Presentation on theme: "Is African politics a racial census? Using new survey data to measure the ethnic political cleavage in new African democracies Robert Ford CCSR, University."— Presentation transcript:
Is African politics a racial census? Using new survey data to measure the ethnic political cleavage in new African democracies Robert Ford CCSR, University of Manchester Nicholas Cheeseman, Jesus College, University of Oxford Contact:
Democracy in Africa: The Third Wave breaks on African shores African politics prior to 1989: (i) Decolonisation (ii) Dictatorship or civil war In 1989, 42 of 47 sub-Saharan African nations were one party states or military dictatorships By 1994, 40 of these 42 had politically liberalised, and 16 had held free elections. By 2007, 24 African nations were rated as electoral democracies Over 200 million people live in African nations which made transitions to democracy since 1989
CountryFounding Election Established Botswana1966 African Zimbabwe Democracies Senegal1981 Namibia1990 Benin, Cape Verde Islands, Zambia1991 Third Wave Kenya, Madagascar, Mali1992 Democracies Lesotho1993 Malawi, Mozambique, South Africa 1994 Ghana1996 Nigeria Dysfunctional or Zimbabwe1980 (RIP: c.2000) Deceased Tanzania1995 Democracies Nigeria1999 (RIP c.2007) Uganda2001
Freedom House Ratings (1-7) Change Benin Mali Ghana South Africa Mozambique Kenya Nigeria Zambia53.54 Senegal Uganda4.5 0 Zimbabwe AB Average
The Afrobarometer project First ever systematic, comparative survey of African social and political attitudes First round in : 12 countries Second round : 16 countries Third round : 18 countries Clustered, stratified, multistage area probability samples from each country. N between 1,200 and 3,600 (Mattes and Bratton, 2007) Personal face-to-face interviews conducted in respondents preferred language by trained interviewers Sample now includes most of Africas democracies, new and old Not a representative sample of Africa or Africans as a whole: authoritarian regimes and countries in conflict are not surveyed
What the Afrobarometer covers The Afrobarometer surveys have gathered information on a wide range of subjects Political: partisanship, voting, attitudes to democracy; opinion of current leaders and government performance; political knowledge and interest; policy priorities; corruption; citizen rights and responsibilities; media consumption Social: Employment, poverty, health, AIDS, attitudes to women, experience of public services, social trust, values Demographics – age, sex, race, ethnicity, class, urban/rural, local conditions Further information, data, results and analysis available at: Interested in research assistance or collaboration? Contact me!
The role of ethnicity in African politics Ethnicity matters in Africa: (i) African states are very diverse: only 2 Afrobarometer countries are ethnically homogenous. Most have five or more significant ethnic groups (extreme examples: Nigeria, Uganda) (ii) Ethnic identities are important resources for mobilising voters in new African democracies, where knowledge and education are low, resources are scarce, and language barriers hinder cross-ethnic mobilisation (Horowitz, 1985; Posner, 2005) (iii) In many countries, ethnicity was (is) an important factor deciding division of political resources – jobs, investment, aid Consensus exists that ethnicity is most important cleavage in most African democracies (though debates over which ethnicities and how mobilised) Aggregate research has examined the link between density of ethnic groups and vote for political parties (examples)
Largest2 nd 3 rd 4 th 5 th 6 th 7 th OtherELF South Africa Zulu 20.0 Xhosa 15.5 Afrikaans 13.5 Setswa 10.2 Spedi 10.3 English 8.7 Sesotho Nigeria Housa 25.2 Yoruba 22.1 Igbo Ghana Akan 52.6 Dagaare 19.4 Ewe 13.5 Ga/ Dangbe Greater London White Brit 59.8 White Other 8.3 Indian 6.1 African 5.3 Caribbean 4.8 Irish 3.1 Bangla ZimbabweShona 78.6 Ndebele England and Wales White Brit 87.5 White Oth 2.6 Indian 2.0 Pakist 1.4 White Irish 1.2 Caribb 1.1 African
Our approach: mapping ethnic cleavages using individual level data Utilising Afrobarometer data, we aim to create new summary measures of ethnic (and other) cleavages in African politics Measures will be comparable between ethnic groups, between nations and between time point First effort to systematically map the strength of ethnic cleavages in new African democracies Some countries omitted due to: Ethnic homogeneity (Lesotho, Cape Verde Islands) Doubts about quality of language data (Tanzania, Madagascar) Doubts about quality of party affiliation data (Benin) Resulting sample: 10 countries,3 waves of data; 3 countries, 2 waves of data
Measures: ethnicity and party affiliation Ethnicity measured using respondents self- reported home language. Advantages: clarity, comparability, respondents own choice Disadvantages: fragmentation; risks inclusion of irrelevant groups; strange responses (non-existent languages); does not reflect most important ethnic cleavages Political affiliation measured using question asking respondents whether they ID with a party. Advantages: clarity, comparability, respondents own volunteered choices; availability (vote data only available in wave 3) Disadvantages: not clear what party ID means to Africans, possibly skewed by other factors (allegiance to a President or other leader); unaffiliated not analysed Non-affiliated to be analysed in forthcoming work
Measuring the ethnic cleavage Why not just look directly at which groups vote for/align with which parties? Problem - this conflates three things: changes in the popularity of different parties the size of different ethnic groups The link between ethnicity and affiliation with a particular party To capture only the link between ethnicity and partisanship, we employ logistic regression methods Logistic regressions capture differences in the odds of different ethnic groups affiliating with a party. Such odds ratios are margin insensitive, so isolate only the strength of the link between ethnicity and partisanship
The kappa score: a summary measure of ethnic cleavages Logistic methods produce a more accurate picture of the strength of ethnic cleavages, but they also produce an awful lot of coefficients – complex, messy, difficult to interpret We need a simple summary index measure of the ethnic cleavage The kappa score (Manza, Hout and Brookes, 1995; Brookes and Manza, 1997): an index of the standard deviation in either regression coefficients or predicted probabilities I.e: a measure of how much the odds of voting for a particular party varies across different ethnic groups Flexible measure: kappas can be aggregated across parties and groups, or disaggregated to look at particular parties and groups Comparable across nations and time points Problems of model specification – which parties/groups to include? Small parties/groups may excessively influence results. Weighting?
Initial Findings Ethnicity is a significant influence on party choice in nearly all Afrobarometer countries. 2 exceptions – Botswana, Senegal – are also the longest established democracies in the sample Strength of ethnic cleavages in Africa seems to be declining overall– 7 of 10 countries with three surveys show evidence of sustained decline in kappa scores Large divergence in kappa levels and trends, with three main patterns: Declining ethnic cleavage, maturing democracy: Malawi, Mali, Mozambique, Namibia, Nigeria, South Africa Low, stable ethnic cleavages, mature democracy: Botswana, Senegal, Ghana Rising ethnic cleavages, unstable democracy? Kenya, Uganda, Zambia, Zimbabwe
Ch Ch Malawi South Africa Mali Nigeria Zambia Kenya * * Namibia Ghana Uganda Mozambique * * Senegal * * Botswana Zimbabwe AB mean (10) AB mean (13) * *-0.032
Initial findings (2) The overall party system trends mask large differences between patterns of support for governing and opposition parties Kappa scores are generally lower for governing parties than opposition Again, several groups of countries emerge: Declining ethnic cleavages: countries where kappa scores are either low or declining for governing and opposition parties, suggesting ethnicity is becoming less relevant overall: Ghana, Mali, Mozambique, Namibia, Senegal, South Africa, Ethnic opposition – ruling party kappa score is low or declining, but opposition parties have high or rising kappa scores, suggesting opposition to regime is concentrated in certain ethnic groups: Botswana, Malawi, Nigeria, Zambia, Uganda Ethnic division – high and/or rising kappa scores for governing and opposition parties, suggesting deep ethnic divisions at the centre of political mobilisation: Kenya, Zimbabwe
Opposition KappaScore High/IncreasingLow/Decreasing Government High/ Increasing Ethnic division Kenya, Zimbabwe Kappa ScoreLow/ Decreasing Ethnic opposition Botswana, Malawi, Nigeria, Uganda, Zambia Declining ethnic cleavage Ghana, Mali, Mozambique, Namibia, Senegal, South Africa
Conclusions and limitations Our measure represents first effort (to our knowledge) to systematically map the strength of ethnic cleavages in African democracies using individual level data Important findings about the strength of ethnic cleavages, and apparent link between maturation of democracy in Africa and decline in ethnic cleavage Specification and robustness: are we including all relevant parties? Ethnic groups? Would results change if we did? Limits of language based ethnicity measure – does this capture the most significant forms of ethnic identity? Limitations of partisanship measure - is it a meaningful measure of political affiliation in Africa? Does the measure mean the same thing in different countries?
Avenues for future research Explaining variation in kappas: why are some African societies more ethnically divided than others? Individual ethnic groups: which groups tend to concentrate their support behind one party, which divide their support evenly? Political disaffection: which countries show the largest divisions in levels in party affiliation, suggesting some groups are dominating party politics while others are excluded? Other cleavages: (a) How does the importance of ethnicity in Africa compare with other cleavages such as class, age, education, religion, urban-rural? (b) Are other cleavages becoming more prominent as democracies mature and ethnic cleavages decline?