Patterns of Turnout in the Ethiopian Election of May 15,
2005 and Implications for Validity of the Results
Preliminary Draft Report, Part 1
June 19, 2005
This report summarizes results from 464 of 517 weredas for which "provisional" results were available from the National Elections Board of Ethiopia (NEBE) as of June 4, 2005. The remaining 53 weredas are not included because the NEBE has posted the winner without releasing the turnout data.
The wereda level data are only a small subset of the database containing results from 30,000 polling stations. Access to the polling station data is necessary in order to properly analyze the data. Without the polling station data, we are limited to examining variability between weredas within a zone. This is hampered by the small number of weredas per zone in many cases, and the presence of large discrepancies between weredas.
Nevertheless, the wereda turnout, as well as the inferred percentage of invalid ballots, and the winning vote percentage, can be explored to identify patterns and try to explain them. Further research would involve linking the election data to socioeconomic data for each wereda or kebele if available. What is presented below is thus only a preliminary evaluation using simple sorting and correlation techniques.
The overall conclusion of this report is that the reported turnout figures for many weredas are not credible. Furthermore, the provisional results exhibit a number of characteristics that are consistent with large-scale manipulation prior to their being reported to the NEBE. In particular, this report makes the following specific findings:
1. Anomalous provisional results are dominated by EPRDF wins by a ratio that is about 30% higher than the ratio of seats won by the EPRDF.
2. Forty-nine of the 324 (15%) weredas won by the EPRDF and its allies were single-candidate elections with no opposition. Turnout in these weredas averaged 93%; far higher than the 83% national turnout. This anomalously high turnout does not appear credible and suggests that a routine practice of ballot padding is prevalent in the country.
3. In most zones across the country, turnout for EPRDF wins exceeded turnout for weredas won by the opposition.
4. In most zones across the country, the percentage of ballots that were invalid was significantly lower in weredas won by the EPRDF as compared to weredas won by the opposition.
5. A relationship between turnout and winning vote percentage is evident in virtually all zones won by the EPRDF but virtually absent in the zones won by the opposition. Such a relationship could be an indicator of vote tampering when combined with the findings in #3 and #4 above.
6. There were 29 weredas where turnout exceeded 90% and the inferred invalid ballot percentage was less than 10%. Twenty-seven of these weredas (93%) were EPRDF wins (excluding Addis Abeba, Tigray, unopposed candidates, and erroneous figures, i.e. turnout > 100, and invalid <= 0).
7. Thirty weredas outside of Tigray were won with a winning percentage in excess of 90% of the vote (excludes unopposed candidates). Twenty-eight of these (93%) were EPRDF wins.
8. The EPRDF sweep of 16 rural weredas in East Gojjam exhibits patterns that are consistent with large numbers of ballots being added illegally or vote totals being manufactured at the wereda or zonal level.
9. The EPRDF won only two contested weredas in Wellega. These two weredas reported the highest turnout in Wellega. The 99.2% turnout in Gelila wereda of East Wellega was more than three standard deviations above the turnout in weredas won by the opposition.
10. The various opposition candidates were unable to field a single candidate in 12 of 39 weredas (39%) in Wellega - an observation that raises serious doubts about whether ballot access was free and fair, particularly since the rest of Wellega was highly competitive, with four different parties and one independent winning seats.
11. The rural weredas of Kelela and Kutaber recorded the highest turnout and the lowest percentage of invalid ballots in all of Wello. Both weredas were won by the EPRDF but the results are anomalous to the degree that they are unlikely to have occurred without post-election intervention.
12. The EPRDF was apparently victimized in Wereda 20 of Addis Abeba, where the EPRDF total was abnormally low, coincident with an abnormally high number of spoiled ballots, and an abnormally high turnout (nearly three standard deviations above average).
13. The election in the Afar Region exhibits such a large number of discrepancies in the number of registered voters, turnout, winning percentage, and invalid ballots that the entire Region warrants special review.
14. Jimma zone was swept by the EPRDF, but turnout in Jima Kersa wereda, which contains the large zonal capital of Jimma was more than two standard deviations below the average in the rest of the zone. In the absence of reasonable explanations for this anomaly, one would have to consider that rural turnout was more likely subject to artificial inflation as compared to Jimma town.