What Personality Actually Predicts

Having established that traits exist and predict aggregates, the obvious question is what they predict and how much.

Conscientiousness does most of the work outside social life. It predicts job performance across nearly every occupation studied, which is unusual: the value of the other factors depends heavily on context, since extraversion helps in sales and can hurt in work requiring sustained solitary concentration. Effort, reliability and following through are useful almost everywhere. It also predicts longevity about as well as some established medical risk factors, through a combination of health behaviours and, presumably, not doing reckless things.

Neuroticism predicts mental and physical health outcomes, and it is the factor where the label does the most damage, since the technical term is considerably narrower and less pejorative than the everyday word. The association with physical health is not merely a reporting artefact, though some of it is: plausible routes include stress physiology, health behaviours and how bodily symptoms are attended to.

Extraversion predicts social outcomes and remarkably little else. Openness predicts creative achievement and educational choices. Agreeableness predicts being liked and, reliably and slightly depressingly, somewhat lower earnings.

On stability, two claims get confused constantly. Rank order is fairly stable from about thirty: the more conscientious thirty-year-old is generally the more conscientious fifty-year-old. Mean levels drift with age, and here the finding is genuinely cheerful. Across cultures, people become more conscientious, more agreeable and less neurotic through their twenties and thirties. It is called the maturity principle, it happens without anybody trying and largely without anybody noticing, and it disposes neatly of the idea that personality is fixed at birth or is a matter of willpower. It is neither.

Which brings us to measurement, and a problem that is rarely stated plainly. Self-report works well for research volunteers with no reason to misrepresent themselves. It works much less well for job applicants, because the socially desirable answer is usually obvious. Faking is easy, only imperfectly detectable, and unevenly distributed: the candidates most willing to game an instrument are not a random sample of candidates. Personality testing in selection is therefore legal, weakly predictive and easy to manipulate, which is a combination worth examining before adopting one.

And there is a downstream effect worth naming, because it is where this material does most harm in practice. A team learns that a colleague scores low on agreeableness and begins interpreting everything that colleague does through it. A modest aggregate tendency is being used to explain individual acts, which is precisely what the person-situation debate established it cannot do. Labels also absorb subsequent observation, once applied, they collect confirming instances and shed disconfirming ones. That is one of the better arguments for not circulating such scores at all.