Bryan Caplan  

Does Personality Matter? Compared to What?

One-Party State watch... Guys in Suits...
I finally found the time to read "The Power of Personality: The Comparative Validity of Personality Traits, Socioeconomic Status, and Cognitive Ability for Predicting Important Life Outcomes."  [new working link!] It's a meta-analysis, so you've really got to trust the authors to be confident in the results.  But if the world works the way the authors say it does, all social scientists ought to be paying attention.

According to this article, you can measure personality with a simple survey, then use it years or decades later to make good predictions about mortality, divorce, and occupational status.  The research team usually reports results from studies with decent control variables, but of course the quality of past research varies.  The paper ends with three key graphs.

The first shows correlations between mortality and (a) socioeconomic status, (b) IQ, and (c) four of the Big 5 personality traits:

In The Bell Curve, Herrnstein and Murray showed that IQ almost always out-predicts SES.  If this paper is right, conscientiousness alone out-predicts IQ for mortality.

The second key graph shows the correlation between divorce and (a) SES, and (b) three of the Big 5 personality traits:
Many of us (Arnold for example) think of divorce as a low-SES problem.  If this paper is right, though, divorce is much more of a low conscientiousness, high neuroticism, low agreeableness problem.

Finally, in the last graph, we see the correlation between occupational status and (a) SES, (b) parental income, (c) IQ, and (d) various personality traits. 

The specific personality traits are unspecified because the desirability of traits varies from job to job - think salesmen versus librarians.  I do wonder, though, why they didn't show a separate bar for conscientiousness, which is supposed to predict job performance in almost any line of work.

Overall, I'm inclined to believe these results.  In my experience, people are highly yet predictably different in their preferences.  Since luck usually averages out in the long run, it seems like these predictable differences should lead to large average differences in people's lives.  The fact that prominent personality specialists will stick their necks out and make these generalizations makes me marginally more confident in my initial intuitions.  Does anyone else want to read the whole piece, and tell us how convincing you found it - and why?

Comments and Sharing

COMMENTS (10 to date)
Steve Reilly writes:


The link to the paper isn't working for me. I tried Google but couldn't find an unguarded version. Do you know of one? Thanks.

Robin Hanson writes:

Gated copy is here.

Justin Wehr writes:

Fascinating stuff. I am wondering if they ran any regressions to test the effects controlling for other variables.

It hardly seems possible that personality traits could have a stronger effect on occupational outcomes than parental income because, presumably, parental occupational outcomes would be influenced by personality traits among other things.

Steve Sailer writes:

Since the 1990s, I've been saying that conscientiousness or work ethic or whatever you want to call it is at least as important as IQ, but it's much harder than IQ to measure quickly and reliably.

I wonder if paper and pencil tests of conscientousness haven't been all that popular because they are liable to occasional catastrophic failures? The danger is that somebody with a high IQ but poor character would use his smarts to figure out what answers on the test would make him sound like the second coming of George Washington. And a high-IQ scoundrel is the last person you want to select.

You could call it the Ahmad Chalabi Problem. The Iraqi convicted embezzler with a Ph.D. in math from the U. of Chicago used his enormous brainpower to figure out how to dupe the neocons into believing that he literally was the George Washington of Iraq, so America should invade his homeland to make him president.

In contrast to character tests, the good news about IQ tests is that they are un-outsmartable. If you can use your brain to figure out what answers the test makers want, well, then you have a high IQ.

Steve Sailer writes:

Dear Bryan:

I like your idea of focusing on Conscientiousness. Unlike the other personality characteristics where different traits are best in different situations, so they are complicated to think about, Conscientiousness is much like IQ in that -- all else being equal -- more is better in most normal situations.

Law school applications are basically measures of IQ and Conscientiousness, with the LSAT score being more weighted toward the former and the GPA being more weighted toward the latter.

Stephen S writes:

I'd like to see the correlations between SES &c. and having a first-degree relative who is OCD.

Arare Litus writes:

The paper is on my "to read" list.

A few questions regarding the data that would be helpful in interpreting the plots:

(1) How is "low" defined? For that matter what is "high"? Do they form some sort of binning? Look at the distance from the mean?
(2) How do they aggregate all the different studies? Could you do a posting on this? [comparing studies, with different metrics, sample sizes, etc. and then testing to ensure the outcome is not sensitive to the input] Does economics have a different convention than other social sciences on how to do this?

Now a few ramblings:

"Many of us (Arnold for example) think of divorce as a low-SES problem. If this paper is right, though, divorce is much more of a low conscientiousness, high neuroticism, low agreeableness problem."

The ideas that SES is the key predictor seems to be commonplace - with the "right" basically using it as a proxy for poor personality traits, and the "left" using it as a proxy for social bias. To me the fact that the bundle of personality traits is stronger than SES is nice to see - it shows, if true, that SES is a poor proxy of either of these interpretations.

It also suggests to me that your statement that "Since luck usually averages out in the long run, it seems like these predictable differences should lead to large average differences in people's lives." is incorrect, and that the timescale of averaging out is long relative to ones life - thus averaging over a large enough group should average out luck, but an individuals life will be highly effected by luck.

The lowdown, to me is (a) personality matters highly - the classic view, (b) luck has a large effect on life - the classic view, (c) SES is a poor proxy for either character or bias - the most important and surprising finding to me. [It is interesting to me to see the first sentence of the study: "The ability of personality traits to predict important life outcomes has traditionally been questioned because of the putative small effects of personality." - how did the field decide personality has a small effect? Is this not at odds with all personal evidence and the classical view of humanity? Answering this question would be a nice case study - since the data seems to support the classic/personal view, what on earth lead the field to believe otherwise?!!?]

I'll let you know what I think post-reading: but coming into it I'm biased to agree - the study fits with common sense and the classic understanding of people. Due to this bias I'm apt to be harder on the study.

The Cupboard Is Bare writes:

The degree to which SES, IQ, conscientiousness, and the other Big Personality Traits exist within a given individual and in those individuals with whom they interact can produce very different outcomes.

In order to be successful, you have to be aware of your shortcomings. You improve those traits that you can and then you position yourself with those individuals whose traits complement yours.

If only I had known this back when I was 18. :)

Dr. T writes:

I am underwhelmed. The correlations between personality and major events are weak.

Mortality: The best correlation was 0.09, which means that factors other than personality had ten times more effect.

Divorce: The best correlation was 0.18, which means that factors other than personality had five times more effect.

Occupational status: The correlation was 0.23, which means that factors other than personality had four times more effect.

I see no ability to predict life events for an individual based on personality tests. Predictions for populations also are weak: all you can say is that a large population of a certain personality type will have a different divorce rate or mortality rate than another large population. I see little utility for these correlations.

Brian H writes:

You remind me of The Eastwood Dictum: "A man's got to know his limitations!"


Comments for this entry have been closed
Return to top