Frequently Asked Questions about the article "Study ties political leanings to hidden biases" appearing in the Washington Post - January 30, 2006

1. Does this research suggest that Bush supporters have racial biases and Kerry supporters do not?

No. On average, the White residents of both districts that went for Bush and districts that went for Kerry reported preferring Caucasians to African-Americans. The White preference was somewhat stronger in districts that showed greater support for Bush. This does not mean that all Bush supporters have stronger White preferences than all Kerry supporters. For example, men are taller than women on average. However, that does not imply that every man is taller than every woman.

2. Does this research suggest anything about George Bush or John Kerry’s own racial preferences?

No. These data cannot be interpreted as saying anything about the candidates’ racial attitudes. To our knowledge, they did not participate in the study.

3. Does this research suggest that Kerry supporters are better or more moral than Bush supporters?

No. Scientific data can only describe relationships among variables, such as this study’s finding that stronger support for Bush in the 2000 and 2004 elections was associated with stronger pro-White preferences. Whether such a relationship means that a person, an attitude, or a behavior is moral or immoral is not a scientific question. Such moral questions are answered by philosophical and religious traditions, or social conventions and standards.

If having racial preferences for one group over another has (negative) moral implications from your perspective, consider that there may be other variables in which Kerry supporters might look ‘less moral’ than Bush supporters. So, even if you are a person who believes that these findings reflect poorly on Bush supporters, it would be imprudent to make a global judgment about moral character of a group of voters from these data.

4. Does this research suggest that all Bush supporters prefer White people to Black people?

No. These data are aggregated by congressional district. So, it is the average of the respondents in a given district. Many Bush supporters (and many Kerry supporters) reported equal liking of African Americans and Caucasians.

5. Are the researchers own biases responsible for the findings?

This is an issue for all scientific research. Researchers usually have a hypothesis about how the data will turn out before conducting the study. Hypotheses are usually based on previous research and theory or researcher intuitions. 40 years of social science research has observed a link between racial attitudes and political orientation. So, there was reason to expect this to persist in the current study.

The possibility that the researchers’ hypotheses influenced the outcome of the study is called “experimenter effects.” These influences can come in the form of how questions are asked, what the experimenter does in interaction with the participants (e.g., subtly encouraging different types of answers), and in how the data are analyzed or interpreted.

In the current study, the questions were straightforward (e.g., How warmly do you feel about Black Americans?), the researchers never interacted with participants (all respondents to the racial attitude measures were volunteers at a public website), and much of the data came from public sources (e.g., the U.S. census). The data analysis procedure is well-established. If there are plausible experimenter effects in the current study, they will need to be identified and tested.

If science is working well, then individual researcher’s preconceptions should be ruled out as influences on results through critical review and replication by other scientists. This is not a threat to scientific research; it is a core value of science – the facts should prevail, even if they are counter to the scientist’s personal beliefs.

6. Can the results from this sample be generalized to the U.S. population?

Scientists must be careful when considering how their samples can be generalized to broader populations. In this case, the sample of implicit and explicit preferences is very large, but it is not a random or representative sample of the U.S. population. As a consequence, the values should not be taken as parameter estimates of what the population is like. At the same time, the analyses test questions within the sample itself. Because the selection pressures of the sample are not likely to vary systematically according to the respondent’s geographical location, we can be more confident in interpreting differences across districts within the sample as having some application to the U.S. as a whole. This is an inference, not something that is guaranteed by the data or the analysis. So, other possibilities could be introduced and tested as alternative explanations. Confidence with these data comes from the fact that it is difficult to generate a plausible selection pressure that could have produced the highly-reliable pattern observed in these data.

7. Does this research suggest that racial attitudes cause voting decisions?

While that may be a plausible interpretation, it is not confirmable with these data because it is not an experiment. We cannot confidently infer cause-and-effect. It does seem unlikely that voting caused racial attitudes in this case. However, it is possible that a third variable caused both, such as political ideology. This is a topic for future research.

Also, even if this relationship were a causal one, these data do not suggest that all voters are using their racial attitudes to decide how to vote. The 2004 election had many prominent issues that were surely important causal factors in voting behavior. If a causal relationship exists, it would still be the case that racial attitudes are just one of many factors that influence voting decisions.

8. Are these data about people or about congressional districts?

The analyses are comparisons across congressional districts. Obviously, the ultimate interest is in the individuals that comprise those districts. While it may be plausible to translate the current findings to the voting behavior of individuals, this inference is subject to further examination.

9. Are these data published?

No. The Washington Post covered a lecture given by Brian Nosek at the Society for Personality and Social Psychology convention in Palm Springs on January 28th, 2006. While other experts have seen summaries of the analyses, the reported findings should be interpreted carefully, especially prior to peer review for publication.

10. Did the study include participants of all ethnicities?

No. The analyses only included White respondents from each district. Including Black respondents would have introduced a biasing factor in that Black respondents tend to favor Black people to White people, and Black people also tend to support liberal candidates. So, if Black respondents were included, the effects could be due to these two conditions, rather than anything to do with other groups' racial attitudes. As you would expect, the differences across districts are exaggerated when you include Blacks in the sample.