Abstract Service companies invest billions of dollars to develop and maintain long-term customer relationships by offering corporate gifts. Yet several questions remain regarding such relationship marketing instruments: What impact do different kinds of gifts have on customers? Which perceptions allow gifts to affect customer behavior? What financial outcomes do these gifts imply for firms? To answer these questions, the authors use data from 1,950 airline customers—combining a longitudinal field experiment with internal customer database information—and explore the effects of different gift designs on customer perceptions and actual spending behavior. The experiment manipulates four gift designs (economic related/unrelated; social related/unrelated) and measures customer perceptions and behavior at different points in time. Multivariate analyses of covariance (MANCOVAs) and spotlight analyses reveal that the congruent combinations of economic related and social unrelated gift dimensions have the strongest effects on customer perceptions of relationship investment. Serial mediation analyses further reveal that the impact of gifts on customer spending is fully mediated by customer perceptions of perceived relationship investment and repurchase intention. Economic related gifts produce the highest contribution margins. Service managers may use these findings to design effective gifts and management processes (e.g., gift success tracking). Keywords relationship marketing, customer loyalty, corporate gifts Many service organizations make extensive use of corporate gifts to improve the relationships with their customers. In the United States alone, companies spend US$22 billion a year on gift items (IBISWorld 2015). Corporate gifts are benefits that a firm confers voluntarily on its customers in its attempt to communicate appreciation and gratitude for their past purchases (Dorsch and Kelley 1994; Liu, Lamberton, and Haws 2015). Gifts are recognized relationship marketing instruments (Palmatier et al. 2009), but in contrast to loyalty programs, they do not require recipients to perform any action (e.g., collecting points, earning miles) to receive them. Unlike direct mailings, gifts reflect dedicated efforts to foster long-term customer relationships (Belk and Coon 1993), and unlike interpersonal communication or preferential treatment, gift exchanges entail ritual and ceremony (Sherry 1983). Previous research has focused on relationship marketing instruments such as loyalty programs and direct mailing (e.g., Berry 1995; Verhoef 2003), yet despite its managerial importance, research on corporate gifts is relatively scant (e.g., Beltramini 1992). Accordingly, key questions related to the use of corporate gifts in relationship marketing contexts remain unanswered, including the impact that different gift designs (i.e., combinations of gift dimensions) have on a gift recipient’s perceptions of the gift and subsequent reactions to that gift. Nor does current research on corporate gifts, granted to both consumers and business customers, provide service managers with a means to assess the financial outcomes of the different gift designs for the firms. This research contributes to the few existing studies on corporate gifts by shedding light on these matters, both conceptually and empirically. We argue that the effect of corporate gifts on relationship outcomes depends on two critical dimensions of the gift design: gift type and gift relatedness. Gift type reflects the basic motivation and relational bonds a gift reflects or targets, and we argue that economic gifts and social gifts represent two major gift types (Belk and Coon 1993; Berry 1995).1 Gift relatedness instead refers to the link between a gift and the gift-giving firm, including its products and services. Accordingly, gift relatedness is high if gifts pertain to the firm and its commercial 1Marketing Center Muenster, University of Muenster, Muenster, Germany 2 University of Augsburg, Augsburg, Germany 3Marketing Cþ, Weilheim, Germany Corresponding Author: Andre´ Marchand, Marketing Center Muenster, University of Muenster, Am Stadtgraben 13-15, 48143 Muenster, Germany. Email: [email protected] Journal of Service Research 1-15 ª The Author(s) 2016 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/1094670516682091 journals.sagepub.com/home/jsr offerings (e.g., company products) but low if those gifts are unrelated (e.g., products from other companies; Keh and Lee 2006; Yi and Jeon 2003).2 Extant research on corporate gifts investigates the effects of an overall gift rather than its design (see Dorsch and Kelley 1994, for an exception); moreover, we know of no gift research that studies economic and social gifts empirically or their interactions with gift relatedness. Building on the concept of fit or congruence, we posit that the combination of the two dimensions (i.e., gift design) matters for customers, rather than the isolated main effects of each dimension. Furthermore, we seek to uncover underlying customer perceptions and test for mediating effects on the link between corporate gifts and behavior by studying two key relationship marketing constructs: perceived relationship investment and repurchase intention (De Wulf, Odekerken-Schro¨der, and Iacobucci 2001). Extant gift research investigates the impact of gifts on customer perceptions (Beltramini 1992; Bodur and Grohmann 2005; Dorsch and Kelley 1994) but not on behavior. In an exception, Haisley and Loewenstein (2011) consider the effects of economic unrelated gifts (gas, restaurant gifts) on the behavior and perceptions of commercial bank customers, but they do not compare such gifts with economic related, social related, or social unrelated gifts, nor do they investigate any mediating processes through perceptions. Evidence of such mediation would support the theory we propose, which is based on a relational mediator framework. To provide these contributions, we compare the effects of different corporate gifts in a randomized field experiment. With a 2 (gift type: economic vs. social) 2 (gift relatedness: related vs. unrelated) between-subjects design, we create four different gifts (economic related, economic unrelated, social related, and social unrelated). The congruent combinations of economic related and social unrelated gift dimensions exert the strongest effects on customer perceptions of relationship investment; the impact of gifts on customer spending is also fully mediated by customer perceptions of perceived relationship investment and repurchase intention. With these findings, managers can form realistic expectations of the outcomes of various gift designs and better track gift success. The results further support and extend limited research on the key mediating role of perceived relationship investment. Finally, corporate gifts can be costly, but we know of no research that considers the financial outcomes of different gift designs for firms. We investigate the contribution margins of different gift designs in a transportation services setting and determine that economic related gifts produce the highest increase in margins, the other gift types even can be unprofitable. Gifts differ in their financial outcomes for firms, and managers and researchers must account for these differences when designing both gift programs and research studies. For managers, this study also helps them identify the most effective combinations of gift dimensions and their contribution margins. Accordingly, this study extends corporate gift literature by investigating the effects of different gift designs on customer relationships, the underlying mediation through perceptions on behavior, and the financial outcomes of different designs for firms. We next present our conceptual model and study design, before we discuss the findings of our experimental study. Theoretical Model Overview Figure 1 presents our theoretical model, which is derived from existing relationship marketing research (Morgan and Hunt 1994; Palmatier 2008), especially that focused on relational mediator frameworks (Bolton, Lemon, and Verhoef 2004; De Wulf, Odekerken-Schro¨der, and Iacobucci 2001; Palmatier et al. 2006; Palmatier et al. 2009). Consistent with this research, the model contains four groups of variables: corporate gifts as a relationship marketing instrument, customer relationship perceptions, customer spending as a form of customer relationship behavior, and gift contribution margin as the firm’s financial outcome. We propose that a gift (X) influences perceived relationship investment directly (path X ) a1; Hypothesis 1) and customer spending (Y) indirectly through customer relationship perceptions, with serial mediation by perceived relationship investment and repurchase intention (path X ) a1 ) b ) c1 ) Y; Hypothesis 2). We account empirically for the potential direct effect of a gift on customer spending (path X ) d0 ) Y). Customer spending translates into incremental revenues minus the costs, such that ultimately it defines the contribution margins (profit). Our rationale for the different gift designs reflects Berry’s (1995) definition of different levels of relationship marketing (economic and social) and the distinction between companyrelated and unrelated rewards (Yi and Jeon 2003). To predict the impact of gifts on customer relationships, we rely on the principle of reciprocity (Beltramini 1992; Dorsch and Kelly 1994), a fundamental social norm that dictates that people repay positive actions, such as gifts or favors (Bagozzi 1995; Gouldner 1960). In our study context, reciprocity means that customers who perceive that a company invests in the relationship with them (e.g., by providing a gift) are likely to develop intentions to return that investment, which ultimately may result in reciprocating behaviors (e.g., purchasing the company’s offerings). In the next two sections, we develop hypotheses for the effects of different gift designs on perceived relationship investment and the mediating process of gifts through perceptions on customer behavior. We refrain from offering a hypothesis on the financial outcomes of gift designs for firms, because gift costs are specific to our study context, so gift effects might be difficult to generalize. Effects of Gift Designs on Perceived Relationship Investment Perceived relationship investment. The different gift designs may exert differential impacts on perceived relationship investment, 2 Journal of Service Research which is defined as ‘‘a customer’s perception of the extent to which a [company] devotes resources, efforts, and attention aimed at maintaining or enhancing relationships with regular customers that do not have outside value and cannot be recovered if these relationships are terminated’’ (De Wulf, Odekerken-Schro¨der, and Iacobucci 2001, p. 35). Such relationship investments foster psychological bonds on the customer’s side that likely encourage him or her to continue the relationship with the firm, stimulating reciprocation (De Wulf, OdekerkenSchro¨der, and Iacobucci 2001). Gifts also initiate both cognitive and affective customer processes (Beltramini 1992; Bodur and Grohmann 2005; Dorsch and Kelley 1994). Customers engage in active reasoning about and update their relationship with the firm through cognitive processing, while affective processes prompt them to experience positive emotions in response to the gifts. Consistent with extant research, we argue that these processes should lead to positive customer perceptions of the gift-giving company, including perceptions of its investment in the relationship (Palmatier et al. 2006; Yoon, Choi, and Sohn 2008). However, the effect sizes may vary across different gift designs, as well as with the congruence between the gift dimensions, which we describe next. Gift designs. Critical dimensions for classifying gift designs include gift type (Belk and Coon 1993; Berry 1995) and gift relatedness (Keh and Lee 2006; Yi and Jeon 2003). Rather than predicting main effects of each dimension, we consider their congruence (or fit), with the prediction that it is the combination of these design elements that matters. The first dimension, gift type, refers to the basic motivation and relational bonds that a gift reflects or targets. Drawing on Berry’s (1995) levels model, we distinguish between economic and social motives that underlie exchange partners’ behavior. Economic gifts focus on financial incentives; social gifts might involve communication with customers or events for social bonding. Social gifts thus entail greater personalization and are more relationship oriented than economic gifts, which make them harder for other firms to imitate (Berry 1995). This classification also appears in other relationship marketing research (e.g., Bolton, Lemon, and Verhoef 2004; Melancon, Noble, and Noble 2011; Verhoef 2003). These gift type dimensions correspond with fundamental models of economic and social exchange (Belk and Coon 1993). An economic exchange of gifts implies that customers perceive the gift as a commodity with economic and utilitarian value. The gift exchange commoditizes the exchange partner, encouraging a market economy logic (Belk 2010; Belk and Coon 1993). A social exchange of gifts instead means that a gift is a token of symbolic value, so the exchange partner becomes part of the customer’s extended self in a social exchange, and the exchange follows a moral economy logic (Belk 2010; Belk and Coon 1993). Some scholars use these fundamental exchange models to characterize different types of service relationships too (i.e., economic–functional vs. social–communal; Goodwin 1996; Goodwin and Gremler 1996). We do not argue here that either economic or social gifts are more effective, because both can enhance customers’ perceptions of relationship investment (De Wulf, Odekerken-Schro¨- der, and Iacobucci 2001). However, providing either economic or social gifts should influence customer perceptions of the kind of exchange or relationship that the company desires. If the company offers an economic gift (e.g., coupon), it suggests Figure 1. Theoretical model of the impact of corporate gifts on customer relationship perceptions, behavior, and financial outcomes. Pre ¼ before the experiment and post ¼ after the experiment. The bold lines represent our hypothesized paths and the dashed lines are the control paths. Marchand et al. 3 an economic exchange; if it offers a social gift (e.g., invitation to a social event), it suggests a more social exchange. The second gift design dimension, gift relatedness, refers to the link between a gift and the gift-giving firm, as well as its products and services, which can be related or unrelated (Keh and Lee 2006; Yi and Jeon 2003). We do not predict that either related or unrelated gifts are more effective per se, because both company related (e.g., free company products) and company unrelated (e.g., free products from other companies) gifts can increase perceptions of relationship investment (Seipel 1971). But providing either related or unrelated gifts should influence the associations that customers assign to the gift (Bodur and Grohmann 2005). With a related gift, the company becomes the center of attention, which may create perceptions of self-interest or commercial intent. If the company instead offers an unrelated gift, it directs customers’ attention to another entity (e.g., another company or product), which should decrease perceptions of self-interest and commercial intent (Belk and Coon 1993; Tesser, Gatewood, and Driver 1968; Trawick, Swan, and Rink 1989). Congruence. Building on the concept of fit—or the perceived degree of congruence between gift dimensions and the perceptions and associations they provoke—we argue that the effectiveness of gift designs depends on whether their dimensions appear congruent or incongruent. Cognitive categorization theory asserts that customers classify cues into categories to facilitate information processing (Cohen and Basu 1987; Sujan 1985). Congruent cues share content and meaning, so unlike incongruent cues, they can reduce customer confusion and may lead to more positive cognitive and affective customer processes that could increase perceptions of relationship investment (De Wulf, Odekerken-Schro¨der, and Iacobucci 2001; Reinholtz, Bartels, and Parker 2015). Perceptions of incongruence might arise if customers misinterpret cues or a firm offers discrepant cues (Bodur and Grohmann 2005). Fit is prominent in many research contexts, including brand extensions (Spiggle, Nguyen, and Caravella 2012) and promotions (Chandon, Wansink, and Laurent 2000), but we know of no study that has addressed fit across gift dimensions and its effect on customers. We argue that customers perceive economic related and social unrelated gifts as more congruent than economic unrelated and social related gifts, and this congruence then should increase the degree of customer-perceived relationship investment. When a company offers an economic gift, it suggests an economic exchange, which fits associations of self-interest and commercial intent prompted by a related gift. With a social gift, a company suggests a more social exchange, which fits the reduced associations of self-interest and commercial intent that come about from an unrelated gift. In contrast, a social related gift may appear manipulative or opportunistic due to the misfit between a social exchange and commercial associations (Belk and Coon 1993; Dorsch and Kelley 1994). As prior research shows, customers respond negatively when they attribute manipulative or opportunistic motives to a company’s social– communal behaviors (Campbell 1995; Deighton and Grayson 1995; Goodwin 1996). Finally, an economic unrelated gift may be ineffective or confusing for customers (Bodur and Grohmann 2005), because the company implies an economic exchange mode but then directs attention to other companies or products, contrary to what customers expect and difficult to encode as a relationship investment (Burke and Srull 1988; Thorndyke and Hayes-Roth 1979). Resource exchange theory, mental accounting, and prospect theory also reinforce such arguments (Smith, Bolton, and Wagner 1999). According to mental accounting principles, customers assign economic and social resources to different mental accounts (Thaler 1985). Resource exchange theory states that customers prefer resources that are from proximal (similar) instead of distal (dissimilar) categories (Brinberg and Castell 1982). Prospect theory predicts that resources get evaluated differently, according to their utility (Kahneman and Tversky 1979). Overall then, these theories suggest that customers perceive higher utility for exchanges involving proximal resources than those involving distal resources. We expect that customers consider economic (social) resources more proximal to related (unrelated) gifts, so economic related and social unrelated gifts should be perceived as having more utility than economic unrelated and social related gifts. Formally, Hypothesis 1: The effect of gift type on perceived relationship investment is moderated by gift relatedness, so economic gifts have a stronger impact when they are related, and social gifts have a stronger impact when they are unrelated. Mediating Perceptions Between Gifts and Customer Behavior Company actions may influence customer behavior directly and/or indirectly. Corporate gifts likely influence customer relationship behavior (i.e., customer spending) indirectly through customer relationship perceptions. Building on extant gift research and the theory of reasoned action (Bodur and Grohmann 2005; Fishbein and Ajzen 1975; Sherry 1983), we focus on two perceptual constructs that represent cornerstones of this mediation between gifts and customer behavior: perceived relationship investment (De Wulf, Odekerken-Schro¨der, and Iacobucci 2001) and repurchase intention (i.e., a customer’s willful choice to deal with a company again; Bagozzi 1983; Oliver and Swan 1989). Specifically, gifts should affect customers’ perceptions of relationship investment, which then influence repurchase intentions, which then affect customer spending. In line with the norm of reciprocity (Gouldner 1960; Palmatier et al. 2009), perceived relationship investment should have a positive impact on repurchase intention (Bagozzi 1995). Reciprocal repurchase intention can be motivated by gratitude, trust, commitment, or even feelings of guilt in response to the company’s relationship investment (Dahl, Honea, and Manchanda 2005; De Wulf, Odekerken-Schro¨der, and Iacobucci 2001; Sherry 1983). We thus expect repurchase intention to 4 Journal of Service Research increase with greater perceived relationship investment (Palmatier et al. 2009). In line with the theory of reasoned action (Fishbein and Ajzen 1975) and empirical marketing research (Bolton, Kannan, and Bramlett 2000; Morwitz, Steckel, and Gupta 2007; Sheppard, Hartwick, and Warshaw 1988), repurchase intention should influence customer spending, a key customer relationship behavior (Zhang and Breugelmans 2012). In contrast with such a mindset-oriented perspective, where gifts influence behavior indirectly through perceptions, a behaviorist perspective predicts that gifts also influence customer spending directly, without changing customer perceptions of the company. Such an additional direct link would imply that gift effects on behavior are only partially mediated by perceptions. Taylor and Neslin (2005) denote this direct link between reward and action a ‘‘rewarded-behavior impact’’ that results from behavioral reinforcements. Stimulated by research into operant (Skinner 1953) and Pavlovian (Pavlov 1927; Rescorla 1988) conditioning, both the behavior modification perspective (Nord and Peter 1980) and behavioral learning theory (Rothschild and Gaidis 1981) assert that customers can act without active reasoning about a reward. Yet we believe this behaviorist perspective may be unrealistic in a gift-giving setting. Corporate gifts communicate appreciation and gratitude, entail ritual and ceremony, and often come as a surprise, such that a mindset likely is always involved in the process of gift receiving (Sherry 1983). We thus posit that an offering of gifts involves customer perceptions and expect a full-serial mediation of gifts on behavior through customer relationship perceptions. Hypothesis 2: The effects of the different gift designs on customer spending are fully serially mediated by perceived relationship investment and repurchase intention. Testing the Model: A Field Experiment Experimental Manipulations To test our model, we conducted a 2 2 between-subjects field experiment in cooperation with an international airline company. The airline industry is one of the largest global industries with US$733 billion in revenues in 2014 [International Air Transport Association (IATA 2015)]. In close cooperation with the airline’s customer relationship managers, we developed realistic manipulations for each of the four combinations of gift type (economic vs. social) and gift relatedness (related vs. unrelated). All four gifts cost the airline about the same. To ensure equivalence in value perceptions from the customer perspective, we ran a qualitative pretest with 11 potential airline customers and four marketing practitioners. During focus group interviews, we presented the focal gifts, along with some slightly different variants of the economic gifts with higher and lower redemption value, and asked the participants whether these gifts were roughly equivalent in value for customers of this airline. At the end of the interview procedures, we selected the four gifts all participants agreed were similarly valuable and preferable (Keh and Lee 2006). The experimental manipulations are displayed in Table 1. The economic related gift offered flight coupons; the economic unrelated gifts provided participants with either a coupon for a free magazine subscription (choice among different titles) or an upgrade from a major car rental company. All the economic gifts had monetary redemption values of 10–20€. Participants in the social related group received an invitation to participate in an exclusive event: an Internet chat with the company’s executive chairperson. The social unrelated gift group received a big chocolate heart by mail. All incentives were immediately redeemable (Keh and Lee 2006); the economic gifts could also be redeemed for a period of up to 24 months. We randomly assigned each participant to one of the four experimental groups or else the control group. In all cases except for the control group, participants received the gifts together with a letter, in which the company thanked them for their loyalty and stressed the importance of having them as customers. The text was equally personalized for each gift. The subject line was consistently, ‘‘A little thank you for being a [airline name] customer.’’ The letter then started by thanking the customer for being loyal to the company and used that loyalty as the motivation for presenting him or her with a gift. All gifts were referred to as a ‘‘small surprise.’’ Next, the letter described the assigned gift. For the economic gifts (coupons), it detailed how to redeem them, and for the social related gift (chat), it described how to set up access to the chat. The letter ended with a general appreciation and a final sentence: ‘‘We are looking forward to continue welcoming you on board of [airline name].’’ Saying ‘‘thank-you’’ and referring to a ‘‘small surprise’’ are typical, ritual elements of gift giving (Sherry 1983). The control group received no communication, because a letter often is an essential part of a gift exchange and can be perceived as a gift or appreciation in itself (Liu, Lamberton, and Haws 2015). Study Design As illustrated in Figure 2, we collected longitudinal data from two surveys and objective spending and cost data from the company’s database, which helped us control for selfselection and common method biases and ensure high internal and external validity. All participants received e-mail invitations to complete an Internet survey at t0. We sent the invitation Table 1. Experimental Manipulations. Gift Type Economic Social Gift relatedness Related Flight coupon Exclusive Internet chat with the company’s CEO Unrelated Coupon for free magazine or car rental upgrade Unbranded chocolate heart Marchand et al. 5 from our university address to emphasize the scientific motivation for the study. Ten weeks after sending the first survey, we started the experimental manipulations at t1 by sending the respective gifts to the participants. Unlike the first message (t0), this letter (t1) came from the airline and did not include any connection to the first survey to avoid response biases. We ended the experiment after 24 weeks, then sent out the second survey (t2). We allowed 8 weeks for responses to arrive and then collected spending data from the database for the next 12 weeks (t3–t4). This 12-week data range seems appropriate for several reasons. First, the coupons were only redeemable during the experiment (until t2), so they did not bias the data during t3–t4. Second, we split the data from the 52 weeks before the experiment (t1) into four quarters, then calculated the mean values of customer spending for each quarter and the whole year (Mq1 ¼ 110.74, Mq2 ¼ 123.34, Mq3 ¼ 126.77, Mq4 ¼ 119.12, and Myear ¼ 119.99). A t-test revealed that the differences between the individual quarters and the year were not significant. Third, the mean number of bookings per customer in these 52 weeks was 6.25, equivalent to about 1.44 bookings every 12 weeks. Therefore, on average, every customer was reasonably likely to book at least one ticket during the 12 weeks. Fourth, the risk of bias due to other company actions or external effects, such as economic changes or actions by competitors, is lower for 12 weeks than for longer periods. In this sense, the 12-week observation of behavioral changes is more internally valid. Fifth, other researchers also use 12 weeks as an appropriate observation window (e.g., Liu 2007; Nunes and Park 2003; Sloot, Fok, and Verhoef 2006). To ensure causality between perceptions and spending behavior, we did not use any spending information before t3 for our postmanipulation spending behavior measure. Data for additional weeks after t4 were not available, because the company’s customer database converted over to a new system thereafter. In addition, we measured preexperiment customer spending behavior during the 52 weeks before the experimental manipulation in t1. Because the company does not treat redeemed coupons as revenue, we do not include them in our customer spending measure. Sample We contacted all customers listed in the airline’s database via e-mail and 9,189 of them filled out the first questionnaire. To restrict our sample to customers who were identifiable and active, we matched these survey data with the company database and retained only those 4,521 customers who used the airline during the 12 months prior to the first survey. To control for nonresponse bias, we compared the data from early and late survey respondents, following Armstrong and Overton’s (1977) procedure. The t-tests indicated no statistically significant differences in customer relationship perceptions, past customer spending, or the status importance of early and late respondents prior to the experimental manipulation. The data thus are sufficiently free of nonresponse bias to permit further analyses. The 1,983 participants who also completed the second survey produced a response rate of 43.9%. After eliminating respondents with implausible answers (i.e., those who claimed to have booked more than 300 flights per year with the airline, who offered the same response for all questions, or who completed the questionnaire in less than 7 minutes), we retained 1,950 participants in the final sample. Participants were randomly assigned to the economic related (n ¼ 385), economic unrelated (n ¼ 401), social related (n ¼ 365), social unrelated (n ¼ 432), and control (n ¼ 367) groups. The mean age of the final sample was 40 years, and 61% were men. Measures We used effect coding for the two independent categorical variables for each gift dimension: gift type with 1 ¼ social, 1 ¼ economic, and 0 ¼ control group and gift relatedness with 1 ¼ unrelated, 1 ¼ related, and 0 ¼ control group. We Figure 2. Study design. 6 Journal of Service Research measured customer spending behavior twice, before and after the manipulation (our dependent variable). Customer spending equaled the actual amount a customer spent on flights with the airline in a given time period, which we gathered from the airline’s customer database. We also measured the relationship perception constructs both before and after the experimental manipulation with established multi-item scales. We therefore could account for changes during the experiment and include the preexperimental values as covariates in the model to rule out alternative explanations. Specifically, to measure perceived relationship investment, we used a reflective, 3-item scale from De Wulf, Odekerken-Schro¨der, and Iacobucci (2001). We measured participants’ repurchase intention with a reflective, 3-item scale adapted from Oliver and Swan (1989). These survey items all used 7-point Likert-type scale, ranging from 1 ¼ strongly disagree to 7 ¼ strongly agree. As an additional covariate, we include status importance, defined as the relevance of a prestigious position in a society (Dre`ze and Nunes 2009; Eastman, Goldsmith, and Flynn 1999). For status importance before the experiment, we used 1 item based on Eastman, Goldsmith, and Flynn (1999) and Priester et al. (2004) measured on a 7-point Likert-type scale, ranging from 1 ¼ not at all important to 7 extremely important. We included customer spending before the experiment as a covariate of customer status level (Dre`ze and Nunes 2009). That is, this measure indicates a customer hypothetically would have qualified for higher status, had the airline offered a tiered frequent flyer loyalty program. We also measured the number of tickets purchased and obtained very similar results; ultimately, we chose to use spending, because it includes more information (frequency transaction size). All items are provided in Table 2. Because the airline was located in Germany, all items were translated into German, using a back-translation approach to ensure conceptual equivalence. Unlike many survey settings (e.g., student samples), we had only one chance to obtain responses during this unique collaboration with the airline. We therefore took every effort to optimize the comprehensibility of the survey (e.g., introductions before the questions) and purify the multi-item scales. In a first pretest, we asked 12 experts (service researchers and airline managers) to comment Table 2. Measures. Construct Measurement Scale Type Based on Customer spending Actual amount a customer has spent on flights with the airline in a given time period. Metric (0 to 1) Airline’s customer database Gift type 1 ¼ social gift, 1 ¼ economic gift, and 0 ¼ control group Effect coding (1,0,1) Berry (1995); Melancon, Noble, and Noble (2011) Gift relatedness 1 ¼ unrelated gift, 1 ¼ related gift, and 0 ¼ control group Effect coding (1,0,1) Keh and Lee (2006); Yi and Jeon (2003) Perceived relationship investment [Company] makes efforts to increase regular customers’ loyalty. [Company] makes various efforts to improve its tie with regular customers. [Company] really cares about keeping regular customers. Reflective (1–7) De Wulf, Odekerken-Schro¨der, and Iacobucci (2001) Repurchase intention It is likely that I will fly [company] regularly. I probably will use [company] for a long time. In the future, I will fly [company] to a lesser extent. Reflective (1–7) Oliver and Swan (1989) Status importance How important is a frequent-flyer program to you? Single item (1–7) Eastman, Gold-smith, and Flynn (1999); Priester et al. (2004) Manipulation Check 1: Social motivation The reward creates a feeling of attachment to [company] or other people there. The reward allows me to have enjoyable interactions with the employees or other customers. Reflective (1–7) Paul et al. (2009) Manipulation Check 2: Economic motivation The reward helps me to save money. Single item (1–7) Paul et al. (2009) Manipulation Check 3: Value Cash value: I find [the gift] valuable. Relevance value: [The gift] was exactly what I wanted. Convenience value: [Company] made it easy to receive [the gift]. Reflective (1–7) O’Brien and Jones (1995); Yi and Jeon (2003) Manipulation Check 4: Psychological benefits Confidence: I will always be able to rely upon that [company] knows what it is doing. Privilege: [Company] gave me the feeling to be a preferred customer. Welcomeness: [Company] gave me the feeling to be a welcomed customer. Single item (1–7) Paul et al. (2009) Marchand et al. 7 on the scale items. For a second pretest, we asked 217 customers of the airline—excluded from the main study—to rate the items, then used their responses to calculate scale validation measures and to assess the reliability and validity of the scales. On the basis of these pretests, we made minor changes in the wording of the introductory text and items. For all reflective constructs, the Cronbach’s as were .88 or higher, the average variances extracted were greater than .88, and the Fornell and Larcker (1981) criterion of discriminant validity—which requires the squared correlation of two constructs to be lower than the variance of both constructs explained by its indicators—was met. We list the descriptive statistics, Cronbach’s as, and bivariate correlations for all model constructs in Table 3. Manipulation Checks To test the effectiveness of our gift manipulations, we asked the participants in the experimental groups to rate, in the second survey, the degree to which the gift they received was economic or social in nature. For this measure, we adapted 2 reflective items for social benefits and 1 item for economic benefits from Paul et al. (2009), then used the mean of the 2 social benefits items (see Table 2). The manipulations were successful, because customers who received a social gift (S) perceived more social benefits (SB) than did those with an economic gift (E; xSB/E ¼ 3.19 < xSB/S ¼ 3.48, F ¼ 11.44, p < .01, Z2 ¼ .009). Customers who received an economic gift perceived more economic benefits (EB) than did those who received a social gift (xEB/E ¼ 3.46 > xEB/S ¼ 2.63, F ¼ 57.07, p < .01, Z2 ¼ .041). Following Yi and Jeon (2003), we refrained from checking the gift relatedness manipulation, considering its obviousness. We conducted confound checks by comparing the effect sizes for unintended and intended manipulations (Perdue and Summers 1986). The differences between unrelated and related gifts were insignificant for both economic (p > .10, Z2 ¼ .002) and social (p > .10, Z2 ¼ .001) gifts, with substantially smaller effect sizes than the intended manipulations, which suggested no confounding problems. Furthermore, we checked whether customers perceived the value of economic versus social and related versus unrelated gifts as equivalent. To capture various value perceptions, we measured value using the mean score of three dimensions from O’Brien and Jones (1995) and Yi and Jeon (2003): cash value, relevance value, and convenience value on a 7-point scale (a ¼ .77). The differences in value between economic versus social (xvalue, E ¼ 3.90, xvalue, S ¼ 3.90; F ¼ .004, p > .10, Z2 < .001) and related versus unrelated (xvalue, related ¼ 3.89, xvalue, unrelated ¼ 3.91; F ¼ .048, p > .10, Z2 < .001) gifts were not significant. In addition, we tested psychological benefits (Paul et al. 2009), a third type of customer benefit in a service relationship context, in the form of confidence, privilege, and welcomeness with 1 item each. We did not seek to manipulate these perceptions, so they should not be affected by the gift type. We find no significant differences (p > .10). The items for the manipulation checks are in Table 2. Model We test our hypotheses with (M)ANCOVAs and a serial mediation analysis (Hayes 2013). Following our theoretical arguments, we model three equations (Equation 1 for Hypothesis 1; all equations for Hypothesis 2 and the contribution margin predictions): M1 ¼ iM1 þ a1X þ X q i¼1 fiCi þ eM1 ; ð1Þ M2 ¼ iM2 þ bM1 þ a2X þ X q i¼1 giCi þ eM2 ; ð2Þ Y ¼ iY þ c1M1 þ c2M2 þ d0 X þ X q i¼1 hiCi þ eY ; ð3Þ Table 3. Descriptive Statistics and Correlations. Variable M SD a 1a 1b 1c 2a 2b 3a 3b 4a 4b 5 1a Gift type .01 .90 n.a. 1 1b Gift relatedness .04 .90 n.a. .032 1 1c Gift type Gift relatedness .03 .90 n.a. .052* .006 1 2a Perceived relationship investment pre 4.23 1.43 .933 .036 .013 .003 1 2b Perceived relationship investment post 4.82 1.38 .937 .009 .022 .062* .489* 1 3a Repurchase intention pre 5.43 1.31 .905 .001 .014 .016 .452* .318* 1 3b Repurchase intention post 5.06 1.35 .886 .009 .016 .023 .278* .487* .496* 1 4a Customer spending pre 396.44 757.62 n.a. .004 .033 .012 .085* .077* .101* .117* 1 4b Customer spending post 62.08 163.38 n.a. .007 .048* .037 .086* .051* .014 .087* .474* 1 5 Status importance 4.58 2.01 n.a. .014 .014 .024 .013 .027 .083* .087* .189* .134* 1 1a 1b 1c 2a 2b 3a 3b 4a 4b 5 Note. a ¼ Cronbach’s a for reflective multi-item scales. Customer spending pre encompasses 52 weeks before the experimental manipulation in t1 and customer spending post covers 12 weeks after the experimental manipulation (t3 to t4). n.a. ¼ not applicable; M ¼ mean; SD ¼ standard deviation. *p < .05 (two-sided). 8 Journal of Service Research where X ¼ the independent variable (gift), M1 ¼ the first mediator (perceived relationship investment post); M2 ¼ the second mediator (repurchase intention post); Y ¼ the dependent variable (customer spending post); C is the set of q covariates (q  1), with coefficients f, g, and h; i ¼ regression intercepts; e ¼ errors in the estimations; and the path relations are a1, a2, b, c1, c2, and d0 as shown in Figure 1. The model estimation is not affected by multicollinearity; all the variance inflation factors are below 1.7. A pooling test for overall homogeneity using the iterative generalized least squares procedure proposed by Gatignon and Reibstein (1986) reveals no significant differences for the error sums of squares obtained across the four separate regressions of Models 1–3 for each gift design and the pooled model with all gift designs. Therefore, pooling appears appropriate. Empirical Results Differential Effects Following a traditional approach, we first conducted a MANCOVA for the dependent variables of perceived relationship investment post, repurchase intention post, and customer spending post, with status importance, perceived relationship investment pre, repurchase intention pre, and customer spending pre as covariates. The interaction of gift type and gift relatedness reveals significant differences in the dependent variables, (Wilks’s L ¼ .994; F(3, 1,939) ¼ 4.064, p < .01) no main effects of gift type and gift relatedness reach significance. We then inspected univariate ANCOVAs for each dependent variable while controlling for the effects of the same covariates. The interaction of gift type and gift relatedness is significant for perceived relationship investment post, F(1, 1,941) ¼ 9.674, p < .01, but not for repurchase intention post, F(1, 1,941) ¼ 2.338, p ¼ .13, or customer spending post, F(1, 1,941) ¼ 2.584, p ¼ .11. We include the control group in all these analyses.3 The MANCOVA provides initial support for our prediction of no main effects of the gift dimensions, their interaction is the key, and the effects only exist for perceived relationship investment. To test the postulated differences in effectiveness across gifts and the control group, we ran further ANCOVAs, as well as simple slope/spotlight analyses, to investigate the interaction effect more deeply (Aiken and West 1991). For the test of Hypothesis 1, we focus on the differential effects of gifts on perceived relationship investment. With the ANCOVAs, we can investigate the effect of the different gifts on perceived relationship investment and use the control group to determine whether gifts outperform a benchmark without gifts. The results (upper part of Table 4) show that the mean perceived relationship investment is significantly higher for all gift groups than for the control group. Figure 3 displays the slope plot for the interaction effects. The spotlight analysis (lower part of Table 4) using Model 1 (perceived relationship investment post) shows that the differences move in the proposed directions and are significant for perceived relationship investment between social unrelated and economic unrelated (t ¼ 3.034) gifts and social unrelated and social related (t ¼ 2.568) gifts. The differences between Table 4. Differential Effects for Perceived Relationship Investment. ANCOVA Comparison With Control Group Cases M SD F p Control group 367 4.544 1.368 Economic related gift 385 4.933 1.347 22.782* <.001 Economic unrelated gift 401 4.806 1.362 9.152* .003 Social related gift 365 4.762 1.389 10.594* .001 Social unrelated gift 432 5.005 1.393 33.974* <.001 All gift groups 1,583 4.881 1.375 28.370* <.001 Simple Slope Analysis At B tp Slope of gift type Related gift .049 1.212 .226 Slope of gift type Unrelated gift .122* 3.034 .002 Slope of gift relatedness Economic gift type .067 1.669 .095 Slope of gift relatedness Social gift type .104* 2.568 .010 Note. M ¼ mean; SD ¼ standard deviation; B ¼ unstandardized coefficient; ANCOVA ¼ analysis of covariance. *p < .05 (two-sided). Figure 3. Slopes for Gift relatedness Gift type. Marchand et al. 9 Table 5. Serial Mediation Analysis Results. Model: Outcome Variables M1: Perceived relationship investment post M2: Repurchase intention post Y: Customer spending post Coefficient SE t Statistic p Value VIF Coefficient SE t Statistic p Value VIF Coefficient SE t Statistic p Value VIF Constant 2.429* .131 18.482 <.001 1.062* .127 8.371 <.001 22.697 17.460 1.300 .194 Perceived relationship investment post .408* .020 2.199 <.001 1.349 2.131 3.004 .709 .478 1.632 Repurchase intention post 8.746* 3.071 2.849 .004 1.631 Gift type Gift relatedness .095* .030 3.164 .002 1.004 .006 .027 .233 .816 1.009 5.706 3.619 1.577 .115 1.009 Gift type .032 .030 1.078 .281 1.006 .025 .027 .952 .341 1.006 .567 3.613 .157 .875 1.007 Gift relatedness .018 .030 .589 .556 1.003 .010 .027 .384 .701 1.003 5.863 3.609 1.624 .105 1.003 Status importance .015 .014 1.116 .265 1.044 .032* .012 2.595 .009 1.044 3.630* 1.654 2.195 .028 1.048 Perceived relationship investment pre .408* .021 19.109 <.001 1.289 .091* .021 4.402 <.001 1.531 4.081 2.812 1.451 .147 1.546 Repurchase intention pre .142* .024 6.027 <.001 1.297 .407* .021 19.309 <.001 1.321 6.071 3.115 1.949 .052 1.575 Customer spending pre .001* .001 2.498 .013 1.067 .001* .001 5.009 <.001 1.071 .098* .005 22.075 <.001 1.084 R2 .259 .387 .235 F 96.732* 153.197* 66.170* Bootstrap Analyses Effect Boot SE Boot CI 95% [LL, UU] Boot CI 99% [LL, UU] Indirect effect of Gift type Gift relatedness through perceived relationship investment post and through repurchase intention post on customer spending post .338 .156 [.107, .741] [.051, .884] Indirect effect of Gift type Gift relatedness through perceived relationship investment post on repurchase intention post .038 .012 [.015, .063] [.007, .071] Direct effect of Gift type Gift relatedness on perceived relationship investment post .095 .030 [.036, .154] [.018, .172] Direct effect of Gift type Gift relatedness on repurchase intention post .006 .027 [.046, .059] [.062, .075] Direct effect of Gift type Gift relatedness on customer spending post 5.706 3.619 [1.391, 12.802] [3.625, 15.036] Note. The coefficients are unstandardized. Because of the effect coding, the interaction term of gift type (1 ¼ social, 0 ¼ control group, and 1 ¼ economic) and gift relatedness (1 ¼ unrelated, 0 ¼ control group, and 1 ¼ related) equals 1 for social unrelated and economic related gifts and 1 for social related and economic unrelated gifts. Thus, it is not possible to interpret the effects of specific gifts from this result; please refer to Table 4 and Figure 3. Boot ¼ bootstrap analyses based on 20,000 resamples; LL ¼ lower limit; UL ¼ upper limit; SE ¼ standard error; VIF ¼ variance inflation factor; CI ¼ confidence interval. *p < .05 (two-sided). 10 economic unrelated and economic related gifts are marginally significant at p < .10 (t ¼ 1.669). Overall, these results provide support for Hypothesis 1. Serial Mediation Analysis To examine the mediation we proposed in Hypothesis 2, we simultaneously applied Models 1–3 in a serial mediation analysis. All three models are significant (FM1 ¼ 96.732, FM2 ¼ 153.197, and FY ¼ 66.170). To investigate the mediating role of perceived relationship investment and repurchase intention, we applied the Preacher–Hayes procedure ‘‘process’’ Version 2.13 (Hayes 2014), with bootstrap analyses based on 20,000 resamples. When the bias-corrected confidence interval (CI) excludes 0, significant mediation exists. As we detail in Table 5, we find a positive, serial mediation effect with a 99% CI [.051, .884] for gifts (Gift type Gift relatedness) through perceived relationship investment post and through repurchase intention post on customer spending post. The direct effect from gift to spending is not significant for a 95% CI [1.391, 12.802], so the results suggest indirect-only (i.e., full) mediation (Zhao, Lynch, and Chen 2010) in support of Hypothesis 2. Revenue and Contribution Margin Predictions Our theoretical model (Figure 1) includes the contribution margins of corporate gifts, which are relevant financial outcomes for firms. Building on the results of the serial mediation analysis and accounting for marginal costs, we investigate the strength of the effect of different gift designs on customer spending by predicting postmanipulation revenues and contribution margin changes. To ensure consistency, we simulated the results of our Y model for customer spending post (Table 5) by calculating the model for each gift design and subtracting the mean spending of the control group, which reveals the incremental gift revenues. Using the economic related gift led to an incremental revenue increase of 19.92€. According to the airline, the effective marginal costs were 2.80€ per customer for economic related gifts (less than the value of the voucher, which not all customers redeemed), suggesting a contribution margin increase of 17.12€ per customer. Economic unrelated gifts instead led to a revenue decrease of 3.22€, even before the marginal costs of 3.09€. Social related and social unrelated gifts prompted revenue increases of 9.64€ and 9.32€, respectively, with marginal costs of .30€ and 1.56€, such that their contribution margins were 9.34€ and 7.76€. Overall, economic related gifts result in the greatest increase in contribution margins. We limit our analysis to marginal costs, because fixed costs become irrelevant across many customers, as is usually the case for mass service providers such as airlines. When accounting for total costs, the costs are highest for the social related manipulation, because it demands the development and implementation of a chat interface (11.80€), which could be used again though. The costs are virtually equal for the other three manipulations (economic related 3.10€, economic unrelated 3.50€, and social unrelated 2.96€). Conclusion and Implications Conclusion Using a longitudinal field experiment that combines experimental manipulations with survey data and actual customer spending information, our study extends the literature on corporate gifts and relationship marketing instruments by examining how different gift designs influence customer perceptions and behaviors, a key issue for relationship marketing theory and practice. The results from a large-scale field experiment offer support for our hypotheses. They also reveal that gift designs differ in their financial outcomes, an important insight for service managers. The impact of gift types on perceived relationship investment is moderated by gift relatedness, such that economic gifts have a stronger impact when they are related, but social gifts have a stronger impact when they are unrelated, confirming that a high fit between gift dimensions is crucial for effective gift, giving in a relationship marketing context. Regarding the serial mediation of gifts on customer relationship perceptions and behavior, we find that the link between gifts and customer spending is fully serially mediated by perceived relationship investment and repurchase intention, confirming a mindsetoriented perspective that builds on the principle of reciprocity. Furthermore, economic related gifts prompt the greatest increase in the contribution margin, a finding in line with research that shows that economic gifts affect customer behavior directly (Berry 1995) and that resources perceived as similar (e.g., coupons and money) are more likely to be exchanged (Brinberg and Castell 1982; Sherry 1983). Implications for Relationship Marketing Management Our results show that gifts can powerfully influence important customer perceptions and spending. They have substantial positive effects on company revenues and contribution margins. Generally speaking, gifts deserve relationship marketing managers’ attention and budget allocations, but managers must also realize that there is no such thing as a ‘‘generalizable’’ effect of gifts. Instead, their results strongly depend on the gifts’ design and underlying dimensions. Understanding the differential effects of various gift designs represents a key task for managers and researchers alike. We recommend that managers select economic related and social unrelated gift designs over economic unrelated (e.g., coupons for products from other companies) and social related (e.g., exclusive events with company chairpersons) ones. This finding is situated in theoretical arguments, so it should transfer to other service industries and settings too. In our study setting, economic related gifts (e.g., coupons) work most effectively in terms of revenues and contribution margins, and social unrelated gifts (e.g., unbranded chocolate heart) appear most effective for relationship perceptions. In contrast, the economic unrelated gift even decreases revenues, which should inform managers’ decisions when designing gifts and allocating budgets. The negative impact for economic Marchand et al. 11 unrelated gifts may indicate that it causes a shift in customers’ interest to other companies, which in our case implied traveling by rental car instead of air travel. Economic gifts seem to function similarly to other monetary incentives, leading customers to adapt their behavior in the way desired by the company, because their loyalty ‘‘pays off.’’ However, managers should be aware that in the long run, repeated (instead of one time) gifts for customer loyalty might be necessary, otherwise, the reciprocity process could wear out. Initial empirical evidence identifies wear-out in loyalty programs (Meyer-Waarden and Benavent 2009). The increasing costs of repeated gifts should enter into profitability calculations with a long-term perspective. Determining how to account for them represents a fruitful challenge for research. In a related sense, attitudinal effects by definition tend to remain more stable over time and may pay off over a longer term, so our study design (which considers customers’ reactions for a limited time) may systematically disadvantage the long-term role of customer perceptions. Social unrelated gifts thus may become just as profitable as economic related gifts over an extended period. We encourage managers to include perceived relationship investment in their customer tracking activities too, together with the more prevalent constructs used to measure their satisfaction and loyalty. Limitations and Further Research The significance of the interaction between gift type and gift relatedness suggests that complex relations are at work, not just main effects, due to individual gift dimensions—a point that deserves more research attention. Similarly, our theoretical model might be extended to other gift designs. For example, we focus on immediate gifts, not delayed gifts, which may exert a ‘‘pressure effect’’ (Liu 2007). Studies that disentangle the direct and indirect effects on customer relationship behaviors thus might generalize our findings about the serial mediation to other gifts or relationship marketing instruments as well as more general situations. Our findings stress the importance of perceived relationship investment as a link between company actions and customer behavior, in support of and extending De Wulf, OdekerkenSchro¨der, and Iacobucci’s (2001) finding that perceived relationship investment is central in customer relationships, such that it depends on the gifts included in our framework but also substantially influences other key customer perceptions. Previous work on perceived relationship investment is relatively limited, so more research on this construct and efforts to measure more specific investment perceptions, such as social bonding, would be valuable. Finally, the context of our empirical work poses some limitations. Our study cannot reveal the ideal frequency of repeated gifts or ideal points to offer them, which are relevant questions for managers. A longer observation window might offer other interesting insights. We also find no effects of customer heterogeneity in a post hoc test of the potential moderating effect of customers’ relationship duration,4 but other forms of heterogeneity might exist. Acknowledgments The authors thank Florian von Wangenheim and Utz Riehl for their helpful comments on this project and the managers of the anonymous airline company for their support. The authors are also grateful to the editor Mary Jo Bitner, the area editor, and the three anonymous reviewers for their constructive comments and suggestions on this research. Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. Notes 1. Economic gifts or relationship marketing instruments have also been referred to as Level 1, financial, monetary, or promotional; social ones have been called Level 2, relational, symbolic, or image enhancing (e.g., Berry 1995; Ga´zquez-Abad, De Cannie´re, and Martı´nez-Lo´pez 2011; Melancon, Noble, and Noble 2011; Verhoef 2003). 2. Related gifts have also been referred to as direct, compatible, or directly related, whereas unrelated gifts have been named indirect, incompatible, or indirectly related. We prefer the terms related and unrelated to avoid confusion with the direct and indirect effects we also investigated in this study. 3. For robustness checks, we calculated the MANCOVA again for only the four experimental groups without the control group. The results are similar for all main effects and the interaction term, (Wilks’s L ¼ .992; F(3, 1,573) ¼ 4.078, p < .01) as they are for the respective analyses of covariance of perceived relationship investment post, F(1, 1,575) ¼ 9.809, p < .01. 4. We test relationship duration as a moderator with a group comparison (t-test) of long-term (self-reported, >12 months at t0) versus new (12 months) customers in each subsample. The effect sizes do not differ significantly.