The strongest predictor of academic achievement is intelligence (g; correlations range between .30 - .70). To date, there is no meta-analysis whereby the association between academic achievement and intelligence has been examined. Earlier meta-analyses (using data from before 1983), only pay attention to the domain of natural sciences and a number of countries do not correct for possible artifacts in the data (for example, range restriction or unreliability). Furthermore, academic achievement has often only been measured by means of achievement tests, whereby school grades are ignored as being a direct measure of academic success. This seems arbitrary, as grades are important in predicting further occupational (and before that, scholastic) success. This meta-analysis focusses on integrating previous research with regard to the correlation between academic success. This is measured by means of school grades and g.
Assess all studies available to date on the subject on a global scale
To present the mean correlation weighted only by the size of the sample, and also to present the true score correlation (corrected for range restriction and unreliability)
Assess possible moderator variables that could affect the correlation.
Boulanger (1981), Fleming and Malone (1983), and Steinkamp and Maehr (1983) all conducted meta-analyses whereby the mean correlation between academic achievement and cognitive abilities ranged between ρ = .34 - .48, which is a little lower than generally denoted in the literature. In the possible moderators that were tested, significant effects were only found on reliability of the outcome measure and grade level. The bulk of the studies considered only students’ achievement in scientific subjects, and not in other domains. Moreover, possible differences in the mean correlation between school grades and g across the domains considered were not analysed.
In regard to the goals set, the authors hypothesize that grades achieved in school have a greater effect on a student’s later school and occupational vocation/ success than other measures used to determine scholastic achievement (for example, school achievement tests or teacher ratings). In this study, only school grades were used as a criterion.
In this, population correlation between school grades and g were investigated with no restrictions on subject, grade level, year of publication or country. Also, the moderating effect of third variables are illustrated.
Five potential moderators were identified:
Type of intelligence test – to test the possible moderating effect of either the nonverbal or verbal character of IQ tests, subgroups were constructed for both these groups.
Subject domains- to test this, school subjects were clustered (i. Biology, physics, and math; ii. Languages; iii. Social studies, history, and geography; iv. Fine arts, music and sports), in order to estimate the population correlation between grades and g.
Grade level – to test this, grade levels were clustered into elementary, middle, and high school.
Gender – to test whether there was an influence of gender as a moderator variable in the relationship between grades and g.
Year of publication – to test whether there has been a change in population correlation with regard to academic achievement and g since the previous meta-analyses that were conducted on the subject, two groups were made: i) studies published before 1983, ii) studies published after 1983.
Inclusion criteria were the following: i) the independent variable of the studies was general intelligence measured by standardized intelligence tests or standardized tests of the same nature, ii) the dependent variable was that school achievement had to be measured by GPA or the grades that a student received on, for example, their report card, iii) studies had to use only primary or secondary school grades, iv) studies had to report zero-order correlations between dependent and independent variables, v) sample sizes had to be mentioned in the paper, vi) the paper had to be written either in English or in German , vii) the studies were not allowed to contain any methodological flaws.
There were 162 primary studies included in the final data set for this meta-analysis. The studies were all published between 1922 and 2014 providing k = 240 independent samples. The total sample consisted of 105,185 (studies were included ranging from 15 – 9776 students). Average age was 13.9 years old (SD = 4.0). The total sample included 50.8% females and a total of 33 countries.
The central finding for the total sample was a mean correlation between school grades and intelligence of ρ = .54. The respective confidence interval did not include zero, thus this can be seen as significant. As this correlation is large, this seems to demonstrate that intelligence is of importance for academic achievement. However, this result is not generalizable as there is substantial residual variance that is not explained by the methodological artifacts that were corrected for.
In sum, the results show that intelligence does substantially influence school grades and can be seen as (perhaps the most) influential variable within this context. Intelligence was found to be a significant predictor on all of the moderator levels that were tested, although some of the scenarios that were tested leaded to even higher validities.
For example, population correlation was found to be the highest for nonverbal and verbal materials, which indicate that a broad measure of g can be seen as the best predictor for school grades. Also, the importance of g rises over grade levels. In other words, intelligence seems to be especially important in more complex educational contexts, whereby the content can only be fully mastered in accordance with an suitable cognitive ability level.
With regard to the learning content, it seems that g has the highest population correlations in Science and Mathematics, Social Sciences, and Languages. This is likely due the fact that the content in these subject domains have a more clear logical structure than the other subject domains that were tested, and the logical structure is often the central element in intelligence testing.
Population correlations found between g and school grades were found to be higher in the studies that were published before 1983 than after that year. However, the relevance of intelligence, although lower now, can still be seen as substantial. This finding could also suggest that the selection procedures for, for example, universities, colleges and further employment, should employ intelligence tests as well as school grades (as intelligence is responsible for incremental validity – the shared variance between school grades and g is lower at these levels).
Finally, population correlation between school grades and intelligence with regard to gender was found to be independent, meaning that gender did not play a role in this. However, only the results that were established within the female population groups were found to be generalizable. This suggests that there are other variables that play a role within the male population groups such as school anxiety, performance avoidance goals, or motivation.
Add new contribution