Modeling Heterogeneity in the Satisfaction,Loyalty Intention,and Shareholder Value Linkage: A Cross-Industry Analysis at the Customer and Firm Levels

Abstract

This study examines the relationship between customer satisfaction, loyalty intention, and shareholder value at the firm and individual customer levels. The authors also explore industry differences by using a multilevel and random-effects approach in which individual customer scores are nested within firm-level data and the estimated interrelationships are treated as random coefficients that are explained by industry characteristics. They compile a unique and detailed data set, which covers 10 years of information on 137 firms and includes a matched sample of 189,069 customers from multiple sources, such as the American Customer Satisfaction Index, the Center for Research in Security Prices, and Compustat, to yield three important insights. First, aggregate firm-level effects may overestimate the impact that satisfaction has at the individual customer level. Second, a consideration of loyalty intention or repurchase intention as the mediator can improve our understanding of the satisfaction–shareholder value relationship and the fact that this relationship can vary across firms. Finally, the influence of satisfaction and loyalty intentions on shareholder value varies by industry. The authors discuss implications of findings for researchers, managers, and investors.

Keywords

customer satisfaction loyalty intention shareholder value heterogeneity industry effects

Marketing executives consistently rank customer satisfaction among their most important priorities (e.g., Aksoy 2013; Anderson Analytics 2009, 2010), and many companies measure and monitor satisfaction at the customer level to understand which customers are satisfied and why. In addition to these individual-level measures, institutions like the American Customer Satisfaction Index (ACSI) track customer satisfaction at the firm level. Academic researchers, managers, market analysts, and investors have used these publicly available aggregate scores to compare satisfaction scores with industry peers or competitors and to perform cross-industry benchmarking with companies they deem “best in class” (www.theacsi.org).

Customer satisfaction (at both the individual and firm levels), however, is typically not in itself the ultimate goal for most managers. The focus on customer satisfaction is driven in large part by the widespread belief that satisfaction leads to customer retention, which in turn results in improved revenue and/or market share (e.g., Anderson and Mittal 2000; Danaher and Rust 1996; Heskett, Sasser, and Schlesinger 1997; Parasuraman, Zeithaml, and Berry 1988; Rust and Zahorik 1993; Rust, Zahorik, and Keiningham 1995; Zeithaml, Berry, and Parasuraman 1996). At the individual customer level, satisfaction is expected to translate into customer loyalty (Aksoy 2013), engagement behaviors (Brodie et al. 2011), and increased customer lifetime value (Rust, Lemon, and Zeithaml 2004; Schulze, Skiera, and Wiesel 2012; Wiesel, Skiera, and Villanueva 2008). Conceptual chain-of-effects models like SERVQUAL (Parasuraman, Zeithaml, and Berry 1988; Zeithaml, Berry, and Parasuraman 1996), return on quality (Rust, Zahorik, and Keiningham 1995), the service–profit chain (Heskett, Sasser, and Schlesinger 1997), and the satisfaction–profit chain (Anderson and Mittal 2000) empirically test the linkages between customer satisfaction and firm performance using the individual customer as a unit of analysis. Customer satisfaction is expected to translate into firm outcomes operationalized at the customer level (e.g., customer profitability; Larivière 2008) through the mediation of customer loyalty. At the firm level, satisfaction is presumed to result in improved shareholder value (SHV).

Prior research has focused on either satisfaction scores within a single firm, in which the individual customer is the unit of analysis (literature stream 1), or aggregate satisfaction scores—typically the ACSI—in which different companies represent the units of analysis (literature stream 2), but generally not both in the same study. Therefore, to better understand the consequences of customer satisfaction, we analyze both individual-level data (i.e., customer-level ACSI and loyalty data, which is not publicly available) and aggregate firm-level data (publicly available ACSI data) for the first time to provide a finer and more granular empirical examination of the ACSI, loyalty intentions, and SHV linkage at both the individual customer and firm levels.

This two-level approach is important for both managers and investors because many managerial decisions and interventions (e.g., types of strategies firms pursue or types of information technology systems firms use) that affect these linkages occur at the customer level. As such, it is important to understand how any insights gleaned using aggregate-level data hold at the individual level¹; more importantly, these effects may be heterogeneous across different industries (e.g., the impact of satisfaction on SHV can deviate from the significance level and the magnitude of the estimated population-averaged effect). Therefore, in this study, we estimate the coefficients of satisfaction and loyalty as random effects for the relationships at both the customer and firm levels; and we further explore and explain these varying or random effects as they relate to industry characteristics (see Gruca and Rego 2005) and research design decisions (e.g., the time lags between satisfaction and loyalty intention). Figure 1 presents the conceptual framework for this study.

Figure 1

CONCEPTUAL FRAMEWORK

To preview the key insights from our comprehensive and granular analysis: We find that for some firms, the influence of satisfaction at the firm level is more pronounced than its impact at the customer level. In other words, the influence of customer satisfaction within, versus between, firms is likely to be different, such that managers cannot rely solely on insights obtained from an analysis of the ACSI–SHV linkage conducted at the firm level or rely only on ACSI releases when making decisions that relate to individual customers or segments. Our results also demonstrate that for some industries, it is equally or even more important to consider loyalty intention in addition to satisfaction. Moreover, it is crucial to take industry characteristics into account to disentangle effects for various types of firms, including the potential mediating role of loyalty intentions. Together, modeling heterogeneity in satisfaction, loyalty intention, and SHV linkage and explaining this randomness in terms of industry characteristics enables us to better interpret some mixed findings from prior studies.

Theoretical Background

Satisfaction, Repurchase Intention, and SHV

Satisfaction and behavioral intentions are some of the most widely measured metrics used by managers, especially in the quantification and management of customer loyalty (Aksoy 2013). In fact, practitioners often treat the combination of very high satisfaction ratings and high intention ratings as loyalty (e.g., Baker-Prewitt 2014). “Satisfaction” is defined as the degree to which the level of fulfillment was pleasant or unpleasant (Oliver 2010, p. 23), representing a more backward-looking metric (Zeithaml et al. 2006), whereas intentions denote a consumer's expected likelihood of engaging in a behavior in the future, therefore representing a forward-looking metric.

Seminal conceptual frameworks like SERVQUAL (Parasuraman, Zeithaml, and Berry 1988; Zeithaml, Berry, and Parasuraman 1996), return on quality (Rust, Zahorik, and Keiningham 1995), the service–profit chain (Heskett, Sasser and Schlesinger 1997), and the satisfaction–profit chain (Anderson and Mittal 2000) conceptualize a chain of effects in which loyalty mediates the relationship between customer satisfaction and firm outcomes, suggesting an indirect link. In addition to these frameworks, other studies have provided similar empirical support (e.g., Larivière 2008; Rust and Zahorik 1993) for the satisfaction → loyalty → firm performance relationship. These studies generally tend to use individual customer-level data to investigate these linkages.

At the aggregate firm level, beginning with Ittner and Larcker (1998), who conducted one of the earliest studies to explore the link between satisfaction and a firm's market value, there has been a great deal of scientific inquiry into the relationship between customer satisfaction and SHV (see Table 1). Although many of these studies have found a direct and positive association between customer satisfaction and SHV creation, others have reported mixed findings. For example, our literature review in Table 1 reveals that 10 (out of 22) studies find evidence of a positive relationship between satisfaction and SHV, whereas the remaining 12 studies report mixed findings.² Table 1 also demonstrates that the indirect chain-of-effects approach that seminal conceptual frameworks have taken at the individual customer level has not been widely adopted by studies at the aggregate firm level. We found two studies that test for the indirect impact of satisfaction on SHV, albeit using variables other than repurchase intention. In a study of airline companies, Grewal, Chandrashekaran, and Citrin (2010) find that sales mediates the relationship between satisfaction and SHV. Since repurchase intentions are related to sales and, in turn, creation of customer equity and customer lifetime value (Gupta, Lehmann, and Stuart 2004; Kumar and Shah 2009), these findings further support the expectation that repurchase intentions will mediate the satisfaction–SHV relationship at the firm level as well. Among studies reviewed, only two (Keiningham et al. 2007; Morgan and Rego 2006) employ both satisfaction and repurchase intentions in the same study when linking to SHV measures. Both studies, however, link satisfaction and repurchase intentions to SHV outcomes directly.

Table 1

SUMMARY OF EMPIRICAL STUDIES THAT INVESTIGATE THE INFLUENCE OF CUSTOMER SATISFACTION AND/OR LOYALTY INTENTIONS ON SHV AT THE AGGREGATE FIRM LEVEL

		Sample Compilation and SHV Metrics		Research Issues Relating to SAT			Research Issues Relating to LI			Analysis and Findings
Study Number	Source	Industries	SHV Metrics	SAT Data	SAT Modeled at Individual Customer Level?	Random Coefficient for SAT (Variation in the Influence of SAT)?	LI Measured?	Random Coefficient for LI (Variation in the Influence of LI)?	Type of Relationship Tested for the Impact of SAT on SHV	SAT Has Impact on SHV?	Random Coefficients Explained by Firm/Industry Characteristics ?
1	Aksoy et al. (2008)	Various industries (ACSI)	Tobin's q, CAPM, FF3, C4	ACSI (1996-2006)	No	No	No	N.A.	Direct	Yes	N.A.
2	Anderson, Fornell, and Rust (1997)	Various industries (SCSB)	ROI	SCSB (1989-1992)	No	No	No	N.A.	Direct	Yes	N.A.
3	Fornell et al. (2006)	Various industries (ACSI)	MVE	ACSI (1994-2002)	No	No	No	N.A.	Direct	Yes	N.A.
4	Ivanov, Kissan, and Babajide (2013)	Various industries (ACSI)	Cumulative abnormal return	ACSI (1995-2006)	No	No	No	N.A.	Direct	Yes	N.A.
5	Luo and Homburg (2008)	Single industry: airline companies (from ACSI database)	Stock value gap to the best performing competitors	ACSI (1999-2006)	No	No	No	N.A.	Direct	Yes	N.A.
6	Merrin, Hoffmann, and Pennings (2013)	Various industries (ACSI)	Stock returns	ACSI (1994-2010)	No	No	No	N.A.	Direct	Yes	N.A.
7	Morgan and Rego (2006)	Various industries (ACSI)	Tobin's q, net operating cash flows, total shareholder return, annual sales growth, gross margin, market share	ACSI (1994-2000)	No	No	Yes	No	Direct	Yes	N.A.
8	Tuli and Bharadwaj (2009)	Various industries (ACSI)	FF3	ACSI (1994-2006)	No	No	No	NA	Direct	Yes	N.A.
9	Grewal, Chandrashekaran, and Citrin (2010)	Single industry: airline companies (from ACSI database)	Tobin's q	ACSI (1997-2005)	No	No	No	N.A.	Indirect (through sales and servicing costs)	Yes	N.A.
10	Luo, Homburg, and Wieseke (2010)	Various industries (ACSI)	Abnormal return, systematic risk, idiosyncratic risk	ACSI (1995-2006)	No	No	No	N.A.	Indirect (through analyst stock recommendations)	Yes	N.A.
11	Ittner and Larcker (1998), Study 1	Single industry, at business level: branches from a financial services provider	Revenues, expenses, margins, return on sales, number of customers	SAT (1995-1996)	No	No	No	N.A.	Direct	Mixed findings	N.A.
12	Ittner, Larcker, and Taylor (2009)	Various industries (ACSI)	Future operating performance, abnormal long-run returns	ACSI (1995-2006)	No	No	No	N.A.	Direct	Mixed findings	N.A.
13	O'Sullivan, Hutchinson, and O'Connell (2009)	Various industries (ACSI)	CAPM, FF3, C4	ACSI (1994-2006)	No	No	No	N.A.	Direct	Mixed findings	N.A.
14	Rego, Morgan, and Fornell (2013)	Various industries (ACSI)	Future market share	ACSI (1994-2006)	No	No	No	N.A.	Direct	No, unless SAT is computed relative to that of firm's rival and customer switching costs are low	N.A.
15	Anderson, Fornell, and Mazvancheryl (2004)	Various industries (ACSI)	Tobin's q, equity price, ratio of price to book value	ACSI (1994-1997)	No	Yes	No	N.A.	Direct	Yes, although a variation in the association between SAT and SHV across firms and industries is found	No
16	Anderson and Mansi (2009)	Various industries (ACSI)	Credit ratings, yield spread	ACSI (1994-2004)	No	Yes	No	N.A.	Direct	Yes, although a variation in the association between SAT and SHV across firms and industries is found	No
17	Graca and Rego (2005)	Various industries (ACSI)	Cash flow growth and variability	ACSI (1994-2002)	No	Yes	No	N.A.	Direct	Yes, although a variation in the association between SAT and SHV across firms and industries is found	Yes
18	Mittal et al. (2005)	Various industries (ACSI), restricted to firms that achieve a dual emphasis (focusing on both revenue expansion and cost reduction simultaneously)	Tobin's q, annualized stock returns	ACSI (1994-2000)	No	Yes	No	N.A.	Direct	Mixed findings	No
19	Ittner and Larcker (1998), Study 2	Various industries (ACSI)	MVE	ACSI (1995-1996)	No	Yes, but through within-industry regressions (subsamples of data; thus, not throughrandom coefficients)	No	N.A.	Direct	Yes, positive impact for the population-averaged effect (Table 6, p. 23); mixed findings for different industries in isolation (Table 7, p. 27)	N.A.
20	Jacobson and Mizik (2009)	Various industries (ACSI)	C4, various portfolio approaches	ACSI (1996-2006)	No	Yes, but through within-industry analyses (subsamples of data; thus, not through random coefficients)	No	N.A.	Direct	Mixed findings for different industries in isolation, which may deviate from the estimate for all firms (e.g., Table 2, p. 6)	N.A.
21	Keiningham et al. (2007)	Various industries (Norwegian Customer Satisfaction Barometer)	Change in revenue	NCSB (2001-2005, 2000-2003)	No	Yes, but through within-industry analyses (subsamples of data; thus, not throughrandom coefficients)	Yes	No	Direct	Mixed findings	N.A.
22	Raithel et al. (2012)	Single industry: cars	C4	Three SAT dimensions from J.D. Power survey data (U. S., U.K., and German markets)	No	Yes, but through within-country analyses (subsamples of data; thus, not throughrandom coefficients)	No	N.A.	Direct	Mixed findings, due to country differences	N.A.
The current study		Various industries (ACSI)	Tobin's q, C4	ACSI (2000-2009)	Yes	Yes	Yes	Yes	Mediation analysis: (1) direct, (2) indirect through the mediation of LI, (3) both direct and indirect, (4) none	Mixed findings: no impact (C4) or indirect impact (Tobin's q) for the population-averaged effect; mixed findings across firms, depending on industry characteristics, the role of LI, and research design (time lag between SAT and LI)	Yes

Notes: SHV = shareholder value; SAT = satisfaction; LI = loyalty intentions; ACSI = American Customer Satisfaction Index; SCSB = Swedish Customer Satisfaction Barometer; FF3 = Fama-French three-factor model; C4 = Carhart four-factor model; ROI = return on investment; MVE = market value of equity; N.A. = not applicable.

Aggregate- versus Individual-Level Effects

When examining the consequences of customer satisfaction on SHV, it is important to determine whether results vary as a function of the level of data used. Zhang, Zyphur, and Preacher (2009) caution against confounding in estimation effects, which can arise at different levels of analysis, and Freedman, Pisani, and Purves (1998) note that the strength of interrelationships is expected to be greater when analyzed at the firm level (i.e., aggregate level) versus the individual level. They argue this occurs because averaging has the potential to mask individual-level variability. Furthermore, the extant literature on the satisfaction–SHV linkage (see Table 1) either studies satisfaction scores within a single firm, whereby the individual customer is the unit of analysis, or employs aggregate satisfaction scores, such as the ACSI, whereby different companies represent the units of analysis. Given the possibility of arriving at disparate conclusions from analyses using differing levels of data, it becomes important to establish the consistency and applicability of insights gained from studying different levels simultaneously. Therefore, this study tests the linkage between satisfaction and SHV by comparing results at the individual customer and aggregate firm levels.

The Influence of Industry Characteristics

As Table 1 reveals, different sample compositions have been used in the literature to date to investigate the relationship between satisfaction at the aggregate firm level and SHV. As such, it is difficult to discern to what extent the portfolio of firms chosen influences the estimated effect that satisfaction has on SHV for the “averaged firm.” To explore these potential effects, in this study, we allow each firm to have its own firm-specific coefficient (or estimate) and explore different industry characteristics as potential explanatory drivers of these random coefficients, consistent with the approach used by Gruca and Rego (2005). We explore three industry characteristics: goods/services classification (e.g., Anderson, Fornell, and Rust 1997), purchase horizon (e.g., Li, Xu, and Li 2005), and durable/nondurable classification (e.g., Larivière and Van den Poel 2004).

In summary, this study contributes to the extant literature on the connections between satisfaction, repurchase intentions, and SHV by (1) examining the links among ACSI, loyalty intentions, and SHV at both the individual customer and aggregate firm levels; (2) employing a multilevel approach in which individual satisfaction scores are nested within aggregate scores (e.g., ACSI) with the aim of unveiling the extent to which insights obtained from aggregate-level investigations can be translated to individual-level contexts; and (3) helping to clarify the extent to which the effects of satisfaction and loyalty intentions on SHV are dependent on industry characteristics.

Data and Research Methodology

Data Sources

The ACSI measures and monitors customer satisfaction and repurchase intention scores for 10 economic sectors, 43 industries, and more than 200 companies and federal and local government agencies.³ The measured companies are broadly representative of the U.S. economy and correspond to roughly 43% of the nation's GDP.

The ACSI publishes customer satisfaction data on an annual basis, with results for a specific set of sectors in each calendar quarter replacing data collected the previous year. There is a lag in the schedule of update releases: typically, results are made publicly available on the third Tuesday of February (fourth-quarter sector results from the previous calendar year), May (first-quarter results), August (second-quarter results), and November (third-quarter results).⁴ For example, the 2006 report on the Coca-Cola Company, a part of the soft drinks industry, was released toward the end of November 2006, replacing the data reported a year prior. Therefore, the first full month for which the 2006 Coca-Cola score is available is December 2006. The repurchase intention scores, on the other hand, are not publicly distributed, although the ACSI collects them at the same time as the satisfaction data.

For our preliminary analysis, we received data on both customer satisfaction and repurchase intentions at the aggregate firm level from ACSI. We matched these data with stock returns from the Center for Research in Securities Prices (CRSP) and with nonnegative book values of equity available from the annual Merged Compustat data, also maintained by CRSP, for a subsample of NYSE-, AMEX-, and NASDAQ-listed U.S. firms that issue ordinary common shares. In addition, we also obtained other accounting data items, such as sales or book value of assets, which are required to calculate profitability measures or valuation multiples, from the Compustat data set and linked them to our data set. There is a difference of approximately two months between the end of a Compustat fiscal year and the actual report date of financial statements for the sample of firms in our study. Thus, we used lagged Compustat data in our analysis, just as with the marketing variables of interest. For example, accounting data for the Coca-Cola Company for the fiscal year ending in December 2006 are considered valid from March 2007 to February 2008. In this study, we use two different SHV metrics: Tobin's q and the Carhart four-factor model (for detailed information on these metrics, see Web Appendix A).

To examine individual-level data (i.e., customer-level satisfaction and repurchase intentions), we accessed the raw data files from which the ACSI is calculated. Because this information is considered proprietary and nonpublic, we received access to the data after we met specific conditions to ensure the data's security.

Sample Description

The final data merged cover the period from June 2000 to December 2009, describing 137 unique firms and a matched sample of 189,069 surveys at the individual customer level. Table 2 provides summary information on size, profitability, and measures of customer satisfaction and repurchase intentions for the sample of firms included in our study.

Table 2

SAMPLE CHARACTERISTICS

A: Summary Statistics
	Mean	Median	SD	10th Percentile	90th Percentile
Book value of assets (billions of dollars)	73.51	17.93	219.27	4.25	133.83
Annual sales (billions of dollars)	27.26	12.43	39.89	4.06	61.25
Return on assets (%/year)	9.57	8.21	6.42	2.62	18.84
Tobin's q (market to book value of assets)	1.47	1.03	1.19	.59	2.98
Satisfaction (range: 1–100)	75.59	76	6.05	68	83
Repurchase intentions (range: 1–10)	8.09	8.06	.68	7.23	8.99
B: North American Industry Classification System (NAICS) Classification of Sectors and Industries
NAICS Code	Sector Definition		Firm-Year Observations		Percentage of Total
22	Utilities		227		24.51
31–33	Manufacturing		267		28.83
44–45	Retail trade		163		17.60
48–49	Transportation and warehousing		51		5.51
51	Information		76		8.21
52	Finance and insurance		90		9.72
72	Accommodation and food services		52		5.62
Total			926		100.00

Testing for Within-Group Agreement

Because customers are nested within firms (i.e., for each firm, a sample of customers rated the company with respect to their customer satisfaction and repurchase intentions), for each firm, we have a sample of individual customer scores that can be grouped. The term “within-group agreement” refers to the degree to which ratings from individuals are interchangeable (Bliese 2000), that is, the degree to which raters provide essentially the same rating (Kozlowski and Hattrup 1992; Tinsley and Weiss 1975).

In this study, “groups” refers to firms, such that within-group agreement denotes a within-firm agreement and reflects the extent to which different customers provided similar satisfaction and repurchase intention scores for the same firm. In case there exists agreement within firms (groups) about both satisfaction and repurchase intentions, the individual responses can be aggregated at the firm level (group level), and these aggregate scores (ACSI and firm-averaged repurchase intentions) can be further employed in the analyses. In case of disagreement, it is essential to account for both levels (individual customer level and aggregate firm level) by means of a multilevel model (Klein and Kozlowski 2000), such that the impact of both satisfaction and repurchase intentions can be studied at both the individual customer level (Level 1) and the aggregate firm level (Level 2).

To test within-firm agreement, we use the r_wg and a_wg agreement indexes proposed by James et al. (1984) and Brown and Hauenstein (2005), in which values ≥.70 are considered indicators of good agreement (Bliese 2013). The r_wg and a_wg values in this study are .590 and .559 for satisfaction, respectively, and .328 and .258 for repurchase intentions, respectively. Therefore, this study adopts a multilevel approach that takes into account both the customer level and the firm level.

Testing the Linkage of Satisfaction, Loyalty Intention, and SHV

To assess whether the total impact of customer satisfaction on shareholder returns (as measured by Tobin's q and the Carhart four-factor model) is (1) direct, (2) indirect (through the mediation of loyalty intentions as suggested by chain-of-effects models), (3) both direct and indirect, or (4) nonexistent, we used a three-level (drawing on Choi and Seltzer 2010) mediation approach (drawing on Zhao, Lynch, and Chen 2010) with Bayesian estimation (drawing on Yuan and MacKinnon 2009).

First, as suggested by Iacobucci (2008) and Yuan and MacKinnon (2009), we estimated the following equations using path modeling. Let i be the different customers under investigation, j the different firms, and t the different observations (i.e., one per year) per firm. Then, the three-level mediation model can be expressed as follows. At the first level,

\begin{matrix} {LI}_{ijt} = β_{1jt} + β_{2jt} {SAT}_{ijt} + β_{3jt} {d2000}_{ijt} + β_{4jt} {d2001}_{ijt} \\ + β_{5jt} {d2002}_{ijt} + β_{6jt} {d2003}_{ijt} + β_{7jt} {d2004}_{ijt} \\ + β_{8jt} {d2005}_{ijt} + β_{9jt} {d2006}_{ijt} {+ r}_{1 ijt}; \end{matrix}

at the second level,

\begin{matrix} β_{1jt} = γ_{10} {+ u}_{1jt}, β_{mjt} = γ_{m0} with m ranging from 2 to 9, \\ {LI}_{jt} = β_{10j} + β_{11j} {SAT}_{jt} + β_{12j} {d2000}_{jt} + β_{13j} {d2001}_{jt} \\ + β_{14j} {d2002}_{jt} + β_{15j} {d2003}_{jt} + β_{16j} {d2004}_{jt} \\ + β_{17j} {d2005}_{jt} + β_{18j} {d2006}_{jt} + β_{19j} {d2007}_{jt} \\ + β_{20j} {d2008}_{jt} {+ r}_{2jt}, \\ {SHV}_{jt} = β_{21j} + β_{22j} {LI}_{jt} + β_{23j} {SAT}_{jt} + β_{24j} {d2000}_{jt} \\ + β_{25j} {d2001}_{jt} + β_{26}_{j} {d2002}_{jt} + β_{27j} {d2003}_{jt} \\ + β_{28j} {d2004}_{jt} + β_{29j} {d2005}_{jt} + β_{30j} {d2006}_{jt} \\ + β_{31j} {d2007}_{jt} + β_{32t} {d2008}_{jt} {+ r}_{3jt}; \end{matrix}

and at the third level,

\begin{array}{l} β_{10j} = γ_{100} {+ u}_{10j}, \\ β_{mj} = γ_{m0} with m ranging from 11 to 20, \\ β_{21j} = γ_{210} {+ u}_{21j}, \\ β_{nj} = γ_{n0} with n ranging from 22 to 32 . \end{array}

SHV_jt is the dependent variable (shareholder value) and refers to both Tobin's q and the Carhart four-factor model. These two dependent variables are used by many of the studies in the satisfaction–SHV linkage literature (see Table 1). The parameters β_2jt and β_11j (“link a” in Figure 1, Panel C: customer satisfaction [SAT] → loyalty intentions [LI]), β_22j (“link b” in Figure 1, Panel C: LI → SHV), and β_23j (“link c” in Figure 1, Panel C: SAT → SHV) represent the key links of the test for mediation in which the first parameter is an estimate at the customer level and the remaining ones at the firm level.

We use the decision tree proposed by Zhao, Lynch, and Chen (2010) for establishing and understanding different types of mediation and nonmediation. More precisely, the parameters of links a and b are used to obtain an additional parameter estimate for link a × link b (Zhao, Lynch, and Chen 2010) and a corresponding Bayesian p-value (Yuan and MacKinnon 2009) at the firm level. Next, the significance levels of the new interaction term link a × link b and of link c are inspected to define the type of relationship satisfaction has on firm performance. Zhao, Lynch, and Chen recommend the following procedure: First, consider the indirect effect via link a × link b; then, inspect link c. The significance levels yield the following possible conditions:

If link a × link b is insignificant and link c is significant, then only a direct effect exists.

If link a × link b is significant and link c is insignificant, then only an indirect effect exists.

If link a × link b is significant and link c is significant, then both a direct and an indirect effect exist.

If link a × link b is insignificant and link c is insignificant, then there is no effect of SAT on SHV.

The parameters β_1jt, β_10j, and β_21j are the random intercepts. The specification of random intercepts allows different Level 1 (i.e., customers) and Level 2 (i.e., firms) units to have different regression intercepts, thereby accounting for unobserved heterogeneity. For instance, for the jth firm, β_10j quantifies the intercept of loyalty intention for the firm's mean-centered value of satisfaction at the year 2009 (i.e., the intercept denotes the value for LI when all variables after the = sign equal 0; note that SAT and LI were group mean–centered, as suggested by Zhang, Zyphur, and Preacher [2009]), where β_21j quantifies the intercept of the SHV metrics for the firm's mean-centered value of satisfaction and loyalty intentions at the year 2009. More precisely, the parameters γ₁₀, γ₁₀₀, and γ₂₁₀ are the population-averaged intercepts, whereas the parameters u_1jt, u_10j, and u_21j enable each firm to have its own intercept. In this part of the analysis, we do not test for the impact of predictors (e.g., industry characteristics) on these random intercepts; however, we do so in the next section, in which we also examine random coefficients.

The remaining βs represent the year dummy variables in each equation (note that the year 2007 serves as the reference category in Level 1, whereas 2009 is the reference year in Level 2)⁵ and account for time effects.

The parameters r_1ijt and r_2jt are the residuals of LI at the individual customer level and the aggregate firm level, respectively, whereas the parameter r_3jt represents the residuals of SHV at the firm level.

Testing for Industry Effects: Random Coefficients Analysis

To examine whether the population-averaged findings of the base models previously described apply across firms and industries, we introduce random coefficients for the three key links (a, b, and c), and thus have four slopes: β_2jt and β_11j (link a, Figure 1, Panel C: SAT → LI), β_22j (link b, Figure 1, Panel C: LI → SHV), and β_23j (link c, Figure 1, Panel C: SAT → SHV), which leads to the following changes at Level 2:

β_{2jt} = γ_{20} {+ u}_{2jt},

and at Level 3:

\begin{array}{l} β_{11j} = γ_{110} {+ u}_{11j}, \\ β_{22j} = γ_{220} {+ u}_{22j}, \\ β_{23j} = γ_{230} {+ u}_{23j} . \end{array}

More precisely, the random coefficients allow different firms to have different regression effects with respect to links a, b, and c, such that each firm may have its own type of mediation by inspecting the firm-specific link a × link b estimate in combination with the firm-specific effect of link c. For instance, for the jth firm, β_11j quantifies the relationship between the mediating variable (LI) and the independent variable (SAT) at the firm level, in which γ₁₁₀ is the population-averaged coefficient, which specifies the averaged effect of satisfaction on loyalty intentions. However, u_11j allows each firm to deviate from this population-averaged coefficient. Thus, each firm can have its own coefficient for link a, which is the sum of the population average and the deviation from it (γ₁₁₀ + u_11j). In a similar vein, u_2jt, u_22tj, and u_23tj represent the deviations from the population-averaged coefficients of links a, b, and c, respectively, for each firm.

In addition, we explore the impact of industry characteristics on the estimated random effects at Levels 2 and 3. We investigate three industry characteristics: purchase horizon (e.g., Li, Xu, and Li 2005), goods/services classification (e.g., Anderson, Fornell, and Rust 1997), and durable/nondurable classification (e.g., Larivière and Van den Poel 2004). We estimate the influence of industry characteristics on the random coefficients as follows:

\begin{matrix} u_{2jt} = β_{33} + β_{34} dService + β_{35} dDurable \\ + β_{36} dPhorizonCont + β_{37} {dPhorizonShort + r}_{4jt}, \\ u_{11j} = β_{38} + β_{39} dService + β_{40} dDurable \\ + β_{41} dPhorizonCont + β_{42} {dPhorizonShort + r}_{5j}, \\ u_{22j} = β_{43} + β_{44} dService + β_{45} dDurable \\ + β_{46} dPhorizonCont + β_{47} {dPhorizonShort + r}_{6j}, and \\ u_{23j} = β 48 + β_{49} dService + β_{50} dDurable + β_{51} dPhorizonCont \\ + β_{52} {dPhorizonShort + r}_{7j} . \end{matrix}

We estimate the influence of industry characteristics on the random intercepts as follows:

\begin{matrix} u_{1jt} = β_{53} + β_{54} dService + β_{55} dDurable + β_{56} dPhorizonCont \\ + β_{57} {dPhorizonShort + r}_{8jt}, \\ u_{10j} = β_{58} + β_{59} dService + β_{60} dDurable \\ + β_{61} dPhorizonCont + β_{62} {dPhorizonShort + r}_{9j}, and \\ u_{21j} = β_{63} + β_{64} dService + β_{65} dDurable \\ + β_{66} dPhorizonCont + β_{67} {dPhorizonShort + r}_{10j} . \end{matrix}

The βs represent the parameter estimates for the four dummy variables to indicate whether firm j's main focus is on goods or services (goods is the reference category) and on durables or nondurables (nondurable is the reference category), and whether the purchase horizon is continuous, long, or short (long purchase horizon is the reference category).

Testing for Time Lags between Customer Satisfaction and Repurchase Intentions

The issue of time lags between satisfaction and SHV has received much research attention in prior studies. In contrast, studies linking both satisfaction and repurchase intentions to SHV are scarce. Therefore, we investigate the impact of different time lags between SAT and LI and their influence on SHV. We investigate five time-lag combinations for each firm performance metric. In line with the previous two sections, we explore both population-averaged effects and random coefficients. The first two alternative models consider contemporaneous time lags for SAT and LI, and the three remaining alternatives consider different time lags between SAT and LI, given that it is possible that the impact of SAT on SHV takes longer than the impact of LI on SHV⁶:

Alternative A investigates the interrelationships between SAT_t-1, LI_t-1, and SHV_t.

Alternative B investigates the interrelationships between SAT_t-2, LI_t-2, and SHV_t.

Alternative C investigates the interrelationships between SAT_t-1, LI_t, and SHV_t.

Alternative D investigates the interrelationships between SAT_t-2, LI_t-1, and SHV_t.

Alternative E investigates the interrelationships between SAT_t-2, LI_t, and SHV_t.

Estimation

We estimate the Bayesian multilevel mediation models using Markov chain Monte Carlo (MCMC) techniques. As Yuan and MacKinnon (2009) note, compared with conventional frequentist mediation analysis, the Bayesian approach does not impose restrictive normality assumptions on sampling distributions of estimates, making statistical inference straightforward and exact. In addition, Gelman and Hill (2007) note that Bayesian estimation provides a more natural and simpler mediation analysis in multilevel models. Therefore, some researchers consider Bayesian inference an ideal approach for mediation analysis, and especially for complex multilevel mediation analysis (Yuan and MacKinnon 2009, p. 302).

For our analyses, we run three independent MCMC chains with different starting points (as suggested by Gelman and Rubin 1992) and 10,000 iterations each, of which the first half is considered the “burn-in” phase and the remaining half is used to estimate the posterior distribution for the parameters, resulting in a distribution based on 15,000 points. To assess the convergence of the MCMC algorithm, we assess the Gelman–Rubin convergence statistic R, autocorrelation plots, and trace plots of the residual variance for each parameter estimate (for details, see Web Appendix B).

FINDINGS

We organize our findings in four parts. First, we present the findings relating to the role of loyalty intentions in the impact that satisfaction has on SHV at the aggregate firm level. Second, we examine whether the influence of satisfaction and loyalty intent as revealed by insights found at the aggregate firm level reconciles with insights found at the individual customer level. Third, we explore the impact of heterogeneity for both satisfaction and loyalty intentions on SHV and further explain it using industry characteristics. Finally, as a robustness test of the interrelationships between satisfaction, loyalty intentions, and SHV, we compare the results of alternative models considering various time lags between satisfaction and loyalty intentions.

Aggregate Firm–Level Results for the Satisfaction, Loyalty Intention, and SHV Linkage

The population-averaged findings with respect to the total impact of customer satisfaction on Tobin's q and the Carhart four-factor model suggest an indirect effect, indicating that the impact of customer satisfaction on firm performance is mediated by loyalty intentions. As such, the logic of conceptual and empirical studies at the individual customer level also holds at the firm level. In other words, aggregate customer satisfaction results in aggregate loyalty intentions, which in turn results in SHV.

More precisely, Table 3 reveals that the direct impact (link c) is not significant, as denoted by the p-values of .416 and .405 for Tobin's q and the Carhart four-factor model. In contrast, the mean of the posterior distribution for the indirect effect (link a × link b) is positive for both Tobin's q (β_j = .027) and the Carhart four-factor model (β_j = .002), and the corresponding one-tailed Bayesian p-values are .008 and .13. One-tailed Bayesian p-values represent the proportion of the posterior distribution that is below zero. For the analysis of Tobin's q, this proportion is below 5% (as indicated with an asterisk in Table 3), whereas for the Carhart four-factor model, this percentage is higher, at 13%. The 95% Bayesian credibility intervals give the 2.5 and the 97.5 percentiles in the posterior distribution. For instance, the 95% credibility interval of the indirect effect of satisfaction on firm performance (link a × link b) for Tobin's q ranges from .005 to .048 (and thus does not include 0), providing strong evidence for a positive indirect impact or mediation effect. In contrast, the posterior distribution of the indirect effect for the Carhart four-factor model and the corresponding p-value reveal that the positive indirect effect for the Carhart four-factor model is less pronounced.

Table 3

BAYESIAN ESTIMATES FOR DRIVERS OF TOBIN'S Q AND CARHART'S FOUR-FACTOR MODEL

	Tobin'sq				Carhart's Four-Factor Model
	Estimate	Posterior SD	One-Tailed Bayesian p-Value	95% Confidence Interval	Estimate	Posterior SD	One-Tailed Bayesian p-Value	95% Confidence Interval
Dependent = LI_ijt
β_2jt SAT_ijt (link a, customer level)	.079^*	<.001	<.001	(.079, .08)	.079^*	<.001	<.001	(.079, .08)
β_3jt d2000_ijt	-.004	.031	.453	(-.066, .056)	-.004	.031	.453	(-.066, .056)
β_4jt d2001_ijt	.01	.025	.348	(-.039, .058)	.01	.025	.348	(-.039, .058)
β_5jt d2002_ijt	-.046	.027	.046	(-.1, .008)	-.046	.027	.046	(-.1, .008)
β_6jt d2003_ijt	-.037	.026	.075	(-.088, .014)	-.037	.026	.075	(-.088, .014)
β_7jt d2004_ijt	-.002	.022	.471	(-.044, .041)	-.002	.022	.471	(-.044, .041)
β_8jt d2005_ijt	-.029	.023	.109	(-.075, .017)	-.029	.023	.109	(-.075, .017)
β_9jt d2006_ijt	-.013	.023	.285	(-.057, .032)	-.013	.023	.285	(-.057, .032)
Dependent = LI_jt
β_11j SAT_jt (link a, firm level)	.08^*	.002	<.001	(.076, .084)	.08^*	.002	<.001	(.076, .084)
β_12j d2000_jt	.032	.028	.121	(-.021, .087)	.032	.028	.121	(-.021, .087)
β_13j d2001_jt	.05^*	.026	.024	(.000, .101)	.05^*	.026	.024	(.000, .101)
β_14j d2002_jt	-.022	.025	.186	(-.071, .028)	-.022	.025	.187	(-.071, .028)
β_15j d2003_jt	.005	.024	.426	(-.043, .054)	.005	.024	.424	(-.043, .054)
β_16j d2004_jt	.022	.024	.173	(-.024, .07)	.022	.024	.173	(-.024, .07)
β_17j d2005_jt	.013	.024	.29	(-.033, .061)	.013	.024	.288	(-.033, .061)
β_18j d2006_jt	.039^*	.024	.047	(-.007, .086)	.039^*	.024	.047	(-.007, .086)
β_19j d2007_jt	.028	.023	.115	(-.018, .075)	.028	.023	.115	(-.018, .075)
β_20j d2008_jt	.009	.023	.348	(-.037, .056)	.009	.023	.347	(-.037, .056)
Dependent = SHV_jt
β_22j LI_jt (link b, firm level)	.334^*	.138	.008	(.065, .605)	.022	.02	.13	(.016, .061)
β_23j SAT_jt (link c, firm level)	.003	.014	.416	(-.024, .03)	.000	.002	.405	(-.003, .004)
β_24j d2000_jt	.835^*	.123	<.001	(.595, 1.077)	-.018	.017	.148	(-.05, .016)
β_25j d2001_jt	.806^*	.115	<.001	(.581, 1.032)	-.022	.016	.082	(-.052, .009)
β_26j d2002_jt	.637^*	.111	<.001	(.422, .855)	-.05^*	.015	<.001	(-.079, -.02)
β_27j d2003_jt	.514^*	.11	<.001	(.3, .729)	-.017	.015	.134	(-.046, .013)
β_28j d2004_jt	.517^*	.107	<.001	(.309, .727)	-.018	.015	.116	(-.046, .011)
β_29j d2005_jt	.576^*	.107	<.001	(.365, .781)	-.021	.015	.072	(-.049, .008)
β_30j d2006_jt	.496^*	.104	<.001	(.291, .696)	-.023	.014	.058	(-.051, .005)
β_31j d2007_jt	.519^*	.103	<.001	(.319, .723)	-.046^*	.014	<.001	(-.074, -.018)
β_32j d2008_jt	.498^*	.103	<.001	(.296, .7)	-.019	.014	.087	(-.047, .009)
Additional Parameter
β_j a x b (link a x link b, firm level)	.027^*	.011	.008	(.005, .048)	.002	.002	.13	(-.001, .005)

p ≤ .05.

Notes: SHV = shareholder value, SAT = satisfaction, LI = loyalty intentions. Link a: SAT → LI; link b: LI → SHV; link c: SAT → SHV.

Contrasting Individual Customer–Level and Aggregate Firm–Level Results for the Satisfaction, Loyalty Intention, and SHV Linkage

The findings displayed in Table 3 reveal that the impact of satisfaction on loyalty intentions (link a) at the firm level (β_11j = .080) holds at the customer level (β_2jt = .079) for the average firm (population-averaged findings). Interestingly, when examining various effects across different firms (Table 4), we observe that at the firm level, the random coefficient for the impact of satisfaction on loyalty intentions is higher for firms that provide durable products/services with a short purchase horizon (e.g., shoe stores, cut-and-sew apparel providers, online bookstores). In contrast, at the customer level, these types of firms tend to have higher loyalty intention scores (i.e., random intercept) but not significantly higher coefficients for satisfaction. Therefore, the influence of satisfaction on loyalty intentions is more pronounced at the firm level, such that the interaction effect (or indirect impact of satisfaction on SHV through the mediation of loyalty as measured by link a × link b) is also positively biased. In summary, for some types of firms, the aggregate insights from satisfaction on SHV are overestimated and therefore less (or not) applicable at the individual customer level.

Table 4

THE IMPACT OF INDUSTRY CHARACTERISTICS ON RANDOM COEFFICIENTS

	Tobin's q		Carhart's Four-Factor Model
	Estimate	One-Tailed Bayesian p-Value	Estimate	One-Tailed Bayesian p-Value
Impact of Industry Characteristics on Random Coefficient of Link a, analyzed at the Customer Level (u_2jt)
β₃₄ dServices	.003	.267	.003	.267
β₃₅ dDurable	.001	.451	.001	.451
β₃₆ dPhorizonCont	-.003	.329	-.003	.329
β₃₇ dPhorizonShort	.001	.463	.001	.463
Impact of Industry Characteristics on Random Coefficient of Link a, analyzed at the Firm Level (u_11j)
β₃₉ dServices	.006	.218	.006	.196
β₄₀ dDurable	.027^*	.001	.027^*	.001
β₄₁ dPhorizonCont	-.007	.194	-.007	.191
β₄₂ dPhorizonShort	.018^*	.048	.017	.053
Impact of Industry Characteristics on Random Coefficient of Link b (u_22j)
β₄₄ dServices	-.335	.228	.003	.479
β₄₅ dDurable	.237	.324	.065	.169
β₄₆ dPhorizonCont	-.613	.125	-.059	.193
β₄₇ dPhorizonShort	.02	.491	-.028	.376
Impact of Industry Characteristics on Random Coefficient of Link c (u_23j)
β₄₉ dServices	-.088^*	.008	-.001	.494
β₅₀ dDurable	.053	.097	-.007	.207
β₅₁ dPhorizonCont	.004	.459	.001	.499
β₅₂ dPhorizonShort	-.014	.405	.005	.327
Impact of Industry Characteristics on Random Intercept for Loyalty Intentions, analyzed at the Customer Level (u_1jt)
β₅₄ dServices	-.046	.312	-.046	.312
β₅₅ dDurable	.299^*	.001	.299^*	.001
β₅₆ dPhorizonCont	-.011	.483	-.011	.483
β₅₇ dPhorizonShort	.219^†	.078	.219^†	.078
Impact of Industry Characteristics on Random Intercept for Loyalty Intentions, analyzed at the Firm Level (u_10j)
β₅₉ dServices	.001	.455	.001	.464
β₆₀ dDurable	-.001	.478	-.001	.483
β₆₁ dPhorizonCont	-.001	.459	-.001	.465
β₆₂ dPhorizonShort	.001	.481	.001	.475
Impact of Industry Characteristics on Random Intercept for Shareholder Value (u_21j)
β₆₄ dServices	-.573^*	.002	.001	.451
β₆₅ dDurable	-.623^*	.004	-.01	.132
β₆₆ dPhorizonCont	-.175	.239	-.001	.461
β₆₇ dPhorizonShort _.	.212	.233	-.003	.389

Significant at .05.

†

Significant at .078.

Notes: Link a: SAT → LI; link b: LI → SHV; link c: SAT → SHV.

Impact of Industry Differences on the Satisfaction, Loyalty Intention, and SHV Linkage

We use a random-coefficient model to investigate whether the indirect effect uncovered (satisfaction → loyalty intentions → SHV) for the average firm (i.e., population-averaged findings) applies to various types of industries.

Our investigation reveals that one cannot generalize the population-averaged findings of the indirect-only effect of satisfaction on firm performance to all firms. More precisely, Table 4 reveals that (1) estimates for link a—and, as a consequence, the indirect effects of satisfaction (link a × link b) on both the Carhart four-factor model and Tobin's q—are higher for firms that provide durable products/services with a short purchase horizon (e.g., shoe stores, cut-and-sew apparel providers, online bookstores); (2) durable services (e.g., energy utilities, cable and satellite television, commercial banks, insurance carriers) have lower values (i.e., intercepts) for Tobin's q than nondurable goods (e.g., beverage manufacturing, food manufacturing, personal care products, grocery stores) at baseline (i.e., irrespective of the influence of satisfaction, loyalty intentions, and year effects); and (3) in contrast to service firms, goods companies also provide evidence of a direct effect of customer satisfaction on Tobin's q. More precisely, with respect to the analysis of Tobin's q, Table 4 reveals a negative and significant impact of service firms on link c; or, in other words, for goods companies, the link between satisfaction and Tobin's q is more decisive.

To better understand the importance of this particular finding, we performed an ad hoc analysis in which we split the total data sample into a goods companies subsample and a service companies subsample. Next, we tested Bayesian multilevel mediation within each subsample. Interestingly, our findings indicate that for service firms, the significance of the indirect effect (link a × link b) and the nonsignificance of the direct effect (link c) reconcile with the population-averaged findings, whereas for the subsample of goods firms, we found evidence for both a significant indirect effect (link a × link b) and a significant direct effect (link c). These results indicate that the total effect of satisfaction on Tobin's q for goods companies is both direct and indirect, whereas service firms follow the population-averaged mean of indirect effect only.

Impact of Time Effects and Time Lags on the Satisfaction, Loyalty Intention, and SHV Linkage

Table 3 reveals the importance of considering fixed year effects. In particular, 11 out of the 25 year dummies for the analysis of Tobin's q are significant at the .05 level, whereas 4 year dummies are found to be significant for the Carhart four-factor model.

In addition, when we consider different time lags between loyalty intentions and customer satisfaction, our findings reveal different types of mediation, suggesting that the type of impact that satisfaction has on firm performance has a temporal/longitudinal character.

With respect to the Carhart four-factor model, we observe a stronger indirect-only effect of satisfaction (Bayesian p-value of .018) when the impact of satisfaction is related to customer loyalty and the Carhart four-factor model for the next year (alternative C in Table 5), as compared with the base model (Bayesian p-value of .130; see Table 3). Notably, we do not find an effect of satisfaction (direct or indirect) on the Carhart four-factor model when the time elapsed between customer satisfaction and the given firm performance metric equals more than one year (alternatives B, D, and E in Table 5) or in situations in which there is no time lapse between customer satisfaction and loyalty intentions (alternatives A and B in Table 5). In addition, the impact of industry characteristics on the observed variation is less pronounced for the Carhart four-factor model than for Tobin's q, indicating that differences between industries have less influence on the different observed values of the Carhart four-factor model. Note that Tobin's q is an accounting-based metric, whereas the Carhart four-factor model takes into consideration market factors such as risk, growth, and momentum. Accounting for these market factors may explain why we observe less variation between customer satisfaction, loyalty intentions, and shareholder returns when we use the Carhart four-factor model.

Table 5

THE IMPACT OF DIFFERENT TIME LAGS BETWEEN SAT AND LI AND THE IMPACT OF INDUSTRY CHARACTERISTICS ON THE SAT, LI, AND SHV LINKAGE AT THE FIRM LEVEL

Tobin's q
	Links Between SAT _t-1 , LI _t-1 , and SHV _t	Links Between SAT _t-2 , LI _t-2 , and SHV _t	Links Between SAT _t-1 , LI _t , and SHV _t	Links Between SAT _t-2 , LI _t-1 , and SHV _t	Links Between SAT _t-2 , LI _t , and SHV _t
Alternative	A	B	C	D	E
Part 1: Investigating the Type of Mediation: Population-Averaged Findings at the Firm Level
Indirect impact of SAT on SHV, through the mediation of LI (link a x link b)	n.s.	n.s.	.004	.003	n.s.
Direct impact of SAT on SHV (link c)	.022	.023	.028	.028	.03
Conclusion about impact (based on Zhao et al. 2010)	Direct only	Direct only	Direct and indirect^a	Direct and indirect^a	Direct only
Part 2: Investigating the Impact of Industry Characteristics at the Firm Level: Laying the Foundation for Different Types of Mediation Across Firms
Impact of Industry Characteristics on Random Slope of Link a
dServices	n.s.	n.s.	n.s.	n.s.	n.s.^‡
dDurable	.033	.036	n.s.	n.s.	n.s.
dPhorizonCont	n.s.	n.s.	n.s.	n.s.	n.s.
dPhorizonShort	n.s.	n.s.	n.s.	n.s.	n.s.
Impact of Industry Characteristics on Random Slope of Link b
dServices	n.s.	-1.364	-.638	n.s.	-.638
dDurable	n.s.	1.221	n.s.	n.s.	n.s.
dPhorizonCont	n.s.	n.s.	n.s.	-.661	n.s.
dPhorizonShort	n.s.	n.s.	n.s.	n.s.	n.s.
Impact of Industry Characteristics on Random Slope of Link c
dServices	n.s.	.075	n.s.	n.s.	n.s.
dDurable	n.s.	-.082	n.s.	n.s.	n.s.
dPhorizonCont	n.s.	n.s.	-.06	-.063	n.s.
dPhorizonShort	n.s.	n.s.	n.s.	n.s.	n.s.
Impact of Industry Characteristics on Random Intercept for
Repurchase Intentions
dServices	n.s.	n.s.	n.s.	n.s.	n.s.
dDurable	n.s.	n.s.	n.s.	n.s.	-.04
dPhorizonCont	n.s.	n.s.	n.s.	n.s.	-.039
dPhorizonShort	n.s.	n.s.	n.s.	n.s.	n.s.
Impact of Industry Characteristics on Random Intercept for Firm
Performance
dServices	-.589	-.626	-.555	-.592	-.584
dDurable	-.577	-.469	-.611	-.515	-.504
dPhorizonCont	n.s.	n.s.	n.s.	n.s.	n.s.
dPhorizonShort	n.s.	n.s.	n.s.	n.s.	n.s.

Carhart's Four-Factor Model
	Links Between SAT _t-1 , LI _t-1 , and SHV _t	Links Between SAT _t-2 , LI _t-2 , and SHV _t	Links Between SAT _t-1 , LI _t , and SHV _t	Links Between SAT _t-2 , LI _t-1 , and SHV _t	Links Between SAT _t-2 , LI _t , and SHV _t
Alternative	A	B	C	D	E
Part 1: Investigating the Type of Mediation: Population-Averaged Findings at the Firm Level
Indirect impact of SAT on SHV, through the mediation of LI (link a x link b)	n.s.	n.s.	.001^†	n.s.	n.s.
Direct impact of SAT on SHV (link c)	n.s.	n.s.	n.s.	n.s.	n.s.
Conclusion about impact (based on Zhao et al. 2010)	No effect	No effect	Indirect only^b	No effect	No effect
Impact of Industry Characteristics on Random Slope of Link a
dServices	n.s.	n.s.	n.s.	n.s.	-.02
dDurable	.034	.036	n.s.	n.s.	n.s.
dPhorizonCont	n.s.	n.s.	n.s.	n.s.	n.s.
dPhorizonShort	n.s.	n.s.	n.s.	n.s.	n.s.
Impact of Industry Characteristics on Random Slope of Link b
dServices	n.s.	n.s.	n.s.	n.s.	n.s.
dDurable	n.s.	n.s.	n.s.	n.s.	n.s.
dPhorizonCont	n.s.	n.s.	n.s.	.135	n.s.
dPhorizonShort	n.s.	n.s.	n.s.	.115	n.s.
Impact of Industry Characteristics on Random Slope of Link c
dServices	n.s.	n.s.	n.s.	n.s.	n.s.
dDurable	n.s.	n.s.	n.s.	n.s.	n.s.
dPhorizonCont	n.s.	n.s.	n.s.	n.s.	n.s.
dPhorizonShort	n.s.	n.s.	n.s.	n.s.	n.s.
Impact of Industry Characteristics on Random Intercept for Repurchase Intentions
dServices	n.s.	n.s.	n.s.	n.s.	n.s.
dDurable	n.s.	n.s.	n.s.	n.s.	-.038
dPhorizonCont	n.s.	n.s.	n.s.	n.s.	-.038
dPhorizonShort	n.s.	n.s.	n.s.	n.s.	n.s.
Impact of Industry Characteristics on Random Intercept for Firm Performance
dServices	n.s.	n.s.	n.s.	n.s.	n.s.
dDurable	n.s.	n.s.	n.s.	n.s.	n.s.
dPhorizonCont	n.s.	n.s.	n.s.	n.s.	n.s.
dPhorizonShort	n.s.	n.s.	n.s.	n.s.	n.s.

†

One-tailed Bayesian p-value for this parameter = .018.

‡

One-tailed Bayesian p-value for this parameter = .059; posterior means = -.019.

Baron and Kenny (1986) refer to this situation as “partial mediation,” whereas Zhao et al. (2010) label it “complementary mediation.”

Baron and Kenny (1986) refer to this situation as “full mediation.”

Notes: Only parameter estimates significant at .05 (one-tailed Bayesian p-value) are reported; n.s. = not significant at .05. SHV = Shareholder Value; SAT = satisfaction; LI = loyalty intentions (repurchase). Link a: SAT → LI; link b: LI → SHV; link c: SAT → SHV.

With respect to Tobin's q, the findings in Table 5 also provide evidence of a direct effect of satisfaction when it is related to loyalty intentions and Tobin's q of the following year (alternatives C and D in Table 5), whereas for the base model (see Table 3), we find only an indirect effect, and no direct effect. In addition, for the remaining time-lag situations (alternatives A, B, and E in Table 5), we find evidence of a direct effect but no indirect effect. As such, for Tobin's q, a direct effect of satisfaction on shareholder returns is dependent on the time lags considered. In particular, if we look at the immediate effect of satisfaction (base model), we find only an indirect effect of satisfaction through loyalty intentions, whereas over the longer term, satisfaction directly impacts Tobin's q.

Next, in line with the base model, we find that Tobin's q values at baseline are higher for nondurable goods, and this finding is supported for all five alternative time-lag models that we examine. This result indicates that the different time lags investigated do not have a significant impact on this particular effect. Finally, we observe that the magnitude of the direct impact of satisfaction on Tobin's q (link c in Table 5) is influenced by industry characteristics such as goods/services classification, durable/nondurable classification, and purchase horizon.

GENERAL DISCUSSION

The goal of this study is to develop a finer understanding of the satisfaction–SHV relationship by considering the mediating role of loyalty. This inquiry was prompted by the observation that although researchers and managers have generally agreed that customer satisfaction affects a firm's performance largely through its impact on customer loyalty, there exists limited research about the effects of loyalty intentions on SHV (only 2 out of 22 studies; Table 1) and whether the mediating role of loyalty on firm performance at the individual customer level holds at the aggregate firm level. Because the majority of existing evidence in the literature is based on aggregate firm-level data, and most managerial decisions pertain to behaviors and outcomes at the customer level, managers need compelling evidence to understand the extent to which insights from the aggregate data would apply to their individual-level decision making. Therefore, we studied the impact of satisfaction and loyalty intentions at both the individual customer and aggregate firm levels and treated these impacts as random and dependent on industry characteristics.

Our main findings reveal that (1) the insights of aggregate firm-level studies may overestimate the impact that satisfaction has at the individual customer level; (2) loyalty intentions increase our understanding of how satisfaction translates into SHV; (3) population-averaged findings do not hold for each firm, such that different interrelationships between customer satisfaction, loyalty intentions, and SHV are likely to exist; and (4) these different interrelationships can be explained by industry differences.

Implications of Our Findings for Researchers, Managers, and Investors

Our findings have important implications that are relevant to managers, investors, and researchers. First, for managers and investors applying these findings within their companies to the customer level, it is important to understand that effects from aggregate firm-level data may be overestimated. More precisely, on the basis of our findings, we expect the impact of satisfaction on SHV to be reduced especially for firms providing durable products and services with a short purchase horizon (e.g., shoe stores, cut-and-sew apparel providers, online bookstores). It is important to take this finding into account when using aggregate firm-level data such as the ACSI to make investment decisions.

Second, we find that for some firms, the satisfaction–SHV linkage is mediated by loyalty intentions, whereas for others the effect of satisfaction on SHV is also direct. More precisely, we find that for service firms, the impact of satisfaction on SHV is indirect (mediated) through repurchase intentions, whereas for goods firms, we find evidence of both a direct and an indirect link. For service firms (vs. goods firms), loyalty intentions appear to be a better predictor of SHV than satisfaction, since this metric is more closely linked to SHV. This may in part be explained by the heterogeneity of performance of service providers’ employees, or there could also be systematic differences in the amount and types of information technology systems and customer relationship management approaches that goods or service firms may be using to interact with their customers (Mithas, Krishnan, and Fornell 2005; Mithas and Rust 2016).

Third, we underscore the need to disentangle firm-specific effects from population-averaged findings, in line with Anderson, Fornell, and Mazvancheryl (2004), Anderson and Mansi (2009), Gruca and Rego (2005), and Mittal et al. (2005). The former matters most to managers, because individual firms can deviate from the “average” firm. Therefore, managers and investors need to be cautious in interpreting findings that are based on population averages and recognize that such findings may not apply to their firm context. In this study, we find that the goods/services classification offers meaningful insight into how satisfaction, repurchase intentions, and SHV interrelate. Managers and investors can easily use this classification when evaluating firm-level satisfaction scores such as the ACSI. In addition, our multilevel and random-effects approach offers avenues for future studies of SHV. Further research can perhaps discriminate between population-averaged and firm-specific effects, making it easier to compare and contrast the findings from different studies.

Fourth, our findings reveal that time effects are likely to influence SHV. In addition, when we investigate the impact of both loyalty intentions and satisfaction, our results show that different time lags between these variables can lead to different conclusions regarding how satisfaction, loyalty intentions, and SHV are interrelated. Thus, it is important to consider the specific time frame when trying to understand how SHV unfolds.

Finally, our findings indicate that the relationship between satisfaction and SHV is sensitive to the metric used to assess performance (in the case of this investigation, Tobin's q and the Carhart four-factor model). If the objective is strictly to predict stock performance (as would be the case for many financial management companies), then much of the effect of customer satisfaction and loyalty intentions on shareholder returns is captured in multifactor financial models. Nonetheless, although our research indicates that this sensitivity to performance metric is less prominent when using multifactor models, satisfaction and loyalty intentions can enhance the linkage to shareholder returns, especially when industry and time effects are taken into consideration.

A Better Understanding of the Mixed Findings from Prior Studies

Although it is often difficult to compare research findings across different prior studies—because they use different compilations of firms that rely on population-averaged findings, consider different time frames (i.e., time effects) and time lags between variables, and use different SHV metrics—this study sheds some light on mixed findings in the prior literature and likely explanations for them. Our literature review (Table 1) reveals 12 studies that report mixed findings. Four of them make use of random coefficients, such that it is likely that some firms/industries have a satisfaction–SHV relationship that deviates from the population mean, in line with our study findings. Interestingly, four other studies do not use random slopes but, rather, examine the effect of satisfaction on SHV for different subsamples of their data set. For instance, Ittner and Larcker (1998) find a positive “population-averaged” effect of satisfaction on the market value of equity (Ittner and Larcker, Table 6, p. 23), but when they consider arbitrarily chosen subsamples of different industries, they variously find no relationship, a positive relationship, and even a negative relationship between satisfaction and SHV (Ittner and Larcker, Table 7, p. 27). In addition, Jacobson and Mizik (2009, Table 2, p. 6) find a positive effect of ACSI on future stock market performance when they consider all firms in their analyses, and they report no relationship (for the subsample of utility firms and the subsample of remaining firms) and positive relationship (for the subsample that combines computer and Internet firms) effects when they explore different subsamples of industries. For instance, it may be that Jacobson and Mizik (2009) do not find a significant impact for the group of remaining subgroup firms because of the “subpopulation”-averaged effect. These authors also do not test whether computer versus Internet firms may have other effects when they are not grouped together. Population-averaged findings may mask the different effects of subgroups. These factors may complement other explanations provided by Fornell, Mithas, and Morgeson (2009) for the different results obtained for subsamples in Jacobson and Mizik's (2009) study.

Differences among prior studies can also be attributed to different SHV or bondholder value metrics, and/or other time frames, being used. For instance, in one study, Anderson and colleagues find that the ACSI–Tobin's q link is stronger for airlines (Anderson, Fornell, and Mazvancheryl 2004), whereas in another study, they find a weaker link between ACSI and yield spread for airlines (Anderson and Mansi 2009). Our findings, which suggest that the impact of satisfaction and loyalty intentions is stronger for Tobin's q than for the Carhart four-factor model, imply some type of risk adjustments that stock market participants may be making, similar to those that concern bondholders. Finally, in interpreting their mixed findings, Ittner and Larcker (1998, p. 17) argue that satisfaction levels may have an indirect effect on accounting performance by attracting new customers. Our study provides evidence for a chain of effects in which satisfaction relates to SHV through the mediation of repurchase intentions. Because repurchase intentions and recommendation intentions are typically highly correlated, our study helps us to better explain the mixed findings of Ittner and Larcker's single-industry (i.e., financial services) study. For service firms (including financial services firms), we find only an indirect impact of satisfaction on SHV through repurchase intentions (which relates to recommendation intentions and, by extension, new customer acquisition). In summary, our conceptual chain-of-effects model at the firm level, in which we consider loyalty as a mediator between satisfaction and SHV, helps us to better understand other mixed findings in the literature.

Limitations and Directions for Further Research

Although this study increases our understanding of how satisfaction relates to repurchase intentions and SHV at a more granular level, it also suggests fruitful avenues for further research. First, this investigation is based on U.S. companies only, and prior research has suggested that satisfaction is likely to differ across cultures and countries (Morgeson et al. 2011). Moreover, in a study of the automobile industry, Raithel et al. (2012) find evidence that different satisfaction dimensions have varying impact on SHV across U.S., U.K., and German markets. Therefore, future studies should examine whether the satisfaction, loyalty intentions, and SHV interrelationships are generalizable to other cultures/markets.

Second, the ACSI primarily monitors large business-to-consumer firms. As such, there is a need to study business-to-business relationships, as well as small and medium-sized enterprises in which satisfaction and loyalty intentions could be linked to firm performance metrics such as sales, market share, and so on.

Third, although repurchase intentions and recommendation intentions are typically considered two items of the same loyalty construct, further research could elaborate on word of mouth and examine how different word-of-mouth dimensions (e.g., online, face-to-face) influence SHV.

Finally, although this study accounts for unobserved heterogeneity, many other factors (e.g., firm strategies, whether revenue or cost focused; types of information technology systems that firms use) have not been explicitly tested for in this study. Therefore, studies that examine these and other potential drivers of SHV, such as customer lifetime value, referrals, and word of mouth, offer other promising avenues for further research.

Footnotes

1

We thank an anonymous reviewer for drawing our attention to this point.

2

Four out of these 12 studies treated the impact of SAT on SHV as random coefficients (i.e., Anderson, Fornell, and Mazvancheryl 2004; Anderson and Mansi 2009; Gruca and Rego 2005; Mittal et al. 2005), and four others investigated the SAT–SHV linkage for various subgroups of industries (i.e., Ittner and Larcker 1998; Jacobson and Mizik 2009; Keiningham et al. 2007) or countries (i.e., ).

3

For more information, please see the ACSI website: .

4

From 2010 onward, ACSI data has been released monthly.

5

The ACSI provided us the unique opportunity to examine both satisfaction and repurchase intentions at both the customer and firm levels. In order to do so, they compiled for us a data set of individual customer responses that covered seven years of data (from 2000 to 2007) for the firms we investigated (i.e., firms linked to stock exchange data).

6

We thank an anonymous reviewer for making this suggestion.

References

Aksoy

Lerzan

(2013), “How Do You Measure What You Can't Define? The Current State of Loyalty Measurement and Management,” Journal of Service Management, 24 (4), 356–81.

Aksoy

Lerzan

, Cooil

Bruce

, Groening

Christopher

, Keiningham

Timothy L.

, and Yalçin

Atakan

(2008), “Long Term Stock Market Valuation of Customer Satisfaction,” Journal of Marketing, 72 (July), 105–22.

Anderson Analytics (2009), MENG Marketing Trends Report 2009, (accessed on November 24, 2015), [available at: http://www.slideshare.net/hubbudunya/meng-marketing-trends-report-2009].

Anderson Analytics (2010), MENG Marketing Trends Report 2010, (Accessed on November 24, 2015), [available at: http://www.scribd.com/doc/27877001/Anderson-Analytics-MENG-2010-Trend-Report-Final].

Anderson

Eugene W.

, Fornell

Claes

, and Mazvancheryl

Sanal K.

(2004), “Customer Satisfaction and Shareholder Value,” Journal of Marketing, 68 (October), 172–85.

Anderson

Eugene W.

, Fornell

Claes

, and Rust

Roland T.

(1997), “Customer Satisfaction, Productivity, and Profitability: Differences Between Goods and Services,” Marketing Science, 16 (2), 129–45.

Anderson

Eugene W.

and Mansi

Sattar A.

(2009), “Does Customer Satisfaction Matter to Investors? Findings from the Bond Market,” Journal of Marketing Research, 46 (October), 703–14.

Anderson

Eugene W.

and Mittal

Vikas

(2000), “Strengthening the Satisfaction-Profit Chain,” Journal of Service Research, 3 (2), 107–20.

Baker-Prewitt

Jamie

(2014), “Advances in Customer Loyalty Measurement,” Burke White Papers No. 0702, [available at http://www.burke.com/Library/WhitePapers/SCI_White_Paper_low_res_pages.pdf]. Cincinnati: Burke.

10.

Baron

Reuben M.

and Kenny

David A.

(1986), “The Moderator–Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations,” Journal of Personality and Social Psychology, 51 (6), 1173–82.

11.

Bliese

Paul D.

(2000), “Within-Group Agreement, Non-Independence, and Reliability: Implications for Data Aggregation and Analysis,” in Multilevel Theory, Research, and Methods in Organizations, Klein

Katherine. J.

and Kozlowski

Steve W.J.

, eds. San Francisco: Jossey-Bass, 349–81).

12.

Bliese

Paul D.

(2013), “Multilevel Modeling in R (2.5): A Brief Introduction to R, the Multilevel Package, and the Nlme Package,” (accessed November 24, 2015), [available at http://cran.r-project.org/doc/contrib/Bliese_Multilevel.pdf].

13.

Brodie

Roderick J.

, Hollebeek

Linda D.

, Juric

Biljana

, and Ilic

Ana

(2011), “Customer Engagement: Conceptual Domain, Fundamental Propositions, and Implications for Research,” Journal of Service Research, 14 (3), 252–71.

14.

Brown

Reagan D.

and Hauenstein

Neil M.A.

(2005), “Interrater Agreement Reconsidered: An Alternative to the rwg Indices,” Organizational Research Methods, 8 (2), 165–84.

15.

Choi

Kilchan

and Seltzer

Michael

(2010), “Modeling Heterogeneity in Relationships Between Initial Status and Rates of Change: Treating Latent Variable Regression Coefficients as Random Coefficients in a Three-Level Hierarchical Model,” Journal of Educational and Behavioral Statistics, 35 (1), 54–91.

16.

Danaher

Peter

and Rust

Roland T.

(1996), “Indirect Financial Benefits from Service Quality,” (with Discussion and Rejoinder), Quality Management Journal, 3 (2), 63–88.

17.

Fornell

Claes

, Mithas

Sunil

, and Morgeson

Forrest V.

(2009), “The Economic and Statistical Significance of Stock Returns on Customer Satisfaction,” Marketing Science, 28 (5), 820–25.

18.

Fornell

Claes

, Mithas

Sunil

, Morgeson

Forrest V.

, and Krishan

M. S.

(2006), “Customer Satisfaction and Stock Prices: High Returns, Low Risk,” Journal of Marketing, 70 (January), 1–14.

19.

Freedman

David

, Pisani

Robert

, and Purves

Roger

(1998), Statistics, 3d ed. New York: W.W. Norton & Company.

20.

Gelman

Andrew

and Hill

Jennifer

(2007), Data Analysis Using Regression and Multilevel/Hierarchical Models. New York: Cambridge University Press.

21.

Gelman

Andrew

and Rubin

Donald. B.

(1992), “Inference from Iterative Simulation Using Multiple Sequences,” Statistical Science, 7 (4), 457–72.

22.

Grewal

Rajdeep

, Chandrashekaran

Murali

, and Citrin

Alka V.

(2010), “Customer Satisfaction Heterogeneity and Shareholder Value,” Journal of Marketing Research, 47 (August), 612–26.

23.

Gruca

Thomas S.

and Rego

Lopo L.

(2005), “Customer Satisfaction, Cash Flow, and Shareholder Value,” Journal of Marketing, 69 (July), 115–30.

24.

Gupta

Sunil

, Lehmann

Donald

, and Stuart

Jennifer Ames

(2004), “Valuing Customers,” Journal of Marketing Research, 41 (February), 7–18.

25.

Heskett

James L.

, Sasser

W. Earl

, and Schlesinger

Leonard A.

(1997), The Service Profit Chain. New York: The Free Press.

26.

Iacobucci

Dawn

(2008), Mediation Analysis. Thousand Oaks, CA: Sage Publications.

27.

Ittner

Christopher D.

and Larcker

David

(1998), “Are Nonfinancial Measures Leading Indicators of Financial Performance? An Analysis of Customer Satisfaction,” Journal of Accounting Research, 36 (Suppl.), 1–35.

28.

Ittner

Christopher D.

, Larcker

David

, and Taylor

Daniel

(2009), “The Stock Market's Pricing of Customer Satisfaction,” Marketing Science, 28 (5), 826–35.

29.

Ivanov

Vladimir

, Kissan

Joseph

, and Babajide

Wintoki M.

(2013), “Disentangling the Market Value of Customer Satisfaction: Evidence from Market Reaction to the Unanticipated Component of ACSI Announcements,” International Journal of Research in Marketing, 30 (2), 168–78.

30.

Jacobson

Robert

and Mizik

Natalie

(2009), “The Financial Markets and Customer Satisfaction: Reexamining Possible Financial Market Mispricing of Customer Satisfaction,” Marketing Science, 28 (September/October), 810–19.

31.

James

Lawrence R.

, Demaree

Robert G.

, and Wolf

Gerrit

(1984), “Estimating Within-Group Interrater Reliability With and Without Response Bias,” Journal of Applied Psychology, 69 (February), 85–98.

32.

Keiningham

Timothy L.

, Cooil

Bruce

, Andreassen

Tor Wallin

, and Aksoy

Lerzan

(2007), “A Longitudinal Examination of Net Promoter and Firm Revenue Growth,” Journal of Marketing, 71 (July), 39–51.

33.

Klein

Katherine J.

and Kozlowski

Steve W. J.

(2000), Multilevel Theory, Research, and Methods in Organizations. San Francisco: Jossey-Bass, pp. 349–81.

34.

Kozlowski

Steve W.J.

and Hattrup

Keith

(1992), “A Disagreement About Within-Group Agreement: Disentangling Issues of Consistency Versus Consensus,” Journal of Applied Psychology, 77 (2), 161–67.

35.

Kumar

and Shah

Denish

(2009), “Expanding the Role of Marketing: From Customer Equity to Market Capitalization,” Journal of Marketing, 73 (November), 119–36.

36.

Larivière

Bart

(2008), “Linking Perceptual and Behavioral Customer Metrics to Multiperiod Customer Profitability: A Comprehensive Service-Profit Chain Application,” Journal of Service Research, 11 (1), 3–21.

37.

Larivière

Bart

and Van den Poel

Dirk

(2004), “Investigating the Role of Product Features in Preventing Customer Churn, by Using Survival Analysis and Choice Modeling: The Case of Financial Services,” Expert Systems with Applications, 27 (2), 277–85.

38.

Chunqing

, Xu

Yinfeng

, and Li

Hongyi

(2005), “An Empirical Study of Dynamic Customer Relationship Management,” Journal of Retailing and Consumer Services, 12 (6), 431–41.

39.

Luo

Xueming

and Homburg

Christian

(2008), “Satisfaction, Complaint, and the Stock Value Gap,” Journal of Marketing, 72 (July), 29–43.

40.

Luo

Xueming

, Homburg

Christian

, and Wieseke

Jan

(2010), “Customer Satisfaction, Analyst Stock Recommendations, and Firm Value,” Journal of Marketing Research, 47 (December), 1041–58.

41.

Merrin

Robert P.

, Hoffmann

Arvid O.I.

, and Pennings

Joost M.E.

(2013), “Customer Satisfaction as a Buffer Against Sentimental Stock-Price Corrections,” Marketing Letters, 24 (1), 13–27.

42.

Mithas

Sunil

, Krishnan

M.S.

, and Fornell

Claes

(2005), “Why Do Customer Relationship Management Applications Affect Customer Satisfaction?,” Journal of Marketing, 69 (October), 201–09.

43.

Mithas

Sunil

and Rust

Roland T.

(2016), “How Information Technology Strategy and Investments Influence Firm Performance: Conjectures and Empirical Evidence,” MIS Quarterly, 40 (1), 223–246.

44.

Mittal

Vikas

, Anderson

Eugene W.

, Sayrak

Akin

, and Tadikamalla

Pandu

(2005), “Dual Emphasis and the Long-Term Financial Impact of Customer Satisfaction,” Marketing Science, 24 (4), 544–55.

45.

Morgan

Neil A.

and Rego

Lopo Leotte

(2006), “The Value of Different Customer Satisfaction and Loyalty Metrics in Predicting Business Performance,” Marketing Science, 25 (September/October), 426–39.

46.

Morgeson

Forrest V.

, Mithas

Sunil

, Keiningham

Timothy L.

, and Aksoy

Lerzan

(2011), “An Investigation of the Cross-National Determinants of Customer Satisfaction,” Journal of the Academy of Marketing Science, 39 (2), 198–215.

47.

Oliver

Richard L.

(2010), Satisfaction: A Behavioral Perspective on the Consumer, 2nd ed. Armonk, NY: M.E. Sharpe.

48.

O'sullivan

Don

, Hutchinson

Mark C.

, and O'Connell

Vincent

(2009), “Empirical Evidence of the Stock Market's (Mis)pricing of Customer Satisfaction,” International Journal of Research in Marketing, 26 (2), 154–61.

49.

Parasuraman

, Zeithaml

Valarie A.

, and Berry

Leonard L.

(1988), “SERVQUAL: A Multiple-Item Scale for Measuring Consumer Perceptions of Service Quality,” Journal of Retailing, 64 (Spring), 12–40.

50.

Raithel

Sascha

, Sarstedt

Marko

, Scharf

Sebastian

, and Schwaiger

Manfred

(2012), “On the Value Relevance of Customer Satisfaction. Multiple Drivers and Multiple Markets,” Journal of the Academy of Marketing Science, 40 (4), 509–25.

51.

Rego

Lopo L.

, Morgan

Neil A.

, and Fornell

Claes

(2013), “Reexamining the Market Share–Customer Satisfaction Relationship,” Journal of Marketing, 77 (September), 1–20.

52.

Rust

Roland T.

, Lemon

Katherine N.

, and Zeithaml

Valarie A.

(2004), “Return on Marketing: Using Customer Equity to Focus Marketing Strategy,” Journal of Marketing, 68 (January), 109–27.

53.

Rust

Roland T.

and Zahorik

Anthony J.

(1993), “Customer Satisfaction, Customer Retention and Market Share,” Journal of Retailing, 69 (2), 193–215.

54.

Rust

Roland T.

, Zahorik

Anthony J.

, and Keiningham

Timothy

(1995), “Return on Quality (ROQ): Making Service Quality Financially Accountable,” Journal of Marketing, 59 (April), 58–70.

55.

Schulze

Christian

, Skiera

Bernd

, and Wiesel

Thorsten

(2012), “Linking Customer and Financial Metrics to Shareholder Value: The Leverage Effect in Customer-Based Valuation,” Journal of Marketing, 76 (March), 17–32.

56.

Tinsley

Howard E.A.

and Weiss

David J.

(1975), “Interrater Reliability and Agreement of Subjective Judgments,” Journal of Counseling Psychology, 22 (4), 358–76.

57.

Tuli

Kapil R.

and Bharadwaj

Sundar G.

(2009), “Customer Satisfaction and Stock Returns Risk,” Journal of Marketing, 73 (November), 184–97.

58.

Wiesel

Thorsten

, Skiera

Bernd

, and Villanueva

Julin

(2008), “Customer Equity: An Integral Part of Financial Reporting,” Journal of Marketing, 72 (March), 1–14.

59.

Yuan

Ying

and MacKinnon

David P.

(2009), “Bayesian Mediation Analysis,” Psychological Methods, 14 (4), 301–22.

60.

Zeithaml

Valarie A.

, Berry

Leonard L.

, and Parasuraman

(1996), “The Behavioral Consequences of Service Quality,” Journal of Marketing, 60 (April), 31–46.

61.

Zeithaml

Valarie A.

, Bolton

Ruth

, Deighton

John

, Keiningham

Timothy L.

, Lemon

Katherine

, and Petersen

Andrew

(2006), “Forward-Looking Customer Focus: Can Firms Have Adaptive Foresight?” Journal of Service Research, 9 (2), 168–83.

62.

Zhang

Zhen

, Zyphur

Michael J.

, and Preacher

Kristopher J.

(2009), “Testing Multilevel Mediation Using Hierarchical Linear Models: Problems and Solutions,” Organizational Research Methods, 12 (4), 695–719.

63.

Zhao

Xinshu

, Lynch

John G.

Jr. , and Chen

Qimei

(2010), “Reconsidering Baron and Kenny: Myths and Truths about Mediation Analysis,” Journal of Consumer Research, 37 (2), 197–206.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.33 MB

0.00 MB