What explains regional variation in election fraud? Evidence from Russia: a research note

Abstract

The December 2011 legislative election was among the most fraudulent national elections in Russia since the communist period. The fraud, however, was not evenly spread across the country. Precinct-level election returns from the 83 regions of the Russian Federation suggest that the level of fraud ranged from minimal or small in some regions to extreme in some others, with moderate to high fraud levels in many regions in between. We argue that in an electoral authoritarian context like Russia, regional variation in fraud can be explained by differences in (a) the perceived need by regional authorities to signal loyalty to the center by “delivering” desired election results; (b) the capacity of regional authorities to organize fraud; and (c) the vulnerability of citizens to political pressure and manipulation. We test the effect of signaling, capacity, and vulnerability on electoral fraud in the 2011 legislative elections with data on the 83 regions of the Russian Federation. We find evidence for all three mechanisms, finding that the tenure of governors in office, United Russia's dominance in regional legislatures, and the ethnic composition of regions are most important for explaining regional variation in electoral fraud.

Keywords::

• Russia
• elections
• election fraud
• regions

Notes

1. In March 2014, two additional regions became part of the Russian Federation – the Republic of Crimea and the federal city of Sevastopol – although these are recognized internationally as part of the territory of Ukraine.

2. For an overview and discussion of different conceptualizations of election integrity, see Van Ham (2014).

3. Not knowing what the actual results would have been in the absence of fraud, how to decide what level of turnout to consider as “anomalous” is subject to some controversy (Klimek et al. 2012). We propose that using the mean plus one standard deviation as a cutoff point is reasonable, especially since we use the actual election data that are likely biased upward due to successful fraud increasing turnout levels. Hence, if anything, our measure is likely to underestimate the extent of fraud that occurred. However, we also test our results using a dependent variable that takes the mean plus two standard deviations as the cutoff point for anomalous turnout. Replication datasets and the codebook are available from the authors.

4. Information about the tenure of governors was retrieved from the websites of the administrations of the 83 federation subjects.

5. Gervasoni (2010) finds that regions in Argentina whose income in part consists of rents in the form of government subsidies are less democratic than regions that do not rely on such rents because they are fiscally independent from their constituents. Because receiving government subsidies makes regions more dependent on the federal center, these recipient regions may also have an incentive to signal loyalty to the center, which they can do by delivering votes in elections. Goode (2007, 384), correspondingly, argues that governors from “debtor regions,” which receive funds from the federal budget, face more pressure than governors from “donor regions” that contribute to the federal budget.

6. Data on gross regional product for the 83 regions are available on the website of Roskomstat at http://www.gks.ru/free_doc/new_site/vvp/vrp98-11.xls.

7. Data on poverty, measured here as the amount of disposable income of households after the most necessary expenses, are available from http://www.mn.ru/multimedia_infographics/20120525/318904205.html; data on the ethnic composition of the federation subjects are available from www.gks.ru/free_doc/new_site/population/demo/per-itog/tab7.xls. The publication Regiony Rossii. Osnovnye kharakteristiki sub”yektov Rossiyskoy Federatsii, available from http://www.gks.ru/wps/wcm/connect/rosstat_main/rosstat/ru/statistics/publications/catalog/doc_1138625359016, contains data for 2011 on urbanization (pp. 58–59), share of employed people working in state enterprises (115–116), and use of the Internet (683–684).

8. Urbanization is left out in these analyses due to multicollinearity with self-generated income.

9. The results for the models reported in Table 1 were also checked leaving out the regions with less than 100 electoral precincts; this did not change the results.