In this paper, we examine the ways in which the notion of 'ethnicity' has been operationalized for quantitative research. We argue that there is a conceptual mismatch between modern theories of ethnicity and the current way in which ethnicity is employed statistically, but that this mismatch can be overcome if we reconceptualise econometrics as using 'ethnicity' as an indicator of a broader notion of 'social distance'. We conceptually subdivide 'social distance' into measures of 'diversity' and 'disparity', and review the strengths and weaknesses of existing measures employed in the literature. We then explore how far different measures of ethnicity are correlated both across stylized distributions and using real world data from districts in Ghana, Uganda, and Indonesia. We find that even measures that purport to pick up very different underlying distributions are, in fact, very highly correlated, and suggest that this should caution us to be more careful in the interpretation of econometric results. We conclude that quantitative research using theoretically more sophisticated measures of ethnicity that link particular distributions with particular political outcomes, combined with qualitative research that allows a closely examination of causal processes, may be the appropriate way forward for empirical research on ethnicity.