Revisiting the old debate: citizens’ perceptions of meritocracy in public and private organizations

ABSTRACT

Public management scholars have long debated public and private sector differences. However, the generalizability of empirical results has been limited due to the shortage of cross-national studies. Using a data set including citizens in twenty-one European countries, we compare citizens’ perceptions of meritocracy (i.e. hard work determines success rather than luck or connections) in public and private organizations and test how such sector differences are linked to country-level macro factors. We find that sector differences in meritocracy exist in almost all European regions. However, sector differences are smaller in countries where NPM practices and meritocratic principles are established.

KEYWORDS:

• Public-private comparison
• sector differences
• Weberian bureaucracy
• meritocracy
• comparative public administration
• management

Introduction

Academics have long debated the significance of differences between the private and public sectors (Lyons, Duxbury, and Higgins 2006). Most scholarship suggests that the characteristics of public sector organizations, including external environments, organizational-level factors, and individual-level behaviours or attitudes, are profoundly different from those of private organizations (Allison 1986; Boyne 2002; Bozeman 1987; Perry and Rainey 1988; Rainey 1989; Williamson 1999). Whether examining the topic at the individual or organizational level, comparative literature on public and private organizations seems to confirm that major differences exist between these sectors (Rainey and Bozeman 2000).

However, several limitations characterize this past research. Although previous studies find significant sector differences, the generalizability of such findings is limited. The majority of studies rely on a single-country analysis of a developed country. With few exceptions, large-N cross-national studies are rare. Furthermore, past research has not examined which countries have larger (or smaller) sector differences than others. It is well known that the characteristics of bureaucracies vary considerably from country to country (Dahlström and Lapuente 2017; Hammerschmid et al. 2019; Pollitt and Bouckaert 2017; Kuhlmann et al. 2019; Painter and Peters 2010). For example, the recruitment and promotion of civil servants operates on the principle of meritocracy in some countries but not in others (Dahlström, Lapuente, and Teorell 2012; Lægreid and Wise 2015). Furthermore, New Public Management (NPM) is said to have blurred public-private distinctions (Bullock, Hansen, and Houston 2018; Diefenbach 2009). However, the prevalence of NPM practices also differs from country to country. Previous studies have made minimal efforts to incorporate differences in national contexts into their analysis (Bullock, Hansen, and Houston 2018).¹ Furthermore, the inclusion of country-level macro variables in the analysis of public management has been called for in recent studies of comparative public management literature (Beagles, Schnell, and Gerard 2019; Haque, Van Der Wal, and Van Den Berg 2021; Meier, Rutherford, and Avellaneda 2017; Roberts 2018). Rather than assuming that ‘all states are alike’ (Milward et al. 2016, 312), scholars have been called on to pay more careful attention to the contextual factors that influence public management’s variables of interest.² Responding to this call, we seek more generalizable explanations for sector differences by employing a cross-national and -regional data set.

Among several dimensions of sector difference, we focus on the degree of meritocracy in public and private organizations. Focusing on a single aspect of sector differences is a limitation of our study. However, this single focus is justifiable given the lack of cross-national data and large-N comparative studies in public administration (Charron, Dahlström, and Lapuente 2016; Jeannot, Van De Walle, and Hammerschmid 2018). Meritocracy can be measured by the extent to which success in public and private organizations is determined by hard work rather than luck/connections. Prior scholarship on mass belief has explored whether people attribute their success to meritocratic elements including hard work, initiative, and ability, or to non-meritocratic elements such as luck, connections, family wealth, or ethnicity (Newman, Johnston, and Lown 2015; Newman 2016; Reynolds and Xian 2014). In a different line of literature, public administration scholars have recognized the importance of meritocracy in recruitment and promotion in civil service systems (Miller 2000; Krause, Lewis, and Douglas 2006; Weber ([1922] 1978); Wilson 1887). Meritocracy in the recruitment and promotion of bureaucrats refers to a system in which ‘personnel are selected by merit-based civil service procedures or some other politically autonomous means’ (Krause, Lewis, and Douglas 2006, 177). Merit-based systems, which appreciate one’s hard work, capabilities, expertise, and achievement, are often contrasted with ‘at-will’ systems, in which personnel are ‘hired and fired at will by political actors’ based on non-merit factors, such as political or personal connections (Krause, Lewis, and Douglas 2006, 177; Charron, Dahlström, and Lapuente 2016).

In comparing these two sectors, we rely on a novel measure of meritocracy from a citizen perspective utilizing the 2013 European Quality of Government survey (EQI) (Charron, Dijkstra, and Lapuente 2015). The EQI survey was conducted through telephone surveys of over 85,000 citizens from 212 regions in twenty-four European countries. The survey was designed to create a general picture of how citizens perceive the quality of their regional government (Charron, Dijkstra, and Lapuente 2015). The external validity of citizens’ perception-based measures of bureaucracy has been demonstrated by Charron, Dahlström, and Lapuente (2016), who found that citizens’ perceptions correlated highly with alternative expert-based measures as well as other objective measures related to country-level meritocracy. Our study examines citizens’ perceptions of meritocracy; in particular, to what extent citizens believe that hard work rather than luck or connections is rewarded by success in public and private organizations.

This study has three research goals. Our first goal is to analyse whether and to what extent differences between the two sectors are found in the twenty-one European countries. Secondly, we explore if such sector differences exist even at the regional level in European countries. Comparative political scientists have been aware of the danger of relying on national-level data for comparative studies, as it creates a ‘whole-nation bias’ (Lijphart 1975; Rokkan 2009). This is especially true for European countries, which have notable within-country variations (Charron, Dijkstra, and Lapuente 2014). We explore within-country variations as to the degree of public-private distinctions in perceived meritocracy. Thirdly, we investigate which country-level factors are associated with larger sector differences in perceived degree of meritocracy. In particular, we focus on the effects of NPM reforms, bureaucratic closedness, and bureaucratic professionalism on perceived sector differences.

For most European countries and regions in our sample, we found that there are statistically significant sector differences in the degree of perceived meritocracy, with private organizations having a higher degree of perceived meritocracy than public organizations. Furthermore, we find that such sector differences are related to the degree of commitment to NPM reforms and meritocratic principles at the national level. In countries with a high degree of commitment to NPM practices or meritocratic principles, public and private organizations are more similar in their degree of perceived meritocracy. However, in countries with lower levels of such commitments, the perceived difference between the public and private sector is greater. This paper first explains the concept of meritocracy and our theoretical framework, along with our hypotheses. The second section explains the data and methods of this study, followed by a third section containing results and analysis. This paper ends with discussion, conclusions, and the limitations of this study.

Theory and hypotheses

The concept of meritocracy

The concept of meritocracy has been widely recognized as a positive system in Western society although some scholars argue that the concept has its origin in ancient Asia (Hobson 2004; Kim and Choi 2017; Poocharoen and Brillantes 2013; Zhang 2015). Meritocracy is understood broadly as a social system in which an individual’s capabilities, hard work, and merits or personal ability lead to wealth, jobs, and power; in non-meritocratic social systems, in contrast, wealth and power are distributed on the basis of an individual’s family, wealth, or social background (Kim and Choi 2017; Madeira et al. 2019; Poocharoen and Brillantes 2013; Zhang 2015). Michael Young, who popularized the term ‘meritocracy’ in 1958 (Castilla and Benard 2010), defined meritocracy as a competitive society that takes talent, merit, competency, motivation, and effort into consideration when distributing income, wealth, and social position (Castilla and Benard 2010; Poocharoen and Brillantes 2013; Young 2008). Today, the concept of meritocracy largely revolves around the idea of equality of opportunity; meritocracy proposes that everyone has an equal chance to advance based on their talents and efforts and regardless of their gender, race, class, or other non-merit-based factors (Castilla and Benard 2010). Meritocratic factors such as hard work, initiative, and ability are often contrasted with external factors such as luck, connections, family wealth, or ethnicity (Newman, Johnston, and Lown 2015; Newman 2016; Reynolds and Xian 2014).

Meritocracy has also been considered one of the core principles of modern civil service systems in many parts of the world (Aucoin 2012; Evans and Rauch 1999; Fukuyama 2013; Moon and Hwang 2013; Poocharoen and Brillantes 2013). Scholarly discussions have often centred on the topic of personnel decisions regarding civil servants, such as hiring, promotion, and firing based on merit rather than ‘at will by political actors’ (i.e. at-will system) (Charron et al. 2017; Krause, Lewis, and Douglas 2006, 770). Meritocratic systems, which select and reward candidates or employees based on their individual merit and hard work rather than for their political connections, has been considered one of the key elements of Weberian bureaucracy (Byrkjeflot, Du Gay, and Greve 2018; Fukuyama 2013; Gualmini 2008). The ideal of meritocracy in bureaucracy values competition, open selection, careful evaluation of qualities, and of having an established recruitment process rather than an arbitrary appointment of individuals to civil service positions (Poocharoen and Brillantes 2013). In practice, the choice of instruments for meritocratic systems varies across countries. However, meritocracy is often associated with merit-based recruitment/promotion via competitive examinations, interviews, and CV screening, as well as the performance-based assessments of civil servants (Poocharoen and Brillantes 2013; Sundell 2014).

Measuring meritocracy in the public sector

Although the idea of meritocracy has been widely discussed in the field of public administration and management, there are few systematic cross-national studies that incorporate administrative characteristics. This is partly due to the lack of comparable cross-national data for public bureaucracies (Fukuyama 2013; Sundell 2014). Despite the number of scholarly discussions on the characteristics of Weberian bureaucracies, efforts to quantify such characteristics have been limited (Dahlström, Lapuente, and Teorell 2012), and quantitatively comparing and measuring national bureaucracies is still a difficult task. However, the recent development of a unique cross-national data set provides tremendous opportunities for researchers to measure and compare national bureaucracies.

To the best of the authors’ knowledge and according to Charron, Dahlström, and Lapuente (2016), there exist only two data sets that measure characteristics of national bureaucracies, including personnel policies, and that cover a broad set of countries. Both data sets are based on expert assessments of national bureaucracies, where data are collected from country experts. The first is Peter Evans and James Rauch’s pioneering work mapping the structures of bureaucracies in thirty-five less-developed countries between 1970 and 1990 (Evans and Rauch 1999; Rauch and Evans 2000). Their work focuses on the degree to which bureaucratic agencies employ meritocratic recruitment and offer long-term careers for bureaucrats, which are scored together as the ‘Weberianness Scale.’ The scale consists of ten survey items, including the importance of formal examination in entry to the public sector and prospects for promotion for bureaucrats who entered the public sector via a formal exam. While the data set provides significant measurements of characteristics of bureaucracy, the data sample is limited to only developing or semi-industrialized countries. However, this data set has led to a number of subsequent empirical studies of Weberian bureaucracy (Cornell, Knutsen, and Teorell 2020; Nistotskaya and Cingolani 2016; Schuster, Meyer‐Sahling, and Mikkelsen 2020).

The second expert, assessment-based cross-national data set is the Quality of Government (QoG) Institute Expert Survey (Dahlström et al. 2015; Teorell, Dahlström, and Dahlberg 2011). The survey was designed based on Evans and Rauch’s work. The first Expert Survey was conducted in 2008–2012 (Teorell, Dahlström, and Dahlberg 2011) while Expert Survey II was carried out in 2014. This survey measured experts’ perceptions of national bureaucracies across 159 countries, focusing primarily on evaluating administrative characteristics such as recruitment, career systems, replacement, compensation, and policy-making and implementation (Suzuki and Demircioglu 2019). Several prior studies using data from this QoG Expert Survey focus on the degree of meritocracy in bureaucratic recruitment, measuring factors such as the importance of the skills of applicants and the role played by political influence and connections in hiring and termination decisions.

Although the reliability of the QoG Expert Survey data has been confirmed by its use and examination in many articles published in highly ranked journals, the major limitations of these data are its reliance on expert opinion (Charron, Dahlström, and Lapuente 2016). The subjectivity of expert-based assessments, measurement issues, and overrepresentation of expert perceptions in cross-national data have often been criticized (Charron, Dijkstra, and Lapuente 2015; Fukuyama 2013). Furthermore, this data set focuses solely on the public sector rather than on sector differences.

In contrast, the European Quality of Government Index (EQI) (Charron, Dijkstra, and Lapuente 2014, 2015; Charron, Lapuente, and Annoni 2019) is a novel data set on bureaucracies in European countries collected ‘from below’ (Charron, Dahlström, and Lapuente 2016, 502); that is, it collects data on the ‘[p]erceptions and experiences of the consumers of public services – ordinary citizens’ as opposed to expert opinions (Charron, Dijkstra, and Lapuente 2015, 317). The EQI data set contains a range of questions regarding characteristics of bureaucracies from a citizen perspective, including levels of corruption, impartiality in bureaucratic decision-making, quality of public services, and degree of meritocracy in public organizations as well as in private organizations. Rather than capturing objective measures of government performance, the EQI survey measures the de facto characteristics and qualities of national bureaucracies, asking citizens about their perceptions of government (Charron et al. 2016).

Charron, Dahlström, and Lapuente (2016) propose a novel measurement of meritocracy in the public sector from a citizen perspective, covering 212 regions in twenty-four European countries. This meritocratic measurement is constructed based on a survey item asking citizens to what extent they think that success in the public sector depends on hard work or on luck/connections. This measure indicates whether citizens believe that the public sector appreciates hard work as opposed to personal connections in recruitment and promotion decisions (502). Their study limits the sample to respondents who are public employees. However, their study also finds a high correlation between public sector employees’ and citizens’ assessments of meritocracy in the public sector (506), meaning that perceptions of the internal workings of the public sector do not differ significantly between respondents who are public employees and those who are not. Their study also finds major between-country and within-country variations in the degree of perceived meritocracy in the public sector. More importantly, the strong external validity of this perceived meritocracy measurement is demonstrated by the authors. Citizens’ perceived meritocracy has a high correlation with ‘alternative measures based on expert opinions, as well as variables from the literature that we would expect to correlate highly with a meritocratic bureaucracy’ (499). Furthermore, many other academic publications have confirmed the EQI data set’s reliability and validity.³ Our paper uses this same survey item to measure the degree of perceived meritocracy of public sector organizations and a similar item to measure the perceived meritocracy of private sector organizations.

Sector differences in meritocracy

Organization theorists have emphasized the commonalities among public, private, and non-profit organizations in terms of employee work motives, rewards, sizes, tasks, and environmental uncertainty (Thompson 1962). Public management scholars, in contrast, tend to focus on sector differences. Based on previous public management literature, we argue that there exist sector differences in the degree of meritocracy. Despite the large number of empirical studies on sector differences, as far as we know, previous studies have not directly compared meritocracy in the public and private sectors. Furthermore, despite an increasing number of public administration studies that utilize data of citizen perceptions of bureaucracies (Bullock, Stritch, and Rainey 2015; Magalhães 2016; Young and Hassan 2020), surveying organizational managers or employees is still the dominant method of data collection in studies of sector differences.

Although we did not find any previous studies that directly compared sector differences in terms of meritocracy, previous studies have identified a variety of factors that can lead to sectoral differences in meritocracy. In the business sector, appreciating employees’ hard work and performance in the form of rewards (such as salary increases and promotions) is prevalent (Burgess and Ratto 2003). The common assumption behind implementing meritocratic personnel systems (such as performance-based payment) in the business sector is that if individuals expect to receive a valued reward for high performance, they are more likely to exert effort (OECD 2005; Vroom 1964). In contrast, many surveys show that public sector employees were not rewarded for achieving a certain level of performance and that public employees perceived a weaker relationship between performance, pay, promotion, and disciplinary action than private employees did (US Merit Systems Protection Board 2012a, b). Therefore, the link between employee performance and rewards is relatively weak in the public sector. Furthermore, seniority-based promotion systems in the public sector remain strong in many countries, especially in those following the traditional Weberian model of bureaucracy (Dahlström and Lapuente 2017; Gualmini 2008). Thus, we test the following hypothesis:

H1: Private organizations have higher perceived meritocracy than public organizations.

National context and meritocracy in the public sector

Empirical studies on sector differences have been conducted predominantly through single-country studies. Therefore, little effort has been made to understand why some countries have large sector differences while others have fewer differences between sectors. There is a shortage of comparative studies in public management literature in general (See Beagles, Schnell, and Gerard (2019), Milward et al. (2016), Roberts (2018), and Suzuki and Hur (2020) for recent critiques). Our goal is to investigate which country-level factors are associated with the degree of sector difference within countries. Although one can think of several country-level factors which may condition the degree of sector difference, due to the limited number of factors that can be included in our study, we focus on three significant country-level characteristics of bureaucracy, including: (1) the prevalence of NPM practices in the public sector, (2) the ‘closedness’ of bureaucratic structures, and (3) the degree of professionalism in bureaucracy. We predict that in countries where NPM-style management is prevalent and meritocracy is strongly in force, there will be fewer differences between the public and private sectors. In contrast, sectoral differences will be greater in countries with traditional closed bureaucracies.

NPM-style management and sector differences

Over the latest four decades, the ideological rise of New Public Management (NPM) has profoundly changed the public sector in many countries (Christensen and Lægreid 2011; Hood 1991). Since the late 1970s (Gruening 2001), traditional, Weberian state bureaucracies have been replaced by results-oriented public organizations, NGOs, and for-profit private firms, or by a marketization of the state itself (Hood 1995; Lapuente and Van De Walle 2020). While NPM reforms originated in Anglo-Saxon countries in the late 1970s (Diefenbach 2009), they quickly spread throughout the world – in particular, to OECD countries, but also to non-OECD countries (Aoki 2015; Aoki and Rawat 2020; Hammerschmid et al. 2019; Lapuente and Van De Walle 2020; Ishihara and Toshihiko 2021).

Although there is no single model of NPM, one common NPM element is the adoption of business management ideas by public sector management (Lægreid 2017; Lapuente and Van De Walle 2020). NPM challenges the traditional idea of public administration, which is that ‘public-sector organizations are “insulated” from the private sector in terms of personnel, structure, and business methods’ (Dunleavy and Hood 1994; Lægreid 2017, 2). The distinctiveness of public sector management and its labour market is one of the characteristics of the classic model of Weberian bureaucracy (Lægreid and Wise 2015). However, the Weberian model has been challenged by reform efforts to make public organizations more like private organizations, most notably through the global rise of NPM (Bullock, Hansen, and Houston 2018; Lægreid and Wise 2015). Therefore, as Lapuente and Van De Walle (2020) put it, “[w]e are in the age of the ‘blurring of the sectors’ (Rainey and Chun 2005). Thus, as Bullock, Hansen, and Houston (2018) argue, it is significant to test if sector differences still hold in countries whose public sectors have experienced NPM-style reforms. However, systematic, cross-national scientific evidence and data collection for studies on NPM and its effect are still limited (Aoki 2019; Hammerschmid et al. 2019; Lapuente and Van De Walle 2020). To the best of the authors’ knowledge, there is no large-N cross-national study on the effects of NPM on public/private sector differences.

We argue that sector differences should be less distinctive in countries where NPM-based management styles have penetrated. As numerous previous studies have shown, in the classic model of Weberian bureaucracy, public organizations tend to constrain the use of extrinsic incentive tools such as pay and promotion for those who perform well. Furthermore, public managers and employees tend to have weaker perceived links between performance and extrinsic rewards including pay, promotion, and job security (Rainey and Chun 2005; US Merit Systems Protection Board 2012a, 2012b). Therefore, public sector managers must rely on intrinsic incentives mechanisms such as public service motivation and involvement in meaningful work (Boruvka and Perry 2020; Rainey and Chun 2005). Furthermore, although promotion tends to be based on merit to safeguard administrative decision-making from political interference in the classic Weberian model of bureaucracy, seniority-based salary and promotion practices remain strong in the pubic sector (Dahlström and Lapuente 2017; Meyer-Sahling, Lowe, and Van Stolk 2012).

In contrast, ‘[e]xtrinsic rewards … are preferred to intrinsic rewards’ in the NPM model of governance (Boruvka and Perry 2020, 573). NPM reforms in many countries have eroded traditional civil service systems characterized by long employee tenures and seniority-based promotions (Lægreid and Wise 2015). Performance-based payment rather than flat salaries and merit-based promotion rather than seniority-based promotion are often regarded as one of the significant NPM-style incentivization tools for public employees (Aoki and Rawat 2020; Dunleavy et al. 2006; Lapuente and Van De Walle 2020; Meyer-Sahling, Lowe, and Van Stolk 2012). In a similar vein, employment strategies aimed at linking merit to employees’ careers (such as performance-based payment systems) are often portrayed as supporting meritocracy (Scully 1997, 413; Castilla and Benard 2010, 545). Thus, we hypothesize that:

H2: Perceived meritocracy in public and private organizations is similar in countries where NPM-style management practices have penetrated.

Bureaucratic closedness and sector differences

Bureaucracies have often been evaluated in terms of whether their systems for recruitment and promotion are ‘open’ or ‘closed’ (Bekke and Meer 2000; Dahlström, Lapuente, and Teorell 2012; Lægreid and Wise 2015). ‘Open’ systems are more flexible in the recruitment and promotion of public officials, with a focus on selecting the best candidate for each position (i.e. position-based systems), and they operate under the regulation of general labour laws. In contrast, ‘closed’ systems feature formalized entries into public service, seniority systems, lifetime employment (i.e. career-based systems), and operate under special labour laws designed specifically to regulate public sector employees (Dahlström, Lapuente, and Teorell 2012; Lægreid and Wise 2015). The classic Weberian model of bureaucracy is more similar to a closed civil service system than to open systems (Dahlström and Lapuente 2017). In closed systems, promotions tend to follow seniority rules rather than merit or performance (Bekke and Meer 2000; Gualmini 2008; Lægreid and Wise 2015). Unlike in the open system, job mobility between public and private sectors in closed systems is limited. Therefore, public sector jobs in closed bureaucracies tend to have higher ‘career distinctiveness’ (Peters 2010) compared to the private sector, meaning that public sector careers have distinct recruitment and promotion patterns and require specific skills and values (Veit 2020).

We argue that sector differences in meritocracy should be high in countries where closed bureaucratic structures remain strong. Closed bureaucracies tend to be highly legalistic, with strictly regulated and formalized entry and promotion systems (Lapuente and Suzuki 2020). In highly formalized systems, little flexibility exists in how a decision is made or what outcomes are due in a given situation; procedures and rewards are dictated by rules (Schminke, Ambrose, and Cropanzano 2000, 96). Furthermore, Napoleonic countries (i.e. France, Belgium, Italy, Portugal, Greece, and Spain), which are ‘often considered to have closed institutions for public employees’ (Dahlström and Lapuente 2017, 78), tend to be more ‘laggard’ in NPM reforms (Orelli, Padovani, and Katsikas 2016, 778; Peters 2008). In fact, compared to countries with more open bureaucratic structures such as Sweden and Finland, public employees’ salaries are not linked to performance in countries with closed bureaucratic characteristics, such as France and Belgium. The above has led us to hypothesize that:

H3: Perceived meritocracy in public and private organizations is more different in countries where closed bureaucratic structures remain strong.

Bureaucratic professionalism and sector differences

The third character of the administrative structure at the national level that we focus on is bureaucratic professionalism. While bureaucratic closedness focuses on the extent to which the hiring and promotion of civil servants are open to the public and whether the rules for hiring and promotion are flexible, bureaucratic professionalism focuses on the extent to which the hiring and promotion of state civil servants is based on meritocracy and subject to political influence (Dahlström, Lapuente, and Teorell 2012; Dahlström and Lapuente 2017). It is well known that some countries are more susceptible to political intervention in the hiring and promotion of civil servants. For example, countries such as Ireland, New Zealand, Norway, Japan, Denmark, and Sweden are ranked high in terms of meritocracy in national bureaucracy while countries including Venezuela, Honduras, Nicaragua, and Bolivia are ranked low (Dahlström, Lapuente, and Teorell 2012). In countries where meritocracy is well established in the national-level bureaucracy, citizens’ perceptions of meritocracy in the public sector are naturally likely to be higher, which is likely to narrow the gap between perceptions of public and private organizations.

H4: Perceived meritocracy in public and private organizations is similar in countries where the meritocratic principle in the public sector is well-established.

Research methods

Data

This study provides a comparative analysis by combining two unique cross-national datasets. The first dataset comes from the 2013 European Quality of Government survey (Charron, Dijkstra, and Lapuente 2015), which includes 56,925 citizens in 174 EU regions. The second dataset comes from the QoG Expert Survey Dataset II (Dahlström et al. 2015), which reports characteristics of national bureaucratic structures based on the opinions of over 1,200 country experts. Our dependent variable, citizens’ perceptions of meritocracy, is from the EQI dataset. Our national-level independent variables are obtained from the QoG Expert Survey.

The EQI survey was conducted via telephone in the local majority language in each country/region in the spring of 2013. More than 85,000 respondents from 212 regions in twenty-four countries answered survey questions. The survey was designed to capture how citizens perceived the quality, impartiality, and corruption of public administration within their region; the survey intended to reflect ‘the perceptions and experiences of the consumers of public services – ordinary citizens’ (Charron, Dijkstra, and Lapuente 2015, 317). Questions focused on the implementation of laws and public policies rather than on law-making or policy decisions. 400 respondents were randomly selected per region, and the regional level is based on the European Union’s NUTS statistical regional level. Regional-level data is significant because even in centralized countries, there exist significant regional variations in characteristics of government, public service quality, and other socio-economic indicators. See Charron, Dijkstra, and Lapuente (2015) for the details of the EQI 2013 survey and Charron (2013) for its sensitivity analysis.

The EQI survey measures a de facto level of meritocracy in public and private organizations, not the de jure side of administrative characteristics or objective government performance data. The survey asks citizens about their perceptions of characteristics of public and private organizations (Charron, Dahlström, and Lapuente 2016). Compared to data collected through employee surveys, the EQI data allow researchers to analyse citizens’ general perceptions of the public and private sectors in their region. Analysing these experienced-based measures offers several advantages. It is widely held that a public sector organization operates differently from a private sector organization at the individual and organizational level; however, very little is known about citizens’ perceptions of the internal characteristics of the public sector and whether they differ from those of the private sector. Using this citizens’ perception-based survey, we can study whether potential sector differences can be associated with macro-level characteristics, thus contributing to the literature of comparative public management.

Methods

To test our three hypotheses about public-private differences in perceived meritocracy and the macro factors associated with these differences, we use t-tests of group means and multilevel regression. For the t-test, we conduct paired t-tests on the country-level means of perceived meritocracy in public and private organizations. We conduct this analysis to see if (1) there are any statistically significant differences in perceived meritocracy between the two sectors, (2) there is any country-level variation in these differences, and (3) which countries have larger differences between the two sectors. We conduct t-tests for our first hypothesis. Then, we conduct multilevel regression models to analyse how three macro factors are associated with sector differences (controlling for other factors). We utilize multilevel model regression as our dataset has a hierarchical structure, with citizens (level 1) nested in country-level factors (level 2); thus, multilevel analysis is the appropriate method (Jones 2008) rather than OLS regressions with country-fixed effects.

We limit our samples to those respondents who had direct contact with public organizations within the last twelve months.⁴ This is because respondents who have not interacted with public organizations may not have appropriate perceptions of public organizations. Our samples include citizens working in the public sector, citizens working in the private sector, and those who are unemployed. We include all respondents in these categories.

Variable operationalization

Dependent variable

Our dependent variable is sector difference/similarities in citizens’ perceptions of meritocracy in public and private organizations. We create the dependent variable by subtracting perceived meritocracy in the public sector from perceived meritocracy in the private sector. Thus, positive values mean the private sector has a higher perceived meritocracy than the public sector. Perceived meritocracy in the public sector is obtained from the survey question: ‘Which statement comes closer to your own views? (1–10 scale).’ Respondents are asked to select a number from 1 to 10, with 1 indicating ‘In the public sector most people can succeed if they are willing to work hard’ and 10 indicating ‘Hard work is no guarantee of success in the public sector for most people–it’s more a matter of luck and connections.’ Then, we reverse the scale. The reversed scale is an ordinal variable, with 1 as the lowest and 10 as the highest degree of perceived meritocracy. Perceived meritocracy in private organization is obtained from the same question. Respondents are asked to select a number from 1 to 10, with 1 indicating ‘In business most people can succeed if they are willing to work hard’ and 10 indicating ‘Hard work is no guarantee of success in business for most people, it’s more a matter of luck and connections.’

Independent variables

Our independent variables are degree of NPM reform penetration, closed bureaucratic structures, and bureaucratic professionalism. These variables are country-level variables and are created from the QoG Expert Survey (Dahlström et al. 2015). The NPM variable is an index variable created based on a principal component analysis of the following statements: (1) ‘The salaries of public sector employees are linked to appraisals of their performance’ and (2) ‘Public sector employees strive to be efficient.’ Cronbach’s alpha is 0.82. The variable ranges from −1.72 to 1.72. The higher the value, the more NPM-style management has penetrated into a country. Closed bureaucracy is also an index variable created based on the following statements: (1) ‘Public sector employees are hired via a formal examination system,’ (2) ‘Once one is recruited as a public sector employee, one remains a public sector employee for the rest of one’s career,’ and (3) ‘The terms of employment for public sector employees are regulated by special laws that do not apply to private sector employees.’ Cronbach’s alpha is 0.78. The variable ranges from −1.84 to 1.46. The higher the value, the more isolated or ‘closed’ public employees are from the practices of the private sector. Finally, professional bureaucracy is an index variable created based on the following statements: (1) ‘When recruiting public sector employees, the skills and merits of the applicants decide who gets the job,’ (2) ‘When recruiting public sector employees, the political connections of the applicants decide who gets the job,’ (3) ‘The top political leadership hires and fires senior public officials,’ and (4) ‘Senior public officials are recruited from within the ranks of the public sector.’ Cronbach’s alpha is 0.90. The variable ranges from −1.67 to 1.78. Higher values indicate a more professional bureaucracy. Figure 1 shows the degree of NPM-style management penetration, closedness, and professionalism of public administrations in our country samples.

Figure 1. NPM reform penetration, bureaucratic closedness, and professionalism by country.

Control variables

Our datasets include individual-, regional-, and country-level controls that may be associated with perceived meritocracy. Individual-level controls include basic demographic characteristics including respondents working sector, gender, age, educational level, and income level. Respondents’ working sector is included to control for a respondent’s knowledge of the internal workings of public or private organizations. It is natural that respondents working in the public sector may know one working sector better than the other. Furthermore, one may tend to perceive one’s own working sector more favourably. We control for such bias by controlling for respondents’ working sectors. We also control for individual-level corruption perception in the public sector, as perceptions of mismanagement in the public sector may also affect perceived sector differences.⁵ Regional-level variables include population size (ln), population density (ln), and GDP/capita (ln). Country-level variables include GDP/capita. As a robustness check, we control for national culture by using Hofstede’s cultural values (power distance, uncertainty avoidance, and individualism) to see if we obtained the same results. Individual-level controls are from the EQI 2013 dataset, and regional-level controls are from Charron et al. (2016). Country-level controls are collected from Teorell et al. (2017) and Hofstede, Hofstede, and Minkov (2010). See Supplementary Material for more details about these variables.

Table S1 in the Supplementary Material shows variable operationalization, and Tables S2 and S3 show the descriptive statistics and correlation matrix. There is no significant correlation among variables except country-level GDP/capita and regional-level GDP/capita. However, we ran an analysis with a model dropping the former to see if we obtained a similar result. Professional bureaucracy and NPM are highly correlated. However, they are not included in the single model. We conducted collinearity diagnostics using Variance Inflation Factors (VIF) based on models 1.3 and 2.3. The mean value of VIF is less than 2.29 and the maximum VIF value is 7.05. These results suggest that the models do not have serious multicollinearity issues.

Results

Sector comparison across countries and regions

We start with mean comparison graphs of perceived meritocracy for public and private organizations in all twenty-one countries in our sample (Figure 2). The left graph reports the country-level mean comparison of meritocracy in public and private organizations sorted by values of private sector meritocracy. Individual-level data are aggregated at the country level. Meritocracy in private organizations ranges from 4.57 to 6.87 with a mean value of 5.45 while public organizations’ meritocracy ranges from 3.07 to 5.51 with a mean value of 4.48. Austria, Germany, Ireland, and the UK record higher values of private sector meritocracy while Croatia, Poland, Hungary, and Romania record lower values. As for public sector values, Sweden, Denmark, Germany, Finland, and the UK rank high while Greece, Croatia, Bulgaria, and Slovakia rank low. As seen in the graph, private organization meritocracy is higher than public organization meritocracy in all countries. However, interestingly, the values of sector differences seem to differ significantly across countries. The right graph in Figure 2 reports the degree of sector differences measured by private-public meritocracy. Note that the larger values indicate that meritocracy in private organizations is much higher than in public organizations. Greece has the highest gap, followed by Austria, Ireland, Bulgaria, Slovakia, Italy, and Portugal. Sweden has the smallest gap, followed by France, Spain, Denmark, Hungary, and Poland. It is notable that some countries have larger sector differences while other countries have smaller differences, ranging from 0.04 to 1.78.

Figure 2. Mean comparison of citizens’ perceived meritocracy in public and private sectors and sector differences.

Notes: N = 53,498. Samples are based on model 3. Post-stratification survey weights are included.

Next, we test if differences in public and private sector meritocracy are statistically different by conducting t-tests by country. Table 1 reports country-level mean values of meritocracy in private and public organizations as well as differences between the two sectors. The results show that, with the exception of Sweden, the mean values of private sector meritocracy are statistically different from and higher than those of the public sector. These results provide empirical support for H1.

Table 1. T-test results: Mean comparison of perceived meritocracy in the public and private sectors by country.

	obs	Private mean	Public mean	Differences	St Err	t value	p value
Austria	2303	6.728	5.332	1.396	.065	21.65	p < .001
Belgium	897	5.446	4.544	.902	.108	8.35	p < .001
Bulgaria	2192	4.895	3.421	1.474	.076	19.45	p < .001
Croatia	684	4.484	3.513	.971	.116	8.4	p < .001
Czech Republic	2614	5.381	4.251	1.13	.066	17.2	p < .001
Denmark	1560	5.809	5.349	.46	.066	7	p < .001
Finland	679	6.279	5.692	.586	.108	5.45	p < .001
France	7075	5.415	5.107	.308	.036	8.5	p < .001
Germany	4120	6.411	5.419	.992	.044	22.3	p < .001
Greece	530	4.894	3.249	1.645	.139	11.8	p < .001
Hungary	918	4.923	4.42	.502	.098	5.15	p < .001
Ireland	584	6.608	4.936	1.671	.127	13.15	p < .001
Italy	5850	5.375	4.155	1.219	.038	31.55	p < .001
Netherlands	2823	5.742	5.024	.717	.066	11	p < .001
Poland	4853	4.814	4.16	.653	.044	15.05	p < .001
Portugal	2072	5.163	3.914	1.248	.065	19.25	p < .001
Romania	2602	4.954	4.015	.939	.069	13.7	p < .001
Slovak Republic	1386	5.106	3.637	1.468	.093	15.75	p < .001
Spain	5157	4.917	4.624	.293	.049	6.05	p < .001
Sweden	980	5.33	5.272	.058	.077	.75	p = .452
United Kingdom	3619	6.409	5.336	1.073	.052	20.75	p < .001

Samples are based on model 1.3.

To further demonstrate the robustness of the findings, we conduct two robustness checks for t-test results. For example, respondents who work in the public sector may respond more favourably as to the public sector’s meritocracy, and vice versa. The amount of knowledge about a particular sector might also vary depending on the sector in which respondents work. It is also possible that respondents’ perceptions of public sector corruption in their area may influence their perception of public sector meritocracy. Therefore, it is necessary to check whether sector differences can be found among respondents with different attributes. As a robustness check, we divide our sample into the following respondents’ sub-group to see if there are still sector differences in each sample; sample groups by respondent’s working sector: 1) respondents working in public organizations, 2) private organizations, or 3) unemployed; and sample groups by respondent’s corruption perception level in their area: 1) low corruption and 2) high corruption.

Table S5 in the Supplementary Material shows the t-test results for the three working sectors. In the public sector sample, sector differences are not statistically significant (p < .01) in seven countries (Belgium, Denmark, Finland, France, Hungary, Romania, Sweden). However, sector differences hold in the fourteen remaining countries. It is interesting that public organizations’ perceived meritocracy is higher than private organizations’ in Spain (p = .002) among the public servants’ sample. Meritocracy in private organizations is higher than in public organizations in thirteen countries. In the private employees’ sample, mean values of private sector meritocracy are statistically different from and higher than those of the public sector. In the unemployed sample, France, Hungary, and Sweden do not show statistically significant sector differences (p < .01). However, the remaining eighteen countries show significant sector differences with private sector values higher than those of the public sector.

Table S6 in the Supplementary Material shows the t-test results of two sample groups by corruption perception. The EQI survey includes items asking for the corruption perceptions of respondents in terms of the local public school system, public health care system, and police force. We created an average value variable for corruption perception in these three public service areas and then created a mean dichotomized dummy variable for corruption level (low/high). Then, we ran a separate t-test within each sample group. Except for Sweden, the mean values of private meritocracy are higher than those of public meritocracy (p < .01). The results of the above robustness checks show that even after limiting the samples by respondent demographics, the mean values of public and private sector meritocracy are still statistically different and are higher for the private sector in most countries.

Having shown statistically significant sector differences at the country level, we look at sector difference at the regional level. Table S4 in the Supplementary Material reports the t-test results of the mean comparison among 181 European regions. The results suggest that meritocracy in private organizations is higher than in public organizations, with statistical significance in 141 European regions out of 181 (p < .01). In all regions with statistically significant sector differences, private sector meritocracy is higher than that of the public sector. This result is quite surprising and demonstrates the robustness of our findings and the generalizability of perceived public-private distinctions even at the regional level. This provides strong empirical support for H1 at the regional level.

Contextual factors: closed public administration

Having demonstrated sector differences, we now examine which country-level contextual factors are associated with the degree of sector differences. We focus on three administrative characteristics: NPM-style management penetration, and closed and professional bureaucratic structures. First, we show bivariate graphs at the country level (Figure 3). X-axis shows administrative characteristics and Y-axis shows sector differences in meritocracy. NPM and professional bureaucracy seem to be negatively associated with sector differences while closed bureaucracy is positively related to sector differences.

Figure 3. Country means of sector differences in citizens’ perceived meritocracy and administrative characteristics.

Notes: N = 21. Samples are based on model 1.3. Post-stratification survey weights are included.

Having shown the bivariate relationship, we now report the results of the multilevel analysis (Table 2). The unit of analysis is individual. Our dependent variable is sector differences in perceived meritocracy. We employ a multilevel mixed-effects linear regression model. Since NPM and professional bureaucracy are highly correlated, we do not include both variables in the same model. Models 1.1–1.3 include NPM and closed bureaucracy, and models 2.1–2.3 include professional and closed bureaucracy. The first includes only our two independent variables. The second model adds individual-level controls. The third model adds regional-level controls.

Table 2. Multilevel Model Results for Impact of Country-Level Factors on Public-Private Distinctions.

	Model 1.1	Model 1.2	Model 1.3	Model 2.1	Model 2.2	Model 2.3
Independent variables
NPM penetration	−0.23***	−0.19**	−0.30**
	(0.07)	(0.07)	(0.09)
Closed bureaucracy	0.05	0.06	0.02	0.10	0.10	0.08
	(0.08)	(0.09)	(0.08)	(0.09)	(0.09)	(0.09)
Professional bureaucracy				−0.19	−0.15	−0.39***
				(0.11)	(0.11)	(0.11)
Individual-level controls
Respondent’s working sector^1 = private sector employee		0.71***	0.71***		0.71***	0.71***
		(0.08)	(0.08)		(0.08)	(0.08)
=Without professional employment		0.34***	0.33***		0.34***	0.33***
		(0.07)	(0.07)		(0.07)	(0.07)
Gender^2 = female		−0.16*	−0.14*		−0.16*	−0.14*
		(0.07)	(0.07)		(0.07)	(0.07)
Age		−0.00	−0.00		−0.00	−0.00
		(0.00)	(0.00)		(0.00)	(0.00)
Educational level^3 = medium education		−0.01	−0.01		−0.01	−0.01
		(0.08)	(0.09)		(0.08)	(0.09)
Educational level = lower education		−0.09	−0.08		−0.09	−0.08
		(0.07)	(0.07)		(0.07)	(0.07)
Income level^4 = Medium income		−0.10	−0.10		−0.10	−0.10
		(0.07)	(0.07)		(0.07)	(0.07)
Income level = Low income		−0.34***	−0.33***		−0.34***	−0.33***
		(0.07)	(0.07)		(0.07)	(0.07)
Individual level corruption perception		0.03*	0.03*		0.03*	0.03*
		(0.01)	(0.01)		(0.01)	(0.01)
Regional-level controls
Population density (ln)			0.02			0.02
			(0.03)			(0.03)
Regional level GDP/capita (ln)			0.03			0.02
			(0.11)			(0.11)
Population size (ln)			0.06			0.06
			(0.04)			(0.04)
Country-level control
Country-level GDP/capita			0.00			0.00**
			(0.00)			(0.00)
Constant	0.93***	0.75***	−0.88	0.93***	0.75***	−1.25
	(0.09)	(0.19)	(1.10)	(0.09)	(0.19)	(1.04)
Between country variance	0.14***	0.15***	0.15***	0.15***	0.16***	0.14***
	(0.04)	(0.04)	(0.05)	(0.04)	(0.04)	(0.05)
Within country between-individual variance	9.95***	9.83***	9.80***	9.95***	9.83***	9.80***
	(0.37)	(0.37)	(0.38)	(0.37)	(0.37)	(0.38)
Intraclass correlation coefficient (ICC)	0.01***	0.02***	0.02***	0.02***	0.02***	0.01***
	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Observations	66,302	55,752	53,498	66,302	55,752	53,498
Number of countries	21	21	21	21	21	21

Robust standard errors in parentheses. Standard errors are clustered at the country level. Post-stratification survey weights are included. *** p < .001, ** p < .01, * p < .05. 1: ref. = public sector employee, 2. ref. = male, 3. ref. = higher education, 4. ref. = higher income.

NPM is negatively associated with sector differences in all three models (p < .001 or p < .01). This means that the more prevalent NPM-style public management is in a country, the fewer differences there will be between public and private institutions. This means that the more entrenched NPM-style management is at the national level, the fewer differences in perceived meritocracy there are between the two sectors. This confirms H2. Closed bureaucracies are positively associated with sector differences, but the effects are not statistically significant in all models. Thus, H3 fails to receive empirical support. Interestingly, professional bureaucracy is negatively related to sector differences, but the association is not statistically significant when we include only country- and individual-level variables. This provides an empirical support for H4.⁶ Other interesting results include the following. Compared to those working in the public sector, private sector employees tend to have larger sector gaps (controlling for other factors). The same can be said of those with unemployed status. Likewise, respondents who perceive more corruption in public organizations in their area tend to perceive that public and private organizations are more different than those who live in less corrupt areas (p < . 05). This makes sense because if respondents think that corruption is prevalent in the public sector in their area, they tend to perceive that meritocracy is not working well in the public sector, which widens sector differences. On the other hand, compared to respondents who belong to the high-income class, low-income class respondents tend to perceive smaller sector differences in meritocracy (p < .001). The reasons for this result are worth considering. Perhaps low-income respondents perceive meritocracy differently from high-income respondents. Alternatively, low-income respondents may access different types of public organizations than high-income respondents.

Robustness checks

We conducted three robustness checks to see if we obtained similar results in terms of the direction of coefficients of our independent variables. First, we ran an analysis without country-level the GDP/capita variable as it is highly correlated to regional-level GDP/capita. Although VIF shows that multicollinearity is not a problem, we double-checked to see if we obtained a similar result. Second, as demonstrated by Bullock, Stritch, and Rainey (2015), national culture may influence work attitudes and motives, which may lead to sector differences. We used Hofstede’s cultural values (power distance, uncertainty avoidance, and individualism) to see if we obtained the same results when controlling for national culture. Table S7 in the Supplementary Material reports the results of these robustness check models. Although professional bureaucracy lost statistical significance when we dropped the national-level GDP, we did not find any significant changes to the results, confirming the robustness of our findings.

Discussion and conclusions

Public management scholars have debated the distinctiveness of internal and external characteristics of public organizations. Despite the breadth of past research, most previous studies examine sector differences only in the context of a single country, and there are few cross-national studies on sector comparison (Bullock, Stritch, and Rainey 2015). Our cross-regional and -national study responds to this research gap. We showed that sector differences in perceived meritocracy exist in most of the EU regions and countries regardless of variations in contextual factors. Thus, our study shows the generalizability of findings from previous studies on sector differences to our sample areas (Baarspul and Wilderom 2011; Rainey 1989; Williamson 1999). We also found that in most regions and countries, perceived meritocracy in private organizations is higher than in public organizations.

However, our empirical analysis shows a large variation in sector differences across regions and countries. We find that in countries where NPM-style management is well established and meritocracy is relatively established in the personnel policies of the national bureaucracy, sector differences were found to be low. Our study also finds greater within-country variation at the regional level. Exploring reasons for within-country variation is outside the scope of this research. However, one explanation could be that such variations are linked to variations in both government systems (such as centralized or decentralized political systems) and institutional factors (such as transparency in decision-making processes) in public and private organizations across regions. The results of our multilevel model analysis show that bureaucratic closedness is not associated with sector differences in citizens’ perceived meritocracy. While the classic literature of Weberian bureaucracies stresses the distinctive characteristics of the public sector, we do not find any significant association between Weberian bureaucracy and sector-differences in perceived meritocracy. Our results also show that individual citizens’ attributes – such as working sector, employment status, gender, education, and income – are significant factors affecting perceived sector differences. This should be explored in future research.

This study offers two substantive insights for research on sector differences and public administration. First, the results of this study suggest the effectiveness of using large-N cross-national data in public management research to generalize the empirical and theoretical conclusions of previous studies. Many previous studies have focused on a single country (particularly the US or European country), and public management theories have been developed based on the results of these studies. It has been found that there are differences between the two sectors, but it has not been known whether these differences exist in all countries. For this reason, it is meaningful to compare countries simultaneously using the same data, as in this study. Second, the results of this study suggest the necessity of examining how the relevance and significance of public management theory differ depending on macro-level institutional factors. Sector differences were found to exist in various countries; however, there are differences within those differences. It is necessary for future public management research to examine the external validity of public management theory and to study the macro factors that affect external validity.

There are limitations associated with our study. First, our study examines only meritocracy as a dimension of sector differences due to the lack of cross-national data. We may see a different picture of sector comparison if we look at other internal characteristics of organization. Second, we use citizens’ perceptions of public and private organizations. The survey items we utilized do not provide clear definitions of public and private organizations. Therefore, the types of organizations that respondents have in mind may depend on the individual’s background. Furthermore, there is a possibility that citizens may not have accurate knowledge of the internal rules of these organizations or that their knowledge may be biased. We have employed research strategies to reduce such bias. First, we limit our samples to those who recently had direct contact with the public sector. Second, we checked the robustness of our t-test results by splitting our samples into sub-groups based on the factors that may influence citizens’ perceptions, such as citizens’ working sector and corruption perception. Third, we include citizens’ socioeconomic status as controls in our multilevel models. We believe that the above research strategies helped to alleviate bias. However, we cannot rule out all potential biases due to our data limitations. Finally, our study relies on cross-sectional data, not time series data. Therefore, we cannot evaluate if sector differences/similarities change over time. Our study reports a snapshot of one point in time, and we do not make causal claims.

Despite these limitations, this study shows how sector differences vary across regions and countries in Europe, and that such differences can be associated with the characteristics of public administrations. We need more comparative research on sector differences that consider various dimensions of organizational characteristics as well as national contexts. In addition, data collection efforts for longitudinal data on sector differences/similarities should be made. Finally, future research could consider experimental survey treatments. Using an experimental survey design approach would provide a basis for causal inferences in understanding how internal and external organizational characteristics and national contexts shape citizens’ perceptions of meritocracy. Scholars are still in the early stages of data collection efforts for comparative bureaucratic research (Van De Walle et al. 2016). As more cross-national data become available, future research should undertake the task of furthering the comparative study of public-private distinctions.

Disclosure statement

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

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/14719037.2021.1895545.

Additional information

Notes on contributors

Kohei Suzuki

Kohei Suzuki is an assistant professor at the Institute of Public Administration, Leiden University. He obtained his Ph.D. in Public Policy from the School of Public and Environmental Affairs at Indiana University, Bloomington. He mainly studies administrative reform and bureaucratic structures with a focus on cross-national settings and Japanese local municipalities. Please see his website for his recent work.

Hyunkang Hur

Hyunkang Hur is an assistant professor at the Department of Public Administration and Health Management, Indiana University, Kokomo. He obtained his Ph.D. in Public Affairs from the School of Public and Environmental Affairs at Indiana University, Bloomington. His research focuses on human resource management and organizational commitment of civil servants.

Notes

1. See Bullock, Hansen, and Houston (2018), Bullock, Stritch, and Rainey (2015), Houston (2011), and Van De Walle, Steijn, and Jilke (2015).

2. We appreciate the reviewers’ comments on the importance of contextual factors in public management literature.

3. For example, Bauhr and Charron (2020), Bubbico, Elkink, and Okolikj (2017), Charron, Harring, and Lapuente (2021), Filippetti and Cerulli (2018), Nistotskaya, Charron, and Lapuente (2015), Rodríguez-Pose and Garcilazo (2015), Suzuki and Demircioglu (2020).

4. The survey asks if respondents or anyone in their immediate family used public school systems or public health care services or contacted security or police forces. Our analysis uses samples who answered ‘yes’ for any of these questions.

5. We thank an anonymous reviewer for pointing out this issue.

6. ICC is very low. The ICC is ‘a measure of the relative distribution of between- and within-group variance of a measure, and a low ICC means that there is relatively little variance between groups’ (Nezlek 2008, 856–857). However, it is recommended that analysts should use a multilevel model when the data they use has a multilevel structure (Nezlek 2008).