On Saturday, October 8th, Georgians will once again go to the polls to elect members for a new parliament. Pre-election surveys, sponsored by Imedi TV and developed by JPM Strategic Solutions, have indicated that the current ruling party, Georgian Dream, has a comfortable lead among all participating parties.
The most recent data, released late last week, indicate that about 33% of eligible likely voters intend to vote for Georgian Dream. In contrast, just 13% reported that they intend to vote for United National Movement, the party that ruled Georgia from 2003 to 2012. No other single party broke the 5% support threshold that would be required to make it into parliament. These results, however, come with a big caveat: a significant number of Georgians tell us that they remain undecided about which party to support. In fact, the data show that undecideds represent about 40% of all likely eligible voters.
Respondents who report undecided/don’t know in quantitative surveys are one of the banes of a pollster’s life. Of course, it is not really the fault of the respondent. Decades of public opinion research has shown that large number of people walk around with conflicting opinions in their heads. When questioned by an interviewer, very often the most appropriate answer seems to be: I don’t know. Yet, when the job of pollsters and researchers is to provide forecasts of election outcomes, undecided voters pose a special problem.
That a certain segment of the population remains undecided about who to vote for is not unusual. According to an aggregation of surveys done by the Huffpost, in the United States just under 6% of likely voters remain undecided about the upcoming presidential elections. This is despite the fact that both Trump and Clinton have been campaigning for over a year and have each spent many hundreds of millions of dollars. Yet, the US election is still over a month away. In Georgia, however, with the election just around the corner, the number of undecided likely voters is nearly 7 times higher. A truly astonishing (and unfortunate) number. Alas, it is also a relatively common number: GORBI’s surveys have documented a relatedly high number of undecideds in every Georgian election since 2006.
The inevitable question is (at least for me!), who are these people? What can we learn about Georgian citizens who say they will almost certainly vote, but remain undecided just days before an election? For political parties, knowing the answer to this question should be a big deal: knowledge about undecided likely voters can help parties to tailor their message to these particular segments of society. Thankfully, there are methods to investigate who these undecided voters are. By using statistical modeling techniques on our recently collected pre-election survey data, we can make inferences about the characteristics of individuals who remain undecided.
Dr. Christopher Anderson, senior data analyst at GORBI, developed a logistic regression model that allowed us to measure the importance of various demographic characteristics (ie. gender, age, education, etc.) on the probability that a respondent reported being undecided. In short, the model can tell us what kind of characteristics are representative of undecided voters. We can then learn if (and how) likely undecided voters differ from likely voters who have already made up their minds.
Our model contained seven characteristics: gender, age, education, income (self-reported), employment status, ethnicity (Georgia, Armenian, Azerbaijani, other), and whether or not the respondent lives in an urban or rural area.
The results of our model show that from the perspective of many demographic characteristics, likely voters who are undecided and decided are generally not that different from each other. There are, however, some exceptions. Unemployed individuals are much more likely to report being undecided than employed individuals. Chart 1 shows that the average unemployed likely voter has a probability of being undecided of .33. That is to say, about 1 in 3 unemployed likely voters is undecided.
However, similarly average, but employed, likely voters have a probability of being undecided of only .26. In other words, about 1 in 4 employed voters is undecided. I think the conclusion is clear: the lack of employment in Georgia has led to a crisis for many voters who have no love for the status quo but lack credible alternatives. It is no surprise that unemployment has topped the list of issues important to voters over the last three elections.
One other variable was statistically significant in our model: ethnicity. In particularly, ethnic Armenians living in Georgia are significantly more likely to report being undecided than members of other groups. Armenian likely voters are 1.75 times more likely than ethnic Georgian likely voters to be undecided. They are over twice as likely as Azeri likely voters to be undecided.
Do these numbers matter? Yes, they absolutely do. These results might seem small and superficial at first, but when extrapolated to the population of the entire country, these percentages translate into tens of thousands of individuals. The inability of political parties to deal with the extremely high levels of unemployment (officially 12%, but some experts consider the level to be between 40-50%) has resulted in large numbers of unemployed voters who remain undecided. The fact that many Armenian voters also remain undecided is a reflection of the fact that in some sense they remain outsiders in Georgian politics. A large number of Armenian voters are ready to participate, but Georgian parties have not yet found a way to appeal to them.
Finally, a prediction. The media in the United States often writes how undecided voters have the power to sway election outcomes. In Georgia, however, despite the larger number of undecided voters, I am confident that our new parliament will either be a coalition led by Georgian Dream or Georgian Dream will win an outright majority of seats. Based on our data, the party’s lead over their competition is so large (and time is so short) that it is extremely unlikely that the undecideds will have a significant influence over the outcome of this election. Rather, these voters are likely to break fairly evenly across participating parties or they will simply stay home. These are voters that the parties should have been courting months ago. If we had a betting business in Georgia, my money would be on Georgian Dream. However, since we don’t, please get in touch with me if you have money to waste…
GORBI is a regional hub for partner organizations and international clients. Since 2003, GORBI remains an exclusive member of Gallup International research network for its two decades of experience in survey research in post-Soviet Union countries, as well as Mongolia and Iraq. This data was provided exclusively to the Financial.