VoxEU Column Economic history Frontiers of economic research

Is race a choice?

Race is usually treated as a fixed, exogenous characteristic in academic studies and policy discussions, but a growing body of evidence calls this assumption into question. This column presents evidence from historical US census data that more than 19% of black males ‘passed’ as white, around 10% of whom later ‘reverse-passed’ to being black. Passing was associated with geographic relocation and with better political-economic and social opportunities for whites relative to blacks, providing prima facie evidence that passing was endogenous.

The relationship between a person’s race and/or ethnicity and her economic, political, and social behaviour has been a focal point for numerous policy discussions. This is because the composition of race and ethnicity are often found to be associated with outcomes such as conflict, earnings, educational attainment, and voting. This is true in many different contexts across the world, including rich countries such as the US, middle-income countries such as Brazil and India, and very poor countries in Sub-Saharan Africa. However, the interpretation that race has a causal impact on the aforementioned outcomes critically depends on whether one believes that race and ethnicity are exogenous and fixed characteristics – i.e. outside the control of an individual and constant over her lifetime.

The literature typically assumes that racial and ethnic identities are fixed and exogenous. For examples, see the reviews of the political economy literature by Alesina and Ferrara (2004) and of studies of the US black-white wage gap by Lang and Lehmann (2012).

Racial identity as a choice

A growing body of evidence from anecdotes, historians, and recent studies suggests that this is not always true. In fact, there may be reason to believe that race is, to some extent, a choice made by an individual because of social-economic and political factors. For example, historians have long noted that Americans with African ancestry often chose to ‘pass’ for white to obtain better economic and political opportunities (e.g. O’Toole 2003 and Sharfstein 2011). Mill and Stein (2012) find that amongst mixed-race siblings, those that identify as white later in life earn significantly higher wages. Recent studies have also documented that individuals have responded to political-economic incentives and changed castes in India (Cassan 2013), manipulated their racial appearance for higher wages in Brazil (Cornwell et al. 2014), or have strategically chosen the official ethnicity of mixed children in China (Jia and Persson 2013). More generally, studies of identity, such as the theoretical work of Akerlof and Kranton (2000), argue that “choice of identity may be the most important ‘economic’ decision people make. Individuals may – more or less consciously – choose who they want to be ... Previous economic analyses of, for example, poverty, labor supply, and schooling have not considered these possibilities”.

To assess the necessity of an overhaul in the way we conceptualise and evaluate the relationship between race and economic and political variables, we need to first know the quantitative importance of race change and whether it is associated with the latter set of variables. Data limitations have made this difficult until now. As recently as 2011, historian Daniel Sharfstein notes that “According to just about anyone who has considered the question, the migration [from black to white in the US] is impossible to reconstruct ... At best, such evidence is scattered across local archives and county courthouses, in library stacks and microfilm reels. Beyond the isolated anecdotes, there seems to be only silence” (Sharfstein 2011: 4).

New evidence on ‘passing’

In Nix and Qian (2015), we address this question with recently available data from the US historical censuses (1880–1940). Specifically, we ask the following questions:

  1. How many Americans of African descent changed their racial identity from ‘black’ to ‘white’ during their lifetime?
  2. How did they achieve race change?
  3. Was the change permanent?
  4. Was it associated with the political-economic and social returns to being white?

These descriptive facts are necessary precursors for assessing the importance of accounting for the endogeneity of race in future research, and for deeper explorations of the linkages between racial identity and other variables.

Using the digitised full population of the US Census data, we link individuals from each census to the next and track whether an individual changes his race from ‘black’ to ‘white’ – i.e. passes for white – or the other way around – i.e. reverse-passes for black. Because women change their surnames upon marriage, we only examine men. The key innovation of our procedure is to randomly select amongst multiple potential matches with the same race such that we achieve higher match rates without affecting the pass rate. Consider the example where person A, who is black in 1900, matches to 100 individuals in 1910, who are all black. In previous studies, person A will typically be dropped from the sample. Instead, we randomly choose one of the 100 individuals to be a match and say that person A did not pass. We also construct upper and lower bounds of the pass rate. We describe the procedure in detail in Section 4 of Nix and Qian (2015).

In the context of our study, there were strong social, economic and political incentives for blacks to pass for white. Extensive racial mixing meant that a large proportion of the ‘black’ population could physically pass for ‘white’. To illustrate the wide gradient of colour for former slaves and the fact that many of those classified as black during 1880–1930 could physically pass for white, see the photographs by M H Kimball in Figure 1. The emancipated slaves shown in this photograph were chosen for a propaganda tour of the North in 1863.1 Passing for white was illegal and required intentional effort by the individual such as relocating to a white community where his past was unknown. There are many biographical accounts of individuals passing for white or reverse-passing back to black.

Figure 1. Emancipated slaves in 1863

We find that more than 19% of black males passed for white during the four census intervals that we examine. These estimates are broadly consistent with the recent genetic evidence on US racial composition from 23andme, which show that amongst the current white population, the number that would be classified as black under the US legal system of the early 1900s is approximately a quarter of the current US black population.

Consistent with historical cases of individuals who temporarily passed for better employment or schooling opportunities, we find that not all passing was permanent. Amongst those who passed for white, approximately one-tenth ‘reverse passed’ to being black in the following census year. Passers were much more likely to geographically relocate than non-passers, particularly to ‘whiter’ locations. In contrast, reverse-passers were more likely to have moved to ‘blacker’ locations than passers who maintained their white identity. These patterns of moving are consistent with the historical evidence that racial identification was typically based on association (i.e. living with and behaving as a white person).

We also document that passing is associated with the political-economic incentives to be white – individuals were more likely to pass for white if the income gap between whites and blacks was larger, or if there were relatively few opportunities for schooling or political enfranchisement for blacks. Although the historical data are limited, these associations provide prima facie evidence that passing is endogenous to social and political economic factors.

In Nix and Qian (2015), we show that race change is not confined to anecdotes or case studies, but is instead a widespread and quantitatively important phenomenon, and likely to be endogenous to variables that the economics literature often examine as outcomes of race.

Concluding remarks

What do these findings mean for policymakers? This partly depends on whether passers behave more like whites or blacks. For example, for political economy studies of the effect of regional racial diversity on voting outcomes, the interpretation would be unchanged if passers who identify as white vote like whites. However, if passers vote like blacks, then current estimates could be misleading. Similarly, for traditional wage regressions that examine the relationship of being black on earnings, the interpretation of the effect of being black depends on the earnings of the average passer relative to the average black worker who does not pass and the average white worker. Thus, another important avenue of future research is to examine the political and economic behaviour of passers relative to non-passers and whites.

The implications of endogenous race change will also depend on the research question that is being asked. For example, that many individuals can choose their race does not affect the interpretation of a the fact that black workers earn less than white workers today if one is interested in the effect of self-identity on earnings. In contrast, if one is interested in the impact of historical enslavement on wages today, then it will be important to somehow account for the fact that many former slaves have since passed for white and to know whether passers have earnings like blacks who did not pass or like whites. It is important to note that passers earning the same amount as whites would not undermine the interpretation that there is discrimination against those who identify as blacks, since such discrimination is likely to be the key motivation for an individual to pass in the first place.


Akerlof, G A and R E Kranton (2000), “Economics And Identity”, Quarterly Journal of Economics 115(3): 715–753.

Alesina, A and E L Ferrara (2004), “Ethnic Diversity and Economic Performance”, NBER Working Paper 10313.

Cassan, G (2013), “Identity Based Policies and Identity Manipulation: Evidence from Colonial Punjab”, CEPREMAP Working Paper (Docweb) 1306.

Cornwell, C, J M Rivera, and I M Schmutte (2014), “Wage Discrimination when Identity is Subjective: Evidence from Changes in Employer-Reported Race”, Working paper, University of Georgia.

Jia, R and T Persson (2013), “Ethnicity in Children and Mixed Marriages: Theory and Evidence from China”, UC San Diego working paper.

Lang, K and J-Y K Lehmann (2012), “Racial Discrimination in the Labor Market: Theory and Empirics”, Journal of Economic Literature 50(4): 959–1006.

Mill, R and L C Stein (2012), “Race, Skin Color, and Economic Outcomes in Early Twentieth-Century America”.

Nix, E and N Qian (2015), “The Fluidity of Race: ‘Passing’ in the United States, 1880–1940”, NBER Working Paper 20828.

O’Toole, J (2003), Passing for White: Race, Religion, and the Healy Family, 1820–1920, University of Massachusetts Press.

Sharfstein, D (2011), The Invisible Line: A Secret History of Race in America, Penguin Group US.


1 They were selected specifically to show the vast differences in skin colour of slaves. The former slaves are described in “White and Colored Slaves” by C C Leigh (Harper’s Weekly, 30 January 1864, p. 71).

3,255 Reads