Research Article: Cause of Death Affects Racial Classification on Death Certificates

Date Published: January 26, 2011

Publisher: Public Library of Science

Author(s): Andrew Noymer, Andrew M. Penner, Aliya Saperstein, Cécile Viboud.

Abstract: Recent research suggests racial classification is responsive to social stereotypes, but how this affects racial classification in national vital statistics is unknown. This study examines whether cause of death influences racial classification on death certificates. We analyze the racial classifications from a nationally representative sample of death certificates and subsequent interviews with the decedents’ next of kin and find notable discrepancies between the two racial classifications by cause of death. Cirrhosis decedents are more likely to be recorded as American Indian on their death certificates, and homicide victims are more likely to be recorded as Black; these results remain net of controls for followback survey racial classification, indicating that the relationship we reveal is not simply a restatement of the fact that these causes of death are more prevalent among certain groups. Our findings suggest that seemingly non-racial characteristics, such as cause of death, affect how people are racially perceived by others and thus shape U.S. official statistics.

Partial Text: The accuracy of official data on birth rates and death rates are often taken for granted. However, recent research has drawn attention to inconsistencies in the recording of race across data sources and the resulting variability in estimates of race-specific death rates in the United States [1], [2]. These analyses have sparked debate among researchers over which measure of race should be considered correct [3]. Rather than focus on identifying errors or inaccuracies in the data, we extend previous research by exploring how the discrepancies in race reporting arise and whether they provide insight into why racial disparities in vital statistics persist. In particular, we use a nationally representative sample of death certificates and matched data from a subsequent survey of the decedent’s next of kin to examine whether cause of death and other non-racial characteristics of decedents are related to their racial classification.

Echoing previous findings of inconsistencies in how an individual is racially classified across different data sources, we find that for 1.1 percent of death certificates in the 1993 National Mortality Followback Survey, the next of kin racially classify the decedent differently than was recorded in official statistics. Levels of inconsistency range from 1 percent among decedents classified as white on their death certificate to 8.8 percent among decedents classified as American Indians.

While previous research has demonstrated inconsistencies in racial vital statistics, the processes creating these discrepancies are not well understood. We explored whether seemingly non-racial characteristics of individuals, such as their cause of death, affect how they are perceived racially by others. Our results demonstrate that otherwise similar Americans whose underlying cause of death was chronic liver disease or cirrhosis were more likely to be classified as American Indian on their death certificate than Americans who died of other causes – even if they were not classified as American Indian by their next of kin in a subsequent survey. A similar pattern exists between dying of homicide and the likelihood of being classified as Black. These findings suggest that the racial information recorded in vital statistics may be affected by the same kinds of social processes that shape racial classification more broadly. Research shows that changes in how people are racially classified over their lifetime are related to changes in social status that conform to widely held racial stereotypes [6]. Just as Americans are less likely to be seen as white by a survey interviewer after they have been incarcerated, unemployed or fallen into poverty, we conclude that stereotypes about who is likely to die a particular kind of death may color our official vital statistics.

To explore the role of cause of death in the racial classification of decedents, we examine the most recent public-release data from the National Mortality Followback Survey (NMFS), conducted in 1993. The NMFS matches data from 22,905 death certificates of U.S. residents drawn from the 1993 Current Mortality Sample to information about the decedent gathered from a followback survey using a proxy respondent. The sample is representative of deaths among the 1993 U.S. resident population, age 15 and older, excluding South Dakota – though, importantly for our purposes, homicides were oversampled to provide more statistical power [8]. We drop the 111 cases that are missing information on death certificate race. In 89 percent of the remaining cases, the proxy respondent is the decedent’s next-of-kin; we introduce controls for the nature of the relationship into the models, but for the sake of simplicity we refer to the proxy respondent as next-of-kin in discussing the results.