HEXACO에 따른 정직성 및 겸손 관련 준거와 성격의 5요인 모델

HEXACO에 따른 정직성 및 겸손 관련 준거와 성격의 5요인 모델

The prediction of Honesty–Humility-related criteria by the HEXACO and Five-Factor Models of personality

Michael C. Ashton a,*, Kibeom Lee b,*

a Department of Psychology, Brock University, 500 Glenridge Avenue, St. Catharines, ON, Canada L2S 3A1

b Department of Psychology, University of Calgary, Calgary, AB, Canada T2N 1N4

HEXACO와 FFM에 대한 두 가지 질문을 연구하였다. 첫째로 HEXACP모델이 FFM을 outpredict하는가? 만약 그렇다면 H-H scale과 근접한 것을 더하면 FFM이 HEXACO model의 predictive validity를 HEXACO수준으로 높여줄까? 자기평가와 관찰자평가 결과를 보면 HEXACO모델이 FFM보다 predictive validity가 더 높았다. FFM에 HH척도가 더해지면 일부 criteria에 대해서는 HEXACO모델만큼 predictive validity가 올라갔지만, 다른 것들에 대해서는 여전히 낮았다.

We examined two questions involving the relative validity of the HEXACO and Five-Factor Models of personality structure. First, would the HEXACO model outpredict the Five-Factor Model (FFM) with regard to several diverse criteria that are conceptually relevant to the Honesty–Humility dimension of personality? If so, would the addition of a proxy Honesty–Humility scale—as computed from relevant facets of the FFM Agreeableness domain—allow the FFM to achieve predictive validities matching those of the HEXACO model? Results from self- and observer ratings in three samples (each N > 200) indicated that the HEXACO model showed considerable predictive validity advantages over the FFM. When a measure of Honesty–Humility derived from the FFM was added to the original five domains of that model, the predictive validity reached that of the HEXACO model for some criteria, but remained substantially below for others.

---

1.4. Purpose of the present research

In the present research, therefore, we aimed to address two main questions. 

Our first question was whether or not the predictive advantage previously observed for the six HEXACO dimensions in comparisons with the Big Five factors would also be observed in comparisons with the dimensions of the FFM. Given the incorporation of some elements of Honesty–Humility within the FFM variant of Agreeableness, one might expect that the latter factor would be a strong predictor of criteria that are conceptually related to Honesty–Humility. In this case, the advantage of the HEXACO model over the Big Five would not be generalizable; that is, there would be little or no advantage for the HEXACO framework over the FFM. On the other hand, however, the results of Ashton and Lee (2005) suggest that at least some Honesty–Humility-related criteria would show only modest associations with FFM Agreeableness. In this case, there would remain a substantial advantage for the HEXACO framework over the FFM.

Pending the answer to our first question described above, our second question was whether or not a proxy Honesty–Humility factor derived from the relevant aspects of FFM Agreeableness—that is, the NEO-PI-R Straightforwardness and Modesty facets—could match the predictive validity of the HEXACO-PI Honesty–Humility factor in predicting a wide variety of criterion variables. The results reported by Ashton and Lee (2005) indicate that such a result should be expected for at least some criteria. But as was also noted by Ashton and Lee, other criteria might be associated with aspects of (low) Honesty–Humility that have weaker links with the Straightforwardness and Modesty facets of NEO-PI-R Agreeableness. For example, the constructs of greed or status-seeking have a limited conceptual overlap with those facets, as do the “outlaw” tendencies that are also relevant to low Honesty–Humility. Thus, a proxy FFM-derived measure of Honesty–Humility might not predict those criteria as strongly as would a broader measure of Honesty–Humility, one that is likely to contain both (a) a larger proportion of common factor variance than does a measure consisting of only two facets and (b) specific variance associated with a wider array of Honesty–Humility traits.

In answering the above questions, we decided to assess several important constructs that appear to be poorly accommodated by the Big Five space, but that also show some conceptual overlap with the HEXACO space, especially the Honesty–Humility dimension. For example, the construct of materialism (e.g., Richins & Dawson, 1992) shows relatively weak associations with the Big Five ( Shafer, 2003 and Sharpe and Ramanaiah, 1999), but resembles traits found at the low pole of Honesty–Humility. Several constructs involving exploitive and/or criminal activity—including unethical business decisions, sexual harassment proclivity, and general delinquency—are relevant to some aspects of the Big Five (specifically, the low poles of Big Five Agreeableness and Conscientiousness) but have an even stronger conceptual overlap with low Honesty–Humility. Some sexuality-related constructs, such as seductiveness ( Paunonen, 2002) and unrestricted sociosexual orientation ( Simpson & Gangestad, 1991)—are conceptually linked to some extent with Big Five (and HEXACO) Extraversion, but also have an element of manipulation or exploitation that suggests low Honesty–Humility. Finally, as noted earlier, another construct involving a form of manipulation is Jackson’s (1970)social adroitness, which is weakly related to the Big Five but is associated with low Honesty–Humility. Thus, all of the constructs listed above are of much interest in the context of predictive validity comparisons between the HEXACO framework and the FFM.

In assessing the personality constructs of the FFM and the HEXACO structure, we used both self-reports and observer reports as our methods of measurement. The validity of self-report measures of personality in predicting objective criteria—such as directly observed behaviors or recorded life outcomes—is well established (e.g., Kolar et al., 1996 and Ozer and Benet-Martinez, 2006). However, there is also some evidence that observer reports of personality, as provided by persons who are well acquainted with the target individual, may provide even more valid assessments than do self-reports (Kolar et al., 1996). In some predictive contexts, moreover, observer report may be the preferred method of personality assessment (for example, in selection settings the target individuals might be motivated to provide highly socially desirable self-reports, and in some clinical settings the target individuals might have limited ability to provide accurate self-reports). The use of observer report measures of personality is particularly useful in the context of the present investigation, because the criterion constructs described above are generally measured by self-report scales. Therefore, by including observer reports as well as self-reports of personality, we would be able to examine the cross-source validity of the FFM and HEXACO dimensions, and to determine whether any predictive advantage of the latter framework would be generalizable across the two sources of personality data.

2.2. Materials

2.2.1. Personality inventories assessing FFM and HEXACO dimensions

We assessed the personality dimensions of the FFM and the HEXACO framework in each of the three samples. 

In the first sample, we obtained self-reports only, using the NEO-PI-R (Costa & McCrae, 1992) and the (full-length) HEXACO-PI (Lee and Ashton, 2004 and Lee and Ashton, 2006). 

In the second and third samples, we obtained both self- and observer reports on shorter versions of these instruments. 

By arrangement with the test publisher, we administered to both samples a subset of the NEO-PI-R items that included the items of the five NEO-FFI scales as well as the remaining items of the Straightforwardness and Modesty facet scales, with items permuted randomly. The second sample completed the half-length form of the HEXACO-PI, and the third sample completed the half-length form of the HEXACO-PI-R.2

2.2.2. Criterion variables

We also assessed the following criterion scales in the three participant samples; except as noted otherwise, all items used 1-to-5 response scales.

In the first participant sample, we administered the seven-item Materialism Centrality scale (Richins & Dawson, 1992), along with an eight-item version of Jackson’s (1970) Social Adroitness scale (as used by Ashton, 1998) and a 21-item Delinquency scale. (The Delinquency scale was based on that of Ashton (1998), but was expanded to include a wide array of delinquent behaviors, including vandalism, impaired driving, workplace safety violations, cheating on exams, speeding, stealing from employers, and vandalism. Each item used a 1-to-8 response scale, with options arranged in increasing order of the amount of delinquent activity, expressed as a frequency or a dollar value. This instrument is available from the authors.)

In the second participant sample, we administered the 15-item SPI Seductiveness scale (Paunonen, 2002), and the five attitude-based items of the Sociosexual Orientation Inventory (Simpson & Gangestad, 1991). For a subsample of 180 participants from Sample 2, we also administered a six-item scale measuring unethical business decisions, the items of which consist of detailed hypothetical scenarios rather than simple statements (adapted from Lee, Ashton, Morrison, Cordery, & Dunlop, 2008; see Appendix A).3 The items of the Unethical Business Decisions scale used a 1-to-4 response scale.

In the third participant sample, we again administered the Materialism Centrality scale and the Unethical Business Decisions scale, as well as a four-item scale assessing willingness to participate in sexual “quid pro quos”. Like the Unethical Business Decisions scale, the Sexual Quid Pro Quos scale is also a scenario-based measure (see Appendix B). The latter scale includes two items that describe a willingness to provide workplace-related favors in exchange for sex, and two items that describe a willingness to provide sex in exchange for workplace-related favors; all of these items use a 1-to-4 response scale.

---

3.1. FFM versus HEXACO variables: Self-reports

We begin by comparing the validity of self-reports on the FFM and HEXACO variables in predicting the criterion variables. Table 2 shows the zero-order and multiple correlations of the self-report personality scales with the criteria, as obtained in each of the three samples.

As seen in the table, the FFM predictors obtained moderately high multiple correlations with the various criteria, with values ranging from .31 to .53. However, the HEXACO predictors generally achieved considerably higher multiple correlations with the same criteria, with values ranging from .39 to .71. The predictive advantage of the HEXACO variables was chiefly due to the Honesty–Humility scale, which in several cases showed zero-order correlations that exceeded the multiple correlations produced by the FFM measures.

With regard to the specific criteria, the greatest differences in predictive validity between the FFM and HEXACO instruments were observed for the Materialism scale (Samples 1 and 3) and the Social Adroitness scale (Sample 1). For both of these criteria, the advantage for the HEXACO model over the FFM was rather large, with differences of about .20 in the observed multiple correlations. A moderately large difference between the predictive validities of the two frameworks was also observed with regard to the Delinquency criterion (Sample 1), for which the multiple correlation obtained by the HEXACO variables was .11 units higher than was that obtained by the FFM scales. A similar pattern was also observed for the Unethical Business Decisions scale, for which the difference in multiple correlations yielded by the HEXACO and FFM variable sets was .15 in Sample 2 and .08 in Sample 3. The HEXACO predictors also showed modest advantages over those of the FFM in predicting the remaining criteria of Samples 2 or 3, all of which involved some element of sexuality—Seductiveness, (Unrestricted) Sociosexuality, and Sexual Quid Pro Quos—with differences in multiple correlations ranging from .05 to .08.

3.2. FFM versus HEXACO variables: Observer reports

Next we compare the validity of observer reports on the FFM and HEXACO variables in predicting the (self-report) criterion variables. Table 3 displays the zero-order and multiple correlations of the observer report personality scales with the criteria, as obtained in both of the samples in which observer reports of personality were collected (i.e., Samples 2 and 3).

The correlations reported in Table 3 are somewhat lower than those of Table 2, but generally are still substantial in size. The multiple correlations obtained by the FFM scales in predicting the criterion variables ranged from .21 to .40; for the HEXACO scales, the multiple correlations with those criteria ranged from .29 to .50. As was observed in the self-report data, the Honesty–Humility measure of the HEXACO framework yielded zero-order correlations with some criteria that exceeded the multiple correlations achieved by the FFM measures.

For the Materialism criterion, the predictive advantage of the HEXACO variable set over that of the FFM was rather large, with a difference of .18 in the observed multiple correlation. A similarly large difference in the multiple correlations was observed in Sample 2 for the Unethical Business Decisions criterion (.18), but in Sample 3 the difference was only moderately large (.10). The predictive advantage of the HEXACO predictors over those of the FFM was small for the Seductiveness and Sociosexuality criteria of Sample 2 (about .05 correlation units), but rather large for the criterion of Sexual Quid Pro Quos of Sample 3, with a difference in multiple correlations of .13.

In general, the predictive advantages of the HEXACO framework over the FFM as observed in self-report data were also recovered in observer report data; in fact, the differences between the multiple correlations obtained by the two models tended to be at least as large in observer reports as in self-reports.

3.3. FFM plus NEO-PI-R Honesty–Humility composite: Self-reports

We next examined the ability of an augmented FFM—one that incorporates a separate variable assessing Honesty–Humility—to predict the various criterion variables. Recall that we computed an ad hoc NEO-PI-R Honesty–Humility composite scale as the mean of the Straightforwardness and Modesty facets of that inventory. For the analyses below, we examine the validity of that variable in combination with the remaining FFM variables, but with the FFM Agreeableness variable being recomputed to remove the Straightforwardness and Modesty content (see Section 2.3).

Within the self-report personality data, the inclusion of the NEO-PI-R Honesty–Humility composite generally added substantially to the validity of the FFM scales in predicting several criterion variables (see Table 2). The increment in observed multiple correlations was especially large (.18) for the criterion of Social Adroitness (Sample 1), but was also moderately large (about .10) for the Seductiveness and Sociosexuality criteria (Sample 2) and for Materialism (in Sample 3). Somewhat smaller increments were achieved in predicting Materialism (in Sample 1), Unethical Business Decisions (Sample 3), and Sexual Quid Pro Quos (Sample 3). The criterion variables for which the addition of the NEO-PI-R Honesty–Humility scale did not produce any appreciable improvement in predictive validity were the Delinquency criterion (Sample 1) and the Unethical Business Decisions criterion (Sample 2).

The predictive validity of this expanded FFM variable set—that is, with the addition of the NEO-PI-R Honesty–Humility composite—can be compared with that of the HEXACO framework. For several criteria—Social Adroitness, Seductiveness, Sociosexuality, and Sexual Quid Pro Quos—the predictive validity of the augmented “FFM-plus-H” framework was about equal to that of the HEXACO framework, as the multiple correlations achieved by the former framework either closely approached or slightly exceeded those achieved by the latter. This was also observed for the Unethical Business Decisions criterion within Sample 3, but the HEXACO model maintained a considerable predictive advantage in Sample 2, with a difference in multiple correlations of about .15. For several other criterion variables, the expansion of the FFM to include the NEO-PI-R Honesty–Humility variable did not produce predictive validities matching those of the HEXACO model; instead, the latter framework produced multiple correlations about .10 units higher in predicting the criteria of Delinquency and Materialism.

3.4. FFM plus NEO-PI-R Honesty–Humility composite: Observer reports

Within the observer report data, the addition of the ad hoc NEO-PI-R Honesty–Humility scale to the FFM variable set again produced some improvements in predictive validity with regard to several criteria (see Table 3). Across the criterion variables of Samples 2 and 3, the augmented “FFM-plus-H” achieved multiple correlations roughly .05 units higher than those of the original FFM. For the criteria of Seductiveness and Sociosexuality, this resulted in predictive validities that matched those yielded by the HEXACO dimensions. However, for the remaining criterion variables of Samples 2 and 3, the advantage of the HEXACO model over the FFM was preserved in spite of the addition of the NEO-PI-R Honesty–Humility composite to the latter framework. In particular, the HEXACO variable set produced multiple correlations with the criteria of Materialism, Unethical Business Decisions, and Sexual Quid Pro Quos that were .06 to .18 units higher than those generated by the “FFM-plus-H” variable set.

Thus, the increment in predictive validity that was afforded by the addition of NEO-PI-R Honesty–Humility to the FFM variable set was found to be generalizable across self- and observer reports of personality. However, the observed differences between the predictive validities of the expanded FFM and the HEXACO framework were somewhat larger in observer reports than in self-reports, with the advantage of the HEXACO framework being maintained across more of the criterion variables within observer report data.

4.1. Predictive validity of the HEXACO and FFM

As shown in Table 2 and Table 3, the six HEXACO dimensions consistently showed substantially higher multiple correlations than did the FFM dimensions in predicting the various criteria. Across such diverse variables as Materialism, Delinquency, Unethical Business Decisions, Sexual Quid Pro Quos, Seductiveness, and Sociosexuality, the multiple correlations yielded by the HEXACO factors were generally about .05 to .20 units higher than were those yielded by the FFM dimensions. These differences generalized across self- and observer reports of the personality predictor variables, and in fact were at least as strong for observer reports as for self-reports.

It should be noted, of course, that the criterion variables examined in this investigation were selected specifically because of their relevance to those aspects of the HEXACO factor space that are largely outside the space of the FFM. In particular, these criteria generally show some clear conceptual link to traits associated with Honesty–Humility. Naturally, the difference in predictive validity reported here would not be observed for criterion variables that are conceptually relevant to the much larger region of the personality space that is spanned by both the HEXACO framework and the FFM. But the fact remains that the various criteria considered in this investigation represent a diverse array of variables that are of obvious importance in human affairs. For example, the serious physical and financial harms caused to individuals and to societies by “common” criminality (as assessed by the delinquency criterion) and by “white-collar” exploitation (as assessed by the Unethical Business Decisions criterion) are self-evident. The prediction of variables such as these is an important aim for personality assessment, and hence the ability of the HEXACO model to predict these variables successfully is an important indication of its practical utility.

Related to this point, we should note that the advantage of the HEXACO model over the FFM in predicting the criterion variables of this investigation was mainly due to the Honesty–Humility factor of the former model. For example, when we added HEXACO-PI(-R) Honesty–Humility to the regression equations involving the NEO-PI-R or NEO-FFI scales, the multiple correlations were in most cases quite similar to those yielded by the six HEXACO-PI(-R) scales.4 In contrast, however, other investigations have examined criteria that show a predictive advantage for the HEXACO model over the FFM, but for which this advantage is not due to Honesty–Humility. Perhaps the best such example is that of the criterion of “phobic tendency” (see Ashton et al., 2008), which was predicted better by the HEXACO-PI Emotionality scale than by the five NEO-FFI scales in combination.5

4 The two exceptions were equations in which observer reports of personality were used as prdictors of Unethical Business Decisions (Study 2) and Sexual Quid Pro Quos (Study 3). In these cases, the multiple correlations achieved by the six HEXACO-PI-R scales exceeded those obtained by the five NEO-FFI scales in combination with the HEXACO-PI-R Honesty–Humility scale, thus indicating that HEXACO dimensions other than Honesty-Humility contributed to the predictive advantage of that model.

5 We should also note that the advantage of the HEXACO model over the FFM in predicting the criterion variables of the present study was chiefly attributable to the common variance shared by the various facets of the Honesty-Humility factor, rather than to the specific variance of any particular facet of that factor. We tested this by computing, for each of the criterion variables, an Honesty–Humility scale that excluded the constituent facet scale having the strongest association with that criterion. For most of the criteria of Tables 2 and 3, the zero-order correlations and multiple correlations yielded by the reduced Honesty–Humility scale continued to exceed substantially those achieved by NEO-FFI Agreeableness (The only exceptions were the Sexual Quid Pro Quos and Unethical Business Decisions criteria as predicted by self-reports in Study 3, in which cases the advantage of the HEXACO variables was eliminated when the Fairness facet was removed from the Honesty–Humility scale. In observer reports, the predictive advantage persisted in spite of the removal of that facet).

5. Conclusion

The Agreeableness domain of the FFM, unlike its counterpart in the classic Big Five framework, incorporates some traits associated with the Honesty–Humility factor of the HEXACO model of personality structure. Nevertheless, several important criterion variables that have conceptual links with Honesty–Humility are better predicted by the HEXACO model of personality structure than by the FFM. When a separate Honesty–Humility variable is computed from measures of the FFM dimensions, some of these criteria can be predicted as effectively as by the HEXACO factors. However, several criteria—particularly those associated with materialism, ethical violations, and criminality—are not well captured by those aspects of Honesty–Humility that are represented in the FFM, and are better accommodated by the dimensions of the HEXACO framework.

---

The prediction of Honesty–Humility-related criteria by the HEXACO and Five-Factor Models of personality

• Michael C. Ashton^a, , , 
• Kibeom Lee^b, , 

 Show more

DOI: 10.1016/j.jrp.2008.03.006

Get rights and content

---

Abstract

We examined two questions involving the relative validity of the HEXACO and Five-Factor Models of personality structure. First, would the HEXACO model outpredict the Five-Factor Model (FFM) with regard to several diverse criteria that are conceptually relevant to the Honesty–Humility dimension of personality? If so, would the addition of a proxy Honesty–Humility scale—as computed from relevant facets of the FFM Agreeableness domain—allow the FFM to achieve predictive validities matching those of the HEXACO model? Results from self- and observer ratings in three samples (each N > 200) indicated that the HEXACO model showed considerable predictive validity advantages over the FFM. When a measure of Honesty–Humility derived from the FFM was added to the original five domains of that model, the predictive validity reached that of the HEXACO model for some criteria, but remained substantially below for others.