Взяточничество в регионах России: аналитическая заметка

Bribery in Russian regions:

A research note

Alexey Bessudnov

Department of Sociology,

Higher School of Economics, Moscow

bessudnov@gmail.com

March 2012

 

This is an early draft. Please do not cite.

1 State of the art and objectives

 

In the last fteen years, the cross-national study of corruption has been a vivid area of research in the social sciences. In a recent review, Treis-man (2007) summarizes the results of this research. Most of the studies used survey data on perceived corruption provided by the Transparency In-ternational. Not surprisingly, at the country level economically developed democracies are perceived as less corrupt. Freedom of the press, openness to trade and the absence of fuel exports are also associated with lower cor-ruption. However, the data reveal an interesting paradox. In contrast to perceived corruption, actual corruption experience is not associated with most of these factors, once controlled for country income. Treisman con-cludes that further research in this eld should be based on the indicators of experienced rather than perceived corruption.

Cross-national research on corruption is complicated by a number of factors. Good quality data are rarely available. Since the available data are cross-sectional, it is hard, if not impossible, to identify causal e ects of independent variables on corruption. Statistical models su er from the problems of omitted variable bias and reverse causation. Treisman (2007) attempted to estimate the causal e ect of income on corruption using the instrumental variable technique, but the choice of the instrument (country income in 1700) is not entirely convincing. In short, we know now about the factors associated with corruption at the country level, but the causal mechanisms for corruption remain unclear.

There is little research of corruption conducted at the cross-regional rather than cross-national level. Heterogeneity between regions within one country is usually lower than between countries. This can help reduce omit-ted variable bias and let us test the hypotheses about the causes of cor-

ruption in a di erent setting. However, it is hard to nd enough variation in corruption between regions of the same country. Corruption is less of a problem in Western Europe and North America, the areas where the social science data are usually of better quality. Many countries outside Europe and North America are simply too small to have enough regional variation in the level of corruption. Besides, cross-regional research requires large samples that would be representative for each region, and a large number of regions to provide enough power for statistical analysis. To the best of my knowledge, the data of this kind do not exist for China, India or Brazil, the countries that could be of interest for researchers. In this sense, Russia pro-vides a unique opportunity for the study of corruption at the cross-regional level. The number of regions is large enough to conduct meaningful statis-tical analysis.

Dininio and Orttung (2005) for the rst time attempted to analyze the di erences in corruption across the Russian regions with the data provided by the Transparency International and the INDEM foundation. This study is based on a survey conducted in 2002 in fourty Russian regions with a sam-ple that included 5,666 citizens and 1,838 representatives of small businesses. Dininio and Orttung based their analysis on the regional scores on the cor-ruption scale released by the TI/INDEM (n=40). They found two factors that were associated with corruption: the size of bureaucracy (positively) and GRP per capita (negatively).

Sharafutdinova (2010) used the same data set to di erentiate between experienced and perceived corruption and found that greater freedom of the press and higher political competition are associated with higher perceived corruption at the regional level, while higher GRP is associated with lower perceived corruption. Interestingly, actual corruption experience was not correlated with perceived corruption after controlling for the freedom of press, political competition, GRP and the proportion of pensioners.

Belousova et al. (2011) estimated OLS models for the TI/INDEM data set separately for experienced and perceived corruption. Wealthier regions were found to be perceived as less corrupt. Other factors that were asso-ciated with perceived corruption were population density and competition between rms (positively), and urbanization (negatively). For experienced corruption, the statistically signi cant predictors were regional wealth (lower experienced corruption in wealthier regions) and dummies for Moscow and St.Petersburg (where corruption was found to be higher).

Apart from the technical problems related to the identi cation of ro-bust statistical association and in particular causal e ects with small sam-ples and cross-sectional data, the studies of corruption that were mentioned above su er from a conceptual problem. They fail to distinguish between large-scale business and governmental corruption (bribes to top-level o – cials, kickbacks, etc.) and low-scale everyday corruption (smaller bribes or, called in a more neutral way, informal payments to doctors, teachers, etc.).

Both phenomena are assumed to have the same theoretical explanation and determinants. This assumption is not entirely convincing without further empirical proof. It may be the case that these two phenomena are governed by di erent social mechanisms.

The study that I propose seeks to improve the analysis by Dininio and Orttung (2005) and others in several ways. First, I use the data of a consider-ably better quality than in previous studies (see a more detailed description in the next section). Second, I have access to the individual-level data that allows me to model both individual and regional predictors of bribery and improves the quality of statistical analysis. Third, while focusing on experi-enced corruption only, I can separate the spheres (such as education, health, etc.) where the acts of bribery happened.

 

2 Data and methodology

The data for the study come from a survey conducted by the Public Opinion Foundation (known as FOM in Russia) in 2003. FOM is a Russian polling rm founded in 1992. It has an extensive experience of conducting surveys in Russia. For details see www.fom.ru (in Russian) or english.fom.ru (in English).

The survey was part of the GeoRating project that is speci cally focused on the comparison of Russian regions. A strati ed random sample was used to survey about 500 people aged over 18 in each of sixty- ve regions (n=32,537). The regions included in the survey represent about 90% of the Russian population (the excluded regions are mainly in the North and in the Caucasus). The interviews were conducted using the face-to-face method. The data are not in public domain, but FOM provided access to the data set.

The survey asked the following question: \Did you have to give money or gifts as a bribe? If yes, when was the last time this happened?” It also asked about the type of the problem people tried to solve with bribes. This allows me to separately look at corruption in the health system, education and the police. The data set also includes information about sex, age, education, income, and supervisory status of respondents and the sector of the economy where they were employed.

At this stage of the analysis, I apply simple linear probability models to model the association between the probability of bribing and individual-level predictors. Then I aggregate the data across the regions and use OLS regression for region-level models. Later both stages of the analysis will be combined within a multilevel model.

 

Table 1: The per-centage of people who bribed (un-weighted)

 

 

%

   
all

14.8

health

5.1

education

2.3

police

1.1

other

4.2

   

First, I present some descriptive statistics. Table 1 shows that in the sample about 15% of the respondents said that they had paid a bribe. Most of these bribes were related to the spheres of health (5%) and education (2%), with bribes to the police being in the third place (1%). Figure 1 shows the distribution of bribes by size. Most bribes are under 5,000 rubles (in 2003 prices), so it is clear that in this data set we mainly deal with low-level bribery, i.e. informal payments to doctors, teachers, the police, etc.

Figure 1: The distribution of bribes by size, FOM 2003

Figure 2 shows the results from the linear probability model for all bribes. The outcome variable equals one if the respondent said that he/she had bribed someone, and zero if he/she had not or refused to answer. The predictors are gender, age, location, income, education, number of people in the household and employment status. All quantitative variables (age, income, number of people in the household) were standardized with mean zero and standard deviation one.

Figure 2: Individual-level predictors of the probability of bribing

As the model shows, there is no di erence in the probability of bribing for men and women. Older people are much less likely to bribe (a one stan-dard deviation change in age reduces the probability of bribing by more than 10%). People living in big cities (the reference category) are more likely to bribe than people living in towns and in the countryside. Wealthier people are more likely to bribe. People with a higher and secondary specialized education also have a higher probability of bribing (compared with the ref-erence category, secondary completed education). Managers are more likely to bribe than clerks (the reference category) and manual workers.

The next three gures (3, 4 and 5) show the models for the individual-level predictors of bribery in the health system, education and the police (these are three elds where bribery is most widely spread, according to the data). The e ects of predictors are di erent for each sphere.

Women are more likely to bribe in the health system (as they probably visit doctors more often and also bring children) and, to a lesser extent, in education. Men are much more likely to bribe the police. People with a higher education are more likely to bribe in all three spheres, although

Figure 5: Individual-level predictors of bribery to the police

Figure 6: The regional intercepts from a multilevel model for the probability of bribing

for the police the e ect does not reach statistical signi cance. Income is positively associated with the probability of bribing in education and to the police, but not in the health system. Older people bribe less in all three spheres. People living in larger households tend to bribe more often (this is probably because larger households are households with children and thus they encounter doctors and teachers more often).

These statistical associations are not entirely surprising, but they con rm reliability of the data.

Next I turn to the analysis at the regional level. Figure 6 shows a map of regional intercepts, estimated in a multilevel model that accounts for all individual-level predictors, but allows the intercepts to vary for each region. The outcome variable is a dummy for having the experience of bribing. The map shows that there is substantial regional variability in the probability of bribing, even when all available individual-level factors are controlled for. In some regions, mainly in the South (such as Rostov, Krasnodar, Stavropol, Belgorod, but also Moscow and Moscow oblast), the probability of bribing is higher than in the others, mainly in the North (Arhangelsk, Perm, the Urals, etc.).

[to be continued]

References

Belousova, V., R.K. Goel, and I. Korhonen. 2011. \Causes of corruption in Russia: A disaggregated analysis.” .

Dininio, P. and R.W. Orttung. 2005. \Explaining patterns of corruption in the Russian regions.” World Politics 57:500{29.

Sharafutdinova, G. 2010. \What explains corruption perceptions? The dark side of political competition in Russia’s regions.” Comparative Politics 42:147{166.

Treisman, D. 2007. \What have we learned about the causes of corruption from ten years of cross-national empirical research?” Annu.Rev.Polit.Sci. 10:211{44.

Источник http://www.units.muohio.edu/havighurstcenter/conferences/documents/Bessudnovpaper.pdf

Следующая статья: /Годовой доклад: Amnesty International 2011/

На главную: /Бизнес иммиграция в США/

Задать вопрос?

×