IDENTITY COMPETITION IN NEW DEMOCRACIES: WHICH IDENTITY WINS IN WHAT CONDITIONS?

The relationship between different identity dimensions of social cleavages and democratic stability has been a renewed topic of considerable interests among scholars of comparative politics in recent years. Following the early works such as the ones of Rabushka and Shepsle, Lijphart, Horowitz, and Bartolini and Mair, a new set of works such as those of Posner, Chandra, Birnir, Norris, and Bates show how social cleavages and identity politics affect electoral stability in particular and democratisation process in general.1 Despite a growing literature on social identity  

Zahid ul Arefin Choudhury, PhD is Assistant Professor and Chairman of the Department of Peace and

Conflict Studies, University of Dhaka. His e-mail address is: zahid.choudhury@du.ac.bd. An earlier version of
this paper was presented in the MPSA Conference held on 1-5 March 2010, Chicago, Illinois, USA; and at the
Asiatic Society of Bangladesh on 23 November 2013.
© Bangladesh Institute of International and Strategic Studies (BIISS), 2015.
 1Alvin Rabushka and Kenneth A. Shepsle, Politics in Plural Societies: A Theory of Democratic Instability, New
York: Pearson Longman, 2009; Arend Lijphart, Democracy in Plural Societies: A Comparative Explanation,
New Haven and London: Yale University Press, 1977; Donald L. Horowitz, Ethnic Group in Conflict, Berkeley:
University of California Press, 1985; Stefano Bartolini and Peter Mair, Identity, Competition, and Electoral
Availability: The Stabilization of European Electorate 1885-1985, New York: Cambridge University Press,1990;
Daniel N. Posner, Institutions and Ethnic Politics in Africa, New York: Cambridge University Press, 2005;
Kanchan Chandra, “What is Ethnic Identity and Does It Matter?”, Annual Review of Political Science, Vol. 9,
2006, pp. 397-442; Johanna Kristin Birnir, Ethnicity and Electoral Politics, New York: Cambridge University
Press, 2007; Johanna Kristin Birnir, “Divergence in Diversity? The Dissimilar Effects of Cleavages on Electoral
Politics in New Democracies”, American Journal of Political Science, Vol. 51, No. 3, 2007, pp. 602-619; Pippa
Norris, Driving Democracy: Do Power-Sharing Institutions Work? New York: Cambridge University Press, 2008;
Pippa Norris, Electoral Engineering: Voting Rules and Political Behavior, Cambridge: Cambridge University

that distinguishes ethnicity from religion in functional terms2 , the vast set of social cleavage literature3 has largely ignored to differentiate ethnic identities (such as the ones based on language or race) from religious identity and to see how each of these categories fares in terms of democratic stabilisation. Posner does ask the question of why political competition comes to be organised along the lines of one ethnic cleavage rather than another, but he does not tease out the context in which religious cleavages vis-à-vis other ethnic cleavages rise or fall.4 Birnir improves upon Posner in treating religion separately, but she leaves out the contextual question as well.5

Basing on these recent works on social cleavages and social identity, the objective of this paper is to suggest and test a theory that captures how religious identity competes with other identities in new democracies, and in what context Muslim identity, a subset of religious identity and the special focus of this paper, rises or falls.  

The paper has been organised as follows. The relevant literature is reviewed to place the research questions of this paper into the broader field of social identity and electoral stability. A theoretical framework then tries to generate a hypothesis followed by a section on case selection and procedures to be used to test the hypothesis. The hypothesis is tested in three steps. First, an ordinary least square (OLS) regression establishes whether ethnic and religious identities explain electoral stability for initial elections for about 58 new democracies. Second, a set of descriptive statistics tracks electoral stability for subsequent elections (up to date) of a subset of the 58 countries. Finally, a few measures from the World Value Survey 1995 and 2000 are used to test whether Muslim identity rises or falls in Muslim majority countries and explain why for yet another subset from the second step. The final section discusses the conclusions and implications to be drawn from this exercise, its potentials and limitations and a few future steps to be considered.  

Social movement literature, particularly the ones on political opportunity and resource mobilisation, suggests that political opportunities facilitate collective mobilisation and political activism of groups who have been excluded from participation. 

6

Press, 2004; and Robert H. Bates, When Things Fell Apart, New York: Cambridge University Press, 2008. 2 Steven Grosby, “Nationality and Religion,” in Montserrat Guibernau and John Hutchinson (eds.),
Understanding Nationalism, Malden, Massachusetts: Polity/Blackwell, 2001; Kolås Tanveer Fazal, “Religion,
Language and Nationhood in Pakistan and Bangladesh”, in Rowina Robinson (ed.), Sociology of Religion in
India, Sage Publications, 2004, pp. 275-301; Jeffrey R. Seul, “Ours is the Way of God: Religion, Identity, and
Intergroup Conflict”, Journal of Peace Research, Vol. 36, No. 5, 1999, pp. 553-569; Rawi Abdelal, Y. M. Herrera,
A. I. Johnston and R. McDermott, “Identity as a Variable,” Perspective on Politics, Vol. 4, 2006, pp. 696-711. 3
 With notable exception of Daniel N. Posner op. cit., and Johanna Kristin Birnir, op. cit. 4
 Daniel N. Posner, op. cit. 5
 Johanna Kristin Birnir, op. cit. 6
 Doug McAdam, Political Process and the Development of Black Insurgency 1930-1970, Chicago: University 

But, the groups that have the ability to mobilise resources will have more advantages over those who do not have such abilities in exploiting the opportunity. As McAdam argues, “A conducive political environment only affords the aggrieved population the opportunity for successful insurgent action. It is the resources of the minority community that enable insurgent groups to exploit these opportunities. In the absence of those resources the aggrieved population is likely to lack the capacity to act even when granted the opportunity to do so.” 7 

McAdam further suggests that any form of social movement is only possible when an indigenous infrastructure or organisation is able to translate the resources into an organised campaign of mass political action.8 Other social movement theorists such as Tilly and Tarrow agree on the centrality of organisational ability of groups in social movement. However, they leave out the question of identity competition, particularly in the context of electoral politics.9 It is not clear from this literature whether, given a political opportunity, minority groups will be able to mobilise more if they organise around their ethnic identity (such as language) rather than their religious identity (such as Islam), if all other things are equal. 

Although the social movement literature is oblivious to the question of identity competition, it is still possible to infer from the powerful general framework it offers. For example, while continuing (from the above quotation) on the central role of organisation in social mobilisation McAdam argues that such organisation is a function of four factors - members, established structure of solidarity incentives, communication network and leaders.10 This indicates that in the context of competing identities (e.g. between ethnic and religious identity), if an ethnic identity has stronger organisation (all four factors of McAdam) than the religious identity then that ethnic identity will be more likely to successfully mobilise people. This was particularly the case in the post-colonial Pakistan, where linguistic identity of the Bengalis of the eastern wing of the country was able to overcome their Muslim identity that eventually allowed Sheikh Mujibur Rahman, the leader of the Bengalis, and his party Awami League to win the 1970 election overwhelmingly leading to the dismemberment of the country in 1971.11 This is because of the fact that language provided the ethnic group a unique identity that separated them from their out-groups with whom they shared the same religion. The normative implication of this analysis is that generally religion has wider audience than do such ethnic cores as language or race. Ethnic identity provides the concerned group the uniqueness it needs to mobilise and organise resources to  

of Chicago Press, 1999; Charles Tilly and Sidney Tarrow, Contentious Politics, London: Paradigm Publishers

2007.
7
 Ibid., pp. 43-47. 8 Ibid., p. 44. 9
 Charles Tilly and Sidney Tarrow, op. cit. 10 Doug Mc Adam, op. cit. 11 Amena Mohsin, “Language, Identity and the State in Bangladesh”, in Michael E. Brown and Sumit Ganguly
(eds.), Fighting Words: Language Policy and Ethnic Relations in Asia, Belfer Center for Science and International
Affairs, Harvard University: MIT Press, 2003.

participate in electoral competitions with its out-groups given a political opportunity of democratic elections. 

The early cleavage literature establishes the fact that social cleavages stabilise electoral politics and democracy. However, such stabilisation of democracy is conditional on the institutional arrangement of the society.12 For example, Lijphart argues that democracy is the best form of government even in the context of pluralism (multiple cleavages). Consociational democracy – a grand coalition representing the cleavages, multi-veto players, proportional representation of the cleavages, and segmented autonomy and federalism – assures democratic stability in the context of multiple cleavages. He further argues that at low levels of pluralism, Westminster (majoritarian) democracy is best but as pluralism increases, consociational democracy is the best.13 Horowitz makes a similar argument that “Split domination—an arrangement in which the key institutions of the society are dominated by different ethnic groups—may provide the basis for a bargain to stabilize this balance of power by recognizing ethnic spheres of influence”.14 Bartolini and Mair take this argument further and show that in the context of electoral politics, cleavages produce “fundamental bias towards stability” (defined as vote stability among parties).15 They empirically show for 38 European countries that with the increase in the level of cultural heterogeneity, level of total electoral volatility decreases. 

Altogether the early social cleavage literature, therefore, claims that social plurality increases democratic stability. The core of the argument is that given a political system that guarantees equity among all political cleavages in political decision-making process, participation of these cleavages in elections provides stability to the overall electoral politics of the political system. Advancing this general argument, Lijphart and Norris focus on the general determinants of social cleavage support for particular parties.16 However, this literature on social cleavages largely leaves out the question of ‘which cleavage?’ Furthermore, they take identities or cleavages as almost fixed entities. Rabushka and Shepsle make such an assumption of fixed cleavage identity even more strongly, although they make a diametrically opposite argument that ethnicity (multiple ethnic cleavages and identity) always destabilises democracies.17 Taking the issue with this assumption of fixedness, Chandra argues, “these families of theories have not demonstrated on analytical grounds that ethnic identity categories, as they classify them, should have an explanatory effect on the outcomes of interest.”18 In sum, the early cleavage literature does not provide with enough clue on identity (or cleavages) competition, and how that might affect (electoral) stability. Cleavage literature also does not provide with suggestions as to  

12 For example, Stefano Bartolini and Peter Mair, op. cit.; Donald Horowitz, op. cit., and Arend Lijphart, op. cit. 13 Arend Lijphart, op. cit., p. 237. 14 Donald Horowitz, op. cit., p. 447. 15 Stefano Bartolini and Peter Mair, op. cit., p. 68. 16 Arend Lijphart, op. cit.; Pippa Norris, op. cit. 17 Alvin Rabushka and Kenneth A. Shepsle, op. cit. 18 Kanchan Chandra, op. cit., p. 422. 

why and in which context some cleavages reduce electoral volatility, while others in some other contexts do not. 

The latest literature on ethnicity and identity politics addresses some of the above questions. For example, Posner distinguishes between tribal versus linguistic identity, and argues that institutions determine when politics revolves around one of the bases of ethnic divisions rather than the other. He argues that “they also shape people’s incentives for selecting one of these potentially salient ethnic identities rather than others, and coordinate these choices across individual so as to produce a society level outcome.”19 Posner concludes that in Zambia tribal identities have tended to dominate during one-party rule, but linguistic identities have tended to dominate under multiparty rule. His argument is based on the general propositions of instrumental rationality: people are interested to get resources from the state, “they believe that having someone in the power position from their ethnic group will facilitate their access to these resources”, and finally, coalition formation is the best way to get someone from their ethnic group into the power position.20 This interesting argument implies that people are not passive bodies on which ethnicity or any other form of identity acts upon, rather they are actors themselves, and they are potent enough to choose between identities to serve their strategic interest. However, Posner leaves out religion as a separate identity from the identity politics.21 Following Bartolini and Mair, Birnir adds to the literature by adding religion as a variable in the context of electoral stability. Using the common measure of volatility -“percentage of votes gained by one party and lost by any other party in each pair of elections” - Birnir shows that “ethnic minority populations, who are represented through ethnic parties, are initially more stable in their voting behaviour than the non-ethnic majority.”22 She concludes that “language cleavages stabilize voting immediately in new democracies, while voting associated with other cleavages, such as religion, takes more time to develop”.23 This is an important addition to the literature. But, Birnir leaves out the questions of context: in which condition is religion powerful and in which it is not? More importantly, she makes the mistake of assuming that the scope of religion as an identity is equal to the scope of language and race as identities. As it is argued in this paper, there exists a fundamental difference between all other identities and religious identity in terms of their respective scope. While all other identities tend to be local and relatively stable24 (not fixed though), religious identity not only tends to be greater in scope (global and adherent population from all other identity groups) but also is able to increase or decrease its scope over time (e.g. through conversion and population increase).25 

19 Daniel N. Posner, op. cit., p. 5. 20 Ibid., p. 3. 21 Daniel N. Posner, op. cit. 22Johanna Kristin Birnir, “Divergence in Diversity? op. cit., p. 609. 23 Ibid., p. 717. 24Johanna Kristin Birnir, Ethnicity and Electoral Politics, op. cit.; Kanchan Chandra, op. cit.; Rawi Abdelal et al.,

Identity as Variable, op. cit. 25 Steven Grosby, “Nationality and Religion,” in Montserrat Guibernau and John Hutchinson (eds.),
Understanding Nationalism, Malden, Polity/Blackwell; Jonathan Fox, "Towards a Dynamic Theory of Ethnoreligious Conflict", Nation and Nationalism, Vol. 5, No. 4, 1999, pp. 431-463. 

Fox treats the general question of religious identity and ethnicity analytically and more comprehensively. He defines religious identity to have four social functions where religion provides (i) a meaningful framework for understanding the world; (ii) rules and standards for behaviour that link individual actions and goals to this meaningful framework; (iii) links individuals to a greater whole and sometimes provides formal institutions which help to define and organise that whole; and (iv) the ability to legitimise actions and institutions.26

Like Fox, Seul also observes that religion provides individuals and groups with meaningful framework of life. He argues that “religions frequently supply cosmologies, moral frameworks, institutions, rituals, traditions, and other identitysupporting content that answers to individuals’ needs for psychological stability in the form of a predictable world, a sense of belongings, self esteem, and even selfactualization”.27 Both the authors suggest that if the “meaningful framework” is threatened religious identity will generate conflicts with the threatening identities (mostly other religions) or groups. Although religious identities tend to highlight the life larger than the earth, they tend to engage in mundane politics on regular basis in their quest to capture the state and vital social resources. Seul makes the point that “Conflicts between religious groups typically are caused by the same material factors and social dynamics that incite and fuel conflict between ethnic, racial, and other identity groups”.28 Fox’s and Seul’s social-psychological analyses of how religion facilitates inter-group competition and conflict make important contribution to the understanding of identity competition.  

However, it is yet to be seen how this understanding of religion can be integrated to the literature on cleavages, ethnicity and electoral stability. This is the task that has been taken up in this paper. What is highlighted here is the importance of investigating how religious identities compete with other sorts of identities such as language or race, not just with other religious identities. 

The existing literature on social cleavages, identity and democratic stability indicates the following points: first, multiple social cleavages and identity groups affect (positively) the democratic (electoral) stability of a political system; second, rational voters take cues from their core identities and want to establish that their groups get the desirable share of the power and the resources of the state; third, since successful social mobilisation relies on the in-group’s organisational strengths, given a political opportunity people want to strategically select their identity based on their perception of their inner-group-organisational strengths; fourth, in the voters’ 

26 Fox, ibid., p. 445. 27 Jeffrey R. Seul, “Ours is the Way of God: Religion, Identity, and Intergroup Conflict”, Journal of Peace

Research, Vol. 36, No. 5, 1999, p. 553. 28 Ibid., p. 564. 

analyses of comparative organisational strengths, religion becomes an effective candidate vis-à-vis other ethnic identities as religious identity can be equally as mundane and salient as its counterparts; fifth, as argued by Birnir, during the initial electoral phases ethnic identity (particularly language) based parties stabilise votes.

Building on the above literature, it is argued that religious identities compete with other sorts of identities. While during the initial phase of electoral political opportunity, core identities (other than religion) tend to stabilise electoral politics, during the later phases religious identities rise over other identities as electorally important instrument. Figure-1 shows why it is the case. 

In Figure-1, ethnic diversity leads to electoral stability, that leads the rise of religious identity based politics. The texts on the causal arrows summarise the mechanisms that ensure the causal link between the elements. Take the first causal link first: between ethnic diversity and electoral stability (the definitions of these and other important terms are given in the measurement section, later in this paper). There are two mechanisms at work that connects ethnic diversity and electoral stability. First, voters get their cues from ethnicity. Birnir makes this argument using a formal model: the main argument is that voters are utility maximisers, and their cost benefit analyses are based on their policy preferences with respect to the political party’s policy preference.29 Since the voters want to have access to the state resources, they make sure that the party they choose represents them almost exclusively so that they get the entire resource that the party can extract from the state. On the other hand, since the party wants to get the median voters within the cleavage that it represents, it will listen to the voters. Ethnic parties will not have constituencies outside their own ethnic boundaries. Second, as argued in the previous section, ethnic organisations mobilise voters to take part in the democratic process once the opportunity comes. They mobilise these voters toward a single party representing them. The paper  

29 Johanna Kristin Birnir, “Divergence in Diversity?”op. cit.  

argues that both of these mechanisms exclude religious identity because voters want exclusive representation, which is generally not possible through religious parties. Religious parties tend to have broader constituency than parties based exclusively on such features as language, race, motherland etc. Religious parties such as Muslim Brotherhood and Jamaat-e-Islam, even tend to draw supports from the national territory. During the initial phase of democratisation, ethnic voters do not have complete information about the resources they will be getting from participating in the national politics, and thus they will not feel secured with such parties since they will fear losing their resources to some unknown or rival groups. As a result, religious organisations will not be able to gather enough resources and support to mobilise politically during the initial period of democratisation. This leads to the following proposition: 

Proposition-1: During the initial phase (first couple of elections) core ethnic identities will fare much better than religious identities in stabilising electoral democracy.

The second causal link states that electoral stability will facilitate religious identity in the long run. This, however, is a conditional statement. Religious identity will be able to attract enough voters only in the context of continued electoral stability. The argument here is that once the ethnic parties are established in a stabilised (electoral) system they become part of the broader institutional status quo. If voters want change, these parties are no longer helpful. This causal link is further conditional upon a number of factors identified by Nasr, which are particularly typical to new Muslim democracies: strong presence of military, emerging private sector (capitalism) and competition over voters.30 Nasr argues that in most Muslim democracies military, even after several free and fair elections, remains a major de facto power player. He identifies three effects of such military presence: “First, it limited the [radical] Islamist’s room to manoeuvre. Second, it gave all parties an incentive to avoid confronting the military while angling for advantage within the democratic forces. Finally, the military’s meddling in politics led to more elections, political realignments, and shifts in coalitions, accelerating and intensifying experimentation with new political formulas.”31  

As a result, on the one hand, radical Islamist parties start to involve in “pragmatic politics,” while on the other hand, ethnic voters start to get disillusioned about their exclusive status in their own political party as well as in the system in general.32 Regarding capitalism, Nasr observes that “the less state-dependent and more integrated into the world economy a country’s private sector is, the more likely is that country to see Muslim Democracy gain traction as a political force … [because] Muslim Democracy combines the religious values of the middle and lower-middle classes with policies 30 Vali Nasr, “The Rise of Muslim Democracy”, Journal of Democracy, 

Vol. 16, No. 2, 2005, pp. 13-27. 31 Ibid., pp. 17-18. 32 Ibid. 

that serves their economic interests.”33 Regarding competition of voters, since no one party can easily dominate the system, all parties try to increase their issue dimensions to attract more voters (even from its own ethnic group) by incorporating issues that are not necessarily salient to the ethnic groups they represent. Nasr observes that “regular competitive elections have both pushed religious parties toward pragmatism and pulled other parties into more diligent efforts to represent Muslim Values.”34 In addition to these three factors, bad governance and system-wide political corruption make voters frustrated. As corruption permeates across ethnic boundaries and the quality of governance remains poor for a considerable period of time despite repeated experimentation with reform measures, voters tend to accept bargains from Islamist parties who offer to bring in system-wide reforms in favour of Shari’a law repudiating all current forms of politics that are inherently corrupt. The net effect of all these is that in the context of stabilised (electoral) democracy, while on the one hand voters across previous boundaries of ethnicity begin to move toward Islam, on the other hand, political parties in order to catch up with the voters and constrained and instructed by the military move toward the Islamic middle-ground, the secularised Islam, or what Nasr calls Muslim Democracy. This leads to the following proposition: 

Proposition-2: As the democracies with Muslim population pass the initial democratisation period (first couple of elections) and institutionalise electoral politics (many ‘free and fair’ elections), religious identity trumps other identities over time. 

Combining propositions 1 and 2, the following hypothesis can be proposed for testing in this paper. 

Hypothesis-1: In diverse new democracies, ethnic identities trump over religious identity, while as the system stabilises electorally religion, especially Islam, gradually overcomes other identities, particularly in countries with substantive number of Muslims population.
 

Before delineating the methodology to be employed in this paper, it is essential to define major concepts and terminologies used in the theoretical framework and arguments of this paper. 

33 Ibid., p. 18. Nasr uses the phrase “Muslim Democracy” to indicate such countries as Bangladesh, Indonesia,

Malaysia, Pakistan (before its 1999 military coup), and Turkey where since the early 1990s political arena
has been opened up allowing Islamic-oriented (but non-Islamist) parties to successfully vie for votes in
elections. According to Nasr, “Muslim Democrats view political life with a pragmatic eye. They reject or
at least discount the classic Islamist claim that Islam commands the pursuit of a Shari‘a state and their
main goal tends to be the more mundane one of crafting viable electoral platforms and stable governing
coalitions to serve individual and collective interests – Islamic as well as secular – within a democratic arena
whose bounds they respect, win or lose.” , ibid., p. 13. 34 Ibid., p. 19. 

According to Lijphart, “Democracy” is a “synonym of what Dahl calls “polyarchy”. It is not a system of government that fully embodies all democratic deals, but one that approximates them to a reasonable degree”.35 In addition, in order to highlight the centrality of election in such a democracy, the paper uses Schumpeter’s minimalist definition of democracy that is marked by the institutional arrangements that ensure “the competitive struggle for people’s vote.”36 New democracy refers to “any newly established democracy, whether the prior authoritarian regime resulted from domestic events or foreign control”.37

According to Bartolini and Mair, electoral stability or volatility is “a measure of the net electoral change between two consecutive elections.”38 According to Birnir, this (electoral volatility) means “the percentage of votes gained by any one party and lost by any other party in each pair of elections.” The standard measure is given by Przeworski:39 

where “volatility between elections at t and t+1 is measured as half the sum of the absolute difference between vote shares (P) of all parties in each election. The shares are taken without their sign in the aggregate and half of the observed difference is used”.40 Here the point of interest is the levels of aggregate volatility of a country for a pair of elections. Birnir’s data is useful; it is derived by using the above measure for the first couple of election for 58 new democracies.41 Later in the paper the same measure is used to derive volatility index for a subset of these 58 countries for all subsequent elections. 

Rise of Islam: Rise of Islam represents what Nasr calls Muslim Democracy. In the context of this paper, this simply means an increased number of people in the (majority) Muslim Democracies who appreciate the value of religion in their daily life.42

Ethnic, Linguistic and Religious Identity and Diversity: Alesina et al. offer measures of fractionalisation for 190 countries.43 Using the following formula they measured ethnic, linguistic and religious fractions:   

35 Lijphart, op. cit., p. 4. 36Joseph A. Schumpeter, Capitalism, Socialism, and Democracy, London and New York: Routledge, 1994, p. 269. 37 Johanna Kristin Birnir, “Divergence in Diversity?”, op. cit., p. 609. 38 Ibid., p. 19. 39 Cited in Ibid. 40 Ibid., p. 21 41 Johanna Kristin Birnir, Ethnicity and Electoral Politics, op. cit. 42 Ibid.

43Alberto Alesina, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and Romain Wacziarg,
“Fractionalization”, Journal of Economic Growth, Vol. 8, 2003, pp. 155-194.

where s2 ij is the share of group i (i = 1 … N) in country j. The index varies from 0 in a perfectly homogenous country to 1 when every individual belongs to a different group. The idea here is to calculate the probability that the two randomly selected individuals belong to two different groups. 

The fractionalisation data of Alesin et al. are used to create major dependent variables of interest: ethnic, linguistic and religious fractions for 58 countries. Ethnic and linguistic fractions from Alesina et al. then used to create an independent measure of ethnic identity for this paper. Due to addition of two 0-1 measures, the new index varies from 0 in a perfectly homogenous country to 2 when every individual belongs to a different group. The religion measure of Alesina et al. is kept intact. The dependent variable here is the electoral volatility or stability, as measured by Birnir for 58 countries.44 The Inter-Parliamentary Union online database is used to gather party-wise national parliamentary elections data for all elections for a selected number of country, that has been then used to calculate electoral stability index for these countries using equation 1 above. To see if people’s appreciation of religion has increased over time, the World Value Survey database 1995 and 2005 are used. Other data such as population, percentage of population of the Muslims, regions, and economic indicators came from the online database of the World Bank and Penn World Table. 

The analysis of this paper is divided into two steps to separately test the propositions mentioned above. For proposition-1, this paper considers 58 cases that include all new democracies. Ordinary Least Square (OLS) regression analysis is done for all these countries to test this proposition.

Proposition-2 is tackled by a descriptive analysis of electoral volatility of a group of countries that are selected by a scheme illustrated in Figure-2. Considering the figure, the goal is to reach to the quadrants of the 2 X 2 table inside the innermost circle, where the circles represent the set of all new democracies and its stable and Muslim majority subsets. The innermost circle includes the countries that are ‘Muslim majority stable democracies’ (initial elections). Since proposition-2 stands only for these countries, other countries are not selected for this paper. Within these Muslim majority stable democracies, the paper allows for variations along lines of two variables, ethnicity and religion, that are measured as continuous indices, as provided by Birnir and Alesina et al. Means of these variables (see table-1) are used as cut points to dichotomise them into high and low, which allows the paper to approximately  

44 Johanna Kristin Birnir, “Divergence in Diversity?”, op. cit. 

segregate countries that are first, stable countries, second, majority Muslim countries, and then countries with high/low religious fractions and countries with high/low ethnic fractions. Detailed categories of countries along these lines are given in Appendix Table-1.

To further illustrate the analysis with a descriptive comparative framework, the paper focuses on the following four countries that have majority Muslim population and are representatives of each of the quadrants of figure-2: (I) Nigeria; (II) Pakistan; and (IV) Turkey and Bangladesh. Note that quadrant (III) has no such country. It is, however, to be acknowledged that given larger space and time one would have done comparative studies on a larger number of cases accounting for all other categories that are left out here. 

Analyses of the data here are presented in two steps: First, cross sectional. OLS to show how electoral stability (dependent variable) is explained by identities (independent variables) for the initial democratisation period (context one) controlling for regime types, electoral systems, and economic indicators (natural log of population, natural log of GDP) and instrumenting with region. The models with expected signs are as follows: 

Electoral volatility = a – b(Religion) –b(Ethnic fraction)

+ b(Regime type) + b(Electoral system)

+ b(Asia) + b(America) + b(Africa)

+ b(Log of GDP per capita)

+ b(GDP growth rate)

+ b(Log of population) + e  

Table 2 demonstrates the results of the model in equation 3. As expected the coefficient for ethnicity is 3.29123, which is statistically significant at 95% confidence level. This means that with the increase in ethnic fractionalisation electoral volatility decreases, controlling for all other variables. On the other hand, the other identity variable, religion, receives the expected sign, but turns out to be statistically non significant. In addition to ethnic identity, only two of the economic measures come out statistically significant: log of per capita GDP and GDP growth rate. All these indicate that during the initial phase (first couple of elections), core ethnic identities fare much better than religious identities in stabilising electoral democracy, all other things being equal. This, then, confirms the proposition-1. 

In the second step, two tasks are at hand: first, to calculate the electoral volatility measure using all available election data for each of the four Muslim majority countries (Figure-2) that according to Birnir were electorally stable during the first two elections.45 The calculation procedure is as follows. The number of parliamentary seats gained by each political party of a country for each election has been collected. Then, for two subsequent elections at a time is calculation of the absolute difference of each party seats, gain or loss. This absolute value is then divided by 2 to get the average seats, gain/loss for that party. To get a measure for the whole country, summed up are such average seats, gain/loss values of all the parties that participated in all the elections. This number has a lower bound, 0; but it does not have any upper bound. Thus it has to be evaluated in relation with all other values of the country. For example, take the case of Bangladesh. For three pairs of elections (1991-1996; 1996-2001; 2001- 2008) this country received three difference scores: 54, 102 and 178. This means, as the country moved from its first election pair to third election pair it became gradually more volatile. The higher the value the higher is the electoral volatility. Similarly, Pakistan’s volatility scores are 65à42à71à60 for five election pairs. This means Pakistan remained relatively less volatile in the subsequent elections vis-à-vis its first election pair. While Nigeria became a little more volatile (32à37), Turkey, compared to its first pair of elections gradually became more volatile over the course of six elections (five election pairs): 114.5à276à243à303à547. In sum, while Pakistan and Nigeria became less volatile, Bangladesh and Turkey became more volatile. In appendix, figures 1 and 2 demonstrate the facts graphically. 

The second task for this step is to compare World Value Survey 1995 with 2000 using a measure of religiosity for all of the above four countries. The survey question is: “is religion important in your life?” The theoretical expectation here is: the countries that have stabilised over time (or became less volatile) should have seen more people responding with ‘yes’ to the survey question in 2000 compared to 1995. That means, the expected sign of mean-difference is ‘plus’. On the other hand, the countries that have de-stabilised over time (or became more volatile) should have seen less people responding with ‘yes’ to the survey question in 2000 compared to 1995. That means, the expected sign of mean-difference is ‘minus’. Table 4 confirms the expectation for Bangladesh, Nigeria and Turkey: over time these countries became relatively more volatile, and therefore, people in these countries have not allowed religious identity to thrive 

This result makes sense. For example, in 2001 the Bangladesh Nationalist Party (right-centre) captured 64 per cent of the seats in the country’s national parliament. People have completely sidelined Jamaat-e-Islami (the Islamist Party). Similarly, in Turkey in 2002, the Justice and Development party (AKP) won 66 per cent of the seats in the parliament, while the Islamist Felicity Party was totally sidelined. In both of these countries, voters had clear Islamist alternative before them, but they did not choose that.46 However, the case of Pakistan did not confirm the expectation. Compared to the 1995 survey, why did less people believe in 2000 that religion was important in their lives? A more in-depth study of Pakistan is required to answer this question. However, looking at the country’s election years may help make an initial speculation: almost none of the government could complete their terms. From 1988 to 1993 they had elections in almost every two years. Due to frequent elections, voters probably did not feel to change their party as the parties in the government did not have enough change to perform. Due to these frequent elections, Pakistan probably received moderate scores in the election volatility table (Table 3). 

In conclusion, these initial results show that the theory and the hypothesis that the paper developed have potential to be developed further and be subjected to more rigorous statistical tests. It is expected that in diverse new democracies, ethnic identities would trump over religious identity, while as the system stabilises electorally religion, especially Islam, would gradually overcome other identities, particularly in countries with substantive number of Muslim population. The OLS regression results confirm that during the initial years ethnic identity does explain much of the electoral volatility of the new democracies. Subsequently, a series of descriptive studies on Bangladesh, Turkey and Nigeria confirmed, although weakly, that if democracies destabilise voters would not go for religious identity. Pakistan appeared to be a deviant case. The paper suggests that Pakistan did not behave as was expected because of its frequent elections that did not allow any government to complete its full term, and as a result, voters instead of increasing their support for religious identity, slightly decreased their support. 

46 Nasr, op. cit.