Democratic Attitudes and Political Participation: An Exploratory Comparison across World Regions. Michael Bratton (Michigan State University)

Transcription

1 Democratic Attitudes and Political Participation: An Exploratory Comparison across World Regions Michael Bratton (Michigan State University) Paper prepared for the Congress of the International Political Science Association, Santiago, Chile, July The globalization of democratic ideals begs interesting research questions. To what extent do popular attachments to democracy vary across the world? What regional differences are observable in mass political behaviors? Are democratic attitudes and political participation related? If so, how strongly? And in terms of causes and effects which comes first? The literature on these topics is nascent, fragmented, and inconclusive. The major contribution to date has been the wave of studies based on the remarkable time series from World Values Survey (WVS), soon to cover a thirty-year period across multiple continents. Although the WVS probes deeply into subjective political, economic and religious values, it has relatively little to say about objective political behavior. 1 Moreover, Inglehart and colleagues subsume political participation as one dimension of "self-expression values," which is a syndrome of interpersonal trust, tolerance and political activism that contributes to the emergence and survival of democracy. Finally, analysis of so-called world values tends to privilege similarity over difference; the distinctive features of the world s various geographical, economic and cultural regions is downplayed in favor of a single human development sequence (Inglehart and Welzel 2005). The Global Barometer Surveys (GBS) have a separate set of strengths and weaknesses. Based on stand-alone barometer surveys for various world regions, South, Latin America, sub-saharan Africa, and the Arab Middle the GBS probes public opinion on the performance of political regimes and has more to say about mass action. So far, however, the GBS contains only a small module of common questions. In depth-analysis is mainly possible within world regions, as illustrated by recent books and reports on how Africans, ns and Latin Americans independently and contextually approach democracy (e.g. Bratton, Mattes and Gyimah Boadi 2005; Chu, Diamond, Nathan and Shin 2008; Corporacion Latinobarometro 2009). Exploratory efforts have only just begun to test prospective global generalizations about voter turnout, popular conceptions of democracy, and the sources of democratic legitimacy (Bratton, Chu and Lagos, 2006; Dalton, Shin and Jou 2007; Chu et al 2008; see also Diamond and Plattner 2008). Predictably enough, research to date suggests that democratic attitudes and democratic behaviors are related. But we know little about precisely which attitudes are connected to exactly which behaviors. Most importantly, the causal direction is unclear. As Brady, Verba and Schlozman have put it, political interest and political efficacy certainly facilitate political activity, but activity presumably enhances interest and efficacy as well. This chicken-and-egg dilemma has usually been addresses at the macro-level using aggregate country data. For example, following Almond and Verba (1963), Inglehart (1990) asserts that civic 1 Under the heading political action, the 2005 WVS data set available at lists variables measuring only petition-signing, boycotting and demonstrating, which hardly exhausts the range of possible participatory acts.

2 culture is a precursor of democracy; by contrast, Muller and Seligson (1994) prefer to see cultural values as a learned response to the experience of living in country that has a democratic regime (635). The few studies that have searched for causality at the micro-level tend to confirm that mass democratic attitudes are learned from participatory behavior, rather than vice-versa (Finkel 1987; Bratton et al 2005). Given the availability of new survey data, the time should be right to settle this matter definitively across more than one country or world region. Based on GBS data, this research note offers a preliminary description of popular democratic attitudes and mass participatory behaviors from several world regions. To the limited extent that one analyst is able to do so (and I emphasize that this is a binding constraint), the paper selectively proposes interpretations of these results. In general, I find that an individual s democratic commitments are linked to various aspects of political participation, though less consistently and more weakly than might be expected. Using an exploratory test, the paper proposes that political behavior seems to drive political attitudes, all the while recognizing that the relationship between attitudes and behavior is deeply endogenous. A key caveat is that analysts must standardize data collection efforts before the question about causal direction can be resolved. As such, this paper can be read as an agenda for the further development of the Global Barometers project. Barometer Surveys: How Global? We have come a long way since the idea of merging regional public opinion surveys into a global barometer was first mooted. Initially, when the Africa and Barometers were just getting started, and the South and Arab Barometers were still in the planning stages, each regional survey team learned extensively from the experiences of our colleagues who had launched comparative survey projects in Latin America and ern Europe. In time, we expanded our ambitions to include cross-continental comparison, which required the development of a common module of core questions for inclusion in each regional instrument. I won t pretend that reaching consensus on the content and format of the core module has been easy; understandably, each regional team has its own research priorities, preferred questions, and sunk costs in particular item wordings and response categories. Nevertheless, by 2005 we had converged on a module of 50 items and 83 variables centered on attitudes to democracy, institutional trust, political participation, media exposure, economic evaluations, and respondent demographics. While many questionnaire items are identical in both wording and response set others are at least functionally equivalent. The module has been applied in 55 countries across five world regions:, South, Latin America, sub-saharan Africa and the Arab Middle (see Table 1). As of July 2009, the GBS has gone on line with an accessible data analysis tool that enables users, even those without statistical software, to produce basic descriptive statistics marginals, crosstabs, graphics for the various countries. This paper employs this data set, though the analysis (starting with a grouping of countries into world regions ) goes beyond what can be accomplished on line. Despite progress in data collection, much remains undone in the quest for reliable cross-national generalization about public opinion. We still have work to do in improving sampling, extending coverage, standardizing survey items, confirming conceptual equivalences, and gradually expanding the module in order to allow greater degrees of analytic freedom. In this section of the paper, I identify additional issues that arise in trying to conduct a cross-region comparison of democratic attitudes and mass political participation. The intent is to seek solutions, not attribute blame. And, as a member of the GBS group, I must include myself in any criticisms. 2

3 Table 1: Global Barometer Surveys: Country Coverage by World Region South Latin America Sub-Saharan Africa Middle (9) Indonesia Japan Korea Mongolia Philippines Singapore Taiwan Thailand Vietnam (5) Bangladesh India Nepal Pakistan Sri Lanka (17) Argentina Bolivia Brazil Chile Colombia Costa Rica Ecuador El Salvador Guatemala Honduras Mexico Nicaragua Panama Paraguay Peru Uruguay Venezuela (18) Benin Botswana Cape Verde Ghana Kenya Lesotho Madagascar Malawi Mali Mozambique Namibia Nigeria Senegal South Africa Tanzania Uganda Zambia Zimbabwe (6) Algeria Jordan Kuwait Lebanon Morocco Palestine Source: GBS data set posted at Horizontal Coverage. Although Global Barometer surveys aim at comprehensive coverage, not all questions are asked across all countries. As Table 2 shows, whole world regions are excluded from numerous key items. For the purposes of the present study, the GBS data set includes seven items mapping democratic attitudes and eight items pertaining to political participation. But data are available for all five regions for only three of these items: support for democracy, rejection of strongman rule, and participation in protest demonstrations (see shaded cells). The gaps in participation are especially debilitating, since protest action is an unconventional form of participation; it would have been far preferable, for example to have complete data on voting. With information available for only 76 percent of the cells in Table 2, analysts continue to confront substantial problems of missing data. Vertical Coverage. Ideally, barometer surveys test the political and economic atmosphere at a common moment in time. But, as Table 3 reveals, the current GBS data set draws on survey results that stretch over a full five years from the onset of Round 1 of the Barometer in 2001 to the conclusion of Round 3 of the Afrobarometer in As such, any attempt to compare these two world regions should take into account changes in the global environment that occurred during this long interval, including the 9/11 attack on the World Trade Center, the resultant war on terror, and an emergent period of economic growth in Africa. One can easily imagine an alternate interval covering the onset of the global financial crisis in late 2007 that could easily skew comparisons across pre- and post-crisis surveys. Thus, in order to minimize the effects of time, the GBS must strive for greater simultaneity in data collection. 3

4 Table 2: Global Barometer Surveys: Availability of Data on Democratic Attitudes and Political Participation ( Horizontal Coverage by Region) South Latin America Sub- Saharan Africa Democratic Attitudes Meaning of Democracy X X X Characteristics of Democracy X X X Support for Democracy X X X X X Rejection of Strongman Rule X X X X X Rejection of Military Rule X X X Satisfaction with Democracy X X X X Suitability of Democracy X X X Political Participation Interested in politics X X X X Discuss politics X X X X Belong to voluntary organization X X X X Voted in last election X X X X Contacted government official X X X X Contacted elected official X X X Contacted traditional leader X X X Protested or demonstrated X X X X X X = data available (76 percent of cells) Highlight indicates questions for which data are available for five world regions Source: GBS data set posted at Middle Table 3: Global Barometer Surveys: Availability of Data ( Vertical Coverage Over Time) World Region Date of Survey Latin America 2003 South 2005 Middle 2005 Sub-Saharan Africa Source: GBS data set posted at 4

5 Data Weights. Although the unit of analysis in GB surveys is the individual citizen, GBS data sets contain higher-level composites countries and regions of vastly different population sizes (ranging from under half a million in Cape Verde to over a billion in India). How then should these units be weighted for purposes of making descriptive comparisons? Consider four alternative schemes for weighting world regions: (a) Simply use the raw data. But because GBS covers more countries in some regions than in others, regional sample sizes are highly unequal. Thus the raw data over-represent sub-saharan Africa (with over 25,000 cases) and under-represent the Middle (with under 7000 cases). (b) Standardize sub-sample sizes. By weighting all regions (e.g. at a mean of 16,400 cases) we could ensure that each region is represented equally within the sample. But this is a somewhat artificial solution, since it assumes that regions with small populations (like the Middle ) have equal influence on global generalizations as regions with vast populations (like South ). (c) Based on this concern, a better solution is to weight world regions according to their relative preponderance in the GBS sample. By this criterion, South ns represents some 46 percent of the current GBS universe and Middle erners just 3 percent. Recognizing that any weighting scheme involves a tradeoff, I nevertheless chose this GBS population weight as the operating norm for the analysis that follows. (d) A fourth solution, not attempted here, would be to weight world regions by their share of total world population. By this criterion, would account for 38 percent, South for 29 percent, and Middle for 7 percent. Although this distribution might seem attractive, it neglects the fact that the current GBS data set (though not all GBS surveys) excludes China, s population behemoth. More importantly, it would give the false impression that countries presently included in the GBS data set were chosen because they somehow represent their regions. As the more open societies in their regions (i.e. those amenable to reliable survey research), they clearly do not. Table 4 illustrates the consequential effects of various weighting schemes on global point estimates of public opinion. Take popular support for democracy. Without weights, some 57 percent of individuals interviewed express support, compared to 59 percent under equal weights and 52 percent using the preferred GBS population weight. I would argue that a realistic estimate of mass support for democracy must represent the population from which it was sampled. As we will see in the next section, the fact that South ns express less support for democracy than the citizens of other world regions tends to pull down the global point estimate. Table 4: Support for Democracy: Comparison of Global Point Estimates by Weighting Scheme No Weight 1 Equal Weight by World Region 2 GBS Population Weight 3 Global Population Weight 4 Prefer democracy Prefer non-democracy Doesn t matter Refused to answer

6 Which of these three statements is closest to your own opinion? A. Democracy is preferable to any other form of government B. In some circumstances a non-democratic government can be preferable C. For someone like me, it doesn t matter what kind of government we have 1. Raw data for five world regions (N= 82,017), with sample sizes ranging from n = 6,940 (Middle ) to n = 25,387 (sub-saharan Africa) 2. Five world regions standardized at n = 16,400 cases per region (N = 82,000) 3. Regions distributed by relative population size in a 55-country Global Barometer data set (N= 82,000) 4. Regions distributed by relative share in world population (N = 91,974) How Global? Beyond technical considerations, the quest for global generalizations raises concerns about concepts and interpretations. One risk is conceptual stretching. Especially in relation to subjective notions like support for democracy, but even with regard to more objective indicators of political participation, we must ensure that analysts and respondents alike, regardless of their region of origin, attach the same meanings to the same concepts. As Richard Rose notes, to ignore the context of attitude formation and political behavior, replaces the minor problem of missing data with the major problem of missing concepts (2007, 297). Another risk is interpretive ignorance. Curtice (2007, quoting Jowell 1998) comments that, the larger the number of countries included, the greater the likelihood that those who attempt to analyze the resulting data do not have sufficient understanding of the social, political, economic and cultural attributes of each country to be able to interpret the data sensitively and sensibly (2007, 900-1). The danger is that the extension of the reach of barometer surveys on a global scale and the publication of data sets in convenient on-line formats may encourage the user(s) to analyze data from countries (or regions) about which he or she may know little or nothing (903). Teamwork is a good solution, with analysis conducted by teams containing experts, preferably nationals, of each region covered. Until then, modesty is the best default position for analysts of global barometer data. In short, unless fortified with deep understanding of country and regional contexts, global generalizations are risky. Attitudes to Democracy This section of the paper, as summarized in Table 5, describes attitudes to democracy, drawing attention to major regional differences. Let us discuss each attitude in turn. Meaning of Democracy. Democracy is a meaningful concept to most citizens. When asked an open-ended question what, if anything, does democracy mean to you? (or words to that effect) 68 percent can offer a substantive definition of the term. Note, however, that only a minority 42 percent is able to propose more than one meaning for this complex concept. Two additional caveats are in order. First, the global mean is a crude estimate, constricted as it is by the absence of data from Latin America and the Middle in the present version of the GBS data set. 2 And, second, there are large cross-regional differences in the popular grasp of the meaning of democracy, ranging from 80 percent in to 47 percent in 2 The meaning of democracy question was apparently not asked in the 2003 Latinobarometro or 2005 Arab Barometer, though it was asked in Latinobarometro 2001as To you, what does democracy mean? What else? (Dalton et al. 2008, 13) 6

7 South. This stark regional gap might close if other countries, notably China, are added to the n sub-sample. 3 The most common meaning of democracy after don t know (40 percent) is a regime of civil liberties, which includes freedom of speech, freedom of association, and freedom in general (19 percent). Fewer than 5 percent respond to the open-ended question by referring to democracy as a regime of social rights or economic development. This common pattern of shared meaning across all world regions offers a prima facie justification for proceeding, at least cautiously, with comparative analysis of other survey questions that employ the word democracy. Table 5: Attitudes to Democracy, by World Region South Latin America Sub- Saharan Africa Middle GLOBAL MEAN Democratic Attitudes Meaning of Democracy Characteristics of Democracy Support for Democracy Rejection of Strongman Rule Rejection of Military Rule Satisfaction with Democracy Suitability of Democracy Source: GBS data set posted at with GBS population weight 1. Percentage of respondents who can offer a meaning to the term democracy (first response) 2. Percentage of respondents who associate democracy with the fulfillment of basic needs 3. Percentage of respondents who prefer democracy 4. Percentage of respondents who reject rule by a strong leader who decides everything 5. Percentage of respondents who reject the prospect that the army comes in to govern the country 6. Percentage of respondents who are fairly satisfied or very satisfied 7. Percentage of respondents who think that democracy is a suitable form of government for their country. Characteristics of Democracy. In a puzzling contradiction, however, the rank order of responses changes once a related question about the characteristics of democracy is posed in a closed-ended format. When asked to choose from a preset list of options, citizens place socioeconomic characteristics above political ones. On average, they associate democracy with the following, in priority order: (a) basic necessities like food, clothing and shelter (28 percent); (b) a small income gap between rich and poor (25 percent); (c) opportunity to change the government through elections (21 percent); and (d) freedom to criticize those in power (8 percent). The tendency toward economic instrumentalism is most marked in South and least so in the Middle (note: this question was not asked sub-saharan Africa and Latin America). But the large population share of the South n countries in the GB sample skews the global 3 Chu et al. (2008) report that, in 2001, only 65 percent of respondents in China could offer an answer to a cognitively demanding question phrased as What does democracy mean to you? China excluded from the current GBS data set apparently for reasons other than the availability of data. 7

8 mean: in both and the Middle, citizens prioritize electoral choice above the provision of basic needs in terms of what they expect from democracy. Support for Democracy. Just over half (52 percent) of the 82,000 citizens interviewed in GB surveys express support for democracy. The prompt is a standard question about whether democracy or a non-democratic government is preferable, with a third optional response being it doesn t matter what kind of government we have. The global mean for support for democracy is relatively reliable in the sense that it is based on data for all five of the world s regions. If it seems unexpectedly low, this outcome is again partly due to the suppressive effects of the heavily weighted South n cases, where almost one-third (31 percent) either don t know what kind of regime they prefer or say that, it doesn t matter. Nonetheless, the GB data reveal an unusually strong popular yearning for democracy in regions where political freedoms are largely lacking, such as sub-saharan Africa (63 percent) and the Middle (79 percent). Rejection of Strongman Rule. Of course, it is all too easy for respondents to pay lip service to support for democracy, a universal ideal. To test the depth of popular regime commitments, it is important therefore to explore whether self-professed democrats also detach themselves from authoritarian regimes. Majorities in all three regions studied (, South and Sub-Saharan Africa) concur that the army should stay in the barracks rather than coming in to govern the country. But people from various regions disagree on the desirability of government by a strong leader in civilian garb. The main contrast here is between South, where barely one quarter reject strongman rule, and the other four regions, where about three quarters do so. South ns apparently associate democracy with weak leadership and harbor a deep nostalgia for effective political management that can demonstrably get things done. Satisfaction with Democracy. We now turn to a concrete question about popular satisfaction with the way democracy (actually) works in each country. On average, fewer than half (46 percent) of all adults interviewed across 55 countries are satisfied either fairly or very with the performance of their elected regimes. In this regard, Africans (46 percent) and South ns (45 percent) are typical of the prevailing global norm. ns are unusual, not only because more than half express satisfaction with democracy, but also because more of them are satisfied than supportive. In this regard, the performance of the regime apparently exceeds low expectations about what democracy can deliver. Latin Americans display an opposite set of characteristics. They are only half as satisfied as supportive, which suggests that the performance of democratic regimes in this region have fallen well short of (perhaps excessively) high expectations. The Suitability of Democracy. Finally, do citizens consider that democracy is suitable, given the existing level of political development of their countries? In other words, do they feel that local circumstances are ripe to adopt democracy? A question on this topic was asked in only three regions, excluding sub-saharan Africa and Latin America. Responses were consistently positive, ranging from 61 percent in the Middle to 78 percent in. On average, two-thirds of all respondents in three world regions say that they find democracy suitable. Thus, notwithstanding diverse cultural legacies and levels of economic development, citizens of different regions apparently share a readiness to try democracy on for size. 8

9 Political Participation Popular political values alone do not a democracy make. In addition to studying attitudes, analysts must also ask whether citizens also behave politically in ways that reinforce and deepen democratic procedures and institutions. Do people actually take advantage of new opportunities for political action that a democratic regime supposedly offers? Conventionally defined, political participation consists of legal activities by private citizens that are more or less directly aimed at influencing the selection of government personnel and the actions they take (Verba, Nie and Kim, 1978, 46). As previously indicated, this paper takes a broad view political participation: it includes showing an interest in politics, discussing politics with others, and joining protest demonstrations; and it centers on conventional core activities like voting, joining voluntary organizations, and contacting political officials. With reference to Table 6, these activities are now discussed in the order they tend to occur in any open, democratic society. Table 6: Political Participation, by World Region South Latin America Sub- Saharan Africa Middle GLOBAL MEAN Political Participation Interested in politics Discuss politics Belong voluntary organization Voted in last election Contacted government official Contacted elected official Contacted traditional leader Protested or demonstrated Source: GBS data set posted at with GBS population weight 1. Percentage of respondents who say they are somewhat interested or very interested 2. Percentage of respondents who claim to discuss politics occasionally or frequently 3. Percentage of respondents who claim membership in a voluntary, including religious, association 4. Percentage who say that they turned out to vote in the last national-level election 5. Percentage reporting contact with a government official during the past year 6. Percentage reporting contact with an elected official during the past year 7. Percentage reporting contact with a traditional leader during the past year 8. Percentage of respondents who say they have participated in a protest or demonstration. Interest in Politics. As a prerequisite for autonomous participation in the political process, an individual must first evince an interest in politics. A count of those who say they are somewhat interested or very interested in politics puts Latin Americans at the forefront (73 percent), with Middle erners lagging behind (40 percent), perhaps reflecting the relative openness of political processes and the availability of opportunities for participation in each of these regions. By these criteria, however, one would expect ns to convey more interest than Africans, which is not the case (51 percent versus 66 percent). It seems likely, therefore, that the openness of society interacts with the intensity of political grievances in determining 9

10 actual levels of political interest. The fact that, as we have seen, ns express much higher levels of satisfaction with democracy than Africans or especially Latin Americans lends credence to this interpretation. Political Discussion. Interest in politics usually prompts people to engage in political discourse. In three world regions (Middle, Africa and ), the proportions of citizens who take part in political discussions with family, friends, neighbors and workmates closely parallel the proportions that express a sense of intellectual engagement. But Latin America appears to be an anomaly: whereas almost three quarters express interest, just over one quarter reports entering discussions; fully 71 percent say they never discuss politics. Unless there is a simple technical reason (for example a coding error in the GBS data set), this disconnection demands interpretation. Why do people in this region, including those who are cognitively predisposed to be active interlocutors, apparently feel constrained to engage in open political discussion? Organizational Membership. Civil society is a common portal into public activism. Voluntary organizations not only serve as venues of political discussion but, as interest groups, they project popular preferences, often in aggregate form, into the political process. Joining with others in a voluntary association, therefore, is itself an act of political participation. Yet the global mean value, which shows about half (51 percent) of all survey respondents engaging in this form of collective action, may be somewhat misleading. The reason is that there is a wider range in organizational membership across world regions than for any other indicator of participation. Whereas 80 percent of Africans claim to belong to a voluntary group usually a church, mosque or religious sect only 17 percent of Middle erners say the same thing. Since it would be hard to argue that Middle erners are less pious than Africans, one is forced to conclude that the people of these regions must have very different views about what it means to join a religious association. Voting. Even though voting lies at the heart of political participation in a democracy, different approaches to measuring voter turnout yield divergent results. Invariably, survey respondents over-report their electoral participation as compared to the official statistics produced by electoral management bodies. The Global Barometer surveys indicate a global mean of 79 percent who voted in their country s last election, a high level that implies inflated responses. Indeed, a comparison with a standard source reveals that a systematic margin of overreporting in three world regions, ranging from 11 percentage points in to 17 percentage points in South and sub-saharan Africa. 4 These discrepancies should be born in mind when comparing descriptive statistics across world regions, especially with the Middle, where over-reporting was negligible (just 2 percentage points). As long as over-reporting is randomly distributed among democrats and non-democrats, however, then self-reported voter turnout may be used as an acceptable proxy in preliminary efforts to explain possible linkages between democratic attitudes and political participation. Contacting Leaders. In periods between elections in representative democracies, citizens attempt to stay involved in politics by contacting political leaders. On average, only about a quarter of survey respondents report making such contacts in the past year, though levels 4 Official turnout was measured as total votes cast divided by voting-age population in the last parliamentary election before 2002 (International IDEA, 2002). This is a relevant comparator since GBS surveys were administered to a cross-section of the voting-age population in the period The official mean voter turnout figures for each region are: = 71 percent, South = 64 percent, sub-saharan Africa = 57 percent, and Middle = 52 percent. 10

11 vary by world region and type of leader. In a discouraging sign for democracy, citizens generally seem to find appointed civil servants or hereditary traditional leaders more accessible than elected officials. And, on average, ns are four times more likely than Middle erners to report such contacts. Finally, while it is obvious what Africans mean by traditional leaders local chiefs and headmen it is less clear (at least this analyst) exactly who occupies equivalent positions in. Protest. The founding human rights charters embody a citizen s right to peacefully rebel against unjust government. Popular protest therefore plays a defining role in attaining and sustaining democracy. On average, some 14 percent of citizens report attending a political protest or demonstration in the course of the previous year, a level exactly represented by Africans and Latin Americans and closely approached by South ns. Perhaps again reflecting relatively deep reservoirs of democratic satisfaction, ns report very little protest activity (just 4 percent take part). This stands in sharp contrast to the 27 percent of Middle erners who claim involvement in protest activity against a background where levels of popular satisfaction with democracy are presently unknown, at least as far as GB surveys are concerned. Democracy and Participation We now turn from description to analysis. With reference to democratic citizenship, are an individual s attitudes related to her behavior? Table 7: Support for Democracy and Rejection of Strongman Rule, Correlations by World Region World Region R Country Range R.173*** Japan Singapore South.038*** India Sri Lanka Nepal Latin America.055*** Paraguay Bolivia Panama Sub-Saharan Africa.178*** Ghana Mozambique Middle.143*** Lebanon Palestine.226*** *** *** *** ***.001 Global.134*** ***p<

12 As a first step in analysis, I explore the depth of personal democratic commitments. As argued earlier, democratic citizenship requires that an individual s professed support for democracy be backed up by a rejection of strongman rule. Table 7 shows that, in every region of the world, preferences for democracy are positively related to denials of dictatorship. The connection is not especially sturdy anywhere, implying that autocratic alternatives can never be counted out, but it is always statistically significant. Africans and ns are most likely to display a consistent commitment to democracy, regularly valuing it above authoritarian rule. The Japanese are the staunchest democrats among ns, as are Ghanaians among Africans. But South ns and, to a lesser extent, Latin Americans are somewhat more ambiguous about democracy since, for some members of these societies, a preference for popular rule apparently coexists with residual nostalgia for the heavy hand of a strongman. Within these regions, democratic commitments are least consolidated in Nepal and Panama, where preferences for democracy are negatively correlated with rejection of strongmen, which suggests that some citizens in these countries apparently perceive autocrats as being democratic! Despite these few exceptions, the overall pattern of correlation permits us to construct a composite measure of commitment to democracy. It is a three-point scale, where 2 = both support democracy and reject strongman rule, 1 = either support democracy or reject strongman rule and 0 = neither support democracy nor reject strongman rule. No such index construction is possible for political participation since we presently possess data for only one variable (protest) across all five regions. Thus Table 8 presents bi-variate correlation coefficients between commitment to democracy and eight single-item indicators of political participation. At this stage, I purposely employ a technique of statistical association that carries no implication of causality in either direction. Noteworthy results are as follows: (a) Of the eight modes of participation, seven are associated with commitment to democracy at meaningful levels of statistical significance. Only contact with government officials is unrelated, which suggests that democrats and non-democrats alike seek to influence bureaucrats. (b) Moreover, all are positively associated, implying that democrats evince a connected syndrome of attitudes and behaviors. The strongest association is between democratic commitments and discussion of politics, which suggests that democrats are active debaters in public settings. Whether commitment begets discussion, or vice versa, remains to be settled. (c) The one striking exception is voter turnout. While positive and significant within four world regions, this relationship is negative and significant at a global level. One possible explanation is that, while committed democrats and voting go together in South, there are so many ambiguous or uncommitted citizens who vote in this region that their presence swamps the link at the global level. This result stands as a stark illustration of the dangers of asserting global generalizations without understanding particular contexts. (d) There are clear distinctions in the attitude-behavior link across regions. In Table 8, the shaded cells indicate the region in which the strongest relationship pertains. The Middle leads other regions in the tie between democratic commitments on the one hand and, on the other, political interest and protest action. South leads with regard to political discussion, voluntary memberships and contacts with government officials. In, the bond is strongest between democratic commitments and other forms of political contact. And it is in sub-saharan Africa where the clearest link between democratic commitment and voting can be found, though this 12

13 link is not especially strong. Indeed, the voting connection is so humble that I am led to the general inference that, alongside committed democrats, many non-democrats are also active participants in open elections. Table 8: Commitment to Democracy: Correlations with Political Participation Global South Latin America Sub- SaharanA frica Middle Interested in politics.055***.118*** -.117***.141***.194*** Discusses politics.187***.065***.277***.100***.176*** Belongs to a voluntary organization.047***.076***.105***.071***.079*** Voted in last election -.025***.029**.052***.055***.048*** Contacted government official ***.163***.077***.028* Contacted elected official.052***.082***.071***.034* Contacted traditional leader.025***.131***.017*.002 Protested or demonstrated.035*** ***.043*** *** Cell entries are Pearson r correlations ***p<.001, **p<.01, *p<.05 Causes and Contexts Ideally, the next step in analysis would be a multivariate causal explanation. But the problem of missing observations in the GBS data set limits our ability to proceed beyond statements about bi-variate association among stand-alone variables. We are inhibited from proceeding to comprehensive causal analysis based on regression methods by at least three considerations. (a) First, it is difficult to construct comprehensive indices that can serve as dependent variables. At present we have only two democratic attitudes and one form of participation (protest) that are measured in the same way in all five regions of the world. (b) Second, regression procedures require a complete data set without missing cases. Most statistical packages automatically delete all cases for which there is even a single missing observation. Under these circumstances, any attempt to run a regression with up to seven or eight predictor variables would rapidly diminish the number of available data points. (c) Third, given that certain survey questions were never asked in entire world regions, the empty cells in the 2005 GBS data set are clearly not missing at random. In this instance, the systematic characteristic of missing-ness makes it impossible (or at least inadvisable) to use techniques of data imputation to fill gaps, whether by simply inserting mean values or employing advanced techniques of multiple imputation. All that is defensible at this stage, therefore, are indicative investigations. In Table 9, I use OLS regression to begin to explore two questions about the development of democracy. Which come first: democratic attitudes or democratic behaviors? And, where does most variation in attitudes and behaviors reside: among individuals or across world regions? For reasons spelled out above, I cannot pretend that these tests are definitive. They simply provide 13

14 hints of relationships that may be embedded in the more complete data sets that have yet to be assembled. The four regression models in Table 9 employ different combinations of eight variables: one democratic attitude (an additive index of an individual s commitment to democracy), two participatory behaviors (indicators citizen involvement in protest and political discussion) and five world regions (each scored as a dummy variable). Indicators for the world regions are added to the pooled data set to control for fixed geographical, cultural and economic effects. The dependent variables for each model are designated with a DV in Table 9. The shaded cells show which independent variables are operating in each model. Finally, an upper case X marks a variable not entered, a lower-case x an excluded dummy variable, and an m stands for missing data. Table 9: Attitudes or Behaviors? Individuals or Regions? Indicative Regression Models Model 1 Model 2 Model 3 Model 4 Democratic Attitude Committed to Democracy DV.050*** DV.171 Political Participation Protested or demonstrated.044 DV X X Discussed politics X X.166 DV World Regions Latin America Sub-Saharan Africa x x South x x Middle m m Constant 1.382***.107***.438*** 1.456*** Variance explained (adjusted r square) DV = dependent variable X = variable not entered x = excluded dummy variable m = missing data Otherwise, cell entries are standardized OLS regression coefficients (beta). All b coefficients (not shown) are statistically significant at***p<.001. Let us begin with the question of causal order: do attitudes drive behavior or vice-versa? In Models 1 and 3, the object of explanation is an individual attitude, namely commitment to democracy. In the other models, the objects of explanation are citizen behaviors: participation in a protest or demonstration (in Model 2) and engagement in political discussions (in Model 4). Which of these sets of models has greater explanatory power? Take the amount of explained variance is a guide: the adjusted r square statistics for Models 1 and 3 are consistently higher than for Models 2 and 4. While hardly conclusive, this outcome implies 14

15 that democratic behaviors (protesting and discussing) have more effect on attitudes (commitment to democracy) than the other way around. In other words, citizens in new democracies around the world seem to learn by doing, that is, to develop their commitments to democracy as a result of having engaged in discussions or demonstrations. Although they may also take action as a result of having previously adopted democratic values (inevitably, the relationship is endogenous), the effect of attitudes on behaviors seems to be weaker than the effect of behaviors on attitudes. Once a more complete data set is available, we will proceed to test this finding more rigorously using simultaneous equations modeling. And what about the source of these variations? At what level of analysis is there greater fluctuation in attitudes and behavior? Does most variance in the attitude-behavior relationship occur among individual citizens or across world regions? In every model in Table 9, there is at least one (and usually more than one) region that has a greater impact on the dependent variable than the operative independent variable. For example, in Model 1, the negative effects on commitment to democracy of the South region are far stronger than the positive effects of protest action. And in Model 3, although the effects of political discussion are quite strong and positive for commitment to democracy, they are relatively weak compared to the powerful positive effects of the regional dummies for, Latin America and sub-saharan Africa (at least when these three regions are compared to the excluded category of the Middle ). In short, the impact of behavior on attitude formation seems to be contextspecific. In becoming a committed democrat, it matters more that one is an African (and not a South n) than whether one chooses to engage in discussion or protest. In other words, when seeking the sources of democratic commitments, we neglect at our peril the regional (or even the country) context in which they are embedded. It is for this reason, that I think there is more analytic leverage to be gained by sustaining a strong emphasis on the contexts in which democratization is occurring than on trying to subsume it into a single human development sequence on a global scale. 15

16 References Almond, Gabriel and Sidney Verba The Civic Culture. Boston: Little Brown Brady, Henry, Sidney Verba and Kay Schlozman Beyond SES: A Resource Model of Political Participation, American Political Science Review, 89, 2: Bratton, Michael, Robert Mattes and E.Gyimah-Boadi Public Opinion, Democracy and Market Reform in Africa. New York: Cambridge University Press. Bratton, Michael, Yun-han Chu and Marta Lagos Who Votes? Implications for New Democracies. Global Barometer Surveys. Chu, Yun-han, Larry Diamond, Andrew Nathan and Doh Chull Shin How ns View Democracy. New York: Columbia University Press. Chu, Yun-Han, Michael Bratton, Marta Lagos, Sandeep Shastri and Mark Tessler Public Opinion and Democratic Legitimacy. Journal of Democracy, 19, 2: Dalton, Russell, Doh Chull Shin and Wally Jou Understanding Democracy: Data from Unlikely Places. Journal of Democracy, 18, 4: Diamond, Larry and Marc Plattner How People View Democracy. Baltimore: Johns Hopkins University Press. Finkel, Steven The Effects of Participation on Political Efficacy and Political Support: A Panel Analysis, Journal of Politics, 49, 2: Inglehart, Ronald Culture Shift in Advanced Industrial Society. Princeton: Princeton University Press. Inglehart, Ronald and Christian Welzel Modernization, Cultural Change and Democracy: The Human Development Sequence. New York: Cambridge University Press. International Institute for Democracy and Electoral Assistance (IDEA) Voter Turnout Since 1945: A Global Report. Stockholm: International IDEA. Latinobarometro Report. Santiago: Corporacion Latinobarometro. Muller, Edward and Mitchell Seligson Civic Culture and Democracy: The Question of Causal Relationships, American Political Science Review, 88, 3: Rose, Richard Perspectives on Political Behavior in Time and Space, Ch.15 in The Oxford Handbook of Political Behavior. Oxford: Oxford University Press, Verba, Sidney, Norman Nie and Jae-On Kin Participation and Political Equality: A Seven-Nation Comparison. New York: Cambridge University Press. 16