The role of cognitive traits in predicting school and life outcomes is well-documented. But what about noncognitive traits such as personality? Do they have an effect and if yes how important is it?
Readers of this blog might remember an earlier article, which showed that personality has little effect on long-term outcomes such as GPAs, but a significant effect on one-time high-stakes test performance.
In this post, I will discuss a paper by Mendolia and Walker (2014), which looks at the effect of personality on exam performance and subject choices in England.
The study looks at around 5,000 high school students born in 1990 observed for 7 years starting in 2004. According to the authors, personality is more malleable than IQ, especially in adolescence. Because, as they say, IQ is more or less finalized by age 8, whereas personality is relatively flexible even during adolescence. I cannot confirm these claims though, but suppose they’re correct. This gives a very good motivation to study the effects of personality, as it may be more easily changed by intervention to achieve better results.
The personality traits considered in this paper are not the usual Big Five traits, but rather the following ones: locus of control, self-esteem and work ethic. The latter two are self-explanatory, while locus of control means one’s belief about how much control one has over one’s life. Having an external locus of control means believing that one is largely not in control, that luck or connections have more to do with success than hard work and one’s own actions. Having an internal locus of control is obviously the opposite.
Most past research indicates that locus of control has some positive effect on the “end results” of schooling decisions such as years of schooling or income later in life. There is not much research on immediate results (grades, subject choices) of the aforementioned personality traits.
As for the Big Five traits, generally conscientiousness and to a lesser extent openness to new experiences are associated with positive schooling outcomes. You can read about this in more detail in the earlier post of mine mentioned in the introduction.
On to our current paper, let us start with the measures of performance. The authors looked at GCSE grades, A-levels subject choices and A-levels grades. Now, this is some British stuff. As I understand, the GCSEs are exams taken at the age of 16 by everyone in most subjects studied (around 10); A-levels subjects can be optionally chosen (kind of like honors or AP courses in the U.S.) at the age of 17 and 18, and of course there are exams afterwards. A-levels are also used for college admissions.
The authors check whether the three personality traits mentioned above can explain any of these performance measures. Of course, they control for various characteristics such as gender, family background, or ethnicity. They use three methodologies to estimate their model, the most interesting of which is propensity score matching, but in the end OLS does a pretty good job as well.
So let’s see the main results by personality trait.
Locus of control. This seems to be the best predictor out of the three traits. The general picture is that an external locus of control is negatively associated with school performance on almost all measures of performance. Students with an external locus of control have on average 8 to 11 percentage points lower grades on A-levels math and science. The effect of locus of control is also very similar across socioeconomic groups, though disadvantaged (poor) students are somewhat more sensitive to having an external locus of control.
Work ethic. Not a bad predictor, a high work ethic is significantly positively associated with about half of the performance measures. This seems to be a better predictor for GCSEs, i.e. earlier in adolescence (age 16), and not as good for A-levels performance or choices. A high work ethic appears to be more important for math and science, raising grades there by 8 to 14 percentage points. As for differences between rich and poor kids, ethics barely predict performance for richer kids, while for poorer kids they’re much more important, especially for the GCSEs.
Self-esteem. This is the worst predictor of the three. A low self-esteem is associated with somewhat worse outcomes on some of the performance measures, but most of these are not robustly significant. The general picture is that math and science performance are hurt somewhat by a low self-esteem. Moreover, for richer students self-esteem does not matter at all, for disadvantaged students there is some negative effect, mostly for GCSE math grades.
The two main conclusions that can be drawn are as follows.
Personality’s important for disadvantaged students. And for students in good financial situation, they’re more negligible. This may be because richer students are pushed more by their parents and environment to succeed in school, take A-levels, and go to college afterwards. So the effects of personality traits may simply not be that important. Disadvantaged students, however, do not have such pressure to perform well in school, so for them personality traits are a better predictor of performance.
Math and science well-explained, English not. The three personality traits examined in this paper explain math and science performance much better than English performance. This may be because an internal locus of control, high work ethics and a high self-esteem are much more important for arguably more difficult (or at least more time consuming) courses such as math and science than for English. Performance in English might be better explained by other traits such as creativity, which are not examined in this paper.
A final thing to note is that the predictive power of personality traits is comparable in magnitude with the effects of such important variables as family income, the presence of older siblings or whether the student lives in a single parent household.
One thing I do not quite understand though is why the authors did not control for cognitive skills. They had all kinds of demographic and family background controls, but not cognitive skills such as IQ or similar. There is the quite possible danger that if one controls for IQ, the effect of these personality traits significantly weakens or even disappears. This is a definite weak point of the study. I don’t recall the authors discussing why they do not control for IQ.