Mar 3, 1999 - The initial chi-square statistic of the three-factor unrestricted model with the 12 items ... prescribed by James, Mulaik and Brett (1982), and Anderson and Gerbing ...... Bearden, William 0. and Jesse E. Teel (1983), âSelected ...
Examining Customer Value, Satisfaction, and Switching Costs in Multiple-Sourcing Purchase Decisions for Business Services Annie H. Liu Purdue University Kenneth L. Bernhardt Georgia State University Mark P. Leach Purdue University
ISBM Report 6-1999
Institute for the Study of Business Markets The Pennsylvania State University 402 Business Administration Building University Park, PA 16802-3004 (814) 863-2782 or (814) 863-0413 Fax
This publication is available in alternative media on request. The Pennsylvania State University is committed to the policy that all persons shall have equal access to programs, facilities, admission, and employment without regard to personal characteristics not related to ability, performance, or qualifications as determined by University policy or by state or federal authorities. The Pennsylvania State University does not discriminate against any person because of age, ancestry, color, disability or handicap, national origin, race, religious creed, sex, sexual orientation, or veteran status. Direct all inquiries regarding the nondiscrimination policy to the Affirmative Action Director, the Pennsylvania State University, 201 Willard Building, University Park, PA 16802-2801; TeL (814) 8654700/V; (814) 865-1150/flY. U.Ed. BUS 99-066
Examining Customer Value, Satisfaction, and Switching Costs in Multiple-Sourcing Purchase Decisions for Business Services
Annie H. Liu, Purdue Universi=y Kenneth L. Bernhardt, Georgia State Universi~’y Mark P. Leach, Purdue Universi=y
Submitted to ISBM Working Paper Series March 3, 1999 Address all correspondence to: Dr. Annie H. Liu Department of Consumer Sciences and Retailing
Purdue University West Lafayette, IN 47907 (765) 494-5009 Fax: (765) 494-0869
email:
aliu(a)purdue.edu
Annie H. Liu is Assistant Professor of Consumer Sciences and Retailing, School of Family Sciences, Purdue University (e-mail: a]iu(~purdue.edu) Kenneth L. Bernhardt is Regents’ Professor of Marketing, J. Mack Robinson College of Business, Georgia State University (email: mktk1b(~iilangate.gsu.edu). Mark P. Leach is Assistant Professor of Agricultural Economics, School of Agriculture, Purdue University (e-mail: 1each(~agecon.purdue.edu). This research was partially funded by grants from the Institute for the Study of Business Markets (JSBM) and Georgia State University.
Examining Customer Value, Satisfaction, and Switching Costs in Multiple-Sourcing Purchase Decisions for Business Services
Customer value is a concept central to marketing. In this study, a model of multiple sourcing organizational
buying behavior and a customer value scale are theoretically
developed for business services. Findings indicate that perceptions of value, satisfaction, and switching costs affect how organizationalbuyers allocate share-of-business to a service supplier.
I
Customer value, or the tradeoff between costs and benefits relative to competition, has been touted as the central driver of repurchase decisions (e.g., Grisaffe and Kumar 1998; Woodruff 1997; Holbook 1994), and customer satisfaction (Woodruff 1997; Reichheld 1996). Recent studies indicate that organizational buyers make purchase decisions based on the perceived value of an offering, rather than quality alone (Grisaffe and Kumar 1998; Gale 1994; Vyas and Woodside 1984). Conceptually, superior customer value has been linked with strong customer loyalty, repeat business, positive word-of-mouth, customer attachment, and growth in market share (Grisaffe and Kumar 1998; Naumann 1995; Buzzell and Gale 1987). Although the notion of providing value to customers is fundamental to marketing and research suggests that providing customer value may establish a competitive advantage (Gale 1994), empirical studies investigating this concept are limited. Marketers are generally unclear about what leads to customer value and how to measure and track value. Specifically, there is very little research devoted to identifying the dimensions of value (Sinha and DeSarbo 1998; Lemmink, de Reyter, and Wetzels 1998; Holbrook 1994), or examining the effect of customer value on organizational buying behavior and channel partnerships (Grisaffe and Kumar 1998). Research on channel relationships has investigated the concepts of perceived switching costs and satisfaction and has found each to affect relationship continuation (e.g., Ganesan 1994; Ping 1994). Specifically, perceived switching costs have been found to act as a barrier to exit (Ping 1994). Satisfaction has been found to enhance a customer’s long-term orientation, intention to continue a relationship, and repurchase intention (Ganesan 1994; Anderson and Weitz 1989; Woodside, Wilson, and Mimer 1992). However, recent studies examining the efficacy of customer satisfaction seem to question the strength of the relationship between satisfaction and customer retention; research has sometimes found a weak or non-existent link between satisfaction and repurchase (Reichheld 1996; Jones and Sasser 1995). Thus, marketers have increasingly advocated the
1
investigation of additional explanatory variables to increase the prediction of repurchase behavior (Woodruff 1997). The purpose of this study is to examine the concept of customer value and its role in organizational buying behavior. As such, it will investigate dimensions underlying the value concept and the relationships among an organizationalbuyer’s level of perceived value, satisfaction, perceived switching costs, and repurchase intentions.
A Model ofShare-ofBusiness Intention
As illustrated in Figure 1, a model of share-of-business intention is proposed for multiple sourcing organizational buyers. This model intends to identify the critical criteria that organizational buyers employ to allocate portions of business among current service suppliers. Prior research on organizational buying behavior has focused mainly on exclusive business exchange relationships (i.e., purchasing from a single supplier) (e.g., Morgan and Hunt 1994; Ganesan 1994; Anderson and Narus 1990; Heide and John 1990; Dwyer, Shurr and Oh 1987). However, in markets where organizational buyers perceive the costs of establishing strong partnerships outweighing the benefits, multiple suppliers are often used. In fact, most firms do not pursue a single source strategy (Presutti 1992), but prefer working with a muted number of suppliers (Presutti 1992; Treleven and Schweikhart 1988). Organizational buyers often allocate business among several suppliers to provide benchmarks (Kapoor and Gupta 1997), to maintain flexibility (Treleven and Schweikhart 1988), and to decrease their dependency on any one supplier (Swift 1995; Heide and John 1988). Treleven and Schweikhart (1988) found that organizational buyers use multiple-sourcing when they fear the disruption of supply, price escalation, inaccessibility to new technologies, or when volatile demand hinders effective inventory scheduling. This study investigates the link between organizationalbuyers’ value perceptions and
2
repurchase intentions when using multiple suppliers. When buyers use multiple outsourcing, shifts in the proportion of business given to one supplier at the expense of others may demonstrate relationship continuance betrer than stay or leave decisions (Gassenheimer, Calantone and Scully 1995). Thus, to better illustrate an organizational buyer’s intention to continue business relationships, Jackson’s (1985) conceptualization of “share-of-business” has been integrated Leuthesser and Kohli 1995). Jackson (1985) suggests that customers having moderate levels of dependence on their suppliers characterize most business-to-business exchange relationships. Organizational buyers may work mainly with a primary supplier to simplify transactions, but also maintain business relationships with a limited number of alternative suppliers to reduce risk (Presutti 1992). As such, buyers are motivated to remain
in
a small number of lasting relationships, but when problems or
competitive opportunities occur, the percentage of total business given to each alternative supplier can shift. From a supplier’s viewpoint, one source of growth comes from taking share-of-business away from competitors working with the same customer. While marketing literature suggests several determinants to relationship continuation, social psychology literature provides a theoretical framework of relationship continuance that guides the development of a conceptual model to help explain share-of-business allocations among several suppliers. In the following section, social exchange theory and the related investment model are presented to provide an arching theory of relationship continuance. Drawing from the investment model (e.g, Rusbult 1980a), customer value (e.g., Anderson and Thomson 1997; Gale 1994), and channel relationship literature (e.g., Heide and John 1988; Ganesan 1994), a theoretical framework is developed and three antecedents of business relationship continuation are identified: customer value, customer satisfaction and perceived switching costs.
3
THEORETICAL BACKGROUND Relevant Theories Social Exchange Theo~y. Thibaut and Kelley (1959) proposed that an individual’s tendency to continue a relationship does not rely on satisfaction with the relationship alone. It depends on his or her level of satisfaction, as well as the comparison level of alternatives (CLalt). The comparison level of alternatives (CLalt) determines the minimum level of relationship outcomes a person will accept in order to continue the relationship; it represents the attractiveness of the best obtainable alternative outside the current relationship. As such, CLMt provides an external comparison
standard to determine the intention to remain in a relationship. Satisfaction is viewed as a function of the discrepancy between the relationship outcome and the individual’s internal comparison level (CL). The concept of the comparison level (CL), or internal comparison standard, is an individual’s assessment of the expected benefits and costs associated with a relationship. Satisfaction, CLalt, and CL have been widely applied to marketing literature in studies examining the disconfirmation paradigm (e.g., Oliver 1980), gap theory (e.g., Zeithaml, Berry and Parasuraman 1996), working partnerships (e.g., Anderson and Narus 1990), and buyer-seller relationships (e.g., Dwyer, Schurr and Oh 1987). Traditionally, customer satisfaction has been viewed as the most important factor that leads to repurchase intent (Oliver 1980; Bearden and Teel 1983). However, consistent with social exchange theory, recent studies in marketing have found that customer satisfaction is a necessary but not sufficient condition to produce long-term customers (Reichheld 1996; Jones and Sasser 1995). The Investment Model. Building on social exchange theory, the investment model (Rusbult 1 980a) shows that satisfaction and alternatives are inadequate to determine relationship continuation. In addition to these two factors, the size of the investment in a relationship affects relationship continuation as well. Invested resources accumulate over time, therefore, as an
4
—
r
individual continues in a relationship, it becomes increasingly costly to leave and lose their investment. Since its development, the investment model has been applied in studies of personal, organizational, and marketing exchange relationships that have consistently demonstrated strong statistical associations among the proposed constructs. For example, satisfaction, alternatives, and investment size have been found to have a direct impact on commitment in friendships (Rusbult 1980b), in close relationships (Rusbult and Buunk 1993; Rusbult 1980a), in abusive relationships (Rusbult and Martz 1992), and in patient-physician relationships (Barksdale, Johnson and Suh 1997). In addition to the three main determinants identified by the investment model, costs and benefits of a relationship have also been shown to affect relationship continuation (e.g., Rusbult and Buunk 1993). As such, research from social psychology examining social exchange theory and the investment model suggests that one~s intent to continue a relationship increases as the benefits increase, as the costs decrease, as satisfaction increases, as alternatives decrease in attractiveness, and as the investment size increases.
Customer Value Past research in marketing on the concept of value can be categorized into three areas: (1) research on consumer consumption value; (2) research on perceived value; and (3) research on customer value or relative value. These areas are briefly reviewed to illustrate previous conceptualizations of customer value and to demonstrate how value in business exchange may be best represented as an organizational buyer’s tradeoff comparison among three elements in the investment model: benefits, costs, and competitive alternatives. Consumer Consumption Value. Hirschman and Holbrook (1982) proposed that the postpurchase consumption experience should entail the experiential consumption value (e.g., symbolic,
5
hedonic, and aesthetic nature of consumption), as well as rational consumption value (e.g., problem solving and need satisfaction). They defined experiential consumption value, or customer value (Holbrook 1994), as “an interactive, relativistic,preference e.~~perience” (p. 27) and advocated that consumers’ fantasies, feelings and fun need to be integrated with the rational consumption value that most of the consumer information processing framework accounts for. Later, Holbrook (1994) created a taxonomic scheme of eight types of customer value by employing three dimensions of the consumption experience (i.e., extrinsic / intrinsic, self-oriented / other oriented, and active reactive
/
consumption value). Perceived Value. Research on perceived value often examines the relationships among
perceived quality, price, perceived value, and purchase intention (e.g., Zeithaml 1988; Dodds, Monroe and Grewal 1991; Chang and Wildt 1994). Perceived value is often viewed as a consumer’s overall assessment of what is received and what is given (Zeithami 1988; p14-)’ and as a tradeoff between perceived quality and its affordability within a choice setting (Monroe and Krishnan 1985; p. 210). Zeithaml suggests that all costs that are salient to customers, such as monetary price and non-monetary price (e.g., time and effort) should be incorporated as perceived costs, and that the benefit components of perceived value should include perceived quality, and other intrinsic and extrinsic attributes. In general, empirical studies on perceived value show that perceived quality or benefits have a positive effect on value, while, perceived costs or sacrifices have a negative effect on value (e.g., Chang and Wildt 1994; Dodds et al. 1991). Additionally, perceived value has been shown to affect purchase intention or willingness to buy (e.g., Sweeney et al. 1997; Chang and Wildt 1994). Selected studies of perceived value and value conceptualization are presented in Table 1. Customer Value. Expanding on the conceptualization of perceived value, recent attention on customer value emphasizes the relativity of competition (Grisaffe and Kumar 1998; Anderson and Thomson 1997; Gale 1994). Specifically, Grisaffe and Kumar found that relative vahie has a greater
6
effect on customers’ intention to recommend than does absolute value. Similarly, Gale (1994) and Sinha and DeSarbo (1998) developed customer value maps (i.e., plot of relative quality and relative price) to demonstrate customer perceptions of an offer relative to competitors’. Customer value in these studies is specified as a relative judgment that is determined by a benefits/costs tradeoff, and by competitive offers. Naumann (1995) suggests that as competitive alternatives increase, customers may come to expect more value from a purchase. Likewise, when many substitutes are available, value perceptions of a current offer are likely to decrease. As such, the availability of competition, whether as a reference or a substitute, exerts a strong influence on customer valne perceptions.
In sum, customer value literature has proposed that benefits and costs are not separate features of an offer, that customers often evaluate the benefits and costs of an offering jointly, and that customers simultaneously compare this assessment with the benefits and costs of competitors. The cognitive comparison between benefits and costs is generally referred to as perceived value (e.g., Dodds et al 1991; Zeithami 1988), and referred to as relative value or customer value when comparisons include competitive alternatives (e.g., Grisaffe and Kumar 1998; Anderson and Thomson 1997; Gale 1994). Thus, customer value can be viewed as a comparative term based on perceived benefits, costs, and competitive alternatives and this comparison is an important determinant to purchase intent (e.g., Grisaffe and Kumar 1998; Chang and Wildt 1994). Interestingly, this conceptualization of customer value incorporates three of the variables proposed in the investment model predicting relationship continuance. However, the investment model conceptualizes benefits, costs and alternatives as separate factors, each having a direct impact on relationship continuation. Similarly, social exchange theory suggests that two types of “comparison levels” (Anderson and Narus 1990) are critical to exchange relations; one is the internal within-relation comparison and the other is the external between-relation comparison. As it is conceptualized here, customer value incorporates both an internal cost-benefit comparison and the
7
external comparison to competitive alternatives. In accordance with equity formulation (Homans 1974), customer value of a supplier can be referred to as benefits-to-costs ratio of a supplier (A), compared to that of competitive alternatives (J). VALA
(BENA / CSTA): (BENJ / CSTJ)
The notion that customer value is relative to competition is especially relevant to the current study since organizational buyers often purchase from multiple suppliers, compare alternative offerings among competing suppliers, and benchmark their current suppliers with competitors (Kapoor and Gupta 1997).
Following the propositions of the investment model, if the benefits/costs comparison of a current supplier is greater than other alternatives available @.e., higher customer value), then, the customer is more likely to continue the relationship. As such, customer value is posited as an antecedent that will positively affect customers’ intention to continue the relationship with a supplier (i.e., allocate share-of-business). Therefore, hypothesis one is as follows: HI:
An organizational buyer’s perceivedlevel of customer value with a supplier will be positively related to the share-of-business given to the supplier.
Perceived Switching Costs Perceived switching costs indicate the ease or difficulty an organizational buyer has in replacing an existing supplier. Developing relationships with new suppliers usually requires, in part, the sacrificing of past investments along with the modification of the established routines (Heide and John 1988). As such, the evaluation of the costs associated with switching can include an assessment of past transaction specific investments as well as perceived potential adjustment costs that the firm will face when establishing a new relationship with a replacement. Transaction specific investments include monetary and non-monetary costs that are irretrievable when the exchange relationship ends (Jackson 1985). Heide and John (1988) suggest
8
that transaction specific investments that cannot be easily transferred to other exchange relations may increase the perceptions of switching costs. Transaction specific investments can arise from procedural knowledge, the establishment of working relationship and routines, as well as idiosyncratic investments on equipment (Heide and Weise 1995; Jackson 1985). Studies on buyerseller relationships show that transaction specific investments can be deemed a relationship pledge that indicate commitment between exchange parties (Anderson and Weitz 1992). Specifically, transaction specific investments between exchange parties can enhance alliances (Heide and John 1990), and facilitate a long-term orientation (Ganesan 1994). The concept of transaction specific investments is similar to the size of investment proposed by Rusbult (1980b), in that each enhance the motivation to continue a relationship. However, individuals can opt to not be in a personal relationship, whereas terminating a supplier in a business exchange usually requires establishing a relationship with a replacement supplier or internalizing the related business function. Similarly, when using multiple suppliers, discontinuing or reducing the amount of business given to a supplier calls for simultaneous adjustment with a replacement. This process would typically involve an evaluation of the resources needed to establish an efficient relationship with the replacement(s). As such, the concept of perceived switching costs, that includes both transaction specific investments and perceived adjustment costs, may better represent the exit barriers considered by organizational buyers when making repurchase decisions. Literature on channel relations suggests that the replaceability of a supplier can be deemed a measure of channel dependence (Heide and John 1988). From an organizational buyer’s perspective, the replaceability of a supplier decreases dependency on that supplier, and therefore, allows alternative choice decisions. From a customer’s perspective, established personal relationships, along with familiar procedures, and contact persons that are present with a current supplier, may be viewed as perceived switching costs. Empirical studies show that high perceived
9
switching costs will decrease organizational search effort, and will limit a customer’s consideration process (Heide and Weiss 1995). Jackson (1985) maintains that organizational buyers are often motivated to stay in existing relationships to economize on switching costs. Perceived switching costs have been found to increase future interactions, commitment (Heide andJohn 1990; Anderson and Weitz 1992), and decrease exit intention (Ping 1994). Similarly, Rusbult (1980a) found that investments, in terms of time, effort and money, can increase the intention to continue a relationship. Thus, hypothesis two is as follows: H2:
An organizational buyer’s level of perceived costs associated with switching from a supplier to another supplier will be positively related to the share-of-business given to the supplier.
Customer Satisfaction Customer satisfaction is defined here as an organizational buyer’s overall post-purchase evaluation toward a service supplier. It is a comparison of expectations and outcome performance. In order to investigate long run relationships between buyers and suppliers, we adapt the concept of relationship satisfaction from social exchange theory and take a cumulative view of customer satisfaction that captures the overall attitude toward the business-to-business service supplier. Literature concerning buyer-seller relationships has found that customer satisfaction influences buyers’ long-term orientation (Ganesan 1994), exit intentions (Ping 1994), and repurchase intentions (Woodside, Wilson and Milner 1992). Specifically, Ganesan (1994) found satisfaction, tmst, and mutual dependence to be the major determinants of a long-term orientation between manufactures and their distributors. Likewise, Ping (1994) found that when satisfaction is low, attractive substitutes influence a channel member’s exit intention; and when satisfaction is high, attractive substitutes have less influence on exit intention. Over the last two decades, research on the disconfirmation paradigm (e.g., Patterson et al.
10
1997; Oliver and Swan 1989; Bearden and Teel 1983; Oliver 1980) has found that customer satisfaction is one of the major determinants of repurchase intention. As such, there is little wonder why many firms have invested in systems and procedures for tracking customer satisfaction. However, customer satisfaction only accounts for a portion of the variance in retention, and recently many of these firms have found customer satisfaction is highiy related to, but not necessarily predictive of relationship continuation (Jones and Sasser 1995). These findings are consistent with the propositions of social exchange theory and the investment model which suggest that while overall satisfaction can enhance intention to continue a relationship, other variables must also be considered. Applying these theories, hypothesis three is as follows: H3:
An organizational buyer’s level of satisfaction with a supplier will be positively related to the share-of-business given to the supplier.
Although many authors suggest that customer value is the driving force for customer satisfaction (e.g., Reichheld 1996; Gale 1994), and that customer satisfaction is the customers’ reaction to the value received (e.g., Woodruff 1997), empirical examination of these relationships has received limited attention (e.g., Lemmink, de Ruyter and Wetzels 1998). Adopting Hartman’s (1967) conceptualization of value G.e., emotional, practical, and logical value dimension), Lemmink and colleges provided evidence to support that value impacts customer satisfaction. In addition, a related stream of literature on value attainment and employee satisfaction provides some empirical evidence that value leads to satisfaction (George and Jones 1996). The results of these studies show that employees who attain high value from the job tend to be more satisfied with the job. Similarly, we have suggested that the concept of value subsumes several of the anetecedents to relationship continuance proposed by social exchange theory and the investment model (i.e., benefits, costs, and alternative comparisons). Both the investment model and the social exchange theory indicate that outcome values (i.e., benefits received minus costs paid) of the current
11
‘V
-
relationship positively affect satisfaction with the relationship. Applying these theories, customers who receive high value from a supplier should be more satisfied than those who receive low value. Thus, customer value is hypothesized to have a positive relationship with customer satisfaction. Hypothesis four is as follows: H4:
An organizational buyer’s perceived level of customer value with a supplier will be positively related to the buyer’s level of satisfaction with the supplier.
RESEARCH DESIGN Research Context The financial staffing service industry provided the context in which to test the research hypotheses. The financial staffing industry is a highly competitive oligopoly with approximately 8O0/o of the market being serviced by six main providers. Prices for financial staffing services have become fairly standardized and, thereby, less influential in purchase decisions. However, service providers have been differentiating themselves through the quality of financial staff the ability to meet the needs of customers, and the relationships that account managers form with key customers. Market demands fluctuate rapidiy and are highly unpredictable; thus, organizational buyers typically employ services from several providers to ensure access to staffing when needed. The high level of multiple-sourcing practices among financial staffing users provides an ideal research setting to investigate the “share-of-business” organizational buyers’ give to their suppliers (Jackson 1985).
Scale Development Both qualitative and quantitative research were conducted to develop measurement scales. Research assistance was solicited from one large financial staffing provider who provided access to account managers (4), customers (5), and sales and marketing executives (2) for eleven qualitative interviews. These initial interviews addressed issues related to organizational purchasing criteria in
12
the competitive, multiple outsourcing financial staffing industry as well as the concepts of value, satisfaction, and switching costs. For space reasons, findings from these interviews are not presented here, but insights from these interviews were used in the development of measurement items. Qualitative interview findings were evaluated and integrated with scales used in past studies to measure customer satisfaction (e.g., Patterson et al. 1997; Oliver and Swan 1989), value perceptions (e.g., Anderson and Thomson 1997; Dodds et al. 1991), switching costs (e.g., Heide and Weiss 1995; Ping 1994), and share-of-business ~ackson 1985), in order to develop a pool of measurement items. Items were subjected to the scrutiny of four judges; two marketing professors, and two sales and marketing executives previously interviewed. These judges were asked to assess the face validity of each item to determine if a measure “looks as if’ it should indicate a particular concept (Heeler and Ray 1972). Ambiguous or inappropriate items were eliminated. All scales were multi-item and items were seven-point Likert scaled and anchored by “strongly disagree” (1) and “strongly agree” (7). For each construct, Table 2 shows the scale items, standardized estimates, scale reliabilities, and measurement model results for both the scale-development stage and the main study. Scale means, standard deviations, and the correlations among the latent constructs appear in Table 3. Statistical tests assessing dimensionality and discrirninant validity are reported in Table 4.
Measures Share-ofbusiness allocation intention refers to amount of business an organizational buyer intends to give a supplier relative to other suppliers the buyer is currently outsourcing with. Four items were developed based on Jackson’s (1985) proposition of share-of-business among multiple suppliers and insight from qualitative interviews. CustomerSatisfaction pertains to an organizational buyer’s overall post-purchase evaluation toward the service supplier. The current study takes a cumulative rather
13
than a transaction specific perspective. Three items were adapted from Oliver and Swan (1989) and Patterson et al. (1997) to measure this construct. Perceived switching costs refer to the costs associated with terminating a current supplier and establishing a new relationship with a replacement supplier (Ping 1994). Four items were used to measure this constmct based on Heide and Weiss’s (1995) and Ping’s (1994) switching costs scale. Customerperceived value pertains to an organizationalbuyer’s evaluation of the costs and benefits of a certain supplier relative to alternatives available. Conceptually, it is a tradeoff judgement among perceived benefits, costs, and close substitutes (Grisaffe and Kumar 1998; Gale 1994). Based on findings from qualitative interviews, customer value for business services encompasses three dimensions: the overall economic value of a supplier, the technical value of the core service, and the value of the working relationship. These dimensions are similar to those proposed by Anderson and Thomson (1997). Economic value refers to an organizational buyer’s overall cost/benefit assessment of a supplier relative to alternative suppliers. Value ofthe core service refers to the technical performance of a current supplier relative to alternative suppliers. Value ofthe relational/support service pertains to the effectiveness of the working relationships with one supplier relative to alternative suppliers. Four items were developed to measure each of these dimensions based on qualitative interviews and on items developed by Dodds and colleagues (1991). Customer perceived value is conceptualized as the higher-order construct represented by these three dimensions.
Quantitative Assessment ofMeasurementScales The scale-development questionnaire was sent by mail to a sample of 735 firms using financial staffing services in the Florida, Georgia, and Washington D.C markets. This sample was provided by a database of clients and prospects maintained by the assisting financial staffing firm. The mailing included the survey, along with a cover letter requesting cooperation and assuring the
14
A
—
academic use of study findings, a $1.00 bill for an incentive, and a self-addressed, prepaid return envelope. Two weeks after the initial mailing, a reminder postcard was sent requesting assistance. Six weeks after the surveys were mailed, 17 surveys had been returned undeliverable, 22 were returned by informants indicating their inability to participate, and one-hundred eighty-eight useable questionnaires were received for an effective response rate of 27~/o (188/696). Although the response rate appears low, it is consistent with recent studies investigating business-to-business relationships (Heide and John 1990) and services (Mishra, Heide and Cort 1998; Zeithaml, Berry and Parasuraman 1996) A single key informant was contacted. Although Philips (1981) criticizes mono-method approaches, such as the single key informant method, and demonstrates that they often lack the ability to provide convergent and discriminant validities for measures, most researchers find the key informant approach efficient, and with little self-report bias (Silk and Kalwani 1982). Various empirical evidence supports key informant techniques, and suggests that they can provide reliable and valid data (John and Reve 1982; Silk and Kalwani 1982). Screening criteria were employed in this study to ensure the qualification of the key informants. Items were adapted from Kumar, Stern, and Anderson’s (1993) competency questionnaire and from Patterson, Johnson and Spreng’s (1997) decision involvement questions. To investigate nonresponse bias, mean comparisons were conducted between early and late respondents (Armstrong and Overton 1977). Analyses of early versus late responses were conducted on thirty-five items, and two significant differences were found. Early respondents were found to spend more on financial staffing services than late respondents (p
=
.04), and tend to order
the service through human resource department more than late respondents (p
.03). Because one
or two significant findings were expected by chance at an alpha of 0.05, and no pattern of significant differences emerged, early and late respondents were assumed to be similar. This provides some
15
evidence that non-response bias does not substantially threaten validity (Armstrong and Overton 1977). A procedure was used to assess the dimensionality, reliability, and validity of the multi-item scales, based on the guidelines of J6reskog (i.e., Jdreskog 1973; Jdreskog and Sbrbom 1979), James, Mulaik and Brett (1982), and Anderson and Gerbing (1988). This procedure involves the evaluation of the items in an “unrestricted model” (J6reskog 1973; Jdreskog and Sorbom 1979), and then the evaluation of items through a confirmatory factor analysis (i.e., measurement) model. Prior to evaluating models including all measurement items, items representing the three theorized dimensions of customer value were evaluated independently of those measuring other concepts. The Unrestricted ModeL The unrestricted model specifies r2 constraints (fixed parameters), where r is the number of latent common factors. Constraints render the model identified for a specified number of factors, and no others, while placing no specific constraints on how manifest variables are related or unrelated to the latent variables. According to Joreskog and Sorbom (1979), to specify an unrestricted model, one measurement item associated with each latent construct is allowed to load freely on that construct while the (r-1) loadings of that item on the remaining latent constructs are fixed to zero. This has the effect of fixing r(r-1) loadings to zero in the factor loading matrix. The remaining loadings in the factor-loading matrix are made free parameters. In addition, each of the r common factor variances is fixed to unity, while the correlations between the common factors are all made to be free parameters. The initial chi-square statistic of the three-factor unrestricted model with the 12 items measuring customer value was statistically significant
(x2
54.80, 33 d.f., p < 0.05), suggesting that,
after accounting for three factors, there remains further non-zero covariation among the manifest variables. To identify problematic items associated with the significance of the unrestricted model, a three-factor confirmatory factor analysis (CFA) model was specified and evaluated
16
(x2 =
110.27, 51
d.f., CFI
=
0.94, GFI
=
0.89, TLI
=
.92,
RMSEA = 0.09). Judicious re-specification procedures
prescribed by James, Mulaik and Brett (1982), and Anderson and Gerbing (1988) identified two items for removal (see Table 2). A three-factor unrestricted model incorporating the remaining ten items fit the data 24.91, 18 d.f., p
=
(x2 =
0.13, n.s.). To further examine the appropriateness of the three-factor
unrestricted model, the three-factor model was compared to unrestricted models with exactly two and four-factors. The results show that the fit of the three-factor unrestricted model was statistically superior to a two-factor model, and not statistically inferior to a four-factor model based on chisquare difference statistics. The MeasurementModeL A measurement model was specified to constrain the ten items to load on three factors as theorized. A chi-square difference statistic between the unrestricted model and the measurement model shows that imposing these additional measurement constraints did not significantly reduce the fit of the model. Additionally, the measurement model showed good overall fit (x2
=
47.09, 32 d.f., CFI
=
0.98, GFI
=
0.95, TLI
=
.97, RIVISEA = 0.05). Together, the
evaluation of the three-factor unrestricted model and measurement model provides evidence of discrimanant and convergent validity for the three dimensions of value1. Next, a six-factor unrestricted model was estimated using the ten retained customer value items, and the eleven items underlying customer satisfaction, perceived switching costs, and shareof-business intention
(~2
=
102.96, 99 d.f., p
=
0.37, n.s.). To further examine the appropriateness
of the six-factor unrestricted model, it was compared to five-factor and seven-factor unrestricted models. A chi-square difference statistic indicated that the fit of the seven-factor unrestricted model
1
Measurement estimates were found to be significant and load on their intended construct indicating convergent validity
(Bagozzi and Yi 1988). The non-significant chi-square difference statistic between the unrestricted and measurement models suggests that the relationships among measurement items and the latent constructs they were not intended to represent are not significantly different from zero. Thus, this finding provides a test of discriminant validity.
17
was statistically better than the hypothesized six-factor model (AX2
31.39, d.f.
=
15, p K 0.05).
Therefore, a six-factor CFA model was evaluated to identify potentially problematic items. The evaluation of the initial CFA model RMSEA
(x2
206.69, 174 d.f., CFI
0.98, GFI
0.89, TLI
.98,
0.03) suggested the removal of two items: one underlying perceived switching costs and
one underlying share-of-business intentions (see Table 2). The six-factor unrestricted model with the 19 retained items fit the data well, was statistically better fitting than the five-factor model, and did not fit statistically worse than a seven-factor model. In addition, the imposed parameter constraints placed on the 19-item, six-factor measurement model did not significantly reduce overall model fit from the unrestricted model, which indicated that items load only on the latent constructs they were developed to measure. The overall fit statistics of the measurement model indicated good fit (x2 GEl
0.92, TLI
.99, RMSEA
142.65, 137 d.f., p
0.35, CFI
0.99,
0.02).
All items were found to load strongly on the intended latent construct, except for one item measuring overall economic value of a supplier (see Table 2). This item was negatively framed and thereby differed from the other items measuring this construct. Scale reliabilities were above .80 for all scales except for the economic value of a supplier. The reliability of this scalewas above .67, which has been deemed acceptable for new scales (i.e., Nunnally 1978)
QuestionnaireAdministration and Sampling Procedure Items from the scales evaluated in the scale-development stage of this study were included on a second survey instrument mailed to a national sample of organizations using financial staffing services. In an attempt to enhance the reliability of the economic value scale, the negatively framed item was reworded. Instead of reading: “compared to other suppliers, this supplier provides low quality for the price”, it was changed to read:
“...
this supplier provides better quality for the price.”
18
The mailing list was obtained from a client and prospect database maintained by the same large financial staffing provider. The sample included organizational buyers using financial staffing from all fifty states except Florida, Georgia, and the metropolitan Washington D. C. area. The revised survey, along with a cover letter, and a self-addressed, prepaid return envelope were mailed to 900 financial staffing customers nationwide. Competency questions were included to qualify key informants’ involvement and experience with the financial staffing selection process. Two weeks after the initial mailing, a reminder postcard was sent requesting cooperation. Six weeks after the initial surveys were mailed, 21 surveys were returned undeliverable, and 27 were returned by informants indicating their inability to participate. A total of 209 surveys were completed and found to be useful for main study data analysis for a response rate of 24% (209/852). Mean comparisons found no significant differences between early and late respondents. By using late respondents as a proxy for non-responding business customer, a level of certainty is attained that the level of non-response bias was low (Armstrong and Overton 1977). Table 5 provides a demographic profile of informants and their firms. A systematic four-step approach was taken to evaluate measurement issues and to test proposed hypotheses. The fit of the unrestricted model was assessed in step-one, and the fit of the measurement model was evaluated in step-two. Thus, the first two procedures mirror those conducted with the scale-development data set. Step-three involved the evaluation of the fit of the structural model, and the final step was to test hypotheses pertaining to the non-zero relationships among latent variables.
Unrestricted Model Results The six-factor unrestricted model was specified with parameters from all but six measurement items (i.e., one item associated with each latent construct) free to load on each latent
19
construct. For each of these six items, the parameter from its associated latent constmct was specified as a free parameter and all others were fixed to zero (Jdreskog and Sdrbom 1979). The sixfactor unrestricted model was estimated using 19 items: ten measuring the three dimensions of customer value, three measuring customer satisfaction, three measuring perceived switching costs, and three measuring share-of-business intention
(z2 =
65.85, 72 d.f., n.s.). A chi-square difference
statistic indicated that the fit of the six-factor unrestricted model was statistically better fitting than the five-factor model, and did not fit statistically worse than a seven-factor model.
MeasurementModel Results The overall fit statistics of the measurement model indicated good fit (z2 p
=
0.08, CFI
=
0.99, GFI
=
=
160.52, 137 d.f.,
0.92, TLI = .99, RMSEA = 0.03). A chi-square difference statistic
between the measurement model and the unrestricted model was not significant at an alpha of 0.10, which suggests that the imposed parameter constraints placed on the measurement model did not significantly reduce overall model fit from the unrestricted model. This indicates that measurement items did load only on the latent constructs they were developed to measure. In addition, all items were found to load strongiy on the intended latent construct. Scale reliabilities were above .82 for all scales. Because the underlying dimensionality of the items was assessed with the unrestricted model, high reliability coefficients provide further evidence of the unidimensionality of latent constmcts (Cortina 1997). Higher-Order FactorMeasurementModeL Next, a second-order construct was integrated into the measurement model to represent the overall concept of customer value. The measurement model was respecified so that the three first-order latent value dimensions G.e., value of the core service, value of support service, and economic value of a supplier) were allowed to underlie a second-order latent construct. In this model, the covariences between the second-order construct (i.e., customer
20
value) and the three first-order latent constructs (i.e., customer satisfaction, perceived switching costs, and share-of-business intentions) were freely estimated. The nine covariences between the three value dimensions and the three first-order latent constructs were each fixed to zero. These added constraints, associated with the theorized second-order conceptualization of customer value, did not significantly reduce the fit of the measurement model (A~2
=
11.81, d.f.
=
6, n.s.).
Therefore, conceptualizing value as a second order construct provides added parsimony, while the constraints placed on the model did not result in lack of model fit.
StructuralModel Results Next, constraints of the theoretical model were imposed. Structural parameters were allowed to vary freely or were fixed to zero in order to impose the hypothesized pattern of relationships in the proposed model. The results of the structural model showed that the proposed second-order stmctural model fit well with the data (CEI .03, x7
=
177.70 and df = 144, p
=
=
.99, GFI
=
.92, TLI
.99, RMSEA =
=
0.03; see Figure 2). All hypothesized parameters were in the
proposed direction. In addition, a chi-square difference statistic between the measurement model and the structural model was non-significant at alpha of 0.10 (AX2 indicates
=
17.18, A d.f.
=
7). This
that imposing the structural constraints did not significantly reduce model fit.
TESTS OF HYPOTHESES Statistical test of hypothesized non-zero parameters were conducted by examining the associated parameter z-scores. Because these hypotheses are tested simultaneously in the stmctural model, a multi-stage Bonferroni procedure involving a more conservative significance level test was imposed. This procedure mimicked that developed by Larzelere and Mulaik (1976) for tests involving multiple correlations. 21
A total of 13 first-order freely estimated measurement parameters, two second-order freely estimated measurement parameters, and five structural parameters were being evaluated simultaneously (see Table 6). As such, the first stage of the Bonferroni multi-stage evaluation had to adjust for a family of 20 tests. The corresponding critical z-value was 2.24 (cc
=
0.05/20/2
=
0.00125). Each parameter in the model was found to be significantly different from zero at this level; rendering further stages unnecessary (Larzelere and Mulaik 1976)2. Thus, all hypotheses are supported
in stage one. Customer value was positively related to share-of-business allocation 5.88, p
intentions (z
intentions (z =
=
