Tuesday, March 30, 2010

Azerbaijan Webcam Redux

As I mentioned in a previous post, I recently composed a paper addressing the effects of election monitoring on reported outcomes in Azerbaijan. This paper is now forthcoming in the journal Electoral Studies. The analysis shows that the presence of webcams in polling stations is consistently associated with depressed turnout figures. Webcams are also associated with lower levels of regime support in the referendum. Because journals have limited space, this post includes supplementary materials that accompany the paper.

Supplementary Table 1 = t-tests for turnout across all reporting periods. The tests for each reporting period show a statistically significant difference between monitored and unmonitored polling stations, with monitored polling stations consistently experiencing lower reported turnout than unmonitored polling stations.

The final turnout report differs from the 7 p.m. report in many cases. In the presidential election, 40% of the 7 p.m. reports did not differ from final turnout, and the standard deviation was 9.2 (in percentage point terms). In the referendum, 76% of the 7 p.m. reports did not differ from final turnout as reported in the protocol, and the standard deviation was 3.5.

The presidential election featured many ballots cast via absentee certificates, creating a challenge for calculating turnout. The periodic reports are based on the number of registered voters; absentee certificate users are not registered in the polling stations. The protocols include more detailed information about the composition of voters reporting to each polling station. Reports could differ due to complications arising from the use of absentee certificates, errors, or misrepresentation.

Supplementary Table 2 = results of the analysis of turnout with unmatched data.

Supplementary Table 3 = results of the analysis of outcomes with unmatched data.

R code for generating and assessing the matched data.

Replication Data (presidential and referendum).