EMPLOYEE DEVELOPMENT: AN EXAMINATION OF SERVICE STRATEGY IN A HIGH-CONTACT SERVICE ENVIRONMENT*

1. Introduction

Many service and manufacturing organizations find that when employees are satisfied, customers are satisfied as well and financial performance is enhanced (e.g., Heskett, Sasser, and Schlesinger 1997; Shellenbarger 1998; Siguaw and Enz 1999; Bernhardt, Donthu, and Kennett 2000). Increasingly, organizations want to measure and understand the relationships among investments in systems and employees, internal performance measures such as employee productivity and satisfaction, and external performance. ACNeilsen, a marketing research firm that ties a portion of management bonuses to employee satisfaction, finds that employee satisfaction, customer satisfaction, and financial performance are correlated (Shellenbarger 1998). Sears tracks similar relationships in their retail stores and observes a time lag from improved employee and customer satisfaction to improved financial performance (Rucci, Kim, and Quinn 1998).

Designing and developing jobs to enhance employee satisfaction and investment in employee training are critical elements in the service strategies of these firms. The relatively high labor content in many service organizations necessitates focusing attention on developing the critical labor component. Within the domain of employee development, design and development of systems to manage service employees help to define an organization's service strategy and play key roles in its delivery. An organization's service strategy is played out (i.e., delivered to customers) during each service encounter. The key component in the realized service strategy is the engagement of service employees in activities that deliver the service concept, service processes, and service outcomes.

Links between employees and customer satisfaction are observed in many industries (Hallowell and Schlesinger 2000) but are especially critical in high-contact service industries such as hospitals, legal services, and consulting. High-contact services are those in which customers directly interact with the service organization or service workers for an extended period (Chase 1981). Kellogg and Chase (1995) define high-contact services as those with high levels of communication time between customers and service employees, intimacy of communication, and richness of information exchanged during contact.

Chase and Tansik (1983) argue that greater employee knowledge and skills are needed in high-contact services because uncertainty (i.e., unpredictability) during the service encounter creates a need for employees who can make continuous and multiple nonprogrammed decisions. Considerable effort must be made by high-contact organizations to develop jobs and work systems, individual knowledge and skills, and interpersonal skills for employees to deliver high-quality services.

In high-contact environments, management of internal service quality is primarily a function of employee development and management. The components of employee development, the internal systems and programs that support service employees, are the key ingredients in a strategy for internal service quality in high-contact settings. Hallowell and Schlesinger (2000) state that "elements of internal service quality include workplace design, job design, employee selection, reward and recognition systems, training, policies and procedures, management style, goal alignment, and communication and tools. . . . Internal service quality is inherently situational in that any number of its different elements may be more or less important in different organizations at different times" (p. 209). It is through development of internal service quality, i.e., internal services that help employees do their jobs, that service organizations lay the foundation for delivery of their service strategies.

A critical component of a high-contact service strategy is management of the interaction between service employees and customers, i.e., the service encounter, which drives customer perception of service quality and customer satisfaction (Brown and Swartz 1989; Schneider, White, and Paul 1998; Gould-Williams 1999). The quality of service encounters is driven by numerous factors including the service workforce and the design of the service process itself. This study focuses on the first element-the workforce-and how workforce management drives internal performance measures that in turn drive external performance. Although relationships between internal service quality and internal performance measures of service capability and job satisfaction have been documented in low-contact services (i.e., insurance firms; Schlesinger and Zornitsky 1991; Hallowell, Schlesinger, and Zornitsky 1996), research in these areas for high-contact services has been limited. Research on these issues has been particularly limited in the health care industry (e.g., hospitals) in which the focus has tended toward the organizational strategy level, relating to issues of market leadership and differentiation in structure, rather than the service strategy level of service design and delivery.

In this study, we investigate the relationships among employee development (i.e., internal service quality), employee outcomes such as productivity and satisfaction, customer satisfaction, and financial performance in hospitals, a high-contact service environment. The basic framework of the relationships studied here originates in the service profit chain (Schlesinger and Heskett 1991b; Heskett et al. 1994; Heskett, Sasser, and Schlesigner 1997; Hallowell and Schlesinger 2000), which is shown in Figure 1. The service profit chain suggests that an operations strategy focused on development of internal service quality enhances employee capabilities and satisfaction, which in turn improve external service quality and customer satisfaction and, ultimately, financial performance. The development of internal service quality, which includes attention to work design, job design, training, and development, is captured here in the Employee Development construct.

Service organizations, particularly in high-contact industries, rely extensively on employees to understand and enact (during service encounters) their strategies. The extensive contact time between employees and customers results in customer perceptions that employees "represent the organization" (Chase and Tansik 1983). The employee development construct developed here reflects this critical role of employees in an organization's service strategy. This study, like the service profit chain, focuses on the criticality of emphasizing employee development to improve service performance.

In this study, data gathered from U.S. hospitals allow us to study a high-contact service environment (Heffring, Neilsen, Szklarz, and Dobson 1986; Bowers, Swan, and Koehler 1994). Hospitals deliver a complex mix of services and exhibit characteristics of encounter, pseudorelationship, and relationship services (Gutek 1995; Gutek, Bhappu, Liao-Troth, and Cherry 1999). Kellogg and Chase (1995) find that service encounters between hospital employees and patients include an array of contact levels. But overall, hospital patients typically exhibit expectations like those of relationship services ( Bowers, Swan, and Koehler 1994) in which the customer and provider (individual) expect to have repeated contact over time. Hospitals seem unique among high-contact services in that there is a relatively short time period in which employees and customers (patients) develop a relationship. Often, high-contact services involve long-term relationships during which service employees and customers develop a personal relationship. Thus, the criticality of service encounters between hospital employees and customers is heightened by high customer expectations of the interpersonal nature of these encounters and the relatively short duration of the employee-customer relationship.

A majority of hospitals are nonprofit organizations, including 85% of the hospitals in this study. The restriction of consumer choice and nonprofit nature of the industry may introduce nontraditional relationships between employees and customers. (Please note that hospital patients are referred to as customers in this article.) Additionally, payment and insurance systems can introduce restrictions in how customers select a hospital, and third-party reimbursement systems mean that hospital customers typically do not directly pay for the services they receive. Hospitals also deliver services high in credence characteristics in which customers may not be able to evaluate the technical quality of medical services they receive (Zeithaml 1981). This inability to adequately evaluate clinical quality shifts customers' attention to service (process) quality and, in particular, their interactions with hospital employees (Meyer and Collier 1998).

These unique hospital and customer characteristics may limit the generalizability of this study, but other services with a similar complex mix of services (e.g., commercial banking), payment systems (e.g., insured auto repair), or credence characteristics (e.g., legal services) may have customer characteristics and their links to performance like those observed in this study.

More detail on the service profit chain and related management theories and concepts are presented in the following section. Then, three competing models to explain the relationships among the studied constructs and the research hypotheses are introduced. A description of the methodology used to test empirically the hypotheses is presented next, including fully specified measurement and structural models. Finally, the results of the analysis and implications for managers, particularly in high-contact service industries, are discussed.

2. Background

The development of a competitive service strategy is modeled by Roth and van der Velde (1991) as a designed match between the "intended" marketing strategy and the "intended" operations strategy. As the service strategy is implemented, the goal is a match between the "realized" service concept or service product bundle, including explicit and implicit intangibles (from marketing), and the "realized" service delivery system (from operations). Roth and van der Velde's (1991) service delivery system includes the structure and infrastructure elements used in the "actual delivery of services" (p. 307). The service concept and service delivery system are designed to fit customer needs and expectations, and the realized service drives business performance.

Two critical elements that define a service strategy for high-contact service organizations are human resources and service delivery systems (Schlesinger and Heskett 1991b). The human resource element includes tangible characteristics such as experience, education, and training and intangibles such as the personal characteristics of the workforce. The service delivery system includes the processes and infrastructure used to deliver the service. The service profit chain, shown in Figure 1, focuses on employee management practices and how their benefits are realized in service provider-to-customer contact. The service profit chain is a model of firm performance that helps managers focus on quantifiable factors that lead to improved customer satisfaction and financial performance. It is premised on employee and customer loyalty as drivers of growth and profitability, rather than at the expense of profitability. In the service profit chain, employees are a critical component of the competitive strategy. Reichheld and Sasser (1990), Schlesinger and Heskett (1991a, 1991b, and 1991c), and Heskett et al. (1994) describe numerous examples of service organizations achieving success by capitalizing on the relationships presented in the service profit chain. However, much of the evidence is based on case studies in retail and fast food industries that have focused on relationships among the service profit chain constructs rather than on development of a full measurement model for the employee components. In the study presented here, a full measurement model is defined and the importance of each construct dimension is measured.

Roth, Chase, and Voss (1997) report the only large sample empirical test of all of the service profit chain construct relationships based on data representing U.S. firms in financial services, health care, hotels, and retail, among others. They identify significant correlations among each contiguous pair of components in the chain. The measurement items used to represent each construct are not reported. Other studies have investigated relationships similar to those proposed in the service profit chain framework. Hallowell and Schlesinger (2000) and Anderson and Mittal (2000) provide reviews of these studies.

3. Research Hypotheses

Using the framework of the service profit chain as a foundation, relationships among employee development, employee outcomes, customer satisfaction, and financial performance are studied here. Alternative models are considered in an effort to explain the relationships among service strategy development, reflected by the employee development construct, and the outcome measures of interest. Three models to evaluate the relationships among the studied constructs are estimated, as shown in Figure 2. Model 1 tests the premises that employee development improves employee outcomes, employee outcomes enhance customer satisfaction, and customer satisfaction leads to improved financial performance as measured by profit and revenue growth. The model constructs are described in further detail in the Methodology section.

Although previous studies address human resource management as a key component of strategic development, they do not address outcomes such as employee satisfaction or customer satisfaction as studied here. In fact, much of the operations management literature has given scant attention to the role of human resources in service design (Cook et al. 2002), while the human resources literature has lacked the customer focus necessary for service organizations (Schneider 1994).

Schneider, White, and Paul (1998) propose that the human resource management practices espoused by the service profit chain provide the "foundation for a service climate." A service climate results when employees perceive that "practices and procedures were in place to facilitate the delivery of excellent service, and management rewarded, supported, and expected excellent service" (Schneider and Bowen 1993, p. 39; Schneider, Wheeler, and Cox 1992). Schneider and Bowen argue that human resources must be managed strategically, providing the systems and policies to support superior service.

Similar research in manufacturing (MacDuffie 1995; Ahmad and Schroeder 2002) and cross-industry samples (Huselid 1995; Delaney and Huselid 1996) evaluates "bundles" of human resource practices and high-performance work practices. MacDuffie's bundles include work systems such as teams, suggestion programs, and job rotation as well as management policies including hiring criteria and compensation systems, components similar to those comprising the employee development construct studied here. He finds these practices are most effective for improving labor productivity and product quality when bundled into a consistent system for employee management and when integrated with policies focused on flexible production. Ahmad and Schroeder (2002) find that most of their studied human resource practices are significantly linked with operational performance.

Siguaw and Enz (1999) study whether improvements in process design, job designs, and rewards for hotel employees influence employee and customer satisfaction. They report case studies that document simultaneous improvement in employee and customer satisfaction as hotels improve check-in and check-out processes. Likewise, Bernhardt et al. (2000) find significant correlation between employee and customer satisfaction in fast-food restaurants. Although these studies treat employee and customer satisfaction as correlated relationships, the study presented here promotes theory and measurement that model employee outcomes as a driver of customer satisfaction.

Huselid (1995) and Delaney and Huselid (1996) study successful implementation of high-performance work practices including organizational structure (e.g., labor-management teams), employee skills (e.g., training), and employee motivation (e.g., linking performance appraisals to compensation). These dimensions are tied to improved financial performance but only employee motivation is linked with productivity.

Based on support from the literature, the research hypotheses tested here evaluate the relationships defined by the structural paths in the models in Figure 2. The first three hypotheses test the relationships presented in Model 1:

H^sub 1^: Employee development has a positive influence on employee outcomes

H^sub 2^: Employee outcomes have a positive influence on customer satisfaction

H^sub 3a^: Customer satisfaction has a positive influence on profit

H^sub 3b^: Customer satisfaction has a positive influence on revenue growth

An alternative model, Model 2 in Figure 2, is used to test whether employee outcomes are a requisite mediator between employee development and customer satisfaction. Model 2 includes all of the paths in Model 1 and a directional path from employee development to customer satisfaction. Because investments in human capital and other systems for employees change the service encounter, we hypothesize a direct effect of employee development on customer satisfaction in addition to a mediated effect through employee outcomes. This hypothesis addresses whether service strategy decisions influence customer satisfaction directly or via the service employee:

H^sub 4^: Employee development has a positive influence on customer satisfaction

In Model 3, the second alternative model shown in Figure 2, a directional path from employee outcomes to financial performance is added to Model 1. Support for this direct link between employee outcomes and financial performance has been observed in some industries (Huselid 1995; Delaney and Huselid 1996) but not in others (Bernhardt et al. 2000). In addition, Delaney and Huselid (1996) observe similar performance effects of human resource management practices in both for-profit and nonprofit organizations. Most of the organizations studied here have nonprofit ownership.

H^sub 5a^: Employee outcomes have a positive influence on profit

H^sub 5b^: Employee outcomes have a positive influence on revenue growth

Model 1 is "nested" in the two alternative models, i.e., Models 2 and 3 contain all of the paths in Model 1 and one additional path each. Because the models are nested, they can be compared using quantitative measures to determine which model provides the best fit for the sample data.

4. Methodology

Structural equation modeling is used to estimate the three alternative models. The significance of structural paths is used to test the research hypotheses. Each of the structural models is estimated twice, first with profit as the financial performance measure and then with revenue growth as the financial performance measure. The measurement model is described in the following section along with data collection, construct development, and common method variance.

4.1. Data Source

Data for this study were gathered simultaneously with those used in a study on quality management strategy and practices in hospitals, and that study provides a further description of the survey methods used [see Meyer and Collier (2001)]. The U.S. general acute care hospitals comprise the study sample. A pilot test of 51 hospitals was conducted to evaluate the reliability of the study constructs. Specialty and veterans hospitals and those with fewer than 60 beds were excluded from the study sample.

In discussions with other researchers, hospital administrators, and quality directors, we determined that quality directors (and similar) are the best source of information and data on the practices and outcomes measured for this study. The study questionnaire was mailed to the Director of Quality, Vice President of Quality, or Quality Manager at 814 hospitals (those remaining in population after exclusions listed previously). Two hundred twenty-nine hospitals completed and returned the questionnaire. Nine questionnaires were excluded because of missing data, and the final sample is 220 hospitals.

Respondents indicated their perception for each measurement item for the employee development, employee outcomes, and customer satisfaction constructs on seven-point Likert-type scales. Although it may be preferable to measure customer satisfaction directly from patients, similar studies show that internal and external measures of customer satisfaction are highly correlated (e.g., Harkey and Vraciu 1992; Licata 1995; Soteriou and Zenios 1999). Specifically, Soteriou and Zenios report a high correlation (r = 0.78) between internal and external customer satisfaction ratings and argue that this high correlation validates the use of internal data when external data are not readily available or easily obtainable. Harkey and Vraciu (1992) and Licata (1995) find similar results measuring customer perceptions of quality and satisfaction in hospitals.

Nonrespondent bias is assessed to ensure hospitals of various sizes (<100, 100-199, 200-299, 300-399, 400-499, and >500 beds) and ownership (for-profit, nonprofit, and government) are adequately represented in the respondent pool. Data on size and ownership for responding and nonresponding hospitals are obtained from a published source [American Hospital Association (AHA) 1997]. Bias by hospital size ([chi]^sup 2^ = 3.41, df = 5, and p = 0.64) and hospital ownership ([chi]^sup 2^ = 4.00, df = 2, and p = 0.14) are not significant.

4.2. Construct Development

Data to measure the constructs for employee development, employee outcomes, and customer satisfaction (see Figure 2) are obtained from the mailed questionnaire. Financial data are obtained from a published source [Health Care Investment Analysis, Inc. (HCIA) 1996, 2000].

The employee development construct is multidimensional as defined by Schlesinger and Heskett's (1991b) internal service quality. Because there are no published scales for internal service quality, its domain is used to formulate measurement of the employee development construct used here. The employee development construct represents a bundle of employee management issues and practices. The dimensions of this construct are similar to those in the Malcolm Baldrige National Quality Award Criteria category for Human Resource Development and Management [National Institute of Standards and Technology (NIST) 1999], which provides a useful framework for definition and measurement of the employee development construct. The Baldrige category includes three dimensions that mirror those addressed by Schlesinger and Heskett (1991b): (1) work systems; (2) staff education, training, and development; and (3) staff well being and satisfaction (NIST 1999). Each dimension is described briefly in the following list. The dimension names used here have been shortened for brevity.

Work systems are the work and job designs that organizations establish for their employees. Jobs should be designed and managed to support organizational strategy and staff plans. Increasing worker flexibility and enhancing decision-making authority for employees help organizations improve their work systems. Compensation and recognition programs for employees are part of the work system (Anderson and Wootton 1991; Counte, Glandon, Oleske, and Hill 1992; NIST 1999, p. 15).

Staff training and development include how these efforts are focused on accomplishing key organizational plans and addressing organizational needs. Building knowledge, skills, and capabilities are the focus of these efforts (Schweikhart and Strasser 1993; NIST 1999, p. 16).

Staff well-being is how the organization maintains an environment and climate that support the well being and motivation of employees. Work environment includes employee safety and health, and work climate includes providing useful benefits to employees. Organizations should have systems in place to measure and evaluate staff satisfaction and turnover (Williams, Sobti, and Aw 1994; NIST 1999, p. 17).

Measurement items for the three dimensions of employee development are shown in the Appendix. A complete discussion of validation and reliability for these items is reported in Goldstein and Schweikhart (2001).

Measures for the employee outcomes construct focus on key desirable results of employee development practices and include employee satisfaction, employee turnover, labor/management relationships, and workforce productivity and efficiency. Measures of the customer satisfaction construct include measures of satisfaction and comparisons against competitors. These items are also shown in the Appendix.

Measures of financial performance are obtained from a published source (HCIA 1996, 2000). For a measure of profit, HCIA assigns each hospital to one of 10 deciles indicating its profitability rank among all U.S. hospitals. Profit is calculated as follows:

(total revenue [operating and nonoperating] - total expenses)/total revenue.

Revenue Growth, which averaged 22% over 4 years for the sample hospitals, is calculated as a proportion as absolute revenue growth measures are skewed toward large hospitals. The years of study are selected based on data availability and to coordinate with collection of the other study data. Revenue growth is calculated as follows:

(2000 operating revenue - 1996 operating revenue)/l996 operating revenue.

4.3. Common Method Variance Analyses

Because data to measure the employee development, employee outcomes, and customer satisfaction constructs are reported by a single individual, several methods for assessing common method variance are addressed. First, a proactive approach of separating measurement items within the mailed questionnaire was used (Podsakoff and Organ 1986; Drolet and Morrison 2001).

Second, a Harman's one-factor test shows that measurement items for employee outcomes have moderate loadings on several factors but most heavily on a factor of employee development items. Although the data indicate these items are related, we know that items to measure employee outcomes are measures of results and the employee development items reflect management intentions and practices. Because we assume that management practices and results would be correlated, the results of this test are not unexpected. Customer satisfaction items load on a separate factor.

Third, a partial correlation procedure is run to further assess common method variance (Podsakoff and Organ 1986). This statistically conservative procedure uses factor analysis to partial from all measurement items the variance accounted for by the first factor (i.e., the common method factor). Next, factor scores are calculated for the study constructs (work systems, staff training and development, staff well being, employee outcomes, and customer satisfaction) from the residuals of the single-factor extraction. Results of this test show the employee outcomes residual factor is significantly correlated with factors for work systems (r = 0.20 and p < 0.01), staff well being (r = 0.16 and p < 0.05), and customer satisfaction (r = 0.67 and p < 0.01). The residual factor for customer satisfaction is significantly correlated with factors for work systems (r = 0.22 and p < 0.01), staff training and development (r = 0.17 and p < 0.05), and staff well being (r = 0.25 and p < 0.01). Factors for employee outcomes and staff training and development are not significantly correlated. These results mean that all but one of the construct relationships remains significant after the common method factor has been statistically controlled. In sum, these tests for common method variance indicate significant relationships among the constructs even after accounting for item correlations that may or may not represent common method variance.

5. Research Results

Structural equation modeling is used to estimate the path weights for the models shown in Figure 2. Employee outcomes, customer satisfaction, and the three dimensions of employee development are modeled as latent variables. Employee development is modeled as a second-order latent variable because this allows the model to capture both the covariation among the measurement items for the dimensions of work systems, staff training and development, and staff well being as well as the covariation among the dimensions themselves, as shown in Figure 2.

5.1. Measurement Model

Before estimating the structural model, the measurement model relating the latent constructs and their measurement items is evaluated. Unidimensionality and reliability of each of the study constructs are assessed in several ways. The proportion of variance extracted by the first factor is calculated (reported in the Appendix) with each factor explaining a significant portion of item variance. Subsequent factors explain significantly less but fairly equal proportions of the remaining variance, an indication of unidimensionality (Carmines and Zeller 1979). Composite reliabilities (Fornell and Larker 1981; Hair, Anderson, Tatham, and Black 1995, p. 642), an indicator of measurement reliability for latent constructs, are reported in the Appendix. All reliabilities are greater than 0.70, acceptable for this type of research (Hair, Anderson, Tatham, and Black 1995).

An assessment of measurement model paths indicates that each path linking a measurement item with a latent construct is statistically significant (p < 0.01), and all path estimates and error terms are within appropriate boundaries. Confirmatory factor analysis is run, resulting in a model with adequate fit [root mean square error of approximation (RMSEA) = 0.077; [chi]^sup 2^ = 470 and df = 203]. The RMSEA is independent of sample size (Steiger 1990; Browne and Mels 1994; Hair, Anderson, Tatham, and Black 1995), and RMSEA < 0.05 represents good model fit; 0.05 < RMSEA < 0.10 is reasonable model fit; RMSEA > 0.10 is poor model fit (Browne and Mels 1994, p. 86-87). The [chi]^sup 2^ statistic is based on the degrees of freedom and sample size, both of which are relatively large in this study.

5.2 Structural Model

A correlation matrix of the measured variables (questionnaire items, profit, and revenue growth) is the input for model estimation using RAMONA (Systat 1995; results verified using LISREL 8.30). Maximum likelihood estimation is used because it generally is the most common estimation method and is the default method in most structural equation modeling software packages. Maximum likelihood is a good estimation procedure when the measurement data are multivariate normal and a relatively large sample size is available, both of which are true for this study.

Model identity is established using Bollen's (1989) two-step rule. First, the number of free parameters in the model is fewer than the number of elements in the correlation matrix. Second, the model is recursive, i.e., it contains no reciprocal causation or feedback loops. These two conditions sufficiently establish model identity.

Estimating Model 1 with profit as the financial performance measure, RMSEA is 0.078 with 90% confidence interval of 0.070-0.086, indicating reasonable model fit. The RMSEA for Model 2 is 0.078 (0.070-0.087) and Model 3 is 0.078 (0.069-0.086). These results are reported in Table 1.

The [chi]^sup 2^ value for overall fit is 577.3 (p < 0.01) for Model 1, meaning that the sample correlation matrix is significantly different from the reproduced correlation matrix that results from model estimation. The Normed [chi]^sup 2^, which is obtained by dividing by the number of degrees of freedom, is 2.34 (577.3/247). A Normed [chi]^sup 2^ of less than 1.0 indicates a model may be overfitted, i.e., estimates too many paths (Joreskog 1969), while a value greater than 3.0 indicates a model may not adequately represent the observed data and may need improvement (Carmines and McIver 1981). The Normed [chi]^sup 2^ of 2.34 obtained here indicates the model adequately represents the data while not overestimating relationships among the data. With Normed [chi]^sup 2^ values of 2.35 and 2.32, Models 2 and 3 have results similar to Model 1, as reported in Table 1. The comparative fit index (CFI) of 0.89 is also reported in Table 1. In sum, the model fit measures indicate a reasonable but not close model fit.

Nested structural models can be compared by evaluating the statistical significance of the difference between the [chi]^sup 2^ values for two models, with degrees of freedom equal to the difference in degrees of freedom for the two models (Steiger, Shapiro, and Browne 1985; Anderson and Gerbing 1988; Hair, Anderson, Tatham, and Black 1995, pp. 643-644). (The null hypothesis is that there is no statistical difference between the two nested structural models.) A comparison between Models 1 and 2, with a [chi]^sup 2^ difference of 0.02 (df = 1, NS), indicates the models do not differ significantly in their fit. The [chi]^sup 2^ difference for Models 1 and 3 is 6.5 (df = 1 and p < 0.05), and the fit of these models differs significantly, as reported in Table 1. Adding paths to a structural model is known to improve model fit, but whether the improvement is theoretically or managerially significant, offering substantive insight into the relationships among the constructs, is the important question. This issue is addressed in the Discussion section.

Models 1, 2, and 3 are also estimated using revenue growth as the financial performance measure, resulting in RMSEA of 0.076 (0.067-0.084) for all three models as reported in Table 2. The [chi]^sup 2^ difference tests comparing the fit of Model 1 with Models 2 and 3 are not significant, indicating that the alternative models do not provide a better fit than Model 1.

Because the health care industry has unique characteristics that may alter the typical relationships observed in other industries, it is useful to evaluate the effects on the financial performance measures of organization and market factors. The significance of commonly cited factors are tested here: hospital ownership (for-profit, nonprofit, and government), size (number of beds), teaching status (major teaching, resident training only, and nonteaching), location (urban and rural), and patient payer mix [proportion of served customers paid by managed care, Medicare, or Medicaid; data sources are AHA (1997); HCIA (1996)]. Profit and revenue growth are regressed on these factors and the following organization and market factors significantly predict one or both financial performance measures: size, teaching status, location, and proportion of customers paid by managed care. Tetrachoric correlations are calculated for the dichotomous variables (location and teaching status; Joreskog and Sorbom 1996), and the four significant factors are added to the structural models described previously as measured variables with direct influence on the financial performance variables. Path weights for these items are reported in the Appendix. The model fit measures for these contingency models include RMSEA point estimates of 0.068-0.70, within the 90% confidence intervals of the noncontingency models. The [chi]^sup 2^ test statistic increases for each estimated model, reflecting the increase in matrix size (i.e., number of variables), and the Normed [chi]^sup 2^ decreases to 2.00-2.07, indicating reduced parsimony (Hair, Anderson, Tatham, and Black 1995, p. 620). We conclude that although the studied organization and market factors explain variability in financial performance and add degrees of freedom to the models, they do not improve model parsimony (i.e., fit per estimated coefficient).

The structural paths from the second-order latent variable employee development to the first-order latent variables range from 0.92 to 0.95 (path weights are consistent in Models 1, 2, and 3), as reported in Table 3. These weights indicate the second-order latent variable employee development explains much of the variance in work systems, staff training and development, and staff well being. The path weights from the other latent variable constructs in Figure 2 to their respective measurement items are reported in the Appendix.

5.3. Hypotheses Tests

Adjusting the financial performance measures for the organization and market factors described previously did not alter the weights of the structural paths of interest; so the hypothesis tests are consistent across the contingency and noncontingency models. Separate models for the two financial performance measures-profit and revenue growth-are estimated and because all paths unrelated to these measures are consistent across all models, the results are discussed simultaneously.

The research hypotheses represent structural paths in Figure 2 and the statistical significance of the structural paths tests the hypotheses. The first three hypotheses test the relationships represented in Model 1. The path from employee development to employee outcomes is 0.90 (critical value = 33.33; p < 0.01), supporting Hypothesis 1. Employee development is a significant predictor of employee outcomes because investments in employee development are realized through improved employee outcomes.

The path from employee outcomes to customer satisfaction is 0.63 (critical value = 11.45; p < 0.01). Therefore, Hypothesis 2 is supported, because employee outcomes is a significant predictor of customer satisfaction.

Hypotheses 3a and 3b, which address the effects of customer satisfaction on financial performance are supported in the model for revenue growth ([beta]^sub 32^ = 0.17, critical value = 2.36, and p < 0.05), but not for profit ([beta]^sub 32^ = 0.0 and critical value = 0.0, NS). These results, including the nonsignificant relationship between customer satisfaction and profit, are discussed in further detail in the Discussion section.

In Model 2, a direct path from employee development to customer satisfaction is added. The path weight of -0.12 (critical value = 0.43, NS) indicates that Hypothesis 4 is not supported. Therefore, the employee outcomes construct, which includes measures of employee satisfaction and productivity, is a significant mediator between employee development and customer satisfaction.

Hypotheses 5a and 5b test the direct influence of employee outcomes on financial performance as indicated in Model 3. This path is significant for profit ([beta]^sub 31^ = 0.26, critical value = 2.57, and p < 0.01), but not for revenue growth ([beta]^sub 31^ = -0.07 and critical value = 0.68, NS).

6. Discussion

This study provides insight into developing a service strategy for high-contact services that enhances employee outcomes, customer satisfaction, and, to some extent, financial performance. Designing and managing employee development is a critical ingredient for building employee productivity, loyalty, and satisfaction. And, particularly in the high-contact environment studied here, the employee outcomes that result from this strategic design are significant predictors of customer satisfaction. The results of this study for the most part support those from previous research in lower-contact service environments such as fast-food, retail, and hotels.

We report a fully specified model of constructs and measurement items to test the relationships shown in Figure 2. The employee development construct is modeled as a multidimensional construct, a second-order latent variable with three dimensions: work systems, staff training and development, and staff well being. Each of these dimensions plays a significant role in developing jobs and work environments that result in increased employee productivity and satisfaction. With internal consistency among the three dimensions, as evidenced by their high loadings on the employee development latent variable, management should focus on all three dimensions. The synergistic effects of these dimensions indicate that none alone is sufficient to achieve the effects on employee-related outcome measures that are observed in the model studied here. Managers in high-contact service environments should account for each dimension in their service strategy, service design, and human resource management plans.

This study focuses on the human resource element of service strategy and its critical role in service encounter management. Good service encounters can be designed and planned so that positive and productive employee-customer interactions are designed into the service delivery, not leaving these encounters to chance. Employees need the "ability and authority to achieve results for customers" (Heskett, Sasser, and Schlesigner 1997, p. 29). Hospital patients indicate that personal interaction with medical staff for medical and nonmedical reasons influences their satisfaction more than their medical outcomes ( Heffring, Neilsen, Szklarz, and Dobson 1986). Hospitals, with characteristics of relationship services (Gutek 2000), should be managed to enhance the customer relationship with the organization. Maintaining employee trust and loyalty to the organization may translate into similar trust and loyalty from customers.

Patients often find it difficult to evaluate medical outcomes because they lack the necessary technical expertise to evaluate these services with credence characteristics (Zeithaml 1981). Similar characteristics are observed in other service industries in which portions of the service delivery are either unobservable or beyond the expertise of most customers (e.g., air travel and some financial services). In such services, customers shift their focus away from the technical portions of the service and focus on the physical environment and their interactions with service employees. As technology replaces previously interactive services, remaining human service encounters should be targeted by management as critical opportunities for enhancing customer satisfaction. Customers focus on these interactions when assessing service quality, and service managers must ensure that the designed service strategy is effectively delivered through these interactions.

Of the three models evaluated here, Model 1 (see Figure 2) provides a parsimonious presentation of the construct relationships (with one exception described next). Model 1, with adequate fit (see Tables 1 and 2), accounts for the relationships among the studied constructs, and Models 2 and 3 do not, for the most part, significantly contribute to our understanding of the tested relationships. The exception is Model 3, using profit as the financial performance measure, which has significantly improved fit over Model 1 based on the [chi]^sup 2^ (see Table 1). However, the results also should be evaluated for practical and managerial significance. In this instance, the path added in Model 3, from employee outcomes to profit, is statistically significant, but with modest explanatory power. Because this is the only relationship that explains a significant portion of profit, it should be considered of managerial interest.

We find employee outcomes to be a necessary mediating construct between investments in employee development and the resulting customer satisfaction. In terms of management practices, this means that organizations should focus on development of work systems, training programs, and services for employee well being as a means to improve employee productivity and satisfaction rather than as a direct means to improve customer satisfaction.

There is no evidence that customer satisfaction is significantly linked with profit in the studied hospitals. The disconnect between these two performance measures differs from results of studies in other service industries (e.g., Roth, Chase, and Voss 1997, variety of service industries; Loveman 1998, banking; Rucci, Kim, and Quinn 1998, retail; Siguaw and Enz 1999, hotels; Bernhardt et al. 2000, fast-food). It is worth noting that in Bernhardt et al. (2000), Rucci, Kim, and Quinn (1998), Loveman (1998), and portions of Siguaw and Enz (1999), the studied facilities are individual service units of a single corporation. Additionally, Nelson et al. (1992) show that patient-perceived quality is related to some measures of profitability for a group of hospitals owned by one corporation. Corporate structures may limit many organizational and environmental factors that could influence the relationship between customer satisfaction and profit. In the study reported here, as well as in Roth, Chase, and Voss (1997), the studied organizations have a variety of owners. Converse to the results of predicting profit, customer satisfaction is a significant but modest predictor (see Table 3) of hospital revenue growth, but employee outcomes are not.

The health care industry has a predominantly indirect financial pathway from customers (patients) to providers (hospitals) through third-party payers (employers, insurers, or government), and customer loyalty can be influenced by payment and referral systems. This unique industry structure may provide insight into the relationships among employee outcomes, customer satisfaction, and financial performance reported here. Organization and market factors have some predictive significance for the financial performance variables, and although many industries have unique market factors that contribute to firm performance, these factors may be particularly significant in the health care industry.

Reviewing the role that profit and revenue growth play in the tested models helps us gain some understanding of potential explanations for these results regarding the prediction of financial performance. The calculation of each performance measure, given previously in the Methodology section, also may be of interest. The calculation for profit includes total revenue and expenses, and the calculation of revenue growth is a proportion based on operating revenue. The direct costs of poor employee outcomes for a hospital are higher turnover (increasing hiring and training costs), lower productivity (increasing operating costs), and lower employee satisfaction (likely increasing both previous costs). The direct cost of poor customer satisfaction is likely to impact negatively revenue as customers seek a different hospital for future services. An argument can be made that employee outcomes are more closely tied to cost functions, and, therefore, profit, and customer satisfaction is more closely tied to revenue, and, therefore, revenue growth. Longitudinally, outcomes associated with employees and customers would be expected to be linked, as shown in this study. Some studies report a lag between improvements in customer satisfaction and realized financial gains (Rucci, Kim, and Quinn 1998; Bernhardt et al. 2000). This study accounts for longitudinal effects of revenue growth but cross-sectional effects for all other measures.

7. Conclusions

Future research in the area of this study should continue to develop models to understand the components of service strategy in high-contact services. Although three dimensions of employee development are measured and modeled here, there may be other dimensions that are not addressed but that significantly influence employee outcomes. Additionally, three structural models are estimated here. There also exist other alternative and mathematically equivalent models that could achieve the same or better statistical fit but that would not be supported by the theories and frameworks studied here. Regardless, we should account for the existence of alternative models (MacCallum, Wegener, Uchino, and Fabringar 1993).

A limitation of this study is that the reported analyses are based mostly on perceptual self-reported data, and although we attempt to address concerns through analysis of a common method variance, it may be preferable to obtain more objective measures of the studied constructs. It is worth noting that the financial performance measures were obtained from a published source. Also, some of the unique characteristics of the health care industry may restrict the generalizability of the results obtained here. However, hospitals provide a useful environment for studying high-contact services, and the observed relationships provide a valuable addition to the service literature, which has focused on lower contact services.

The results of this study provide managerial insight into the strategies and systems that should be designed and implemented to support employees in high-contact service environments and show how influential these systems are in predicting employee outcomes including satisfaction, productivity, efficiency, and turnover. Service organizations that invest (or cut investment) in employee development systems including work systems, training and development, and staff well being can predict an increase (or decrease) in employee outcomes given the close relationship between these two constructs. Employee outcomes serve as a useful predictor of customer satisfaction as well. It is noteworthy that this study finds significant relationships among employee development, employee outcomes, and customer satisfaction that are similar to those observed in other service industries, despite the unique characteristics of hospital patients as customers.

The contributions of this study to the service design literature include the illustration of the need for service strategies that embrace employee development systems as a means for driving organizational performance. These relationships are empirically supported in a high-contact service environment with a large sample of independently owned organizations. In addition, fully specified measurement and structural models are reported here so that the measurement items that capture each construct domain and the relationships among those constructs are explicitly reported.

The author thanks three anonymous reviewers and coeditors Aleda V. Roth and Larry J. Menor for their helpful comments and suggestions, which contributed to improvements in this article. Much of the work on this article occurred while the author was at the Carlson School of Management, University of Minnesota.