Personality and Job Performance in Financial Services Managers

PERSONALITY AND JOB PERFORMANCE 91 Personality and Job Performance in Financial Services Managers JesuÂs F. Salgado* and AndreÂs Rumbo This article presents research in which the Five Factor Model of personality was tested as a predictor of job performance. 125 financial services managers who had enrolled in a potential evaluation programme were given the NEO-FFI, a questionnaire designed for measuring the Big Five. Job performance was assessed using nine rating scales and they were grouped into two components: job problem-solving ability and job motivation. Also, one single scale for measuring global job performance was used. The results show that Neuroticism and Conscientiousness correlated with the two components and with the global measure of job performance. Extraversion, Openness and Agreeableness are correlated with one facet or with the global rating of job performance. Taken together, the results suggest that the Five Factor Model is a valid predictor of job performance. The implications of the results for practice and future research are discussed. raditionally, personality has been seen as a variable with low validity for predicting job T performance, and some reviews of criterionrelated validity appear to confirm such a belief (Ghiselli 1973; Guion and Gottier 1965; Schmitt, Gooding, Noe and Kirsch 1984). Two reasons might explain these findings. First, in the area of personality, there have been considerable controversies for a long time (e.g. crosssituational consistency; ideographic vs. nomoethic approach, etc.), and there has never been a universally accepted personality model. Secondly, in the review articles (e.g. Ghiselli 1973; Guion and Gottier 1965; Schmitt et al. 1984), the validity coefficients were integrated across variables obtaining a sole validity coefficient for all personality measures, thereby masking the predictor±criterion construct relationships (Hough, Eaton, Dunnette, Kamp and McCloy 1990). However, in the 1980s the five-factor model (FFM) of personality was consolidated (Digman 1990), and it became a strong paradigm in the area. In effect, the actual research appears to show that only five factors of personality generalize across subjects, observers, variables, factor-analytic algorithms and languages (Borkernau 1992; John 1990). These factors have received different names, but the most used are: Neuroticism or emotional stability (N), Extraversion or surgency (E), Openness or culture (O), Agreeableness (A) and Conscientiousness (C). However, some authors criticized the five-factor model for different motives. For example, Block (1995) criticizes the dependence of the model on the factor ß Blackwell Publishers Ltd 1997, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA. analysis method, Eysenck (1992) sustains that three factors are sufficient and, in the field of I/O Psychology, the most relevant criticisms were made by Hogan (1991) and Hough (1992). For example, Hogan (1991) sustains that, in organizational settings, it could be more convenient to use specific or small factors than variables as complex as the Big Five. Specific factors could lead to a better prediction of job performance. For his part, Hough (1992) affirms that the Big Five are very heterogeneous and incomplete factors, and that additional factors are necessary (e.g. locus of control and achievement motivation). Recently, integrative research on the relation between personality and job performance was conducted using the `Big Five' model as a framework of analysis (see Robertson 1993, 1994). The findings of the research carried out suggest that the personality measures can be predictors of job performance. Five studies using meta-analytic techniques were carried out by Barrick and Mount (1991), Hough et al. (1990), Mount and Barrick (1995), Salgado (1997) and Tett, Jackson and Rothstein (1991). The findings of these quantitative reviews show that the most relevant personality factors which predict job performance are Conscientiousness and Neuroticism, in this order. For example, in a large-scale meta-analysis, Barrick and Mount (1991) found that Conscientiousness (p = 0.22) is a consistently valid predictor for all occupational groups and all criterion types, but Neuroticism, Extraversion, Openness and Agreeableness do not appear to be relevant predictors of job performance, except for specific Volume 5 *Address for correspondence: JesuÂs F. Salgado, Dept. Psicologõ a Social y Ba sica, Universidad de Santiago de Compostela, 15706 Santiago de Compostela, Spain. Number 2 April 1997 INTERNATIONAL JOURNAL OF SELECTION AND ASSESSMENT 92 occupations and criteria. In a partial update of this study, Mount and Barrick (1995) found that the validity for Conscientiousness was underestimated in their prior meta-analysis, and they suggest that a value of 0.31 is closer to its true value. For their part, in a small-scale metaanalysis, Tett et al. (1991) found that Agreeableness is the most relevant, followed by Neuroticism and Openness, and, with lower validity, Extraversion and Conscientiousness. Hough et al. (1990) found that, for job proficiency criteria, adjustment (Emotional Stability) and dependability (Conscientiousness) show an observed (uncorrected) validity of 0.13, and for training criteria the observed validity for the same predictors is 0.16 and 0.11 respectively. A characteristic of these reviews is that they were conducted using only studies carried out in the USA and Canada. Salgado (1997), including only studies conducted in the European Community, found that Neuroticism (p = 0.19) and Conscientiousness (p = 0.25) are valid predictors across jobs and criteria. Therefore, taking these reviews together, it seems that Neuroticism (Emotional Stability) and Conscientiousness generalize validity across jobs, criteria, organizations and countries. More recently, after the meta-analyses by Barrick and Mount (1991), Hough et al. (1990) and Tett et al. (1991) were completed, some single studies were conducted to check the validity of the Five Factor Model. Cortina, Doherty, Schmitt, Kaufman and Smith (1992) evaluated the predictive validity of the Big Five using the Minnesota Multiphasic Personality Inventory (MMPI) and the Inwald Personality Inventory (IPI) (Inwald, Knatz and Shusman 1983) as measures. They found that personality showed a moderate validity across criteria for Conscientiousness, Neuroticism and Agreeableness, but the evidence for Openness and Extraversion was very small. However, those measures did not add significantly to the predictive efficiency of the Civil Service Examination consisting of three tests: reading, accuracy of observation and a written examination. Similarly, Lillibridge and Williams (1992) applied the five-factor model to predict management potential using the Guilford± Zimmerman Temperament Survey (GZTS) (Guilford, Zimmerman and Guilford 1976) and the Edwards Personal Preference Schedule (EPPS) (Edwards 1959). The results of this study showed that E and A were valid predictors for management potential. A third study was conducted by Salgado, Rumbo, Santamarõ a and Losada (1995). In Salgado et al.'s study, the 16 factors of the 16PF were grouped for measuring the `Big Five' and were correlated with several measures or facets of job performance. Neuroticism was significantly correlated with a Volume 5 Number 2 April 1997 global measure of job performance, and Extraversion, Openness and Agreeableness were correlated with some facets of the criterion. For their part, Van der Berg and Feij (1993) applied three personality questionnaires for measuring four factors: Emotional Stability, Extraversion, Sensation Seeking and Achievement Motivation. Their results showed that Emotional Stability and Extraversion were significantly correlated with a measure of self-appraised performance. Taking these studies as a whole, the Big Five appear to be valid predictors of job performance, but the findings are not conclusive. A characteristic of these studies is that measures originally not developed to assess the Big Five were clustered in these factors using conceptual criteria. However, this method of grouping the personality measures (obtained by instruments that were not developed using the Big Five model) into five factors has some problems. For example, several researchers arrived at different clusters for the same questionnaires. In a factor analytical study of the MMPI, Costa, Zonderman, McCrae and William (1985) found that the MMPI only provided measures of four factors, with Conscientiousness excluded from the model. In another study, correlating the MMPI and the Big Five as they are measured by NEO-PI (Costa and McCrae 1985), Costa, Busch, Zonderman and McCrae (1986) found that the MMPI measured Neuroticism, Extraversion, Openness and Agreeableness, but not Conscientiousness. For their part, Johnson, Butcher, Null and Johnson (1984) found four factors but, in their study, the Agreeableness factor was lacking. More recently, Cortina et al. (1992) found that MMPI assessed Neuroticism, Extraversion, Agreeableness and Conscientiousness but not Openness. Moreover, the correlation between the MMPI measure for the Openness and the IPI measure for the same factor was 0.01, and the correlation between the MMPI and the IPI for Agreeableness was 0.13. These findings showed a lack of convergent validity for these measures. Cortina et al. (1992) suggested that Openness may not be represented in either the IPI or the MMPI, or in one but not the other. The same explanation may be extended to Agreeableness. Also, in the Lillibridge and Williams (1992) study, there are problems with regard to their method of measuring the Big Five. These authors assessed four of the five factors, with Neuroticism absent. For their part, Van der Berg and Feij (1993) failed to obtain a measure of Agreeableness. The lack of convergent validity and the fact that these instruments were not developed into the frame of the Big Five model could explain the inconsistencies in the findings previously reported. However, at present there are several ß Blackwell Publishers Ltd 1997 PERSONALITY AND JOB PERFORMANCE questionnaires that were developed to assess the Big Five. The first inventories using the Big Five as the model of personality were the Hogan Personality Inventory (HPI) (Hogan 1982, 1986) and the NEO-PI (Costa and McCrae 1985, 1992). Recently, Barrick and Mount (1993), and Barrick, Mount and Strauss (1993), assessed the `Big Five' using the PCI, an inventory for the comprehensive description of the five personality constructs. Currently, there are some other `Big Five' questionnaires. For example, recent instruments were developed by Bartram (1993), Caprara, Barbaranelli and Borgogni (1994) and Salgado (1994), and the number is growing. Several studies have been conducted using these questionnaires to examine their predictive validity. For example, Hogan and colleagues have developed some indices to use in organizational environments, such as an index for Service Orientation, or for Reliability, Stress Tolerance, Management Potential, Sales Potential and Clerical Potential (Hogan 1991; Hogan and Hogan 1989; Hogan, Hogan and Busch 1984). For their part, Barrick and Mount (1993) found that Conscientiousness and Extraversion were related to job performance in a managerial sample, although the level of job autonomy is a moderator of the validity. In another study, Barrick et al. (1993) found that Conscientiousness is a valid predictor of job performance in sales representatives. Also, their results show that autonomous goal setting and goal commitment mediate the relationship between Conscientiousness and two measures of job proficiency. However, a characteristic of these last studies is that they are not focused on testing the Big Five model as a whole. Therefore, part of the validity research with the five-factor model has been carried out to test specific aspects (e.g. Barrick and Mount 1993; Barrick et al. 1993; Hogan and Hogan 1989; Hogan et al. 1984), while the rest of the studies have been carried out with questionnaires based on other personality models than the five-factor model (e.g. Cortina et al. 1992; Lillibridge and Williams 1992; Salgado et al. 1995; Van der Berg and Feij 1993). Thus, it appears necessary to conduct studies for checking the model as a whole with questionnaires based on the five-factor model. This research has as its primary goal the testing of the validity of the Big Five for predicting job performance in financial services managers, and we will use a questionnaire specifically designed to assess the Big Five: NEO-FFI (Costa and McCrae 1992). Based on the results of the meta-analyses of the personality validity, we hypothesized that the Big Five will be valid predictors of job performance. More specifically, we stated the ß Blackwell Publishers Ltd 1997 93 following predictions: (a) Neuroticism will be negatively correlated with job performance; (b) Conscientiousness will be positively correlated with job performance; (c) Conscientiousness will be the dimension that will show higher validity; and (d) a Big Five composite will show higher validity than any single personality dimension. Method Sample The subjects were 125 middle managers from a Spanish financial services organization (a savings and loan institution) with around 2700 employees. All the subjects were males. Their ages ranged from 25 to 57 years. Middle managers carry out the following functions as their main duties: providing financial services to customers, directing and coordinating a group of employees, assisting in cash management activities, examining documents prepared by subordinates and ensuring that the security procedures are followed. They examine, evaluate and process loan applications, and prepare, type and maintain records of financial transactions. The middle manager is in charge of the office when the Director is absent. Predictors The `Big Five' are the predictors used in this research, and the NEO-FFI (Costa and McCrae 1992) is the tool used to measure them. This questionnaire has 60 items that assess Neuroticism, Extraversion, Openness, Agreeableness and Conscientiousness. Each factor is measured by 12 items. For this research, a Spanish translation was carried out and the process of adaptation was as follows: first, the senior researcher translated the NEO-FFI to Spanish; once the translation was complete, a backtranslation was conducted by a bilingual person that was unfamiliar with the English version of the NEO-FFI; then the two versions were forwarded to the authors and PAR Inc. for review and suggestions for further revision. When the Spanish version was accepted by the authors and Par Inc., data collection commenced. In the present sample, the reliability (internal consistency) for N, E, O, A and C was 0.76, 0.72, 0.58, 0.58 and 0.74, respectively. Criterion measures The criterion measures in this study were nine rating scales that assessed the competency of the individuals in nine characteristics of job performance. The scales had five points: deficient, insufficient, sufficient, notable and excellent. The characteristics assessed were: knowledge, effi- Volume 5 Number 2 April 1997 INTERNATIONAL JOURNAL OF SELECTION AND ASSESSMENT 94 ciency, problem comprehension, adaptability to the job, leadership, ability for relations, aspiration level, initiative and attitude. A nine-point scale was added to this set of scales in order to assess global job performance. In this study, four supervisors served as raters, but each subject was assessed by only one judge who rated him/her on the scales. The interrater reliability cannot, therefore, be estimated. The average test±retest reliability, computed from two measures made two months apart, is 0.58. Although the interrater reliability for each single scale was not computed, it is known that it is very low. Hunter and Hirsh (1987), and King, Hunter and Schmidt (1980), have shown that for a single rating on a particular trait, the interrater reliability is 0.31, and that the reliability of summated ratings by a single supervisor is 0.43. Nothing suggests that the reliability is greater in this study. Furthermore, in the present study the correlations between the single scales were high, suggesting the presence of a Halo effect, and that they are not discrete and separable facets of the criterion. These two reasons suggest that a good solution would be to factor analyze the correlation matrix of the criterion facets in order to decide what dimensions of work performance are to be employed and to improve the reliability of the criteria. The criterion ratings were factorized using component principal analysis and Varimax rotation (an oblique solution was also tried, but the results were essentially the same as the results found in the orthogonal solution). In fact, the correlation between factors was 0.14, and this value indicates an angle of 82º. Therefore, this result is very close to the orthogonality. We used three criteria to decide the number of factors to retain: Parallel Analysis (Horn 1965), eigenvalues greater than one (Kaiser 1960) and Table 1: Rotated Factor Loadings of Criterion Measures Knowledge Efficiency Problem comprehension Adaptability to job Leadership Ability for relations Aspiration level Initiative Attitude VP Factor 1 Factor 2 0.797 0.758 ÿ0.167 0.227 0.884 0.700 0.700 0.554 0.570 0.700 ÿ0.088 4.102 ÿ0.020 0.253 0.072 0.251 0.646 0.432 ÿ0.922 1.665 Note: VP = Variance explained by the factor. The VP is computed as the sum of squares for the element of the factor's column in the factor loading matrix. Volume 5 Number 2 April 1997 the Scree Test (Cattell 1966). The three criteria agreed in that two components should be rotated (see Table 1). The eigenvalues for the nine components in the unrotated solution were 4.45, 1.32, 0.88, 0.62, 0.44, 0.41, 0.35, 0.29 and 0.25 respectively. The two retained components accounted for 64.07% of variance. To interpret the results of the factor analysis we took into account only weights greater than 0.50, and we assumed that the weight of each variable is only in the factor with the higher loading. In Table 1 it can be seen that the first component has the following facets of job performance: knowledge, efficiency, problem comprehension, adaptability to the job, leadership, ability for relations and initiative. We named this factor as a job problem-solving ability. The second factor is composed of aspiration level and attitude, and they appear to represent job motivation. Therefore, job performance may be correctly represented by three scores: (a) global job performance, (b) job problem-solving ability and (c) job motivation. The internal consistency for the first composite is 0.87, and for the second is 0.58. Procedure The subjects of this study were participants in a process of potential evaluation conducted by two researchers of the company. Therefore, the performance ratings were collected for administrative purposes. The raters were given a onesession training course in which the motives of the programme were explained, as were the system characteristics and the appraisal procedure. Over the following days, the supervisors rated the subjects. When they finished the evaluation of the subjects, they had an individual session with those responsible for the process. In this session, all the ratings were reviewed in order to reduce errors. The NEO-FFI was answered collectively in small groups. The researchers of the company conducted this process on site. Results The validity of the NEO-FFI scales for predicting job performance was appropriately estimated by correlations. In Table 2, the intercorrelations between the NEO-FFI and the single facets of the criteria are shown. The correlations between the NEO-FFI and Global Job Performance (GJP), and the two dimensions of work performance, are also shown. Regarding the single criterion measures, in general the direction of the correlations was as expected. Neuroticism shows negative correlations with all criterion measures and Conscientiousness shows positive correla- ß Blackwell Publishers Ltd 1997 PERSONALITY AND JOB PERFORMANCE 95 Table 2: Correlations Between the NEO-FFI Scales and the Criterion Measures N E O A C K EF PC AJ LI AR AL IN AT JPA JM GJP Na Eb Ð 54 02 ÿ34 ÿ38 ÿ13 ÿ06 ÿ14 ÿ13 ÿ08 ÿ15 ÿ30** ÿ12 00 ÿ16* ÿ22* ÿ23** Ð 00 39 47 ÿ01 ÿ15 ÿ06 ÿ02 ÿ06 22** 16 ÿ02 07 ÿ02 15 13 Ob Ð 03 ÿ14 ÿ15 ÿ19* ÿ19* ÿ20* ÿ06 ÿ15 ÿ15 ÿ09 14 ÿ20* ÿ04 ÿ11 Ab Ð ÿ16 05 10 ÿ03 ÿ23** ÿ01 ÿ22* ÿ12 ÿ03 ÿ18* ÿ07 ÿ17 ÿ03 Ca Ð 08 19* 14 07 09 19* 25** 14 19* 17* 27** 32** K Ð 55 63 45 48 24 39 45 ÿ11 72 23 43 EF Ð 62 47 51 33 57 57 15 79 48 65 PC Ð 58 54 53 47 57 ÿ06 84 31 52 AJ Ð 33 54 46 59 13 76 39 44 LI Ð 34 44 47 06 69 34 41 AR Ð 36 37 15 63 32 46 AL Ð 65 45 65 91 62 IN Ð 25 78 58 56 AT Ð 12 77 37 JPA Ð 52 67 JM Ð 61 GJP  X SD Ð 17.1 32.1 24.1 32.1 36.1 4.9 4.8 4.9 4.7 4.7 4.8 4.5 4.3 5.4 33.2 9.8 6.3 7.3 6.1 5.8 5.3 5.4 0.8 1.0 0.8 0.9 0.8 0.9 1.1 1.0 0.7 4.7 1.6 1.0 Note: N = Neuroticism; E = Extraversion; O = Openness to Experience; A = Agreeableness; C = Conscientiousness; K = Knowledge; EF = Efficiency; PC = Problem Comprehension; AJ = Adaptability to Job; LI = Leadership; AR = Ability for Relations; AL = Aspiration Level; IN = Initiative; AT = Attitude; JPA = Job Problem-Solving Ability; JM = Job Motivation; GJP = Global Job Performance; a = one-tailed test; b = two-tailed test; *p < 0.05; **p < 0.01. tions with the same measures. For their part, Extraversion, Openness and Agreeableness are positively correlated with some facets and negatively with the rest. Furthermore, Conscientiousness shows the highest correlations of the five personality factors. However, taking into account the low level of reliability of the single criterion measures, it is preferable to centre the analysis on the criterion dimensions and on the Global Job Performance measure. The first dimension of the criterion, the Job Problem-Solving Ability (JPA), is significantly correlated with Neuroticism, Openness and Conscientiousness. Job Motivation (JM), the second criterion dimension, is significantly correlated with Neuroticism and Conscientiousness. For its part, Global Job Performance is significantly correlated with Neuroticism and Conscientiousness. Therefore, Neuroticism and Conscientiousness are the only factors that show significant correlations with these three criterion measures. The average uncorrected validity is 0.20, 0.09, 0.12, 0.09 and 0.25 for Neuroticism, Extraversion, Openness, Agreeableness and Conscientiousness, respectively. Consequently, regarding the size of the coefficients, both Neuroticism and Conscientiousness show acceptable validity for use in personnel assessment. It is necessary to take into account that all reported validity coefficients are affected by measurement errors in predictor and criterion. As we are interested in the true validity between the Big Five and job performance, those validity coefficients should be corrected for measurement errors in the criterion in order to obtain an unbiased estimation of validity (Society for Industrial and Organizational Psychology (SIOP) (1987). ß Blackwell Publishers Ltd 1997 The correction of validity for the measurement errors may be made using the appropriate coefficients. For the criterion, the appropriate reliability estimate should be the interrater agreement (Guilford and Fruchter 1979; Schmidt and Hunter 1995) or the temporal stability coefficient (Schmidt, Hunter and Urry 1976). We cannot compute any estimate of the interrater reliability because each subject was rated by a single rater. Alternatively, we know the test±retest reliability for each supervisor and the average test±retest reliability. The average test±retest reliability is 0.58, and this value is close to the value of 0.52 found by Rothstein (1990) for the interrater agreement when a single rater assesses the subjects. Barrick and Mount (1991) use Rothstein's value for the supervisory rating criterion and Tett et al. (1991) use a value of 0.50 for the same criterion. For his part, Salgado (1997) used a value of 0.62, and the same author found a mean value of 0.575 for the supervisory rating in the studies conducted in Spain (Salgado 1995). Here we will use 0.58 as an estimate of interrater reliability for correcting validity coefficients. The downward bias in the interrater reliability made by this estimate, if it is possible, would be very small in all cases. The corrected validity coefficients among the five factors, and the criterion dimensions and Global Job Performance, are shown in Table 3. When the validity coefficients are corrected for measurement errors, all personality factors that appear to be valid in the prior analysis now show coefficients with a medium effect size, according to Cohen's (1988) rule that coefficients around 0.30 are medium size. Again, Neuroticism and Conscientiousness have the highest validity coefficients, Openness is a valid Volume 5 Number 2 April 1997 INTERNATIONAL JOURNAL OF SELECTION AND ASSESSMENT 96 Table 3: Validity of the NEO-FFI Scales Corrected for Measurement Error in Criterion JPA JM 90%CI rc Neuroticisma Extraversionb Opennessb Agreeablenessb Conscientiousnessa ÿ0.21 ÿ0.03 ÿ0.26 ÿ0.09 0.22 0.034 0.823 0.021 0.433 0.026 ÿ0.40 to ÿ0.02 ÿ0.22 to 0.16 ÿ0.45 to ÿ0.08 ÿ0.28 to 0.10 0.41 to 0.04 GJP rc ÿ0.29 0.20 ÿ0.05 ÿ0.22 0.35 90%CI 0.005 0.088 0.655 0.052 0.000 ÿ0.47 to ÿ0.11 0.28 to 0.01 0.25 to ÿ0.14 ÿ0.41 to 0.04 0.53 to 0.18 Note: JPA = Job Problem-Solving Ability; JM = Job Motivation; GJP = Global Job Performance; measurement error in criterion; = probability of rc; 90%CI = 90% Confidence Interval. predictor of Job Problem-Solving Ability (p = 0.021; two-tailed test) and Agreeableness is a valid predictor of Job Motivation (p = 0.052; two-tailed test). In connection with the hypotheses previously stated, these results confirm the first three hypotheses. In effect, Neuroticism presents a negative correlation with job performance and Conscientiousness shows positive correlations. Furthermore, Conscientiousness has the highest validity coefficients. No hypotheses were stated for Openness and Agreeableness, and both dimensions correlated negatively with the criterion components. However, a characteristic of this study is that it is a small-sample one. This type of study has a problem related to sampling error. The smaller the sample, the larger the sampling error. This error may affect the significance tests in both directions, accepting or rejecting the null hypothesis, and significance tests cannot control for this. Hunter and Schmidt (1990, p. 31) suggest using confidence intervals as an alternative to significance tests at the level of single studies. Confidence intervals are generated by the standard error of the mean effect size and they reflect the extent to which sampling error remains in the estimate of effect size (Whitener 1990). Because we use the correlation as an estimate of the validity, we must use the standard error of a correlation to generate the confidence interval around the validity found for each personality dimension. The 90% confidence intervals are shown in Table 3. If the results of the corrected validity are interpreted with the confidence intervals, the conclusions are very similar to when significance tests were used. We made three confidence intervals for each personality dimension, one interval for each criterion measure, although we only comment on the results for Neuroticism and Conscientiousness. In the case of Neuroticism, the lower values of 90% confidence interval are: JPA = ÿ0.02; JM = ÿ0.11 and GJP = ÿ0.12. For Conscientiousness, the lower values are: JPA = 0.04, JM = 0.18 and GJP = Volume 5 Number 2 April 1997 rc a = one-tailed test; ÿ0.30 0.17 ÿ0.14 ÿ0.04 0.42 b 90%CI 0.004 0.141 0.108 0.738 0.000 ÿ0.48 to ÿ0.12 0.36 to ÿ0.02 0.33 to ÿ0.05 ÿ0.23 to 0.15 0.59 to 0.25 = two-tailed test; rc = validity corrected for 0.25. Therefore, for Neuroticism and Conscientiousness, none of the 90% confidence intervals include 0, and therefore it is not a reasonable possibility that the validity is 0 for these two personality dimensions. On the contrary, the confidence intervals suggest that the corrected validity is significantly different from zero (Whitener 1990). The last hypothesis stated in the introduction is to check the FFM as a whole, and one method is to compute the multiple correlation between the Big Five as a composite and the criteria. As can be seen from Table 2, the Big Five are intercorrelated but, from a theoretical point of view, they are orthogonal personality factors. Could these correlations change the above conclusions in respect to Neuroticism and Conscientiousness? Could these correlations be explained by the presence of a single common factor in the predictor variables? The correlations between the personality dimensions are not a strange finding. For example, Costa and McCrae (1992, p. 100) showed that the Big Five, as they are assessed by the NEO-PI-R, are correlated (e.g. Neuroticism±Conscientiousness, r = ÿ0.53; Extraversion±Openness, r = 0.40). However, these correlations are using raw data. If the correlations are estimated at the factor level, the Big Five are orthogonal. According to Costa and McCrae (1995), these correlations mean that the NEO-PI-R scores are not perfect measures of the FFM. In the same article, Costa and McCrae indicate that the raw scale scores of the NEO-FFI also show modest intercorrelations in most samples, but `The item factor structure of the NEO-FFI has been confirmed in both Canadian . . . and German . . . samples', and `although these brief scales are not completely uncorrelated, they appear to give useful approximations to five orthogonal factors' (Costa and McCrae 1995, p. 218). Obviously, if the Big Five assessed by the NEO-FFI are independent factors, and they contribute independently to explain the criteria, the multiple correlation between the five factors and the criteria must be remarkably higher than the correlation between one single dimension ß Blackwell Publishers Ltd 1997 PERSONALITY AND JOB PERFORMANCE 97 Table 4: Multiple Regression Analyses using Big Five as Independent Variables and the Criteria Measure as Dependent Variables Criterion R RA R2 90%CI Rc JPA JM GJP 0.51 0.41 0.45 0.48 0.38 0.41 0.26 0.17 0.20 0.33 to 0.63 0.23 to 0.53 0.26 to 0.56 0.63 0.50 0.54 Note: JPA = Job Problem-Solving Ability; JM = Job Motivation; GJP = Global Job Performance; R = multiple correlation coefficient; RA = adjusted multiple correlation; 90%CI = 90% Confidence Interval using RA as centre of the interval; Rc = adjusted multiple correlation coefficient corrected by criterion reliability. (e.g. Neuroticism or Conscientiousness) and the criteria. However, if the intercorrelations are due to a single common factor in the predictor variables, then a personality composite will show a validity of similar size or only slightly higher than the single factors. In Table 4, it is possible to see the results of three multiple regression analyses using Job Problem-Solving Ability, Job Motivation and Global Job Performance as dependent variables, respectively. The results of the multiple regression analyses clearly show that a personality composite has greater validity than any one dimension considered individually. Furthermore, 90% confidence intervals indicate that the composite validity is significantly different from zero. Therefore, (a) the intercorrelations do not suggest one single common factor and (b) using the five factors together in a composite, the validity is remarkably high, even though the multiple correlation was corrected for shrinkage. The size of the validity is similar to the validity found for ability composites (Hunter and Hunter 1984). These results confirm our fourth hypothesis, according to which the Big Five composite would have higher validity than the single personality dimensions. Thus, the present test of the FFM to predict job performance suggests that the Big Five may be reasonably included in a personnel selection battery. Discussion The objective of this study was to test the Five Factor Model of personality as a predictor of job performance. Based on previous research using the Big Five, we thought that Neuroticism would be negatively correlated with performance criterion measures and that Conscientiousness would be positively correlated. Also, we hypothesized that Conscientiousness would show the highest validity coefficient. With respect to the major goals of this research, the findings confirm the stated hypotheses. A close examination of the results shows that they are consistent with the findings ß Blackwell Publishers Ltd 1997 of previous research conducted with the Big Five. For example, Barrick and Mount (1991) found validities (corrected for unreliability and range restriction) ranging from 0.04 for Openness to 0.22 for Conscientiousness. Based on different procedures than those of Barrick and Mount (1991), Tett et al. (1991) found that the `Big Five' are valid predictors and that the validities ranged from 0.16 for Extraversion to 0.33 for Agreeableness. In their meta-analysis, Hough et al. (1990) found uncorrected validities of 0.13 and 0.16 for Neuroticism and Conscientiousness, and Salgado (1997) found validities ranging from 0.02 for Agreeableness to 0.25 for Conscientiousness. For their part, Cortina et al. (1992) found that Neuroticism, Agreeableness and Conscientiousness, as measured by the IPI, are correlated significantly with supervisor ratings for police recruits. In the Cortina et al. study, neither Openness nor Extraversion resulted in valid predictors. Similarly, Van der Berg and Feij (1993) found that Emotional Stability (r = 0.21) and Extraversion (r = 0.20) correlated significantly with self-appraised performance. Also, Salgado et al. (1995) found that Neuroticism showed a significant correlation with a measure of Global Job Performance, although Extraversion, Openness and Agreeableness are also correlated with some facets of the criterion. The sizes of the validities found here are similar to those of Cortina et al. (1992), Salgado et al. (1995), Tett et al. (1991) and Van der Berg and Feij (1993), and slightly higher than those reported by Barrick and Mount (1991), Hough et al. (1990) and Salgado (1997). Specifically, in the actual research, Neuroticism shows a validity with a high coefficient, if it is compared to the validity found by Barrick and Mount (1991), the most extensive Big Five meta-analysis to date. Although one single study is not sufficient evidence to contradict the Barrick and Mount conclusion (based on 117 single studies) that Conscientiousness is the only factor that generalizes validity across persons, jobs and situations, the actual findings, along with those meta-analyses by Hough et al. (1990) and Salgado Volume 5 Number 2 April 1997 INTERNATIONAL JOURNAL OF SELECTION AND ASSESSMENT 98 (1997), and the single studies of Salgado et al. (1995) and Van der Berg and Feij (1993), suggest an alternative explanation. It may be possible that there are cross-cultural differences in the relevance attributed to Neuroticism in order to perform a job acceptably. Perhaps Neuroticism is viewed differently in Europe than in America, and in this way Neuroticism would have a more local or less generalizable validity in the USA than in Europe. In other words, the Big Five personality factors would show a different cross-culturally generalizable validity, Conscientiousness being more generalizable and Neuroticism less generalizable in (more local) validity. However, this hypothesis cannot be checked in this study. In connection with the fourth hypothesis, the findings suggest that the Big Five may be used as a composite predictor in a similar way to the ability composite. Used in this way, the Big Five shows an impressive validity ranging from 0.50 to 0.63. Any single personality predictor reaches a similar validity size. These findings are also proof of the independence of the Big Five, although they are not completely uncorrelated. Another aspect of this study concerns the reporting of the results. The most usual form is to report the probability level of the observed (uncorrected) validities and to describe the corrected validities as the best point estimation of the validity. We report the probabilities of the corrected validities using the standard error estimates, applying the formulas provided by Bobko and Rieck (1980). Thus, it is possible to apply a significance test to the hypotheses concerning the corrected validities as well as the uncorrected validities. Also, we reported the confidence intervals of the corrected validities because a growing number of authors suggest that significance tests cannot appropriately control the (Type II) error (Cohen 1994; Hunter and Schmidt 1990). By reporting the significance tests along with the confidence intervals, the reader has a better picture of the results of the study. In the future, we suggest using both the significance test and the confidence intervals in reporting the findings of single validity studies. The findings of this research have certain implications for the practice of personnel selection. First, they suggest that personality measures are valid predictors of job performance, and therefore should be used in personnel selection. Secondly, the results also suggest that the five-factor model appears to be a good model for predicting job behaviour and that assessment instruments could be developed with this model. In addition, a few words about the questionnaire NEO-FFI used in this research appear necessary in the light of the recent research by Schmit and Ryan (1993). The inventory is a brief version of the NEO-PI, Volume 5 Number 2 April 1997 which is the model suggested by Matarazzo (1992) for the personality inventories for the 21st century. The findings of the present research show that the NEO-FFI is an acceptable instrument for predicting job performance in financial services managers. However, according to the results of Schmit and Ryan (1993), the assessment situation (anonymous vs. nonanonymous as in personnel selection) dramatically affects the structure of the NEOFFI by changing it, and the change in structure may result in low validity for the NEO-FFI scales. However, the present results contradict Schmit and Ryan's (1993) suggestion, and they confirm that the NEO-FFI is a valid procedure for personnel selection. Finally, we should comment on the criterion measures used here. In this research we used a criterion being collected for administrative purposes, and this can result in a higher `Halo effect' and higher extremism in the ratings than criteria obtained for research purposes (Guion 1965; Veres, Field and Boyles 1983; Warmke and Billings 1979). Furthermore McDaniel, Whetzel, Schmidt and Maurer (1994) found that the validity coefficients based on research criteria are generally greater than the validity coefficients based on administrative criteria. Therefore, the administrative measures would affect the validity coefficient found here, downwardly biasing their sizes. In summary, it may be said that in contradiction to the pessimism of the classical reviews of personality measure validities, the recent research shows opposite evidence. Personality measures can be valid predictors of job performance (Day and Silverman 1989; Robertson and Kinder 1993; Salgado 1996a; Salgado et al. 1995; Van der Berg and Feij 1993) and the `Big Five' is a relevant model to use in personnel selection. Three or four factors are significantly associated with job performance. Also, this research shows that the NEO-FFI is a brief measure of these, with acceptable reliability and criterion validity. We suggest using the NEO-FFI when a rapid measure of personality is needed. Acknowledgement This research was supported by the XUGA Grant 21104A95 from the Xunta de Galicia (Spain) to the first author. References Barrick, M.R. and Mount, M.K. (1991) The Big Five personality dimensions and job performance: a meta-analysis. Personnel Psychology, 44, 1±26. ß Blackwell Publishers Ltd 1997 PERSONALITY AND JOB PERFORMANCE Barrick, M.R. and Mount, M.K. (1993) Autonomy as a moderator of the relationships between the Big Five personality dimensions and job performance. Journal of Applied Psychology, 78, 111±118. Barrick, M.R., Mount, M.K. and Strauss, J.P. (1993) Conscientiousness and performance of sales representatives: test of the mediating effects of goal setting. Journal of Applied Psychology, 78, 715±722. Bartram, D. (1993) Validation of the `ICES' personality inventory. European Review of Applied Psychology, 41, 207±218. Block, J. (1995) A contrarian view of the five-factor approach to personality description. Psychological Bulletin, 177, 187±213. Bobko, P. and Rieck, A. (1980) Large sample estimators for standard errors of functions of correlation coefficients. Applied Psychological Measurement, 4, 385±398. Borkenau, P. (1992) Implicit personality and the fivefactor model. Journal of Personality, 60, 295±327. Caprara, G.V., Barbaranelli, C. and Borgogni, L. (1994) The Big Five Questionnaire. Firenze (Italy): OS Organizzazione Speciali. Cattell, R.B. (1966) The scree test for the number of factors. Multivariate Behavioral Research, 1, 245± 276. Cohen, J. (1988) Statistical power analysis for the behavioral sciences. 2nd edition. Hillsdale, NJ: Erlbaum. Cohen, J. (1994) The earth is round (p < 0.05). American Psychologist, 49, 997±1003. Cortina, J.M., Doherty, M.L., Schmitt, N., Kaufman, G. and Smith, R.G. (1992) The `Big Five' personality factors in the IPI and MMPI: predictors of police performance. Personnel Psychology, 45, 119±140. Costa, P.T. and McCrae, R.R. (1985) The NEO-PI Personality Inventory. Odessa, FL: Psychological Assessment Resources. Costa, P.T. and McCrae, R.R. (1992) The NEO-PI Personality Inventory. Odessa, FL: Psychological Assessment Resources. Costa, P.T. and McCrae, R.R. (1995) Solid ground in the wetlands of personality: a reply to Block. Psychological Bulletin, 117, 216±220. Costa, P.T., Zonderman, A.B., McCrae, R.R. and William, R.B. (1985) Content and comprehensiveness in the MMPI: an item factor analysis in a normal adult sample. Journal of Personality and Social Psychology, 48, 925±933. Costa, P.T., Busch, C.M., Zonderman, A.B. and McCrae, R.R. (1986) Correlations of MMPI factor scales with measures of the five-factor model of personality. Journal of Personality Assessment, 50, 640±650. Day, D.V. and Silverman, S.B. (1989) Personality and job performance: evidence of incremental validity. Personnel Psychology, 42, 25±36. Digman, J.M. (1990) Personality structure: emergence of the five-factor model. Annual Review of Psychology, 41, 417±420. Edwards, A.L. (1959) Edwards Personality Preference Schedule. New York: Psychological Corporation. Eysenck, H.J. (1992) Four ways five factors are not basic. Personality and Individual Differences, 6, 667± 673. ß Blackwell Publishers Ltd 1997 99 Ghiselli, E.E. (1973) The validity of aptitude tests in personnel selection. Personnel Psychology, 26, 461± 477. Guilford, J.P. and Fruchter, B. (1979) Fundamental statistics in psychology and education. New York: McGraw-Hill. Guilford, J.S., Zimmerman, W.S. and Guilford, J.P. (1976) The Guilford±Zimmerman Temperament Survey Handbook: twenty-five years of research and applications. San Diego: EdITS Publishers. Guion, R.M. (1965) Personnel Testing. New York: McGraw-Hill. Guion, R.M. and Gottier, R.F. (1965) Validity of personality measures in personnel selection. Personnel Psychology, 18, 135±164. Hogan, R. (1982) A socioanalytic theory of personality. In M. Page (ed.), Personality: current theory and research. Nebraska Symposium on Motivation. Lincoln, Nebraska: Nebraska University Press. Hogan, R. (1986) Manual for the Hogan Personality Inventory. Minneapolis, MN: National Computer Systems. Hogan, R. (1991) Personality and personality measurement. In M.D. Dunnette and L.H. Hough (eds.), Handbook of Industrial and Organizational Psychology, Vol. II. Palo Alto, CA: Consulting Psychologists Press. Hogan, J. and Hogan, R. (1989) How to measure employee reliability. Journal of Applied Psychology, 74, 273±279. Hogan, R., Hogan J. and Busch, C. (1984) How to measure service orientation. Journal of Applied Psychology, 69, 157±163. Horn, J.L. (1965) A rationale and test for the number of factor. Psychometrika, 30, 179±185. Hough, L.M. (1992) The `Big Five' personality variables-construct confusion: description versus prediction. Human Performance, 5, 139±155. Hough, L.M., Eaton, N.K., Dunnette, M.D., Kamp, J.D. and McCloy, R.A. (1990) Criterion-related validities of personality constructs and the effect of response distortion on those validities. Journal of Applied Psychology, 75, 581±585. Hunter, J.E. and Hunter, R.F. (1984) Validity and utility of alternate predictors of job performance. Psychological Bulletin, 96, 72±98. Hunter, J.E. and Hirsh, H.R. (1987) Applications of meta-analysis. In C.L. Cooper and I.T. Robertson (eds.), International Review of Industrial and Organizational Psychology. Chichester, UK: Wiley. Hunter, J.E. and Schmidt, F.L. (1990) Methods of metaanalysis. Newbury Park, CA: Sage. Inwald, E.E., Knatz, H.F. and Shusman, E.J. (1983) Inwald Personality Inventory Technical Manual. New York: Hilson Research. John, O.P. (1990) The `Big Five' factor taxonomy: dimensions of personality in the natural language and in questionnaires. In L. Pervin (ed.), Handbook of Personality: theory and research. New York: Guilford. Johnson, J.H., Butcher, J.N., Null, C. and Johnson, K.N. (1984) Replicated item level factor analysis of the full MMPI. Journal of Personality and Social Psychology, 49, 105±114. Kaiser, H.F. (1960) The applications of electronic computers to factor analysis. Educational and Psychological Measurement, 20, 141±151. Volume 5 Number 2 April 1997 INTERNATIONAL JOURNAL OF SELECTION AND ASSESSMENT 100 King, L.M., Hunter, J.E. and Schmidt, F.L. (1980) Halo in a multidimensional forced-choice performance evaluation scale. Journal of Applied Psychology, 65, 507±516. Lillibridge, J.R. and Williams, K.J. (1992) Another look at personality and managerial potential: application of the five-factor model. In K. Kelley (ed.), Issues, Theory and Research in Industrial Organizational Psychology. Amsterdam: NorthHolland. Matarazzo, J.D. (1992) Psychological testing and assessment in the 21st century. American Psychologist, 47, 1007±1018. McDaniel, M.A., Whetzel, D.L., Schmidt, F.L. and Maurer, S.D. (1994) The validity of employment interviews: a comprehensive review and metaanalysis. Journal of Applied Psychology, 79, 599± 616. Mount, M.K. and Barrick, M.R. (1995) The Big Five personality dimensions: implications for research and practice in human resources management. In K.M. Rowland and G. Ferris (eds.), Research in personnel and human resources management, 13, pp. 153±200. Greenwich: CT: JAI Press. Robertson, I.T. (1993) Personality assessment and personnel selection. European Review of Applied Psychology, 43, 187±194. Robertson, I.T. (1994) Personality and personnel selection. In C.L. Cooper and D.M. Rousseau (eds.), Trends in Organizational Behavior, 1. London: Wiley. Robertson, I.T. and Kinder, A. (1993) Personality and job competences: the criterion-related validity of some personality variables. Journal of Occupational and Organizational Psychology, 66, 225±244. Rothstein, H.R. (1990) Interrater reliability of job performance ratings: growth to asymptote level with increasing opportunity to observe. Journal of Applied Psychology, 75, 322±327. Salgado, J.F. (1994) Manual Tecnico del IP/5F. Departamento de Psicologõ a, Universidad de Santiago de Compostela (Spain). Manuscrito no publicado. [Technical Manual for IP/5F. Dept. of Psychology, University of Santiago de Compostela, Spain. Unpublished manuscript.] Salgado, J.F. (1995) Fiabilidad intraevaluador e interevaluador de la valoraciones de rendimiento en el trabajo. Unpublished manuscript. Salgado, J.F. (1996) Personality and job competences: a comment on Robertson and Kinder's (1993) study. Journal of Occupational and Organizational Volume 5 Number 2 View publication stats April 1997 Psychology, 69, 373±375. Salgado, J.F. (1997) The five-factor model of personality and job performance in the European Community (EC). Journal of Applied Psychology 82, 30±43. Salgado, J.F., Rumbo, A., Santamarõ a, G. and Losada, M.R. (1995) El 16PF, el modelo de personalidad de cinco factores y el rendimiento en el trabajo. Revista de PsicologõÂa Social Aplicada, 5, 81±94. [The 16PF, five-factor model of personality and job performance.] Schmidt, F.L. and Hunter, J.E. (1996) The measurement errors in the applied research. Answers to 27 scenarios. Psychological Methods, 1, 199±223. Schmidt, F.L., Hunter, J.E. and Urry, V.W. (1976) Statistical power in criterion-related validation studies. Journal of Applied Psychology, 61, 473±485. Schmit, M.J. and Ryan, A.M. (1993) The Big Five in personnel selection: factor structure in applicant and nonapplicant populations. Journal of Applied Psychology, 78, 966±974. Schmitt, N., Gooding, R.Z., Noe, R.D. and Kirsch, M. (1984) Meta-analyses of validity studies published between 1964 and 1982, and the investigation of study characteristics. Personnel Psychology, 37, 407±422. Society for Industrial and Organizational Psychology, Inc. (1987) Principles for the validation and use of personnel selection procedures. 3rd edition. College Park, MD: Author. Tett, R.P., Jackson, D.N. and Rothstein, M. (1991) Personality measures as predictors of job performance: A meta-analytic review. Personnel Psychology, 44, 703±742. Van der Berg, P.T. and Feij, J.A. (1993) Personality traits and job characteristics as predictors of job experiences. European Journal of Psychology, 7, 337±357. Veres, J.G., Field, H.S., and Boyles, W.R. (1983) Administrative versus research performance ratings: An empirical test of rating data quality. Public Personnel Management, 12, 290±298. Warmke, D.L. and Billings, R.S. (1979) Comparison of training methods for improving the psychometric quality of experimental and administrative performance ratings. Journal of Applied Psychology, 64, 124±131. Whitener, E.M. (1990) Confusion of confidence intervals and credibility intervals. Journal of Applied Psychology, 75, 315±321. ß Blackwell Publishers Ltd 1997