Sunday, September 20, 2009

Race and the Big Five

* p < .05, two-tail test, comparison with white counterpart.

The General Social Survey asked ten questions in 2006 about personality traits--two questions for each of the Big 5. Two questions per trait is not nearly enough (typical performance by sociologists) but we'll use what we've got.

For extraversion, the questions asked about outgoingness, and for agreeableness, people were asked if they agreed that they didn't find fault in others and trusted them. On conscientiouesness, people were asked if they agreed that they were hardworking and thorough. Questions for emotional stability asked about agreeing that you are relaxed or prone to nervousness.

Respondents were asked if they are imaginative and artistic, so I called this measure "creativity." Answers to these pairs of questions were summed. You see the means in the table.

Whites are significantly more extraverted than Mex-Ams, but do not differ from blacks. Blacks are more agreeable than whites, but Mex-Ams are less so. The group do not differ on conscientiousness.

Blacks are more emotionally stable than whites. This contradicts findings from my last post, but keep in mind that the measure I used before tapped depression more than anything else, while this one focused on how relaxed someone is.

Finally, the groups do not differ on creativity.

One comment about the data: it's clear that respondents lack some objectivity because a disproportionate number of people describe themselves as having what I'm sure people see as the positive quality. Few people admitted that they are lazy and fail to do a thorough job. Conscientiousness was more skewed than any other measure, but the others were skewed as well. There is a tendency to think you're outgoing, agreeable, conscientious, relaxed, and creative, regardless of the facts. I don't know if groups differ in these biases, but there is an overall bias, no doubt.


ken said...

You can never trust someone to rate themselves. I read one study (fuzzy on the numbers) where 90% of people said they were in the 10% of social agreeableness. For intelligence, I think 80% of people think they're above average.

Anonymous said...

You actually can trust people to rate themselves... and you can trust them to err in predictable ways. It is difficult to make cross-group comparisons with self-report scales because everyone is likely to rate themselves according to what is normal for one's own group (whatever the respondent considers that to mean). Therefore, the more racialized the respondent, the more the answers are normed to the race average. Generally, the more insular the group, the more idiosyncratic the responses.

Also, blacks unfortunately have more egotistic/narcissistic self-views (you may judge this from their enormous self-esteem as a group, or on any proxy variable of said attribute). They will come up glowing on any self-report that is transparently polarized into a better or worse response. Conscientiousness and stability fit the bill. If you're black, are you going to tell someone that you're unstable and unreliable? No, no, no.

Steve Sailer made a similar observation once, on why emotional intelligence is so wonderful and egalitarian a tool: because the questions which judge EQ are transparent.

So trust, but trust carefully with self-report measures. Be slow to make cross-group comparisons with them.

bgc said...

I agree with anonymous that the take home message of self-rated personality scales is that they work pretty well within homogenous groups and where there is no incentive to lie. But they certainly do not work across cultures - the international comparisons are self-rated personality are just rubbish.

Richard Lynn has told me that although he used to publish international comparisons of self-rated personality, he has now concluded that these are not valid comparisons - and now uses behavioural comparisons instead (e.g. measuring crime rates instead of 'Psychoticism').

The way ahead will be to use the within-group data to vaidate genetic markers, and then use the markers between populations.

In the mean time it might be possible to use self-reported behavioural markers in GSS (and other similar surveys) which correlate within groups to calibrate the self-rating between groups.

So, find out how many job changes, or periods of unemployment, or level of insurance, or drug use (or whatever other objective correlate of C might be devised) corresponds to 'average' self-rated conscientiousness levels within various homogeneous groups.