Human resources (HR) practitioners and industrial psychologists normally use a combination of personnel selection methods and measures, and this combination constitutes a personnel selection model (Ployhart & Schneider, 2012). There are a number of personnel selection methods, which include application forms, graphology, reference checks, job knowledge tests and the selection interview, most of which have been found to have low validity in predicting job performance (Dale, 2003; Gatewood, Feild, & Barrick, 2016; Motowidlo, 2003; Schmidt & Hunter, 1998). On the contrary, there are psychological personnel selection measures such as CI, EI, personality and assessment centres, which are more scientific with better predictive validity (Anastasi & Urbina, 1997; Cote & Miners, 2006; Gregory, 2004; Hunter & Hunter, 1984; Joseph & Newman, 2010; O’Boyle et al., 2011).

Personnel selection involves the use of multiple methods and measures to be able to predict the multi-nature of the job performance criteria (Ployhart & Schneider, 2012). The underlying assumption of personnel selection models is that, as personnel selection measures are added to the model, the predictive validity is improved by way of incremental validity (Hattrup, 2012). Certain criteria must therefore characterise personnel selection models for them to predict job performance with fidelity. For example, each of the personnel selection methods or measures constituting a personnel selection model should be able to predict job performance at significant levels (Joseph & Newman, 2010). In addition, each personnel selection method or measure should be theoretically distinct from the constructs represented by other personnel selection methods or measures in the model to avoid redundancy (Joseph & Newman, 2010).

From our review of the literature, we classified personnel selection models into two major categories. First, there are models whose objective is to automate the decision-making processes such as resume screening (Kaluginaa & Shvyduna, 2014; Shehu & Saeed, 2016). We labelled these models ‘efficiency personnel selection models’. Efficiency-based personnel selection models suffer the drawback of being silent on the predictive power of the models, which is an important aspect of personnel selection. The second group consists of models that use regression techniques to determine the predictive power of different personnel selection measures and methods (Joseph & Newman, 2010; O’Boyle et al., 2011; Ployhart & Schneider, 2012). We termed these models ‘predictive personnel selection models’. The proposed personnel selection model for our research was, thus, predictive in nature.

Nguwi (2014a) states that many organisations in Zimbabwe still depend on the traditional personnel selection methods, which include the review of professional resumes and the selection interview. This is despite the fact that research has already established the ineffectiveness of the aforementioned practices (Nguwi, 2010). Nguwi (2010) also points out that the pitfalls of personnel selection methods such as the unstructured interviews emanate from the subjectivity of the judgements and conclusion arrived at because of the lack of standardisation of the interview process, which results in biases. In addition, most organisations in Zimbabwe engage in nepotism by employing their relatives and use unscientific personnel selection measures and methods (Dumbu & Chadamoyo, 2012; Nyamubarwa et al., 2013, Zinyemba, 2014). Thus, if Zimbabwean organisations fail to adopt valid personnel selection models, they may continue to employ unsuitable people, resulting in reduced employee performance capability. The study therefore sought to propose a scientific and valid model for personnel selection.

Job performance refers to the engagement in behaviour aimed at achieving tasks and duties that contribute to the economic value of an organisation (Bozionelos & Singh, 2017; Dan, Zhang, & Li, 2015; Motowidlo, 2003; Rich, Lepine, & Crawford, 2010). Job performance encapsulates both results and behaviour. The results-based approach represents the outcome of job performance (Rich et al., 2010). On the contrary, the behavioural approach to job performance is based on the notion that the behaviour is observable and provides a medium through which organisational results can be achieved (Jundt, Shoss, & Huang, 2014; Motowidlo, 2003; Tufail, Bashi, & Shoukat, 2017). Motowidlo (2003) argues that defining job performance from a behavioural point of view assists industrial psychologists to develop measures to assess how the results are achieved. Thus, although the ultimate outcome of job performance (results) is the evidence of job performance, the behaviour engaged in by employees may also be important in explaining how the results were achieved, leading to the notion that job performance is a multi-factor criterion (Williams & Anderson, 1991).

In the present study, job performance is conceptualised according to the multi-factor model and in terms of two main domains, namely task and contextual performance (Bozionelos & Singh, 2017; Williams & Anderson, 1991). Task performance or in-role behaviour refers to the effectiveness in performing activities that contribute to the organisation’s technical core tasks and focusses on core production activities (Borman & Motowidlo, 1993; Bozionelos & Singh, 2017; Jiao & Hardie, 2009; Miao, Humphrey, & Qian, 2017; Varela, Salgado, & Lasio, 2010; Williams & Anderson, 1991). The term contextual performance, or organisational citizenship behaviour, refers to extra-role behaviours that contribute to organisational performance because they positively affect the psychological, social and organisational contexts of work (Jiao & Hardie, 2009; Matula & Uon, 2016; Motowidlo, 2003). According to Motowidlo (2003), contextual performance may take at least two forms: organisation citizenship behaviour directed towards the organisation (OCBO) and organisation citizenship behaviour directed towards other individuals (OCBI). Organisation citizenship behaviour directed towards the organisation refers to individual effort aimed at assisting the organisation to achieve its objectives, and thus indirectly contributing to the economic value of the organisation (Bozionelos & Singh, 2017; Motowidlo, 2003; Williams & Anderson, 1991). Organisation citizenship behaviour directed towards other individuals occurs when individuals engage in behaviours that result in the improvement of their work colleagues’ job performance, for example assisting them with job tasks (Borman & Motowidlo, 1993; Bozionelos & Singh, 2017; Williams & Anderson, 1991).

The concept of CI refers to the ability to think rationally and to solve factual, novel and complex problems and is regarded as a major determinant for the survival of species (Gottfredson, 1997; Jensen, 1998; Lam & Kirby, 2002; Rindermann, 2007). Gregory (2004) conceptualises CI as the ability to reason (acquiring knowledge and using it for future purposes), to learn and to adapt to one’s environment. Spearman (1923) demonstrated that an individual’s performance on a variety of CI tests or subtests of intellectual effectiveness was determined by the general factor (g) and another factor (s) specific to the test or subtests (Willis, Dumont, & Kaufman, 2011). According to Spearman (1923, 1927a), performance on the subtests of CI was not influenced by a single specific factor (s), meaning that the more pervasive g was responsible for the performance (Floyd, McGrew, Barry, Raphael, & Rodgers, 2009; Willis et al., 2011). Gregory (2004) argues that the g supports the s by way of providing a common supply of mental energy as and when it is needed. Theories proposed by Thurstone (1938), Vernon (1950), Sternberg’s (1985), Gardner (1983, 1993, 1999), Guilford (1967) and Cattell (1941, 1971) support Spearman’s (1904, 1923, 1927a) g and s. Specifically, most of the theories acknowledge the presence of a g of CI that is pervasive and assists people to adapt to their environments. For this reason, we conceptualised CI as a unitary concept represented as general mental ability in line with Spearman (1904, 1923, 1927a).

The concept of EI has attracted attention over the last three decades (Gooty, Gavin, Ashkanasy, & Thomas, 2014; Joseph & Newman, 2010; O’Boyle et al., 2011; Wu, 2011). The emergence of the EI theory arose because of the realisation that CI alone could not explain all antecedence of job performance and occupational outcomes (Joseph & Newman, 2010; O’Boyle et al., 2011). Thus, the EI theory was developed mainly to explain the importance of non-cognitive components of intelligence to occupational outcomes including job performance (Murphy & Janeke, 2010).

Murphy and Janeke (2010) conceptualise EI as individuals’ traits or skills that help them to adapt to the environment by interpreting, managing and using theirs and others’ emotions to solve problems. Some researchers such as Mayer, Salovey and Caruso (2000) define EI as an ability to process emotional information cognitively, whereas others such as Schutte, Malouff and Bhullar (2009) and Schutte et al. (1998) argue that EI is a dispositional trait. The foregoing discourse resulted in two divergent conceptualisations of EI, that is EI as an ability and EI as a trait.

The conceptualisation of EI as an ability is based on the premise that EI is a set of cognitive abilities required for the processing of emotional information and the regulation of emotions (Mayer & Salovey, 1997; Wong, Law, & Wong, 2004; Wong, Wong, & Law, 2007; Zeidner, Matthews, & Roberts, 2004). The main difference between ability EI and other types of EI is that ability EI test items require the choice of a correct answer (Mayer et al., 2000). Mayer et al. (2000) conceptualise EI in terms of the four branches, namely identification, understanding, usage and regulation of emotions. Although ability EI is divided into the four-branch model, it is, however, unitary in nature, just like CI (Mayer & Salovey, 1997; Mayer, Salovey, & Caruso, 2004; Wong et al., 2004).

Petrides, Pita and Kokkinaki (2007) define trait EI as a collection of emotion-related dispositions that can be assessed through self-reports. Although theorists may vary in terms of its conceptualisation, trait EI can generally be described in terms of four factors, namely perception of emotion, managing own emotions, managing others’ emotions and utilisation of emotion (Schutte et al., 2009). In contrast to ability EI, the definition of trait EI is straightforward because it is a non-cognitive affective construct and acknowledges that emotions are subjective (Petrides et al., 2007). Petrides and Furnham (2003) argue that the conceptualisation of trait EI is robust because the construct is distinct from CI and personality.

The theory of psychological type (Jung, 1921) underpins the personality type theory of Myers and Briggs (Myers, 1987). The concept of psychological types was developed from a combination of the two attitudes (extraversion and introversion) and the four psychological functions (thinking, feeling, intuiting and sensing) (Burger, 2014; Feist & Feist, 2009). Myers and Briggs (Myers, 1987) extended Jung’s (1921) theory by adding an auxiliary function (judging and perceiving) and the auxiliary function’s balancing role in extraversion and introversion (Kirby & Myers, 2000). The personality type theory, thus, consists of two attitudes and six psychological functions (Myers, 1987). The attitudes and functions can be classified into four dichotomies and these are extraversion (E) versus introversion (I), sensing (S) versus intuition (N), thinking (T) versus feeling (F) and judging (J) versus perceiving (P) (Kirby & Myers, 2000). The four dichotomies produce 16 personality types, which describe personality in terms of a four-letter code (e.g. ENFP) and described in more detail by Kirby and Myers (2000). According to Myers (1987), the four dichotomies do not develop as balanced because this would result in undifferentiated personality. Rather, one of the four dichotomies becomes dominant, giving direction to differentiated personality (McCaulley & Martin, 1985, Myers, 1987).

Weak correlations have been found between ability EI and trait EI (Brackett & Mayer, 2003; Goldenberg, Matheson, & Mantler, 2006; Mayer, Salovey, & Caruso, 2002). Low correlations have also been reported between ability EI and personality (Austin, Farrelly, Black, & Moore, 2007; Joseph & Newman, 2010). With regard to CI, correlations between ability EI and CI have been found to be moderate to high (Austin, 2010; Cote & Miners, 2006; Joseph & Newman, 2010). In terms of job performance, Gooty et al. (2014) found that ability EI was related to task performance. Greenidge, Devonish and Alleyne (2014) found that the four dimensions of ability EI were positively correlated with organisational citizenship behaviour. A study by Sony and Mekoth (2016) shows that ability EI is positively correlated with job performance. Fallon et al. (2014) found ability EI to be a predictor of job performance through better decision-making. However, meta-analysis has revealed that in predicting job performance, ability EI measures have very low incremental validity beyond CI (O’Boyle et al., 2011).

Concerning the relationship between trait EI and personality type as measured by the Myers–Briggs Type Indicator (MBTI), significant positive correlations have been found between trait EI and sensing–intuition, thinking–feeling and extraversion–introversion (Higgs, 2001; Leary et al., 2009). Meta-analyses conducted by Joseph and Newman (2010) and O’Boyle et al. (2011) reveal weak relationships between trait EI and CI as well as ability EI. Research also provides evidence for trait EI from both explanatory and predictive viewpoints (Hui-Hua & Schutte, 2015; Mikolajczak, Roy, Verstrynge, & Luminet, 2009). For example, Wu (2011) found that trait EI significantly predicts job performance. Jung and Yoon (2012) found trait EI to predict organisational citizenship behaviour at significant levels. Joseph and Newman’s (2010) and O’Boyle et al.’s (2011) meta-analyses revealed that trait EI has incremental validity beyond personality better than ability EI. Schutte et al. (1998) note that people with high trait EI tend to perform better on difficult and frustrating tasks than those low on trait EI. Trait EI has also been found to predict job performance better in high emotional than in low emotional labour jobs (Joseph & Newman, 2010; O’Boyle et al., 2011).

Perry and Ball (2005) found a significant correlation between intrapersonal and interpersonal intelligence and the MBTI personality types, with no significant relationship with trait EI. This seems to indicate that personality types and trait EI are two distinct constructs. A study by Virmozelova and Dimitrova (2013) revealed that sensing and introversion were negatively correlated with sharing emotions and empathy and optimism, whereas extraversion correlated positively with sharing emotions and empathy. Johnson and Miller (2003) showed that extraversion is significantly positively correlated with the total EI score as measured by the Global Personality Inventory of Schmit, Kihm and Robie (2000), as well as the subscales of self-awareness, social skills, motivation and empathy. Johnson and Miller (2003) also found a low-to-moderate negative correlation between MBTI intuition and the total Global Personality Inventory as well as on all the subscales. The observed negative and low correlation between the personality types and the rest of the Global Personality Inventory EI scales suggests that trait EI and MBTI conceptualisation of personality may be, to a larger extent, distinct constructs. It is therefore worth investigating the MBTI in personnel selection contexts as it may have incremental validity in a personnel selection model.

Furnham, Dissou, Sloan and Chamorro-Premuzic (2007) note that studies on the relationship between personality types and CI are very few. Furnham et al. (2007) demonstrated that personality as measured by the MBTI showed that intuition and perceiving scores were positively and moderately correlated with both fluid and crystallised intelligence. This seems to be consistent with an earlier study by Kaufman, McLean and Lincoln (1996) in which individuals with high scores on intuition tended to score higher on CI. Myers and McCaulley (1985) also found intuitive individuals to have higher CI. One would, therefore, argue that if some personality types are correlated with other psychological constructs relevant to occupational settings, then it is worth testing the personality type theory in personnel selection contexts.

Research on the relationship between personality types and job performance is limited (Feist & Feist, 2009). Vincent, Ward and Denson (2013) found that the MBTI’s intuition is associated with leadership performance. The MBTI types have also been found to improve the quality of decision-making and problem solving (Prince, 2015; Sample, 2017), which is expected to lead to a better job performance. Engineers high on judging and intuition were found to perform better on the job (Carr, De la Garza, & Vorster, 2002). Because the area of personality types and job performance seems under-researched, more research is needed to determine the utility of personality types in personnel selection. Concerning the predictive power of CI, there is ubiquitous evidence, suggesting that CI predicts job performance across most, if not all, occupations (Cote & Miners, 2006; Gottfredson, 1997; Jensen, 1998; Joseph & Newman, 2010; Lam & Kirby, 2002; O’Boyle et al., 2011; Rindermann, 2007).

Based on a review of the research literature, the following research hypothesis was formulated:

The study was exploratory in nature and utilised a cross-sectional survey design (Leavy, 2017). The joint predictive power of CI, ability EI, trait EI and personality types in terms of job performance is not well understood and an under-researched phenomenon.

The research participants consisted of a convenient sample (N = 299) of supervisory-level employees and professionally qualified and experienced specialists from five private and publicly listed organisations in Zimbabwe. Black African males (61.2%) and females (38.8%) represented the sample. Almost half of the participants had more than 10 years’ tenure (49.5%) and the other half had less than 10 years’ tenure (50.5%). The 22–36 years age group (60.5%) was predominant in the sample. Thus, a relatively young workforce of experienced and qualified specialists formed a majority of the sample.

The primary researcher, a registered industrial or occupational psychologist with the Allied Health Practitioners Council of Zimbabwe, collected the data by means of scheduled contact sessions of at most 20 people, with participants completing the four measuring instruments. The JPS was also emailed to participants’ supervisors, for them to provide information on the participants’ performance levels. The researcher recruited participants via an email message.

The data were statistically analysed by using IBM SPSS version 22 (IBM, 2013) and the SAS version 9.4 (SAS, 2013) software programs. A Harman’s single-factor test and confirmatory factor analysis (CFA one-factor solution) was conducted to test for common method variance (CMV) (Tehseen, Ramayah, & Sajilan, 2017). Cronbach’s alpha coefficients and composite reliability were computed for each scale. Descriptive statistics, bivariate correlations, CFA and structural equation modelling (SEM) were performed. Best fit indices included the root means square error of approximation (RMSEA) and standardised root mean residual (SRMR) of below < 0.08, and the comparative fit index (CFI) above > 0.90. The MBTI personality types were treated as categorical data: E/I, S/N, T/F and J/P in the statistical analyses.

Permission to conduct the research was obtained from the participating organisations. The research institution (University of South Africa) also granted ethics clearance. Voluntary participation was honoured, and informed consent to utilise the data for research purposes was obtained from the participants. Confidentiality and anonymity were respected and maintained.

This article followed all ethical standards for carrying out research without direct contact with human or animal subjects.

The Harman’s one factor test revealed a lack of CMV for the AES (8.59%), JPS (5.53%), MBTI (9.16%) and the WEIS (3.47%). The one-factor CFAs (see Table 1) for these scales also indicated poor data fit which provided additional evidence that CMV was not a serious threat to the interpretation of the findings. The GAMA was excluded because it was a timed test and as a result had a number of missing responses.

SAS version 9.4 (SAS, 2013) was used to test the discriminant validity of the overall measurement model. Discriminant validity is important to assess for the presence of multi-collinearity among the constructs. Three models (one factor solution, multi-factor solution and optimised multi-factor solution) were run as part of efforts to ascertain the validity of the measurement model. The optimised multi-factor CFA model (see Model 3 in Table 2) showed good model fit (i.e. lack of multi-collinearity) and as such provided evidence of discriminant validity among the various scales’ constructs: chi-square = 318.99/df = 213; chi-square/df = 1.50; p ≤ 0.001; RMSEA = 0.05; SRMR = 0.07; CFI = 0.90). We used the CALIS statistical procedure with a maximum likelihood (Levenberg–Marquardt optimisation method) to optimise the model fit. Confirmatory factor analysis was not performed for the MBTI because the scale has dichotomous subscales. The absence of multicollinearity shows that the predictor constructs were theoretically distinct and thus satisfied the criteria for a personnel selection model (Hattrup, 2012; Joseph & Newman, 2010).

Table 3 shows the means and standard deviations and internal consistency reliability coefficients of the scales. The AES, JPS, GAMA and WEIS had some of their subscales reliabilities below 0.70, which is the cut-off point for good internal consistency reliability (Costa, Van, Abbott, & Krass, 2015). For the overall scales reliabilities, only the WEIS had the Cronbach’s alpha of below 0.70 (0.55). Internal consistency reliability coefficients of ≥ 0.50 can, however, be acceptable for research purposes (Taber, 2017). The alpha for the overall WEIS scores did not pose a serious threat to the interpretation of the results, but the low alpha was considered in the interpretation of the results.

The MBTI scores were categorical, meaning that the bivariate correlations for the dependent and independent variables had to be computed by using non-parametric tests. The Spearman’s correlation (SAS version 9.4; SAS, 2013) was therefore performed.

The practical effect sizes for the bivariate correlations were interpreted in line with Rashid, Mondol, Rahman and Noman (2016). In terms of CI and ability EI, a negative correlation (r = −0.31, p ≤ 0.001; moderate practical effect) was observed between the overall GAMA and WEIS scales. For the CI and personality, positive correlations were observed between the GAMA and sensing–thinking (r = 0.18, p ≤ 0.05; small practical effect) and the GAMA and intuition–thinking (r = 0.18, p ≤ 0.05; small practical effect). Negative correlations were, however, observed between the GAMA and sensing–feeling (r = −0.18, p ≤ 0.05; small practical effect) and the negative GAMA and intuition–feeling (r = −0.19, p ≤ 0.05; small practical effect). For CI and EI, there were no significant correlations between the overall GAMA and the AES (trait EI) and between the overall WEIS (ability EI) and AES. Concerning ability EI and personality, there were positive correlations between the WEIS and intuition–feeling (r = 0.17, p ≤ 0.05; small practical effect) and the WEIS and sensing–feeling (r = 0.18; small practical effect, p ≤ 0.05). The WEIS was negatively correlated with the sensing–thinking dichotomy (r = −0.16, p ≤ 0.05; small practical effect). A significant positive correlation was found between trait EI (AES overall score) and the MBTI intuition–feeling (r = 0.15, p ≤ 0.05; small practical effect). In summary, no correlations between the predictor variables were above r = 0.80, indicating the absence of multicollinearity and further demonstrating a discriminant validity of the measurement model (Cohen, Cohen, West, & Aiken, 2013).

The correlation between CI (GAMA) and JPS was significant (r = 0.35, p ≤ 0.001; moderate practical effect). For ability EI (WEIS), the regulation of emotions was the only subscale that had a significant correlation with overall job performance (r = 0.18, p ≤ 0.05; small practical effect) and OCBI (r = 0.23, p ≤ 0.01; small practical effect). Trait EI (overall AES) was significantly correlated with only task performance (r = 0.15, p ≤ 0.05; small practical effect), and the perception of emotion subscale was significantly correlated with both overall job performance (r = 0.15, p ≤ 0.05; small practical effect) and task performance (r = 0.20, p ≤ 0.05; small practical effect). In terms of personality, there was significant positive correlation between the sensing–thinking dichotomy and overall job performance (r = 0.15, p ≤ 0.05; small practical effect), showing that the thinking type is significantly correlated with job performance. The results revealed a significant positive correlation between the sensing–feeling dichotomy and OCBO (r = −0.15, p ≤ 0.05; small practical effect), demonstrating that the feeling type is negatively correlated with job performance.

SAS version 9.4 (SAS, 2013) was used to perform SEM and path analysis. The CALIS procedure of covariance structure analysis with a maximum likelihood estimation was utilised. The various measures (CI, ability EI, trait EI and job performance) were treated as overall constructs and the MBTI personality types as categorical construct types: E/I, S/N, T/F and J/P. Four SEM models were run and the results are reported in Table 5. The four models had good fit statistics because the RMSEA and SRMR were below 0.08, and the CFI was above 0.90.

Table 6 shows the various path analyses for the four SEM models using the latent variables. Practical effect sizes were reported in line with Cohen et al. (2013), and the pathways were all positive. For SEM Prediction Model 1, only CI (GAMA) (β = 0.80; R² = 0.65; large practical effect, p ≤ 0.001), ability EI (WEIS) (β = 0.64; R² = 0.41; large practical effect, p ≤ 0.001) and personality (MBTI) (β = 0.22; R² = 0.05; small practical effect, p ≤ 0.05) loaded significantly onto job performance. Concerning SEM prediction model 2, only CI (GAMA) (β = 0.73; R² = 0.54; large practical effect, p ≤ 0.001) and ability EI (WEIS) (β = 0.32; R² = 0.10; small practical effect, p ≤ 0.01) loaded significantly onto the job performance criterion. Structural equation modelling prediction model 3 shows that both types of EI did not load significantly onto job performance. In SEM prediction model 4, only personality type (β = 0.24; R² = 0.06; small practical effect, p ≤ 0.05) loaded significantly onto the job performance variable. The results show that CI was the best predictor of job performance, followed by ability EI, and then by personality. Trait EI did not account for variance in job performance.

The next step was to ascertain the best fit final SEM model, and this was achieved by computing the chi-square comparisons tests. Table 5 shows that the fit statistics of model 1 and model 2 were relatively better than those of model 3 and model 4. The results showed that SEM prediction models 1 and 2 contained the majority of the predictor variables and had the highest variance in job performance explained by the predictor variables compared with models 3 and 4. Thus, the chi-square comparison test was only performed for models 1 and 2 as illustrated in Table 7.

Table 7 shows that model 1 and model 2 (p = 0.64) were not significantly different, indicating that both models had relatively equal strengths in predicting job performance. However, choosing model 2 (with three predictor variables) ahead of model 1 (with all the four predictor variables) would result in reduction of the predictive power. This is because the incremental validity of 11% (65%  – 54%) from CI, 31% (41%  – 10%) from ability EI and 5% (absent in model 2) from personality would be lost. Thus, we chose model 1 as the final SEM prediction model for the study. Figure 1 illustrates SEM model 1. In summary, the results provided partial evidence in support of the research hypothesis.

The study enhanced personnel selection theory by identifying predictor variables that best predict job performance. Although HR practitioners in Zimbabwe and beyond could adopt the proposed personnel selection model, the research findings provided preliminary evidence of the utility of CI, ability EI, trait EI and personality, paving way for more research the area.