This article explores two prominent narratives through an analysis of election data. The first narrative is that Yanukovych’s win was legitimate. While fraud may have been present, its scale was small and it was not decisive. The second narrative suggests that Ukraine’s major operational political cleavage separates eastern and western regions, rendering the central region of the country a crucial prize for candidates to secure victory in presidential contests.
Analysis of earlier presidential elections has noted that Leonid Kuchma was victorious in 1994 by garnering votes in the eastern half of the country along with central regions in the western half, and in 1999 by capturing the west and some areas east of the Dnipro River.
2.
The article uses eight macro-regions defined by historical, social, and political similarities (Lowell W. Barrington and Erik S. Herron, "One Ukraine or Many? Regionalism in Ukraine and Its Political Consequences," Nationalities Papers 32:1[2004]: 53-86). The macro-regions are East (Donetsk and Luhansk), Eastcentral (Dnipropetrovsk, Kharkiv, and Zaporizka), Crimea (Crimea and Sevastopol), South (Kherson, Mykolayiv, and Odesa), Northcentral (Chernighiv, Cherkassy, Kirovohrad, Kyiv City and Oblast, Poltava, and Sumy), West (Ivano-Frankivsk, L’viv, and Ternopil), Westcentral (Khmelnitsk, Rivne, Vinnytsya, Volyn, and Zhytomyr), and Southwest (Chernivtsy and Zakarpatska). For clarity, the macro-regions will be capitalized while general directional terms will not be capitalized.
3.
This analysis uses data from 33,683 polling stations compiled from results provided by Ukraine’s Central Electoral Commission. The total number of polling stations slightly differs between the two rounds, likely due to additional ships serving as special polling stations in Odesa during the second round. (The first round featured 33,673 polling stations.) Analysis using polling station data must be interpreted with caution as these data do not reveal the attitudes or behaviors of individual voters, rendering the interpretation particularly vulnerable to problems of ecological inference. However, polling station data can reveal variation in micro-regional results and can thereby provide additional insights into election dynamics.
4.
See http://www.osce.org/documents/odihr/2010/02/42679_en.pdf.
5.
See http://www.interfax.com.ua/eng/press-conference/31465/.
6.
Both campaigns traded allegations about fraud, especially in the first round.
7.
See http://www.osce.org/documents/odihr/2010/02/42816_en.pdf.
8.
Cf. Mikhail Myagkov and Peter C. Ordeshook, "The Trail of Votes in Russia’s 1999 Duma and 2000 Presidential Elections," Communist and Post-Communist Studies 34:3(2001): 353-70. See also Mikhail Myagkov, Peter C. Ordeshook, and Dmitry Shakin, "Fraud or Fairytales: Russia and Ukraine’s Electoral Experience," Post-Soviet Affairs 21:2(2005): 91-131.
9.
Erik S. Herron and Paul Johnson, "Fraud before the ‘Revolution’: Special Precincts in Ukraine’s 2002 Parliamentary Election," in Ingmar Bredies , Valentin Yakushik, and Andreas Umland, eds., Aspects of the Orange Revolution III: The Context and Dynamics of the 2004 Ukrainian Presidential Elections (Stuttgart, Germany: Ibidem-verlag, 2008).
10.
A.A. Sobyanin and V.G. Sukhovol’skiy, Demokratiya, Ogranichennaya Falsifikatsiyami: Vybory i Referendumy v Rossii v 1991-1993 Gg ( Moscow: Project Group for Human Rights, 1995).
11.
Valentin V. Mikhailov, Osobaya Zona: Vybory v Tatarstane ( Ulyanovsk, Russia: Kazan Branch of the International Assembly to Protect Rights, 2000); and Mikhailov, "Kolichestvo Demokratii: Analiz Vyborov Prezidenta RF 1996 G, v Regionakh,"Armageddon3:1(1999): 134-53.
12.
The first-digit probabilities in the Benford distribution are 1 = .301, 2 = .176, 3 = .125, 4 = .097, 5 = .079, 6 = .067, 7 = .058, 8 = .051, and 9 = .046.
13.
Walter R. Mebane, "Election Forensics: Vote Counts and Benford’s Law" (Paper presented at the 2006 Summer Meeting of the Political Methodology Society, University of California-Davis , July 2006); and Mark J. Nigrini, "Monitoring Techniques Available to the Forensic Accountant," Journal of Forensic Accounting 7:2( 2006): 321-44.
14.
Some aspects of election data present challenges to the application of Benford’s Law. Data in polling stations are constrained by the lower bound of zero but an upper bound that varies from place to place, depending on the number of individuals registered. This feature of polling station data produces different "available" digits across polling stations (i.e., more 1s are available in a precinct with 2,000 registered voters than in a station with 1,000 registered voters). Polling stations can also produce single-digit or zeros as results, undermining analysis of first and second digits. The application of Benford’s Law confronts not only statistical problems, but also issues related to the interpretation of results. The presence of anomalous data does not constitute proof of fraud. Indeed, seemingly anomalous results could be produced in natural strongholds for candidates or by a candidate’s own poor performance. Lastly, Benford’s Law is unlikely to identify diffuse or low scale fraud.
15.
Erik S. Herron, Elections and Democracy after Communism? (New York, NY : Palgrave Macmillan, 2009).
16.
In previous elections, some national-level data failed to conform to a Benford-type distribution (notably the Communist Party in 2002’s constituency races and Viktor Yanukovych in the third round of the 2004 presidential election). Nonconforming results were present in the East, Eastcentral, South, and Northcentral regions in past elections, but this analysis was conducted at the oblast level (see ibid.).
17.
The correlation coefficients are all significant at the .05 level and are .1800 for Tymoshenko, -.1536 for Yanukovych, and -.3142 for "against all."
18.
The decisive votes were cast in prisons (Herron and Johnson, "Fraud before the ‘Revolution’").
19.
The distribution of polling stations reporting 100 percent turnout is as follows: 12 in Crimea, 15 in Vinnytsya, 8 in Volyn, 53 in Dnipropetrovsk, 88 in Donetsk, 6 in Zhytomyr, 1 in Zakarpatska, 20 in Zaporizka, 8 in Ivano-Frankivsk, 9 in Kyiv Oblast, 8 in Kirovohrad, 43 in Luhansk, 27 in L’viv, 8 in Mykolaivsk, 53 in Odesa, 15 in Poltava, 13 in Rivne, 11 in Sumska, 12 in Ternopil, 25 in Kharkiv, 6 in Kherson, 9 in Khmelnitsk, 4 in Cherkaska, 3 in Chernivets, 4 in Chernighiv, 13 in Kyiv City, and 3 in Sevastopol.
20.
For example, the author has witnessed polling station commissions evaluate ballot papers with text such as "they are all thieves" written on the paper in lieu of a recorded vote.
21.
Sobyanin and Sukhovol’skiy, Demokratiya, Ogranichennaya Falsifikatsiyami.
22.
In the second round of the 2004 presidential election, the author observed a polling station commission deem ballots with additional marks valid for Viktor Yanukovych, but analogous ballots invalid for Viktor Yushchenko.
23.
The results of a paired two-sample t-test indicated that the outcomes are significantly different: t = 19.54.
24.
Polling stations with low levels of ballot invalidation were distributed in the following way: 76 in Crimea, 240 in Vinnytsya, 228 in Volyn, 47 in Dnipropetrovsk, 196 in Donetsk, 232 in Zhytomyr, 29 in Zakarpattya, 81 in Zaporizka, 90 in Ivano-Frankivsk, 133 in Kyiv Oblast, 142 in Kirovohrad, 105 in Luhansk, 333 in L’viv, 122 in Mykolaivsk, 156 in Odesa, 172 in Poltava, 122 in Rivne, 123 in Sumska, 259 in Ternopil, 132 in Kharkiv, 91 in Kherson, 227 in Khmelnitsk, 137 in Cherkaska, 22 in Chernivets, 136 in Chernighiv, 39 in Kyiv City, 10 in Sevastopol, and 54 abroad.
25.
Polling stations with high levels of ballot invalidation were distributed in the following way: 12 in Crimea, 14 in Vinnytsya, 5 in Volyn, 17 in Dnipropetrovsk, 11 in Donetsk, 16 in Zhytomyr, 11 in Zakarpattya, 8 in Zaporizka, 14 in Ivano-Frankivsk, 27 in Kyiv Oblast, 12 in Kirovohrad, 11 in Luhansk, 10 in L’viv, 6 in Mykolaivsk, 27 in Odesa, 14 in Poltava, 10 in Rivne, 11 in Sumska, 5 in Ternopil, 11 in Kharkiv, 7 in Kherson, 11 in Khmelnitsk, 10 in Cherkaska, 7 in Chernivets, 7 in Chernighiv, 5 in Kyiv City, and 1 in Sevastopol.
26.
Polling stations reporting 5 percent or greater "improvements" in ballot completion were distributed in the following way: 12 in Crimea, 28 in Vinnytsya, 13 in Volyn, 49 in Dnipropetrovsk, 19 in Donetsk, 23 in Zhytomyr, 36 in Zakarpattya, 30 in Zaporizka, 10 in Ivano-Frankivsk, 33 in Kyiv Oblast, 19 in Kirovohrad, 23 in Luhansk, 24 in L’viv, 23 in Mykolaivsk, 48 in Odesa, 32 in Poltava, 22 in Rivne, 20 in Sumska, 20 in Ternopil, 26 in Kharkiv, 15 in Kherson, 36 in Khmelnitsk, 25 in Cherkaska, 9 in Chernivets, 22 in Chernighiv, 14 in Kyiv City, and 1 in Sevastopol.
27.
Polling stations reporting 5 percent or greater increase in the ballot invalidation rate were distributed in the following way: 7 in Crimea, 13 in Vinnytsya, 4 in Volyn, 12 in Dnipropetrovsk, 11 in Donetsk, 13 in Zhytomyr, 3 in Zakarpattya, 7 in Zaporizka, 5 in Ivano-Frankivsk, 22 in Kyiv Oblast, 7 in Kirovohrad, 9 in Luhansk, 6 in L’viv, 3 in Mykolaivsk, 21 in Odesa, 14 in Poltava, 7 in Rivne, 9 in Sumska, 4 in Ternopil, 9 in Kharkiv, 4 in Kherson, 7 in Khmelnitsk, 8 in Cherkaska, 3 in Chernivets, 6 in Chernighiv, 6 in Kyiv City, and 1 in Sevastopol.
28.
Correlation coefficients are significant at the .05 level and are -.107 for Tymoshenko and -.142 for Yanukovych.
29.
Cf. Mikhailov, Osobaya Zona.
30.
Polling stations reporting 2 percent or greater vote loss for Tymoshenko were distributed in the following way: 1 in Cherkaska, 1 in Dnipropetrovsk, 6 in Donetsk, 1 in Ivano-Frankivsk, 2 in Kharkiv, 1 in Kyiv Oblast, 13 in Luhansk, 1 in Mykolaivsk, 1 in Odesa, 2 in Poltava, 1 in Sevastopol, 1 in Sumska, and 3 in Zaporizka.
31.
Polling stations reporting 2 percent or greater vote loss for Yanukovych were distributed in the following way: 1 in Chernighiv, 1 in Dnipropetrovsk, 4 in Donetsk, 2 in Kharkiv, 2 in Luhansk, 1 in Odesa, 1 in Sevastopol, and 3 in Zaporizka.
32.
Cf. Ralph S. Clem and Peter Craumer, "Shades of Orange: The Electoral Geography of Ukraine’s 2004 Presidential Elections," Eurasian Geography and Economics 46:5(2005): 364-85; and Paul Kubicek, "Regional Polarisation in Ukraine: Public Opinion, Voting and Legislative Behaviour," Europe-Asia Studies 52:2(2000): 273-94. See also Clem and Craumer, "Orange, Blue and White, and Blonde: The Electoral Geography of Ukraine’s 2006 and 2007 Rada Elections," Eurasian Geography and Economics 49:2(2008): 127-51.
33.
Three hundred ninety polling stations showed gains in turnout three or more standard deviations from the mean; 1,059 showed gains two or more standard deviations from the mean.
34.
This figure includes results for polling stations located abroad.
35.
Tymoshenko’s district-level gains (measured by raw votes) are 3 standard deviations from the mean in 5 districts (4 in L’viv, 1 in Ternopil), 2 to 3 standard deviations from the mean in 6 districts (2 in Ivano-Frankivsk, 3 in Kyiv City, 1 in Volyn), 1 to 2 standard deviations from the mean in 20 districts (1 in Khmelnitsk, 6 in Kyiv City, 4 in L’viv, 4 in Ivano-Frankivsk, 1 in Rivne, 1 in Vinnytsya, 1 in Sumska, 1 in Chernivets, 1 in Kyiv Oblast), 0 to 1 standard deviation from the mean in 50 districts, 0 to -1 standard deviation from the mean in 134 districts, and beyond -1 standard deviation from the mean in 11 districts (2 in Donetsk, 9 in Luhansk).
36.
Yanukovych’s district-level gains (measured by raw votes) are greater than 2 standard deviations from the mean in 9 districts (1 in Kharkiv, 3 in Odesa, 2 in Donetsk, 1 in Dnipropetrovsk, 2 in Luhansk), 1 to 2 standard deviations from the mean in 32 districts (7 in Dnipropetrovsk, 2 in Sevastopol, 6 in Kharkiv, 2 in Mykolaivsk, 4 in Zaporizka, 3 in Crimea, 1 in Luhansk, 6 in Donetsk, 1 in Odesa), 0 to 1 standard deviations from the mean in 52 districts, 0 to -1 standard deviations from the mean in 100 districts, and beyond -1 standard deviations from the mean in 33 districts (1 in Zakarpattya, 2 in Vinnytsya, 2 in Khmelnitsk, 3 in Rivne, 5 in Ternopil, 4 in Volyn, 8 in L’viv, 6 in Ivano-Frankivsk, 1 in Chernivets, 1 in Sumska).
37.
The correlation coefficients are all significant at the .05 level. They are -.2458 for Tymoshenko, .1944 for Yanukovych, and .6126 for "against all" votes.
38.
The correlation coefficients are all significant at the .05 level. They are .0986 for Tymoshenko, -.2931 for Yanukovych, and .7214 for "against all" votes.
39.
The correlation coefficients for Tymoshenko and "against all" votes are significant at the .05 level. They are -.1594 for Tymoshenko and .4352 for "against all" votes.
40.
The correlation coefficients are all significant at the .05 level. They are .5150 for Tymoshenko, -.5295 for Yanukovych, and .1575 for "against all" votes.
41.
The correlation coefficients are all significant at the .05 level. They are .5638 for Tymoshenko, -.5545 for Yanukovych, and -.1249 for "against all" votes.
42.
The correlation coefficients are all significant at the .05 level. They are -.4061 for Tymoshenko, .4025 for Yanukovych, and .0526 for "against all" votes.
43.
The Southwest is not included in Tymoshenko’s areas of support due to results in Zakarpattya, which have been attributed to the involvement of Viktor Baloha.
44.
Southwestern districts in Chernivtsy and Zakarpattya also registered turnout declines.
45.
The Northcentral region is excluded as the comparison category.
46.
The data were analyzed using regression with robust standard errors and seemingly unrelated regression (SUR). In the former approach, the data were also clustered by district to account for error correlation. Supplementary tables with full results are available at http://vse-na-vybory.blogspot.com.
47.
The interpretation of the table requires several caveats. The results are point predictions but are accompanied by confidence intervals. Thus, the point estimates are an approximation. Moreover, as noted in the text, other explanatory variables are held at zero to emphasize the independent effects of turnout and region on results.
