Acquaintance reports of personality and academic achievement: A case for conscientiousness

The present article examines personality as a predictor of college achievement beyond the traditional predictors of high school grades (HSGPA) and SAT scores. In an undergraduate sample (N=131), self and informant-rated conscientiousness using the Big Five Inventory (BFI; John, Donahue & Kentle, 1991) robustly correlated with academic achievement as indexed by both freshman GPA and senior GPA. A model including traditional predictors and informant ratings of conscientiousness accounted for 18% of the variance in freshman GPA and 37% of the variance in senior GPA; conscientiousness alone explained unique variance in senior GPA beyond the traditional predictors, even when freshman GPA was included in the model. Conscientiousness is a valid and unique predictor of college performance, and informant ratings may be useful in its assessment for this purpose. Acquaintance reports 3 Acquaintance reports of personality and academic achievement: A case for conscientiousness The question of what makes a good student “good” lies at the core of a socially-relevant discussion of college admissions criteria. While past research has shown personality variables to be related to school performance (e.g. Costa & McCrae, 1992; De Raad, 1996), academic achievement is still widely assumed to be more a function of intellectual ability than personality. The purpose of this study is to address two ambiguities that trouble past research in this area: the choice of conceptually appropriate outcome measures and the overuse of self-report data. A highly influential meta-analysis by Barrick and Mount (1991) concluded that conscientiousness is a robust and valid predictor of job performance across all criteria and occupations. Soon after, Costa and McCrae (1992) provided evidence that conscientiousness is likewise related to academic performance. This finding has been replicated by others (recently, Chamorro-Premuzic & Farnham, 2003a and 2003b). Moreover, conscientiousness appears to be free of some of the unwanted complications associated with ability as assessed by the SAT: Hogan and Hogan (1995) reported that personality inventories generally do not systematically discriminate against any ethnic or national group, and thus may offer more equitable bases for selection (see also John, et al., 1991). Still, skepticism remains. Farsides and Woodfield (2003) called the relationship between personality variables and academic performance in previous literature “erratic and, where present, modest” (p. 1229). Green, Peters and Webster (1991) found academic success only weakly associated with personality factors; Rothstien, Paunonen, Rush and King (1994) found that the Big Five factors failed to significantly predict academic performance criteria among a sample of MBA students; Allik and Realo (1997) and Diseth (2003) found most of the valid variance in achievement to be unrelated to personality. Acquaintance reports 4 The current study seeks to address two pervasive obstructions to conceptual clarity in the previous literature: 1) Lack of consistency in the measurement of “academic achievement.” Past studies have used individual exam grades, final grades in a single course, semester GPA, year-end GPA, GPA at the time of the study, or variables such as attendance or participation. The present study uses concrete and consequential outcomes: freshman cumulative GPA (fGPA; the measure most commonly employed in previous research) and senior cumulative GPA (sGPA; a final, more comprehensive measure of college success.). 2) Near-exclusive use of self-report personality measures. Reliance on self-reports can be problematic because what one believes of oneself may or may not be an accurate or complete assessment of one’s true strengths and weaknesses. Thus, the present research utilizes ratings of personality provided by the self and by informants. As the personality inventories used in these analyses were administered up to three years prior to the measurement of the dependent variable, finding a meaningful relationship between the two will provide evidence that one’s traits – evaluated by someone else and a number of years in the past! – are consistent enough to serve as useful predictors of a real and important outcome. Within the confines of these parameters, and based upon previous literature, it is hypothesized that: conscientiousness will fail to show the mean differences in ethnicity problematic of SAT scores; both selfand informant-rated conscientiousness will be positively and significantly related to both outcome measures; and finally, conscientiousness will be capable of explaining incremental variance in both outcome measures beyond what is accounted for by the traditional predictors. Method Acquaintance reports 5 Participants This study examined the predictors of academic achievement in a sample of 131 target participants (54.2% female, 45.8% male), who were those among an original sample of 217 undergraduates with sufficiently complete data for the present analyses, as described below. Due to the “minority majority” status of the UCR campus population, the diverse sample included substantial proportions of Asians or Asian Americans (43.5%), Hispanics or Latin Americans (19.8%), Caucasians (16.0%), African Americans (12.9%), and students of other ethnic descent (7.6%). The study also includes 258 informants who described participants with whom they were acquainted. Each target participant and informant was paid $10 per hour. The larger data set was originally designed to explore issues of accuracy in personality judgment. Other analyses, completed and planned (see Letzring, Block & Funder, 2004; Letzring, Wells & Funder, in press; Vazire & Funder, 2006), address different topics and do not overlap with those in the current study. Targets & Informants To deal with missing data, all participants in the larger sample who were lacking any one of the predictor variables (SAT score, HSGPA, or either selfor informant-rated Conscientiousness) were dropped (reducing the N from 217 to 153 at this stage of selection). Among the remaining participants, 21 were missing sGPA (i.e., had not yet graduated at the time the GPAs were collected from the University) but had a junior-level GPA; for these, a regression using junior GPA to predict sGPA was performed (r = 0.96 between the two) and the resulting score was imputed. 22 participants had neither sGPA nor a junior GPA; these last were dropped, leaving the final N = 131 for target participants. Means and standard deviations for both the Acquaintance reports 6 dependent and predictor variables in this smaller sample were comparable to those of the larger group from which they were drawn. Each participant provided contact information for two people who knew him or her best and would be willing to provide information about him or her. 127 participants in our target sample recruited the requested 2 informants, while 4 participants recruited only 1, for a total of 258 informants. Measures Traditional Predictors Participants completed a release form granting access to their academic records; HSGPA and SAT scores were later obtained from the UCR Registrar’s Office. The Registrar provided either an SAT score or an SAT score converted from an American College Testing (ACT) score. Only the total score (rather than the separate verbal/quantitative sub-scores) was used. Personality In order to assess traits at a global level, participants provided self-reports and informants provided peer ratings using the Big Five Inventory (BFI; John, Donahue & Kentle, 1991), which assesses extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience. BFI-scale reliabilities and other psychometric properties have been shown to be similar to those of the much longer scales of Costa and McCrae’s (1990) NEO-FFI (John, et al. 1991). Where two informants were available (all but 4 cases), a composite of their ratings was created by averaging the conscientiousness scale scores. Reliability of the averaged informants’ conscientiousness rating was .59. Academic performance Acquaintance reports 7 Cumulative fGPA and sGPA were collected from the campus Registrar. While the data collection phase of the original, larger project began a few years before the analyses completed for this study and all of the participants had progressed in their academic standing, not all of them had yet completed their senior year. Participants missing GPA data were handled as described above. Results Analyses examined mean differences among ethnic groups and correlations between each of the potential predictors and the two outcome measures. A final set of analyses entered the predictor variables into hierarchical regressions predicting GPA. Descriptive Statistics Mean differences by ethnicity in HSGPA, SAT scores, and BFI scores were examined with one-way ANOVAs (see Table 1). Members of the different groups were admitted to UCR with approximately the same incoming HSGPA (M = 3.51) and very little variation (SD = 0.37), F(4, 126) = 0.68, p = 0.609. There was, however, a significant difference between ethnicities in their entering SAT scores, F(4, 126) = 5.56, p = 3.7 x 10, with Caucasians the highest and African Americans the lowest. As predicted, there were no significant differences in conscientiousness across ethnicities. Correlations There were no significant correlations between gender and any of the variables included in this study. HSGPA and SAT scores – the two traditional predictors – are only modestly related in this sample: r(131) = 0.12, n.s., indicating that they are independently capable of explaining variance in college GPA. sGPA, containing all the variance of fGPA, is thus well correlated with it, r(131) = 0.68, p < .05. Correlations between academic performance and the Acquaintance reports 8 hypothesized predictors of performance (HSGPA, SAT scores, and conscientiousness) are presented in Table 2. While the traditional predictors are correlated with both fGPA and sGPA, conscientiousness shows a relationship with the outcome variables as well. While none of the predictor/freshman outcome relationships are significantly different from the predictor/senior outcome relationships, each of the predictor variables is more highly correlated with sGPA than with fGPA. The true utility of each predictor may be revealed increasingly towards the latter years of college; also, sGPA would be expected to be more reliable than fGPA, because it includes more data. While the relationship between GPA and self-reported conscientiousness is not statistically different from the relationship between GPA and informant-reported conscientiousness, a high self-rating on conscientiousness is less related to academic success than the assessment by one’s acquaintances. In either case, these results support the hypothesis of a positive linear relationship between a student’s conscientiousness and his or her GPA. Regression In order to assess the effectiveness of personality as a predictor of achievement, hierarchical regression analyses were performed on both fGPA and sGPA, with informantreported conscientiousness entered last. Predicting fGPA The incremental utility of conscientiousness is illustrated in Table 3. Both SAT (β = .23, p = .005) and HSGPA (β = .23, p = .006) retain their predictive power at the final step, but conscientiousness (β = .22, p = .007) provides an extra 4.8% of variance, bringing this model’s total percentage of variance explained up to 18.1%.