Content

3 Unpacking Corruption in Europe in:

Ina Kubbe

Corruption in Europe, page 99 - 159

Is it all about Democracy?

1. Edition 2015, ISBN print: 978-3-8487-2347-8, ISBN online: 978-3-8452-6451-6, https://doi.org/10.5771/9783845264516-99

Series: Comparative Politics - Vergleichende Politikwissenschaft, vol. 6

Bibliographic information
Unpacking Corruption in Europe The theoretical approaches, empirical studies, variables and hypotheses described earlier are included in this study’s bathtub model of corruption that combines economic and sociological approaches into an interdisciplinary framework and integrates certain variables that are linked to a society’s democratic development at the macro and micro level. For the purpose of identifying the causes of corruption in Europe, I fill the concrete situations at each level of the model by empirical data. According to that, I assume that economic, political, socio-cultural and historical variables affect the extent of corruption at the macro level that includes the characteristics of certain countries (situation 1 in the model) whereas, at the micro level certain characteristics of individuals such as socio-demographic factors, values, norms, and attitudes are expected to impact the extent of corruption. The aggregation of all individual corrupt actions (“aggregational logic”), in turn, leads to situation 2 on the macro level, where corruption can be measured by certain indices such as the CPI. The macro and micro level are examined by panel and multilevel analyses to explain the extent of corruption in European states across time at the macro level via the aggregated individual actions at the micro level (situation 2 in the model). The extent of corruption at the macro level is measured by the Corruption Perceptions Index and additionally by the Control of Corruption Index for a period of 19 years (1995-2013). Subsequently, I conduct an crosscultural multilevel analysis at the micro level from representative population surveys around Europe. The dependent variable at the individual level is measured by the item ““How widespread do you think bribe taking and corruption is in this country?” of three waves from the World Values Survey (1994-1999, 1999-2004, 2005-2009). Unfortunately, the fourth wave of the WVS (2010-2014) does not include the item for European states. Nonetheless, especially the interrelation between macro and micro level implies theoretical as well as methodological challenges. The methodological difficulties of uniting micro and macro level data have been taken up in a series of papers (King et al., 1994; King et al., 2004). Particularly, problems arise where general assumptions are made about countries that are internally diverse. Western (1998, p. 1255) calls this “the fundamental problem of comparative research.” More precisely, wherever contextual 3 99 variables are invoked, differences in causal processes within countries are related to characteristics that vary across them. I will take these problems into account to avoid ecological fallacies, reconstructing and drawing conclusions on individual behavior from country-level data, and individualistic fallacies, occurring when observations at the individual level are generalized to the macro or meso level (Robinson, 1950). Explaining Corruption at the Macro Level The empirical analysis performed at the macro level follows a panel-data research design that provides a rich and powerful study of a set of people or countries. Panel data are repeated measures on individuals or countries, observed for several time periods. In particular, panel data analyses can encompass a wide variety of certain approaches. The most common types are independently pooled panels, fixed effects and random effects models. Among these types of models, additional dynamic panel, robust, and covariance structure models exist (Wooldridge, 2007).50 Panel data analyses offer a large number of advantages compared with other conventional quantitative methods such as cross-sectional analyses. For instance, they enable researchers to run regression analysis considering both the spatial and temporal dimension of data. While the spatial dimension refers to a set of cross-sectional units of observation, the temporal dimension pertains to periodic observations of a set of variables characterizing these cross-sectional units over a particular time period. Therefore, panel data include high informative values that are more variable, less collinear and include more degrees of freedom leading to more efficient estimates than standard regression analyses. The increased precision in estimation is particularly a result of an increase in the number of observations owing to combining or pooling several time periods of data for each individual or country. In addition, with the help of panel analyses, variance within individuals or societies can be achieved and, in contrast to crosssectional analyses, causal effects of several independent variables on the dependent variable can be determined. Hence, with repeated observations of enough cross-sections, panel analyses facilitate learning more about dy- 3.1 50 For more details on this see also Cameron and Trivedi (2005); Stock and Watson (2007) or Wooldridge (2009). 3 Unpacking Corruption in Europe 100 namics with short time series than is possible from a single cross-country analysis. The combination of time series with cross-sections can achieve a quality and quantity of data that is impossible using only one of these two dimensions (Gujarati and Porter, 2009). Furthermore, to handle the problem of serial dependence, estimations can be based on panel-corrected standard errors. So, the linear regression models with panel-corrected standard errors51 take autocorrelation into account. However, even for linear regression, standard panel analyses use a wider range of models and estimators than is the case with cross-section-data. The usual formular of panel analysis reads as follows: yit = a + bxit + ϵit  , where y is the dependent variable corruption, x is the independent variable, a and b are coefficients, i (i = 1, ...,N) and t (t = 1, ...,T) are indices for units of observations (here: countries) and time (here: 1995-2010). The error ϵit  needs special attention in the analyses. For instance, assumptions about the error term determine whether we speak of fixed effects or random effects.52 For the analysis of the causes of corruption in European states, the focus lies on data from a short panel which means large cross-sectional units of observations (such as countries, states, counties) for a few time periods, rather than a long panel such as a small cross section of countries for many time periods (Cameron and Trivedi, 2005; Wooldridge, 2007). In this case, it would be possible to concentrate on time-series analyses. Due to data availability, my investigation encompasses a time horizon of 19 years for 37 countries. To detect the causes of corruption at the country level, I run several linear regression models with panel-corrected standard errors for estimating variance in these models, each including economic, political, socio-cultural and historical variables at the macro level. These variance estimates return to the assumption of many observations per panel but allow for panellevel heteroskedasticity and contemporaneous correlation of observations between the panels. For valid statistical inference it is important to control 51 In Stata, pooled OLS regressions with panel-corrected standard errors can be estimated with the xtpcse command. As a result, OLS parameter estimates along with the panel-corrected variance estimates are received. 52 In a fixed effects model, ϵ��  assumed to vary non-stochastically over �  or �  making the fixed effects model analogous to a dummy variable model in one dimension. In a random effects model, ���  is assumed to vary stochastically over �  or �  requiring special treatment of the error variance matrix. 3.1 Explaining Corruption at the Macro Level 101 for likely correlation of regression model errors over time for the given unit of observation. Suggesting to rely on OLS coefficient estimates with panel-corrected standard errors, Beck and Katz (1995) convincingly demonstrate that their large T asymptotics based standard errors, which correct for contemporaneous correlation between the subjects, perform well in small panels. After checking for multicollinearity53, I analyze these variables (economic, political, socio-cultural and historical factors) are related to corruption in order to identify significant relationships. Finally, I take the significant variables and put them in an overall multilevel model presenting the causes of corruption for European states. Measuring Corruption at the Country Level To measure the dependent variable “Extent of Corruption” I use the Corruption Perceptions Index from Transparency International for a period of 19 years (1995-2013). The CPI-scales are rescaled to a range of 0 to 10, where 0 indicates low corruption and 10 the highest level. In particular, this allows for interpretation and comparison of findings with assessments gained by the Control of Corruption Index provided by the World Bank. Obviously, I can only systematically include countries in the analysis that provide useful data. On the basis of data availability, my investigation includes 37 European countries for the period between 1995 and 2013. Excluded states are either not considered by most other data sources such as Andorra, Liechtenstein, Malta, and San Marino, or are outliers within the dataset, such as Turkey or Russia. For some missing corruption data, I have used the average of one year and the previous year. As a result, the data set is referred to as a balanced panel, meaning that there are hardly any missing values. For comparative reasons and to filter out the specific European determinants of corruption, I ran all calculations with two samples: a European sample and a sample including countries outside Europe. The following illustrations (Figure 4 and Table 2) demonstrate that corruption varies widely across different European countries. The average ex- 3.1.1 53 Multicollinearity appears if two or more independent variables are highly correlated. This was not the case. See also Appendix E in Kubbe (2013). 3 Unpacking Corruption in Europe 102 tent of corruption (1995-2013) in 37 European states is 3.94. The highest levels of corruption are found in Ukraine (7.6), Albania (7.1), Moldova (7.1), Georgia (7.1) and Bosnia and Herzegovina (6.8), followed by Macedonia (6.8), Belarus (6.9) and Georgia (6.6), whereas the countries with the lowest extent of corruption turn out to be the Scandinavian countries: Denmark (0.51), Finland (0.57) and Sweden (0.8), followed by Iceland (0.89), the Netherlands (1.19), Switzerland (1.20), Norway (1.24), Luxembourg (1.59) and the United Kingdom (1.74). Germany, Austria, Ireland, France, Belgium, Portugal, Spain, Estonia, Slovenia and Cyprus score between 2.0 and 4.0 and Hungary, Lithuania, Czech Republic, Poland, Italy, Greece, Slovakia and Latvia score between 4.0 and 6.0. On the whole, Europe is characterized by widely diverging corruption values. More precisely, it is striking, that there are still significant differences between Western and Eastern European states. The average score of the Western countries54 is 2.06. With this score, Western Europe is found at the bottom of corruption values in Europe. Contrary to this, the average corruption level of Eastern states55 is 6.37 and thereby considerably higher. A comparison of Northern56 (2.28) and Southern Europe57 (5.33) show a similar picture. Notably, levels of corruption are on average not exceptionally lower in Southern Europe than in post-communist societies. Countries such as Greece (5.66), Italy (5.44), Portugal (3.67) or Spain (3.67) are similarly rated by the Corruption Perceptions Index as post-communist countries such 54 The sample of West Europe includes Belgium, Germany, France, Luxembourg, Netherlands, Austria and Switzerland. 55 Bulgaria, Moldova, Romania, Slovakia, Ukraine, Hungary, Poland, Czech Republic, Belarus and Georgia are part of the Eastern European sample. 56 Northern Europe includes Denmark, Estonia, Finland, Iceland, Ireland, Lithuania, Latvia, Norway, Sweden and United Kingdom. 57 Albania, Bosnia and Herzegovina, Croatia, Greece, Italy, Portugal, Cyprus, Slovenia, Spain and Macedonia are part of the Southern European sample (United Nations Statistics Division, 2015). 3.1 Explaining Corruption at the Macro Level 103 as Romania (6.63), Hungary (4.98), Slovenia (3.92) or Estonia (3.83) (Corruption Perceptions Index, 2015). Corruption across Europe (Average Level: 1995-2013) Table 2 gives an overall overview of the descriptive statistics of the dependent variable of the country level, the transformed Corruption Perceptions Index. It illustrates the number of observations, means, standard deviations, minimums, and maximums of the data. Corruption in Europe (Corruption Perceptions Index transformed) Countries Observations Mean (1995-2013) Standard Deviation Min Max Ukraine 16 7.60 .33 7.20 8.50 Albania 13 7.14 .39 6.60 7.70 Moldova 15 7.12 .42 6.40 7.90 Belarus 15 6.94 .87 5.20 8.00 Georgia 13 6.69 1.12 4.80 8.20 Bosnia and Herzegovina 11 6.68 .45 5.80 7.10 Romania 17 6.63 .53 5.60 7.40 Macedonia 12 6.57 .70 5.60 7.70 Bulgaria 16 6.23 .36 5.90 7.10 Croatia 15 6.12 .52 5.20 7.30 Figure 4: Table 2: 3 Unpacking Corruption in Europe 104 Countries Observations Mean (1995-2013) Standard Deviation Min Max Latvia 16 5.85 .71 4.70 7.30 Slovakia 16 5.80 .48 5.00 6.50 Greece 19 5.66 .54 4.65 6.60 Italy 19 5.44 .65 4.50 7.01 Poland 18 5.44 .82 4.00 6.60 Czech Republic 18 5.38 .49 4.63 6.30 Lithuania 15 5.22 .44 4.30 6.20 Hungary 19 4.98 .32 4.50 5.88 Cyprus 11 3.94 .47 3.40 4.70 Slovenia 15 3.92 .41 3.30 4.80 Estonia 16 3.83 .46 3.20 4.50 Spain 19 3.67 .81 2.90 5.69 Portugal 19 3.67 .32 3.03 4.44 Belgium 19 3.14 .77 2.40 4.75 France 19 3.07 .29 2.50 3.70 Ireland 19 2.34 .48 1.43 3.10 Austria 19 2.20 .49 1.30 3.10 Germany 19 2.09 .26 1.73 2.70 United Kingdom 19 1.74 .46 1.30 2.60 Luxembourg 18 1.59 .46 1.00 3.15 Norway 19 1.24 .31 .50 2.10 Switzerland 19 1.20 .22 .89 1.60 Netherlands 19 1.19 .20 .97 1.70 Iceland 16 .98 .56 .30 2.20 Sweden 19 .80 .19 .50 1.20 Finland 19 .57 .35 .00 1.10 Denmark 19 .51 .27 .00 1.00 European Countries (Total) 625 3.94 2.28 0 8.50 Moreover, looking at the patterns of corruption development over time (1995-2013), it is striking that the extent of corruption has continuously risen in European states. While the average score of corruption was 2.91 in 1995, it has continuously grown to 4.0 in the year 2013. Notably, the average extent of corruption suddenly increased in Europe in 1999 to 4.19 (see Figure 5). This might be due to the fact that Transparency International in- 3.1 Explaining Corruption at the Macro Level 105 cluded and ranked post-communist countries such as Belarus, Bulgaria, Estonia, Latvia, Slovakia, or the Ukraine for the first time in 1998 (see also Table 3). Extent of Corruption in Europe across Time Table 3 illustrates, in particular, that countries such as Austria, Belarus, the Czech Republic, Germany, Greece, Iceland, Ireland or the United Kingdom have declined in their degrees of corruption. That is an annoying progress and emphasizes the importance of examining the extent and dynamic of corruption. However, there are also countries that could improve their scores of corruption. For instance, states such as Albania, Bulgaria, Estonia, Georgia, Latvia, Lithuania or Romania have decreased in the extent of corruption (Corruption Perceptions Index, 2015). Development of Corruption (Corruption Perceptions Index transformed) Countries 1995 (first survey) 2000 2005 2010 2013 Albania 7.7 (1999) - 7.6 6.7 6.9 Austria 2.8 2.3 1.3 2.1 3.1 Belarus 6.1 (1998) 5.9 7.4 7.5 7.1 Belgium 3.1 3.9 2.6 2.9 2.5 Figure 5: Table 3: 3 Unpacking Corruption in Europe 106 Countries 1995 (first survey) 2000 2005 2010 2013 Bosnia and Herzegovina 6.7 (2003) - 7.1 6.8 5.8 Bulgaria 7.1 (1998) 6.5 6.0 6.5 5.9 Croatia 7.3 (1999) 6.3 6.6 5.9 5.2 Cyprus 3.9 (2003) - 4.3 3.7 3.7 Czech Republic 4.6 (1996) 5.7 5.7 5.4 5.2 Denmark 0.6 0.1 0.5 0.7 0.9 Estonia 4.3 (1998) 4.3 3.6 3.5 3.2 Finland 0.8 0 0.3 0.8 1.1 France 3 3.3 2.5 3.2 2.9 Georgia 7.7 (1999) - 7.7 6.2 5.1 Germany 1.8 2.4 1.8 2.1 2.2 Greece 5.9 5.1 5.7 6.5 6.0 Hungary 5.8 4.8 5 5.3 4.6 Iceland 0.6 (1998) 0.8 0.3 1.5 2.2 Ireland 1.4 2.8 2.6 2.0 2.8 Italy 7.0 5.4 5.0 6.1 5.7 Latvia 7.3 (1998) 6.6 5.8 5.7 4.7 Lithuania 6.2 (1999) 5.9 5.2 5.0 4.3 Luxembourg 3.1 1.4 1.5 1.5 2.0 Macedonia 6.7 (1999) - 7.3 5.9 5.6 Moldova 7.4 (1999) 7.4 7.1 7.1 6.5 Netherlands 1.3 1.1 1.4 1.2 1.7 Norway 1.3 0.8 1.1 1.4 1.4 Poland 4.4 (1996) 5.9 6.6 4.7 4.0 Portugal 4.4 3.6 3.5 4.0 3.8 Romania 6.5 (1997) 7.1 7.0 6.3 5.7 Slovakia 6.1 (1998) 6.5 5.7 5.7 5.3 Slovenia 4.0 (1999) 4.5 3.9 3.6 4.3 Spain 5.6 3.0 3.0 3.9 4.1 Sweden 1.1 0.6 0.8 0.8 1.1 Switzerland 1.2 1.4 0.8 1.3 1.5 Ukraine 7.2 (1998) 8.5 7.4 7.6 7.5 United Kingdom 1.4 1.3 1.4 2.4 2.4 3.1 Explaining Corruption at the Macro Level 107 Taking a closer look at each country itself, the different developments of the extent of corruption in European states also becomes evident and illustrates various dynamics in the certain countries (Figure 6). It is notable, that there are countries that have scarcely changes their degrees of corruption over time such as the Scandinavian countries (e.g. Sweden, Denmark) or the Netherlands. Otherwise, it is illustrated that nations exist that show strong decreases in the development of corruption for the time period of 1995-2013. The reasons for these dynamics and various extents of corruption in Europe are still unexplained. Thus, the question remains: “What affects (perceived) corruption in European states over time and across and within countries?” 3 Unpacking Corruption in Europe 108 Extent of Corruption in European States across TimeFigure 6: 3.1 Explaining Corruption at the Macro Level 109 3 Unpacking Corruption in Europe 110 3.1 Explaining Corruption at the Macro Level 111 3 Unpacking Corruption in Europe 112 Empirical Findings: The Impact of Country Characteristics on Corruption Economic Factors The economic factors include a country’s economic development and its integration in the European Union. Figure 7 illustrates the average development of the economic development, compared to the average score of corruption in Europe, over time. The EU-membership is presented by boxplots, because it is designed as a continuous variable (dummy variable). Extent of Corruption and Economic Factors across Time Economic Development Data on the economic development is taken from the World Bank and measured by the GDP per capita based on purchasing power parity (PPP). PPP GDP is the gross domestic product converted to international dollars using purchasing power parity rates. “An international dollar has the same purchasing power over GDP as the U.S. dollar has in the United States. GDP at purchaser's prices is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making 3.1.2 3.1.2.1 Figure 7: 3.1 Explaining Corruption at the Macro Level 113 deductions for depreciation of fabricated assets or for depletion and degradation of natural resources” (World Bank, 2014). A descriptive analysis reveals that the average rate of a country’s economic development (logarithmized) in European states for the period from 1995-2013 gradually increases. While in 1995 the average logarithmized GDP score was 9.4, the level has improved to 9.9 in 2013. The lowest economic development is assigned to Moldova (7.5), followed by Georgia (7.9) and the Ukraine (8.3). The highest development has Luxembourg (11.1), followed by Norway (10.8) and Switzerland (10.6). The following scatterplots indicate a negative relationship between the extent of corruption and a country’s economic development, suggesting that the extent of corruption will be higher in European countries with lower levels of economic development. This is confirmed by a negative correlation coefficient of -0.82 and initially conforms the assumed relationship. Correlation between Extent of Corruption and Economic Development (logarithmized) Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. Economic Development (logarithmized) is measured by GDP per capita based on purchasing power parity. Figure 8: 3 Unpacking Corruption in Europe 114 Correlation between Extent of Corruption and Economic Development across European Countries (Average) Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. Economic Development (logarithmized) is measured by GDP per capita based on purchasing power parity. EU-Membership To measure EU-membership, I constructed a dummy variable for the membership in the European Union. I take the data from the respective websites. In fact, this means that the country has the number 1) if it is a member of the respective organization; and 0) if the country is a nonmember. Most European countries are members of the EU. 75%, which means 28 out of 37 countries of the European sample, are members (status as of June 2015). Belgium, France, Germany, Italy, Luxembourg and the Netherlands were founding members of the EU in the 1950s. Later, Denmark, Ireland and UK joined (1973), followed by Greece in 1981 and Portugal and Spain in 1986. Austria, Finland and Sweden became members in 1995. Most of the post-communist countries joined the European Union in 2004, such as Cyprus, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia. In 2007, Bulgaria and Romania Figure 9: 3.1 Explaining Corruption at the Macro Level 115 also became members of the European Union. In 2013, Croatia joined the EU. It is striking that in most countries that joined the EU during the period of this investigation (1995-2013), such as Cyprus, Czech Republic, Estonia, Latvia, Lithuania, Poland, Slovakia and Slovenia improved their corruption scores after becoming member in the EU (Corruption Perceptions Index, 2015). The boxplots illustrate the statistical distribution of the data, comparing the countries that are member and non-members of the European Union (Figure 10). A correlation coefficient of -0.44 suggests a negative relationship between the extent of corruption and EU-membership and corresponds to the hypothesis the extent of corruption will be higher, if the country is not a member state of the European Union. Boxplots of the Extent of Corruption and EU-Membership Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. EU-Membership is measured by 1 = EU-Member and 0 = Non-EU-Member. The following table reports the overall summary of the descriptive statistics of the economic variables used in the empirical analysis. It presents the number of observations, means, standard deviations, and minimums and maximums of used data. The number of observations results from the Figure 10: 3 Unpacking Corruption in Europe 116 time period (1995-2013) and the number of included European countries. Overall, I collected the data from different sources such as the World Bank or International Labour Organisation, supported by the Quality of Government dataset of the University of Gothenburg in Sweden (see Appendix B). Economic Factors (1995-2013) Variables Observations Mean Standard Deviation Min Max Economic Development 696 9.7 0.89 7.33 11.62 EU-Membership 703 .55 .49 0 1 Political Factors The political factors include the degree of democracy and the percentage of women in parliaments. The following figure 11 visualizes the average development of these variables, compared to the average score of corruption in Europe. Extent of Corruption and Political Factors across Time Table 4: 3.1.2.2 Figure 11: 3.1 Explaining Corruption at the Macro Level 117 Degree of Democracy To measure the degree of democracy, I use an average by Freedom House (political rights and civil liberties)58 that is transformed to a scale of 0-10 and Polity IV59 that is transformed to a scale of 0-10 as well. Thereby, the scale of the variables ranges from 0-10 where 0 is least democratic and 10 most democratic. Moreover, the version of Polity IV has imputed values for countries where data on Polity is missing by regressing Polity on the average Freedom House measure. Hadenius and Teorell (2005) show that this average index performs better, both in terms of validity and reliability than its constituent parts (Quality of Government, 2015). A preliminary descriptive analysis suggests that the average degree of democracy in European states gradually increases for the period from 1995-2013. While in 1995 the average score was 8.58, the level has improved to 9.28 in 2013. As figure 12 illustrates, the least democratic country in Europe is Belarus (1.51), followed by Bosnia and Herzegovina (5.10) and Georgia (6.57). In 58 Freedom House evaluates each country’s political rights as well as civil liberties with a rating from 1 to 7, where 1 represents the most free and 7 the least free. Political rights ratings are based on an evaluation of the electoral process, political pluralism and participation, and the functioning of government. Civil liberties ratings are based on an evaluation of freedom of expression and belief, associational and organizational rights, rule of law, personal autonomy and individual rights. The ratings of political rights and civil liberties are determined by the total number of points (up to 100). Each country receives 10 political rights questions and 15 civil liberties questions; countries receive 0 to 4 points on each question, with 0 representing the smallest degree and 4 the greatest degree of freedom. The average of the political rights and civil liberties ratings determines the country`s overall status: Free (1.0 to 2.5), Partly Free (3.0 to 5.0) or Not Free (5.5 to 7.0). Freedom House also assigns upward or downward trend arrows to countries that saw general positive or negative trends during the year that were not significant enough to result in a ratings change (Freedom House, 2015b). 59 The Polity IV Project carries data collection that especially includes indices that measure the degree of democracies and autocracies for purposes of comparative, quantitative analyses. The Polity score covers all independent states with a total population greater than 500.000 for a time period from 1800 to 2010 and captures the regime authority spectrum on a 21-point scale ranging from -10 (hereditary monarchy) to +10 (consolidated democracy). The indicators are primarily based on subjective interpretations of historical material and similar sources by experts Polity IV. 3 Unpacking Corruption in Europe 118 contrast, the most democratic nations are Austria (10.00), Finland (10.00) or Ireland (10.00)60 (Freedom House 2015b; Polity IV, 2015). As assumed, the following scatterplots indicate a negative relationship between the extent of corruption and the degree of democracy, suggesting that high degrees of democracy may hinder the growth of corruption in European countries. The correlation coefficient is -0.58 and conforms the assumed corruption-democracy-nexus that more advanced democratic structures lead to a lower extent of corruption in European countries. Correlation between the Extent of Corruption and the Degree of Democracy Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. Degree of Democracy is measured by an average of Freedom House and Polity IV scaled from 0 (“least democratic”) to 10 (“most democratic”). Figure 12: 60 Countries such as Cyprus, Denmark, Iceland, Luxembourg, Netherlands, Norway, Portugal, Sweden and Switzerland also received an average of 10.00 points by Freedom House/Polity IV and were ranked as well as established democracies. 3.1 Explaining Corruption at the Macro Level 119 Correlation between the Extent of Corruption and the Degree of Democracy across European Countries (Average) Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. Degree of Democracy is measured by an average of Freedom House and Polity IV scaled from 0 (“least democratic”) to 10 (“most democratic”). Women in Parliaments To investigate the assumption that the extent of corruption will be lower in countries that have higher levels of female participation in parliaments, I particularly use data from the World Bank and the United Nations to measure the percentage of parliamentary seats in a single or lower chamber held by women. A preliminary descriptive analysis demonstrates that the percentage of women in European parliaments increases steadily. In 199761, the average score was 15.97%, whereas it increased to 25.24% in 2013. The Ukraine has with 7.25% the lowest percentage of women in parliaments and is followed by Georgia with 7.78% and Albania with 9.73%. Sweden has with 44.41% the highest percentage of women in par- Figure 13: 61 There is no data available on the percentage of women in parliaments in European states for the years 1995 and 1996. 3 Unpacking Corruption in Europe 120 liaments, followed by Finland with 38.58% and Denmark with 37.23% (World Bank Indicators, 2015). The following scatterplots indicate a negative relationship between the extent of corruption and the percentage of women in parliaments (Figure 14 and 15). The cases are relatively well distributed along the negative regression line. This suggests that a high share of women in parliaments leads to a decrease in the extent of corruption in European states. This is substantiated by a negative correlation coefficient of -0.66 between both variables and initially confirms the assumed relationship. Correlation between the Extent of Corruption and Women in Parliaments Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Women in Parliaments is measured by the percentage of parliamentary seats in a single or lower chamber held by women. Figure 14: 3.1 Explaining Corruption at the Macro Level 121 Correlation between the Extent of Corruption and Women in Parliaments across European Countries (Average) Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Women in Parliaments is measured by the percentage of parliamentary seats in a single or lower chamber held by women. The following table 5 gives an overall overview of the descriptive statistics of the political variables used in the analysis. It illustrates the number of observations, means, standard deviations, and minimums and maximums of used data. The number of observations is also based on the time period (1995-2013) and the number of included European countries. Most of the data comes from international organizations such as Freedom House, the World Bank or United Nations (see Appendix B). Political Factors (1995-2013) Variables Observations Mean Standard Deviation Min Max Degree of Democracy 667 8.96 1.81 1.08 10 Women in Parliaments 623 21.20 10.80 3 47.3 Figure 15: Table 5: 3 Unpacking Corruption in Europe 122 Socio-Cultural Factors The socio-cultural factors include four variables: a society’s dominant religion (Catholics, Orthodox, Protestants, Muslims). The following graphic illustrates the average development of the socio-cultural variables over time (1995-2013), compared to the average score of corruption in Europe. Extent of Corruption and Socio-Cultural Factors across Time Religion To examine the influence of religion on the extent of corruption, I primarily took percentage data from the Worldmark Encyclopedia of the Nations, the Statistical Abstract of the World and the United Nations to measure a society’s dominant religion. Percentage of Catholics The average percentage of Catholics in European states is 35%. European countries with a large proportion of Catholics are, for instance, Spain, Poland and Croatia (Worldmark Encyclopedia of the Nations, 2015). Countries with low percentages of Catholics are the Scandinavian countries 3.1.2.3 Figure 16: 3.1 Explaining Corruption at the Macro Level 123 such as Denmark, Norway or Sweden. The regression line of the following scatterplots indicates a weak negative relationship between the extent of corruption and a society’s percentage of Catholics (see Figure 17 and 18). This is substantiated by a correlation coefficient of -0.05. Correlation between the Extent of Corruption and Percentage of Catholics Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Catholics is measured by the percentage of population. Figure 17: 3 Unpacking Corruption in Europe 124 Correlation between the Extent of Corruption and Percentage of Catholics across European Countries (Average) Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Catholics is measured by the percentage of population. Percentage of Orthodox The average percentage of Orthodox in European states is with 0.22% very low. Countries with a large proportion of Orthodox are especially found in Greece, Moldova and Romania (Worldmark Encyclopedia of the Nations, 2015). Countries with low percentages of Orthodox people are especially Western and Northern societies such as Belgium, France and the United Kingdom. The regression lines of the following scatterplots indicate a positive relationship between the extent of corruption and a society’s percentage of Orthodox. This is confirmed by the correlation coefficient of 0.62, suggesting that a high percentage of Orthodox may lead to an increase in the extent of corruption (Figure 19 and 20). Figure 18: 3.1 Explaining Corruption at the Macro Level 125 Correlation between the Extent of Corruption and Percentage of Orthodox Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Orthodox is measured by the percentage of population. Figure 19: 3 Unpacking Corruption in Europe 126 Correlation between the Extent of Corruption and Percentage of Orthodox across European Countries (Average) Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Orthodox is measured by the percentage of population. Percentage of Protestants The average percentage of Protestants in European is 17%. In particular, Denmark, Finland, Iceland, Norway and Sweden are countries with a high proportion of Protestants (Worldmark Encyclopedia of the Nations, 2015). The regression line of the following scatterplots suggests a negative relationship between the extent of corruption and a society’s percentage of Protestants (Figure 21 and 22). The correlation coefficient is -0.65 and initially verifies the assumed relationship that countries with higher levels of Protestants are likely to be less corrupt. Figure 20: 3.1 Explaining Corruption at the Macro Level 127 Correlation between the Extent of Corruption and Percentage of Protestants Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Protestants is measured by the percentage of population. Figure 21: 3 Unpacking Corruption in Europe 128 Correlation between the Extent of Corruption and Percentage of Protestants across European Countries (Average) Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Protestants is measured by the percentage of population. Percentage of Muslims The average percentage of Muslims in European states is approximately 6%. Albania, Bosnia and Herzegovina, and Macedonia are societies with a high percentage of Muslims (Worldmark Encyclopedia of the Nations, 2015). The regression line of the following scatterplots indicates a positive relationship between corruption levels and a society’s percentage of Muslims. The correlation coefficient is 0.28 (Figure 23 and 24). Figure 22: 3.1 Explaining Corruption at the Macro Level 129 Correlation between the Extent of Corruption and Percentage of Muslims Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Muslims is measured by the percentage of population. Figure 23: 3 Unpacking Corruption in Europe 130 Correlation between the Extent of Corruption and Percentage of Muslims across European Countries (Average) Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. The variable Muslims is measured by the percentage of population. The following table illustrates the descriptive statistics of the socio-cultural factors used in the following analysis. It illustrates the number of observations, means, standard deviations, and minimums and maximums of the included data. The number of observations is also based on the time period (1995-2013) and the number of European countries. The data was mainly collected from the Worldmark Encyclopedia of the Nations, the Statistical Abstract of the World, United Nations, the World Bank, OECD or United Nations (see Appendix B). Socio-Cultural Factors (1995-2013) Variables Observations Mean Standard Deviation Min Max Religion Percentage of Catholics 703 35.37 35.77 0 94 Percentage of Orthodox 703 22.28 34.59 0 98 Percentage of Protestants 703 17.00 28.75 0 95.2 Percentage of Muslims 703 5.80 13.72 0 70 Figure 24: Table 6: 3.1 Explaining Corruption at the Macro Level 131 Historical Factors The historical factors include two independent variables: a country’s years of democracy and the communist past (“communist legacies”). The following graphic illustrates the average development of a country’s years of democracy, compared to the average score of the extent of corruption in Europe. The variable communist past is a dummy variable and therefore illustrated by boxplots. Extent of Corruption and Historical Factors across Time Years of Democracy To measure a country’s democratic history, I used the number of consecutive years since 1930 when the system had been democratic, as classified by (Beck et al., 2001). This is adapted from Beck et al.’s variable “tensys”, which just measured tenure of the system, whether democratic or authoritarian. Democracies are those with a 6 or higher on Beck et al.’s Executive Index of Electoral Competitiveness. I take the data from the Quality of Government Dataset (Quality of Government, 2015). The average score of democratic years in Europe is 37. Georgia (18 years), Belarus (19 years) and Latvia (21 years) are comparatively young 3.1.2.4 Figure 25: 3 Unpacking Corruption in Europe 132 democracies in Europe. Particularly, Belgium, Denmark, Finland or France are with approximately 80 years the oldest democracies in Europe. The regression line of the scatterplots indicates a negative relationship between the extent of corruption and years of democracy, confirming the hypothesis that European states with longer democratic histories will have lower levels of corruption. This is confirmed by a negative correlation coefficient of -0.83, suggesting a strong negative relationship between a country’s experience of democracy and the extent of corruption (Figure 26 and 27). Correlation between the Extent of Corruption and Years of Democracy Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. Years of Democracy are measured by the number of consecutive years since 1930 when the system had been democratic. Figure 26: 3.1 Explaining Corruption at the Macro Level 133 Correlation between the Extent of Corruption and Years of Democracy across European Countries (Average) Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. Years of Democracy are measured by the number of consecutive years since 1930 when the system had been democratic. Communist Past To measure the history of communist rules in a country, I use a dummyvariable (1= communist past; 0= no communist past).62 I primarily collected data from the Worldmark Encyclopedia of the Nations. More than half of the European societies examined, 19 out of 37 countries, have a communist past. Countries with a post-communist past are Albania, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Finland, Georgia, Hungary, Latvia, Lithuania, Macedonia, Moldova, Poland, Romania, Slovakia, Slovenia and the Ukraine. A correlation coefficient of Figure 27: 62 Using dummy-variables does not present the best analysis approach, because it can be assumed that there are larger differences in the intensity of communist legacies between countries. However, currently, it is common practice in political science research (Pop-Eleches and Tucker, 2011). 3 Unpacking Corruption in Europe 134 0.69 indicates a positive relationship between the extent of corruption and a society’s communist past, suggesting that the extent of corruption in countries will be higher, if the country has a communist past. Boxplots of the Extent of Corruption and Communist Past Note: Dependent Variable: Extent of Corruption: Corruption Perceptions Index transformed: 0= low corruption; 10=highest level of corruption. Communist Past is measured by a dummy variable: 1= Communist Past; 0= no Communist Past. The following table reports the descriptive statistics of the historical factors used in the analysis. It illustrates the number of observations, means, standard deviations, and minimums and maximums of used data. The number of observations arise from the time period (1995-2013) and the number of European countries. The data was mainly collected from the Worldmark Encyclopedia of the Nations, the Statistical Abstract of the World, the United Nations and Transparency International (see Appendix B). Figure 28: 3.1 Explaining Corruption at the Macro Level 135 Historical Factors (1995-2013) Variables Observations Mean Standard Deviation Min Max Years of Democracy 592 35.78 28.01 0 80 Communist Past 592 .51 .50 0 1 The Impact of Country Characteristics on Corruption Table 8 presents the empirical findings of panel analysis at the macro level. The model includes 37 countries and 536 observations. Due to comparison reasons and for easily interpretation, the variables were standardized to a scale from minimum 0 to maximum 1. Therefore, the original variables are subtracted by its minimum value and subsequently divided by its maximum values. Furthermore, they include a time-lag of two years – the difference in time by which one observation lags behind or is later than another. The following model shows that all included independent variables have significant relationships with corruption, measured by the transformed CPI, in Europe and confirms the hypotheses of the macro level. Overall, the explained variance of the model (r²) is 0.87, meaning that almost 90% of the extent of corruption in Europe can be explained by the chosen economic, political, socio-cultural and historical factors. These results confirm previous studies and demonstrate the robustness of these factors in explaining corruption also for the European states. Macro Model of Corruption Variables Extent of Corruption in Europe Economic Development -0.425*** (0.116) EU-Membership -0.019*** (0.006) Degree of Democracy -0.123*** (0.036) Women in Parliaments -0.094*** (0.026) Percentage of Catholics 0.088*** (0.009) Percentage of Orthodox 0.176*** (0.013) Table 7: 3.2 Table 8: 3 Unpacking Corruption in Europe 136 Variables Extent of Corruption in Europe Percentage of Protestants -0.112*** (0.017) Percentage of Muslims 0.141*** (0.020) Years of Democracy -0.234*** (0.013) Communist Past 0.062*** (0.013) Constant 0.754*** (0.053) Observations 536 R-squared 0.87 Number of Countries 37 Note: Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Dependent Variable: “Extent of Corruption” (Corruption Perceptions Index transformed); 0= low corruption; 10= highest level of corruption. In this model, holding other factors constant, a country’s economic development is the most important contributor to the reduction of corruption levels in Europe. This result confirms the hypothesis that the extent of corruption will be higher in countries with lower levels of economic development. It identifies that when economic conditions improve, European countries are likely to improve their corruption scores, and experience less corruption. Furthermore, my findings indicate, that a country’s EU-membership tends to hinder an increase in corruption in European states. Especially after becoming a member in the European Union, the corruption scores of several countries such as Cyprus, the Czech Republic, Estonia, Latvia, Lithuania, Poland, Slovakia and Slovenia have significantly improved. This provides initial support that countries that are more integrated into international networks such as the EU are more exposed to economic pressure, maybe normative pressures as well, against corruption (Sandholtz and Gray, 2003; Kostadinova, 2012). This also implies that rules such as the Copenhagen Criteria, which define whether a country is eligible to join the European Union, seem to have a significant influence on a country’s level of corruption. The establishment of the rule of law in a country is not compatible with widespread levels of corruption, and if countries attempt to join the EU, they have to minimize corrupt activities as far as possible. Therefore, the admission of countries into organizations with high anti-corruption standards such as the European Union seems to 3.2 The Impact of Country Characteristics on Corruption 137 be an overall efficient anti-corruption instrument because international pressure tends to produce behavioral changes in countries regarding their corruption levels. This also confirms Kostadinova’s (2012, p. 240) assumption that “the desire to join the European Union was a much more effective driving force for implementation of anticorruption policies. […] Ironically, many people in the admittedly more corrupt Romania and Bulgaria think that only the Union can save them from corrupt politicians”. Moreover, the degree of democracy and the percentage of women in parliaments are influential in explaining the levels of corruption in Europe. The degree and duration of democracy are important contributors to the decrease in a country’s corruption levels. This result assists the assumption, in particular, of Hill (2003), Shah (2007), Billger and Goel (2009) and Saha et al. (2009). As expected, the model indicates that the extent of corruption is lower in countries that have higher levels of female participation in parliaments. This result is attributed to the fact that democracies are strongly related to greater gender equality. Gender equality may be conducive to democracy by promoting a less hierarchical cultural milieu for decision-making (Norris et al., 2002). Moreover, in democratic states, principles such as equality, fairness, transparency, checks and balances, and accountability are more strongly fostered than in authoritarian regimes that are characterized in particular by strong hierarchies. Therefore, specific components of liberal democracies such as the protection of women’s political rights and democratization processes include necessary conditions for honest governments because their institutions provide a fairer system and tend to include corruption-restraining mechanisms (Sung 2003). In other words, more advanced democratic structures and institutions and high percentages of women in parliaments lead to lower levels of corruption. These findings affirm the second question of the book “Is it all about democracy”. Moreover, my result confirm cultural approaches that claim democratic societies and polities are often committed to norms and values of justice and equal opportunities that are in opposition to corruption norms (Uslaner, 2006). Additionally, I analyze the interaction between a country’s degree of democracy and the percentage of women in parliaments in terms of reducing the extent of perceived corruption by multilevel models (see Table 11). Furthermore, the model illustrates that religion is a strong predictor of corruption levels in Europe and confirms my assumption that that countries with higher levels of Protestants are likely to be less corrupt. My findings indicate that societies with a higher percentage of individuals of 3 Unpacking Corruption in Europe 138 Catholic, Orthodox, and Muslim faiths show higher levels of corruption, while the relationship in Protestant societies such as Denmark, Sweden or Norway seems to be the opposite. My results lend credence to the argument that “In more hierarchical systems (for example, Catholicism, Orthodoxy and Islam), challenges to the status quo are less frequent than in more egalitarian or individualistic religions” (Dreher et al. 2007, p. 448). Theoretically, this association is often ascribed to egalitarian and individualistic features of Protestantism that facilitate the extent to which officeholders are held accountable for their actions. Thus, compared to other religions such as the Orthodox and Catholic churches as well as Islam, Protestant societies show less hierarchy and are less prone to tolerance towards power abuses and corrupt behavior. As a result, societies that indicate more egalitarian and individualistic features are more likely to show lower levels of corruption. This also suggests for the argument that democratic values such as equality decrease corruption levels. As assumed, there are significant relationships between historical factors such as the durability of democratic systems and a country’s communist past and the extent of corruption. This finding implies that democratic structures do not only decrease levels of corruption, but that this effect is also strengthened by the duration of democratic principles. In other words, the longer a democracy lasts, the less corrupt it is. The relationship between the duration of democracy and the extent of corruption is even stronger than the relationship with a country’s degree of democracy. I also analyze the interaction between the degree and duration of democracy with regard to the extent of perceived corruption at the individual level (Table 2). To conclude, democracies in Europe are not free of corruption per se and do not necessarily exhibit honest governments and politicians, but they have fewer problems with corruption reflecting the duration of democratic rule. Consequently, reducing levels of corruption would imply a change of specific practices and habits that are deeply embedded in a society’s culture and its institutions. In contrast to this, a country’s communist past significantly increases the extent of corruption in Europe. This indicates that a country’s communist past fosters the growth of corruption levels and that post-communist countries seem to still be susceptible to corrupt practices. This is in line with Skaaning (2009, p. 226) who assumes that ”as culture only changes slowly, the corrupt traditions have arguably survived the end of communist regimes. Communism is thus likely to have established a negative legacy. New bureaucracies were not created 3.2 The Impact of Country Characteristics on Corruption 139 from scratch, large extents of the personnel carried over, and enterprises as well as private people in general had 'internalized' certain practices.” Overall, these results strongly support sociological approaches that highlight cultural norms and values and focus on actors’ social behavior in institutions and societies. Robustness Check I have implemented several approaches to check the robustness of my results through empirical investigation. These attempts include, in particular, re-estimating the models with the Control of Corruption Index (transformed) as a dependent variable as well (Kubbe, 2013). The findings are very similar to each other (Appendix I1 - I10 in Kubbe, 2013). Moreover, where possible, I tested several alternative specifications of the independent variables such as a country’s international integration, the degree, and duration of democracy to reduce the danger of misspecification. The evidence has shown that the results are very similar, sometimes nearly identical. As supplementary test, I have also run a series of several OLS regression models and different forms of panel analyses as fixed-effects models63 aiming to eliminate unobserved constant factors, or random-effects models that take within- as well as between country variations into account (see Appendix K, L, M in Kubbe, 2013). Furthermore, I have run additional models with country and year-dummy variables to find out specific geographic and temporal characteristics within Europe. The panel analysis including country dummy variables makes clear, that country differences exist. Unfortunately, the model including year-dummies was insignificant because of collinearity problems. Finally, the linear regression models with panel-corrected standard errors performed in the best way and the main findings of the models remain robust throughout the changes. 63 A problem with fixed-effects models is that it excludes all countries without variation in the dependent variable. However, a Hausman test reveals that the unique errors are correlated with the independent variables, hence suggesting that fixedeffects models are preferable to random-effects models. 3 Unpacking Corruption in Europe 140 Explaining Corruption at the Micro Level Multilevel models, also known as mixed models, hierarchical linear or nested models, present an appropriate analytical procedure for analyzing corruption at certain levels. They are considered as generalizations of linear models, but can also be extended to non-linear models. As standard regression models multilevel modeling aims to study the relationship between a dependent variable and a set of independent variables. Overall, the data structure is hierarchical, and the sample data are viewed as multistage sample from this hierarchical population (Hox, 2002).64 By allowing for residual components at each level, multilevel modeling takes the existence of the hierarchical data structure into account. The most common types of multilevel modeling are random intercepts, random slopes, and random coefficient models.65 There are a number of reasons for using multilevel models in the research of corruption. Primarily, multilevel analyses afford researchers the opportunity to use data with certain levels of analysis simultaneously and focus on questions of how individuals are affected by their context, and of how higher levels structures merge from lower level events. Moreover, they allow studying effects that vary by the units of observations, and esti- 3.3 64 Hox (2002, p. 1) defines multilevel analysis as follows: “The general concept is that individuals interact with the social contexts to which they belong, meaning that individual persons are influenced by the social groups or contexts to which they belong, and that the properties of those groups are in turn influenced by the individuals who make up that group. Generally, the individuals and the social groups are conceptualized as a hierarchical system of individuals and groups, with individuals and groups defined at separate levels of this hierarchical system. Naturally, such systems can be observed at different hierarchical levels, and variables may be defined at each level. This leads to research into the interaction between variables characterizing individuals and variables characterizing groups, a kind of research that is now often referred to as 'multilevel research'.”. 65 Random intercepts models imply that the intercepts are allowed to vary across different groups or countries, after controlling for covariates. Assuming that the slopes are fixed, it implies that the scores on the dependent variable for every individual observation are predicted by the intercept that varies across groups. In contrast, in random slope models, the slopes are allowed to vary across certain groups or countries, implying that the slopes are different across these groups and that intercepts are fixed across different contexts. Random coefficient models include both random intercepts and random slopes and allow both to vary across groups, meaning that they are different in different contexts (Steenbergen and Jones, 2002; Jones, 2008). 3.3 Explaining Corruption at the Micro Level 141 mate group level averages. Of particular note is that multilevel research allows researchers to measure and use variables directly at their natural and not at the aggregate level (Rabe-Hesketh, 2012).66 In this manner, multilevel modeling offers an advanced instrument to explain why individuals within and across countries vary in their perception of corruption in Europe. The usual formular of multilevel analysis reads as follows: Yij = β0j + β1j(X1ij) + β2j(X2ij) + rij (Level 1 regression equation). Referring to this formular, Yij is the dependent variable for an individual observation at Level 1, subscript i refers to individual cases, j refers to the group /country, Xijr indicates the level 1 independent variable, β0j refers to the intercept of the dependent variable in country j at Level 2, β1j refers to the slope for the relationship in country j (Level 2) between the independent and the dependent variable at Level 1; rij refers to the random errors of prediction for the Level 1 equation. At the individual level, both the intercepts and slopes in the countries can be either fixed, meaning that all groups have the same values; nonrandomly varying, implying that the intercepts and / or slopes are predictable from an independent variable at Level 2; or randomly varying, meaning that the intercepts and / or slopes are different in the different groups, and that each have their own overall mean and variance. With regard to the Level 2 regression equation, the dependent variables are the intercepts and the slopes for the independent variables at Level 1 in the groups of Level 2. The formula used here are: β0j = γ00 + γ01Wj +u0j and β1j = γ10 + u1j. In this context, γ00 refers to the overall intercept. This is the grand mean of the scores on the dependent variable across all the groups when all the predictors are equal to 0. Wj refers to the Level 2 predictor; γ01 corresponds to the overall regression coefficient, or the slope, between the dependent variable and the Level 2 predictor; u0j refers to the random error component for the deviation of the intercept of a group from the overall intercept. Whereas γ10 corresponds to the overall regression coefficient, or the slope, between the dependent variable and the Level 1 predictor, u1j refers to the error component for the slope, implying that the deviation of the group slopes from the overall slope. 66 See also Raudenbush and Bryk (2002) and Hox and Roberts (2011). 3 Unpacking Corruption in Europe 142 Measuring Corruption at the Individual Level To measure corruption at the individual level, I use data from the World Values Survey that refer to the perception of corruption by individuals from multiple countries. The World Values Survey includes data of almost a hundred countries in five different rounds and allows the comparison of numerous countries worldwide. The dependent variable of the micro level is generated by asking citizens “How widespread do you think bribe taking and corruption is in this country?” Responses were recorded on a four-point scale where “1” implies “no public officials engaged in it”; 2 = “a few are”; 3 = “most are” and 4 implies “almost all public officials are engaged in it” (World Values Survey, 2015). This is in contrast to the data from the macro level based on survey data by experts. For this reason, I call the dependent variable of the micro level “extent of perceived corruption”. Using subjective perceptions while measuring culture dimensions and corruption are prone to bias. That is why, results have to be interpreted cautiously. This also applies for the interpretation of the macro level data. However, both dependent variables (macro and micro level) are highly correlated. Charron (2015) also demonstrates for Europe that corruption country rankings of experts are highly correlated with perceptions of citizens, even when controlling for a country’s economic performance, size of government, partisanship and population. For the following analyses, I use available data from three waves from the World Values Survey: 1994-1999; 1999-2004, and 2005-2009. This is nearly equivalent to the time period that is used at the macro level (1995-2013). Between these years, polls were conducted in 25 European societies. These countries encompass Albania, Bosnia and Herzegovina, Bulgaria, Belarus, Croatia, Czech Republic, Estonia, Finland, Georgia, Germany, Hungary, Latvia, Lithuania, Macedonia, Moldova, Norway, Poland, Romania, Slovakia, Slovenia, Spain, Sweden, Switzerland, the Ukraine and United Kingdom. In this context, I focus on the maximum numbers of countries that have sufficient observations at both levels macro and micro level, for all the variables being considered. In this way, I acquire corruption estimates from almost 30.000 respondents in 25 coun- 3.3.1 3.3 Explaining Corruption at the Micro Level 143 tries.67 The following illustrations (Figure 29 and Table 9) demonstrate how European societies distribute over the extent of perceived corruption. Extent of perceived corruption in Europe Similar to the extent of corruption at the macro level, figure 29 demonstrates that the perception of corruption varies widely across different European countries. The overall mean score on the four point scale in these 25 European countries is 2.88. The highest extent of perceived corruption are found in Macedonia (3.39), Lithuania (3.33), and Georgia (3.31), whereas the countries with the lowest extent of corruption turn out to be, in particular, the Scandinavian countries Norway (2.01) and Finland (2.18), followed by Switzerland (2.30), and Sweden (2.31). It is again striking, that there are significant differences between Western and Eastern European states. The average score of the extent of perceived corruption in Western European countries68 is 2.41. On the contrary, the average Figure 29: 67 Contrary to the panel analyses at the macro level, countries such as Austria, Belgium, Cyprus, Greece, Denmark, France, Iceland, Italy, Ireland, Luxembourg, Netherlands, Portugal and Spain have to be excluded. 68 The sample of Western Europe only includes Germany and Switzerland. As a result of the exclusion of a lot of Western countries such as Belgium, France or Luxembourg this sample is comparatively underrepresented. This only serves as an illustration. 3 Unpacking Corruption in Europe 144 corruption level of the Eastern societies69 is 3.09, and thereby comparatively higher. A comparison of Northern70 (2.63) and Southern Europe71 (2.89) show a similar picture. Notably, levels of corruption are not exceptionally lower in Southern Europe than in post-communist societies. These findings are very similar to the descriptive corruption results at the macro level. On the whole, European societies are characterized by diverging values of perceived corruption extent. Furthermore, table 9 gives an overall overview of the descriptive statistics of the dependent variable, including the number of observations, means, standard deviations, and minimums, maximums and coefficients of variance of the data. Extent of perceived corruption in Europe Countries Observations Mean (1994-2009) Standard Deviation Min Max Macedonia 900 3.39 .75 1 4 Lithuania 950 3.33 .68 1 4 Georgia 1850 3.31 .68 1 4 Ukraine 2539 3.30 .68 1 4 Belarus 1962 3.24 .62 1 4 Bulgaria 792 3.17 .71 1 4 Latvia 1137 3.11 .63 1 4 Slovakia 1009 3.08 .81 1 4 Moldova 905 3.08 .75 1 4 Czech Republic 1069 3.07 .81 1 4 Croatia 1027 2.91 .64 1 4 Poland 995 2.91 .78 1 4 Bosnia and Herzegovina 1006 2.90 .71 1 4 Estonia 911 2.89 .67 1 4 Spain 1125 2.88 .81 1 4 Hungary 615 2.88 .77 1 4 Albania 846 2.75 .67 1 4 Table 9: 69 Belarus, Bulgaria, Czech Republic, Moldova, Romania, Slovakia, Slovenia and the Ukraine and belong to the sample of Eastern Europe. 70 Northern Europe includes Estonia, Finland, Lithuania, Latvia, Norway and Sweden. 71 Albania, Bosnia and Herzegovina, Croatia and Macedonia and Slovenia belong to the sample of Southern Europe (United Nations Statistics Division, 2015). 3.3 Explaining Corruption at the Micro Level 145 Countries Observations Mean (1994-2009) Standard Deviation Min Max Romania 1101 2.71 1.08 1 4 Germany 1955 2.53 .66 1 4 Slovenia 893 2.52 .74 1 4 United Kingdom 1044 2.45 .70 1 4 Sweden 991 2.31 .76 1 4 Switzerland 1111 2.30 .73 1 4 Finland 910 2.18 .84 1 4 Norway 1099 2.01 .67 1 4 European Countries (Total) 28742 2.88 .82 1 4 The differences among the respondents in their estimation are still unexplained. According to cultural and economic approaches, it is assumed, that corruption is likely experienced differently depending on socio-demographic factors, social norms, values and attitudes. The following analysis takes also into consideration that people are not interchangeable: they are individuals and different inside and therefore respond differently to the same external factors. Not everyone is obliged to pay bribes, for example, and not everyone believes is endemic (Elster, 1989; Banuri and Eckel, 2012). These descriptive results presented in figure 29 and table 9 also illustrate the similarities to the levels of corruption at the macro level. More precisely, compared these corruption parameters at the individual level to the transformed CPI index, where on a 10 point scale, the mean score is 3.92 and the standard deviation is 2.33 in European states. Moreover, of particular importance in comparing the dependent variables of the macro and micro level is their high correlation. As already described in the second chapter about measuring corruption (chapter 2.3) the mean estimates of perceived corruption at the micro level correlate with estimates of corruption from the macro level indices such as the Corruption Perceptions Index. These high correlations reveal that experts and citizens whose aggregated views comprise the country score are in broad agreement regarding the extent of corruption in certain countries (see also Charron, 2015). My analysis confirms these results. The correlation between the Corruption Perceptions Index (transformed) and the aggregated item “Extent of perceived corruption” of the World Values Survey is 0.84. The correlation between the Control of Corruption Index (transformed) shows a similarly high value of 0.82. In fact, the correlation of both levels, macro and micro, 3 Unpacking Corruption in Europe 146 especially indicates the linkage between the country and the individual level offered by the bathtub model of corruption. Empirical Findings: The Impact of Socio-Demographic Characteristics, Values, Norms, and Attitudes on Corruption The multilevel models of corruption include seven independent micro level variables: gender, age, employment status, level of income, level of interpersonal trust, satisfaction with the financial situation and justification of bribery.72 The following table reports the overall summary of descriptive statistics of the individual-specific explanatory variables including socio-demographic factors, values, norms and attitudes used in the empirical analysis. It presents the number of observations, means, standard deviations, and minimums and maximums of the original data. The number of observations results from the time period (1995-2009) and the number of included European countries. The data was mainly collected from the World Values Survey. Socio-Demographic Characteristics, Values, Norms, and Attitudes (1994-2009) Variables Observations Mean Standard Deviation Min Max Socio-Demographic Characteristics Gender 59075 1.52 0.49 1 2 Age 59052 1955.58 17.15 1902 1991 Employment Status 55489 3.16 2.16 1 8 Level of Income 51304 128330.2 26.4143.6 -99 826040 Values and Norms Level of Interpersonal Trust 56381 0.29 0.45 0 1 Attitudes Satisfaction with Financial Situation 56491 5.08 2.59 1 10 Justification of Bribery 56874 1.82 1.74 1 10 3.3.2 Table 10: 72 One can easily think of better variables or items to cover the explaining factors of the “extent of perceived corruption.” However, these were the best available data and items in the WVS and the only items that were measured throughout the three consecutive waves. 3.3 Explaining Corruption at the Micro Level 147 Socio-Demographic Characteristics Gender To measure the variable gender, I use the three waves of the World Values Survey, covering a period from 1994-2009. The categories are 1) Male and 2) Female. A correlation coefficient of 0.02 illustrates a slightly positive relationship between the extent of perceived corruption and an individual’s gender in European states. This result does not confirm the assumed corruption-gender-nexus that there is a significant relationship between perceived corruption and an individual’s gender. Yet, figure 30 slightly indicates that on average females perceive a higher extent of corruption than males. Extent of perceived corruption and an Individual’s Gender Note: Dependent Variable: “Extent of perceived Corruption” is generated by asking “How widespread do you think bribe taking and corruption is in this country?” Responses were recorded on a four-point scale: “1” =“no public officials engaged in it”; 2 = “a few are”; 3 = “most are” and 4 = “almost all public officials are engaged in it.” Gender is measured by 1 = Male and 2 = Female 3.3.2.1 Figure 30: 3 Unpacking Corruption in Europe 148 Age To measure the variable age, I use the item “Year of Birth” from the World Values Survey (World Values Survey, 2015). Similar to the variable gender, the correlation coefficient is with 0.01 very low, indicating a weak positive relationship between the extent of perceived corruption and the age of individuals. However, this does not correspond to hypothesis that there is a significant relationship between perceived corruption and an individual’s age. The following graphic (Figure 31) illustrates that most of the respondents, independently of age, perceive high levels of corruption in their country. Those people think that most public officials are engaged in corrupt activities. However, almost 35% of elderly people, born between 1901 and 1925, assume that only “a few” public officials act corruptly. Extent of perceived corruption and an Individual’s Age Note: Dependent Variable: “Extent of perceived Corruption” is generated by asking “How widespread do you think bribe taking and corruption is in this country?” Responses were recorded on a four-point scale: “1” =“no public officials engaged in it”; 2 = “a few are”; 3 = “most are” and 4 = “almost all public officials are engaged in it.” The variable age is measured by “Year of Birth.” Figure 31: 3.3 Explaining Corruption at the Micro Level 149 Employment Status To measure an individuals’ employment status, I use data from the World Values Survey that includes the item: “Are you employed now or not? If yes: About how many hours a week? If more than one job: only for the main job.” The categories encompass, for example, “Full time”, “Part time”, “Self-employed”, “Retired”, “Housewife”, “Students” or “Unemployed” (World Values Survey, 2013). Yet, the correlation coefficient is, similar to the variables gender and age, also very weak with 0.04, implying that the hypothesis that an individual’s employment status influences the extent of perceived corruption initially has to be rejected. The following graphic (Figure 32) also illustrates that there is no significant relationship between a person’s (un)employment status and the extent of perceived corruption. Extent of perceived corruption and an Individual’s Employment Status Note: Dependent Variable: “Extent of perceived Corruption” is generated by asking “How widespread do you think bribe taking and corruption is in this country?” Responses were recorded on a four-point scale: “1” =“no public officials engaged in it”; 2 = “a few are”; 3 = “most are” and 4 = “almost all public officials are engaged in it.” The item “Employment Status” is measured by the following categories: “Full time”, “Part time”, “Self-employed”, “Retired”, “Housewife”, “Students” or “Unemployed.” Figure 32: 3 Unpacking Corruption in Europe 150 Level of Income To measure a person’s income, I take the following item from the World Values Survey: “Here is a scale of incomes. We would like to know in what group your household is, counting all wages, salaries, pensions and other incomes that come in. Just give the letter of the group your household falls into, before taxes and other deductions.” The categories encompass the decides for society: 1= Lowest decide, 10= Highest decide (World Values Survey, 2015). A correlation coefficient of -0.11 indicates a slightly negative relationship between the extent of perceived corruption and an individual’s level of income. Consequently, the hypothesis that there is a significant relationship between perceived corruption and an individual’s level of income in Europe is initially rejected. The data on level of income cannot be presented graphically because it is country-specific and therefore very complex. Values and Norms Level of Interpersonal Trust To measure an individual’s level of interpersonal trust, I use the WVS item: “Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?”. The categories encompass: “1 Most people can be trusted; 0 Can’t be too careful.” (World Values Survey, 2015). A correlation coefficient of -0.17 indicates a slightly negative relationship between the extent of perceived corruption and an individual’s interpersonal trust. This does not initially confirm the hypothesis that the level of interpersonal trust influences the extent of perceived corruption. Yet, figure 33 illustrates that individuals with low levels of interpersonal trust tend to perceive higher levels of corruption. 3.3.2.2 3.3 Explaining Corruption at the Micro Level 151 Extent of perceived corruption and an Individual’s Level of Interpersonal Trust Note: Dependent Variable: “Extent of perceived Corruption” is generated by asking “How widespread do you think bribe taking and corruption is in this country?” Responses were recorded on a four-point scale: “1” =“no public officials engaged in it”; 2 = “a few are”; 3 = “most are” and 4 = “almost all public officials are engaged in it.” Level of Interpersonal Trust is scaled from 1 (“Most people can be trusted”) to 0 (“Can’t be too careful”). Attitudes Satisfaction with Financial Situation To measure an individuals’ satisfaction with his or her financial situation, I use the WVS-item: “How satisfied are you with the financial situation of your household? If '1' means you are completely dissatisfied on this scale, and '10' means you are completely satisfied, where would you put your satisfaction with your household's financial situation?” (World Values Survey, 2015). A correlation coefficient of -0.27 suggests a slightly negative relationship between both variables. This is illustrated by figure 34. As a result, the hypothesis that that there is a significant relationship between percei- Figure 33: 3.3.2.3 3 Unpacking Corruption in Europe 152 ved corruption and an individual’s satisfaction with the financial situation cannot initially be confirmed. Extent of perceived corruption and an Individual’s Satisfaction with the Financial Situation Note: Dependent Variable: “Extent of perceived Corruption” is generated by asking “How widespread do you think bribe taking and corruption is in this country?” Responses were recorded on a four-point scale: “1” =“no public officials engaged in it”; 2 = “a few are”; 3 = “most are” and 4 = “almost all public officials are engaged in it.” The item “Satisfaction with Financial Situation” is scaled from 1 (“Completely dissatisfied”) to 10 (“Completely satisfied”). Justification of Bribery To measure an individual’s justification of bribery, I use the following item from the World Values Survey: “Please tell me for each of the following statements whether you think it can always be justified, never be justified, or something in between, using this card. (Read out statements. Code one answer for each statement): Someone accepting a bribe in the course of their duties.” The item is scaled from 1 (“Never justifiable”) to 10 (“Always justifiable”) (World Values Survey, 2015). A correlation coefficient of 0.08 indicates that there is no relationship between the extent of perceived corruption and an individual’s justification of bribery. Figure 34: 3.3 Explaining Corruption at the Micro Level 153 That is why the hypothesis that the level of the justification of bribery influences the extent of perceived corruption has initially to be rejected. Extent of perceived Corruption and an Individual’s Justification of Bribery Note: Dependent Variable: “Extent of perceived Corruption” is generated by asking “How widespread do you think bribe taking and corruption is in this country?” Responses were recorded on a four-point scale: “1” =“no public officials engaged in it”; 2 = “a few are”; 3 = “most are” and 4 = “almost all public officials are engaged in it.” The item “Justification of Bribery” is scaled from 1 (“Never justifiable”) to 10 (“Always justifiable”). The Impact of Individual Characteristics on Corruption To explain perceived corruption at the individual level, I run several multilevel models. After checking for multicollinearity, I specify a random intercept model, including micro level variables such as an individual’s socio-demographic characteristics, values, norms, and attitudes (Model 1), a model that integrates the significant micro level variables of model 1 and the macro level model (Model 2), and finally I estimate a model that additionally includes four cross-level variables (Model 3).These cross-level variables encompass an individual’s satisfaction with financial situation and a country’s degree of democracy; an individual’s interpersonal trust Figure 35: 3.4 3 Unpacking Corruption in Europe 154 and a country’s degree of democracy; the percentage of women in parliaments and a country’s degree of democracy; and a country’s duration and degree of democracy. As it is usual in multilevel models, the individual-level variables (except for dummies) are centered on country means, whereas the country-level variables are centered on the global mean. In summary, I standardize all variables into the same number format in order to establish comparability between variables that are originally measured in different schemes. Moreover, to adjust the estimation for the unequal probability of selection, sampling weights are assigned at one or both levels in the two-level model.73 The dependent variable at the micro level is measured by the WVSitem “How widespread do you think bribe taking and corruption is in this country?” (1994-2009). It has to be again stressed survey data may have inherent bias that complicate analyses and interpretations of findings and have to be seriously taken into account. Micro Model of Corruption Variables Model (1) Model (2) Model (3) Individual Level Gender 0.012 (0.016) Age 0.000 (0.000) Employment Status 0.008* (0.005) 0.007** (0.003) 0.005** (0.002) Table 11: 73 Sample sizes vary between countries. This biases estimates when individual-level data are pooled, giving countries with the larger samples more weight. The bias is undesirable because differences in sample size exist for reasons (such as available funds) that are of no theoretical interest. To eliminate the bias, samples must be weighted. In this context, Welzel (2013) suggests two possibilities. One possibility is to weight each country sample for the proportion of the world population it represents. This approach is appropriate when the universe to which one infers is defined as the world population. The alternative is to weight country-samples to equal size. This approach is appropriate when country-level conditions are analyzed as a source of variation of how people act and think. Then, population size is irrelevant (Welzel (2013, p. 86). I follow this approach throughout my multilevel analysis. Hence, in all pooled individual-level analyses, I weight country-samples to equal size without changing the overall number of cases. 3.4 The Impact of Individual Characteristics on Corruption 155 Variables Model (1) Model (2) Model (3) Level of Income -3.29 (2.31) Level of Interpersonal Trust -0.243*** (0.065) -0.168*** (0.024) -0.136*** (0.018) Satisfaction with Financial Situation -0.072*** (0.0119) -0.033*** (0.004) -0.034*** (0.004) Justification of Bribery 0.038*** (0.009) 0.021*** (0.006) 0.023*** (0.005) Country Level Extent of Corruption (CPI (transformed)) 0.100*** (0.037) 0.099** (0.040) Economic Development -0.077 (0.187) -0.191 (0.131) EU-Membership 0.062 (0.073) 0.014 (0.083) Degree of Democracy 0.037* (0.019) 0.039 (0.098) Women in Parliaments 0.007 (0.008) -0.002* (0.001) Percentage of Catholics -0.000 (0.001) -0.002* (0.001) Percentage of Orthodox -0.001 (0.003) -0.006*** (0.002) Percentage of Protestants -0.007*** (0.002) -0.012*** (0.002) Percentage of Muslims -0.003 (0.004) -0.002 (0.002) Years of Democracy 0.002 (0.004) 0.024* (0.014) Communist Past 0.090 (0.091) 0.098 (0.083) Cross Level Satisfaction with Financial Situation x Degree of Democracy 0.000 (0.001) Interpersonal Trust x Degree of Democracy -0.026*** (0.007) Women in Parliaments x Degree of Democracy -0.029*** (0.008) Duration x Degree of Democracy -0.008* (0.004) Constant -0.111 (0.072) 0.387 (1.109) 1.061 (0.912) Observations 19,289 21,210 21,210 3 Unpacking Corruption in Europe 156 Variables Model (1) Model (2) Model (3) Number of Countries 20 21 21 Variance (cons) 2.37 (1.02) 3.48 (1.11) 5.71 (4.80) Variance (Residual) .59 (.07) .55 (.03) .54 (.06) Between Country-Variation of Dependent Variable 81% 86% 92% Within Country-Variation of Dependent Variable 19% 14% 8% Note: Robust Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Dependent Variable: “Extent of perceived Corruption” is generated by asking “How widespread do you think bribe taking and corruption is in this country?”; Responses were recorded on a four-point scale: “1” =“no public officials engaged in it”; 2 = “a few are”; 3 = “most are” and 4 = “almost all public officials are engaged in it.” Individual-level variables (except for dummies) are centered on country means; Country-level variables are centered on the global mean; all variables are standardized into the same number format; Sampling weights are assigned at one or both levels in the two-level model. Models calculated with STATA 12.1. Data cover all societies and respondents from four waves of World Values Survey (1994-2009). Average sample size per country (standard deviation thereof). The results of all multilevel models demonstrate that, holding other factors constant, an individual’s employment status, level of interpersonal trust, satisfaction with the financial situation, and the justification of bribery are significant in the explanation of the extent of perceived corruption. While an individual’s employment status and justification of bribery show a positive relationship with the extent of perceived corruption, the variables level of interpersonal trust and satisfaction with the financial situation are negatively associated with perceived corruption (Model 1). Socio-demographic characteristics such as an individual’s gender, age and level of income do not show significant relationships with the perception of corruption. This means, that in terms of the gender variable, the assumption that there is a significant relationship between perceived corruption and an individual’s gender cannot be confirmed and contradicts studies by Swamy et al. (2001) and Dollar et al. (2001). Both of these authors demonstrate that women are less involved in corrupt transactions and are less likely to condone bribe-taking than men. This result rather supports the study of Alatas et al. (2009) and their argument that the perception of corruption is more culture-specific than gender-dependent. 3.4 The Impact of Individual Characteristics on Corruption 157 Interpersonal trust is constantly the strongest predictor of the extent of perceived corruption. This implies that people who have high levels of interpersonal trust show lower levels in the perception of corruption. These findings also corroborate the study of Moreno (2002) who has demonstrated a negative relationship between corruption permissiveness and support of democracy and interpersonal trust. Based on this result, my analysis indicates that trust seems to be a good control mechanism of corruption within a society. Generally, trust is a central component of social capital and a value that expresses the belief that others are part of your moral community (Uslaner 2006; Putnam 1993). Similarly, Manzetti and Wilson (2007) provided evidence that citizens living in countries with weak democratic institutions tend to support corrupt governments. As a result, these countries are likely to continue to maintain the status quo even though corruption is visible. In all models, an individual’s satisfaction with his or her financial situation indicates a negative relationship with the extent of perceived corruption. This implies that people who are unsatisfied with their financial situation perceive a higher extent of corruption of public officials. However, this result does not necessarily confirm the study of Torgler and Valev (2006) who demonstrated that people who are dissatisfied with their financial situation tend to be more willing to act illegally. It merely shows that these people perceive higher levels of corruption to exist or are more sensitive towards corrupt actions. Moreover, the coefficients of the variable do not show high values. For instance, as one moves up one unit on the variable satisfaction with the financial situation, the extent of perceived corruption is expected to decrease by around 0.07 points in model 1, and 0.03 in model 2 and 3. Model 2 that includes the significant variables of the macro level demonstrates that the extent of corruption and Protestantism are explanatory variables in terms of the extent of perceived corruption. While the extent of corruption, measured by the transformed CPI, has a positive association with the extent of perceived corruption, Protestantism reduces the perception of corrupt actions. In terms of the significant relationship of a country’s extent of corruption this implies that people living in countries with high levels of corruption also perceive higher levels of corruption. While in Protestant countries with a high number of women in parliaments that generally show lower levels of corruption, the individual perception of corruption is less likely as well. This again confirms the results of the macro model and demonstrates their robustness. 3 Unpacking Corruption in Europe 158 The findings illustrate that both the extent of corruption and its perception are culturally influenced and determine individuals’ actions. It seems that people have greater expectations and a higher estimated probability that, for instance, a given public official will engage in corrupt acts in societies with high levels of corruption (Fisman and Miguel, 2007). These results also clearly demonstrate the cultural transmission of corruption, which implies that individuals from societies in which corrupt transactions are quite common are more likely to engage in corruption and expect others to engage in it as well (Hauk and Saez-Marti, 2002). Additionally, there are significant relationships between three cross-level variables and the extent of perceived corruption (an individual’s interpersonal trust and a country’s degree of democracy; the percentage of women in parliaments and a country’s degree of democracy; and a country’s duration and degree of democracy). However, there is no significant relationship between the cross-level variable an individual’s financial and a country’s degree of democracy and corruption (Model 3). This confirms again the assumption that the extent of perceived corruption is strongly linked with a country’s degree and duration of democracy and is associated with a high level of interpersonal trust. These factors reduce the extent of corruption, while an individual’s financial satisfaction plays a minor role in explaining the perception of corrupt actions. As a result, corruption is a cultural, multilevel phenomenon that can be explained very well by sociological approaches that highlight cultural norms and values and focus on actors’ social behavior in communities, institutions and societies. Thereby, my analysis also confirms that culture interacts with corruption through two channels, formal (democratic) institutions, and informal institutions such as interpersonal trust (Banuri and Eckel, 2012). Furthermore, the models reveal that approximately between 80-90% of the variance is attributed to differences between countries and 10-20% to differences between individuals. This reveals that the roots of corruption seem to lie on the country level rather on the individual level. 3.4 The Impact of Individual Characteristics on Corruption 159

Chapter Preview

References

Abstract

International studies often point to Europe for low levels of corruption. However, recent scandals in nearly all European states illustrate that corruption continues to be on the rise. The author investigates the causes of corruption in Europe. The analysis indicates that a country’s contextual conditions such as the economic development, the degree and duration of democracy or historical factors like the post-communist past strongly influence Europe’s level of corruption. Furthermore, corruption is likely experienced differently depending on interpersonal trust and the justification of bribery. The findings reveal that a bundle of factors adding up to a specific “democratic culture” hinders the growth of corruption by generating strong democratic institutions and fostering citizen norms and values aimed at monitoring and sanctioning corrupt actors. As a result, democracy promotion is the best remedy against corruption spread in Europe.

Zusammenfassung

Auch wenn europäische Staaten vergleichsweise geringe Korruptionswerte aufzeigen, verdeutlichen Skandale immer wieder, dass Korruption ein großes Problem darstellt, mit dem auch Europa stark zu kämpfen hat. Die Autorin untersucht daher die Ursachen von Korruption auf dem europäischen Kontinent. Verschiedene Analysen zeigen, dass Kontextfaktoren eines Landes wie dessen ökonomischer Entwicklungsstand, der Demokratisierungsgrad und die jeweilige Dauer oder historische Faktoren wie die kommunistische Vergangenheit das Auftreten von Korruption stark beeinflussen.

Darüber hinaus spielen interpersonales Vertrauen und die Rechtfertigung von Bestechungszahlungen eine erhebliche Rolle in der Wahrnehmung von Korruption. Insgesamt zeigen die Befunde, dass letztendlich eine „demokratische Kultur“ der Schlüssel im Kampf gegen Korruption in Europa ist. Diese fördert demokratische Institutionen sowie Normen und Werte, die darauf abzielen, korrupte Akteure zu kontrollieren und sanktionieren.