Wednesday, March 18, 2009

Some Thoughts on Turnout in Azerbaijan

The CEC's preliminary results show overwhelming support for all of the questions in the referendum. In October's presidential election, the main question did not surround the outcome, however, but rather turnout. While no minimum threshold threatened the outcome, observers thought that President Aliyev's partisans would ensure a strong turnout to show support. Turnout measures mobilization and is the primary issue for the referendum as well.

The graphic to the left shows a simple box plot of the precinct-level turnout data published by the CEC. The CEC posted turnout data from 5,367 precincts, although some precinct-level data are incomplete (e.g., TEC 29 has three precincts with no results reported). As a visual representation of simple descriptive statistics, the box plot does not provide rigorous analysis. However, it shows some interesting results and suggests that some cases might be outliers. The 50th percentile is represented by the line in the rectangle; the 25 and 75th percentiles by the upper and lower bounds of the box. Observations outside the "whiskers" are outlying cases worthy of further investigation.

At 10 a.m., two precincts reported over 75% turnout (TEC 96, PEC 45 and TEC 98 PEC 27). These precincts are small, and could be located in special precincts where turnout can be managed. Turnout management could be benign (i.e., hospital patients may be transported to the precinct and may vote early) or questionable. At the closing of the polls, 60 precincts reported 100% turnout. These precincts also report questionable results. In TEC 41, PECs 37 and 38 both had 1,500 registered voters and all registered voters reportedly cast ballots. According to the published protocols, all of the voters in both precincts cast affirmative ballots for each of the 29 questions on the ballot. While not impossible, these results are highly improbable. Indeed, one would expect that among 3,000 voters, at least one voter might have made an error on a ballot with 29 questions, invalidating it.

This commentary begs the question: what is the normal pace of turnout? In my earlier post, I suggested that 30% could be a "reasonable" upper bound for turnout figures at 10 a.m. The figure of 30% was purely hypothetical; my main point was that a reasonable upper bound at 10 a.m. is lower than a reasonable upper bound at 5 p.m. (while the lower bound could always, in principle, be zero). As the day progresses, increasing variance in turnout reports is not necessarily surprising. No research, to my knowledge, has identified a pace of turnout that conforms with free and fair practices. Indeed, the pace of turnout is likely to be affected by many factors: the perceived closeness of the race, the level of citizen interest, and if election day is on a work day or holiday, among other factors. A forthcoming article by Pacek, Pop-Eleches, and Tucker in the Journal of Politics shows that the perceived importance of an election can strongly affect turnout in post-communist societies (my own research concurs with their finding). The referendum was portrayed as important enough for citizens to be motivated to show regime support. Yet, the preliminary data also raise some red flags about manipulation. I will post additional analysis as I look more deeply at the data.

1 comment: