Traits

Mini Review

• A trait is a difference between people in what they feel, do, and think that holds across situations and over time.
• The Big Five are the five broad traits that consistently emerge from factor analysis of personality scores.
• Trait scores come from four types of measures, and validity is the accuracy of a score for a particular use.
• Some particular uses of Big Five scores have been tested in more research than almost anything else in psychology: how well they predict certain life outcomes…

Prediction

• Prediction: Using a person’s score on one variable to estimate their score on another.
  • Ex: Using a student’s conscientiousness score to estimate their GPA.
  • Ex: Using today’s high temperature to estimate tomorrow’s.
• The correlation between the two scores is the number that says how accurate the prediction will be.

The Same Correlation, a New Use

• So far, we’ve used correlations to describe and measure traits.
  • To describe traits, we used factor analysis to condense thousands of correlations among trait words into a few broad factors.
  • To measure traits, we used the correlation between two measures of the same trait to check score reliability (the two halves of the class, or the same questionnaire taken twice, weeks apart) and score validity (your ratings and my self-report).
• Now, we’re talking about a correlation between a trait score and a different outcome, and using it to estimate that outcome for people we have not measured on it.
• The statistic is the same each time. What changes is what the two variables are and what we use the number for.

Where a Prediction Correlation Comes From

• A study has a trait score and an outcome for each of many people, with the trait score from earlier and the outcome from later.
  • Ex: Conscientiousness scores from 500 first-year students, then their GPA four years later.
  • Ex: Big Five scores collected decades ago, matched to the death records that exist today.
• The correlation describes that group. It says how closely higher trait scores went with higher outcomes across those people.
• Nothing is predicted about the people in the study. Both of their scores are already known.

Using the Correlation on a New Person

• Prediction is when we use a correlation found in a research study to estimate the outcome of a person who was not in that study.
  • Ex: A new first-year student takes a conscientiousness questionnaire. Can we use her score to predict her GPA at graduation four years from now?
• If conscientiousness and GPA are not correlated, the best prediction we can make is that she will have an average GPA.
  • Why? Because most people are in the middle. So if you don’t know anything else, guess the middle!

Using the Correlation on a New Person, Continued

• But if conscientiousness and GPA are correlated, we can do better than the average.
  • If her conscientiousness score is higher than average, we predict a GPA higher than average. If it is lower than average, we predict a GPA lower than average.
  • The stronger the correlation, the further from the average we predict, and the closer her actual GPA is likely to be to our prediction.

Scatterplots at Four Correlation Sizes

• Each dot is one person in a study. The line is the prediction for a new person at each trait score.
• At r = .10 the line is nearly flat, so the prediction is barely different from the average, and the dots land far from it.
• At r = .50 the line is steeper and the dots land closer to it, so the prediction is more accurate. Most personality findings fall between the two middle panels.

How Big Is a Correlation?

• Rules of thumb for research on people (2019): r = .05 is very small, r = .10 is small, r = .20 is medium, and r = .30 is large. Correlations above .40 between a trait and a life outcome are rare.
• A correlation runs from −1 to 1. The sign gives the direction, and the size says how accurate the prediction is. A correlation of −.20 predicts as well as one of .20; the difference is which end of the trait goes with the outcome.
  • Ex: Neuroticism and well-being correlate at about −.46. Higher neuroticism scores go with lower well-being, and the prediction is as accurate as it would be at .46.

Conscientiousness and GPA at r = .20

• Each curve is the GPA of students who have one conscientiousness score. The dashed line inside each curve is the predicted GPA for that score.
• The prediction for a 4 is above the prediction for a 2, but only a little, and the two curves cover almost the same GPAs.

The Same Two Scores at Four Correlation Sizes

• As the correlation grows, the predictions move apart and the curves overlap less.
• Most personality findings are in the middle two panels, but even at r = .50 the curves overlap more than they separate.

What the Correlation Says About a New Person

• The prediction for a new student who scored 4 is a little above the average GPA. Her actual GPA can be anywhere under her curve.
• A correlation of .20 moves the prediction a little and narrows the curve almost not at all. Knowing her score changes where the prediction starts, but it barely changes how sure we can be.
• Even the largest correlations in personality research, around .40, leave a wide curve.

Why a Small Prediction Still Matters

• For one student, r = .20 changes the prediction only a little.
• Across many students, a little adds up. A college that uses the score for 2,000 applicants a year makes a better guess than the average for each one, and those small gains are repeated 2,000 times.
• A trait score is rarely the only information. Conscientiousness, high school grades, and test scores each predict a little, and together they predict more than any one of them.
• The other choice is to predict the average GPA for everyone, which uses no information at all.

Poll

Two questions for each outcome: which Big Five factor predicts it best, and how large is the correlation?

1. Job performance
2. College grades
3. How long a person lives
4. How satisfied a person is with their life
5. Score on an intelligence test

Click Here to Answer

Poll Results

The Life Outcomes Replication Project (2019)

• The replication project (2019): 78 published findings that a Big Five trait predicts a life outcome were retested, preregistered, in new samples of US adults with a median of 1,504 people per test.
  • Finding: 87% of the 78 findings replicated, but the replicated correlations were only 77% as large as the ones first reported.
• Every finding on the slides that follow either survived this test or comes from a meta-analysis published after it.

Personality and School Grades (2022)

• The academic performance meta-analysis (2022): 267 samples and 413,074 students, 48% from Europe and 40% from North America, with the rest from Australia, Eastern Europe, the Middle East, and Asia.
  • Finding: Conscientiousness predicted grades at r = .20, more than any other factor, and the prediction held after cognitive ability was accounted for.
  • Finding: Openness was second at r = .13, but most of its prediction was explained by cognitive ability.
  • Finding: Agreeableness (r = .07), extraversion (r = .01), and neuroticism (r = −.02) predicted grades weakly or not at all.

Personality and How Long People Live (2017)

• The fifteen-sample mortality study (2017): Big Five scores were measured years before death records were checked for 44,094 people in 15 long-running samples, 9 from the US and 6 from the UK, Sweden, the Netherlands, and Australia.
• This study reports its results as chances of dying during the study period, not as correlations. Each finding compares a person near the top of the trait with a person near the bottom.

Personality and How Long People Live (2017), Continued

• Finding: The highest scorers on conscientiousness were 44% less likely than the lowest scorers to have died by the end of the study.
• Finding: The highest scorers on agreeableness and extraversion, and lowest scorers on neuroticism, were all 20-25% less likely to have died than the lowest scorers.
• Finding: Openness made no difference.

Personality and Well-Being (2020)

• The well-being meta-analysis (2020): 462 studies and 334,567 people.
• All five factors predicted well-being (or life satisfaction) to some extent.
  • Finding: Neuroticism was the strongest predictor, at r = −.46.
  • Finding: Extraversion (r = .37) and conscientiousness (r = .36) were next, in the positive direction.
  • Finding: Agreeableness (r = .25) and openness (r = .19) predicted well-being less strongly.

Conscientiousness at Work (2019)

• The century review of conscientiousness (2019): 92 meta-analyses covering 175 work variables, more than 1.1 million people, and more than 2,500 studies were combined. The review does not report the countries separately.
  • Finding: Conscientiousness was related in the expected direction to 98% of the variables, with an average correlation of r = .20.
  • Finding: The prediction of job performance was weaker in highly complex jobs.

Extraversion at Work (2019)

• The extraversion review (2019): 97 meta-analyses covering 165 work variables were combined. The review does not report the countries separately.
  • Finding: Extraversion was related in the expected direction to 90% of the variables, with an average correlation of r = .14.
  • Finding: The correlations reached r = .20 or higher for motivation, positive mood, working with others, and performance.
    • Nuance: The facets differed. Positive emotions predicted well across variables, but sociability predicted little, and sensation seeking predicted in the opposite direction.

Agreeableness (2022)

• The agreeableness review (2022): 142 meta-analyses covering 275 variables, more than 1.9 million people, and more than 3,900 studies were combined. The review does not report the countries separately.
  • Finding: Agreeableness was related in the expected direction to 93% of the variables, with an average correlation of r = .16.
  • Finding: The strongest predictions were for investment in relationships, working in teams, and following social norms.

Openness and Intelligence (2023)

• The personality and ability meta-analysis (2023): 1,325 studies from more than 50 countries, with data on millions of people, were combined.
  • Finding: Openness predicted scores on intelligence tests at r = .26, and its intellect and ideas facets at r = .28.
  • Finding: The other four factors were near zero as whole factors, with neuroticism the largest of them at r = −.07.

Why Cross-Cultural Predictions Are Harder to Compare Than Trait Structure

• Most of the evidence reviewed above comes from Western countries, where the outcomes studied tend to take similar forms.
• The Big Five structure has appeared in many cultures, but this shows how trait descriptions cluster, not whether those traits predict the same outcomes everywhere.
• Predictions are comparable only when the outcome itself is comparable. Grades, office performance, and life-satisfaction ratings all reflect local institutions, practices, and values.
• Where schooling ends earlier, work is primarily agricultural, or occupations are strongly divided by gender, “grades” and “job performance” may mean something different or may not be measured at all. The same trait may therefore predict a different outcome, or no recorded outcome.

Gender and Big Five Scores

• The 27-country study (2026): Samples representative of the adult population in 27 countries, 143,313 people, completed the same Big Five questionnaire.
  • Finding: Women scored higher than men on average on neuroticism and agreeableness, by a medium-sized difference that was larger in more individualistic countries, with smaller differences on conscientiousness and openness and none on extraversion.

Gender and Big Five Scores, Continued

• The 74-country study (2025): Gender and personality were compared across 74 countries and four questionnaires, at the level of factors, facets, and single items.
  • Finding: A person’s gender predicts any single factor score only weakly, but adding up many small differences across many narrow traits identifies gender fairly accurately.
• Overall: The first study describes the average difference between the two groups, and the second describes what that difference lets you predict about one person. A medium average difference means the two groups overlap a lot on each factor, so a factor score cannot reliably predict a person’s gender, and gender cannot tell you a person’s score.

Self-Reports and What Other People See

• Most of the studies above relied on self-report measures.
• The observer-ratings meta-analysis (2010): 263 samples and 44,178 people who rated themselves and were rated by others.
  • Finding: Self-ratings and ratings by people who know the person well agreed at a medium to large level, and other people’s ratings predicted job performance and grades somewhat better than self-ratings did.

Self-Reports and What Other People See, Continued

• The self-other knowledge study (2010): 165 people rated themselves, were rated by friends and strangers, and then completed behavioral tests.
  • Finding: The self was the more accurate judge of inner traits such as anxiety, but friends were the more accurate judges of traits such as intelligence and creativity, and both were equally accurate on extraversion.
• Overall: Self-reports and other people’s reports agree more than they disagree, so both measure the same personality. Where they differ, the self is better at what is felt inside, and other people are better at what is shown, which is why other people’s ratings predict outcomes that other people judge, such as job performance and grades, a little better.

What These Predictions Do Not Show

• The findings above are predictions. In each study, researchers correlated a trait score with a later outcome.
• A prediction says nothing about whether the trait caused the outcome.
  • Ex: Conscientiousness predicts grades at r = .20. That is true whether conscientiousness raises grades, whether something else raises both, or some of each.

Causation

• Causation: A relationship in which (1) two variables are correlated, (2) one of them comes before the other in time, and (3) their relationship is not due to any third variable.
• NOTE: In a study, “predicts” means “is correlated with.” When a study says a trait predicts an outcome, requirement 1 is met, but nothing has been said about requirements 2 and 3.

Why Personality Research Shows Prediction Rather Than Causation

• The findings above meet requirement 1: the trait and the outcome are correlated.
• A study meets requirement 2, the trait coming before the outcome in time, when it measured the trait first, as the mortality study and many of the others did.
• Requirement 3, that no third variable explains the relationship, is the one personality research cannot fully meet. A study can rule out one third variable at a time, such as cognitive ability in the grades meta-analysis, but ruling out all of them takes randomly assigning people a personality, and we cannot do that.