As part of my PhD thesis, I have generated a dataset that includes all corresponding questions across four rounds of the Afrobarometer. After having submitted my PhD thesis, I have extended the dataset up to round 5. Beginning with data from round 4, for each question the corresponding question from earlier rounds was merged (and later extended to round 5). In some occasions, codings had to be changed. If this was the case, the process is described in the dataset. From all rounds of the Afrobarometer, the first round is characterized by the least comprehensive data collection. Some questions that are indispensable to the PhD hypotheses are not included. Due to the sparse data availability, round 1 of the Afrobarometer is excluded from the analysis.
As part of my PhD thesis, I have generated a dataset that merges the respondents ethnic self-classification in the Afrobarometer to the Ethnic Power Relations (EPR) data family, i.e. the dataset matches individuals from the Afrobarometer data to their corresponding ethnic group in the EPR dataset. First, I coded the individuals ethnic group. The question used to identify the respondents’ ethnic group is: "What is your tribe? You know, your ethnic or cultural group". Since in the EPR data, ethnic groups are given in rather broad categories, in a second step, I coded sub-ethnic groups as to belong to their higher-order ethnic group. For each of the ethnic groups that were named by the respondents in the Afrobarometer, I first checked whether the ethnic group can also be found in the EPR dataset. If this was not the case, I reviewed various ethnographic sources whether the ethnic group is a sub-group of one of the EPR coded ethnic groups. The most used ethnographic sources were Gates and Appiah (2010), Diagram Group and Giles (2000), Murdock (1967), Olson (1996), Stokes (2009), and Yakan (1999).
Originally, for the four rounds in the Afrobarometer, slightly more than 70% of the respondents (43,401 of 61,330) in the Afrobarometer could be matched to their ethnic group in the EPR data. Note that I only could match the rounds in which raw data on the ethnic identity of individuals is available, i.e. rounds 3, 4, and some of the additional rounds that were not part of the regular cycle (Namibia in 2002 and 2007, Nigeria in 2001, South Africa in 2004). Only 8,505 respondents did not reply to the question of their ethnic identity or answered it as “Other” or “Don’t know”. Another 9,424 respondents classified themselves as being ethnic and named an ethnic group, however, no corresponding ethnic group was found in the EPR dataset. The dataset was later extended to also include round 5 of the Afrobarometer.