Quality of local government and social trust in European cities

Contextual foundations of social trust

Previous research on social trust using comparative survey data has largely focused on economic and institutional factors (Alesina and La Ferrara, 2002; Delhey and Newton, 2005). For economic factors, the underlying rationale is that in contexts of economic wealth, people are, on average, equipped with sufficient economic resources, which makes them less vulnerable to non-reciprocal or trust-breaching behaviours. Another reason for assuming a positive link between wealth and social trust is that economic development alters long-term value priorities where people increasingly prioritise values related to individualism and self-direction over security and conformity (Inglehart, 1997). Individualists are more likely than collectivists to pursue collective group interests out of intrinsic motivation (Yamagishi, 1988) and to trust others beyond ingroup boundaries (Berigan and Irwin, 2011). Besides wealth, negative consequences of income inequality have been discussed with reference to social trust (Blake et al., 2015; Uslaner and Brown, 2005). In unequal societies, people increasingly engage in self-interested behaviour, compete for social status, develop anxieties about moving down the social ladder and interact less with people from other social classes (Buttrick and Oishi, 2017; Newman et al., 2015).

In a similar vein, studies on neighbourhood effects have found that economic deprivation of neighbourhoods or communities is a critical factor hampering social trust, and that group-based inequalities and residential segregation represent further trust-inhibiting factors (Abascal and Baldassarri, 2015; Letki, 2008; Sampson, 2012; Ziller and Spörlein, 2020). Moreover, residential instability is substantially related to lower levels of local social ties and friendship, and of social trust (Sampson, 1988, 2012).

With reference to institutional foundations, explanatory approaches have emphasised fairness or impartiality as a main feature of an institutional setup that facilitates social trust (Rothstein and Stolle, 2008; Rothstein and Teorell, 2008). The idea is that if public officials apply regulations equally to all citizens alike, which means that they operate in an unbiased and non-corrupt fashion, this signals the reliability of the institutional and moral fabric of a society. In turn, the reputation of institutions operating impartially reassures people not only when dealing with public authorities, but also in social interactions with other citizens. If citizens share the expectation that dishonest and exploitative behaviour will be sanctioned, then they will have a higher incentive to comply and will be less inclined to commit trust-breaching behaviours. Iteratively, this should lead to less social conflict, more cooperation and higher rates of social trust in the long run (Sønderskov and Dinesen, 2016).

Theorising about institutional quality making the social world more predictable and thus facilitating social trust rests on the assumption that people recognise the institutional framework they live in. Since citizens often have limited political sophistication, it is unrealistic to expect them to accurately observe and evaluate national policies and institutional regulations (Campbell, 2012; Soss and Schram, 2007). Instead, several authors have argued that people learn about the institutional conditions they live in by having contact with intermediary agents such as police officers, teachers and public officials, as well as by experiencing (and assessing) specific welfare services such as childcare and health services (Kumlin, 2004; Sønderskov and Dinesen, 2016). By focusing on the local context of municipalities’ public service provision, our study provides a more direct test (compared to cross-national designs) of how variations in the performance of local institutions affect people’s perceptions of co-residents being trustworthy.

The role of quality of local government

Local governments are responsible for a range of services, including water and energy supply, waste collection, child and elderly care, education, public transport, maintenance of roads and public buildings and spaces, youth service and cultural offerings (Narbón Perpiñá and De Witte, 2018). Local governments differ in the way they provide such services both with regard to their quality and in terms of efficiency, that is, the administrative performance through which a municipality creates relevant output given available resources (Borge et al., 2008). Providing high-quality services represents a desirable goal of citizens and public officials as the quality of public services is positively linked to residents’ intention to stay in the city or neighbourhood instead of moving out (Shinohara, 2018), levels of civic participation (Chatzoglou et al., 2013) and collective efforts to reach common goals (Mugion et al., 2018; Sampson, 2012).

In terms of efficiency, administrations typically lack incentives to be cost-effective in providing services. However, political competition, financial constraints and the implementation of new management tools can improve the cost-efficient allocation of service provision (Boyne and Walker, 2004). In addition, social, demographic, economic, spatial and historical factors have been found to shape variation in local government efficiency (see Narbón Perpiñá and De Witte, 2018, for a review). From the perspective of ‘institutional reflexivity’ (Lampe, 2017), efficiency in providing public services reflects administrative skills and capabilities that are relevant for addressing challenges such as changes in administrative tasks, territorial reform, demographic change including immigration, budget cuts, the provision of additional housing, pollution control, and renewals or upgrades of infrastructure.

Apart from public service quality and local government efficiency, impartiality or fairness of bureaucracies is typically considered as a central dimension of quality of government (Charron et al., 2014; Rothstein and Teorell, 2008). At the national level, impartiality becomes apparent in an absence of corruption, which is substantially related to mass education and economic development (Uslaner, 2017). Local variations in institutional fairness can be attributed to historical development, economic characteristics (e.g. inequality) and political factors (e.g. anti-corruption regulations, incentives for public officials) (Charron et al., 2014; Olsen et al., 2019). Although public service quality, local government efficiency and institutional fairness represent different dimensions of the overall quality of local government, we expect substantial interrelations due to common origins. Ultimately, all three aspects are expected to positively relate to social trust, assuming the following underlying mechanisms.

In order to link public service quality and efficiency in public service provision and social trust, we draw on the literature on (contextual) determinants of social trust and employ two arguments. First, public service quality, efficiency and fairness of local governments should shape how reliable they are perceived to be by citizens, which in turn reduces uncertainty in social interactions and facilitates residents’ social trust. This argument draws from the literature on the institutional foundations of social trust according to which shared perceptions of reliable local institutions strengthen people’s beliefs that exploitative behaviour will be sanctioned, which, in turn, facilitates social trust and readiness for collective action (Brehm and Rahn, 1997; Rothstein and Stolle, 2008). People learn about the quality and efficiency of service provision by ‘consuming’ or using these services and infrastructures in everyday life. Community residents may also notice the problem-solving capacity of their local government, particularly while having contact with public officials, and by relying on second-hand experiences of co-residents. This noticing can be related to the administrative output as well as the perceived procedural justice, which may result in a rather symbolic satisfaction with the administrative process (Yang and Holzer, 2006). The expectation that the perceived reliability of local governments strengthens social trust refers to a rational notion of social trust (Hardin, 2002), according to which, information and regulations make the social world more predictable and thus facilitate social trust.

Second, an intact local infrastructure provides better opportunities for people to meet and interact with each other (and thus improves social trust) compared to places characterised by poor public service provision. For example, people are more likely to stop and talk to neighbours where cues of physical and social disorder are absent (Ross et al., 2001), and people are more likely to use playgrounds (and have contact with others there) if the playgrounds are intact (Miles, 2008). In contrast, there are many examples of how a lack of infrastructure may inhibit social contact and connectedness. People will avoid public places and parks that are not well maintained; a lack of public transport will lead to higher rates of people commuting by cars; and a poor educational infrastructure will cause more resourceful parents to transfer their children to better schools, which increases segregation (and thus lack of contact) along socioeconomic lines (Cass et al., 2005; Douglas et al., 2017). In a related vein, Sampson (2012: 158) notes that ‘the civic infrastructure of local organizations and voluntary associations helps sustain a capacity for social action in a way that transcends traditional personal ties’. The expectation that public service provision may enhance social contacts and thus social trust between residents refers to both a rational and a communitarian notion of trust formation that heightens the role of repeated social interactions and common social identity (Braithwaite, 1998; Glanville et al., 2013).

Institutional performance and social trust: Issues related to causality and measurement

The theoretical arguments presented presume that institutional performance causally determines social trust, which is in line with several prior studies (Freitag and Bühlmann, 2009; Kumlin and Rothstein, 2005; Rothstein and Stolle, 2008; Sønderskov and Dinesen, 2016). Nevertheless, others have argued that high levels of aggregated social capital and social trust (e.g. within sub-national regions) determine well-functioning political institutions. In a seminal study, Putnam (1993) argues (and finds empirical evidence) that the civic culture of Italian communities positively relates to the performance of regional governments. Similarly, studies find that high levels of aggregated social capital indicators, which include trust, predict lower levels of government corruption and higher institutional quality (Bjørnskov, 2010; Coffé and Geys, 2005; La Porta et al., 1997).

While a reciprocal causal relationship – where institutional performance and social trust determine each other – is plausible in terms of theory, such endogenous loops hamper inference from statistical models that typically assume exogenous predictor variables. Studies addressing endogeneity that employ models on longitudinal data or instrumental variables are scarce (but see Coffé and Geys, 2005; Sønderskov and Dinesen, 2016). The present study makes use of three-wave panel data at the level of European cities. This enables us to tackle issues related to endogeneity by modelling reciprocal empirical relationships using cross-lagged panel models (Allison et al., 2017).

With reference to the measurement of public service provision, we argue that aggregated survey responses represent the quality of public service provision and local government efficiency in a more reliable way than accounts that use structural indicators on municipality characteristics. The assessment of service quality using structural indicators remains problematic, as the definition of quality criteria is necessarily a subjective decision. Attempts in this regard include approximations using technical efficiency measures (Zafra-Gómez et al., 2010; Ziller and Goodman, 2020) or indicators of collective behaviour such as electoral turnout (Balaguer-Coll and Prior, 2009). While structural indicators of local governments are typically comparable within the same country, indicators from different countries are often hardly comparable or too unspecific to map local public service provision (Borge et al., 2008).

A more direct measure of quality refers to perceptional indicators looking at citizens’ satisfaction with local services (Balaguer-Coll and Prior, 2009), expert ratings (Kaufmann et al., 2009), crowd ratings of service users (Hendrikx et al., 2018) or a combination of expert ratings and aggregated public opinion data (Charron et al., 2014). Given inconsistencies in structural local government indicators, perception-based indicators appear to be better comparable when taking data from more than one country context into account. From a conceptual point of view, taking people’s perceptions into account addresses an oft-cited issue in community research, and research on contextual effects more generally; namely that context characteristics will for the most part influence behaviours and attitudes to the extent that they are perceived (e.g. Koopmans and Schaeffer, 2016; Sampson et al., 2002). In technical terms, it is reasonable to assume that a municipality’s actual quality of local government shapes residents’ perceptions thereof, which in turn possibly impacts social trust. It is important to note that such a perception-based indicator (as we intend to use) necessarily refers to (aggregated) municipality-level differences and not individual differences. ¹ In addition, a large number of observations for building aggregated perception indicators increases their reliability and validity (Schunck, 2016).

Data and variables

To test the stated hypotheses, we use repeated cross-sectional survey data from the project Quality of Life in European Cities (European Commission, 2016), which collects survey data on European cities with an urban centre of at least 50,000 inhabitants, containing probability samples of about 500 residents per city and wave. For the main analyses, we use the two most recent waves from 2012 and 2015, as the included indicators are comparable. For additional tests focused on causal identification of the key variables, we additionally use the 2009 wave. To maintain comparability in terms of institutional and historical trajectories, we exclude cities of post-communist countries. ² Although the survey represents no panel at the level of individual respondents, repeated observations at the city level enable us to test longitudinal relationships of city-level indicators (Fairbrother, 2014).

The outcome variable social trust refers to the perceived trustworthiness of fellow city residents (‘Generally speaking, most people in [CITY NAME] can be trusted’) and is measured with a 4-point answer scale which was recoded to range between 1 (strongly disagree) and 4 (strongly agree).

We operationalise quality of local government in three ways. First, we measure the general perceived efficiency of local governments using the item ‘The administrative services of [CITY NAME] help people efficiently’ (recoded 4-point scale, ranging from 1 ‘strongly disagree’ to 4 ‘strongly agree’). Second, we gauge institutional fairness using an item on the perceived trustworthiness of local administrations (‘Generally speaking, the public administration of [CITY NAME] can be trusted’, recoded 4-point scale, ranging from 1 ‘strongly disagree’ to 4 ‘strongly agree’). Third, we measure public service output quality as satisfaction with public services using indicators on how satisfied respondents are in general with public transport, sports, facilities, healthcare provision, cultural offers (e.g. concert halls, theatres, museums, libraries), the state of streets and buildings in the neighbourhood and public spaces (e.g. markets, squares, pedestrian areas). Answer scales (4-point) range from 1 ‘not at all satisfied’ to 4 ‘very satisfied’ (recoded), and factor scores from a confirmatory factor analysis (standardised coefficients are all >0.45, CFI = 0.984, RMSEA = 0.044, SRMR = 0.021) serve as an index. ³

In addition to individual perceptions, we built city–year-level variables of the three perceptual indices by aggregating the individual-level variables, which in the empirical analysis reflect the contextual effect – that is, a city-level characteristic for the given survey year net of compositional differences across cities. Given a sufficiently high number of respondents (in this case about 500 per city and wave), these aggregated indicators are assumed to represent a valid measure of the actual quality of local government of a given municipality in a given year. We assess the convergent validity of the perception-based indicators by assessing correspondence to expert ratings (see Results section).

As control variables, we include indicators of respondents’ socio-demographic status: age, sex and education in years. Economic status is measured with a dichotomous measure indicating whether or not respondents have encountered problems in paying bills at the end of the month during the past 12 months (0 = never, 1 = from time to time, or most of the time). Time living in the city of residence is measured with five dummy variables (have lived in city entire life, more than 10 years, between 5 and 10 years, between 1 and 5 years, less than 1 year). As time-varying city-level variables, we use unemployment rates and proportion of immigrants obtained from the Eurostat database (Eurostat, 2019), as well as GDP per capita and population density (2012 and 2015) obtained from the OECD city database. ⁴

A list of included cities, descriptives of the used variables and zero-order correlations between individual-level as well as macro-level variables is presented in the online appendix (Tables A1–A3).

Method

To test the stated hypotheses, we employ multilevel models using survey data of the 2012 and 2015 waves. The central predictor variables are city–year aggregates of quality of local government indicators. We include in all models individual-level control variables and a wave dummy variable to account for general shifts in the outcome over time. The models include a random intercept at the city–year level as well as city fixed effects which control for potential confounding from unobserved time-constant heterogeneity at the city level. This longitudinal analysis provides a test on how a change in city-level institutional quality relates to a change in social trust, and is more credible than results obtained from models without unit fixed effects. In additional models, we include city–year variables as control variables that are available only for a subset of cities. The functional form of the employed multilevel model is given as:

Y itjk = α 0 + β X itjk + γ Z tjk + δ Ci t jk + θ T t + u tjk + e itjk

Index i indicates individuals, index j relates to cities and index k indicates countries. Index t refers to the survey year. X itjk is a vector of individual-level variables, and β is the corresponding vector of coefficients. Z tjk refers to time-varying city-level variables, including indicators of quality of local government. Models include city dummies Ci t jk (or city fixed effects). They absorb at the same time the entire variance belonging to countries, and leave time-variant information only for estimation of the coefficients (Allison, 2009; Fairbrother, 2014). The time variable denotes as T t . u tjk is the random intercept of the city level, and e itjk is the idiosyncratic error term. We use an identity link function for estimating the main models. Since the outcome variable has only four categories, we also present models using an ordered logit link function as a robustness test. To facilitate the interpretation of the results in terms of standard deviations, we z-standardised all continuous variables.

To assess causal ordering of the hypotheses-relevant variables, we employ cross-lagged panel models with unit fixed effects. Previous methods using cross-lagged models have been criticised for not sufficiently taking unit-specific unobserved heterogeneity into account, which has been solved by recent methodological developments (Allison et al., 2017; Hamaker et al., 2015). With regard to our model specification, we employ maximum likelihood structural equation models using aggregate survey responses of the key variables from three available waves. We include as predictors a lag-1 variable of the predictor and outcome and a contemporary variable of the predictor, as recommended by recent approaches (Leszczensky and Wolbring, 2019).