Petukhova Elena
“Niccolo M” Group of Companies

Quantitative methods in predicting voting result.
(On the case of governor’s election campaign, Russia).

The prediction of results of elections is a problem not only in Russia but also in other countries. We propose the procedure of predicting the results of election worked out on the base of governor’s election in one of Russian regions (Tula) held in March of 2001.

Specialists of the Center “Niccolo M” participated in that election as consultants. The election campaign had lasted for 3 months until the Election Day – 8^th of April. Every two weeks “Niccolo M” conducted the quantitative survey on a sample about 2500 respondents all over Tula Region. So, information was gathered out of 10 surveys.

1. Popularity of candidates.

Figure 1.

The first indicator of candidates’ potential to win is their popularity or how many people know them. As for those three candidates in electoral campaign in Tula Region the most famous was the incumbent governor Star. His popularity was more than 90%. Other two candidates, Sokol and Samoshin were less known. Sokol (red line, figure 1) started from 50%. Samoshin (blue line, figure 1), the head of his District, was known for only 10% of electorate. The diagram demonstrates candidates’s election campaigns. Sokol and Samoshin were very active in their campaigns. At the end of the campaign all three candidates had equal popularity.

The second indicator is trust – respondents answering question: “Do you trust this candidate?” The figure 2 shows the changing attitude towards each candidate from 10th of Jan. to 2nd of Apr.

Figure 2.

As we can see attitude towards Star and Sokol were almost constant, no dramatic changes had happened. There was about 30% of electorate who trusted Sokol. And at the end of electoral campaign more people confided in Star. The most significant change took place in people’s attitudes towards Samoshin (blue line). His intensive campaign firstly led him on the top of the list (10^th of February), and then it decreased to the level of Sokol. These changes were determined by the direction of candidates’ electorate campaigns. Till March they conducted the positive campaigns, they told about their business, family, future plans and so on. In March they started to fight.

The next figure shows the most important data – voting. Voting data is close to data on “Trust”. The pick of popularity of Samoshin is 27% in the middle of February. After that he started to loose his preference. So in this period something happened that changed the situation. In that period Star became more and more popular. To the end of the campaign his electorate was twice more than his competitors’. The final result of survey dated 2nd of April, showed the following: Star had 34% of voters, Sokol had 17% and Samoshin had 18%.

Figure 3.

The question “Which candidate you will never vote for?” helps to determine the percent of people who will never vote for the candidate. As data shows in that campaign Star was absolutely inappropriate for more than a third of electorate. He started from 37% of negative voting, at 18^th of February his anti-electorate was the biggest and then he started to win positions. At the end of the campaign about 27% of electorate said that they would never vote for him. As for other two candidates, they had much less “anti-voters” than Star. But their dynamics were negative. The number of people who did not want to vote for them has been increasing, especially after 18^th of February. As was mentioned above, these two candidates started to fight with each other while Star was not taken into consideration as a leader.

Figure 4.

Rating of candidates’ ascribed chances (Question: “How do you think who will win the elections?”) gives the opportunity to calculate and predict the decision making of people who will come for voting but till the last day have not decided yet who they will vote for. About 5-10% of voters make their decision on polling station. Part of them make rational choice. Their reason is not to vote for predetermined looser. So, Star was perceived as a winner of elections for the whole campaign. He took the lead at the rating of 23% and finished at 37%. Other candidates were perceived as winners mostly in the middle of campaign (Figure 5).

Figure 5.

Figure 6.

Voting results depends on people’s electorate activity. Analysis of candidates’ electorate showed that their activity is different. Sokol and Samoshin were aspired to the same electorate that was not very active in compare to Star’s electorate. Their electorate consisted of people with higher education, young and middle aged, which live mostly in towns. Star’s electorate was people, which lived in suburbs, older-aged, which get used to vote since Soviet times. So, predicting of the turnout is the starting point of predicting the results of elections.

The number of people, who said in a last survey that they surely vote, was 72% (last column, Figure 6). This number seemed to be very high. Russian local elections are usually attended by less than 50% of electorate. Also we asked people if they had voted in previous elections. The statistics says that Tula turnout on presidential election in 2000 was 67% while 83% of respondents said they had voted. The difference is 18%! The turnout on Parliament elections in 1999 was 61% while 71% respondents said they had voted. The same situation was with previous governor’s elections: the turnout was 58% and survey data was 70%. This increasing of previous turnouts can be explained like that: the act of voting is socially approved. During election campaign people hear every time that if they do not vote it will be the waste of federal and local budget money. Taking into account the differences between statistics and answers of respondents we crossed people who surely will vote with those who knew the exact date of elections (8^th of April) and the result of that calculation was from 46% to 54% turnout, average – 50%.

Figure 7.

The next stage is to discover the activity of each candidate’s electorate and core electorate which could be found out from data of anti-voting and pair voting (situation of second round of elections, when there are only two candidates).

The results of voting for other candidates, Sokol and Samoshin, were equal at any level of turnout. So we could assume that they will have about 20% of electorate each if 50% will come to polling stations. As for Star his rating depended on turnout. It increased while turnout decreased. At turnout 100% he could get 34%. At the level of 70% he could get 39%. If the turnout will be 50% he will get 46% of electorate (Figure 7).

Election, 8^th of Apr.

At 8^th of April the turnout was 48%. Star won the elections with 49%. Sokol has got 19% of electorate, Samoshin – 21%.

The data analysis was compared to dynamics of the whole electoral situation for three months. On the results of the analysis we produced the procedure of predicting voting results. The procedure allows predicting voting results rather precisely. Thereby, election campaign was rather typical governor’s election in Russia and at the same time it had some peculiarities that made the campaign interesting from scientific and practice points of view. We consider this procedure as multi-purposed and could be used for different levels of elections (federal, regional, local).

WAPOR ,USA, june, 2002