The Effects of Corruption on Satisfaction with Local and National Governments. Does Corruption ‘Grease the Wheels’?

Abstract

A causal association was found between higher levels of corruption and lower levels of satisfaction with local and national governments. However, interaction term analysis provides suggestive evidence that: first, as public service satisfaction increases, the effects of corruption change from negative values to positive values for both local and national governments; second, as political situations improve, the negative effects of corruption diminish for both local and national governments; and, third, as the economic situation improves, the negative effects of corruption diminish for local, but not for national governments.

Citizens’ evaluation of their government’s performance is associated with the degree of legitimacy and support that they extend to it (Lipset 1959; Weatherford 1992; Przeworski et al. 1993; Camacho 2014; Stockemer & Sundström 2013). Citizens tend to support their government if they view its performance as satisfactory. Conversely, they are likely to withdraw their support if they view the government’s performance as unsatisfactory. Thus, satisfaction with government performance has crucial implications with respect to peoples’ beliefs about the legitimacy of the state, its ruling politicians and political parties, as well as with respect to their faith in democracy as a system of governance (Weitz-Shapiro 2008; Harmel & Yeh 2011; Corral & Orcés 2013).

In turn, one of the most important factors affecting satisfaction with government performance is corruption. In well-established democracies, corruption is linked to a reduction in government support (Bailey & Paras 2006; Wagner et al. 2009). Importantly, perceptions regarding the honesty and fairness of government are found to be more important than other aspects of government performance such as efficiency and effectiveness (Putnam 1993; Pharr et al. 2000). Anderson and Tverdova found that corruption is one of the most important aspects of government performance used by people ‘to judge political institutions’ (Anderson & Tverdova 2003, p. 104). According to Holmes (2006), incidents of corruption have led to the collapse of governments in a number of established democracies.

However, no consensus has been reached regarding the effects of corruption in post-communist and developing countries. An important strand of the literature emphasises the negative effects of corruption by supporting the so-called ‘sands the wheels’ hypothesis. According to this idea, corruption undermines trust in institutions and increases tolerance of offering and accepting bribes in public institutions (Habibov et al. 2017). It has been both empirically and theoretically proven that corruption has an immensely negative effect on development around the world (Chitakunye et al. 2015). Several authors assert that corruption tends to discourage investment and economic growth, aggravate the problems of informal economic activities, create obstacles to economic and political reform, exacerbate the inequality between the rich and the poor, significantly reduce satisfaction with public services and, in the long run, cause considerable degradation of living standards (Justesen & Bjørnskov 2014; Habibov 2016; Habibov & Cheung 2016). In addition, corruption results in negative citizen perceptions of public service performance, thus reducing their overall level of trust in the government (Park & Blenkinsopp 2011). In post-communist countries, corruption is also linked to lower levels of support for the current regime (Grosjean 2011). Furthermore, lower levels of citizen satisfaction with the government undermine civil society, which tends to be fragile in post-communist countries, and increase the likelihood of engagement in informal economic activities (Wallace & Latcheva 2006; Zakaria 2013).

Following the ‘sands the wheels’ hypothesis, corruption should also have a negative effect on government satisfaction. Indeed, this has been reported by several studies that have focused directly on the link between corruption and government satisfaction. Thus, in South Korea, perceived corruption has led to the erosion of satisfaction with the government (Park & Blenkinsopp 2011). Studies of Latin American countries have reported similar findings (Hiskey & Seligson 2003; Weitz-Shapiro 2008). In China, corruption has been reported to have a strong negative effect on satisfaction with local governments, although the central government was found to be shielded from blame (Harmel & Yeh 2011).

Another important strand of the literature stresses the positive effects of corruption. According to this ‘grease the wheels’ hypothesis, corruption should be tolerated inasmuch as it provides an effective way of mitigating the distortions produced by ineffective bureaucratic procedures (Habibov et al. 2017). This hypothesis can be traced to the works of Leff (1964) and Huntington (1968), who advanced two arguments explaining the positive effects of corruption. First, corruption is a way of avoiding bureaucratic obstacles, especially in countries with burdensome and often changing regulations. Second, bribes provide an important incentive to bureaucrats to work, particularly in countries where public service is not well developed (Aidt 2003; Green 2011). Such competition may allow for a more efficient distribution of scarce resources.

The positive effects of corruption in terms of increased efficiency were markedly noticeable under the command-style economic system in communist countries (Holmes 2000). Management-level corruption increased the efficiency of central planning by mitigating its rigidness (Nye 1967). Additionally, well-developed forms of corruption were used successfully to alleviate shortages of goods caused by the inefficient distribution of goods and distorted prices (Levy 2007). High levels of corruption have led to demands for economic and political reforms in post-communist countries (Grosjean 2011).

Along the same lines, recent studies report that even if the effects of corruption are negative, the magnitude of the negative effect is lower in nations, such as in post-communist countries, where government institutions are weak (Anderson & Tverdova 2003; Méon & Weill 2010). Furthermore, the effects of corruption are less damaging if citizens are satisfied with the services they receive (Lavallée et al. 2008). Following the ‘grease the wheels’ hypothesis, it is not surprising that citizens in developing and post-communist countries support highly corrupt governments if they perceive that the services they receive are effective, or if they believe that the political and economic situation is improving (Manzetti & Wilson 2006; Weitz-Shapiro 2008; Winters & Weitz-Shapiro 2013).

Yet another perspective, articulated as the ‘culture of corruption’ hypothesis, suggests that corruption is heavily influenced by cultural norms (Miller et al. 2001; Fisman & Miguel 2007; Lee & Guven 2013). As such, a ‘culture of corruption’ can lead to the situation where corruption may be viewed as the norm rather than as a crime and, consequently, is widely accepted and tolerated (Smith 2008; Turex 2011). In post-communist countries, the levels of corruption inherited from the communist era increased during transition and came to be associated with insufficient remuneration for the government sector, ever-changing and burdensome regulations, and blurred lines between the private and public realms (Skaaning 2009; Libman & Obydenkova 2013). Because post-communist countries have high levels of petty corruption, such as having to pay or give gifts to workers in the public health and education sectors, such corruption is typically not viewed negatively and is generally accepted. Post-communist countries generally have higher levels of corruption in comparison with other countries with similar income levels (Suhrcke 2000; Bowser 2001). Since the ‘culture of corruption’ is seen as routine and even essential, it may not significantly affect the level of citizen satisfaction with the government.

Thus, in general, a variety of studies has focused directly on the link between corruption and government satisfaction in developing and post-communist countries. Some of these studies have yielded contradictory results, reporting that corruption has either a positive or negative effect on government satisfaction, while other studies report no effect at all. Furthermore, the majority of these studies used single-stage regression methods with cross-sectional data. Such an approach cannot identify whether corruption affects government satisfaction or vice versa, or whether a third unobservable variable causes simultaneous changes in government satisfaction and corruption. Also, most of these studies examined the effects of corruption on only one level of government (such as local), while the effects of corruption on national governments have not been thoroughly investigated.

This article intends to contribute to the literature in the following ways. First, we attempt to transcend the limitations of previous studies by using instrumental variable (IV) regression to address reverse causality and unobserved variable problems to highlight the causal association between corruption and satisfaction with government. We also include interaction terms in the estimated IV models to test directly how the marginal effects of corruption on satisfaction with government vary according to the extent of satisfaction with services, the economy and politics, as well as support for democracy. Second, we distinguish between the effects of corruption on satisfaction with government at both the local and national levels. To the best of our knowledge, this is one of the first studies to analyse the effects of corruption on government satisfaction using a wide and varied sample of 29 post-communist countries in Eastern Europe and the former Soviet Union.

By focusing our analysis on both corruption and satisfaction with the government, our article contributes to the newly emerging literature on the link between corruption and satisfaction with government (Manzetti & Wilson 2006; Weitz-Shapiro 2008; Park & Blenkinsopp 2011; Winters & Weitz-Shapiro 2013). In broader terms, it also contributes to the literature on popular preferences and satisfaction with government policies (Alesina & Fuchs-Schüendeln 2007; Dimitrova-Grajzl et al. 2012; Habibov 2015), as well as the determinants of public support for economic and political reform (Fidrmuc 2000; Kim & Pirttilä 2006; Grosjean 2011; Rovelli & Zaiceva 2013).

Method

Study sample

The data sets we used are from the second round of the Life in Transition Survey (LITS) conducted in 2010 by the European Bank for Reconstruction and Development (Steves 2011). Our study covers 29 post-communist countries: Albania, Armenia, Azerbaijan, Belarus, Bosnia & Hercegovina, Bulgaria, Croatia, the Czech Republic, Estonia, Georgia, Hungary, Kazakhstan, Kosovo, Kyrgyzstan, Latvia, Lithuania, Macedonia, Moldova, Mongolia, Montenegro, Poland, Romania, Russia, Serbia, Slovakia, Slovenia, Tajikistan, Ukraine and Uzbekistan. The LITS employs a two-stage clustered stratified design to select households for interview. In each country, the list of primary sample units (PSU) created by the respective national governments was employed to stratify each country according to its regions. Through the application of the probability proportion to size rule, approximately between 50 and 75 PSUs were chosen from the list for each country, depending on the size and population density of the country. Next, the random walk technique was used to chose approximately 20 households for interview in each PSU. One respondent from each household was then randomly selected to complete the survey. Approximately 1,000 respondents per country were interviewed by specially trained interviewers.

Outcome variables and predictor

Our focus was to test our hypotheses about the relationship between government satisfaction and corruption at both the local and national levels. Hence, there were two outcome variables: satisfaction with the local government; and satisfaction with the national government. The LITS measures the rating of government satisfaction according to a five-point ordinal scale where 1 is ‘very bad’ and 5 is ‘very good’.

Our predictor variable was corruption. Corruption was measured according to the answers provided by LITS respondents as to whether they had made an unofficial payment or gift (other than any legitimate fees that may have been applicable) while using public services in the last 12 months. Public services include traffic police, civil courts, the issue of official documents (such as passports and property deeds), public primary and secondary education, public vocational education, public healthcare, the provision of unemployment and other social security benefits. Summing up positive answers for each of eight services, we computed the composite index of corruption. The value of this index varied from 0 to 8, with higher values indicating higher degrees of corruption (Cornbach’s α = 0.85). The summary information of outcome, predictor and covariates can be found in Table 1.

TABLE 1 Descriptive Statistics

Variables	Description	Mean	Standard. deviation	Yes (%)	No (%)	Min	Max
Outcome variables
Satisfaction with local government	Rating of local government satisfaction on a five-point scale from ‘very bad’ as 1 to ‘very good’ as 5	3.042	0.959			1	5
Satisfaction with national government	Rating of national government satisfaction on a five-point scale from ‘very bad’ as 1 to ‘very good’ as 5	2.827	0.996			1	5
Predictor
Corruption	Index of corruption with unofficial payment or gift during interaction with public	0.366	0.800			0	8
Covariates
Public service satisfaction	Index of satisfaction with quality and efficiency of public services received	0.965	1.106			0	8
Economic situation improved	Respondent believes that economic situation is now better than around 4 years ago			23	77	0	1
Political situation improved	Respondent believes that political situation is now better than around 4 years ago			24	76	0	1
Support for democracy	Respondent believes that democracy is preferable to any other political system			47	53	0	1
Middle wealth household	Respondent lives in middle wealth 33.3% household in her/his country as measured by wealth index			34	66	0	1
Wealthiest household	Respondents lives in wealthiest 33.3% household in her/his country as measured by wealth index			30	70	0	1
Age	Age of respondent in years	45.474	17.492			17	99
Female	Female respondent			61	39	0	1
University	University educated respondent			20	80	0	1
Instrument
Index of ‘solicited bribe’	Paid a bribe because was asked to pay by the official	0.0964	0.412			0	6

Explanation and operationalisation of covariates influencing satisfaction with the government

One of the most important factors influencing satisfaction with the government is the satisfaction people have with the public services they received (Lavallée et al. 2008). To control for the potential effects of satisfaction with public services, we use an additive index that includes seven questions about the respondent’s satisfaction with the quality and efficiency of public services as outlined above. By adding up the responses ‘satisfied’ and ‘very satisfied’, we estimated a continuous variable, ‘public service satisfaction’, which varied from 0 to 8, where higher values indicated a higher degree of public service satisfaction.

We also controlled for socio-economic status, namely, the educational attainments of the respondent and the wealth status of the household. On the one hand, individuals with higher socio-economic status in terms of education and wealth may express higher levels of government satisfaction. At the same time, higher socio-economic status can be a proxy for higher expectations regarding government performance (Lewis & Pattinasarany 2009). Consequently, higher status leads to higher expectations. If, however, the expectations of individuals with a higher socio-economic status exceed actual government performance, then higher economic status will become associated with a lower level of government satisfaction.

In addition, we controlled for socio-demographic characteristics, namely, age and gender. The definitions and statistics of these covariates are described in Table 1 together with the outcome and predictor variables. Finally, to control for unobservable characteristics on the country and PSU levels, we included a set of country and PSU dummies for all the estimated models.

Analytical strategy

As noted in the introductory section, we used IV regression to address reverse causality and omit variable problems. The estimation of IV regression requires an instrument that is correlated with the predictor, corruption. At the same time, such an instrument should not affect the outcome variable, government satisfaction, other than through the predictor. Usefully for our study, the LITS asked individuals if they had paid bribes (our predictor variable) and then asked those who answered positively to clarify the reasons for doing so by choosing one of four options provided: (a) the respondent paid a bribe that was directly solicited by a public official; (b) the respondent was not directly asked to pay, but knew that an informal payment in this situation was expected; (c) the respondent was not directly asked to pay, but wanted to express gratitude; or (d) the respondent offered to pay to get things done.

We chose to use ‘solicited bribe’ as an instrument. This instrument was defined as a binary variable that equalled 1 if the respondent had paid a bribe because he or she had been directly solicited by a public official, and equalled 0 if the respondent initiated the bribe, or had not paid a bribe at all. We selected this instrument specifically because it was unlikely to have an effect on the outcome variables other than through the predictor. This may be illustrated by a situation in which the public official solicits a bribe rather than the respondent making an initial offer. It is unlikely that the official would be aware of the respondent’s individual level of satisfaction with local and national governments (Clausen et al. 2011). Furthermore, it is improbable that the official would choose to either solicit or not solicit a bribe based on the respondent’s level of satisfaction with respect to local and national governments. Thus, it is unlikely that the instrument had a direct effect on the outcome variables. Finally, if a public official requested a bribe but the respondent chose not to pay it, then no incidence of corruption would have occurred. In such a case, the possible reduction of trust caused by the bribe that had been solicited (but not paid) is directly controlled by the predictor (corruption = 0), and then transmitted to the outcome via the predictor. Hence, even in this case also, it is unlikely that the instrument had a direct effect on the outcome variable (Habibov & Cheung 2016).

Inasmuch as the LITS inquired about corruption in eight types of public services, we summed up the above-described binomial variable ‘solicited bribe’ in an additive index that encompassed all services evaluated by the LITS. The value of the index may vary from 0 (if no bribe had been solicited by public officials) to 8 (if bribes had been solicited in all services). Following this, we correlated this index with the outcome variable, satisfaction with local and national governments. The correlation was negligible, namely, −0.01 for local and 0.00 for national governments, suggesting that the instrument was not directly associated with outcome (Mukaka 2012).

In addition, we looked at the situation in which a respondent with a lower level of government satisfaction and, as such, lower expectations of public services, may have been more likely to interpret an uncertain interaction with a public official as a demand for a bribe. However, given such a scenario, the ultimate decision to accept or reject the offered but not actually requested bribe lay with the public official rather than the respondent. To test whether a respondent with a lower level of satisfaction was more likely to interpret an ambiguous interaction with a public official as a request for a bribe, we correlated satisfaction with local and national governments with being asked to pay a bribe. The correlation was low: −0.03 for local and 0.01 for national governments.

Finally, we considered an alternative situation in which a respondent with a low level of satisfaction with the government may choose not to interact with public institutions at all. As a result, such an individual is less likely to be asked to pay any bribes. In reality, however, avoiding contact with public institutions such as law enforcement, traffic police, courts and government offices issuing land and property titles may disadvantage the respondent more than if the respondent had interacted with the institutions and paid the requested bribes. Indeed, the correlation between the frequency of use of public services and satisfaction with local and national governments was negligible, namely, 0.00 for local and 0.02 for national governments.

Thus, the instrument passed the empirical tests required for IV models (Wooldridge 2010). The first-stage F statistic is considerably higher than the rule of thumb value of 10, while the minimum eigenvalue statistic is higher than Stock and Yogo’s critical values, suggesting that our instrument is not weak (Stock et al. 2002; Stock & Yogo 2005). At the same time, significant results of the Durbin–Wu–Hausman tests of endogeneity indicate that IV results should be preferred over single-stage OLS.

Statistical analysis

Since our outcome variables are in the form of an ordinal scale, we estimated the classic IV regression model (Wooldridge 2010). Given that we have two outcome variables, satisfaction with the local and the national government, we estimated two IV regression models. All estimations include country and community dummies to control for country and community level fixed effects. The baseline IV regression model can be described as:(1) $\begin{array}{r}\mathrm{M}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{g}\mathrm{e}:\\ & Satisfactionwithgovernmen{t}_i=constant+{\beta }_{1}corruptio{n}_i\\ & +{\beta }_{2}covariate{s}_i+{\beta }_{3}countryfixedeffect{s}_i\\ & +{\beta }_{4}communityfixedeffect{s}_i+erro{r}_i\end{array}$(1) (2) $\begin{array}{r}\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{s}\mathrm{t}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{g}\mathrm{e}:\\ & Corruptio{n}_i=constant+{\theta }_{1}instrumen{t}_i+{\theta }_{2}covariate{s}_i\\ & +{\theta }_{3}countryfixedeffect{s}_i\\ & +{\theta }_{4}communityfixedeffect{s}_i+erro{r}_i\end{array}$(2) where the instrument is the index of ‘solicited bribe’ over eight public services.

To study the complementary effects between corruption and public service satisfaction, improvement in economic situation, improvement in the political situation and support for democracy, we included the interaction terms (corruption × public service satisfaction, corruption × economic situation improved, corruption × political situation improved and corruption × support for democracy) into the IV regression model one by one by treating the interaction term as a potential endogenous variable. Following the example of Murnane and Willett, the baseline IV regression model should be rewritten as (Murnane & Willett 2010, pp. 247–49):(3) $\begin{array}{r}\mathrm{M}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{g}\mathrm{e}:\\ & Satisfactionwithgovernmen{t}_i=constant+{\beta }_{1}corruptio{n}_i+{\beta }_{2}interactionter{m}_i\\ & +{\beta }_{3}covariate{s}_i+{\beta }_{4}countryfixedeffect{s}_i+{\beta }_{5}communityfixedeffect{s}_i+erro{r}_i\end{array}$(3) (4) $\begin{array}{r}\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{s}\mathrm{t}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{g}\mathrm{e}\left(\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{o}\mathrm{f}\mathrm{t}\mathrm{w}\mathrm{o}\mathrm{e}\mathrm{q}\mathrm{u}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s}\right):\\ & Corruptio{n}_i=constant+{\theta }_{1}instrumen{t}_{1,i}+{\theta }_{2}instrument{s}_{2,i}\\ & +{\theta }_{3}covariate{s}_i+{\theta }_{4}countryfixedeffect{s}_i\\ & +{\theta }_{5}communityfixedeffect{s}_i+erro{r}_i\end{array}$(4) (5) $\begin{array}{rl}Interactionter{m}_i=& constant+{\lambda }_{1}instrumen{t}_{1,i}+{\lambda }_{2}instrumen{t}_{2,i}+{\lambda }_{3}covariates\\ & +{\lambda }_{4}countryfixedeffects+{\lambda }_{5}communityfixedeffects+erro{r}_i\end{array}$(5) where the first instrument (instrument_1,i) is the index of ‘solicited bribe’ and the second instrument is the interaction term between the first instrument and the term (namely, public service satisfaction, economic situation improved, political situation improved or support for democracy) interacting with the predictor in the main stage. Since we treat the interaction term as a potential endogenous predictor variable, it must also satisfy the usual assumptions of the IV regression model.

Findings

The results of the baseline IV regression model

The results of the main stage of IV regressions are reported in models 1–2 of Table 2. Corruption has a negative and statistically significant effect on satisfaction with both local governments (β = −0.18; p-value = 0.000) and national governments (β = −0.11; p-value = 0.000). In terms of covariates, the direction and significance of the regression coefficients in the IV regression models for satisfaction with local and national governments are similar. Thus, increased satisfaction with public services has a positive effect on satisfaction with local and national governments. An improved economic situation and support for democracy also have a similar positive influence on satisfaction with both levels of government. In contrast, an improved political situation has less of an effect on satisfaction with local governments than it does on satisfaction with national governments. Older individuals and females expressed more satisfaction with both levels of government, although the effects of age and gender are rather low. The effects of university education and household wealth are not significant.

TABLE 2 Results of Baseline Models

	IV		OLS
	Model 1 Local	Model 2 National	Model 3 Local	Model 4 National
Panel A: Result of main-stage
Corruption	−0.176*** (0.013)	−0.113*** (0.013)	−0.098*** (0.007)	−0.070*** (0.007)
Public service satisfaction	0.105*** (0.005)	0.083*** (0.005)	0.101*** (0.005)	0.080*** (0.005)
Age	0.002*** (0.000)	0.001*** (0.000)	0.002*** (0.000)	0.001*** (0.000)
Female	0.049*** (0.011)	0.061*** (0.011)	0.051*** (0.011)	0.062*** (0.011)
University	−0.001 (0.014)	0.014 (0.014)	−0.002 (0.014)	0.013 (0.014)
Middle wealth household	−0.014 (0.013)	−0.015 (0.013)	−0.017 (0.013)	−0.017 (0.013)
Wealthiest household	−0.016 (0.014)	−0.016 (0.014)	−0.025 (0.014)	−0.021 (0.014)
Economic situation improved	0.214*** (0.017)	0.231*** (0.017)	0.217*** (0.017)	0.233*** (0.017)
Political situation improved	0.178*** (0.016)	0.356*** (0.016)	0.180*** (0.016)	0.358*** (0.016)
Support for democracy	0.046*** (0.011)	0.039*** (0.011)	0.047*** (0.011)	0.040*** (0.011)
Country dummies included	Yes	Yes	Yes	Yes
PSU dummies included	Yes	Yes	Yes	Yes
N	28,336	26,788	28,336	26,788
F statistic	23.25***	43.09***	23.38***	43.24***
First-stage regression summary statistics
F statistics for the first-stage	12,000	11,000
Minimum eigenvalue statistic	12,000	11,000
Stock and Yogo’s critical values	16.38	16.38
Tests of endogeneity
Durbin–Wu–Hausman chi-sq test	49.10***	14.75***
Panel B: Results of first-stage
Index of ‘Solicited bribe’	1.009*** (0.009)	1.003*** (0.010)
Public service satisfaction	0.062*** (0.003)	0.063*** (0.003)
Age	−0.001*** (0.000)	−0.001*** (0.000)
Female	−0.018* (0.007)	−0.020** (0.008)
University	0.019 (0.010)	0.021* (0.010)
Middle wealth household	0.024** (0.009)	0.020* (0.009)
Wealthiest household	0.063*** (0.010)	0.059*** (0.010)
Economic situation improved	−0.014 (0.012)	−0.007 (0.012)
Political situation improved	−0.013 (0.011)	−0.018 (0.011)
Support for democracy	−0.002 (0.008)	−0.008 (0.008)
Country dummies included	Yes	Yes
PSU dummies included	Yes	Yes
N	28,336	26,788

Notes: Standard errors are in brackets. *p < 0.05; **p < 0.01; ***p < 0.001.

The results of OLS regression models

We also estimated OLS models with the same set of variables as the IV regression model. The results of OLS estimations are reported in models 3–4 and can serve as a benchmark. Again, corruption leads to a lower level of satisfaction with local governments (β = −0.01; p-value = 0.000) and national governments (β = −0.07; p-value = 0.000). As observed, IV regression models suggest a higher magnitude with respect to the corruption effect than do OLS models. Thus, IV regression shows that the effect of corruption with respect to satisfaction with the national government is approximately 1.6 times larger than OLS suggests the effect would be. Regarding satisfaction with the local government, the difference between the suggested effect of corruption between IV regression and OLS is even larger, at about 1.8 times. These findings highlight a higher endogeneity bias in the estimations of satisfaction with local governments than with national ones.

Interestingly, corruption has a stronger negative effect with regard to local governments than national ones, regardless of the model used for estimation. Hence, IV suggests that the negative effect of corruption on local governments is 55% larger than the effect on national governments. Similarly, OLS suggests that the negative effect of corruption on local governments is 40% larger than the effect on national governments.

The results of IV with interaction terms

The results of the main stage of IV regression estimations with interaction terms are reported in models 5–12 of Table 3. The results in Table 3 include the full baseline specifications from Table 2, although only the main variables of interest are reported to conserve space.¹

TABLE 3 Results of with Interaction Term Models

	Local				National
	Model 5	Model 6	Model 7	Model 8	Model 9	Model 10	Model 11	Model 12
Corruption	−0.214*** (0.016)	−0.197*** (0.016)	−0.199*** (0.016)	−0.164*** (0.017)	−0.135*** (0.016)	−0.128*** (0.016)	−0.131*** (0.016)	−0.100*** (0.017)
Public service satisfaction	0.082*** (0.007)				0.069*** (0.007)
Corruption × public service satisfaction	0.047*** (0.009)				0.027** (0.009)
Economic situation improved		0.189*** (0.019)				0.214*** (0.019)
Corruption × economic situation improved		0.062* (0.025)				0.042 (0.024)
Political situation improved			0.151*** (0.018)				0.336*** (0.018)
Corruption × political situation improved			0.067** (0.024)				0.053* (0.024)
Support for democracy				0.056*** (0.014)				0.049*** (0.014)
Corruption × support for democracy				−0.026 (0.022)				−0.027 (0.022)
Dummies country included	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Dummies PSU included	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Socio-demographic characteristics	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	28,336	28,336	28,336	28,336	26,788	26,788	26,788	26,788
F statistic	23.30***	23.17***	23.19***	23.15***	42.97***	42.91***	42.92***	42.91***
First-stage regression summary statistics
F statistics	3,591.01***	5,635.58***	5,723.84***	5,857.72***	3,433.95***	5,221.85***	5,263.28***	5,453.83***
Stock and Yogo’s critical values	7.03	7.03	7.03	7.03	7.03	7.03	7.03	7.03
Tests of endogeneity
Wu–Hausman F statistic	19.54***	25.64***	25.71***	25.35***	6.61**	8.22***	8.29***	8.02***
Durbin–Wu–Hausman chi-sq test	39.36***	51.63***	51.77***	51.05***	13.32**	16.58***	16.73***	16.30***

Notes: Standard errors are in brackets. *p < 0.05; **p < 0.01; ***p < 0.001.

Corruption has a negative effect on satisfaction with government at the local and national levels in all models. Similar to the baseline models, corruption has a stronger negative effect on local governments than on national governments, regardless of the model estimated. Indeed, the negative effects of corruption on satisfaction with local governments vary between model 8 (β = −0.16; p-value = 0.000) and model 5 (β = −0.21; p-value = 0.000). In comparison, the negative effects of corruption on satisfaction with national governments are relatively lower in model 12 (β = −0.10; p-value = 0.000) and model 9 (β = −0.14; p-value = 0.000).

The interaction terms ‘corruption × public service satisfaction’ also have a positive effect both on local governments (β = 0.05; p-value = 0.000) and national governments (β = 0.03; p-value ≤ 0.05), as do the interaction terms ‘corruption × political situation improved’ for local governments (β = 0.07; p-value ≤ 0.05) and national governments (β = 0.05; p-value ≤ 0.01). In contrast, the interaction terms ‘corruption × economic situation improved’ are significant only in estimations for local governments (β = 0.06; p-value ≤ 0.01), while ‘corruption × support for democracy’ is not significant in any of the estimations.

To assist us further in interpreting the results, a detailed visual illustration of the effect of the interaction terms is provided in Figures 1 and 2 (Brambor et al. 2006).

FIGURE 1. Marginal Effect of Interaction Term Corruption × Public Service Satisfaction

Notes: Local gov―Local government; 95% Upper C.I.―95% upper confidence intervals; 95% Lower C.I.―95% lower confidence intervals; Marginal effect of corruption is based on model 5 for local governments and model 9 for national governments in Table 3.

FIGURE 2. Marginal Effect of Corruption × Political Situation Improved and Corruption × Economic Situation Improved

Notes: Local gov―Local government; National gov―National government; 95% Upper C.I.―95% upper confidence intervals; 95% Lower C.I.―95% lower confidence intervals; Political situation improved is based on model 7 for local governments and model 11 for national governments in Table 3; Economic situation improved is based on model 6 for national governments in Table 3.

Figure 1 illustrates the marginal effect of the interaction term ‘corruption index × public service satisfaction’. As public service satisfaction increases, the effect of corruption changes from negative values to positive values for both local and national governments. The first diagram illustrates that the effect of corruption on satisfaction with local governments went from approximately −0.21 to approximately 0.16 as public service satisfaction increased. The second diagram shows that the effect of corruption on satisfaction with national governments went from approximately −0.14 to approximately 0.08 as public service satisfaction increased.

The first diagram in Figure 2 illustrates the effect of the ‘corruption index × political situation improved’ interaction term. With respect to local governments, the effect of corruption is −0.13 for respondents who believe that the political situation has improved, whereas the effect of corruption is −0.20 for respondents who disagree. With regard to national governments, the effect of corruption is −0.08 for respondents who believe that the political situation has improved, where the effect of corruption is −0.13 for respondents who disagree.

The second diagram in Figure 2 depicts the effect of the interaction terms ‘corruption index × economic situation improved’ for local governments. The effect of the corruption is −0.13 for respondents who believe that the economic situation has improved, whereas the effect of corruption is −0.20 for respondents who disagree.

Conclusion

The empirical evidence in this study provides a rather complex picture. On the one hand, we found a significant and negative effect of corruption on satisfaction with government, regardless of the statistical specifications used. As far as estimations of the instrumental variables can address reverse causality and omitted variable problems, our findings suggest that the direction of the effect goes from corruption to government satisfaction. In general, we also reject the ‘culture of corruption’ hypothesis by showing that corruption has a significant effect on satisfaction with the government in all estimations. Although corruption is widespread in post-communist countries, it is not viewed by citizens as a ‘rule of engagement’ or an everyday occurrence.

On the other hand, further analysis with respect to the interaction terms provides suggestive evidence for the ‘grease the wheels’ hypothesis inasmuch as the negative effects of corruption are reduced in an environment characterised by perceptions of improvement in the political and economic situation, as well as in satisfaction with public services. The partial validation of the ‘grease the wheels’ hypothesis suggests that many citizens in post-communist countries are ready to tolerate corruption, which they regard as a lesser evil, providing that the economic and political situation, as well as public services, are seen to be improving. The opposite is also true. The negative effects of corruption become more damaging if citizens feel that there are no positive changes in the political and economic situation, and satisfaction with public services is low.

Finally, we also found that the negative effects of corruption are higher with respect to local rather than national governments. The explanation could be twofold (Harmel & Yeh 2011). On the one hand, many public services—for instance, healthcare and education—are managed by local governments. Consequently, corruption in those services is blamed on local governments. On the other hand, it could be the case that many citizens in post-communist countries believe in the good intentions of their national governments—intentions that are distorted and ruined by corrupt practices at the local level.

Additional information

Notes on contributors

Nazim Habibov

NAZIM HABIBOV, Associate Professor, School of Social Work, University of Windsor, 167 Ferry Street, Windsor, Ontario N9A0C5, Canada.

Lida Fan

LIDA FAN, Associate Professor, School of Social Work, Lakehead University, 955 Oliver Road, Thunderbay, Ontario P7B 5E1, Canada. Email: [email protected]

Alena Auchynnikava

ALENA AUCHYNNIKAVA, School of Social Work, University of Windsor, 167 Ferry Street, Windsor, Ontario N9A 0C5, Canada. Email: [email protected]

Notes

1 The complete results are available from the authors upon request.