Introduction

Despite the intensive involvement of the European Union (EU) in facilitating the transformation of its post-communist neighbours, Eastern European transitions continue to be plagued by clientelism and problems with the fight against corruption. In particular, the blurry demarcation between formal institutions and personal networks remains embedded in post-communist contexts (Aliyev 2015). Moreover, citizens continue to often support openly malfeasant officials electorally and in daily life.

The predominant EU strategy for triggering change is via institution-building and training relevant officials at the national level. However, this institution-focused approach misses context-specific normative local understandings (Belloni and Strazzari 2014) relevant to the investigation of indirect impacts of EU exposure. As a result of the national-level emphasis of EU intervention, the existing literature mainly examines the link between exposure to EU training/work during enlargement and state-level public officials’ attitudes (Meyer-Sahling and Van Stolk 2015), with very few investigating this at the municipal level where EU exposure is more diffuse (Fagan and Sircar 2020).

The current study seeks to address this gap by gauging normative understandings of corruption and informality by municipal officials, and by examining whether these attitudes are affected by exposure to involvement in EU-compliant policy implementation. The following two questions are explored below:

  1. (1)

    Do municipal officials view informal payments differently, depending on whether they are involved in EU-compliant policy implementation?

  2. (2)

    Do levels of support by municipal officials for corrupt political leaders during periods of economic prosperity differ between those involved in EU-compliant policy implementation versus others?

This article thus complements the primary focus of the existing literature on Europeanisation and Eastern enlargement by concentrating on the possible indirect impact of the EU (Börzel and Risse 2012a) on local-level behaviours, instead of the direct intensive state-level processes of socialisation (Checkel 2005). The type of EU influence explored here is even more indirect, since the article examines whether involvement in implementing EU-compliant policies in the environmental sector has a knock-on effect on more general attitudes towards corruption and informal payments. In short, the article examines whether involvement in implementing policies—which officials acknowledge are compliant with EU standards—triggers emulation of attitudes towards corruption and informality.

To answer the research questions, an original survey of Serbian municipal-level officials in the environmental sector was conducted. The questionnaire included a previously used vignette-based survey experiment on ‘corruption voting’ (Klašnja and Tucker 2013) to estimate whether, on average, respondents’ electoral punishment of corrupt officials would be attenuated in times of economic prosperity. The aim is not to gauge electoral behaviour, but rather to use voting intention as a proxy for estimating the extent to which respondents are willing to tolerate corruption around them. The analysis will also examine whether respondents exposed to EU-compliant policy implementation viewed informal payments differently from other respondents.

The article uses the case of Serbia, because it represents a particularly difficult challenge for tackling corruption attitudes and where informal clientelist practices are especially deeply embedded (Babović et al. 2016). Serbia thus provides an especially challenging or ‘crucial case’ comparatively (Eckstein 2009) of the potential impact of indirect influence via EU-compliant policy implementation on municipal officials’ corruption attitudes. A recent European Commission (EC) Progress Report concluded that though institutional and legal arrangements are ‘broadly in place’ (EC 2022, p. 33), there is still no anti-corruption strategy and action plan to ‘thoroughly address corruption’ (EC 2022, p. 28).

This article will look specifically at the environmental sector, which is an important optic in investigating the negative impacts of sub-optimal governance. Environmental policies can be beset with issues related to corruption which can lead to the destruction of natural habitats, degradation, and severe negative impacts on local communities (Nemesio 2015, pp. 327–328). Although the problem of corruption by Serbian officials in healthcare or law enforcement is more acute (UNODC 2011), studying municipal environmental officials also provides important insights due to officials’ involvement in issuing environmental permits and environmental impact assessments, and the potential for clientelism in these processes (Fagan and Sircar 2015a).

The remainder of the article will be divided into four sections. In the next section, the trade-off between political corruption and economic voting is explored, how this is different in post-communist versus Western European countries, and how this trade-off might be impacted in the former by EU exposure through involvement in policy implementation. The review of these literatures informs the research hypotheses. In the third section, details are provided about the survey of Serbian municipal environmental officials. The fourth section presents the results of the two parts of the study (corresponding to the two research questions): the association between EU exposure and attitudes towards informal payments in exchange for public services; and a survey experiment conducted separately for those exposed to EU-compliant policy implementation and those who were not, to examine whether they trade-off elite corruption and competence differently. The final section discusses the results in a broader context, particularly about the impact of (weak) EU exposure on corruption attitudes at the municipal level in enlargement countries.

Political support, corruption, and indirect EU influence

The analysis below brings together the insights from the literatures on economic voting, attitudes towards political corruption, and indirect EU influence.

Lewis-Beck and Nadeau (2011, p. 288) summarise the central tenet of economic voting theory as follows: ‘in good economic times voters reward the incumbent, in bad economic times they punish the incumbent’.

An important debate in the economic voting literature is whether an individual’s personal recent experience of economic prosperity or general economic conditions affects voting behaviour. Kinder and Kiewiet (1979) found that voters’ choices for US Congressional elections were associated more strongly with feelings about general business conditions rather than personal experience with unemployment. Generally, voter behaviour is driven more by perceptions of collective (‘sociotropic’) conditions rather than personal (‘pocketbook’) experience, with little comparative evidence to the contrary (Lewis-Beck and Stegmaier 2013). The prevailing intuitions regarding economic vote theory mainly draw on advanced democracies, but the relationship between politics and the economy is qualitatively different for post-communist countries. Economic voting behaviour seems to gradually develop (Bochsler and Hanni 2019), perhaps due to citizens willing to endure difficult reforms, blaming economic hardship on the previous regime, or slow development of informed democratic citizens (Duch 2001).

Like economic voting, the literature on political accountability shows that across the world, perceived corruption is also strongly associated with personal electoral punishment (Cain et al. 1990; Ferraz and Finan 2008). The examination of the relationship between political malfeasance and electoral support can be thought of as ‘corruption voting’ (analogous with economic voting as outlined above), where personal experience (‘pocketbook corruption voting’) and general perceptions (‘sociotropic corruption voting’) form two distinct channels affecting voting behaviour (Klašnja et al. 2016). The term corruption voting is used below to describe the association between corruption and electoral support. Roberts (2008) suggested that incumbents in post-communist Europe are electorally punished in response to widespread political corruption, though there is insufficient comparative data to explore the thesis. It is thus instructive to focus on the relationship between corruption and political support in the region (Jastramskis et al. 2021).

Though there is some empirical support for ‘corruption voting’ in comparative contexts, an enduring puzzle in political behaviour is why voters retain corrupt elected officials. One possibility is that voters in emerging democracies like in Eastern Europe become disaffected and abstain rather than engage electorally (Kostadinova and Kmetty 2019; Kostadinova 2009). Rundquist et al. (1977) offer three broad reasons why voters sometimes continue to support corrupt political elites: lack of information; explicit exchange (that is, direct individual financial payoffs); and implicit exchange. In this article, economic performance will be the form of implicit exchange that mediates corruption voting (Jastramskis et al. 2021).

Winters and Weitz-Shapiro (2013) explore two potential explanations for continued support for corrupt officials, which they call the information hypothesis and trade-off hypothesis. The information hypothesis is that voters continue to support corrupt politicians, since they are unaware of their misdeeds, that is, the lack of information in Rundquist et al. (1977). By contrast, and relevant for this article, the trade-off hypothesis is that voters overlook corruption as elected officials provide in other ways, such as delivering public goods or economic growth, or ideological alignment, as forms of implicit exchange. Winters and Weitz-Shapiro (2013) focus specifically on completing public works as a mark of competence and conclude that there is little support for the trade-off hypothesis. On the other hand, voters in Spain were willing to overlook corruption in exchange for economic side benefits during the housing boom (Fernández-Vázquez et al. 2016). Zechmeister and Zizumbo-Colunga (2013) concluded that both personal and collective economic conditions significantly mediated the corruption vote in the US. Relevant to this article, Jastramskis et al. (2021) find that corruption voting is attenuated when unemployment is lower in post-communist Central and Eastern European countries.

It is clear from a survey of the literature that there is considerable variation in findings depending on the country or countries under study. Different places, unsurprisingly, may have different norms governing social, economic, and political relationships. It is hence insufficient to focus solely on institutional approaches but crucially to also examine the attitudes of public officials who work within these institutions. Given the substantial power of informal linkages between political elites and other citizens in the Western Balkans and other parts of Eastern Europe, it is also important to examine local understandings of social exchange (Torsello 2014) in the current study.

Norms around social exchange tend to crystallise at a formative stage and tend to be determined by local practices rather than externally driven attempts for change. In this article, we focus on Europeanisation, that is, the influence of the EU to facilitate changes in political institutions, policies and discourses through processes of ‘construction, diffusion, and institutionalization’ (Radaelli 2003, p. 30). Europeanisation is often seen as a form of socialisation—‘a process of inducting actors into the norms and rules of a given community’ (Checkel 2005). Checkel (2005) identifies three mechanisms to induce international (including EU) socialisation: strategic calculation, role-playing, and normative suasion. The first mechanism refers to rational behaviours based on costs and benefits—or external incentives (Sedelmeier and Schimmelfennig 2005)—without internalising norms and rules. Role-playing occurs when contacts between agents are intense and long-standing, leading to recipients acting in appropriate ways without necessarily internalising norms and rules. Normative suasion is the culmination of international socialisation, and recipients fully internalise (that is, take for granted) the socialiser’s norms and rules without prolonged contact. Strategic calculation emanates from a logic of consequences, while role-playing and normative suasion assume a logic of appropriateness (following March and Olsen 1998).

These mechanisms pre-suppose some form of direct contact between the EU and relevant actors in recipient (enlargement) countries. EU intervention during integration processes—for example, through conditionality, negotiations, and technical assistance—tends to primarily target state-level elected officials and civil servants in the candidate country. Indeed, existing research most often sees Europeanisation as the direct influence of the EU on domestic structures (Börzel and Risse 2007). In addition to socialisation, Börzel and Risse (2012a) outline other mechanisms of EU direct influence, including coercion, conditionality, capacity-building, and persuasion.

Hence, the EU attempts to influence the rule of law in enlargement countries in the Western Balkans almost entirely through direct interactions with state-level actors (Fagan and Sircar 2015b), namely forms of leverage (conditionality) and linkage (political dialogue and assistance) (Way and Levitsky 2007). However, what is neglected in studying EU integration in the Western Balkans is potential indirect EU influence in contexts where there is little or no contact from the EU, such as the day-to-day policy implementation work by municipal officials. Börzel and Risse (2012a) identify three types of potential influence in the absence of direct EU contact, all of which are forms of ‘deliberate emulation’ (Lavenex and Uçarer 2004): competition, lesson-drawing, and mimicry. Individuals may compete with each other to succeed on particular performance indicators. In the absence of such indicators, individuals may still draw lessons from EU best practices for improvement. Alternatively, individuals may mimic EU practices, since it is seen as the right thing to do. The latter two forms of emulation differ since lesson-drawing relies on rationality and mimicry on appropriateness (Börzel and Risse 2012b). This article complements studies of direct EU influence by investigating whether there are impacts where there is little direct EU contact and hence indirect EU influence via emulation among municipal officials. In other words, the mechanism at play is whether municipal officials involved in implementing EU-compliant policy implementation begin to emulate attitudes around corruption and informality differently from those not exposed to this type of policy work.

It can be argued that the fight against corruption and bolstering the rule of law are not good tests for EU influence, since local stakeholders sometimes identify other international actors—such as USAID and OSCE—as more influential (Fagan and Sircar 2015b). However, EU integration processes have helped align non-European donors so that they play a complementary role in EU-driven reforms of the rule of law (Mendelski 2015, p. 330).

The lead role of the EU has been further reinforced by the most recent phase of enlargement policy, during which programme funding and conditionality ‘have emphasised the rule of law and the control of corruption as main aims’ (Bartlett 2021, pp. 193–194). The European Commission (EC) ‘new approach’ to enlargement rests on the prioritisation of the rule of law, opening the relevant accession chapters (23 and 24) early and closing them towards the end of negotiations to ensure a clear track record of progress (EC 2011). The EU approach to enlargement was further clarified more recently as ‘fundamentals first’, which prioritises the rule of law, along with economic and public administration reform (EC 2020). The EU is the largest donor overall in the Western Balkans, with €700 million dedicated to the rule of law in the second round of the EU Instrument for Pre-Accession (IPA II) 2014–2020 (European Court of Auditors [ECA] 2022). Although the rule of law is a wide concept, the EC recognises that it is ‘intrinsically linked’ to efforts tackling corruption (Council of the EU 2019).

More broadly, the EU frames itself as a ‘normative power’ (Manners 2002) in areas related to the rule of law—including the fight against corruption—during EU enlargement processes and further afield. Articulating its external relations, the EU aims to promote ‘political reform, rule of law, economic convergence and good neighbourly relations in the Western Balkans and Turkey’ (EEAS 2017, p. 24), and ‘pursue tailor-made policies to support inclusive and accountable governance, critical for the fight against terrorism, corruption and organised crime’ globally (EEAS 2017, pp. 25–26). Along these lines, a strategic evaluation of international EU anti-corruption efforts recommended that the Union should take a ‘proactive, global leadership role, engaging as an advocate for the Rule of Law and the fight against corruption’ (EC et al. 2022, p. 105).

Thus, investigating the connection between potential EU indirect influence and municipal public officials’ corruption attitudes is an important addition to existing research, since it resonates with the EU’s recent prioritisation of the rule of law (including tackling corruption) in its enlargement policy, and its aim to be the global leader in promoting anti-corruption practices.

If involvement in EU-compliant policy implementation has indirect influence, then it is expected that there would be differences in attitudes towards informality between those who have been exposed via policy work those who have not. In particular, in line with Western norms around financial and non-financial informal payments:

H1

Municipal officials involved in EU-compliant policy implementation will be less accepting than other respondents regarding informal payments.

The second investigation in this article applies the experimental approach of Klašnja and Tucker (2013) and explores one of their key findings. The authors conducted two separate vignette-based experiments on citizens from a low-corruption (Sweden) and a high-corruption (Moldova) context. The scenarios contained hypothetical instances of positive and negative depictions of sociotropic or pocketbook corruption, against the backdrop of economic prosperity or recession. Thus, there were eight randomly assigned vignettes in the two countries. Perhaps the most surprising result is that economic prosperity completely mediates sociotropic (but not pocketbook) corruption voting in the high-corruption context (Moldova). That is, there is no significant difference in political support between a corrupt and non-corrupt Mayor during a period of economic boom in Moldova. By contrast, there is strong evidence of sociotropic corruption voting regardless of the state of the economy in Sweden. This leads to the second area of enquiry in this study, whether those involved in EU-compliant policy implementation act differently from others in a high-corruption context.

In other words:

H2

Municipal officials involved in EU-compliant policy implementation will support corrupt politicians less than non-corrupt ones in periods of economic prosperity. Other respondents will not support corrupt politicians less than non-corrupt ones in periods of economic prosperity. Since this hypothesis is framed in terms of general economic conditions, it focuses on sociotropic corruption voting.

The current study differs from Klašnja and Tucker (2013) in two important respects. First, instead of examining how corruption voting differs between low-corruption and high-corruption settings, what is investigated here is the potential difference between groups within a high-corruption case based on whether they have been influenced indirectly by the EU via involvement in policy implementation.

Second, and crucially, the respondents in the survey experiment are municipal officials instead of the broader citizenry. Although this greatly restricts the potential sample, the attitudes of the respondents might have a more profound and explicit effect on the distribution of resources and local governance more generally. Thus, there is a lot at stake if they are ambivalent towards corruption or clientelism. This article does not seek to measure whether the respondents are corrupt, but rather examine their attitudes in the face of informal payments and hypothetical malfeasance.

There has been a recent ‘explosion’ in research using experimental methodology on public officials, with over 50 such studies on policy responsiveness published 2011–2017 (Costa 2017). This is because experimental methods offer the possibility of interpreting results in a causal manner (Grose 2014). However, an experimental approach is not without drawbacks. First, the pool of potential respondents is far smaller than in research on the overall population. Even with reasonable response rates, sample sizes are relatively low, so the statistical power may be insufficient to detect an effect. Second, there are ethical concerns related to using public officials as subjects. On the one hand, McClendon (2012) argues that as officials carrying out public duties do not have the same protections regarding confidentiality since finding out their conduct is of public interest. On the other hand, this permissive approach may have consequences working with local officials, where their involvement in the research may lead to an erosion of trust from citizens and affect respondents’ livelihoods. These methodological and ethical trade-offs need to be kept in mind when conducting experimental research on public officials.

The following section outlines how the data were collected for this research.

Data

The analysis below is based on survey data collected from municipal Serbian officials responsible for environmental protection. Due to GDPR restrictions, the researchers were unable to access a contact list. However, the Serbian Standing Committee for Cities and Municipalities (SKGO) agreed to distribute the link to the web-based questionnaire to relevant officials via email. The survey questions were translated from English into Serbian and then administered via a web-based questionnaire.

For the survey experiment, respondents were divided into blocks depending on their self-reported involvement in EU-compliant policy implementation. The officials were asked whether their work at the city/municipality involve implementation of EU policies, with those answering ‘yes’ assigned to one block and the rest to the other block. Within each block, respondents were randomly assigned one of two vignettes (see Table 1). In both vignettes, the Mayor of a hypothetical medium-sized Serbian town has overseen economic growth. In Vignette A, the Mayor condones corruption within the city administration, while he cracks down on malfeasance by his officials in Vignette B. Respondents were asked how likely the protagonist (‘Marko’) would be to vote for the incumbent Mayor (using a 1–7 scale with higher scores indicating greater likelihood). The experimental manipulation thus examines respondents’ attitudes towards sociotropic corruption voting.

Table 1 Text of vignettes (differences italicised)
Full size table

Following questions similar to a previous survey of Serbian citizens on corruption (UNODC 2011), respondents were also asked how acceptable it was to give money or gifts to doctors or teachers. In addition to these questions, the following background items were included: gender; year started working at the city/municipality; whether net monthly salary exceeded 10,000 RSD; education; and ethnicity.

SKGO sent invitation email messages on its distribution lists of municipal environmental officers, which included a link to the web-based questionnaire, in April 2019. SKGO sent follow-up messages in May 2019 and September 2019. Of the 174 cities, municipalities, and city-municipalities, 90 responses were obtained. The sample was further reduced to responses where the survey experiment and question about involvement in EU-compliant policy implementation were non-missing. This resulted in a final sample size of 82 (which corresponds to a 47% response rate).

After the initial data collection, the education and ethnicity variables were dropped, since a large majority of the respondents were Serb (85%) and reported the highest category of education (95%), and thus, there was little variation in the sample for these variables.

Results

Nearly two thirds of the respondents (63.4%) reported that they were involved in EU-compliant policy implementation as part of their work in the city/municipal administration. The sample was also predominantly female (77.6%). Those involved in the implementation of EU policies had worked in their respective municipalities approximately 2.7 years longer on average than other respondents.

As outlined above, it is important to examine the patterns of social exchange norms among public officials. Although the relevant questions in the current study draw on UNODC (2011), the questionnaire extends the earlier study by separating money from other gifts, since these may be viewed differently by respondents. The results are presented in Table 2. Overall, the practice seen as least acceptable across both groups is when a doctor requests money or gifts in exchange for a service. On the other hand, offering a gift or money to a doctor tended to be condoned more.

Table 2 Acceptability of gifts or money for doctors or teachers (%)
Full size table

The options in the UNODC survey are not exactly the same as in the current study, but the earlier survey of Serbian citizens is broadly consistent with the responses from the municipal officials. However, municipal officials on the whole appear to condone these types of informal payments more than the Serbian citizenry. Approximately 70% of Serbian citizens felt that offering money/gifts to a doctor was ‘not acceptable’, whereas around 40% among municipal officials deemed it ‘never acceptable’. Doctors proactively requesting money or gifts was seen highly unfavourably by Serbian citizens, with around 85% finding it unacceptable, compared with 76% of the municipal officials. The Serbian population seemed to be particularly stringent regarding money or gifts for teachers, with over 90% saying it was not acceptable, compared with only 58% of municipal officials. However, if the responses for ‘rarely acceptable’ are included for municipal officials (which is not one of the options in the UNODC study), then 83% feel that gifts/money are never or rarely acceptable.

There are two further observations to note. The first is that, unlike in the UNODC report, giving or requesting gifts is separated from money in the current study. The percentage of municipal officials saying that doctors asking for money is unacceptable is similar to that for gifts (75–77%). However, there is a more substantial difference for giving gifts or money. For doctors, 60% of the respondents felt that it was never acceptable to offer money, compared with 21% giving the same response for giving gifts. When it comes to teachers, 78% of the municipal officials responded that giving money was never acceptable, compared with 38% for gifts. As noted above, however, there was the same percentage of respondents (38%) that felt that giving gifts to teachers was rarely acceptable. If these categories are combined, then 91% of the officials surveyed deemed offering money to a teacher never or rarely acceptable versus 75% for giving gifts instead.

As the item on offering gifts to teachers shows, the responses are not always all-or-nothing, or in other words, respondents also use the intermediate response categories. This is related to a second observation regarding the results in Table 2. Looking separately at those exposed to EU-compliant policy implementation versus other respondents for each of the questions, municipal officials in the sample whose work involves EU-compliant policy implementation tended to have more nuanced attitudes towards money and gifts for doctors and teachers. That is, with the exception of offering a gift to the doctor, those with exposure to EU-compliant policy implementation were more likely to say that these behaviours were ‘sometimes acceptable’ compared with the rest of the sample. The inverse is also true: those involved in implementing EU policies were less likely to deem the request or offer of gifts/money as ‘never acceptable’.

The reason for this is not entirely clear and is discussed in the conclusion. The differences between those exposed to EU-compliant policy implementation and others may rest on other assumptions they had regarding the behaviours mentioned not in the text of the survey questions. For example, were the offers of gift/money before or after a successful surgery or at the end of the school year? If it is afterwards, then these exchanges could indicate gratitude. However, if the offer or request for some informal compensation occurs beforehand, then the teacher or doctor may have more corrupt motives.

In sum, the questions on the different forms and behaviours around informal payment showed that municipal officials tended to condone gift-giving more than money-giving, and offering such compensation rather than being asked. Compared with the sample of Serbian citizens from UNODC (2011), municipal officials were less likely to respond that these behaviours were always unacceptable. Moreover, officials who implement EU-compliant policies as part of their work were more likely to envision scenarios where it was sometimes acceptable to give gifts or money.

Turning to the other focus of this study, an independent samples t test (with equal variances assumed) was conducted on the data from the survey experiment. Recall that respondents within each block (EU exposure versus others) were randomly presented with one of two vignettes, with the Mayor cracking down on corrupt local officials in Vignette B but not in Vignette A. The difference in the mean scores is thus an estimate of the average effect of sociotropic corruption voting at the local level when the Mayor removes corrupt officials in the local administration (using a scale ranging between 1 and 7). Looking at the results of the t test in Table 3, both blocks show statistically significant mean effects. For municipal officials involved in EU-compliant policy implementation and for other respondents, the average score for Vignette A was 3.292 and 3.733, respectively. In both blocks, respondents viewed Vignette B significantly more favourably on average than the other vignette. The mean difference is 1.923 points for those involved in EU-compliant policy implementation in their work, which is 0.2 more than for the other respondents.

Table 3 Independent samples t test by vignette and block
Full size table

The t test results show that likely political support for the Mayor increases significantly within each block when he fires the corrupt officials, and that this is a substantial difference of nearly two points on the seven-point scale. However, it is not immediately evident whether the mean effect is significantly higher for those involved in EU-compliant policy implementation compared with others. To estimate this, the analysis was conducted using OLS regression, with the score on the likelihood of Marko voting for the Mayor as the outcome variable (on a scale ranging between 1 and 7). The regression model is:

$$ {\text{score }} = \alpha + \beta _1 ^* {\text{EU}} + \beta _2 ^* {\text{vignette}} + \beta _3 ^* {\text{EU}}^* {\text{vignette }} + {\varvec{\beta}} _{\mathbf{x}} ^* {\mathbf{X}} $$
(1)

where EU is whether the work of the respondent involved EU-compliant policy implementation (1 = yes, 0 = no), vignette is the version of the scenario that was presented (1 = B, 0 = A), and X represents the control variables. Initially, the models were fitted without the demographic controls. Looking at the regression coefficients, the estimate for the expected difference in the score between the vignettes for those involved in EU-compliant policy implementation is β2 + β3, and is β2 for the rest of the respondents. The results of the regression analysis are presented in Table 4.

Table 4 OLS regression models
Full size table

The results in Model 1 are equivalent to those in the t test. The expected effect of positive sociotropic corruption behaviour by the Mayor is 1.733 points for respondents other than those involved in EU-compliant policy implementation, while it is 1.733 + 0.189 or 1.922 points for those involved in EU work. Importantly, the interaction term represents how much more or less the mean effect of positive sociotropic corruption behaviour is for those who have been exposed to EU-compliant policy implementation. In Model 1, the coefficient estimate is 0.189, that is, those exposed to EU work react more favourably to positive sociotropic corruption behaviour by 0.189 points on average compared with other respondents. However, the coefficient is not statistically significant.

The control variables that were added are: an indicator for female (1 = female, 0 = male); whether the respondent’s net monthly income exceeds 10,000 RSD (1 = yes, 0 = no); and the year the respondent started working in the city/municipal administration. Some of these items were missing for the respondents, so the sample was first restricted to cases where all of the demographic items had been answered. This complete-case analysis reduced the sample size to 68.

Model 2 is the same regression model, but using the restricted sample. Although the coefficient estimates change slightly, the findings remain the same. The expected effect for respondents other than those implementing EU policies is 1.599 points, and is 0.544 points higher on average for those exposed to EU work. However, again, the interaction term is not significant. When the controls are included for gender, years of experience, and salary (Model 3), the estimated coefficient remains statistically significant and increases to 1.794. The interaction term changes to 0.444 and is still not significant. In the complete-case analysis, the values of the demographic variables are quite similar between respondents involved in EU-compliant policy implementation and others (see Table 5).

Table 5 Demographic characteristics by block for complete cases (n = 68)
Full size table

Thus, across the regression models, the main finding is that there is a significant positive average effect related to positive sociotropic corruption behaviour by the hypothetical Mayor. In other words, for the municipal officials surveyed, average political support was significantly higher for the Mayor who cracked down on malfeasance, despite the improved economic situation in the municipality. The estimated differences were 1.6–1.9 points on a seven-point scale, so were quite substantial as well. Moreover, this effect was more pronounced for those whose work involved EU-compliant policy implementation, but the difference was not statistically significant.

Discussion and conclusions

These results show the limited indirect influence of the EU on the emulation of attitudes towards corruption and informality, and they contrast with both of the initial hypotheses. Individuals involved in EU-compliant policy implementation would be expected to favour informal payments less and have a far sharper notion regarding their unacceptability. For example, low-corruption settings like Sweden are quite negative towards any evidence of corruption compared with a high-corruption case like Moldova (Klašnja and Tucker 2013). However, contra H1, those who are involved in EU-compliant policy implementation have more nuanced attitudes towards informal payments for doctors and teachers compared with other respondents. An additional area for future investigation is how money and gifts are viewed differently—in particular, money is seen as a less acceptable informal payment across respondents.

One possible explanation for these results is that money and gifts, especially for health care services, are seen as an act of gratitude in post-socialist settings rather than a bribe (Grødeland 2013; Stepurko et al. 2013). This grey area between gratitude and corruption is not unique in post-communist Europe. Although the euphemisms used to describe informal payments perhaps displays an ‘aesthetics of deception’ (Makovicky and Henig 2018, p. 126), local understandings are not universally clear-cut. In post-communist Europe, while in Slovakia the term pozornosť (‘attention’) is ambiguous (Školkay 2018), aploksne (‘envelope’) only has negative connotations in Latvia (Kažoka and Kalnins 2018). The Persian term baksheesh sits uncomfortably between ‘tip’, ‘alms’, and ‘bribe’, and is found in the Middle East, South Asia, and in the Balkans (McLeod-Hatch 2018). Thus, the norms around informal payments can be ambiguous not only in the Western Balkans but elsewhere as well. Hence, it is perhaps less surprising that we find little evidence of emulation around informal payments, which requires future research.

Turning to the other investigation in this article, the aforementioned results by Klašnja and Tucker (2013) showed that in a high-corruption setting, during times of economic prosperity, political corruption does not significantly alter political support. By contrast, there is a strong corruption voting effect in a low-corruption setting. If indirect EU influence impacts sociotropic corruption voting in Serbia, it would be expected that those involved in EU-compliant policy implementation will punish corrupt officials despite increased economic prosperity, while other respondents would not. Interestingly, these conjectures in H2 are only partially realised. Indeed, those socialised through EU-compliant policy implementation displayed significant sociotropic corruption voting behaviour despite economic prosperity. However, this is also true for those not involved in EU-compliant policy implementation. Moreover, although those subjected to indirect EU influence show more evidence of corruption voting, the difference between the two groups is not statistically significant. In other words, differently from the results from Moldova, environmental municipal officials in Serbia withdraw support from corrupt officials on average, irrespective of exposure to EU-compliant policy implementation.

Indirect EU influence through policy implementation thus does not appear to significantly impact corruption attitudes among Serbian municipal officials in the environmental sector. The result is perhaps not surprising, given that the mode of EU influence studied is indirect and very weak. This is not to say that the EU could not exert indirect influence in other policy sectors at the municipal level. Emulating the EU is seen as the driving force for regional integration and cooperation in other parts of the world (Lenz 2012). Future research could focus on EU assistance programmes for sectoral development and regional cooperation (that may operate on the local level), which would provide an example of a stronger channel of EU influence.

On the other hand, any indirect influence may only exist alongside direct forms of EU influence (Börzel and Risse 2012b). At the municipal level, direct influence could take the form of training or tutelage, which has been shown to be pivotal in triggering change more generally in enlargement countries (Bomberg 2007; Papadimitriou and Phinnemore 2004). Yet more recent research showed no significant impact of EU training for national-level civil servants in new Member States in Central and Eastern Europe (Mayer-Sahling et al. 2016), and on corruption attitudes among Serbian municipal officials (Fagan and Sircar 2020).

On a more positive note, even in the high-corruption setting of Serbia, respondents not exposed to indirect EU influence also strongly sanction corrupt behaviour and do not look the other way when there are collective economic benefits. This finding drives to the heart of the underlying notions related to EU and other international influence, which seek to improve behaviours of ‘bad’ officials. However, officials’ attitudes towards corruption voting already seem in line with Western/EU standards. The question then becomes why, despite congruent norms, does Serbia experience high levels of corruption?

The answer could be due to domestic factors. Börzel and Risse (2012a) identify domestic structures as scope conditions for direct and indirect EU influence. It is these domestic factors, particularly the cost–benefit calculus for ruling elites, that explain variation in the reform of the rule of law in the Western Balkans (Noutcheva and Aydin-Düzgit 2012). Thus, explaining outcomes in the transformation of norms may involve the interplay between EU ‘norm diffusion’ and domestic ‘norm contestation’ in recipient countries (Dandashly and Noutcheva 2021).

Moreover, the discrepancy between revealed survey behaviours and realities on the ground is not unique to Serbia. Boas et al. (2019) found that Brazilian voters tended to punish malfeasant officials in hypothetical scenarios but not in real-life municipal elections. The authors conclude that the gap is due to the fact that individuals punish corruption for a hypothetical Mayor, but not when it is their own Mayor, especially if they have overseen a period of job creation and public services.

Another possible confounder is social desirability, and whether it is equally prevalent among both groups of respondents. Self-reported data as in the online surveys used are susceptible to social desirability bias (Neuberger 2016), and in particular, social desirability bias is accentuated when the behaviour surveyed is related to moral decision-making (Randall and Fernandes 1991). There are two reasons why social desirability may not be problematic in the current study. The first is that if respondents answer in socially desirable ways, particularly those exposed to indirect EU influence, then one would expect these respondents to consistently appear more critical of corrupt and clientelist behaviour. However, those involved in EU-compliant policy implementation have a less black-and-white attitude towards informal payments. Second, following Nederhof (1985), self-administered questionnaires with closed-ended items (as in the current study) are less vulnerable to social desirability.

As noted above, another drawback with survey experiments on public officials is that the sample is limited. First, the restricted sample also affects the statistical power of the tests used. Although those involved in EU-compliant policy implementation do not display significantly higher corruption voting behaviour than others, this could be that the mean effect is not large enough to be detected given the sample size. Second, the randomisation process estimates the mean effect size in the two sub-samples, not the overall population of environmental municipal officials. Given the data protection around the respondents, it would not be possible to gauge the representativeness of the sample from the population of municipal officials.

An avenue of further research, where there is evidence of significant emulation of EU norms, is to explore whether the form of emulation is lesson-drawing or mimicry. Börzel and Risse (2012b) concede that ‘lesson-drawing and mimicry are sometimes hard to distinguish in the real world’. To uncover the type of emulation, more detailed qualitative data not available in the current research would be needed in future studies to explain mechanisms of EU indirect influence.

Despite these caveats, the research has thrown up an interesting result: strong average anti-corruption attitudes by municipal officials in a high-corruption country, irrespective of EU exposure. This has implications for the assumptions of direct and indirect influence by the EU and other international actors not only in Serbia, but also international efforts to influence local governance and the rule of law elsewhere.