Perceptions and attributions of race differences in health risks

Jonathan Baron, Bethany Neiderhiser, Oscar H. Gandy, Jr.[+]
University of Pennsylvania

Abstract

We asked African-American (black) and non-black students and staff of a major university to estimate the frequency of 18 causes of death, and the black/white ratio of these causes. Consistent with past research, subjects were somewhat insensitive to the magnitude of differences among causes of death. Blacks showed this effect more than nonblacks. However, blacks were more knowledgeable about those causes that were less frequent for blacks, such as suicide and motor-vehicle accidents. Causes frequently mentioned in the news appeared to be overestimated. This affect was smaller for older subjects. Otherwise, accuracy was unaffected by sex, race, or student vs. non-student status.

In this study and in a second study in which we presented the attributions and asked for agreement/disagreement, we asked subjects for explanations (attributions) of that affected blacks more or whites more. The most frequent attributions were in terms of correlations without further causal explanation, e.g., blacks are poor and poor people are more (less) affected by some cause. In both studies, blacks were more likely than nonblacks to attribute race differences to the behavior of the more affected group. In the second study, blacks were less likely to endorse genetic attributions.


Knowledge of health risks is important both for personal protective behavior and for understanding the social and environmental forces that affect human health. In the United States, African Americans (blacks) differ from others in their health (Rogers, 1992), just as they differ in other indicators of social well-being such as income. The nature of these differences is relevant to questions about the nature of these more general differences. For blacks, as for other groups, knowledge of health risks peculiar to their group is also personally important: advice about health given to the general public should be somewhat selectively absorbed as a function of risk factors, and group membership is one of these factors. In particular, black death rates from homicide, cancer, heart disease, and stroke (among other things) are higher than for the population in general, but rates are lower for suicide and motor-vehicle accidence (among other things).

The purpose of the present study is to assess blacks and others' knowledge of risks in general and of racial differences in risk. We are also interested in how blacks and other explain the race differences. Finally, we ask whether exposure to news media affects our measures of knowledge or attribution (explanation) (see Gandy et al., 1997, for discussion of presentation of race differences in risk in the news).

Lichtenstein et al. (1980) studied perceptions of judged frequency of causes of death. They found that judgments were fairly accurate, with two general exceptions. First, low-frequency causes were overestimated and high-frequency causes were underestimated. This pattern could result from regression to the mean, so it is of little interest. Second, causes that were frequently mentioned in newspapers were overestimated, relative to other causes of the same frequency. Our study extends this study by asking whether perception of racial differences in risk is also influenced by media coverage of racial difference and by the extent to which the subject is exposed to this coverage.

Flynn et al. (1994), in the same tradition, examined race differences in perception of risks -- identified by their causes rather than their effects -- such as cigarette smoking and nuclear power plants. They found that white males considered most risks to be less serious than did white females, black males, or black females. The latter three groups did not differ. Our study cannot replicate this one, because we did not ask about perceived seriousness, and we gave subjects a standard on which they could anchor their risk estimates.

Perceptions of race differences were assessed, for a different purpose, by McCauley and Stitt (1978), who found that white subjects were generally accurate in their perception of the black/white ratio in a number of social indices, such as percent of each group on welfare or percent of illegitimate births. We extend this study to potentially less controversial questions such as the relative risk of death from heart disease.

Ryan (1995, 1996) extended the McCauley and Stitt study by including black subjects -- college students -- as well as whites. She asked about positive and negative stereotypical traits of blacks and (by comparison) whites, such as ``athletic,'' ``sexually aggressive,'' ``academically intelligent,'' and ``self-centered.'' She assessed accuracy by comparing subjects' estimates of percentages to the subjects' own self-ratings on the same traits. She found that both blacks and whites exaggerated the magnitude of group differences and that blacks did this more than whites.

Study 1

We asked subjects to indicate the death rates for whites, and the black/white ratio, for 18 causes of death. These included the most frequent causes of death, plus others chosen because blacks had lower rates than whites (e.g., suicide) or because they are frequently discussed in the news (e.g., drugs and alcohol).

We also asked our subjects for their explanations (attributions) of race differences in risk. We ask this both for causes of death that they considered more likely for blacks than for whites and for causes that they considered more likely for whites. We have no ``right answer'' for these responses. Even the most recent studies of these issues find it difficult to unconfound the possible causes. For example, Krieger and Sidney (1996) found that reported experience with racial discrimination correlated with high blood-pressure in blacks, but it is possible that those more prone to anger are also more prone to report such experiences, regardless of the facts, and anger may cause high blood-pressure.

Method

Subjects were solicited at the University of Pennsylvania by posting signs asking for people interested in completing questionnaires for pay ($6/hour for this questionnaire and others), and by distributing questionnaires to staff of the University, to be returned by campus mail. Staff were not paid individually, but, if they sent in a cover sheet with their name and address -- to be separated from the questionnaire -- $1.50 was put in a kitty, with $75 given to one person out of each 50 who returned the sheet (plus the remainder to one more person).

The 158 subjects consisted of 29 African Americans, henceforth called black, 122 non-Hispanics, henceforth considered nonblack; Hispanic subjects were excluded for analysis of race differences (but Asian Americans were included as nonblack). Otherwise, 52% of the sample were female and 54% were staff (vs. 46% students) of the University of Pennsylvania; 34% of the staff members were employees of the Hospital of the University of Pennsylvania (HUP). Age ranged from 18 to 59 (mean 26, median 21) and did not differ between students and staff.

The questionnaire used the causes of death shown in Tables 1 and 2. Table 1 shows the statistics on the white death rate from Gardner and Hudson (1996, referred to as CDC henceforth), along with the answers of nonblack and black subjects (as described later). Table 2 shows the CDC statistics on the black/white ratio, and, again, the answers of nonblack and black subjects. In the questionnaire, only the causes were provided (except for the overall total of 485), and subjects filled in their own estimates of the white rate and black/white ratio.

Results

Blacks were more knowledgeable about the causes of death on which blacks were lower. They were otherwise less sensitive to differences among causes. Other variables, including exposure to the news, had no effects on accuracy. Blacks were more likely to attribute differences in risk to the behavior of the group with the higher risk.

Accuracy

For the causes of death on which the ratio was low the black-white difference was statistically significant (t=4.24 p=.000). Blacks know more about extreme differences in risk between the races. In particular, blacks know that blacks are much less likely to kill themselves. The only other race difference was in the slope of the function relating the logarithm of the estimate to the logarithm of the true value. (Lichtenstein et al. fit a curvilinear function to these values, but we could not reject a linear function with only 18 risks.) The mean slopes were .27 for blacks and .44 for nonblacks (t=2.68, p=.008); both groups were undersensitive to differences (as found by Lichtenstein et al., 1978), but blacks were less sensitive. This difference in slope did not account for the difference in knowledge of the low risks.

To examine the effect of news coverage, we first looked to see which of our causes were covered most extensively in the news. We searched Lexis/Nexis and the Vanderbilt University Television News Archive (http://tvnews.vanderbilt.edu) for stories about each of the causes in our list, using alternate terms when appropriate. The causes most frequently mentioned (in association with ``death'') in both databases were cancer, HIV infection, homicide, and drugs and alcohol. We defined the error rate for each cause as the departure (residual) from the best fitting line relating the log of the subject's response to the log of the CDC rate. Mean error rates were significantly positive (overestimates) for cancer (t=3.22, p=.002), HIV (t=6.12, p=.000), and drugs and alcohol (t=4.29, p=.000); the mean error rate for homicide was not different from zero. In general, these results support the conclusion of Lichtenstein et al. (1978), although we do not have enough events to test across events.

The mean of sum of these error rates across the four causes was positive, of course (t=5.29, p=.000). This measure was unaffected by black, sex, staff membership, or exposure to television or newspaper news. It was, however, negatively correlated with age (r=-.24, p=.003; this relationship emerged in several regression analyses as well).

Attributions

We asked each subject to explain the reasons for the two causes of death with the highest black/white ratio and the two with the lowest ratio, according to their judgments. Explanations were categorized as follows:

First, we classified attributions as correlational or causal. Typical correlational explanations were that blacks were poorer and poverty was correlated (positively or negatively) with the cause in question. The subject gave no explanation for the correlation. We did not analyzes correlational explanations further. Causal explanations were classified according to whether the cause was the other race (than that with the higher frequency of death from the cause at issue), the same race or the individuals involved, people in general (including a ``racist society''), or nature (e.g., genetic differences). Finally, we classified causal judgments as moral or nonmoral. ``Racism'' counted as moral, as did ``lack of self-control.'' These express moral judgments. We classified each subject as giving or not giving an explanation of each type.

Most of the attributions were correlational: 70% of the subjects gave such attributions for high risks (for blacks), and 73% for low risks. Attributions to the other race were 7% for high, 0% for low; attributions to people in general were 2.5% for high, 0.6% for low; moral attributions were 10.1% for high, 1.9% for low; and attributions to own behavior were 15.2% for high and 12.7% for low.

The overall effect of black on attributions was significant (p=.001 by an overall χ square test). Blacks were more likely to attribute high risks to the behavior of the group with the higher risk (34% vs. 11%, p=.003), more likely to attribute low risks (for blacks) to the behavior of the group with the higher risk (31% vs. 7%, p=.001). Blacks also tended to be less likely to attribute high risks to correlations (55% vs. 74%, p=.069, Fisher test), and less likely to attribute low risks to correlations (59% vs. 77%, p=.060). Otherwise, blacks did not differ from nonblacks. These effects remained when other variables (sex, age, news exposure, staff vs. student) were taken into account, and these other variables had no significant effects on attributions.

Study 2

In the second study, instead of asking the subject to explain the causes that he or she designated as showing the highest and lowest ratio, we asked for attributions about four causes of death, two on which blacks were higher, heart disease and breast cancer, and two on which blacks were lower, suicide and motor-vehicle accidents. We also presented causes, based on subjects answers in Study 1, that corresponded to the scoring categories we had used.

Method

The 121 subjects consisted of 20 African-Americans, called ``black,'' and 101 others, who were all treated as ``nonblack.'' All the blacks were U.S. citizens, and all but 8 of the nonblacks. The median ages were 20 for the nonblacks and 23 for the blacks, with 70% of the blacks and 87% of the nonblacks under 30. 90% of the blacks and 60% of the nonblacks were female. Subjects were solicited by direct email and by posting the questionnaire on local news groups, especially those that blacks would be likely to read. To compensate subjects for completing the survey, we placed $1 in a kitty for each response and gave $50 to one subject, picked at random, in each group of 50 responses (and $21 to a subject in the group left over).

The survey on ``Race and health'' presented four facts about race. Subjects were asked to classify each of several explanations of each fact as true, probably true, unsure or don't know, probably false, or false. Subjects were told, ```Whites' refers to Americans of European or East Asian descent and `Blacks' refers to African Americans. All facts are based on statistics collected by the U.S. Government. Each fact can have more than one cause.'' Table 3 shows the facts (1-4) and the explanations, in brief, with a code name for each. (An ``other'' category was also provided but few subjects used it).

Subjects were also asked their sex, race, date of birth, nationality, the number of times they read a newspaper and watched TV news in the last week, and how many times per week they usually do these things.

Results

Table 4 shows the mean responses to the explanations provided. In general, subjects attributed the difference in heart disease largely to blacks eating more fat, genetic differences, and blacks not going to doctors as much. They attributed the difference in suicide to a greater sense of community among blacks; the difference in accident rate to blacks not driving as much; and the difference in breast cancer to protective measures and affording medical care. In general, blacks were less inclined to favor genetic explanations and more inclined to favor explanations in terms of blacks' behavior or of prejudice.

News exposure had no effect on any of the items concerning explanations, and no significant effect overall in a multivariate regression. Older people had greater news exposure through both newspapers (r=.27) and television (r=.22, p<.02 for both), but black and sex were unrelated to news exposure.

To examine effects of race, sex, and age on the ratings of causes -- since black was confounded with age and sex -- we performed a multivariate regression of all the explanation ratings as a function of black, sex, and age. The overall regression was significant (p=.000), but there was only a single significant canonical correlate, indicating that the three predictors did not have distinct effects on different items. Overall, the effect of black was significant (p=.009). Individually, blacks were more likely to agree with HFAT (p=.020), SCOPE (p=.017), and ACARS (p=.002), and less likely to agree with BGENE (p=.001).

We also formed composite measures of the explanations according to our theoretical categories: genetics, blacks' behavior, society, and prejudice. We formed these separately for the two causes for which blacks were at higher risk (heart disease and breast cancer) and the two for which blacks were at lower risk (suicide and accidents). The eight categories were predictable from black (p=.006) in a multivariate regression with sex and age. In particular, blacks were less likely to endorse genetic explanations of high risks (composite of HGENE and BGENE; p=.007), and more likely to endorse societal explanations of low risks (composite of SPOOR, SFAIL, SBLAME, and ACARS; p=.012). Interestingly, as found in Study 1, blacks were more likely to endorse explanations of high risks in terms of black behavior (composite of HFAT, HEXC, HDOCS, BBSE, BBEH,and BLUMP; p=.083, and p=.016 when age and sex are left out of the regression).

The somewhat surprising result here is that, while blacks show no great tendency to blame everything bad on whites or everything good on blacks, they were strongly opposed to genetic explanations in general. (This result may have been missed in Study 1 because very few genetic explanations were offered spontaneously by anyone, so we did not have a good measure of most subjects' attitude toward them.) We can think of two explanations of this result. One is that blacks are more familiar with the behavioral tendencies among blacks that they tended to prefer, such as eating high-fat diets and generally not worrying about prevention. Thus, blacks may, in essence, see the truth. Another explanation is that blacks may be resistant to genetic explanations of anything because of the use of such explanations to account for the black-white gap in test scores and academic achievement. They may rather keep the door to genetics entirely closed than have to decide on a case-by-case basis whether it is relevant to statistical differences between blacks and whites. It would be interesting to see whether priming the subjects with a clear case that is also non-stigmatizing -- sickle-cell anemia -- would affect blacks' willingness to endorse genetic explanations of other health differences.

Conclusion

Our results are mostly negative, but negative results may be of some interest here. In particular, we did not find any evidence of self-serving attributions or beliefs involving wishful thinking. Blacks differed little from non-blacks in any measures of beliefs, although they were better informed about the risks that were lower for them and apparently less well informed (or less expressive of their knowledge) about overall differences among causes of death.

Although we found that causes with frequent news coverage tended to be overestimated, this effect did not differ as a function of self-reported exposure to news.

Consistent with the findings of McCauley and Stitt (1978), we found no tendency to overestimate race differences. Indeed, as shown in Table 2, race differences were seriously underestimated by both races in every case in which differences were present, with the sole exception of blacks' estimates of differences in suicide. In fact, we are led to suspect that this underestimation phenomenon may be a general result of regression to the mean. In this case, McCauley and Stitt's finding of apparent accuracy for apparently more controversial aspects of the African-American stereotype may reflect a combination of a regression phenomenon and truly exaggerated stereotypes, leading to approximate accuracy on the whole, although they also found considerable underestimation of black/white ratios.

References

Flynn, J., Slovic, P., & Mertz, C. K. (1994). Gender, race, and perception of environmental health risks. Risk Analysis, 14, 1101-1108.

Gandy, O. H., Kopp, K., Hands, T., Frazer, K., & Phillips, D.\ (1997). Race and risk: Factors affecting the framing of stories about inequality, discrimination, and just plain bad luck. Public Opinion Quarterly, 61, 158-182.

Gardner, P., & Hudson, B. L. (1996). Advance report of final mortality statistics, 1993. Monthly Vital Statistics Report, 44 (7), supplement. Centers for Disease Control and Prevention, National Center for Health Statistics.

Krieger, N., & Sidney, S. (1996). Racial discrimination and blood pressure: The CARDIA study of young black and white adults. American Journal of Public Health, 86, 1370-1378.

Lichtenstein, S., Slovic, P., Fischhoff, B., Layman, M., & Combs, B. (1978). Judged frequency of lethal events. Journal of Experimental Psychology: Human Learning and Memory, 4, 551-578.

McCauley, C., & Stitt, C. L. (1978). An individual and quantitative measure of stereotypes. Journal of Personality and Social Psychology, 36, 929-940.

Pincus, F. L. (1996). Discrimination comes in many forms: Individual, institutional, and structural. American Behavioral Scientist, 40, 186-194.

Rogers R. G. (1992). Living and dying in the U.S.A.: sociodemographic determinants of death among blacks and whites. Demography, 29, 287-303.

Ryan, C. S. (1995). Motivations and the perceiver's group membership: Consequences for stereotype accuracy. In Y.-T.\ Lee, L. J. Jussim, & C. R. McCauley (Eds.), Stereotype accuracy: Toward appreciating group differences, 189-214. Washington, DC: American Psychological Association.

Ryan, C. S. (1996). Accuracy of Black and White college students' in-group and out-group stereotypes. Personality and Social Psychology Bulletin 22, 1114-1127.

Table 1.
CDC statistics on causes of deaths for whites, and geometric mean corrected response of nonblack and black subjects.

Cause of death White rate Nonblack Black
0. All 485
1. Heart 173.9 65.4 37.5
2. Stroke 24.5 27.1 18.9
3. High BP 1.7 20.6 21.0
4. Cancer 129.4 46.5 40.8
5. Chronic lung 21.9 15.5 18.9
6. Car accidents 16.1 26.6 21.4
7. Other accidents 13.5 13.9 19.1
8. Pneumonia/flu 12.9 6.5 6.4
9. Diabetes 11.0 8.4 10.4
10. HIV infection 10.5 22.1 29.4
11. Suicide 12.0 9.3 11.6
12. Suic./firearms 7.3 5.7 9.0
13. Homicide 6.0 9.0 15.6
14. Hom./firearms 4.0 9.6 14.7
15. Chronic liver 7.9 8.5 9.3
16. Alzheimer's 2.4 6.4 9.5
17. Drugs/alcohol 10.6 16.6 25.0
18. Pregnancy, etc. 4.8 6.2 8.4

Table 2.
CDC figures and geometric mean responses of nonblack and black subjects to the question about the black/white ratio (with the CDC statistics in the first column). Numbers in bold are the largest ratios, numbers in italics are the smallest.

Cause of death B/W ratio Nonblack Black
1 Heart 1.49 1.10 1.35
2 Stroke 1.84 1.04 1.23
3 High BP 4.06 1.20 1.89
4 Cancer 1.37 1.04 0.95
5 Chronic lung .81 1.08 0.86
6 Car accidents 1.01 1.01 0.75
7 Other accidents 1.63 1.15 0.79
8 Pneumonia/flu 1.44 1.14 0.85
9 Diabetes 2.44 1.05 1.60
10 HIV infection 3.96 1.43 1.39
11 Suicide .60 0.88 0.52
12 Suic./firearms .60 0.97 0.59
13 Homicide 6.82 1.96 1.90
14 Hom./firearms 7.95 2.10 2.55
15 Chronic liver 1.43 1.17 0.96
16 Alzheimer's .67 0.90 0.73
17 Drugs/alcohol 1.96 1.60 1.59
18 Pregnancy, etc. 4.27 1.44 1.05

Table 3.
Health facts, with the code word used for each.

1. Blacks are more likely to get heart disease.
HFAT eat more high fat foods
HEXC get less exercise
HJOBS work at jobs that are more stressful to the heart
HPREJ suffer from stress caused by racial prejudice
HGENE genes
HDOCS do not go to doctors as much
HWARN doctors and other health professionals do not warn blacks
2. Whites are more likely than blacks to commit suicide.
SCOPE too busy coping with the effects of racial prejudice
SPOOR rich people commit suicide more often than poor people
SFAIL whites have to live up to higher expectations from others
SCOMM have a better sense of community
SBLAME Whites blame themselves, blacks blame others
3. Whites are more likely to die in automobile accidents.
ACOPS drive more carefully, more afraid of the police (police prejudice)
ACARS don't drive as much, live in cities, cannot afford it
APREJ don't drive as much, prejudice
ACOOR better drivers because they are more coordinated, genetically
4. Black women are more likely to die of breast cancer.
BGENE genetically more susceptible to breast cancer
BBSE less likely to take preventative steps (self-examination, requesting x-rays)
BAFFD less likely to be able to afford high-quality care
BDOCS doctors are less concerned
BBEH less likely to take steps to reduce risk
BLUMP less likely to go to the doctor when they find a lump

Table 4.
Mean ratings for blacks and whites of each explanation, on a scale where 0 is false, 2 is neutral, and 4 is true.

Explanation Whites Blacks
1. Blacks are more likely to get heart disease.
HFAT 2.43 3.20
HEXC 1.29 1.80
HJOBS 1.72 2.10
HPREJ 2.08 2.85
HGENE 2.58 2.15
HDOCS 2.34 2.65
HWARN 1.20 1.75
2. Whites are more likely than blacks to commit suicide.
SCOPE 0.48 1.05
SPOOR 1.80 1.75
SFAIL 1.72 1.10
SCOMM 2.28 2.90
SBLAME 1.53 1.05
3. Whites are more likely to die in automobile accidents.
ACOPS 0.98 0.85
ACARS 2.83 1.90
APREJ 0.48 0.45
ACOOR 0.72 0.50
4. Black women are more likely to die of breast cancer.
BGENE 2.44 1.30
BBSE 2.69 3.05
BAFFD 2.94 3.15
BDOCS 1.00 1.70
BBEH 1.82 2.50
BLUMP 1.78 2.15

About this document ...

Perceptions and attributions of race differences in health risks

This document was generated using the LaTeX2HTML translator Version 96.1-c (Feb 29, 1996) Copyright © 1993, 1994, 1995, 1996, Nikos Drakos, Computer Based Learning Unit, University of Leeds.

The command line arguments were:
latex2html -html_version 3.1 -split 0 -no_subdir -ascii_mode fog.

The translation was initiated by Jonathan M Baron on Tue Dec 30 15:31:11 EST 1997

...Jr.
This research was supported by a grant from the Annenberg Foundation. Send correspondence to Jonathan Baron, Department of Psychology, University of Pennsylvania, 3815 Walnut St., Philadelphia, PA 19104-6196, or (e-mail) baron@psych.upenn.edu. O. G. is in the Annenberg School.
 


Jonathan M Baron
Tue Dec 30 15:31:11 EST 1997