Since the 1990s, state practices of counting and classifying populations by ethnicity have come under increased scrutiny within the social sciences (Arel 2002 ; Kertzer and Arel 2002 ; Nobles 2000 ; Perlmann and Waters 2002 ; Petersen 1997 ; Statistics Canada and U.S. Census Bureau 1993 ). A number of excellent case studies have provided critical insights into how and why ethnic enumeration is pursued in particular times and places. ¹ However, with some notable exceptions (Morning 2008 ; Rallu et al. 2006 ), little attention has been given to theorizing or empirically testing a global model of ethnic classifi cation and counting. Consequently, there is a limited understanding about the general conditions that impede or encourage state recognition of ethnicity in the national census and the forms that such recognition takes.