in the literature, whose features fit less well with mainstream theory (e.g., around social protection). Given that all cases in this pathway are “shaky typical cases,” this points to some gaps in the dominant theoretical explanations of the institutional drivers of high punitiveness.

Case-oriented reflection

QCA encourages post-analysis case-based reflection, not least because this can strengthen the explanatory potential of QCA-generated solutions by offering insights into causal mechanisms associated with a given pathway, or highlight factors omitted from the QCA (see Note 2) (Oana et al., 2021: 181–182).

New Zealand shares many features with the punitive Anglophone LMEs, but is omitted from the pathway because it has proportional representation. Lacey (2012) argues that New Zealand’s 1996 transition to proportional representation occurred in the context of a collapse of public confidence in politics. She suggests that the introduction of referenda allowed single issues such as crime to shape political priorities, while the rise of “law and order” style politics among newly empowered small parties created an adversarial institutional profile quite distinct from that of the proportional representation nations of continental Europe. This suggests our “majoritarian voting” set could be adapted to capture the extent to which institutions foster an adversarial political culture.

The neglected punitive LMEs do not feature prominently in the comparative literature on punitiveness and may point to factors under-theorised in debate, particularly as none of the cases in the pathway have majoritarian voting systems. However, Korea, like New Zealand, may be misclassified here: while its parliament is elected via a semi-proportional multi-member system, it has a directly elected president chosen via first-past-the-post. There are good reasons to focus on its presidential branch given the strength of the president and the adversarial contestation of the post (Han, 2021). For Portugal and Spain, whose voting system classification is much less contentious, their relatively recent (1970s) transition from authoritarian to democratic politics may be an important contextual factor in unpacking long-term political culture. For example, Makin (2013) has argued that initial moves away from authoritarian-era high punitiveness in Spain were short-lived and that it soon returned as part of political contestation, with populist agendas becoming a key part of election-winning formulas for left and right.

While the neglected punitive LMEs run counter to theoretical expectations because they are not classed as having weak social support, qualifications must be made. Korea’s placement outside the weak social protection set might be seen as surprising given its low social spending (Yang, 2020), but our analysis reflects it combines a low net replacement rate with relatively strong employment protection. However, this indicator may not capture the nuances of labour precarity in Korea, where levels of self-employment and platform work are comparatively high (Lee, 2023). Classic arguments about the model of welfare in southern Europe (e.g., Ferrera, 1996), which highlight significant gaps in the social safety net and strong status differentials, may also leave some surprised at the classification of social protection in Portugal and Spain. Both (particularly Portugal) have relatively high replacement rates in our income protection indicator. However, alternative indicators of cash protection that exclude their relatively generous insurance-based payments – such as the OECD’s “adequacy of minimum income benefits” indicator – would have placed them towards the bottom of the sample. Overall, this group of cases map uneasily against many common indicators of welfare state effort, underlining the value of embedding case-oriented reflections into QCA analysis and pointing to future research agendas that unpack the case nuance further in order to understand the institutional dynamics fuelling high punitiveness in these neglected LME cases.

Pathways to low punitiveness

Necessary and sufficient conditions

The principle of asymmetrical causation means that routes to low punitiveness cannot be presumed to mirror those to high punitiveness. Underlining this, while our QCA identified LME membership as a necessary condition for high punitiveness, no conditions were identified as necessary for low punitiveness.

With respect to sufficient conditions, the enhanced intermediate approach resulted in model ambiguity when minimising the truth table, with two potential models identified. QCA good practice recommends that model ambiguity is always reported and a clear justification is provided for choosing one model over another (Oana et al., 2021). The two models heavily overlapped, and we selected the model with the strongest overall consistency and coverage, which also resonated more clearly with arguments in the theoretical and case-based literature, all contributing to our view that it was the more credible model (see Note 2).

Overall solution

Table 3 summarises the low punitiveness solution. Pathway 2a (~IN* ~ WS*EC), “low punitiveness egalitarian CMEs,” combines low inequality, stronger social protection, and high economic coordination (i.e., a CME) and covers a large number of cases: Austria, Belgium, Denmark, Finland, France, Italy, the Netherlands, Norway, Sweden, and Switzerland. This fits well with key theoretical arguments and covers key cases usually cited in discussions of low punitiveness.

Table 3. Enhanced intermediate solution for low punitiveness

Note: ☑ indicates the presence of a condition.

☒ indicated the absence of a condition.

– indicates the condition does not matter.

Pathway 2c (~WS* ~ MV*EC), “low punitiveness CMEs with proportional representation,” heavily overlaps with Pathway 2a. It includes high economic coordination and stronger social protection, with non-majoritarian voting rather than low inequality. It adds two further cases to the overall solution – Germany and Greece – while featuring all countries in Pathway 2a apart from France.

Finally, Pathway 2b (WS* ~ TR* ~ MV), which we label neglected low punitiveness cases with proportional representation, covers just two cases. These cases are less commonly cited as exemplars of welfarist approaches – Ireland and Japan – and the pathway fits uneasily with theoretical expectations, combining weak social protection, low levels of trust, and a non-majoritarian voting system. This pathway is more troubling for the dominant arguments in the literature.

The overall solution – the combination of the three pathways (~IN* ~ WS*EC + WS* ~ TR* ~ MV + ~WS* ~ MV*EC) – is a strong one, being highly consistent (0.905) with a high level of coverage (0.748). The sufficiency plot (Figure 2) shows there are no deviant cases consistency in kind (bottom right quadrant) and no deviant cases consistency in coverage (top left quadrant).

Figure 2. Sufficiency plot (outcome = ~PR).

We applied the same robustness checks for this solution, and it remained highly consistent (RF_cons 0.976), with strongly robust coverage (RF_cov 0.701), producing an RCR of one out of four, the “best” outcome in Oana et al.’s (2021) framework, indicating no cases are “shaky.” The absence of deviant or shaky cases points to a very robust and consistent explanation of the routes to low punitiveness, providing strong support for comparative political economy perspectives. However, the uneasy relationship between the neglected low punitiveness cases and proportional representation pathway to the core theoretical expectations requires explanation.

Case-oriented theoretical reflections

Pathway 2b to low punitiveness – neglected low punitiveness cases with proportional representation – comprises just two cases—Japan and Ireland—and provides a number of puzzles, including that it features weak social protection and low social trust as conditions, both contrary to theoretical expectations. “Irish exceptionalism” has been a growing focus of recent discussion (Marder and Hamilton, 2023; Brangan, 2020, 2021). Brangan (2021) draws attention to cultural factors, arguing a “pastoral penality” legacy in Ireland eschews a “correctional” approach, linked to the historic influence of Catholicism on societal values. Earlier work on Japan’s distinctively low crime rates often “invoke[d], in one way or another, the unique characteristics of Japanese culture and the resulting strong informal social control mechanisms” (Roberts and Lafree, 2004: 179). However, we should be cautious of explanations that rely on simplistic conceptions of culture: these early perspectives on crime in Japan have been critiqued as lacking empirical grounding (Huang et al., 2012) and for an ethnocentrist reliance on “othering” stereotypes of the “Far East” (Bui and Farrington, 2019: 23). Moreover, the Confucian values sometimes referenced in regards to Japan’s low punitiveness have also been offered to explain Korea’s high punitiveness (Lee, 2020), emphasising the complexities of interpreting cultural factors and the need for great care in unpacking societal values data (Hudson et al., 2014; Karstedt, 2012).

France also presents some significant analytic challenges. Its majoritarian voting system makes it an outlier in the low punitiveness egalitarian CME pathway, resulting in the QCA splitting the low punitiveness CMEs into two pathways. However, France’s voting system is difficult to classify because its second ballot system is neither a conventional first-past-the-post model nor true proportional representation (Dolez and Laurent, 2010). We score it “more in than out” of the majoritarian set (0.67), but a case can be made for placing it closer to the 0.5 crossover of maximum ambiguity. France is also something of a borderline case in our low punitiveness set, appearing just after our crossover point. Wacquant (2009) cited France as an exemplar of the punitive turn and some of the alternative test solutions computed for our robustness tests placed France in the high punitiveness set. When doing so, voting systems became more important in our solutions: low punitiveness pathways 2a and 2c collapse into a single pathway (CMEs with strong social protection and proportional representation), while for high punitiveness a third pathway is added that France joins (majoritarian voting and low trust). However, if it was judged that France is better classified as being just inside the high punitiveness set, this would place a CME in the high punitiveness set and mean being an LME would no longer be a necessary condition for high punitiveness. Hall and Soskice (2001) identified France as a potentially ambiguous CME case, suggesting it might be seen as a CME-LME hybrid, so it might better be placed at the point of maximum ambiguity for economic coordination too. All-in-all, France might be seen as a hybrid case that defies easy categorisation.

The position taken on key issues around France’s classification influences whether our QCA suggests egalitarianism matters or not. Moving France to the high punitiveness set would collapse low punitiveness pathways 2a and 2c into a single pathway, CMEs with strong social protection and proportional representation, in which inequality does not matter for punitiveness. Only two cases in this possible pathway – Germany and Greece – would be outside the low inequality set. Notably, both are classed as having strong social protection – that is, elements of egalitarianism – and it may be that our chosen inequality measure, which captures differences in income before tax and benefits, points to the decisive role that the welfare state can play a key role in ameliorating the impacts of higher market inequality (see Note 2).

Concluding discussion

The causal complexity unpacked by QCA’s configurational approach can make findings difficult to digest. To aid movement from case details to the “big picture” questions of political economy that theorists have grappled with, Table 4 compares the findings against theoretical expectations. With the exception of social trust, which is likely a weak proxy for culture, our QCA findings provide strong support for key dimensions of the political economy and punitiveness thesis. As Lacey et al. (2018) suggested, the clusters identified in the VoC thesis – and related institutional complementarities – map strongly against high and low punitiveness cases. With regard to culture/societal values, we might also note that this dimension features much less strongly in the VoC literature than in other strands of the comparative capitalisms literature such as the Business Systems Theory approach (see Wood et al., 2020 for an overview)

Table 4. Overview expectations and findings

However, our analysis also adds further nuance to this debate. While the larger pathways – which covered common exemplars of high and low punitiveness – matched theoretical expectations well, the smaller pathways captured neglected cases and underexplored or unexplored institutional configurations that sat less easily with theoretical expectations. These cases and configurations merit greater attention and may point to the need for some refinement of theoretical propositions in comparative political economy debates about cross-national variations in punitiveness. For example, Ireland may offer important insights into the wider dynamics that can steer LMEs towards low punitiveness; notably, it shares many features with the high punitiveness Anglophone LMEs, but differs in having well-established proportional representation.

Moreover, our fsQCA provides the first systematic empirical cross-national assessment of Lacey et al.’s (2018) suggestion that comparative political economy provides a strong framework for explaining cross-national variations in punitiveness and, caveats in Table 4 notwithstanding, the findings provide clear endorsement of their thesis. Our findings raise the question of whether penal policy forms one of the institutional complementarities found in each VOC; Lacey et al. (2018: 212) suggested as much when pointing to the strengths of a “comparative political economy model, with its close attention to institutional particularities and their concrete shaping of incentives.”

The VoC thesis originated as an actor-centred political economy approach that presents the institutional context in which firms and other organisations operate as an incentivising driver for their strategic interaction with their environment. In turn, this makes the institutional context a key explanatory factor in the development of different models of capitalism (Hall and Soskice, 2001). At first sight, the relevance of firm-level incentives for criminal justice policy seems limited. While Wood et al. (2014) have suggested a weakness of the comparative capitalisms literature is its failure to explore how overlapping state supported military-industrial and penal complexes drive some sectors of the USA’s LME, few theorists have drawn such links. Indeed, on the contrary, Cavadino and Dignan (2013: 291) expressed some scepticism about the link between VoC and penality, questioning whether decision-makers are really incentivised to connect distant factors such as economic coordination and penal punitiveness. How might we reconcile such prima facie instincts with the findings from our systematic cross-national analysis?

In articulating the role of mutually reinforcing institutional complementarities across distinct social, economic, and political spheres, the VoC thesis holds open the possibility of what we dub here “nested institutional complementarities,” whereby disparate spheres of political economy can be connected in an “indirect” manner. In particular, we note that widely cited comparative political economy studies of criminal justice have argued that welfare (state) regimes shape punitiveness (notably Cavadino and Dignan, 2013) and, at the same time, widely cited comparative political economy studies of social protection have argued that cross-national variations in welfare map against different VoC (notably Estevez-Abe, Iversen and Soskice’s (2001) suggestion that different modes of economic coordination are accompanied by complementary “welfare production regimes”). Connecting the two develops a theoretical rationale for viewing penal policy regimes as one of the institutional complementarities found in different VoCs, with welfare regimes acting as a bridge between penal policy and modes of economic coordination.

In offering this notion of “nested institutional complementarities” our analysis provides a further step on the road in demonstrating the value of applying the VoC framework to penal policy analysis, which as Gallo (2021: 268) argues, helps “systematise our accounts of penal variation, moving beyond description of difference and towards an understanding of stable and recurring dynamics that might be resilient to political changes” by placing institutional complementarities at its core. More detailed case-based analysis in future research is required to articulate these connections further, but existing detailed national case studies of penality already offer insights that might be built on; for example, Pratt’s (2007) study of the development of Scandinavia’s penal exceptionalism underscores that an expansive welfare state was central in embedding the culture and values that led to this approach.

Future work, including case-oriented analysis, might also adopt a broader view of punitiveness than captured by our outcome measure (see Note 2); incarceration rates offer a narrow view of punitiveness, and a wider perspective would include better understanding of how punishment, penality, and policing interact in different countries (Newburn and Jones, 2022), issues we were unable to explore due to lack of suitable comparative data. National incarceration rates also miss important variations within countries that feature prominently in high-profile case studies of punitiveness. Notably “The role of race in US crime policy […] is sometimes posited as the primary factor in American punitive exceptionalism” (Lappi-Seppälä, 2018: 246), and data limitations are a barrier to cross-national analysis here too, although Anderson’s (2023) recent work on ethnoracial disproportion in incarceration rates across 16 countries around the year 2016 shows what might be possible in future as researchers develop new datasets.

Other factors that might fruitfully be considered in future work include: variations within countries arising from federalism, devolution, and decentralisation (Brangan, 2020; McAra, 2011); the role differences in political discourse play in shaping the dynamics of criminal justice policymaking (Gallo, 2021); and a richer temporal analysis of long-run and short-run factors, including ongoing changes in our five institutional domains. The latter might capture dynamic changes within pathways and the movement of cases between them, addressing the risk that studies of political economy and punitiveness can produce overly deterministic accounts (Gallo, 2021). Such an approach might usefully combine with focuses on the under-explored cases and configurations of conditions – particularly those that challenge the theoretical assumptions – that we identify here.