The Interaction Between Conscientiousness and General Mental Ability: Support for a Compensatory Interaction in Task Performance

Abstract

We propose a compensatory interactive influence of conscientiousness and GMA in task performance such that conscientiousness is most beneficial to performance for low-GMA individuals. Drawing on trait by trait interaction theory and empirical evidence for a compensatory mechanism of conscientiousness for low GMA, we contrast our hypothesis with prior research on a conscientiousness-GMA interaction and argue that prior research considered a different interaction type. We argue that observing a compensatory interaction likely requires (a) considering the appropriate interaction form, including a possible curvilinear conscientiousness-performance relationship; (b) measuring the full conscientiousness domain (as opposed to motivation proxies); (c) narrowing the criterion domain to reflect task performance; and (d) appropriate psychometric scoring of variables to increase power and avoid type 1 error. In four employee samples (N₁ = 300; N₂ = 261; N₃ = 1,413; N₄ = 948), we test a conscientiousness-GMA interaction in two employee samples. In three of four samples, results support a nuanced compensatory mechanism such that conscientiousness compensates for low to moderate GMA, and high conscientiousness may be detrimental to or unimportant for task performance in high-GMA individuals.

Appendices

Appendix 1

Additional Analyses for Sample 2

Because personality and GMA batteries in sample 2 were used to select employees and included data on all applicants, but only performance data for hired employees were available, there was evidence of direct range restriction in sample 2. Minimum values for conscientiousness and GMA in the full applicant data (conscientiousness: Min. =  − 2.58, Max = 4.00, SD = .92; GMA: Min =  − 2.46, Max = 1.99, SD = .87) were substantially lower than were minimum values among hired employees (conscientiousness: Min =  − 1.22, Max = 3.77, SD = .84; GMA: Min =  − 1.74, Max. = 1.99, SD = .82). To determine whether range restriction might have affected findings, we conducted post-hoc analyses that included multivariate imputation by chained equations using the ‘mice’ package (van Buuren, 2018; Pfaffel et al., 2016) to impute performance data for the applicants not hired.

Results of analyses using imputed data to account for range restriction still did not show evidence of a significant conscientiousness-GMA interaction. However, results did show a marginally significant quadratic interaction, b =  − 0.03, p = .082, and GMA accounted for a proportion of the variability in the conscientiousness-performance relationship (45%) comparable to that of other samples. Results are shown in Table 3 below.

Additionally, to evaluate the potential impact of limited power on our findings in sample 2, we conducted three Monte Carlo simulations using parameter estimates from samples 1, 3, and 4, respectively, and sample size from sample 2 (N = 261). Across 1000 replications, simulations suggested that when sample size is reduced to that of sample 2, the true positive rate is .68 for sample 1, .22 for sample 3, and .24 for sample 4. In other words, if the findings for sample 1, 3, or 4 were true in the population, the sample size in sample 2 is under-powered.

Table 3 Results of task performance predicted by conscientiousness and GMA in sample 2 corrected for range restriction

Appendix 2

Top-Down Selection Simulation Analyses

For each dataset, 1000 samples of 100 people each were randomly chosen as pools of “applicants.” We then used the coefficients for the “standard” linear model (i.e., using GMA and conscientiousness as linear predictors of performance) to calculate predicted performance and, separately, the coefficients from the quadratic interaction model to calculate predicted performance for each sample. Using these predicted values, we selected (i.e., “hired”) both the top 10% and top 20% of applicants. We then used the actual, observed performance for each selected applicant to calculate (a) mean performance, (b) maximum performance (i.e., the performance score for the top performer), and (c) the minimum performance (i.e., the performance score for the lowest performer). Then, we determined the “winner” of the two models regarding their selection decisions. For the minimum and maximum, the winner was the model that picked the “better best performer” and the “better worst performer,” respectively. For the mean, we considered the mean scores to be different only if they showed greater than a .10 SD difference (i.e., one standard error for the population of 100 applicants). Table 4 shows the results for simulated selection scenarios. The number outside the parentheses is the percentage of times (of the 1000 simulated applicant pools) that the quadratic interaction model was the winner for the selection ratio. The number inside the parentheses is the percentage of times the standard linear model was the winner⁶

The results suggest that, in most cases, the quadratic interaction model is better at predicting top performance, especially when the selection ratio is 20%. The quadratic interaction model generally resulted in higher average performance, the selection of the better top performer, and the selection of the better worst performer. In most instances, the quadratic interaction model won more than twice as many times as the linear model, indicating its general superiority in making top-down selection decisions. Thus, the simulation results suggest that, despite small effect sizes using traditional metrics (e.g., ΔR²), the quadratic interaction has substantial practical value in its ability to more effectively predict performance. Notably, this is true even for sample 2, despite the non-significance of the interaction. These results further point to the limited power of sample 2 as a potential explanation for the non-significance of findings

Table 4 Comparison of the quadratic interaction model and the standard linear model efficacy in simulated top-down selection