Measuring cultural diversity: ethnic, linguistic and religious fractionalization in the OECD

Abstract

Many claim that national economic success depends upon cultural homogeneity. We collect new time-series data and develop new measures of ethnic, linguistic and religious fractionalization for the OECD countries. We show that cultural diversity may vary by type across countries and over short periods of time. We also show that our measure of ethnic fractionalization is a significant predictor of economic performance in OECD countries despite the fact that they constitute a particularly homogeneous and economically advantaged group.

Keywords:

• Ethnic categories
• political economy
• inter-ethnic social distances
• cross-national analyses
• national identity
• data collection

Acknowledgements

Thanks for comments on an earlier version of this paper go to Michael Allen, Matthew Lange, Kathryn Lively, Kathy Sherrieb, Mike Smith and two reviewers.

Notes

1. Issues of cultural homogeneity are also of concern to students of nationalist insurgency (Cederman and Girardin 2007), civil war (Fearon, Kasara and Laitin 2007; Wimmer, Cederman and Min 2009), the provision of public goods (Habyarimana et al. 2009) and more.

2. Japan presented a special case insofar as some people reported that they were both Buddhist and Shintoist, which meant that the percentages totalled more than 100 per cent. Because calculating a fractionalization index requires that percentages total 100 per cent we adjusted the percentages accordingly by re-scaling them. That is, we kept the same proportions among groups but reduced the percentages to add up to 100 per cent. This slightly reduced religious diversity in Japan.

3. The standard deviation for each measure is presented in parentheses.

4. All correlations reported are Pearson correlations whose significance is tested with two-tailed tests.

5. As noted above, the linguistic data are the weakest. And the data on religious fractionalization are probably inflated by the inconsistent measurement across countries of non-religious and atheist groups.

6. Growth rates are missing in 1989 and 1990 for Poland and the Czech Republic. Therefore, their averages are for the years 1991–9 only.

7. Ireland's standardized regression residual is greater than 3.

8. Slovakia's standardized regression residual is greater than 2.5.

9. GDP per capita is in constant 2000 US dollars as reported in the World Bank World Development Indicators (2009) for the beginning of each period, 1985 and 2000.

10. This is the log of (1 + average years of education) at the beginning of each period, 1985 and 2000. Educational data come from Barro and Lee (2001).