What Do First Names Signal? Evidence on Race, Class, and Unobserved Characteristics from a New Dataset

Racially distinctive names are frequently used in field research on discrimination to signal the race of the name’s bearer. Although race is correlated with socioeconomic status (SES), most studies choose names based on racial distinctiveness alone, raising concerns about confounding: distinctively Black names may signal lower socioeconomic status as well as race, making it difficult to isolate the effect of racial signaling alone. While this confound has long been recognized, implementing a solution has been challenging because birth certificate data – the primary source for racially distinctive names – lacks direct SES measures. Using California birth certificate data from 1982 to 2017, we extend Fryer and Levitt (2004) (FL)’s Black Name Index (BNI) and create a Socioeconomic Name Index (SNI) that summarizes the SES signaled by each name, providing the first comprehensive measure of the socioeconomic status signal conveyed by a name. We find complete SES polarization among common racially distinctive names: the SNI of all common names with a strong white signal (BNI of 0.10 or below) exceeds the SNI of all common names with a strong Black signal (BNI of 0.90 or above). This means that researchers cannot simultaneously achieve high racial distinctiveness, SNI balance, broad SNI coverage, and reliable signal value through common names. We document some tradeoffs that researchers face in addressing this problem: relaxing distinctiveness thresholds improves SES overlap but risks using names whose racial signal is unreliable. The optimal tradeoff will depend on the goals of the research. We provide the Names Dataset, with SNI and BNI values for common first names, for use in choosing names for research. Even when names are balanced on SNI and BNI, they may still convey information about hard-to-observe traits such as personality. Using the same data, we find that while SNI and BNI significantly predict adult economic outcomes in an observational regression, both effects disappear when a novel sibling fixed effect is added. This suggests that names proxy for unobservable family traits beyond race and SES, implying that names may not be perfect proxies for either. Researchers should consider supplementing name selection with pretesting or existing datasets capturing perceived race, SES, and personality signals.