This is the kind of subject that deserves the most careful consideration, but usually doesn't get it.
David Laitin (Stanford University), Daniel Posner (UCLA)
[appeared in APSA-CP: The Comparative Politics Newsletter 12 (Winter 2001)]
In recent years, ethnic fractionalization has emerged as a central variable in quantitative analyses of outcomes ranging from economic growth rates (Easterly and Levine, 1997) and the quality of governance (La Porta et al, 1999) to ethnic conflict (Kay et al, 2000) and the frequency of coups d’etat (Londregan and Poole, 1990). Almost all such analyses employ, either alone or in combination with other measures, the same measure of ethnic fractionalization. This index, called ELF (for Ethno-Linguistic Fractionalization), is available for 129 countries – indeed, its broad coverage is the principal reason for its widespread adoption – and reflects the likelihood that two people chosen at random will be from different ethnic groups. It is calculated using the Herfindahl concentration formula from data compiled in a global survey of ethnic groups published in the Atlas Narodov Mira (1964) and subsequently included in Taylor and Hudson (1972).
Users of the ELF index have analyzed their results, to their peril, without any regard to the constructivist findings in the literature on ethnicity. Constructivist findings would make the standard ELF index suspect for four different reasons.
First, the users of the ELF index assume that a country’s degree of ethnic fractionalization is fixed, analogous to its topography or its distance from the equator. To the extent that a country’s boundaries do not change, it is assumed, its ELF score should remain constant. Constructivist theories of ethnicity, however, would compel us to challenge this assumption. They would lead us to expect changes in the level of ethnic fractionalization over time, as people over generations assimilate, differentiate, amalgamate, break-apart, immigrate and emigrate.
Take the case of Somalia.
It used to be said that Somalia was a lucky country because the colonialists had happened to draw the boundaries right and thus there was only one tribe in Somalia, divided up merely into clans. But then in the late 1980s the country fell apart, and suddenly we started hearing about the deep divisions between clans and sub-clans.
At independence, Isaaqs (from former British Somaliland) and Hawiyes (from former Italian Somalia) insisted they spoke the same language, and any survey of linguistic diversity undertaken at the time would have reflected this. In recent years, however, Isaaqs have begun consciously differentiating their speech forms from those of the Hawiyes as part of an attempt to justify recognition for their secessionist republic – much as Croat and Serb intellectuals and linguists have done over the past fifteen years in the Balkans (Greenberg, 2000). A linguistic survey conducted today would thus produce a quite different accounting of linguistic divisions in both former Yugoslavia and former Somalia.
Clan distinctions in Somalia have undergone a similar metamorphosis. With the decline of the dictatorship of Mohammed Siyaad Barre in the late 1980s, what had previously been considered one of the most ethnically homogeneous countries in Africa became severely divided by inter-clan fractionalization, with a concomitant change in the level of aggregation that is considered appropriate by political analysts. Studies of Somalia in the 1960s that focused on clan-based divisions tended to concentrate their analysis at the highest level of division (the clan family), of which there are three.
But amid the fractionalization caused by the civil war that broke the country apart a decade ago, more recent analyses have tended to emphasize distinctions among clans and even sub-clans. Thus, due to the civil war, a survey of ethnic fractionalization today would yield a substantially larger number of clans (and a correspondingly higher fractionalization index value) than one undertaken forty years ago. Contrary to the assumptions of most users of the ELF index, levels of ethnic fractionalization in Somalia have been dynamic over time, not stable givens of the landscape. Constructivist findings would thus seem to demand that fractionalization scores be provided over a time series to accommodate such changes.
For example, is there a single Republic operative within the Beltway who ever thinks about how the current racial/ethnic categories used by the federal government in dispensing goodies drives political behavior and how those categories could be altered for the long-run benefit of the GOP?