The Relationship Between Emotional Intelligence and Contextual Performance: Application to Automated and Vending Service Industry Executives

Abstract

This study examined the relationship between emotional intelligence (EI) and contextual performance among members of the National Automatic Merchandising Association (NAMA), representing leaders of the vending, coffee service, and food service management industries. Canonical analysis was used to investigate the relationships between three dimensions of EI—IN, OUT, and RELATIONSHIPS—and two dimensions of contextual performance—Interpersonal Facilitation and Job Dedication. Canonical analysis of the responses of 191 individuals demonstrated that there is a positive relationship between EI scores and contextual performance of NAMA members, as predicted.

KEYWORDS:

• Emotional intelligence
• contextual performance
• confirmatory factor analysis
• canonical analysis
• vending
• coffee service

INTRODUCTION

The topic of emotional intelligence (EI) recently has received much attention in the areas of human resources, management, leadership, and organizational behavior, since Goleman (1995) popularized the concept of EI first proposed by Salovey and Mayer (1990) in the book Emotional Intelligence: Why It Can Matter More Than IQ. Goleman (1995, 1998) claimed that EI is expected to contribute to effective performance at work as well as to success in personal life. In particular, he asserted that nearly 90% of the difference in the profiles of star performers and average performers in senior leadership positions was attributable to EI factors rather than to cognitive abilities (Goleman, 2000).

Researchers have reviewed methodological and theoretical issues in the EI constructs and have developed and validated a new model of EI that can be utilized practically in workplace settings (Law, Wong, & Song, 2004; Matthews, Roberts, & Zeidner, 2004). Cichy, Cha, and Kim (2007) also addressed this challenge by developing a new EI model, testing it, and modifying it into three dimensions—IN, OUT, and RELATIONSHIPS—using confirmatory factor analyses of the responses of 926 private club leaders. The revised three-factor structure of this new EI model, comprising 20 items, showed evidence of convergent and discriminant validity. Cichy et al. (2007) believe that the newly developed EI model is psychometrically sound and sufficiently short to be a useful tool for application in real organizational settings such as the private club industry. Research also acknowledges the need to further test empirically the relationship between EI and work outcomes such as job performance. Cichy et al. (2007) argued that “the validation process (of constructs) should be done first, even before claiming the utilization of EI with other outcome variables such as job performance, leadership effectiveness, and life success” (p. 6). Thus the next necessary step, after validating the EI scale, is to examine the effects of EI on work outcome variables in different organizational settings.

Study Objectives

The primary objective of this study was to examine the relationship between EI and an important work outcome variable, namely contextual performance, among members of the National Automatic Merchandising Association (NAMA), representing leaders from the vending, coffee service, and food service management industries.

Furthermore, this study investigated which dimension of EI (i.e., IN, OUT, RELATIONSHIPS) contributes most to which dimension of contextual performance (i.e., Interpersonal Facilitation and Job Dedication) based on the relationship between these two sets of variables.

LITERATURE REVIEW

Emotional Intelligence

Salovey and Mayer (1990) were pioneers in developing the concept of emotional intelligence, defined as “the subset of social intelligence that involves the ability to monitor one's own and others’ feelings and emotions, to discriminate among them and to use this information to guide one's thinking and actions” (p. 189). Their model subsequently was revised to include four dimensions of EI, including the ability to perceive accurately, appraise, and express emotion; the ability to access or generate feelings when they facilitate thought; the ability to understand emotions and emotional knowledge; and the ability to regulate emotions to promote emotional and intellectual growth (Mayer & Salovey, 1997).

Based on Salovey and Mayer's (1990) work, Goleman (1995) popularized the EI concept applied to workplace settings. He argued that it takes more than traditional cognitive intelligence to be successful at work. He explained that cognitive skills open a company's doors, while emotional skills help people thrive once they are hired. His EI model consists of four general abilities: self-awareness, self-management, social awareness, and relationship management (Goleman, 1995, 1998).

Later, Bar-On (1997, 2000) developed EQ-i, consisting of a total of 15 competencies in five composite scales, including intrapersonal, interpersonal, adaptability, stress management, and general mood. His definition of EI is “an array of personal, emotional, and social abilities, and skills that influence an individual's ability to cope effectively with his or her given environmental demands and pressures” (p. 1). Bar-On's definition is broader in scope than other models of EI developed by Salovey and Mayer (1990) and Goleman (1995).

Cichy et al. (2007) developed a new model of EI, comprising IN, OUT, and RELATIONSHIPS, and tested the reliability and validity of this three-factor EI model. Their EI model was tested with private club leaders, including chief operating officers (COOs) and general managers. IN is defined as one's ability to sense and lead one's own emotions, while OUT is one's ability to be aware of or relate to and understand others’ emotions. The RELATIONSHIPS construct represents one's ability to integrate emotional experiences with his or her actions and thoughts while interacting with others.

Contextual Performance

More recently, research on EI in the workplace suggests that EI is expected to contribute to positive work outcomes and organizational activities. Researchers have discussed and demonstrated that EI skills and capabilities are contributors to effective leadership (Carmeli, 2003; Gardner & Stough, 2002; George, 2000; Goleman, Boyatzis, & McKee, 2001). Most importantly, the effects of EI on job performance in the workplace are of great interest to researchers (Goleman, 2001; Law et al., 2004; Sala, Druskat, & Mount, 2005; Wong & Law, 2002; Zeidner, Matthews, & Roberts, 2004).

Until early 1993, researchers tended to investigate aspects of job performance emphasizing formal and core job responsibilities while ignoring a set of interpersonal and volitional behaviors (Conway, 1999). Borman and Motowidlo (1993) initially expanded job performance into two dimensions—task performance and contextual performance—arguing that contextual performance should be differentiated from task performance. Borman and Motowidlo (1993, 1997) recognized that task performance does not capture the full range of the individual's work roles. Task performance consists of expected and required job-specific behaviors, including formal and core job responsibilities, while contextual performance includes patterns of behaviors beyond job-specific duties (Borman & Motowidlo, 1993; Motowidlo & Van Scotter, 1994). The concept of contextual performance is almost identical with Organ's organizational citizenship behaviors (1988, 1997).

While both prescribed (known as task performance) and discretionary (known as contextual performance) behaviors in job performance are important to the success of the organization, leaders interacting with many levels of individuals within the organization are likely to be successful when engaged in extrarole behaviors beyond formal role requirements. Van Scotter and Motowidlo (1996) divided contextual performance into two dimensions: interpersonal facilitation, defined as “cooperative, considerate, and helpful acts that assist co-workers’ (staff) performance,” and job dedication, defined as “self-disciplined, motivated acts such as working hard, taking initiative and following rules to support organizational objectives” (p. 525).

Leaders with high EI are expected to show high levels of sensitivity toward discretionary organizational behaviors supporting staff and to put extra effort into their jobs. Emotionally intelligent individuals are likely to offer their staff more empathic responses to both personal and work-related problems, since they pertain to abilities or capabilities of understanding, using, regulating, and managing their own and others’ emotions (Carmeli & Josman, 2006). Several researchers have shown empirical support for the positive relationship between EI and contextual performance or organizational citizenship behaviors in work settings (Carmeli & Colakoglu, 2005; Carmeli & Josman, 2006; Charbonneau & Nicol, 2002). Hence the following hypothesis was proposed:

Hypothesis 1. There is a positive relationship between the three dimensions of EI and the two dimensions of contextual performance among NAMA members.

Since there was a lack of empirical support for quantifying the strength of the relationship among the three dimensions of the EI model and two dimensions of contextual performance in the current literature, the following research question was addressed:

Research Question 1. Which dimensions of EI and contextual performance are relatively more important than others?

METHODS

Data Collection and Sample Characteristics

Data were collected from a Web-based survey from May to August 2006. The target sample was members of the NAMA. NAMA is the leading trade association for the vending and coffee service industries. An invitation email was sent directly from NAMA's president and chief executive officer (CEO) to 1,395 valid email addressees. Prior to the main survey, all NAMA members received a prealert email explaining this online survey. Furthermore, an announcement of this online survey was included in a trade journal relating to the vending and coffee service industries to encourage NAMA members’ participation. For those who did not respond to the survey initially, two reminder emails were sent at 2-week intervals. Among the NAMA members, 213 members responded to the survey, resulting in a 15.1% response rate. After cleaning the incomplete data, the final number of usable responses used for further analyses was 191 (13.7%).

As Table 1 indicates, the majority of respondents (89%) were male, ages of half the respondents ranged between 46 and 55 (42%), and more than half of the respondents (62%) indicated more than 4 years of college. Their managerial experience was 19 years on average, and almost half (47%) had 20 or more years experience as a manager. Their total experience in the vending and coffee service industries was 22 years on average, and 48% of respondents had 20 or more years experiences in the industry.

Table 1 Profile of Respondents (n = 191)

Variables	Descriptions	Respondents
Gender
	Male	88.9%
	Female	11.1%
Education
	High school or less	8.5%
	Some college or associate (2-year) degree	29.8%
	Baccalaureate (4-year) degree	44.1%
	Postgraduate work	17.6%
Age
	35 years or younger	5.9%
	36–45 years	24.6%
	46–55 years	42.2%
	55–65 years	24.1%
	66 years or older	3.2%
Number of years as a manager
	4 years or less	4.1%
	5–9 years	8.6%
	10–19 years	39.9%
	20–29 years	30.9%
	30–39 years or more	16.5%
Average		18.5 (SD = 2.9)
Number of years as a manager in the industry
	4 years or less	3.1%
	5–9 years	6.6%
	10–19 years	42.2%
	20–29 years	31.4%
	30–39 years or more	16.7%
Average		21.9 (SD = 2.7)
Note. SD = standard deviation.

Measure Development and Scales

Scales of EI and contextual performance used in this study were adopted from a previous empirical study conducted by Cichy et al. (2007). The scale development process is described in detail in Cichy et al. (2007), explaining how items were generated and modified. A pilot study was conducted with 39 NAMA members who participated in the NAMA Executive Development Program in January 2005 on the campus of Michigan State University (Cichy, Geerdes, & Cha, 2006). Those participating in the pilot study provided feedback and comments regarding the clarity of survey questions.

All items for EI and contextual performance were measured on a 5-point Likert-type scale, ranging from 1 (very seldom or not true of me) to 5 (very often or true of me). The EI scale consisted of 37 items in three dimensions: IN (12 items), OUT (9 items), and, RELATIONSHIPS (8 items). Examples of IN, OUT, and RELATIONSHIPS are “I am able to sense my own feelings,” “I understand and appreciate emotions of others,” and “I am able to express and use feelings to communicate emotions with others,” respectively. Two dimensions of contextual performance were measured with 15 items—Interpersonal Facilitation (7 items) and Job Dedication7break; (8 items)—modified from Van Scotter and Motowidlo's (1996) contextual performance scale. Examples of Interpersonal Facilitation and Job Dedication are “I praise staff members when they are successful,” and “I pay close attention to important details in my work,” respectively.

Statistical Analyses

A confirmatory factor analysis (CFA) was used to validate the scales of EI and contextual performance using AMOS 5.0 software. CFA provides evidence of scale reliability, convergent validity, and discriminant validity (Bollen, 1989; Byrne, 1994; Kline, 1998). Next, canonical correlation analysis was used for this study. A canonical correlation is the correlation of two canonical variables, one representing a set of independent variables (in this example, the three dimensions of EI) and the other representing a set of dependent variables (in this example, the two dimensions of contextual performance) (Tabachnick & Fidell, 2006). This multivariate statistical method was used for investigating the relationships between two sets of variables. Whereas multiple regression is used to predict a single dependent variable (e.g., either Interpersonal Facilitation or Job Dedication) from a set of multiple independent variables (e.g., IN, OUT, RELATIONSHIPS), canonical correlation simultaneously predicts multiple dependent variables from multiple independent variables. Thus, in situations with multiple dependent and independent variables, canonical correlation is the most appropriate and powerful multivariate technique (Hair, Anderson, Tatham, & Black, 1998).

FINDINGS AND DISCUSSIONS

Factor Structure of EI and Contextual Performance

Table 2 showed overall fit indices for two EI models: an initial model and the revised model. According to overall fit indices, the initial model consisting of 37 items does not show acceptable fit indices, χ² (626) = 1639.7, p <.001 (χ²/df = 2.62, comparative fit index[CFI] = .85, nonnormed fit index [NNFI] = .82, root mean square error of approximation [RMSEA] = .07), although its chi-square ratio (<3) and RMSEA (<.10) are within the acceptable range. The initial model was modified and revised, based on the modification index. The revised model showed that all fit indices suggested a good fit of data, Δχ² (116) = 272.8, p <.01 (χ²/df = 2.4, CFI = .905, NNFI = .904, RMSEA = .05). The modified model surpasses the proposed model on all fit criteria, which confirmed that the modifications were meaningful. Using this modification process, the modified model of EI resulted in three dimensions (17 items total) consisting of IN (6 items), OUT (6 items), and RELATIONSHIPS (5 items).

Table 2 Comparison of Overall Fit Indices for Proposed and Modified Models of Emotional Intelligence (n = 191)

Model	χ^2	df	χ^2/df	NNFI	CFI	RMSEA	Δχ^2
Model A ^a	1639.7	626	2.62	.829	.840	.070	—
Model B ^b	272.8	116	2.35	.904	.905	.050	1366.9∗
Note. NNFI = nonnormed fit index; CFI = comparative fit index; RMSEA = root mean square error of approximation; Δχ^2 = χ^2 difference between two models (i.e., χ^2 of Model 1 minus χ^2 of Model 2 = 562.3). Criteria to accept model: χ^2/df should be less than 3; NNFI and CFI should be greater than .90; RMSEA should be less than .05.

Contextual performance was evaluated via the same process as in the EI construct with 15 items, including two factors (i.e., Interpersonal Facilitation and Job Dedication). Table 3 shows overall fit indices comparing the initial and modified contextual performance model. An initially proposed model was tested, but did not produce a good fit with data, χ² (89) = 355.3, p < .001 (χ²/df = 4.0, CFI = .65, NNFI = .59, RMSEA = .13). Seven items were removed for the revised model via modification procedures. Results for Model B showed that all fit indices suggested a good fit of data, Δχ² (18) = 26.5, p < .01 (χ²/df = 1.5, CFI = .98, NNFI = .96, RMSEA = .05). The modified model was improved from the proposed model on all fit criteria. The modified model of contextual performance resulted in two factors (8 items) consisting of Interpersonal Facilitation (4 items) and Job Dedication (4 items).

Table 3 Comparison of Overall Fit Indices for Proposed and Modified Models of Contextual Performance (n = 191)

Model	χ^2	df	χ^2/df	NNFI	CFI	RMSEA	Δχ^2
Model A ^a	355.3	89	3.99	.592	.654	.125	—
Model B ^b	26.5	18	1.47	.962	.975	.045	328.8∗
Note. NNFI = nonnormed fit index; CFI = comparative fit index; RMSEA = root mean square error of approximation; Δχ^2 = χ^2 difference between two models (i.e., χ^2 of Model 1 minus χ^2 of Model 2 = 562.3). Criteria to accept model: χ^2/df should be less than 3; NNFI and CFI should be greater than. 90; RMSEA should be less than. 05.

Reliability and Validity of Two Constructs

In addition to the model fit test, the reliability (calculated by Cronbach's alpha) was assessed. Means, standard deviations, and reliabilities are presented in Table 4. Standardized coefficient alpha for two constructs (EI and contextual performance) are as follows: IN = .86, OUT = .87; RELATIONSHIPS = .77 for EI; Interpersonal Facilitation = .89, Job Dedication = .77 for contextual performance. According to Nunnally (1978), Cronbach's alpha should be greater than .7 to be reliable. All scales of both constructs demonstrate acceptable and good reliability.

Table 4 Means, Standard Deviations, and Reliabilities for EI and CP Constructs in Modified Model (Each of Model B) (n = 191)

	Number of			Reliability
Factors	variables	Mean	SD ^a	(Cronbach's alpha ^b )
Emotional intelligence
IN	6	3.90	.57	.858
OUT	6	3.83	.61	.868
RELATIONSHIPS	5	3.96	.57	.771
Contextual performance
Interpersonal facilitation	4	3.87	.76	.892
Job dedication	4	3.92	.34	.772
Note.

Convergent validity is evidenced if different indicators used to measure the same construct obtain strongly correlated scores. In structural equation modeling (SEM), convergent validity can be assessed by reviewing the size and t test for factor loadings (Kline, 1998). For all three dimensions of EI and two dimensions of contextual performance, all factor loadings are statistically significant at p < .05. The size of factor loadings are greater than .5, all ranging from .57 to .90 in the EI construct and .53 to .96 in the contextual performance construct, which support evidence of convergent validity.

Discriminant validity was examined by the pairwise correlations between factors obtained from the revised models (Anderson & Gerbing, 1988). As a rule of thumb, Kline (1998) suggests that each pairwise correlation between factors should not exceed .85. As Table 5 shows, estimated correlations between factors were not excessively high, ranging from .24 to .78, providing support for discriminant validity.

Table 5 Correlations Among Factors In Model B for Examining Discriminant Validity (n = 191)

	IN	OUT	RELS ^a	IF ^b	JD ^c
IN	1
OUT	.308	1
RELS	.290	.244	1
IF	.418	.504	.778	1
JD	.457	.595	.660	.603	1
Note.

Hypothesis Testing: Canonical Analysis Results

Deriving the Canonical Function

The maximum number of canonical functions that can be extracted from the sets of variables equals the number of variables (e.g., IN, OUT, RELATIONSHIPS) in the smallest data set, independent or dependent variables. Since the current study has three independent variables and two dependent variables, the maximum number of canonical functions that can be extracted is two. Hair et al. (1998) recommended three criteria to decide which canonical functions should be interpreted, specifically: (1) the level of statistical significance of the function, (2) the magnitude of the canonical correlation, and (3) the amount of explained variance.

Level of Statistical Significance of the Function

Wilks’ lambda statistically tests whether two sets—three dimensions of EI with two dimensions of contextual job performance—of variables are associated by canonical correlation. Table 6 shows Wilks’ lambda test, determining the level of statistical significance of two functions. The result shows that two sets (three dimensions of EI and two dimensions of contextual performance) of data are significantly associated by canonical correlation.

Table 6 Wilks’ Lambda Test for Level of Significance (n = 191)

Root no.	Wilks’ lambda	Approximate F	Significance of F
1	.64	15.52	.000
2	.95	5.10	.007

Magnitude of Canonical Correlation

The magnitude of canonical correlation provides an estimate of the strength of the relationship between the predictive and the criterion set of variables. As shown in Table 7, the first canonical correlation is .57 (p < .05), while the second canonical correlation is .23 (p < .05). As usual, the first canonical correlation is far more important than the second.

Table 7 Eigenvalue and Canonical Correlation (n = 191)

Root			Cumulative	Canonical	Squared
no.	Eigenvalue	Percent	percent	correlation	correlation
1	.483	89.844	89.844	.571	.325
2	.055	10.156	100.000	.227	.052

Amount of Explained Variance

Values of squared canonical correlation give an indication of the percentage of variance in the dependent canonical variate that can be explained by the independent canonical variate. In this particular data set, for the first canonical function, dimensions of EI explain 33% of the variance in the dimensions of contextual performance.

Taken together, these results provide support for Hypothesis 1, with a significant positive relationship between three dimensions of EI and two dimensions of contextual performance among NAMA members (n = 191).

Interpreting Standardized Canonical Coefficients

It was determined that the canonical relationship is statistically significant and that the magnitudes of the canonical root and the amount of explained variance are acceptable. These findings indicate that there are significant relationships between three dimensions of EI and two dimensions of contextual job performance when examining these relationships simultaneously. To analyze in greater depth the relationship between individual dimensions of EI with those of contextual performance dimensions (Research Question 1), canonical weights need to be interpreted. Specifically, to determine the relative importance of each of the original variables in the canonical relationships, standardized canonical coefficients were interpreted. That is, the standardized canonical coefficients indicate the relative importance of the contribution of individual variables to a given canonical variable. Generally speaking, variables with relatively larger weights contribute more to variates, and vice versa. For dependent variables (job dedication and interpersonal facilitation for contextual job performance), both sizes of standardized canonical coefficients are considerably ample, indicating that both job dedication and interpersonal facilitation contribute to the given canonical variable. More specifically, the standardized canonical coefficient for job dedication (.62) is larger than the standardized canonical coefficient for interpersonal facilitation (.57), as shown in Table 8. On the other hand, for independent variables (IN, OUT, and RELATIONSHIPS) of EI, the RELATIONSHIPS (.68) construct contributes most to the given canonical variables followed by IN (.31) and OUT (.092).

Table 8 Standardized Canonical Coefficients for EI Dimensions and Contextual Performance Dimensions (n = 191)

Dependent	Function	Independent	Function
variables	no. 1	variables	no. 1
IF ^a	.574	IN	.301
JD ^b	.623	OUT	.092
		RELS ^c	.684
Note.

CONCLUSION

This study explored the relationship between EI and contextual performance among NAMA members, representing leaders of the vending, coffee service, and food service management industries. Canonical analysis supported that there is positive relationship between EI scores and contextual performance among NAMA members. Moreover, the RELATIONSHIP dimension is the most important independent variable, while both Interpersonal Facilitation and Job Dedication are important dependent variables for this positive relationship between these two sets of variables. Although one is not able to claim a casual relationship between EI and contextual performance, this study's findings are that there is a clear connection between EI and contextual performance among NAMA members.

Although the findings of this study contribute to a growing body of literature in the areas of EI and contextual performance, several limitations constrain the interpretation of the study's findings. A major limitation of this study was reliance on self-evaluation of EI and contextual performance. To have a more rigorous and objective test of the relationship between EI and contextual outcomes, evaluations from others need to be employed (Carmeli & Josman, 2006).

Future research is required to investigate whether the magnitude of the relationship between EI and contextual performance may differ, depending on other critical variables such as individuals’ attitudes, gender, or other affective outcome variables, suggesting investigations of moderating variables. Besides contextual performance, future research needs to extend investigation of the effects of EI on other work outcome variables.

Authors’ note: This research was supported in part by a grant from the NAMA Foundation. Mi Ran Kim and Julie Tkach, doctoral students in the Department of Community, Agriculture, Recreation, and Resource Studies at Michigan State University, were research assistants on this project.

Notes

^aModel A: Proposed initial model containing 37 items.

^bModel B: Modified model obtained by deleting 20 items from Model A, resulting in 17 items remaining.

^aModel A: Proposed initial model containing 15 items.

^bModel B: Modified model obtained by deleting 7 items from Model A, resulting in 8 items remaining.

^aStandard deviation.

^bCronbach's alpha provides an estimate of the interitem reliability or consistency.

^aRelationships.

^bInterpersonal Facilitation.

^cJob Dedication.

^aInterpersonal Facilitation.

^bJob Dedication.

^cRELATIONSHIPS.