Research Proposal Research topic Determinants and Pay-offs of Electoral Fraud in Russia Objectives 1. To investigate demographic, socio-economic and political determinants of electoral fraud within regions during 2003-2011 Parliamentary Elections. 2. To evaluate the monetary and non-monetary pay-offs from electoral fraud for the regions and local authorities. The main objective of the research is to detect electoral fraud during last 3 Federal Parliamentary Elections (2003-2011) and to explain the differences in its levels across regions and years. Electoral fraud is defined as an efforts to influence or control elections through illicit, clandestine manipulation of electoral processes" (McDonald 1972, p. 81). We consider electoral fraud to be similar to corruption according to the definition of World Bank (1997) as "a misuse of public authority for the private gain". Here the gain may come in various forms: political, social or monetary. We believe that similar to the research on rent-seeking activities electoral fraud can be explained by socio-economic and political factors. To measure the degree of fraud we employ a statistical approach based on previous studies by Myagkov et. al. (2005), Myagkov et al. (2007), Myagkov and Ordeshook (2008), Spilkin (2011), Vorobyev (2011). We use correlation between turnout and share of votes for the governing party (United Russia) as an indicator of electoral fraud. The measure of fraud is calculated as an estimated percentage of suspiciously assigned votes for the governing party United Russia. 1
Main determinants are derived from the suggested theoretical frameworks and are classified as motivational, deterrents and available resources and include various factors as: governor's affiliation to governing party "United Russia" and number of terms served in the position, government size, population, quality of law-enforcement, press freedom, size of middle class, budget expenditures, salaries of teachers, etc. Next we evaluate the possible pay-offs from electoral fraud. As suggested by Jarocinska (2010) who analyses regional data in Russia, high electoral results for the incumbent is strongly associated with future intergovernmental transfers and grants to the region. We expect that not only higher percentage of vote for the main Russian party (Edinaya Rossiya, ER) but especially the illicit manipulations that are rewarded afterwards with large grants and subsidies to the region. It is consistent with the idea of quota system that is the most common form of the electoral fraud according to McDonald (1972, p. 91). McDonald defines quota system as a situation when local authority takes an explicit or implicit cue from the incumbent that additional electoral support from this location will be rewarded, or in other words, "good" relations demand a certain "reciprocity". The hypothesis has been coroborated in the working paper of Kalinin and Mebane (2011) but with a different estimation technique and measure of fraud for federal elections during 1996-2008. As suggested by Kalinin and Mebane (2011) another motivation to commit fraud is a guarantee for the regional governor to keep his position. Therefore, we estimated the influence of fraud on governor's chances for re-appointment. Hypothesizes 1. A regional governor who belongs to United Russia supplies more of electoral fraud. 2. A regional governor in his first term does not have enough of experience to manipulate electoral outcomes. 2
3. Free mass media deters electoral corruption. 4. Big government is associated with more electoral fraud. 5. Budget expenditures provide more resources for falsifications. 6. Electoral fraud being a measure of regional loyalty to the center is associated with more intergovernmental grants. 7. Governors of the regions with falsified elections have a lower risk of not being appointed for the next term. Practical contribution Historically elections in USSR used to served primarily to allow elites to maintain social control" (Hermet et al. 1978). For a long time the main purpose of elections to keep government legitimate and accountable was persistently disregarded. There is no surprise that the beginning of the transition from authoritarian regime was marked by a low trust in such political institutions as elections and political parties (Pammet 1999, p. 46). Last ten years has not succeeded in building confidence of Russian voters: according to the opinion survey 1 performed by Levada Center in July of 2011 about 53% of respondents did not believe in fairness of upcoming Parliamentary election. After the elections in December the problem of public mistrust has out grown into massive protest meetings for fair elections, and additional measures against falsifications at the next federal elections in March. However even despite online monitoring during Presidential elections there were still electoral frauds 2. According to the survey by Fund of Public Opinion about evaluation of fairness of Presidential elections about one quarter of respondents asserted the elections as unfair 3. There are several reasons to believe that electoral fraud did occur to a certain extent. 1 http://www.bbc.co.uk/russian/rolling_news/2011/07/110727_rn_levada_duma_elections.shtml (in Russian) 2 As Vladimir Putin has mentioned during his meering with opposition on March 6: "elections did not go without some falsifications". 3 Available at http://fom.ru/politika/10360 (in Russian). 3
GOLOS association 4 gathered and documented multiple cases of violations and falsifications during last elections. Different evidence was provided by Shpilkin (2011) who employed statistical analysis to reveal abnormalities in the official electoral data. This method is not new and has previously been used by Myagkov et al. (2005) and Myagkov and Ordeshook (2008) as an argument to consider Russian elections to be suspiciously conducted. The lack of confidence in democratic institutions and occasional cheating during elections represent a persistent problem. But to improve the situation with elections it is extremely important to analyse determinants of electoral fraud. Our research will provide a theoretical framework and a solid econometric evidence on this issue. The comprehension of what causes electoral fraud is a necessary condition for building a strategy of actions. For example, if the connection between unfair elections in the region and the intergovernmental grants is established, it will raise an exigency of policy revision towards new criteria on federal budget reallocation. The future research can also supply additional argument of the importance of regional press freedom. Certain social-economic conditions when found to be correlated with the fraud may be used as indicator to increase the law-enforcement effort in the region to secure fair elections. And finally we will test if strong middle class is associated with fair elections as a democratic institution. Literature review From the beginning of 21 st century economic and political literature on the topic of electoral fraud has experience a rapid growth. And the reason is a formation of stable political systems in countries of the third wave of democratization. Corrupt election under any regime can be seen as rent-seeking activity. The theoretical studies see electoral fraud as a mechanism in a struggle of incumbent for 4 Available at the website http://www.kartanarusheniy.org/ (in Russian). 4
power. Sutter (2003) examines the effect of monitoring on the degree of fraud suggesting that less costly monitoring and high risk of protest in case of fraud can make elections cleaner. A working paper by Simpser (2008) attempts to explain very high victory margins of incumbent as a way to deter activity of opposition. Another paper by Schedler (2002) employs game-theory approach to model democratization process via elections. Then electoral fraud is one of strategic choices of incumbent to secure his position. Most of empirical research can be classified in two groups. First one comprises descriptive analysis of the facts of electoral corruption (e. g. McDonald, 1972; Pammet, 1999; Cox and Kousser, 1981, Reynolds, 1993). Second one is dealt with methods to detect and measure the level of fraud. Recently owning to advance in mathematical software and data technologies it became possible to thoroughly inspect and analyze electoral results with regard to statistical abnormalities. Then detected artifacts in data are referred to potential manipulations during electoral process. This direction of research is presented by Myagkov et al. (2005), Myagkov et al. (2007), Myagkov and Ordeshook (2008), Spilkin (2011), Vorobyev (2011) who draw the attention to the Russian elections and suspicious correlation between voter turnout and votes for the governing party. The Benford law that formulates the specific non-uniform distribution of digits in random data is another proposed statistical approach to test fairness of elections and it was examined by Deckert et al. (2011), Shikano and Mack (2011). To our knowledge the only econometric study that investigates determinants of electoral fraud is a working paper by Kalinin and Mebane (2011). They find a significant support to the hypothesis about connection between intergovernmental transfers and fraudulent elections among Russian regions. At the same time their measure of fraud lacks mathematical argumentation and statistical analysis heavily rely on cross-section estimations that require strong assumptions about homogeneity (in non-experimental research it is likely to be violated). We propose a better measure for falsifications based on previous econometric studies and employ a panel data to perform more robust fixed effects estimations. At the same time the main focus of our research will be broader 5
compared to Kalinin and Mebane (2011) as we will look at social-economic and political determinants of electoral corruption. Methodology a) Data The main variable of interest is a measure of electoral fraud. We develop a methodology in line with Myagkov et al. (2005), Myagkov et al. (2007), Myagkov and Ordeshook (2008), Spilkin (2011), Vorobyev (2011). Analysing three Parliamentary Elections in 2003, 2007, and 2011 one can notice a statistical anomaly when percentage of votes for the United Russia (UR) strongly correlates with voter turnout. Figure 1 illustrates that effect as a abnormal distribution of UR's votes over turnout across polling (ballot) stations in Russia. The asymmetric upturn on the right side indicates that votes for UR are driven by additional participation of voters. As a comparison Shpilkin (2011), Vorobyev (2011) demonstrate that in countries as Canada, Finland, Sweden where are elections are properly conducted the distributions of votes are symmetric and close to normal. Myagkov and Ordeshook (2008) explain the right upturn as a consequence of fraud in favour of UR during elections. This fraud occurs by the stuffing of ballot boxes with falsified ballots, compelling voters to vote who otherwise might prefer to stay home via various threats and forms of intimidation, or by simply adding to a candidate s total in official summaries without regard to votes actually cast (p. 6). When fraud as ballot stuffing occurs at a polling station the turnout rate necessarily increases and this polling station moves to the right tail in turnout distribution. Consequently all polling stations depicted on the right side of distribution give the incumbent an advantage in comparison to the ones on the left. Now it is possible to match left and right polling station to detect correlation driven by illegal manipulation, since any objective correlation between turnout and 6
incumbent's share of votes is equally present on both sides. Figure 1. Distribution of votes share for Governing party at polling station for 2003, 2007, 2011 Parliamentary Elections As suggested by Shpilkin (2011) we calculate the empirical coefficient of true correlation between vote share for UR and total turnout. For example, the Figure 2 is a scatter plot of data at the level of polling stations for Parliamentary Election in 2011 (approximately 96 thousands across Russia), it is possible to see that before 50% UR votes are proportionately to votes for other parties. Further we employ this coefficient to replicate the normal distribution of the left side to extract true magnitude of votes share for UR. 7
Figure 2. Scatter plot for Parliamentary Election of 2011. Figure 3 and Figure 4 demonstrate original and corrected realistic distribution of UR votes over turnout in 2011 elections. One can notice a distribution close to normal for the corrected data. For our measure of electoral fraud across regions we will employ this method to compute correct estimation of votes for UR and then the difference between original and corrected values will be the measure of fraud. Original data is available at the website of Central Election Commission of Russian Federation. The correction will be performed by commercial mathematical software Polyanalist. 8
Figure 3. Distribution plot of original votes over turnout in 2011 Parliamentary Election. Figure 4. Distribution plot of corrected votes over turnout in 2011 Parliamentary Election. 9
Information about research variables and data sources is present in Table 1. Table 1. Summary of Data Sources Variable Name of Data Source Website Governor's Affiliation with Governing Party Edinaya Rossia http://er.ru/ Number of Governor's terms served Analytical and Informational Resource http://governors.ru/?razdel=vdat®mode=regions Freedom of Mass Media Glasnost Defence Foundation http://www.gdf.ru/ Government size Federal State Statistic Service http://gks.ru/ Budget expenditures Federal Treasury of Russia http://roskazna.ru/ Intergovernmental grants and transfers Federal Treasury of Russia http://roskazna.ru/ Average income Federal State Statistic Service http://gks.ru/ Income inequality Federal State Statistic Service http://gks.ru/ Population size Federal State Statistic Service http://gks.ru/ Urbanisation rate Federal State Statistic Service http://gks.ru/ Unemployment rate Federal State Statistic Service http://gks.ru/ Ethnic polarisation National Census http://www.perepis-2010.ru/ http://www.perepis2002.ru/ Education level National Census http://www.perepis-2010.ru/ http://www.perepis2002.ru/ b) Theoretical framework We expect electoral fraud to be determined by following scheme: Motivation Leverage Electoral Corruption=. (1) Risks We assume electoral fraud to be sanctioned by local authorities, in particular regional governor. That allows the motivation behind falsification to be determined by certain characteristics of the governor such as his affiliation to United Russia and his age. Another part of motivation is an expected gain to the region for special effort in a form of intergovernmental grants. Leverage indicates the means to carry out a successful manipulation. For this reason available budgetary resources and the share of state-employed population matters, especially since low level of subordinates in the electoral process is responsible for implementation of the ballot stuffing. In this respect low salaries of teachers will decrease the cost of committing fraud. The category of risks comprises deterrents and restrictions for manipulation. Strong middle class and free mass media 5 tangibly increase the risk of being caught. It is also important to account for law-enforcement 6 in the region since falsification in Russia is a criminal act and supposed to be punished if detected. 5 Similar to Brunetti and Weder (2003) we expect mass media to deter fraud. 6 See Alt and Lassen (2010) for description of law-enforcement effect. 10
c) Estimation Panel data allows us to perform fixed effect (FE) estimation with clustered errors which disposes time invariant unobserved heterogeneity (Woolridge 2001, p. 267). Year specific effects will be addressed by year dummies to remove the correlation across regions for the same year which is likely due to the nature of our data. The regression analysis will be performed in STATA 11 software (xtreg -fe-, ro command). d) Expected research output The expected output will include significant estimates of our independent variables to support our proposed hypothesizes. References Alt, J. E. and Lassen, D. D. (2010).Enforcement and Public Corruption: Evidence from US States. (2010-08), Economic Policy Research Unit (EPRU), University of Copenhagen. Department of Economics. Brunetti, A. and Weder, B. 2003. A free press is bad news for corruption. Journal of Public Economics 87(7-8) 1801-1824. Cox, G. W. and Kousser, J. M. 1981. Turnout and Rural Corruption: New York as a Test Case. American Journal of Political Science 25(4) pp. 646-663. Deckert, J., Myagkov, M. and Ordeshook, P. C. 2011. Benford's Law and the Detection of Election Fraud. Political Analysis 19(3) 245-268. Hermet, G., R. R. R. A. 1978. Elections Without Choice. Wiley, New York. Jarocinska, E. 2010. Intergovernmental grants in Russia. Economics of Transition 18(2) 405--427. Kalinin, K. and Mebane, W. R. 2011. Understanding Electoral Frauds Through Evolution of Russian Federalism: From "Bargaining Loyalty" to "Signaling Loyalty". SSRN elibrary. McDonald, R. H. 1972. Electoral Fraud and Regime Controls in Latin America. The Western Political Quarterly 25(1) pp. 81-93. Myagkov, M., Ordeshook, P. and Shakin, D. 2007. The Disappearance of Fraud: The Forensics of Ukraine's 2006 Parliamentary Elections. Post-Soviet Affairs 23(3) 218--239. Myagkov, M., Ordeshook, P. and Shakin, D. 2005. Fraud or Fairytales: Russia and Ukraine's Electoral Experience. Post-Soviet Affairs 21(2) 91--131. Myagkov, M. and Ordeshook, P. C. 2008. Russian Elections: An Oxymoron of Democracy. SSRN elibrary. 11
Pammett, J. H. 1999. Elections and democracy in Russia. Communist and Post-Communist Studies 32(1) 45-60. Schedler, A. 2002. The Nested Game of Democratization by Elections. International Political Science Review 23(1) pp. 103-122. Shikano, S. and Mack, V. 2011. When Does the Second-Digit Benford s Law-Test Signal an Election Fraud? Facts or Misleading Test Results. Journal of Economics and Statistics (Jahrbuecher fuer Nationaloekonomie und Statistik) 231(5-6) 719-732. Shpilkin, S. 2011. Mathematics of Elections 2011. Troitskii Variant 94(25) 2-4. Simpser, A. 2008. Cheating Big: On the Logic of Electoral Corruption in Developing Countries. Mimeo, University of Chicago. Sutter, D. 2003. Detecting and Correcting Electoral Fraud. Eastern Economic Journal 29(3) 433-451. Vorobyev, D. (2011).Towards Detecting and Measuring Ballot Stuffing. (wp447), The Center for Economic Research and Graduate Education - Economic Institute, Prague. Wooldridge, J. M. 2001. Econometric Analysis of Cross Section and Panel Data. Vol. 1 The MIT Press. World, B. 1997. World Development Report: The State in a Changing World. Oxford University Press, New York. Project Participants Nikita Zakharov project leader. He recently graduated from Freiburg University where conducted an econometric research on corruption in Russia. Responsible for theoretical and econometric part. Dmitry Kogan project participant. He is currently employed as a senior software engineer at Software company. Responsible for the mathematical and programming part of the research, in particular for methodology and computation of the electoral fraud measure. Project Timetable Collection of data (2012 August October 7 ) Preparation of data and econometric estimations (2012 October-December) Composing a draft version (2012 December 2013 February) Competition of working paper (2013 February April) 7 Federal State Statistic Service will publish all necessary data for the year 2011 by October. 12