Conceptualization and Measurement Strategies
Both ethnicity and state capacity are much-debated concepts that have generated a large and often controversial literature on how to conceptualize and measure them. In this review chapter we flag major differences and similarities in the conceptualization and measurement of these two concepts, which allows us to situate political economy works on the ethnic diversity and public goods provision hypothesis within the broader literature and put different bodies of scholarship into dialogue.
At the most basic level, ethnicity refers to relations of membership structured around a belief in a shared culture and common ancestry. These relations are considered significant by members and outsiders, and are constituted through processes of self-identification and external ascription (Barth 1969; Jenkins 1997). Yet there exist a variety of different approaches on how to disaggregate this conceptualization of ethnicity and make it operational for empirical research (Lieberman and Singh 2012; Wimmer 2012a).
The majority of studies that examine the ethnic diversity and public goods provision hypothesis approach ethnicity in demographic terms. This research tradition is primarily concerned with the relative size and number of distinct ethnic groups within a unit of analysis (a country, or a sub-national unit, whether metropolitan areas or municipalities). Accordingly, its measurement strategies seek to identify the extent of ethnic fragmentation or heterogeneity. In so doing, studies usually rely on an ethnic fragmentation index, which measures the probability that two randomly selected individuals in a given country, city, or county belong to different ethnic groups. Specifically, cross-national analyses in this line of work tend to focus on ethnolinguistic fragmentation (ELF) and employ various data sets, most importantly the Atlas Narodov Mira (Atlas of the Peoples of the World), an ethnolinguistic fractionalization index first developed by researchers in the Soviet Union during the 1960s, which increases its value with the number of different language groups and the relatively equal size of these groups (Easterly and Levine 1997, pp. 1218-23; La Porta et al. 1999, p. 238). Other studies construct an ethnic fractionalization index based on self-identified racial identity from census data (Alesina et al. 1999, pp. 1254-5) or surveys (Luttmer 2001, pp. 502-5). Sub-national studies of ethnic diversity in developing countries equally replicate the ELF-based approach and build ethnic fractionalization indices based on their own survey data (Miguel and Gugerty 2005, pp. 2341, 2364) or on existing census data (Banerjee et al. 2005, pp. 641-3).
Yet, this is not the only way to operationalize ethnicity. A comprehensive look at the literature reveals that there are at least two other major approaches to conceptualization and measurement. One of them combines ethnic demography with a focus on relative group power. This perspective compares ethnic groups in terms of their control over economic and socio-political resources (Baldwin and Huber 2010; Cederman et al. 2010; Cederman and Girardin 2007; Chandra and Wilkinson 2008; Lieberman 2003; 0stby 2008) and differentiates between the extent of ethnic exclusion that prevails in a given social arrangement. Another body of scholarship focuses on ethnic mobilization, which broadly refers to collective action that draws on a sense of shared origins and identification with a joint way of life as the basis for political claims-making (McAdam et al. 2001; Olzak 1983; Yashar 2005). Ethnic mobilization manifests itself in a variety of institutional and organizational vehicles, including political parties, social movements, civic associations, and the like. Meaningful variation at the aggregate level can be traced to different intensities in the collective action of ethnic groups, distinct strategies (e.g. violent and non-violent tactics), but also differences in the nature of demands and geographical location (e.g. rural vs. urban) of ethnic mobilization.
Taken together, there are at least three contrasting approaches to move from root concept to a more specified conceptualization of ethnicity. As we will show below, each of these approaches tends to lead to distinct causal mechanism hypotheses about how ethnicity affects state capacity.
-  See, for example, Chandra and Wilkinson (2008) and Lieberman and Singh (2012) fordistinct approaches to the conceptualization and measurement of ethnicity. Hendrix (2010),Soifer (2012), and Saylor (2013) are recent methodological discussions on the measurement ofstate capacity.
-  This conceptualization avoids compartmentalization by treating 'race' and nationhood assubtypes of ethnicity, and not as analytically distinct phenomena (Wimmer 2008, 2012). Thefusion of beliefs in a common culture and shared ancestry with claims for national sovereigntyand the control of a state is characteristic of nations, while the association of phenotype withcommon descent delimitates racial groups and categories.
-  Here we follow Adcock and Collier's (2001) distinction between root concept and fullyspecified conceptualization.