Two structural equation models will be used to test the theory underlying the service profit chain. As discussed in section 4.2, two models are used instead of one due to the sample frame of this study. The first model, the employee model, will test the plausibility of the first half of the service profit chain; specifically, it will investigate the relationship between internal service quality, employee satisfaction, employee loyalty and employee productivity. The second model, the customer model, will explore the linkages between total retail experience, value, customer satisfaction and customer loyalty.
This chapter is organized in the following manner. Section 5.1 will detail why structural equation modeling is the methodology of choice for this research. Included in this section is a discussion of the power of the two models. Section 5.2, the employee model, will be broken into two sections. The first section, section 5.2.1 focuses on the composition of internal service quality as a second order factor. Once this higher level construct is established, section 5.2.2 explores how it relates to employee outcome factors. The individual hypotheses proposed in chapter 2, and summarized below, are tested, along with a discussion of overall model fit.
H1a: Internal service quality is positively associated with employee satisfaction.
H1b: Internal service quality is positively associated with employee loyalty.
H1c: Employee satisfaction is positively associated with employee loyalty.
H1d: Employee satisfaction is positively associated with employee productivity.
H1e: Employee loyalty is positively associated with employee productivity.
Both sections, 5.2.1 and 5.2.2, end with a discussion of the contributions this research makes in regards to the two respective areas.
The organization of Section 5.3 follows the same basic logic of Section 5.2 only it looks at the customer model. Section 5.3.1 details the measurement of total retail
experience, a second order factor similar to internal service quality. Section 5.3.3 investigates the relationship between total retail experience, value, customer satisfaction and customer loyalty. The hypotheses laid out in Chapter 3 are treated independently, followed by an analysis of overall model fit.
H2a: Total retail experience is positively associated with value.
H2b: Total retail experience is positively associated with customer satisfaction.
H2c: Value is positively associated with customer satisfaction.
H2d: Customer satisfaction is positively associated with customer loyalty.
Both sections end with a discussion of the contributions of this research.
5.1. Structural equation modeling
Structural equation modeling is a multivariate technique that allows for very powerful statistical analysis. As a technique, it provides many advantages over simpler statistical methodologies like analysis of variance (ANOVA) or regression. It is because of these advantages, detailed below, that structural equation modeling is selected as the data analytic tool of choice.
First, structural equation modeling incorporates the use of latent variables. A latent variable cannot be measured directly but rather must be represented or measured by two or more variables (Hair et al, 1998). This definition becomes clearer with an
example from this research. In order to measure a multi-faceted concept such as training and coaching it is necessary to ask questions regarding initial training, on-going training, length of training, quality of training, etc. It is impossible to measure this concept using only a single question; that is, a single indicator. As such, several questions are used to build a representation of the training and coaching an employee receives. In our study we use six questions. Using six questions has many advantages over using a single question.
The six question method allows for a much more comprehensive rendering of employees’
perceptions of the training they have received. Furthermore, because reliability is a function of the number of indicators used, using six questions increases the reliability of the survey instrument. By definition, a single item measure has zero reliability.
A second major advantage to using structural equation modeling is that it allows for a variable to act both in a dependent and independent role simultaneously. In the service profit chain model, many variables fit this description. For example, customer satisfaction acts as a dependent variable in the equation containing the total retail
experience construct, equation (1), while also acting simultaneously as an independent variable in the equation containing the customer loyalty construct, equation (2).
Customer Satisfaction = _1 * Total Retail Experience + _2 * Value + Error (1) Customer Loyalty = _3 * Customer Satisfaction + Error (2) This advantage also makes it possible to determine both the direct and indirect effects of variables within the service profit chain model. Even more importantly, this advantage allows for the determination of the causal nature of the relationships within the model;
findings which neither regression nor ANOVA can provide.
The primary concern of using structural equation modeling is obtaining large enough samples to achieve reasonable power (MacCallum et al, 1996; Fan et al, 1999;
Jackson, 2001). Various researchers have recommended that between five and ten observations are needed for each path estimate within the structural equation model (Hair et al, 1998). As discussed in section 4.2, our population frame allows us to select a sampling plan that results in very large sample yielding substantial power for our statistical analyses. Using the framework proposed by MacCallum et al (1996), the power for both structural equation models used in this research approaches 1.0. Degrees of freedom for the employee and customer models are 619 and 455 respectively. Sample size is 872 for the employee model and 1,076 for the customer model.
5.2. Employee Model
The employee model concentrates on both the direct and indirect relationships between internal service quality and employee related outcome variables – satisfaction, loyalty and productivity. Specifically, our rendering combines both a second order
measurement model of internal service quality as well as a structural model linking it to employee satisfaction and employee loyalty. Employee satisfaction is also directly linked to employee loyalty and productivity. Finally, employee loyalty is linked to employee productivity. A generic representation is given in Figure 5.1.
As discussed in Chapter 2, all the links in the chain are hypothesized to be positive.
SYSTAT 10.2’s structural equation modeling software, RAMONA, is used to perform the analysis.
5.2.1. Composition of internal service quality
The measurement portion of the employee model consists of constructing a second order internal service quality factor. As discussed in section 2.2, an eight dimensional representation of internal service quality is used. The eight first order factors include: training and coaching, goal management, teamwork, empowerment, work design, organizational support – management, organizational support – tools and rewards and recognition. The results of the second order construct development are
Internal
Service Quality
Employee Satisfaction
Employee Loyalty
Employee Productivity
+ +
+
+
+
Internal
Service Quality
Employee Satisfaction
Employee Loyalty
Employee Productivity
+ +
+
+
+
Figure 5.1. Generic representation of employee model
illustrated in Figures 5.1 and 5.2. Two figures are used because of spatial limitations. In reality, only one model is employed and all eight dimensions feed to a single internal service quality construct. Ovals represent latent variables, rectangles represent manifest variables. First and second order factor loadings are given in the illustration. The variances of all the latent variables are set to 1.0 for identification purposes. This specification results in standardized path coefficients that can be compared in terms of magnitude. Finally, all path coefficients, first and second order alike, are significant at the p<.001 level. The following notational schema, shown in Table 5.1, will be used throughout the remainder of the data analysis.
Notation Explanation Example of
Notation
Description of example XX# Capital letters followed by
number
TC1 The first question used to measure the TC (training and coaching) construct
exx# Lowercase “e” followed by two lowercase letters followed by number
etc1 The error term associated with the first question used to measure the TC construct (sometimes called unique variance of question TC1)
exxf Lowercase “e” followed by two lowercase letters followed by lowercase “f”
etcf The error term associated with the latent TC factor (training and coaching)
XXF Capital letters followed by a capital “F”
TCF The training and coaching factor
Table 5.1. Notational abbreviations used in this research
TC2
etc2
TC3
etc3
TC4
etc4
TC5
etc5
TC7
etc7
TC1
etc1 .615
.709 .805 .703 .807 .845
Training
&
Coaching
1.0 etcf
G3
eg3
G4
eg4
G5
eg5
G2
eg2 .640
.728 .810 .865
1.0 egf
T2
et2
T3
et3
T1
et1 .617
.865 .836
Teamwork
1.0 etf
E2
ee2
E3
ee3
E5
ee5
E1
ee1 .864
.916 .784 .626
Empower ment
1.0 eef
Internal Service Quality
1.0 .837
.843
.626
.679
TC2
etc2
TC3
etc3
TC4
etc4
TC5
etc5
TC7
etc7
TC1
etc1 .615
.709 .805 .703 .807 .845
Training
&
Coaching
1.0 etcf
G3
eg3
G4
eg4
G5
eg5
G2
eg2 .640
.728 .810 .865
Goal Management
1.0 egf
T2
et2
T3
et3
T1
et1 .617
.865 .836
Teamwork
1.0 etf
E2
ee2
E3
ee3
E5
ee5
E1
ee1 .864
.916 .784 .626
Empower ment
1.0 eef
Internal Service Quality
1.0 .837
.843
.626
.679
Figure 5.2. Internal Service Quality composition, part I
Figure 5.3. Internal Service Quality Composition, part II
Several important considerations can be taken away from the measurement portion of the employee model. First, the eight dimensional representation of internal service quality appears to be a very strong rendering. All eight dimensions exhibit large second order factor loadings, ranging from 0.626 (teamwork) to 0.843 (goal
management). These findings suggest that employees do indeed develop a broad conceptualization of their work surroundings, very similar to what previous researchers have called organization culture, organizational climate and/or human resource
management.
It is also interesting to note that the magnitudes of the factor loadings of the internal service quality dimensions fall into two groups. The first group consists of training and coaching (0.837 factor loading), goal management (0.843), organizational support – management (0.807) and organizational support – tools (0.815). It appears that employees attribute the most weight to these four dimensions when assessing the quality of their work environment. The second group is comprised of teamwork (0.626),
empowerment (0.679), work design (0.658) and rewards and recognition (0.644). None of the confidence intervals around the factor loading parameter estimates from the first group overlap with those from the second group. For example, the 95% confidence interval around the factor loading parameter estimate for training and coaching is (0.812, 0.861); the interval around the teamwork factor estimate is (0.583, 0.669). Because these two intervals do not overlap, it can be concluded that training and coaching is a more salient shaper of employees’ perceptions of their internal service quality than teamwork is. This argument can be extended to the rest of the elements in each of the two groups.
The measurement portion of the employee model makes two very important contributions. First, it is the most thorough development of a comprehensive internal service quality construct. The literature review in Chapter 2 is the most scrupulous examination of internal service quality literature to date, integrating theory from several different disciplines. This review provides for an eight dimensional representation that exhibits high degrees of both content and face validity. We also provide the most rigorous statistical construction of the internal service quality factor. Section 4.4 details the reliability, uni-dimensionality and discriminant validity of each individual element;
the first such study to do so. Moreover, Section 5.2.2 explores the nomological and predictive validities of the internal service quality construct by examining its relationship with employee satisfaction and employee loyalty. Again, this study is the first to explore thoroughly these relationships using a second order factor model.
The measurement portion of the employee model also gives practicing managers a valuable tool that can be used in resource allocation. The second order factor loadings of the eight internal service quality dimensions indicate the weightings that employees place on each individual dimension. Managers can use these weights, when faced with limited budgets, as an allocation instrument. If limited funds are available to improve working conditions, a manager will want to select any of the four first group items: training and coaching, goal management, organizational support – management and organizational support – tools. Improving any of these four areas will yield the greatest overall increase in internal service quality.
5.2.2. Linking internal service quality to satisfaction, loyalty and productivity
As discussed in Chapter 2, the following hypotheses are all embedded within the employee portion of the service profit chain:
H1a: Internal service quality is positively associated with employee satisfaction.
H1b: Internal service quality is positively associated with employee loyalty.
H1c: Employee satisfaction is positively associated with employee loyalty.
H1d: Employee satisfaction is positively associated with employee productivity.
H1e: Employee loyalty is positively associated with employee productivity.
Structural equations will test each of these hypotheses independently as well as testing the overall fit of the employee portion of the service profit chain.
The structural framework tested along with the results are illustrated in Figure 5.4. Due to spatial limitations, the eight internal service quality dimensions, along with their indicators, are omitted from the diagram, however, their equations are included in the model (i.e. internal service quality is still a second order factor made up of eight first order factors all made up of their individual indicators). The paths that are significant are so at the p<.001 level. The two paths that are not significant are not so at the p<.05 level.
As noted earlier, the variances of all latent variables are set to 1.0 for identification purposes, rectangles represent indicators and ovals represent latent variables. Table 5.2
which is presented immediately after Figure 5.4 details the point estimate, 90%
confidence interval, standard error and t-value for each path parameter.
Figure 5.4. Structural equation results for employee model
Internal Service Quality
1.0
Employee Loyalty
Employee Productivity
1.0
1.0
eelf
eepf
EL1 EL2 EL3
eel1 eel2 eel3
EP1 EP2 EP3
eep1 eep2 eep3
Employee Satisfaction
ees1
.696
EL4
eel4
EP4
eep4
.213
.600
ns
ns
.800 .861 .863 .870
.705 .885 .868 .720 Internal
Service Quality
1.0
Employee Loyalty
Employee Productivity
1.0
1.0
eelf
eepf
EL1 EL2 EL3
eel1 eel2 eel3
EP1 EP2 EP3
eep1 eep2 eep3
ees1
.696
EL4
eel4
EP4
eep4
.213
.600
ns
ns
.800 .861 .863 .870
.705 .885 .868 .720
Table 5.2. Structural equation results for employee model
Before discussing model fit each hypothesis is treated individually. H1a theorizes that there is a positive relationship between internal service quality and employee
satisfaction. The beta coefficient for the path between these two constructs is 0.697, 90%
confidence interval of (0.665, 0.730). The standard error of the path estimate is 0.020.
The t value associated with the path is 35.46 which is significant at the p < .001 level.
These results provide empirical validation of hypothesis H1a – there is a positive relationship between internal service quality and employee satisfaction. The confirmation of this hypothesis demonstrates that high quality support services and organizational policies, such as goal management, support – management, support – tools, rewards and recognition, etc., lead to employee satisfaction. In other words, employees notice and value the developmental HR practices of their organizations.
These findings resemble those in closely related fields. Whether the research has used the term organizational culture (Schneider, 1990; O’Rielly et al, 1991; Sheridan, 1992), organizational climate (Schneider et al, 1980; Rogg et al, 2001), high performance work systems (Huselid, 1995), high commitment human resource management (Arthur 1992, 1994; Whitener, 2001), innovative human resource practices (MacDuffie, 1995), quality of work life (Havlovic, 1991; Lau et al, 2001) or perceived organizational support (Eisenberger et al, 1986; Rhoades and Eisenberger, 2002) the results have been the same – employees are grateful for the efforts of the organization’s commitment to provide them an excellent working environment that not only treats them with respect but also develops their capabilities.
A similar result occurs when testing hypothesis H1b – Internal service quality is positively associated with employee loyalty. The path coefficient between these two
variables is 0.212, 90% confidence interval of (0.150, 0.274). The standard error of the coefficient is 0.038, yielding a t value of 5.62 which is significant at the p < .001 value.
These findings do lend support for Hypothesis H1b: internal service quality is positively associated with employee loyalty. The same arguments made above, linking internal service quality to employee satisfaction, can be made here. Employees do recognize and value an excellent working environment where their potential is utilized and developed.
As social exchange theorists argue, employees will feel a certain degree of reciprocity for their organization’s support – one way the reciprocity will reveal itself is through a heightened sense of commitment (Homans, 1961; Blau, 1964; Schneider et al, 1980;
Wayne et al, 1997; Rhoades and Eisenberger, 2002).
Before moving to the next hypothesis, a special comparative note should be made in regards to the direct effect of internal service quality on employee satisfaction and employee loyalty. The magnitude of the effect of internal service quality on employee satisfaction is over three times greater than the magnitude of the effect of internal service quality on employee loyalty, beta weights of 0.697 versus 0.212. This is the first research that explicitly allows for this comparison. We believe there are two underlying reasons for this result. First, there are probably fewer contextual effects that mediate the
relationship between internal service quality and employee satisfaction than there are that mediate the effect between internal service quality and employee loyalty. For example, an employee who is working in a retail environment while pursuing a college degree may value and be highly satisfied with their organization’s internal work environment but their career goal of pursuing a more permanent job in specialized field may lead to a less pronounced effect of their working environment on their intent to remain with the
organization in a similar position, i.e. loyalty. Second, from a temporal sense, internal service quality may have a more immediate effect on satisfaction than it does loyalty.
Giving employees a better working environment may dramatically increase their immediate satisfaction, but it may take a certain amount of time for that satisfaction to yield an increase in a more long term concept like loyalty. This concept is partially accounted for by exploring the indirect path from internal service quality to employee loyalty through the employee satisfaction variable.
The third hypothesis embedded within the service profit chain is H1c which states that there is a positive association between employee satisfaction and employee loyalty.
The beta coefficient between these two variables is 0.601, 90% confidence interval of (0.547, 0.656). The standard error for this coefficient is 0.033 resulting in a t value of 18.10, which is significant at the p < .001 level. These results lend support for hypothesis H1c and reflect the findings a vast amount of previous research suggesting the link
between employee satisfaction and employee loyalty is quite strong (see section 2.3 for a review of this literature). Indeed meta-analyses by Petty et al (1984) and Griffeth et al (2000) concluded that employee satisfaction is the most significant predictor of employee loyalty – the high beta weight, 0.601, of this study, serves to validate this concept.
The result of the structural equation model indicates that there is no evidence of the expected positive relationship between employee satisfaction and employee
productivity or of the relationship between employee loyalty and employee productivity;
hence, hypotheses H1d and H1e are not supported. These results contradict a vast
amount of previous research into these relationships, see section 2.3 and Appendix E for a review of this research. We believe that our results stem not from a truly insignificant
relationship among the three variables but rather from the construction of the employee productivity measure. In the main survey instrument, four items are used to measure employee productivity:
• I feel that I am a productive associate.
• Within my store I am a top seller.
• My average sales per hour is among the best in the store.
• My productivity has increased the longer I have worked in the store.
Although these questions have proven to be valid in previous studies (Denison et al, 1995; Huselid, 1995; Spreitzer et al, 1997; Silvestro and Cross, 2000) after deeper investigation of the operating policies of the specialty retailer used in this study, it
appears they may not be appropriate measures for this specific setting. First, the specialty retailer employs associates whose sole job responsibilities are stocking the front room and inventory control in the back room. As such, questions pertaining to sales are not applicable. Second, sales associates work in store zones and are instructed to hand customers over to other associates when the customer leaves a zone. Due to this teaming approach, individual sales figures are not tracked, therefore, employees do not actually know their average individual daily sales. Jointly these two store operating
characteristics, along with the fact that anonymous surveys are utilized, precluding the possibility of using manager’s perceptions of employee productivity, may be the
underlying cause of the insignificance among the three employee measures: satisfaction, loyalty and productivity. At the very least, this finding warrants future investigation and research.
The difficulty in measuring employee productivity in the service industry is a common one to service management scholars (Vuorinen et al, 1998; Van Looy et al,