Phenotypical, Linguistic or Religious? On the Concept and Measurement of Ethnic Fragmentation

Malaysian Journal of Economic Studies, Jun-Dec 2003 by Yeoh, Kok Kheng

Rae and Taylor (1970: 96) had shown that the XC equation (see footnote 16) can be rewritten as

XC = 2FC - F1 - F2

where F1 is the fragmentation of cleavage X1, F2 is that of cleavage X2, and FC is the probability that any two individuals are in different groups in at least one of the cleavages The derivation of F1 or F2 (following the computational procedure of Rae and Taylor's index of fragmentation described earlier) is relatively straightforward. However, to calculate FC, more detailed data will be required, for instance the proportion of members in each linguistic segment who belong to each religious category and vice versa.

This equation is important in showing how cross-cutting is in fact closely related to the fragmentation of the relevant cleavages. Since FC can take on several values for given fixed values of F1 and F2, the latter do not completely determine XC. However, F1 and F2 do constrain the possible values of XC. Rae and Taylor (1970) showed that if F1 and F2 are both very low (i.e. the cleavages are not very fragmented), FC must also be low. Conversely, FC is high if both F1 and F2 are very high. Hence, from the relation XC = 2FC - F1 - F2, the measure XC must be low in both of these cases. This means that there cannot be much crosscutting whenever F1 and F2 are both very low (near 0) or both very high (near 1 ). High values of XC can only occur when one of the cleavages has low fragmentation and the other has high fragmentation (ibid: 99-103). It can be seen in Table 5 that ethnolinguistic-ethnoreligious cross-cutting is prevalent in Laos, Cambodia, Thailand, the Philippines, Indonesia and probably Burma while such ethnic cleavages are obviously mutually reinforcing in countries like Malaysia, Brunei, Fiji and probably Singapore. The predominant Roman Catholic/animist East Timor, which is not included in the list, is another country exhibiting cross-cutting characteristics, with its high degree of linguistic fragmentation (0.667) assuaged by the lack of sharp religious differentiations.

The above shows that, Malaysia, which is characterised by its reinforcing racial, linguistic and religious cleavages, should be considered more fragmented in terms of overall ethnic structure, than another country that happens to have similar degrees of racial, linguistic and religious differentiations but where such cleavages are cross-cutting. The same can be said of Fiji. Therefore, to reveal the true picture of ethnic fragmentation, the levels of EFIs should ideally take into consideration the effects of cross-cutting.

5. Concluding Remarks

This paper proposes an index of ethnic fractionalisation that comprises all three major types of non-class cleavages in society - racial (phenotypical), linguistic and religious. Whereas the existing studies on public policy and ethnicity either included only one of these components (Mueller and Murrell's work (1986) which employed linguistic groups as the units of measurement) or considered them as separate variables (McCarty's 'ethnic variance' and 'religious variance')17, this paper regards these components as different manifestations of one single characteristic. In other words, racial (phenotypical), linguistic and religious characteristics represent different markers of ethnic (or socioracial) distinction (often more loosely termed ethnic markers). To treat them as separate variables or to employ one to the exclusion of the others inevitably leads to the mismeasurement of the degree of fragmentation. There are two ways to encompass all these three ethnic markers. The first option is to construct a composite index based on three separate indices measuring racial, linguistic and religious diversities respectively. Although technically simple, this option is not adopted in this paper due to the high risk of mismeasurement, as there is no way to accurately gauge the relative weight of the three separate types of fragmentation, especially in the light of the possible cross-cutting or reinforcing link between them. On the contrary, the approach followed here is to employ solely the most significant ethnic marker of a country as the unit of measurement, for instance, race (phenotype) in Rwanda, language in India and religion in Bosnia-Hercegovina. Such an approach can of course be said to be as arbitrary as the first option as it disregards the other 'less significant' ethnic cleavages. However, on close scrutiny it emerges as the most accurate way to measure ethnic diversity since in reality it is the most prominent cleavage that counts in the polarisation of society, though it is in itself often a symbol for social mobilisation finding its root in some politico-economic differentiation. It also has the advantage of not having to rely on arbitrary weighting of different indices as required by the first option and avoiding excessive assumptions (Occam's razor). Therefore, while the index of ethnic fractionalisation (EFI) proposed in this paper represents the degree of fragmentation in terms of one of the following cleavages: racial, linguistic and religious (with the possibility of some conceptual overlapping among them), exactly which type of cleavage is selected depends on the particular context of the country concerned. For instance, ethno-religious cleavages provide a more accurate picture of the situation in Bosnia-Hercegovina - so do racial differences in Rwanda and Burundi - than linguistic ones, since linguistic homogeneity of these countries is far from reflecting the true degree of their ethnic fragmentation. Ideally, the effect of cross-cutting and reinforcing influences between the different markers should also be taken into consideration but again it is practically impossible to accurately measure such complex links (quantitative measurements of the degree of crosscutting or reinforcing such as Rae and Taylor's XC index would require detailed field survey in each country (which is beyond the scope of this paper) to determine the proportion of the members of a type of ethnic group who also belong to some other types of ethnic groups). Instead of arbitrarily assigning values for such influences, it serves to provide a more accurate measurement of the overall ethnic diversity and demographic heterogeneity, for practical purposes, by not taking them into consideration. While the existence of such influences cannot be denied, a comparison of individual countries' social histories easily reveals that such influences are not as significant as to alter the relative degree of fragmentation between countries.