Regional patrons and hegemonic party electoral performance in Russia

Abstract

A political scientist examines how regional elites shape the electoral fortunes of Russia's hegemonic party, United Russia (UR). Using original data on regional legislative elections from 2003 to 2011, we show that UR performs better in those regions where regional governors control strong political machines. Russia's leadership undercut its own electoral strategy by replacing popular elected governors with colorless bureaucrats who struggled to mobilize votes on behalf of United Russia. This is one of the reasons for United Russia's poor performance in recent elections.

Keywords:

• United Russia
• elections
• regional politics
• political machines
• political parties
• governors

View correction statement:

Erratum

Acknowledgements

This article is an output of a research project implemented as part of the Basic Research Program at the National Research University Higher School of Economics (HSE).

Notes

^1. To be sure, not all accounts of ruling parties omit elites from the equation. Brownlee (2007) argues that successful dominant parties are those in which rival elite factions put their differences to bed and cooperate within the framework of the ruling party. Slater (2011) also highlights the importance of elite collective action in the construction of a dominant party and further posits that elites will engage in this collective action when they feel threatened by endemic contentious politics.

^2. On the USA, see Banfield and Wilson (1963). On Latin America, see Kern (1973), and on Africa, see Clapham (1982). On southern Europe, see Chubb (1982), and on the former Soviet Union, see Hale (2003). And on South Asia, see Weiner (1967) and on South-east Asia, see Scott (1972).

^3. As nondemocracies, hegemonic party regimes tend to be characterized by (1) low levels of programmatic linkages between parties and voters, (2) the socio-economic dependence of citizens on the state, and (3) a dearth of nationwide collective action. This makes hegemonic party regimes comparatively fertile breeding grounds for both clientelism and regionalism.

^4. My analysis takes the existence of a nascent ruling party as a given (Reuter 2011, 2010). Other work directly tackles the question of why ruling parties emerge in the first place and why individual elites choose to affiliate with the new ruling party.

^5. At the time of writing, the number of regions had been reduced to 83.

^6. As Reuter and Remington (2009) point out, the communist opposition was on the decline when UR was formed.

^7. For an important exception to the lack of empirical work, see Golosov (2011).

^8. Of 166 elections in the sample, 23 were elected fully on the basis of party lists, and 23 were mixed with a higher proportion of party list seats. The remainder were held with an even balance between party-list and SMD seats.

^9. In the Appendix, I test the robustness of key results to an alternative specification that uses the total share of seats won by UR as the dependent variable.

^10. For elections held between 2003 and 2005, election results are averaged for the first and second halves of the year, respectively.

^11. Many accounts exist. See Fish (2005) for some examples.

^12. Thirty-four percent of appointed governors worked in a different region immediately prior to appointment (Buckley et al. 2012).

^13. Polling is not conducted in Altay Republic, Republic of Tuva, Chukotskiy Avtonomnyy Okrug, Republic of Sakha (Yakutia), the North Caucasus republics, or any of the autonomous okrugs except Khanty-Mansiysk. The private survey data were purchased from FOM.

^14. Results are robust only to analyzing elections from 2003 to 2010.

^15. For an important exception, see Greene (2010).

^16. Greene also analyzes the effect of bureaucratic quality on dominant party performance using expert rankings of meritocracy in the civil service. Such data are unavailable for the Russian regions.

^17. The method works by first estimating the following equation on regional monthly panel data stretching from December 2003 through December 2010: y _it  = βy _it − 1+βy _it − 2+f _iz +γt+u _i +ϵ _it , where i identifies regions, t is real time in months, and y is the logarithm of total government spending. Seasonality is controlled for by including a set of fixed effects f _iz for each of the 12 calendar months in each region. And γt is a linear term that accounts for increased spending over time. Region-specific deviations in spending are accounted for with a series of region-fixed effects, u _i . βy _it − 1 and βy _it − 2 are the first and second lag of the dependent variable. In other words, this equation estimates current spending in a region as a function of seasonality, time trends, previous spending, and region-specific patterns. The amplitude of the electoral budget cycle is defined as the residual of this equation for the month prior to the regional election.

^18. In the Appendix, I also test whether levels of, as opposed to changes in, regional government spending affect UR's vote shares and find no evidence for this proposition.

^19. This measure differs from Golosov's slightly. See the Appendix for an explanation and for robustness checks using Golosov's original measure.

^20. Opposition Fragmentation is calculated as one over the sum of squared vote shares received by the opposition (), where v _i is the share of the opposition votes received by the ith party. The quantity v _i is not the share of the total vote received by party i, but the share of the vote won by the opposition. Thus, this measure is similar to the Laakso–Taagepera index of the effective number of parties, with the sole exception that it restricts the calculation to the opposition's vote total.

^21. In democracies, it is well known that the number of ethnic groups is positively associated with the number of parties (Cox 1997).

^22. Nine regions held only one election in this period, three held three elections, and one, Sverdlovskaya Oblast, held five.

^23. Recent studies in political science with high N to T ratios such as ours have used this clustered Standard Error approach. See, for example, Golder (2006).

^24. The conditional coefficients in Column 2 are a good example of why Brambor et al. (2006, 76) warn that it is ‘extremely difficult and often impossible to evaluate conditional hypotheses using only the information provided in traditional results tables.’ As they note, ‘it is perfectly possible for the marginal effect of X on Y to be significant for substantively relevant values of the modifying variable Z even if the coefficient on the interaction term is insignificant’ (2006, 72). These results illustrate this point because the sign on Margin of Gov. Victory X URMember in the main results table is insignificant, and yet, from the conditional coefficients we see that this is due to the large standard error on Governor's Margin of Victory when URMember = 0. When URMember = 1, Governor's Margin of Victory has a substantively and statistically significant effect. This information could not be gleaned from the coefficient and standard error on Margin of Gov. Victory X URMember alone. Results are similar when interacting the other two measures of governor authority with URMember.

^25. In Table A3, I show results where Regional Ties and Governor Popularity are included in the same model. Both remaining significant.

^26. FOM conducted four surveys in 2003, two in 2004, two in 2005, four in 2006, four in 2007, four in 2008, four in 2009, and four in 2010. Surveys were conducted in 71 regions, although this number varies slightly by year.

^27. Results are robust to three commonly used alternative estimation methods for this type of data: (1) robust standard errors clustered on region, (2) a generalized estimating equation with exchangeable panel-specific error structure, and (3) Newey-West HAC standard errors (with errors correlated up to four lags).