¹ The Afrobarometer project’s website, 〈www.afrobarometer.org〉, gives further information.

² The institutions in question are judicial courts, independent media, and elected legislatures.

³ Stephen C. Craig, Jason Gainous, Michael D. Martinez and James G. Kane, ‘Winners, Losers, and Perceived Mandates: Voter Explanations of the 1998 Gubernatorial and 2000 Presidential Elections in Florida’ (paper presented at the Annual Meeting of the American Political Science Association, Chicago, 2004), p. 2.

⁴ There are numerous studies of winner–loser gaps in advanced industrial democracies. For examples see: Christopher J. Anderson, André Blais, Shaun Bowler, Todd Donovan and Ola Listaug, Losers’ Consent: Elections and Democratic Legitimacy (Oxford: Oxford University Press, 2005); Christopher J. Anderson and Andrew J. Lotempio, ‘Winning, Losing and Political Trust in America’, British Journal of Political Science, 32 (2002), 335–51; Christopher J. Anderson and Yuliya V. Tverdova, ‘Winners, Losers, and Attitudes About Government in Contemporary Democracies’, International Political Science Review, 22 (2001), 321–38; Christopher J. Anderson and Yuliya V. Tverdova, ‘Corruption, Political Allegiances, and Attitudes toward Government in Contemporary Democracies’, American Journal of Political Science, 47 (2003), 91–109; Susan A. Banducci and Jeffrey A. Karp, ‘How Elections Change the Way Citizens View the Political System: Campaigns, Media Effects and Electoral Outcomes in Comparative Perspective’, British Journal of Political Science, 33 (2003), 443–67; Harold D. Clarke and Alan C. Acock, ‘National Elections and Political Attitudes: The Case of Political Efficacy’, British Journal of Political Science, 19 (1989), 551–62; Craig et al., ‘Winners, Losers, and Perceived Mandate’; Dieter Fuchs, Giovanna Guidorossi and Palle Svensson, ‘Support for the Democratic System’, in Hans-Dieter Klingemann and Dieter Fuchs, eds, Citizens and the State (Oxford: Oxford University Press, 1995), pp. 323–54; Ola Listhaug and Matti Wiberg, ‘Confidence in Political and Private Institutions’, in Klingemann and Fuchs, eds, Citizens and the State, pp. 298–323; Richard Nadeau and André Blais, ‘Accepting the Election Outcome: The Effect of Participation on Losers’ Consent’, British Journal of Political Science, 23 (1993), 553–63; Richard Nadeau, André Blais, Neil Nevitte and Elisabeth Gidengil, ‘Elections and Satisfaction with Democracy’ (paper presented at the Annual Meeting of the American Political Science Association, Washington, D.C., 2000); Pippa Norris, Critical Citizens: Global Support for Democratic Government (Oxford: Oxford University Press, 1999).

⁵ Citizens who deem the system illegitimate may not take up arms against the state, but they will also not act as a buffer to those who do seek to alter the political system from within or from without.

⁶ Anderson et al., Losers’ Consent; Allan E. Lind and Tom R. Tyler, The Social Psychology of Procedural Justice (New York: Plenum, 1988); Tom R. Tyler, Jonathan D. Casper and Bonnie Fisher, ‘Maintaining Allegiance toward Political Authorities: The Role of Prior Attitudes and the Use of Fair Procedures’, American Journal of Political Science, 33 (1989), 629–52; Tom R. Tyler, Why People Obey the Law: Procedural Justice, Legitimacy, and Compliance (New Haven, Conn.: Yale University Press, 1989).

⁷ The authors of Losers’ Consent use data from Eastern Europe and Mexico in their analyses of new democracies; Bratton and his colleagues have done some important work on this topic in Africa; and Norris’s analysis in Critical Citizens includes Mexico, India and Chile. For more information from these works see: Anderson et al., Losers’ Consent; Michael Bratton, Robert B. Mattes and Emmanuel Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa (Cambridge: Cambridge University Press, 2005); Wonbin Cho and Michael Bratton, ‘Electoral Institutions, Partisan Status, and Political Support in Lesotho’; Electoral Studies, 25 (2006), 731–50; Norris, Critical Citizens.

⁸ For a detailed description of a test of mediation, see Reuben M. Baron and David A. Kenny, ‘The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual Strategic and Statistical Considerations’, Journal of Personality and Social Psychology, 51 (1986), 1173–82.

⁹ There is a related debate about whether evaluations of procedural justice are the causes or consequences of perceived institutional legitimacy of the US Supreme Court. For a review, see Jeffery J. Mondak, ‘Institutional Legitimacy and Procedural Justice: Reexamining the Question of Causality’, Law and Society Review, 27 (1993), 599–608.

¹⁰ However, the effect of expectations of electoral fraud and perceptions of political corruption on opposition support in Mexico were not significant in a multivariate analysis in James A. McCann and Jorge I. Dominguez, ‘Mexicans React to Electoral Fraud and Political Corruption: An Assessment of Public Opinion and Voting Behavior’, Electoral Studies, 17 (1998), 483–503.

¹¹ Staffan I. Lindberg and Minion K. C. Morrison, ‘Exploring Voter Alignments in Africa: Core and Swing Voters in Ghana’, Journal of Modern African Studies, 43 (2005), 565–86.

¹² Mondak, ‘Institutional Legitimacy and Procedural Justice’.

¹³ Round 2 did not include the question about electoral fairness and neither Round 2 nor Round 3 included the questions about defending democratic institutions.

¹⁴ For further details see Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa, p. 17; Larry Jay Diamond, ‘Thinking About Hybrid Regimes’, Journal of Democracy, 13 (2002), 21–35. In the year of the first round Afrobarometer survey (1999, 2000 or 2001 accordingly), Freedom House rated the following countries as free: Botswana (political rights = 2, civil rights = 2), Mali (political rights = 2, civil rights = 3), Namibia (political rights = 2, civil rights = 3), and South Africa (political rights = 1, civil rights = 2). Freedom House rated the following countries as partially free: Ghana (political rights = 3, civil rights = 3); Lesotho (political rights = 4, civil rights = 4); Malawi (political rights = 3, civil rights = 3); Nigeria (political rights = 4, civil rights = 4); Tanzania (political rights = 4, civil rights = 4); Uganda (political rights = 6, civil rights = 5); Zambia (political rights = 5, civil rights = 4); and Zimbabwe (political rights = 6, civil rights = 5). See Freedom House, Freedom of the World: The Annual Review of Political Rights and Civil Liberties (Freedom House, 2005 [cited 8 March 2005]); available from 〈http://www.freedomhouse.org/research/freeworld/2005/table2005.pdf〉.

¹⁵ The data on the most recent elections and the electoral system comes from an updated version of Lindberg’s dataset: Staffan I. Lindberg, Democracy and Elections in Africa (Baltimore, Md.: Johns Hopkins University Press, 2006).

¹⁶ Vanessa A. Baird, ‘Building Institutional Legitimacy: The Role of Procedural Justice’, Political Research Quarterly, 54 (2001), 333–54, p. 334.

¹⁷ James L. Gibson, Overcoming Apartheid: Can Truth Reconcile a Divided Nation? (New York: Russell Sage Foundation, 2004), p. 294.

¹⁸ James L. Gibson and Gregory A. Caldeira, ‘The Legitimacy of Transnational Legal Institutions: Compliance, Support, and the European Court of Justice’, American Journal of Political Science, 39 (1995), 459–89, p. 471.

¹⁹ M. Stephen Weatherford, ‘Measuring Political Legitimacy’, American Political Science Review, 86 (1992), 149–66, p. 149.

²⁰ See the Appendix for exact question wording and coding rules. The regression analyses are based on an unweighted pooled sample of 21,531 respondents. Descriptive statistics, including means and frequency distributions, are calculated using a weighted sample to correct for disproportionate subsamples within countries and to standardize country samples at n = 1,200. Frequency distributions record proportions of valid responses.

²¹ The alpha coefficient for institutional trust is 0.77, indicating a fair degree of reliability. To capture support for institutions rather than individuals, the analysis includes only institutions that were not commonly associated with a particular individual. In most cases citizens have experienced only one president and one Member of Parliament under their current system. Thus it is difficult to assess from survey questions, for example, whether citizens trust the institution of the presidency or the current president. An index variable that excluded trust in the electoral commission was also used to ensure that this institution alone was not responsible for the findings. The results were the same for the key variables although the size of the coefficients and the statistical significance were somewhat smaller.

²² The alpha coefficient for defending democracy is 0.80 indicating a high degree of reliability.

²³ The alpha coefficient for the four indicators of institutional legitimacy is very low (0.17) indicating that these are distinct attributes rather than different measures of a single coherent structure.

²⁴ The variable is named Identifies with Winning Party in Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa.

²⁵ Bratton et al. acknowledge that perhaps not all citizens truthfully or accurately report their partisan attachments but argue that the measure is still valid: ‘Of course, some respondents may rewrite their personal histories by reporting voting records deemed politically correct. Despite the possibility that we were sometimes intentionally misled, we still expect that being a self-proclaimed “winner” increases one’s loyalty to incumbent leaders and reduces one’s willingness to criticize their performance’ (Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa, p. 259).

²⁶ Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa, pp. 256–61.

²⁷ The questions that make up the measures of legitimacy were not always asked in every country. Where a question was not asked, the country is eliminated from the analysis for that measure only. Figures 2 through 5 show which countries are represented (there are bars in the chart above the country name) and which countries are not represented (no bars above the country name).

²⁸ All three trust questions were not asked in Uganda, so the article discuses the eleven countries for which there are data.

²⁹ On average, winners said that they can trust their government institutions in ten of the eleven countries (Nigeria being the exception), although the means are significantly above 0 in only seven of those countries. Average winner trust is not significantly different from 0 in Lesotho, Nigeria and Zimbabwe. In contrast, losers are trusting in only five countries, with only two being significantly positive. Average loser trust is not significantly different from 0 in Mali and Zambia.

³⁰ For losers, the mean value of granting authority is significantly negative in every country. For winners, the mean is not significantly different from 0 in Botswana, Tanzania, Uganda and Zambia. The winner mean is significantly positive in Lesotho, Namibia and South Africa, and significantly negative in Malawi, Mali and Zimbabwe.

³¹ The average winner is significantly positive in all countries. The average loser is positive in all countries but not significantly so in Zambia and Zimbabwe.

³² Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa, p. 260.

³³ To facilitate interpretation, the results of ordinary least squares (OLS) regressions are presented in this article. Ordered logit analyses were also conducted because the dependent variables are categorical variables. The statistical significance of the key independent variable (winner status) and intervening variable (free and fair election) remains the same in every model, with the exception of Model B predicting external efficacy, where the p-value for the coefficient on winner–loser status changes from 0.003 in OLS to 0.000 in ordered logit.

³⁴ See the appendix of this article for question wording and coding. Except where otherwise indicated, the construction and coding of these and other variables mirror those described in Appendix A of Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa, p. 260, pp. 355–91.

³⁵ For examples, see Anderson and Tverdova, ‘Corruption, Political Allegiances, and Attitudes toward Government in Contemporary Democracies’; Anderson et al., Losers’ Consent; Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa; Clarke and Acock, ‘National Elections and Political Attitudes’; Nadeau and Blais, ‘Accepting the Election Outcome’; and Nadeau et al., ‘Elections and Satisfaction with Democracy’.

³⁶ The measure evaluating of economic performance used in this article includes four of the five variables included in the measure used in Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa. The analysis in this article excludes the measure of whether the economic conditions of one’s own group are worse, the same as, or better than other groups in the country. Dropping this variable from the index does not alter the main results and it allows data from Uganda to be used in several of the equations; that particular question was not asked in Uganda.

³⁷ Previous research shows that educated Africans tend to be more critical of their government, but initial expectations about the effects of gender and age are more ambiguous. For example, see Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa; Robert Mattes and Michael Bratton, ‘Learning About Democracy in Africa: Awareness, Performance, and Experience’, American Journal of Political Science, 51 (2007), 192–217; Devra C. Moehler, Distrusting Democrats: Outcomes of Participatory Constitution-Making (Ann Arbor: University of Michigan Press, 2008).

³⁸ The excluded category is Botswana.

³⁹ The difference between winners and losers is significant at the 0.00 level, except for Zambia, which is significant at the 0.05 level.

⁴⁰ The winner mean in Zimbabwe is not significantly different from 0. The loser means for Lesotho, Mali and Tanzania are not significantly different from 0.

⁴¹ Notably, when the same model is run on each country separately, the coefficients on winner status are all significant with 99 per cent confidence. The substantive effects are largest in Malawi and Tanzania.

⁴² For the models that predict defending democracy, the coefficient on winner status increased (or became less negative) in the presence of the intervening variable. In other words, the reverse gap between winners and losers narrowed when electoral evaluations were controlled for. For all the other dependent variables, the coefficient on winner status decreased (or became less positive) when free and fair election was included in the equation.

⁴³ As mentioned above, the evidence presented is consistent with the initial hypotheses but the available data do not allow us to establish conclusively the direction of causation between partisanship, fairness evaluations and perceived legitimacy. Panel, experimental or qualitative data would help to establish that the causal pathways are as hypothesized.

⁴⁴ Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa, p. 260.

⁴⁵ Analysis revealed insignificant relationships between the beta coefficients on winner status for all the four measures of legitimacy and (1) whether an election was judged free and fair by outside observers; (2) if the parliamentary electoral system is PR instead of majoritarian or plurality; and (3) the level of civil and political rights at the time of the election. However, the very small samples could be responsible for the insignificant results. In addition, graphical depictions of the beta coefficients and 95 per cent confidence intervals by national characteristics did not reveal notable patterns.

⁴⁶ This assessment is in accordance with Bratton et al.’s ‘learning theory of cognitive rationality’. They argue that cognitive awareness and performance evaluations are far more critical to understanding public opinion in Africa than cultural, sociological or institutional ties and attributes. Furthermore, this article’s results with respect to institutional legitimacy mirror Bratton et al.’s empirical analyses on attitudes towards political and economic reforms: performance evaluations are more influential than partisan attachments. For reference, see Bratton, Mattes and Gyimah-Boadi, Public Opinion, Democracy, and Market Reform in Africa.

⁴⁷ There are minor variations in question wording by country. For the exact wording, see the Round 1, 12-country merged codebook (1999–2001) at 〈http://afrobarometer.org/round1m.html〉. All missing data and non-responses were dropped from the analysis using list-wise deletion.