Is self-discipline more important than IQ in predicting academic performance?

Question

There is one highly cited (>1K citations in Google Scholar) study titled "Self-discipline outdoes IQ in predicting academic performance of adolescents", and quoting from its abstract:

Self-discipline measured in the fall accounted for more than twice as much variance as IQ in final grades, high school selection, school attendance, hours spent doing homework, hours spent watching television (inversely), and the time of day students began their homework. The effect of self-discipline on final grades held even when controlling for first-marking-period grades, achievement-test scores, and measured IQ. These findings suggest a major reason for students falling short of their intellectual potential: their failure to exercise self-discipline.

As a side-note, the author also gave a highly watched TED talk.

Given the usual replication issues in psychology, has this result been replicated, corroborated, criticized etc.?

And to put my doubt in a bit more perspective, in one large UK study (5-year prospective longitudinal, 70,000 of children) the correlation between IQ at age 11 and academic test score (GCSE) five years later was 0.81. And the same study setup repeated four years later on even larger sample of 175K obtained a similar correlation of 0.83. Alas the British studies didn't look at any measures of self-discipline, personality etc. The story in the US is not that different: ACT was found to correlate 0.77 with IQ on a large sample, while the correlation of SAT with IQ was 0.82 (0.87 when corrected for nonlinearity). I didn't dig into the paper to see at what age the IQ was tested in this case. On smaller samples, the correlation reported was less, for some reason that is unclear to me right now. However, one criticism of SAT/ACT which probably translates to the GCSE as well is that these tests themselves are a thin proxy for IQ testing (i.e. not enough knowledge is being tested in them.)

Answer 1

(Partial answer)

While researching the topic of high failure rates in introductory computer science/programming classes, I found a paper (from researchers urelated to ogrinal grit paper) which determined that grit was indeed a much better predictor of succes in such a class than the ACT score (which is taken as a much closer proxy to IQ by the study's authors). Using logistic regression they say that

Our data suggest that there is a positive association between grit and student success in introductory programming courses (β = 2.332, p = .002). That is, grittier students earn higher grades. This result provides an affirmative answer to the first research question -- grit does predict student performance in high failure-rate courses as it does in low-failure rate ones.

The results on the influence of ACT were mixed. Though the correlation matrix shows an insignificant link between ACT and grades in introductory programming courses, the relationship is statistically significant in the regression analysis (β = 0.256, p = .028). However, ACT score, as a proxy of student intelligence, has much less influence on course grade than grit. Thus, the answer to our second research question is also affirmative – grit is a more powerful predictor of course grade than student intelligence.

So if I'm allowed a bit of pun here, Spolsky's "smart and gets things done" is more of a case "damn persistent and somewhat smart", at least at this introductory programming level.

Also, it's interesting that although grit supposedly predicts [other] school grades, high-school grade-point-average could not predict success in CS1:

Data also shows that high-school GPA (HSgpa) did not significantly predict grade (OR = 1.448, β = 0.370, p = .650).

---

And on a more general note, there's one recent (2016) meta-analysis, which notes in its abstract:

Our results based on 584 effect sizes from 88 independent samples representing 66,807 individuals indicate that the higher order structure of grit is not confirmed, that grit is only moderately correlated with performance and retention, and that grit is very strongly correlated with conscientiousness. We also find that the perseverance of effort facet has significantly stronger criterion validities than the consistency of interest facet and that perseverance of effort explains variance in academic performance even after controlling for conscientiousness. In aggregate our results suggest that interventions designed to enhance grit may only have weak effects on performance and success, that the construct validity of grit is in question, and that the primary utility of the grit construct may lie in the perseverance facet.

I have yet to read the whole paper, but (amusingly) the way I found this was wondering/googling whether I should have added "passion" to my Spolsky-like take on the CS1 predictors. But from this meta-analysis, apparently it's not that important... for the way grit is measured by current psychometric tests, such as Grit-8; just browsing the questions there, only one of the strikes me as being obviously about passion; in the rest of the questions I guess it was hoped it would be implicit/correlated, but from the meta-analysis it seems that's not the case much, so as currently measured grit is mostly "perseverance of effort", which I'd rather call persistence or tenacity.

I should add that Duckworth has responded to this meta-analysis... on NPR.