The re-election of corrupt mayors: context, relational leadership and level of corruption

ABSTRACT

This study examines why citizens re-elect corrupt mayors, adopting a conceptual framework that considers two mitigating factors in punishing corrupt leaders: first, a strong relationship between the mayor and the citizens; second, a low level of corruption that may be considered negligible by the citizens. Both factors are contingent on local contexts. Thus, the study investigates in clustered local contexts, the impact of variables related to the mayor-citizen (taking account of the duration of the political career, party membership and trasformismo/party-switching) and the level of corruption on mayoral re-election. The results shows that the relationship between the mayor and the citizens is a mitigating factor that works homogeneously across contexts, albeit based on different factors. With regard to low levels of corruption as a mitigating factor, evidence of this is found only in medium and highly developed communities. The theoretical and policy implications are examined.

KEYWORDS:

• Context
• corruption
• mayoral leadership
• re-election

1. Introduction

The paradoxical fact that corruption is unpopular while corrupt politicians are in many cases popular at the local level has often been highlighted. Particularly, devolution and the consequent concentration of power in their hands seem to have increased the opportunities for corruption. However, the lack of a backlash against corruption on the part of the electorate is well established. As a result, it is not surprising that numerous studies have focused on the re-election of corrupt mayors (e.g., Ferrer 2020).

A number of reasons have been put forward to explain the re-election of corrupt mayors, exploring heterogeneities within the electorate, summarised in two main factors that mitigate the proclivity of the citizens to punish corrupt politicians (Klašnja, Lupu, and Tucker 2021). First, the citizens may not be aware of, or interested in, corruption scandals (Ferrer 2020). Second, voters may condone the behaviour of mayors if they share the same ideology, or there may be a tacit trade-off. This is evident in expectations on the part of the voters to benefit from clientelistic networks (Benito et al. 2019), and in cases in which despite cases of corruption, the administration is nevertheless efficient (Muñoz, Anduiza, and Gallego 2016). For the sake of simplicity, we call this the ‘condoning hypothesis’.

This approach has three limitations. Although mitigating factors have often been applied to cases in which corrupt mayors are re-elected (e.g., Muñoz, Anduiza, and Gallego 2016), they do not have specific characteristics relating to the role of mayors. Rather, they seem to be valid for any political leader, underestimating the specificity of roles at different levels. Many studies show that direct elections give rise to a strong relationship between the mayor and the citizens, with both actors playing a distinct but unified role, with an influence on voting behaviour (Hartley 2018). This relational perspective derives from leadership studies, focusing on the ability of mayors to exert influence based on their political experience, personal networking, degree of autonomy from the political parties at national level, and a willingness on the part of voters to confer electoral legitimacy on corrupt mayors (Crosby and Bryson 2018).

Furthermore, re-election is not unconditional: the level of corruption can play a key role in the re-election of corrupt mayors, since corruption may be more visible due to the small size of the municipalities compared to regions and state. It can be easily traceable along a continuum ranging from less serious cases of corruption, in all probability deemed to be tolerable, to more serious cases that should work against re-election. However, previous studies consider corruption as a stereotypical event, with few studies identifying the seriousness of corruption as a factor influencing the re-election of corrupt political leaders (Costas-Pérez, Solé-Ollé, and Sorribas-Navarro 2012).

Finally, although the mitigating factors in the literature emphasise the differences between various voting communities, they capture some contextual features but not the context as a whole, consisting of a set of environmental and broader social and economic factors capable of providing overarching explanations for voting behaviour (Kernalegenn and van Haute 2020).

This study is intended to address these shortcomings by assessing the relationship between political leaders and the citizens, with low levels of corruption seen as additional mitigating factors in relation to the sanctioning of corrupt mayors in line with the corruption condoning hypothesis.

This helps to address explanations focused on the re-election of mayors in spite of evidence of corruption. The relation is specific to the mayor as key feature of local government as well as not homogeneous among mayors. It is up to the Mayor’s ability to be re-elected and the propensity on the part of the citizens to re-elect him/her.

The willingness to re-elect corrupt mayors is not unconditional, as it takes count of different levels of corruption by identifying cases that can be condoned as they do not give rise to high costs for the electorate, falling below a certain threshold. These factors can produce different outcomes according to the context where re-election occurs (Troisi and Alfano 2019).

Consequently, providing insights into the specific local conditions under which re-election occurs can cast light on why certain corrupt leaders are popular in their communities, and why fighting corruption is a ubiquitous challenge, with ambivalent results in various local contexts.

In order to identify the effect of these two mitigating factors on the sanctioning of corruption in local contexts, this study focuses on Italian mayors involved in corruption resulting in judicial proceedings during their term of office. We test whether in different local contexts re-election is conditional on the relationship between the mayor and the citizens (considering the duration of the political career, party membership and trasformismo), and on the degree of seriousness of the corruption. In the first phase, we cluster municipalities into sub-groups considering socio-economic features together with social capital indicators. We then use within-cluster binomial logistic regression for a comparison-based appraisal to ascertain whether the relationship between the mayor and the citizens, and the seriousness of the corruption, affect the likelihood of re-election. The results show that the relationship between the mayor and the citizens is a mitigating factor that works homogeneously across contexts albeit based on different factors. The low-level corruption seriousness tends to be condoned in highly developed municipalities, while in less developed municipalities mayors are more likely to be re-elected even in cases of high-level corruption.

The relationship between the mayor and the citizens works as a mitigating factor in a more homogeneous way: representation seems to be a matter of similarity rather than responsibility, with the mayor primarily representing the identity of the community and, to a lesser extent, its interests.

This article is structured as follows: Section 2 consists of the literature review, then Section 3 outlines the methodology used. The results of the analysis are presented in Section 4, and Section 5 consists of the discussion and conclusions.

2. Literature review: re-electing corrupt politicians

The literature on the re-election of corrupt politicians focuses on individual characteristics with a potential impact on voting behaviour. Informational asymmetries have been widely studied as a mitigating factor. Thus, if the electorate is not well informed, this helps to explain the lack of political accountability. However, this has seldom been empirically validated in terms of the positive impact on the likelihood of the re-election of corrupt politicians (Ferraz and Finan 2011). Information on corruption might simply reduce voter turnout (Costas-Pérez, Solé-Ollé, and Sorribas-Navarro 2012). Condoning corrupt politicians shows that voters may be aware of corrupt behaviour but turn a blind eye. Support for corrupt politicians may be justified by voters who share the same ideology, or when there is no alternative candidate from an ideological point of view (Charron and Bågenholm 2016). However, the importance of party loyalty is controversial: Konstantinidis and Xezonakis (2013) reported on cases in which party loyalty did not modify the sanctioning of corrupt Greek politicians by the electorate. Ideological considerations appear to be predominant among voters who are less politically sophisticated (Muñoz, Anduiza, and Gallego 2016).

A further important consideration is that voters may knowingly re-elect corrupt politicians due to an implicit trade-off (Sandholtz and Koetzle 2000). Overlooking cases of corruption may depend on the mayor’s positive performance in terms of access to and distribution of resources. Barberá, Fernández-Vázquez, and Rivero (2016) show that short-term economic benefits arising at the same time as the corruption lead to the re-election of corrupt incumbents. In other words, voters expect to gain personal benefits within a clientelistic network (Benito et al. 2019).

Furthermore, a complementary strand of the literature emphasises the difference between various voting communities in the re-election of corrupt mayors, by highlighting specific contextual features. Contexts characterised by high unemployment, low levels of education, and a high ageing index have always been considered fertile ground for electoral manipulation, for the construction of clientelistic networks to control through political budgetary cycles and particularistic-driven policies (Vergne 2009; Weitz‐shapiro 2012; Meza 2015) and for mitigating the impact of corruption scandals (Van der Eijk and Franklin 2009). However, a number of studies show that strong economic growth (e.g., Zechmeister and Zizumbo-Colunga 2013) tends to weaken a backlash by the electorate against politicians known to be corrupt. Social capital is equally crucial in terms of the tendency of different groups to react strongly to corruption. At local level, high levels of social capital have been shown to reduce the impact of corruption by means of electoral competition. Voters with values and beliefs that foster cooperation are more likely to elect candidates who put social welfare before their narrow personal interests (Nannicini et al. 2013).

We posit that in the case of mayors, the response of the citizens to corruption depends on factors at local level, both individual and contextual. The relation between political leaders and the community and the low level of corruption seriousness within specific contexts are proposed as additional mitigating factors and discussed in Section 2.1.1.

This view departs from explanations based on the role of information asymmetries due to the higher visibility of cases of corruption in municipalities compared to regional or central government. We examine the seriousness of corruption in terms of its continuity and magnitude, resulting in greater visibility. However, the focus on relations at municipal level reduces the importance of traditional ideology, as the strong link between mayors and the citizens is considered to encourage engagement on the part of the public in response to the decline of traditional political parties, and the diminishing importance of ideology, especially at local level (Fava 2017). Finally, the relationship between the mayor and the citizens, and the seriousness of the corruption, are not necessarily related to any implicit trade-off. The mayor’s influence and the legitimacy conferred by the citizens play a key role in voting behaviour. Similarly, minor cases of corruption may be passed over even by voters not dependent on the mayor or benefitting from clientelism.

2.1. The conceptual framework

In this section we outline our conceptual framework based on three key concepts: the relationship between the mayor and the citizens, the level of corruption, and local context in understanding the re-election of corrupt mayors. Moreover, we highlight their specific features leading to two hypotheses.

Figure 1 outlines the conceptual framework.

Figure 1. Conceptual framework.

2.2. The relationship between mayors and the citizens. Lessons from leadership studies

The interaction between the leader’s influence and the legitimacy conferred by the citizens is the first mitigating factor in the re-election of corrupt mayors. Previous studies have tended not to focus on this factor. Certain communities condone corruption since they identify with a particular political leader and are willing to overlook certain shortcomings. The strong bond between mayors and the local citizens derives from direct mayoral elections, conferring legitimacy with political significance for their communities far beyond formal powers. This relationship has been analysed in terms of the positive effects it produces (Hartley 2018). However, limited empirical research has been carried out into how voters contribute to consolidate a corrupt leadership by tolerating or even promoting it through re-election.

Three factors of the relationship between mayors and the citizens can have a positive impact on the likelihood of re-election of corrupt mayors. The first is the duration of the mayor’s political career. A lengthy term of office can enable mayors to build a ‘personal brand’ with the citizens, developing community networks and weakening the opposition (Egner et al. 2018). A prolonged term of office can significantly strengthen the link between the mayor and the citizens, as pointed out by Aragón and Pique (2020), who found that political seniority increases a politician’s chances of re-election. Mayoral experience and a long career in local politics have been found to increase the chances of re-election in cases of corruption compared to corrupt mayors with a shorter political career (Ryšavý 2013), and in the case of corrupt mayors in municipalities where the opposition parties were under-represented (Britto and Fiorin 2020; Ferrer 2021).

The distance from the central parties can strengthen the relationship with a positive impact on re-election, as illustrated in two ways. First, consider the membership of a local party with no connections to a central party. Mayors supported at local level for who they are rather than for their political affiliation have a greater chance of re-election because of this direct relationship with the citizens (Egner et al. 2018). In Italian municipalities, mayors running on local lists seem more likely to be re-elected (Nese and Troisi 2019), particularly in small municipalities (i.e., 85% of all municipalities in Italy) with the highest percentage of voters in the less developed areas (Vampa 2015). These arguments can be used in the case of corrupt mayors running for a civic list. More specifically, non-partisan actors are more likely to be re-elected than partisan actors at local level, based on the idea of a shared programme and identity group that is stronger at local level.

Another important indicator of the autonomy of mayors and their direct relationship with the citizens with a positive impact on the chances of re-election is trasformismo, which describes how throughout their political career, mayors may switch parties repeatedly, including mayors quitting the party for which they were elected for different public positions. This phenomenon has increased with the decline of party affiliation, leading to an unstable environment of party-switching politicians with autonomous control of the local electorate (Bianchi 2015). At central level it has been shown that transformisti involved in corruption are more likely to be re-elected than party politicians involved in scandals thanks to their autonomous electoral control and the absence of a veto on the candidacy by their party (Calamandrei 2017). This argument could be extended to the local level.

As for the specific local conditions under which re-election occurs, a number of studies focus on how the relationship between the mayor and citizens can vary according to the context. In the case of the duration of the term of office, some studies emphasise specific socio-economic features as covariates in the re-election of corrupt mayors. In high-income municipalities, it has been shown that the chances of corrupt mayors with a long career being re-elected are greater than in low-income municipalities. This may be because long-serving mayors may favour established networks that can boost job opportunities and incomes within the network (Vukovic 2020). However, some other studies have shown that additional covariates including low educational levels and low rates of employment favour the re-election of long-serving mayors (Ferraz and Finan 2011).

Mayors belonging to civic lists involved in cases of corruption tend to be re-elected with a voter turnout comparable to the first term in contexts where education and social capital are low (Giommoni 2021). This can be interpreted as a choice in opposition to ideological parties, typical of less sophisticated socio-demographic groups. In these environments, civic lists promote local values rather than ideologies associated with the traditional parties (Fava 2017).

Finally, based on anecdotal evidence, it has been posited that trasformismo can increase the likelihood of re-election of corrupt mayors mainly in less developed communities. This is mainly due to the fact that in these contexts it is easier to cultivate ties with the citizens, reinforced by means of clientelistic measures, that are unaffected when mayors switch parties (Magone 2021).

Since specific contexts produce contrasting results in terms of the effects on the length of the mayor’s political career in favouring re-election, while in less developed contexts civic lists and trasformismo tend to favour re-election, the following hypothesis may be formulated:

H1a

The duration of the mayor’s political career should have a positive impact on re-election

H1b

In less developed contexts, running for a local list and trasformismo should have a positive impact on re-election

2.3. The level of corruption

The re-election of mayors is not unconditional: as not every corruption scandal has the same electoral impact, re-election may depend on the seriousness of the corruption. Low levels of corruption, that may be condoned or tolerated by the citizens, can result in re-election since they do not entail high costs for the taxpayer.

A small number of studies identify a cut-off point above which corruption does not lead to re-election as it is particularly significant or evident (Vuković 2020). However, other studies support the implicit trade-off theory and identify a perceived threshold, which varies according to environmental conditions. In this connection, during recessions, corrupt rents used for social spending can be deemed to be mild corruption as they generate positive externalities. In contrast, the same case can be seen as serious in environments where corruption is a salient issue with little focus on beneficial side effects (Ecker, Glinitzer, and Meyer 2016). Other studies rely on information theory, with the extent of media coverage considered as a proxy for the seriousness of a scandal, since the focus is on cases that are morally unacceptable. Cases of local corruption reported in national newspapers, deemed to be the most serious, diminish the chances of re-election (Praino, Stockemer, and Moscardelli 2013).

This study does not consider perceived levels of corruption, but objective levels combining the structural elements of the corruption with the sanctions imposed by the Italian penal code. We posit that the more objectively serious the corruption, the more visible it will be within a small community. The structural factors of corruption concern the number of actors involved, referred to the extent of the corruption, and the duration of the corrupt behaviour (Meza and Pérez‐chiqués 2021). In particular, Ferraz and Finan (2011) have shown that second-term mayors engage in more complex kinds of corruption than first-term mayors. Moreover, we consider the sanctions imposed by the Italian penal code for each kind of corruption, in terms of average length of custodial sentence. We consider objective elements of corruption to be preferable at micro level for two reasons. The alternative perception-based indicators are usually available only at national level, and tend to overlook local-focused features that can better identify the level of corruption. Finally, some authors criticise perception-based indicators due to the lack of experience of real cases of corruption of the respondents or bias due to general sentiment (Klašnja 2015).

Although, to the best of our knowledge, there is no evidence of how these objective factors affect the chances of re-election, they can be interpreted in the same way as the perceived seriousness of corruption. Studies relying on subjective indicators show that the more widespread corruption is perceived to be, the more the citizens tends to punish the politicians involved in scandals (Costas-Pérez, Solé-Ollé, and Sorribas-Navarro 2012).

Finally, we consider how the effect of the level of corruption on the re-election of the mayor varies across contexts. We consider studies that focus on the level of corruption as a whole rather than the result of features contributing to the definition of the level of corruption. Bågenholm and Charron (2020) found that in contexts with higher rates of unemployment, involvement in more serious cases of corruption did not reduce the chances of re-election and had insignificant effects on voter turnout. Ferraz and Finan (2011) found an inverse relationship between the seriousness of corruption, in monetary terms, and the re-election of corrupt mayors in contexts characterised by higher levels of education. Moreover, some studies show that corruption is deemed to be less serious in more developed communities with a higher civic sense (Costa and Kahn 2003). Overall, the findings support the hypothesis that the seriousness of corruption is not necessarily a constraint on re-election in less developed contexts. However, serious cases of corruption can reduce the chances of re-election in highly developed contexts where only low-level corruption would tend to be condoned or tolerated.

These arguments lead to the following hypothesis:

H2

In highly developed contexts, high levels of corruption are likely to constrain the re-election of corrupt mayors while low levels are likely to be compatible with re-election

3. Methodology

3.1. Data sources

This study drew on judgements of the Italian Cassation Court between 2012 and 2019, with a content analysis of approximately 700 sentences aimed at identifying cases of mayoral corruption. Sentences were randomly selected from the official website¹ from among approximately 2500 cases using the search term ‘mayor’. Records of the judgements provide information about the judicial proceedings. Information not contained in the judgements, such as information about the mayor’s political career or socio-economic characteristics of the municipalities, was obtained from other sources, such as the Ministry of the Interior, and the National Institute of Statistics (ISTAT). A second randomised dataset relating to approximately 500 non-corrupt mayors was constructed following a search of the Ministry of the Interior database to compare the re-election of corrupt and non-corrupt mayors.

3.2. Variables

This section outlines the variables used in the cluster analysis to define the local contexts. We then discuss the dependent variable of re-elections, and the independent variables referring to the relationship between the mayor and the citizens and level of corruption, used in the logistic regressions. The context variables focus on the socio-economic characteristics and social capital indicators. The socio-economic characteristics of municipalities were described using indices provided by ISTAT, including the ageing index, vulnerability index, levels of educational attainment, level of urbanisation (inhabitants/km²), unemployment rate and mean per capita income for the municipality. Voter turnout, the percentage of employees working for non-profit bodies, and the crime rate (the number of complaints to law enforcement agencies per 100,000 inhabitants) were taken as proxies for social capital. These parameters have been used to define the socioeconomic conditions of municipalities in previous local development studies (Nannicini et al. 2013).

The variables in our logistic regressions are as follows. The dependent variable was the re-election of the mayor. This is a dichotomous variable, with a value of 1 if the mayor is re-elected, and 0 if not.² In the sampling phase, mayors not standing for re-election were excluded (approximately 3%).

(i) Variables relating to the relationship between mayors and the citizens

The proxies for the relationship between mayors and the citizens are based on data from the Ministry of the Interior website. We consider the duration of the mayor’s political career in years. Party political membership is operationalised as a dichotomous variable assuming a value of 1 for mayors elected for a local party, 0 for mayors who were members of a central party. Trasformismo is expressed as a count variable measuring the number of changes between parties during the mayor’s career.

(ii) Variables relating to the municipality’s economic performance

In controlling whether implicit exchanges come into play, we focused on two proxies for mayoral performance: investment in public order and environmental development (mainly waste management) during the mayor’s term of office. Data were taken directly from the annual accounts of the municipality and expressed in euros per capita.³ These variables create one main potential issue: as stated in many studies on political budget cycles, the amount of investment is an endogenous variable and may be the result of attempts by mayors to increase their electoral appeal as local elections approach (Vergne 2009). To deal with this issue, the average amount of investment during the term of office is considered.

(iii) Variables relating to corruption

The second set of variables relates to the structural factors of corruption as described in the Court rulings. We consider the duration and the number of actors, distinguishing between cases where the corruption was engaged in individually, in an ad-hoc group, or by an ongoing criminal association. This variable is categorical, assuming a value of 1 if the corruption was individual, 2 if ad hoc, and 3 if in a stable association. The sentence was codified as a categorical variable with three groups in ascending order according to the average length of custodial sentence as laid down by the penal code: (1) up to three years; (2) 3–6 years; and (3) 6–9 years. The final dummy variable was the focus of the crime, classified as public works contracts, licenses/permits, elections, and human resource management.

Finally, to reduce the endogeneity deriving from an uncontrolled confounding variable such as the weakness of the opposition, we introduce into the model the difference of votes between the elected mayor and the second-place candidate as a proxy for political competition (Pereira, Melo, and Figueiredo 2009).

3.3. Research method

A two-step method was used: first, an initial cluster analysis of the characteristics of the contexts; second, a logistic regression within each cluster for a comparison-based assessment. The relationship between the mayor and the citizens as well as the corruption variables may have a different impact within the communities identified according to the socio-economic and social capital characteristics. Clustering the communities based on these two factors enables macro-groups of municipalities to be identified in terms of similarities. Our cluster analysis generated six clusters, three for each dataset of corrupt and non-corrupt mayors, respectively. Using R-project software, hierarchical cluster analysis was performed. Ward’s minimum variance clustering method was used, applying squared Euclidean distance. The number of clusters was then defined by a dendrogram and silhouette coefficient analysis.

In the second step, a multivariate logistic regression was performed for each individual cluster to evaluate the impact of the variables concerning the relationship between the mayor and the citizens as well as the level of corruption on the likelihood of re-election of corrupt mayors. Running individual regressions on each cluster produces more reliable results than traditional regression models as it is based on a more uniform data structure (Klimberg, Ratick, and Smith 2017). The binomial logistic regression is formally described by the following equation:

$logit\left(p\right)=log\left(\frac{p\left(y=1\right)}{1-p\left(y=1\right)}\right)={\beta }_0+{\beta }_1\ast x_1+\dots +{\beta }_k\ast x_{\begin{array}{c}k\end{array}}$

in which y is the dependent variable defined as a binary variable, in this case re-election. The ${\beta }_i$ coefficients, using maximum likelihood estimation, describe the effect of each independent variable on the logarithm of the odds ratio. Intuitively, in the case of non-corrupt mayors, only variables relating to mayoral power were used. The results from the cluster and logit models are outlined in Section 4.

4. Results

Ward’s criterion provided the most clearly defined groups within the dendrogram analysis (Figures 2–3) and the greatest silhouette coefficient.

Figure 2. Dendrogram for corrupt mayors.

Figure 3. Dendrogram for non-corrupt mayors.

Table 1 shows the summary statistics of the characteristics of the local context for the six clusters, Clusters A1 to A3 regarding corrupt mayors and Clusters B1 to B3 regarding non-corrupt mayors, respectively. Cluster A1 groups observations in which local contexts have the least favourable socio-economic characteristics. For this cluster, local context shows the least advantageous values for almost all the socioeconomic and social capital variables, with the lowest voter turnout (63.08%) the highest unemployment (40.53%) and vulnerability index (102.72). Since the vulnerability index was used to minimise the variance with respect to the value 100 (ISTAT), even a small difference indicates a large difference in the level of vulnerability.

Table 1. Cluster analysis results. Mean values of the variables (standard deviation in parenthesis).

	Sample 1 (corrupt mayors)			Sample 2 (non-corrupt mayors)
	Cluster A1	Cluster A2	Cluster A3	Cluster B1	Cluster B2	Cluster B3
Ageing index	119.04 (45.15)	188.93 (89.66)	176.22 (71.82)	150.33 (62.20)	181.75 (85.01)	185.43 (95.37)
Unemployment rate	26.31 (6.79)	14.38 (6.40)	12.12 (5.14)	14.86 (6.30)	14.52 (5.91)	12.06 (5.45)
Low education rate	40.53 (5.37)	36.14 (6.74)	33.68 (5.10)	38.70 (6.98)	37.93 (5.51)	33.66 (5.91)
Vulnerability index	102.72 (2.05)	99.89 (1.63)	97.68 (6.37)	101.71 (1.25)	101.15 (1.13)	99.70 (1.57)
Voter turnout	63.08 (9.36)	66.65 (10.53)	71.37 (5.82)	58.56 (4.55)	60.90 (9.79)	72.03 (6.22)
Employees in non-profit firms	4.34 (0.05)	4.76 (1.43)	4.48 (1.21)	5.01 (0.76)	4.77 (0.82)	4.21 (0.87)
Urbanisation (inhabitants/km^2)	428.44 (525.93)	292.31 (325.01)	2251.65 (1515.90)	227.50 (219.96)	1527.27 (694.82)	212.63 (204.76)
Municipality mean income	39908.37 (23717.87)	31281.13 (9766.88)	33773.06 (8942.19)	34654.77 (10223.24)	36410.59 (10690.40)	41711.17 (24422.22)
Crime rate	3689.67 (947.29)	3171.34 (523.75)	4534.61 (1428.19)	3535.55 (1109.64)	3530.67 (967.33)	3246.91 (773.38)
Number of observations	109	468	126	264	52	184

Cluster A3 contains observations predominantly relating to more favourable socio-economic and social capital variables (e.g., the lowest vulnerability index and unemployment rate, 97.68% and 12.12%, respectively). Finally, Cluster A2 consists of municipalities characterised by socioeconomic and social capital variables between those of Clusters A1 and A3. The cluster analysis for corrupt mayors resulted in the identification of three clusters corresponding to: (1) medium-sized municipalities with a lower level of socioeconomic development (A1); (2) small municipalities with an intermediate level of socioeconomic development (A2); and (3) larger municipalities with a higher level of socioeconomic development (A3).

Three clusters were also obtained for non-corrupt mayors relating to lower (B1), intermediate (B2) and higher (B3) degrees of socioeconomic development, respectively. The differences between certain socioeconomic variables were less marked than for corrupt mayors.

The results of the logistic regression models and the diagnostic test are shown in Table 2. Models 1-2-3 refer to Clusters A1-A2-A3, respectively, while Model 4 covers the entire sample of corrupt mayors.

Table 2. Binomial logistic regression results.

	Corrupt mayors				Non-corrupt mayors
	Model 1 Cluster A1	Model 2 Cluster A2	Model 3 Cluster A3	Model 4 Total sample	Model 5 Total sample
(Intercept)	−12.55*** (2.967)	0.654 (0.557)	1.328 (5.271)	0.700 (0.462)	1.610** (0.492)
Political career	0.283** (0.094)	0.050* (0.023)	0.704* (0.277)	0.122*** (0.021)	0.024 (0.014)
Party membership	2.895*** (0.783)	0.850*** (0.217)	2.509* (1.216)	1.051*** (0.178)	−1.857 *** (0.242)
Trasformismo	1.471* (0.685)	0.910*** (0.216)	−2.453* (1.445)	0.459** (0.178)	−1.791 *** (0.248)
Public order expenditure	0.041** (0.013)	−0.003 (0.002)	0.022 (0.022)	−0.002 (0.002)	−0.001 (0.002)
Environmental development expenditure	−0.001 (0.001)	−0.0002 (0.000)	−0.0001 (0.001)	−0.0004 (0.000)	0.002*** (0.000)
Stability of the criminal partnership	1.354** (0.469)	−0.708*** (0.164)	−2.570* (1.297)	−0.469*** (0.130)
Severity of the corruption	1.877** (0.592)	0.112 (0.170)	−1.943* (0.812)	−0.142 (0.131)
Contracts for public works	−0.464 (0.774)	−0.584* (0.282)	−3.365* (2.644)	−0.737*** (0.221)
Licensing	−0.993 (0.963)	−0.017 (0.284)	−0.612 (1.611)	0.139 (0.238)
Elections	0.134 (1.062)	0.822 (0.443)	−1.242 (1.555)	0.535 (0.336)
Difference of votes	0.133 (0.091)	0.045 (0.027)	0.027 (0.146)	0.023 (0.022)	−0.025 (0.021)
Pseudo R^2	0.53	0.13	0.80	0.15	0.18
Number of observations	109	468	126	703	500

Significance codes: ***p < 0.001, **p < 0.01, *p < 0.05.

The logit linearity assumption was checked by the Box-Tidwell test; absence of multicollinearity was verified using the variance inflation factor.

Model 5 refers to non-corrupt mayors. For non-corrupt mayors, logistic regression produced similar results for each cluster. Due to space constraints, Table 2 provides the results for the entire sample of non-corrupt mayors. Table 2 shows that the impact of the independent variables has different levels of significance, and differences in signs of the beta coefficients depending on the different socioeconomic contexts (i.e., clusters). This is particularly evident for the corruption crime variables.

Regarding the variables for the relationship between the mayor and the citizens, the length of a political career is positively related and significant for all models related to corrupt mayors (0.122, p < 0.001 for Model 4), supporting H1a. Longer political careers increase the likelihood of re-election of corrupt mayors. Party membership is equally positively related and significant for the re-election of corrupt mayors in all environments (1.051, p < 0.001 for Model 4), not supporting hypothesis H1b. This suggests that corrupt mayoral candidates with a local party affiliation are more likely to be re-elected than mayoral candidates who are members of a central party, regardless of context specificities. However, in contrast with findings in the literature, this variable is negatively related in the case of non-corrupt mayors (−1.857, p < 0.001): re-election is less likely for mayors affiliated only to a local party.

When lower and intermediate contexts are considered, trasformismo is positively related to the re-election of mayors, in line with hypothesis H1b. (1.471, p < 0.05 and 0.910, p < 0.001 respectively). In highly developed contexts, trasformismo was negatively related (−2.453, p < 0.05) further supporting H1b. This means that in the less developed areas in socioeconomic terms, a mayoral career with considerable mobility between parties increases the likelihood of re-election; conversely, trasformismo in more developed contexts reduces the chances of re-election.

Public order expenditure is significantly positively related to the chances of re-election only in less developed contexts (0.041, p < 0.01): corrupt mayors allocating funds to public order measures are more likely to be re-elected. However, environmental spending is not significant in the case of corrupt mayors, although it is significantly positively related to the re-election of non-corrupt mayors (0.002, p < 0.001). In other words, high environmental expenditure is more likely to result in the re-election of non-corrupt mayors.

The stability of criminal affiliation is significant in all models for corrupt mayors, but in contrasting ways. In Model 1 the stability of the criminal affiliation is positively related to the likelihood of re-election (1.354, p < 0.01), while in Models 2 and 3, it is negatively related (−0.708, p < 0.001 and −2.570, p < 0.05 respectively). These results confirm hypothesis H2. In more highly developed contexts, the probability of re-election of corrupt mayors involved in an active criminal organisation is lower, whereas in less developed municipalities, the probability of re-election increases, consistent with H2.

The severity of a criminal conviction is significantly positively related to the re-election of corrupt mayors in less socioeconomically developed environments (1.877, p < 0.01). In these municipalities, mayors are more likely to be re-elected even after committing serious crimes. In contrast, in more socioeconomically developed contexts, criminal convictions are negatively related to re-election (−1.943, p < 0.05). In these contexts, mayors involved in corruption are less likely to be re-elected. Both findings are consistent with H2.

With regard to the nature of the criminal conviction, taking human resource management as a benchmark, re-election is significantly negatively correlated to public works contracts in Models 2 and 3 (−0.584, p < 0.05 and −3.365, p < 0.05 respectively). In clusters A2 and A3, the chances of re-election of mayors engaged in corruption relating to public works contracts are lower, which is consistent with hypothesis H2.

5. Discussion and conclusions

The aim of this analysis was to offer insights into the re-election of corrupt mayors. It contributes to the literature by extending the analyses of the mitigating factors of the punishment of corrupt mayors within the corruption condoning hypothesis with two further factors: the relationship between the mayor and the citizens, and the level of the corruption. The two factors better focus the re-election of the mayor by taking into account the ability of mayors to gain support and the willingness of the citizens to offer their support, and the differing costs of corruption within a specific community.

This study provides considerable support for the research hypotheses. It shows that mayoral re-election is founded on local power and legitimacy within an exclusive relationship that is only in part contingent on local contexts. The factors that strengthen the link between the mayor and the citizens, and particularly the duration of the mayor’s political career and party membership, impacted positively on re-election in all the clusters. These results emphasise the legitimacy of leaders who, in spite of corruption, are recognised because they act with experience and independence, strengthening ties with the citizens and weakening those with political parties. However, trasformismo produces varying results, reducing the likelihood of re-election in highly developed contexts, while in the other two clusters, it is positively correlated to re-election. The negative beta for highly developed contexts concurrently with the positive betas for intermediate and less developed contexts are in line with the existing literature (Magone 2021) confirming that in the less developed contexts strong ties with the citizens, possibly strengthened with clientelistic measures, exist regardless of party switching on the part of the mayor.

An implicit exchange based on the mayor’s overall performance does not seem to strengthen the relationship between the mayor and the citizens, as our results are not significant in any of the contexts. This contrasts with the results for the non-corrupt mayor clusters, where performance variables have a positive impact on re-election, apart from less developed contexts where expenditure on public order has a positive impact on mayoral re-election. In these contexts, the ratio between the crime rate and population size is high, with public order perceived as a priority.

As for the low level of seriousness of corruption as a mitigating factor, we show that it is contextual to medium and highly developed communities, but not less developed ones. The difference in the beta sign between highly developed (negative) and less developed municipalities (positive) confirms our hypothesis. In highly developed contexts, cases of petty corruption (subject to low-level of sanctions) are overlooked, whereas cases of serious corruption (criminal affiliations and corruption in public works contracts) seem to constrain re-election. The stability of the criminal affiliation, in terms of resources, the number of actors involved, and continuity over time, leads to serious cases of corruption with extremely high social costs and high visibility. In more highly educated, high-income communities with a strong awareness of the public good, more stringent control is exerted over the more serious cases of corruption probably due to an awareness of the consequences, greater respect for the welfare of the community, and a greater ability to monitor the actions of the mayor. Similarly, public works constrain re-election: high visibility and seriousness reflect the fact that they are material works representing a significant part of the duties discharged by the mayor. In contrast, in less developed municipalities, both the stability of the criminal affiliation and the serious nature of the crime are positively related to the re-election of corrupt mayors. Where there is no space for the seriousness of the corruption to be a mitigating factor, there seems to be an implicit trade-off. In these municipalities, corrupt mayors are better able to use the profits from the corruption to cultivate stable relationships with the citizens, that pay off in electoral terms in spite of the type of corruption (Benito et al. 2019).

This analysis has three main limitations. First, electoral accountability is only one of the many mechanisms that helps prevent mayoral corruption. The justice system has a key role to play in preventing the re-election of corrupt mayors. However, as an institutional issue, it is often considered as a macro variable, whereas this study focuses on micro variables, assessing their impact at local level. Although many studies have underlined territorial divergence in terms of efficiency, in Italy the judicial system is inefficient compared to other European systems. The data collection method in this study precluded a more dynamic investigation, since only corruption cases involving mayors were considered and crimes in which a corrupt mayor was engaged prior to being elected as mayor, in another political role, were not part of this study. However, comprehensive data including crimes committed over the political career would have entailed considerable extra research and adequate data would not have been available. Second, we believe that our overview of corruption reveals its objective features as reported in the cases, rather than relying on questionnaires for framing corruption according to less reliable perception indices. Third, another limitation is that we focused on corruption specifically involving Italian mayors. The findings may not be generalisable to other illegal or unethical actions or to other jurisdictions. However, our research design could aid and contextualise further studies on related activities in other countries.

Our results have two theoretical implications. First, when considering different contexts, explanations for the re-election of corrupt politicians cannot be generalised. The low level of seriousness results in the condoning of corruption only in highly developed communities, not in less developed contexts. However, in less developed contexts, the implicit trade-off hypothesis seems to be confirmed. Implicit trade-offs relating to the positive performance of mayors were not substantiated in this study. Above all the relationship between the mayor and the citizens seems to be more homogeneously supported. Our results show that representation seems to rely more on the principle of similarity than responsibility, with the mayor defending the identity of the community and to a lesser extent its interests. The re-election of corrupt mayors seems to be a case of the community protecting the mayor’s interests, not vice versa. This could explain why the corrupt use of power is enabled by the sub-optimal use of democracy. Electoral defeat in serious corruption cases seems to be the exception, though this is not the case in less developed communities where the relationship seems to be stable even in negative circumstances. The second theoretical implication concerns the distortion of the relationship between mayors and the citizens, insofar as the interests of the electorate are not properly safeguarded, as an undervalued effect. Such distortion can be absolute, with voters totally under the influence of the mayor: these voters ‘protect without being protected’, re-electing corrupt mayors regardless of the severity of the corruption. Alternatively, it can manifest itself in a less severe way with voters who, in the event of corruption, break the link with the mayor and vote for another candidate.

Finally, the study has policy implications. A contextualised understanding of the re-election of corrupt mayors may help in some way to understand the limitations of current anti-corruption strategies. Relying on generalisations about political agents and corruption will inevitably lead policy-makers to adopt anti-corruption strategies that are not conducive to preventing or combatting corruption. Examining the specific relationship between the mayor and the citizens and its capacity to produce undesirable effects would arguably identify more specific and efficient solutions.

Disclosure statement

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

Additional information

Notes on contributors

Roberta Troisi

RobertaTroisi is an associated professor of Organization and Management. Her research interests lie in the area of organizational corruption both through theoretical and empirical analysis. She has collaborated actively with researchers in several other disciplines of social science, particularly economics and political science. She is the director of the monitoring centre of the territorial development at University of Salerno (ITA). She is member of the workgroup “fight against crime and terrorism” Agency for the support to the European research (APRE) aimed at defining the national suggestions on Annex 3, Horizon Europe.

Gaetano Alfano

GaetanoAlfano is a PhD Student in Political Sciences at the University of Salerno. He received his master degree in Civil Engineering from the University of Salerno in 2016. In the last year he collaborated as research assistant at the monitoring centre of the territorial development, University of Salerno (ITA).

Notes

1. http://www.italgiure.giustizia.it/sncass/.

2. In Italy unsuccessful candidates in municipal mayoral elections are automatically elected as councillors (Legislative Decree 267/2000). These candidates were classified as not re-elected.

3. Clientelistic networks cannot be controlled as they are occluded. In the discussion we argue that they were likely to occur in certain circumstances.