The Effects of Tangible Rewards Versus Cash Rewards in a Sales Tournament: A Field Experiment

  

Khim Kelly

CPA Ontario Research Fellow

University of Waterloo

kokelly@uwaterloo.ca

 

Adam Presslee

University of Pittsburgh

apresslee@katz.pitt.edu

 

Alan Webb

KPMG Leadership Fellow

University of Waterloo

a2webb@uwaterloo.ca

 

June 2015  

We thank the Company’s owner and operations manager for allowing us access to data and assisting with conducting the experiment. Thanks to Steve Albrecht, Vic Anand, Jason Brown, Willie Choi, Mike Drake, Harry Evans, Josh Gunn, Darren Henderson, Max Hewitt, Scott Jackson, Jason Kuang, Marlys Lipe, Patrick Martin, Jeff McMullin, Doug Prawitt, Vaughan Radcliffe, Joe Schroeder, Geoff Sprinkle, Steve Smith, Bryan Stikeleather, Monte Swain, Tyler Thomas, Kristy Towry, Brad Tuttle, Jeff Wilks, David Wood, Mark Zimbelman, and workshop participants at Brigham Young University, Emory University, Indiana University, Nanyang Technological University, University of Pittsburgh, University of South Carolina, and Western University, and conference participants at the 2015 Management Accounting Section Midyear Meeting for comments on an earlier draft. We also thank Weiming Liu for his research assistance.


ABSTRACT

Firms frequently use tangible rewards to reward employees but minimal research has examined their performance effects relative to cash rewards. Our study investigates the effects of tangible rewards versus cash rewards in consecutive sales tournaments. We conducted a field experiment at a rug wholesaler where two consecutive three-month sales tournaments were held for its retailers. Retailers with similar prior year’s rug sales were organized into competition groups of six or seven. The top three retailers in each competition group after each three-month tournament received either cash rewards or tangible rewards (gift cards) to be distributed to their sales staff. We find that retailers have higher sales when they are competing for gift cards than cash in the second tournament but not in the first tournament. Moreover, the results are driven by the losers of the first tournament. Consistent with our prediction, tournament one losers competing for gift cards significantly increased sales from the first tournament to the second tournament compared to a non-significant change in sales for tournament one losers competing for cash. Of interest to both practice and theory, our results suggest that compared to cash rewards, tangible rewards may be more effective at sustaining effort for competitors who are unsuccessful in an initial competition.


I. INTRODUCTION

            We examine the performance effects of using tangible rewards versus cash rewards in a tournament incentive scheme implemented over two consecutive tournament periods. Tangible rewards are non-cash incentives that are restricted in their use but have a non-trivial monetary value (e.g., travel, merchandise, and gift cards) (Presslee, Vance, and Webb 2013). There is a considerable and growing use of tangible rewards to motivate and recognize good performance in organizations. A recent survey indicates about 75% of U.S. businesses use tangible rewards with estimated annual spending of nearly $77 billion (Incentive Federation Inc. 2013).[1] Proponents of tangible rewards claim they are more motivating than cash rewards because they are typically hedonic in nature, can involve social recognition, and are more distinctive and memorable relative to other compensation elements (e.g., salary) (Jeffrey and Shaffer 2007).[2]

            Although the use of tangible rewards is common in organizations, there has been little research examining their impact on performance. Moreover, the evidence in existing literature is mixed regarding the performance effects of tangible rewards relative to cash rewards in a tournament incentive scheme. For example, in lab experiments using tournament incentive schemes, Jeffrey (2009) reports weak evidence that tangible rewards lead to larger performance improvements than cash rewards, while Shaffer and Arkes (2009) find no effects of reward type on performance. In an effort to provide evidence as to when and how reward type influence behavior in natural settings, we conduct a field experiment where winners in two consecutive sales tournaments received either cash rewards or gift cards that can be spent on hedonic items.

We use a repeated tournament setting to examine possible effects of reward type on performance for three primary reasons. First, tournaments offer the potential to observe direct effects of reward type on effort and performance since tournament outcomes are based solely on relative performance (Berger, Klassen, Libby, and Webb 2013). The ability to observe direct effects of reward type on effort is important given prior research showing that tangible rewards can negatively affect effort and performance indirectly by influencing the difficulty of self-selected performance goals (Presslee et al. 2013). Second, the tournaments used in prior research on reward type employed design features that may have inadvertently limited the likelihood of observing performance effects of tangible versus cash rewards. In particular the low percentage of winners and limited choice of tangible rewards in Shaffer and Arkes (2009) and the absence of feedback during the tournament in Jeffrey (2009) may have reduced competitors’ motivation. As such, there is considerable scope for research employing tournament schemes with stronger motivational properties. Finally, there are theory-based reasons for expecting that reward type will influence the subsequent effort of competitors who perform poorly (e.g., lose) in an initial tournament and employing repeated tournaments allows us to directly evaluate this possibility, which has not been considered in prior research.

Economic theory and mental accounting theory lead to opposite predictions regarding the effects of reward type on performance in the first tournament. Economic theory suggests cash will be more attractive and motivating than tangible rewards because it is perfectly fungible and can be used to purchase anything (Waldfogel 1993). Conversely, mental accounting theory suggests tangible rewards are likely to be more attractive and motivating than cash rewards (Thaler 1985, 1999). Because tangible rewards are distinct from cash earnings they are more likely to be ‘recorded’ in a mental account separate from cash earnings (Helion and Gilovich 2014; Presslee et al. 2013). As a result of this different mental accounting tangible rewards are more likely to be spent on hedonic (fun, pleasurable) items whereas as cash rewards are more likely to be spent on utilitarian things (e.g., groceries, housing) (Helion and Gilovich 2014; Thaler 1985). Because hedonic spending results in stronger affective responses, tangible rewards are likely to be more attractive and memorable than cash, thus inducing greater effort. Given these strong competing theories about which reward type will be more motivating our first prediction is a null hypothesis that reward type (cash versus tangible) will not have a significant effect on performance in the first tournament.

Psychology theory suggests reward type will affect the subsequent effort choices of competitors who perform poorly in the first tournament. This is an important issue given prior research showing that weaker performers often reduce effort (or adopt risky task strategies) in the later stages of a tournament or in subsequent tournaments upon receiving feedback that they are trailing the leaders or that they have lost (Berger et al. 2013; Hannan, Krishnan, and Newman 2008). Of direct relevance to our setting, Rottenstreich and Hsee (2001) show that the type of outcome, affect-rich versus affect-poor, affects the weighting of the probability that those outcomes will occur. Individuals tend to overweight small probabilities when outcomes are affect-rich versus affect-poor because of the high attractiveness of the affect-rich outcomes. In our setting this would result in weaker performers in the first tournament eligible for tangible rewards overweighting the likelihood of succeeding in the second tournament relative to their counterparts eligible for cash rewards. As a result, our second prediction is that any effort reductions from the first tournament to the second tournament by losers of the first tournament will be smaller for those competing for tangible versus cash rewards. Finally, because our first hypothesis is a null prediction and we cannot predict the magnitude of the effects related to our second hypothesis we pose a research question examining the overall performance effects of reward type in the second tournament.

To test our predictions we conducted a field experiment at 54 home furnishings retailers that sell specialty area rugs supplied by a wholesale company (hereafter “the Company”). We worked with the Company to design a tournament sales contest, first organizing the retailers into eight competition groups, with six or seven retailers that had similar prior year’s rug sales in each group. Retailers competed against other retailers in their own competition group, and not against retailers in other groups, in two consecutive three-month tournaments. We then randomly assigned each competition group to one of the two reward-type conditions (four groups in each of the cash and gift cards conditions), holding the reward type assigned to a group fixed across the two consecutive tournaments. At the end of each three-month tournament, the top three retailers in each competition group in terms of rug sales dollars were awarded a prize equal to 15% of total sales, paid either in cash or gift cards from a set of choices. Within each tournament monthly feedback was provided to all retailers including their own total sales and relative ranking to that point in the competition. Winners/losers of each tournament were announced shortly after the end of the third month along with the final sales figure and relative ranking.

             Results from the first tournament fail to reject our null hypothesis that performance will not differ between reward type conditions. Although the descriptive results show retailers eligible for cash rewards outperformed those eligible for tangible rewards, the difference is not significant. We find strong support for our second hypothesis in that losers of the first tournament eligible for tangible rewards improved performance from the first tournament to the second tournament significantly more compared to retailers eligible for cash rewards. Moreover, the magnitude of this effect was sufficiently strong so as to result in losers of the first tournament eligible for tangible rewards outperforming their counterparts in the cash reward condition in the second tournament.

Our study makes three main contributions to the literatures on reward type and tournament incentive schemes. First, our results suggest that tangible rewards in the form of gift cards for hedonic items, when used in conjunction with repeated tournaments, may be an effective means of sustaining effort for weaker performing competitors.  As such our study provides evidence of when and how tangible rewards compared to cash rewards affect performance in organizational settings. In particular, our repeated tournament design illustrates that the effects of reward type may take time to emerge, an outcome that could not be observed in the single tournament settings employed in prior research. Second, we are not aware of other research showing that effort reduction in subsequent tournaments by losers of initial tournaments can be attenuated by the use of tangible rewards. Based on Rottenshtreich and Hsee (2001), we provide an improved theoretical understanding of the factors that influence subsequent effort by poor performers in initial tournaments. Finally, our findings are important from a practical perspective. The use of sales tournament schemes by organizations is common and our results show that tangible rewards, readily implementable in practice, can enhance performance in such settings. Moreover, while prior research has identified several implementable tournament features that can sustain the motivation of stronger performers (e.g., moderate percentage of winners, graduated rewards, precision of feedback) our results speak directly to actions that compensation scheme designers can take to motivate weaker performers. 

The next section presents our research setting, theory and hypotheses. Following that we describe our research method, present our results, and conclude with a discussion of our findings and the implications. 

II. RESEARCH SETTING, THEORY, AND HYPOTHESES

Research Setting and Tournament Incentives

To establish the context for our hypotheses, we first describe our research site and the two experiment conditions used in our research design. The Company is a privately owned wholesaler and distributor of area rugs to independent retailers in Canada and the United States, and has been in business for almost 30 years. The Company’s operations manager approached the researchers for assistance in designing a one-off tournament incentive scheme to increase the sales of the Company’s rugs at select retailers. A tournament represents a competition whereby individuals are rewarded based on their relative performance over a specified period of time (Hannan et al. 2008). Because rewards are based on relative performance, tournaments represent an effective means of filtering out the effects of common uncertainty that can impact the performance of all competitors (Lazear and Rosen 1981). There is an extensive literature on tournament incentive schemes, which generally shows they have positive effects on effort and performance (Casas-Arce and Martinez-Jerez 2009; Matsumura and Shin 2006; Orrison, Schotter and Weiglt 2004). Prior to implementing the incentive scheme in this study, the Company had never used sales incentives for its retailers.

A total of 54 independent Canadian retailers participated in the study.  Retailers were organized into eight competition groups comprising of between six or seven retailers per group with similar prior year sales (see Method section for details). Retailers competed only against other retailers in their assigned competition group so they are likely to perceive a reasonable chance of winning. If retailers were competing with other retailers who had significantly higher rug sales, the competition would not motivate these retailers because they would have a very low likelihood of winning. The eight competition groups were then randomly assigned to either the cash rewards condition or the tangible rewards condition in which winners earned gift cards.

There were two consecutive three-month tournament periods. Retailers within each competition group were ranked based on the cumulative invoice dollar value of the Company’s rug products sold in their store during each three-month tournament period. The top three performers in a competition group received a reward (cash or gift cards) equal to 15% of cumulative sales for the three-month period. Providing rewards for the top three performers in each group represents 43% (50%) of the seven (six) retailers in the group. We chose a moderate proportion of winners to be consistent with prior research suggesting that too high a proportion of winners (e.g. 75%) can reduce effort because competitors perceive a high probability of winning (Orrison et al. 2004; Harbring and Irlenbusch 2008) while too low a proportion of winners (e.g. 25%) can also have negative effects on effort since many competitors perceive a low probability of winning (Harbring and Irlenbusch 2008). Rewards were determined as a percentage of sales to create a graduated reward scheme where the value of the rewards increases with performance, which research suggests is more likely to motivate effort among top performers than a fixed reward scheme where all winners receive the same reward regardless of performance (Becker and Huselid 1992; Lynch 2005). These two design features, as well as having retailers with similar sales compete against each other, make it likely that our tournaments induced sufficient effort from retailers, regardless of reward type, to avert any floor effects that would limit the ability of reward type to differentially motivate effort.

Theory and Hypotheses

Tangible Rewards

            As a means of motivating and rewarding good performance, many organizations use tangible rewards as an element of their compensation package (Long and Shields 2010). Distinct from employee recognition, tangible rewards have a non-trivial monetary value and come in various forms including merchandise, travel, or gift cards (Presslee et al. 2013).[3] Recent surveys of compensation practices by U.S. firms indicate about 75% of respondents use tangible rewards and of those users, almost 90% give gift cards to offer employees choice in how they can spend their reward (Incentive Federation Inc. 2013; Incentive Research Foundation 2012). Despite the relatively widespread adoption of tangible rewards, surprisingly little research has examined their effects in settings where rewards are performance-based.

Two studies have used proprietary data from field sites to examine the performance effects of reward type but neither employs a tournament incentive scheme. First, Alonzo (1996) uses sales data from sales associates at 900 tire retailers who were eligible to receive either cash or tangible rewards (merchandise and travel items) based on the number of tires sold. Alonzo (1996) finds that retailers eligible for tangible rewards outperformed those eligible for cash rewards by 46%. However, retailers in the tangible rewards condition were not made aware of the specific monetary value of the available merchandise leaving open the possibility that they valued them higher than the cash awards available in the other condition.[4] Second, Presslee et al. (2013) find that call center employees eligible for cash rewards outperformed those eligible for tangible rewards (merchandise and travel items).[5] Rewards in Presslee et al. (2013) were contingent upon goal attainment with employees selecting their own performance goal from a menu of three choices with the reward value increasing in goal difficulty but no rewards could be earned for performance in excess of the chosen goal. Compared to those eligible for cash rewards, employees eligible for tangible rewards were more committed to their chosen goal but selected less challenging goals. This reduced level of goal difficulty in turn led to lower performance for employees eligible for tangible versus cash rewards. Thus in Presslee et al. (2013) the link between reward type and effort is mediated by the difficulty of the chosen performance goal.

            Only two studies have directly compared the use of tangible versus cash rewards where receipt of the rewards is contingent upon relative performance in a tournament setting. First, Shaffer and Arkes (2009) find no performance effects of reward type (cash rewards, tangible rewards, choice of cash versus tangible rewards) in a lab experiment where student participants solved anagrams. However, the generalizability of Shaffer and Arkes (2009) is limited by their use of a winner-take-all tournament scheme with only one individual (3% of competitors) eligible for a reward (value of $250) in each condition, which may have negatively impacted motivation for the majority of their participants (Harbring and Irlenbusch 2008). Also, there was a limited choice of tangible rewards (three possible items: Apple I-pod, Sirius satellite radio receiver and 6 month subscription, Palm Tungsten E handheld organizer) raising the possibility that some participants did not find them attractive.

Second, results from Jeffrey’s (2009) lab experiment show that tangible rewards (candy bars or massages) result in a larger performance improvement than cash rewards of an equal value for adult participants performing a word creation game. Participants were paid based on relative performance with graduated rewards increasing in value from $2 (20th percentile) to $100 (95th percentile) with performance. However, Jeffrey’s (2009) inferences regarding the performance effects of reward type are no longer significant after controlling for the significant effect of participants’ age (p. 150). Further, Jeffrey (2009) did not employ repeated tournaments nor did he provide participants with any feedback while the task was on-going to permit social comparisons.

The non-significant (weak) findings of these two lab studies may be attributable, in part, to the small proportion of winners, lack of feedback during the competition, and potentially unattractive rewards, which are tournament features likely to have negative effects on motivation. Moreover, neither study: 1) examines the effect of reward type on tournament performance in an organizational setting; 2) uses gift cards as tangible rewards, despite their popular use in practice (Incentive Research Foundation 2013) [6], or; 3) examines the effects of reward type over more than one period. Accordingly, there is considerable scope for further research using tournament incentive schemes designed to provide greater potential to observe the performance effects of tangible versus cash rewards.

Reward Type and Overall Performance in Tournament One

Economic theory suggests that the greater fungibility of cash represents its key motivational advantage over all other restricted financial rewards of equivalent market value (Offenberg 2007; Waldfogel 1993). Simply put, employees can purchase almost anything with cash, a characteristic not possessed by tangible rewards. Indeed, prior research indicates that when people jointly evaluate cash and tangible rewards, they focus on the fungibility of the rewards and thus indicate a preference for cash (Hein and Alonzo 1998; Jeffrey 2009; Shaffer and Arkes 2009).[7]  Thus from a purely economic perspective, cash should be more attractive than tangible rewards resulting in higher motivation and effort.

Conversely, mental accounting theory and related evidence would predict that tangible rewards will be more attractive and thus induce more effort than cash rewards. Mental accounting refers to the coding, categorization and evaluation of outcomes (realized and potential) when making choices (Thaler 1999). Of particular relevance to our setting is the categorization process, which relates to the way in which both potential and realized financial transactions (e.g., revenues and expenses) are ascribed to particular mental accounts (Thaler 1985). Mental accounting theory offers that individuals categorize cash rewards to a different mental account than they do tangible rewards. Potential cash rewards, because of their similarity to other forms of cash earnings (e.g., salary), are likely to be categorized to a mental account that includes cash salary (Jeffrey 2009; Thaler 1999). Potential tangible rewards with hedonic properties are more likely to be categorized to a mental account distinct from cash earnings (Helion and Gilovich 2014; Thaler 1999). Consistent with individuals using different mental accounts for different types of rewards, Presslee et al. (2013) report that employees at their research site considered tangible rewards as more distinct (separate) from other sources of income compared to cash rewards of an equivalent value.[8]

The different mental accounting used to record potential cash versus tangible rewards is likely to influence effort. Research on mental budgeting finds that people discipline their spending by using different mental accounts to set up budgets and track expenses for different categories of expenditures (Cheema and Soman 2006; Heath and Soll 1996). Cash rewards that are categorized in a cash earnings mental account are more likely to be budgeted for necessities and utilitarian items (Thaler and Shefrin 1981). Conversely, tangible rewards are likely categorized into a less utilitarian mental account and budgeted for luxuries and hedonic items. For example, Helion and Gilovich (2014) find that individuals in their experiment were more likely to purchase hedonic items when paid in gift cards as opposed to cash of an equal amount. All the gift card tangible rewards in our research setting are, by design, used to purchase items with hedonic attributes. Research shows that items with hedonic attributes (e.g., dinner at a nice restaurant) lead to a stronger positive affective response than more utilitarian items (e.g., groceries) of equal value (Shaffer and Arkes 2009). The stronger affective response associated with how tangible rewards can be used is likely to make them more attractive and memorable than cash rewards and as a result, induce more effort to attain them. Consistent with the motivational benefits of affect-rich rewards, Presslee et al. (2013) provide evidence consistent with tangible rewards being more attractive than cash rewards of an equal market value. Controlling for ability, employees at their research site were more committed to attaining self-selected performance goals when rewards were tangible versus cash.

Economic theory and mental accounting theory result in opposite predictions regarding the performance effects of cash versus tangible rewards. Because we have no clear basis for predicting which, if either, theory will dominate behavior, and given the equivocal findings from prior research reviewed above, we state the following null hypothesis:

H1 (null):        In the first tournament overall performance will not differ between retailers eligible for cash versus tangible rewards.

 

Reward Type and Effort Reduction

            A problem observed in multi-period tournament settings is that weaker performing individuals often reduce their effort or adopt riskier strategies that negatively impact performance in the later stages of a tournament, or in subsequent tournaments after the initial tournament has ended. This observed effort reduction has been attributed to a low expectancy (probability) of winning (Berger et al. 2013; Casas-Arces and Martinez-Jerez 2009; Hannan et al. 2008). Economic-based theory assumes that expectancy (probability) assessments that a particular event will occur (e.g., win a gamble, win a tournament, etc.) are independent of the type of outcome that will ensue (i.e., monetary payoff versus non-monetary payoff) (Rottenstreich and Hsee 2001). Based on this assumption of probability-outcome independence, any effort reduction by individuals trailing within a tournament or in response to losing a tournament should not differ between those eligible for cash versus tangible rewards.

However, psychology theory and related evidence clearly suggest that probability assessments are not independent of outcomes. Specifically, individuals are more likely to overweight a low probability outcome (e.g., winning a tournament) when the outcome is affect-rich (i.e., more desirable) versus when the outcome is affect-neutral (e.g., cash) (Loewenstein, Weber, Hsee, and Welch 2001). According to Rottenstreich and Hsee (2001) “the emotional response generated by affect-rich outcomes tends to overwhelm the nuanced effect of probabilities” (p. 817). In other words, the desirability of affect-rich outcomes, can affect individuals’ assessments of the likelihood of those outcomes. For example, Rottenstreich and Hsee (2001) find that undergraduate students value a 1% chance at winning a $500 coupon redeemable for a trip to Europe (affect-rich outcome) significantly higher than they value a $500 coupon towards university tuition (affect-neutral outcome) and McGraw, Shafir and Todorov (2010) document similar findings.[9] Thus, psychology research suggests that the type of probabilistic outcome (i.e., affect-rich versus affect-neutral) can affect the weighting of the probabilities associated with those outcomes such that low probabilities are over-weighted for affect-rich outcomes (Rottenstreich and Hsee 2001).

            As described earlier, mental accounting theory predicts that tangible rewards will be categorized in a mental account separate from cash rewards, resulting in budgeted expenditures on more affect-rich (hedonic) outcomes than cash rewards. Thus, based on Rottenstreich and Hsee (2001) weaker performing individuals eligible for tangible rewards are more likely to overweight the low probability of winning a subsequent tournament, than those eligible for cash rewards. As a result of the overweighting, weaker performing individuals eligible for tangible versus cash rewards will maintain a higher expectancy of winning the subsequent tournament, resulting in smaller decreases in effort. We identify weaker performers as those retailers who lose the first tournament. Losing is an unambiguous outcome and thus we think it is reasonable to assume that losers of the first tournament, will on average, have a low expectancy of success in the next tournament (Berger et al. 2013). This reasoning leads to our second hypothesis:

H2:      Weaker performers (losers) of an initial tournament will decrease their performance less in the next tournament when they are eligible for tangible versus cash rewards.

 

Two issues regarding H2 bear emphasis. First, because the development of a reasonably well-calibrated expectancy of failure (or success) is likely to take time, we make no prediction analogous to H2 within the first tournament given the short duration of the competition and the provision of limited feedback (i.e., rank only). Second, there is evidence that top performers or winners may become complacent and reduce their effort in later periods of a competition or in a subsequent tournament because they perceive a high expectancy of winning even with a lower effort level (Berger et al. 2013; Casas-Arces and Martinez-Jerez 2009). However, the tournament used at our research setting has two features likely to mitigate complacency effects: (1) a graduated payout scheme is used to reward winners (15% of sales) creating a financial disincentive for complacency; and (2) feedback is only given on rank rather than actual sales making it difficult for top performers to assess how much scope exists for complacency. Therefore, we do not believe there is sufficient scope for reward type to differentially influence any complacency that may arise in our setting.

Reward Type and Overall Performance in Tournament Two

Strong competing theories result in H1 being a null hypothesis while H2 predicts losers of the first tournament eligible for tangible rewards will show smaller performance decreases in the second tournament than their counterparts eligible for cash rewards. Whether there is an overall performance advantage for cash or tangible rewards in the second tournament therefore depends on both the overall outcome in the first tournament as well as the magnitude of the effects predicted in H2. Because we have no theoretical basis for predicting either of these, we pose the following research question regarding overall performance in the second tournament:

RQ1:    Will overall performance differ in the second tournament between retailers eligible for cash versus tangible rewards?

 

III. METHOD

Participating Retailers and Procedures

The Company initially identified 66 Canadian retailers to participate in the study, of which 54 were finally included in the study. The Company included only Canadian retailers to avoid differences between the Canadian and U.S. economies that might influence sales. The Company excluded retailers with no associated sales representatives because communication with these retailers was more difficult.  The Company also excluded large retailers with multiple locations because these retailers tended to have their own sales incentives and would be less interested or motivated by the Company’s incentive program. The 66 retailers were contacted by the Company by email two weeks before the study began to provide general information about the sales competition; retailers who were not interested in participating were asked to notify the Company by email. Of the 66 retailers, 59 retailers agreed to participate in the sales competition. However, in the year prior to the study (2012), five of these retailers had sales that were, on average, more than three times greater than the average sales of the other 54 retailers. To avoid the potentially large impact these five retailers may have had on the results of the tournaments, they were not assigned to either of the reward type conditions.

The remaining 54 retailers were assigned to competition groups of retailers with similar prior year sales in the following manner. The Company first ranked all participating retailers based on their prior year’s (2012) total rug sales. Retailers were then organized into eight competition groups comprising of between six or seven retailers per group based on their sales ranking. Retailers ranked first to seventh formed the first group, retailers ranked eighth to fourteenth formed the second group, and so on. The eight competition groups were then randomly assigned to either the cash rewards condition or the gift card rewards condition (see “Independent Variable” sub-section for details). Each reward type condition had three competition groups with seven retailers each and one competition group with six retailers (i.e., 27 retailers in each condition).

After retailers were assigned to their respective conditions and one week before the study began, the Company sent a second email to each retailer with more detailed information about the sales competition and their assigned condition. This second email also included links to two videos. The first video provided general information on the Company and the area rugs distributed by the Company. The second video provided information on how sales personnel could use a small rack of rug samples as part of their sales pitch to customers.[10] Each retailer had a unique link to the two videos. A software program captured the number of times a retailer accessed the video on how to use the small rack of rug samples.

During each three-month tournament period, the Company provided each retailer with their cumulative rug sales performance and ranking vis-à-vis the other retailers in their competition group at the end of each month and also at the end of the three-month tournament. A sample of the emails sent to retailers with feedback on their rug sales and ranking is provided in Appendix 1. At all times, the identities of all retailers in each tournament group were kept anonymous and each retailer’s sales performance was known only to that retailer. The end-of-month feedback was sent within 18 days of the following month and the rewards were sent to the winning retailers within 90 days of the end of the tournament.

Sample Characteristics

As shown in Table 1, average annual rug sales in 2012 were $2,890 (n = 27, standard deviation = $1,907) in the cash rewards condition and $2,761 (n = 27, standard deviation = $1,334) in the gift card rewards condition and are not significantly different between the two conditions (p = 0.77).[11]  Unit sales per month averaged 17.6 (standard deviation = 12.7) in the cash condition and 16.6 (standard deviation = 8.4) in the gift card condition and the difference is not significant (p = 0.74). Of the participating retailers in the two reward type conditions, 70% (38 out of 54) were in English-speaking provinces while the remaining 30% (16 out of 54) were in a French-speaking province (Quebec). The number of French-speaking versus English-speaking retailers does not differ significantly across conditions (Pearson Chi-Square = 1.42, p = 0.23). All communications with French-speaking retailers were translated by a professional translator and then verified by a second independent translator.  Finally, based on responses to our post-experiment survey, the cash and gift card conditions were similar in terms of the number of full-time (respectively 4.3 versus 4.6) and part-time sales staff (respectively 0.3 versus 0.8) and the percentage of female sales staff (respectively 56% versus 60%) (all p values > 0.33).

Table 1: Comparison of Reward Conditions

 

 

Means (Standard Deviation)

Full-Year 2012 

Total Sales 

Unit Sales 

Price/Unit 

Cash (n=27)

$2,890 (1,907)

17.6 (12.7)

$165

Tangible (n=27)

$2,761 (1,334)

16.6 (8.4)

$166

 

 

 

 

 

Percent

English Speaking 

 

 

Cash (n=27)

77%

 

 

Tangible (n=27)

63%

 

 

 

 

 

 

 

Means

 

Demographics 

Full-Time

Sales Staff1 

Part-Time

Sales Staff1 

Percentage Female1

Cash (n=7)

4.3

0.3

56%

Tangible (n=9)

4.6

0.8

60%

Independent Variable

As discussed earlier, retailers were first organized into groups of competitors with similar rug sales. Each competition group was then randomly assigned to either the cash rewards condition or the tangible rewards condition in which winning retailers received gift cards. The use of gift cards to operationalize tangible rewards biases against H2 since monetizing rewards makes them more like cash and thus less likely to invoke different mental accounting processes (Jeffrey and Shaffer 2007). However, gift cards were easy to administer by the Company, are common in practice, and eliminate any uncertainty regarding the actual value of the tangible reward. Of those firms that use tangible rewards, almost 90% of them use gift cards (Incentive Research Foundation 2012).

Retailers in both conditions participated in two consecutive sales tournament periods. Each tournament period comprised of three months (March to May 2013, and June to August 2013). Rug sales by retailers are seasonal in nature, with higher sales in the fall and winter seasons. The Company chose to implement the sales competition during spring and summer where sales tended to be lower to motivate higher sales in those seasons. The top three retailers in each competition group received a prize equal to 15% of the total invoice dollar value of rugs sold during each 3-month tournament period, in the form of cash or gift cards.[12] We worked with the Company to select gift cards that would have hedonic properties and be attractive to sales staff at the retailers. The Company asked winning retailers to indicate their sales staff’s choice of gift cards from twelve locations that included popular bookstores, cinemas, food and beverage establishments, and retail shops.[13] Average payouts across both reward type conditions were $204 and $201 respectively for the first and second tournaments. Cash and gift cards for each tournament period were distributed about 3 months after the end of each tournament period. Communications sent to the retailers by the Company repeatedly informed them that the cash and gift card rewards should be equally distributed among the sales staff responsible for the rug sales.[14]

Dependent, Control and Other Measures

We used the total rug sales dollars for the three months in each tournament period as well as the ranked total rug sales dollars as the dependent variables. The Company indicated that the retailers and retail market for home furnishings in French speaking province of Quebec were distinct from other provinces; such that retailers in Quebec may respond differently to the sales competition compared to retailers in other provinces. Retailers in Quebec tended to be smaller and owner-operated, and the retail market is more interior-designer focused. Therefore, we controlled for whether the retailer is in Quebec or not. We also controlled for the number of times the retailer accessed the link to the video on how to use the small rack of rug samples since the Company intended this video to be helpful in improving sales.

Retailers were invited by the company to complete a brief survey at the end of the second tournament. The link to the survey was provided in emails to retailers after the end of the second tournament. The survey contained questions on the size of the retail store, the nature of compensation and incentives for sales personnel in the store, the attractiveness of the Company’s tournament incentives, etc.

IV. RESULTS

Hypothesis 1

            Given strong competing theories, our first hypothesis predicts no difference in total sales between retailers competing for cash versus tangible rewards in the first tournament. We use total sales for the three-month period covering the first tournament as the dependent variable, with reward condition (RewardType: Tangible = 1, Cash = 0), retailer language (Language: French = 1, English = 0), the cumulative number of times the retailer viewed the video on how to use the small rack of rug samples (VideoViews) during the first tournament period, and prior year sales (March to May 2012) as independent variables. Because of the potential for extreme observations to influence our inferences in both tournament periods given our small sample size and relatively high variation in monthly sales, we use OLS regressions, robust regressions, and rank regressions to test our hypotheses.[15] For the rank regressions we ranked all retailers across the two reward type conditions according to their total sales for the tournament period with a lower numeric rank indicating higher sales. 

Descriptive results for the first tournament (March to May 2013) are shown in Table 2, Panel B; median sales for the cash condition are higher than those for the tangible condition ($437 versus $368). Table 3 (Panels A – C) presents the results of the regression models for the first tournament. RewardType is not significant in any of the models with all p-values > 0.29.[16] Overall, consistent with competing theory-based behavioral mechanisms likely at work under different reward types, results from the first tournament fail to reject the null.[17]  

Table 2: Descriptive Statistics for Tournament Results

Panel A: Total Sales Both Tournaments (March to August 2013)

 

 

2012

Mean

(Standard

Deviation)

 

Bottom

Quartile

 

 

Median

 

Top

Quartile

 

 

Range

Cash (n = 27)

$1,819 (1,674)

$615

$1,541

$2,104

$0 - $6,856

Tangible (n = 27)

$1,298 (1,050)

$447

$   946

$1,882

$0 - $3,456

 

 

 

 

 

 

2013

 

 

 

 

 

Cash (n = 27)

$1,668 (2,240)

$447

$   627

$2,049

$0 - $8,637

Tangible (n = 27)

$1,338 (1,116)

$308

$1,177

$2,051

$0 - $4,246

 

 

Panel B: Tournament One Sales (March to May 2013)

 

 

2012

Mean

(Standard

Deviation)

 

Bottom

Quartile

 

 

Median

 

Top

Quartile

 

 

Range

Cash (n = 27)

$908 (1,348)

$129

$318

$1,215

$0 - $6,121

Tangible (n = 27)

$476    (490)

$    0

$288

$   813

$0 - $1,556

 

 

 

 

 

 

2013

 

 

 

 

 

Cash (n = 27)

$935 (1,208)

$228

$437

$815

$0 - $3,831

Tangible (n = 27)

$619    (652)

$159

$368

$904

$0 - $2,457

 

2013: Tournament One Winners

 

 

 

 

 

Cash (n = 12)

$1,809 (1,383)

$631

$1,123

$3,194

$437 - $3,831

Tangible (n = 13)

$945 (769)

$318

$670

$1,525

$129 - $2,457

 

2013 Tournament One Losers

 

 

 

 

 

Cash (n = 15)

$236 (170)

$0

$288

$348

$0 - $487

Tangible (n = 14)

$316 (311)

$0

$231

$506

$0 - $947

 

 

Panel C: Tournament One Payouts1

 

Mean (Standard Deviation)

 

Cash

Tangible

Total

1st Place

$398.10 ($206.60)

$209.93 ($142.46)

$304.01 ($192.63)

2nd Place

$252.83 ($216.90)

$160.99 ($103.80)

$198.32 ($164.89)

3rd Place

$163.31 ($175.70)

$72.18 ($74.55)

$112.68 ($129.08)

Overall

$271.33 ($207.53)

$141.89 ($115.37)

$204.03 ($175.37)

 

Payout

*Total Payout [n=25] = $5,101

*Cash Payout [n=12] = $3,256

*Tangible Payout [n=13] = $1,845

Table 2: continued

Panel D: Tournament Two Sales (June to August 2013)

 

Mean

(Standard

Deviation)

 

Bottom

Quartile

 

 

Median

 

Top

Quartile

 

 

Range

2012

 

 

 

 

 

Cash (n = 27)

$912    (831)

$189

$725

$1,460

$0 - $3,070

Tangible (n = 27)

$821 (1,012)

$154

$343

$1,241

$0 - $3,456

 

 

 

 

 

 

2013

 

 

 

 

 

Cash (n = 27)

$733 (1,172)

$    0

$329

$974

$0 - $4,806

Tangible (n = 27)

$719    (638)

$129

$795

$966

$0 - $2,466

 

2013 Tournament One Winners

 

 

 

 

 

Cash (n =12)

$1,475 (1,458)

$522

$1,063

$1,505

$329 - $4,806

Tangible (n =12)

$1,207 (571)

$839

$1,024

$1,497

$468 - $2,466

 

2013 Tournament One Losers

 

 

 

 

 

Cash (n = 15)

$139 (174)

$0

$119

$258

$0 - $517

Tangible (n = 15)

$328 (363)

$0

$248

$795

$0 - $966

 

 

Panel E: Tournament Two Payouts2

 

Mean (Standard Deviation)

 

Cash

Tangible

Total

1st Place

$302.63 ($286.47)

$256.09 ($104.30)

$279.35 ($201.13)

2nd Place

$247.61 ($248.71)

$161.06 ($48.57)

$204.33 ($172.22)

3rd Place

$113.55 ($79.15)

$126.11 ($40.52)

$119.83 ($58.60)

Overall

$221.26 ($218.72)

$181.09 ($85.72)

$201.18 ($163.75)

 

Payout

*Total Payout [n=24] = $4,828

*Cash Payout [n=12] = $2,655

*Tangible Payout [n=12] = $2,173

 

 

1 There is no significant difference in payout in Tournament One across conditions at either rank level (i.e., 1st place cash vs. 1st place tangible) (p > 0.32) or overall (p = 0.12).

2 There is no significant difference in payout in Tournament Two across conditions at either rank level (i.e., 1st place cash vs. 1st place tangible) (p > 0.26) or overall (p = 0.16)).

 

Table 3: Tournament One Analysis (March to May 2013)1

Panel A: OLS Regression (n = 54)

 

Model: SalesMarchtoMay20132i = b0 + b1RewardType3i + b2Language4i +b3VideoViewsMarchtoMay20135i + b4MarchtoMaySales2012i6 + ei

 

 

Coefficient

t-statistic

p-value

Constant

690.04

3.31

0.013

    Reward Type

-313.26

-1.12

0.299 

   Language

-52.68

-0.22

0.834 

   VideoViewsMarchtoMay2013

184.15

1.60

0.155

   MarchtoMaySales2012

0.16

2.44

0.045

Adjusted R2

12.8% 

 

 

 

Panel B: Robust Regression (n= 54)  

Model: SalesMarchtoMay2013i = b0 + b1RewardTypei + b2Languagei + b3VideoViewsMarchtoMay2013i + b4MarchtoMaySales2012i + ei

 

 

Coefficient

t-statistic

p-value

Constant

284.24

3.09

0.003

    Reward Type

19.76

0.19

0.425 

   Language

-178.18

-1.57

0.123

   VideoViewsMarchtoMay2013

215.63

5.94

< 0.001

   MarchtoMaySales2012

0.06

0.80

0.430

 

Panel C: Rank Regression (n = 54)7 

 

Model: SalesRankMarchtoMay2013i = b0 + b1RewardTypei + b2Languagei + b3VideoViewsMarchtoMay2013i + b4MarchtoMaySales2012i + ei

 

 

Coefficient 

t-statistic 

p-value 

Constant  

29.10

13.34 

< 0.001

    Reward Type

2.05 

0.41 

0.696 

   Language

3.89 

0.97 

0.362

   VideoViewsMarchtoMay2013

-3.45 

-2.10 

0.074

   MarchtoMaySales2012

-0.01 

-3.68 

0.008 

Adjusted R2

16.2% 

 

 

 

 

1 All p-values are two-tailed. OLS regressions use robust standard errors clustered on tournament.

Total sales March to May 2013.

3  Reward Type: 0 = cash; 1 = Tangible.

4  Language: 1 = French; 0 = English

5  Number of times the video link on how to use the small rack of rug samples was accessed by the retailer March to May 2013.

6 Total sales March to May 2012.

7 Sales rank based on relative sales performance across all retailers with a lower rank indicating higher sales.

Hypothesis 2

H2 predicts that weaker performing retailers (losers) in the first tournament eligible for tangible rewards will show smaller performance decreases from the first tournament to the second tournament compared to their counterparts eligible for cash rewards. Consistent with prior research that has examined effort changes in tournament settings (e.g., Berger et al. 2013; Hannan et al. 2008) we use the change in performance between the two tournaments (SalesChange) as the dependent variable and with one exception, the same set of independent variables as used to evaluate H1. Instead of using prior year sales as a control variable, we use sales from the first tournament since they should impound current factors affecting sales in the second tournament (e.g., local competitive environment, economic conditions). We employ the same three regression analysis techniques as described above for H1.  

            Descriptive results for SalesChange for tournament one losers are shown in Table 4, Panel A.  Consistent with H2, the median SalesChange for tournament one losers competing for cash rewards is $0 compared to a median positive change of $202 for tournament one losers competing for tangible rewards. The regression analyses results shown in Table 4, Panels B – D, show that RewardType has a significant effect on the change in performance between tournaments for losers of the first tournament in the expected direction (all one-tailed p-values < 0.05).[18] For losers of the first tournament, those competing for tangible rewards show a significantly larger improvement in performance from the first tournament to the second tournament, relative to cash rewards retailers. Thus, consistent with H2, losers of the first tournament were considerably less likely to reduce effort, and in fact increased effort, when they were competing for tangible rather than cash rewards.

Table 4:  Analysis of Sales Change from Tournament One to Tournament Two1

Panel A: Descriptive Statistics

 

Sales Change (Tournament Two – Tournament One)

 

Mean

(Standard

Deviation)

 

Bottom

Quartile

 

 

Median

 

Top

Quartile

 

 

Range

Tournament One Losers2

 

 

 

 

 

   Cash (n = 15)

 <$42>   (300)

<$348>

$   0

$159

<$487> -    $517

   Tangible (n = 14)

    $311    (474)

$   0

$202

$576

<$409> - $1,376

 

 

 

 

 

 

Tournament One Winners2

 

 

 

 

 

   Cash (n = 12)

<$402> (1,154)

<$1,285>

<$349>

$512

<$2,937> - $   975

   Tangible (n = 13)

<$127>    (742)

<$  392>

<$129>

$339

<$1,613> - $1,084

 

 

Panel B: OLS Regression for Tournament One Losers (n=29) 

Model: SalesChange3i = b0 + b1RewardTypei + b2Languagei + b3VideoViewsMarchtoAugust2013i + b4MarchtoMaySales2013i + ei

 

 

Coefficient

t-statistic

p-value

Constant

0.46

0.01

0.997

    Reward Type

394.79

2.63

0.017* 

   Language

-36.85

-0.29

0.783

   VideoViewsMarchtoAugust2013

114.97

0.82

0.437

   MarchtoMaySales2013

-35.44

-1.18

0.275

Adjusted R2

25.5% 

 

 

 

 

Panel C: Robust Regression for Tournament One Losers (n=29)

Model: SalesChangei = b0 + b1RewardTypei + b2Languagei + b3VideoViewsMarchtoAugust2013i + b4MarchtoMaySales2013i + ei

 

 

Coefficient

t-statistic

p-value

Constant

-49.29

-0.27

0.791

    Reward Type

334.29

1.99

0.029* 

   Language

42.47

0.19

0.852

   VideoViewsMarchtoAugust2013

146.13

0.93

0.360

   MarchtoMaySales2013

-0.25

-0.64

0.526

 


 

Table 4: continued

 

Panel D: Rank Regression for Tournament One Losers (n=29) 

Model: RankSalesChange4i = b0 + b1RewardTypei + b2Languagei + b3VideoViewsMarchtoAugust2013i + b4MarchtoMaySales2013i + ei

 

 

Coefficient

t-statistic

p-value

Constant

27.86

8.26

< 0.001

    Reward Type

-10.72

-2.57

0.018* 

   Language

-1.95

-0.63

0.549 

   VideoViewsMarchtoAugust2013

-2.78

-0.64

0.542

   MarchtoMaySales2013

0.01

1.27

0.244

Adjusted R2

26.7% 

 

 

 

 

1See Table 3 for variable definitions. All p-values are two-tailed except for Reward Type, which is one-tailed when the coefficient is consistent with our H2 directional prediction. One-tailed p-values are indicated with *. OLS regressions use robust standard errors clustered on tournament.

2Losers (winners) did not (did) finish in the top three in their competition group in the first tournament.

3SalesChange: Tournament Two sales – Tournament One sales.

4RankSalesChange: rank of Tournament Two sales – Tournament One sales. Lower values indicate more positive (less negative) changes. 

For completeness we also examine whether the theory underlying H2 influences effort choices within tournament one. Results of additional analysis (not tabulated) show that for retailers trailing the leaders (i.e., ranked 4th or worse) after month one or two of the first tournament, reward type does not significantly affect the change in performance in the subsequent month in any of the regression models used to evaluate our findings (all p-values > 0.24). We also examine our assumption that the tournaments used in our setting mitigate possible complacency effects whereby winners of the first tournament reduce effort in the second tournament. Descriptive results for SalesChange for tournament one winners are shown in Table 4, Panel A.  We do find some evidence consistent with overall complacency effects (i.e., reduced effort) (Berger et al. 2013) as the median SalesChange for tournament one winners is a negative change of $259 (cash = -$349; tangible = -$129) and this decrease in performance of tournament one winners in tournament two is marginally significant (p = 0.09). However, in keeping with our assumption that reward type will not influence any complacency arising after the first tournament, results from regression analyses (not tabulated) show that RewardType does not have a significant effect on SalesChange for tournament one winners in any of the regression models used to evaluate our findings (all p-values > 0.90).[19]

Research Question

Finally, we examine our research question regarding the effects of reward type in the second tournament. Descriptive results in Table 2, Panel D show that median sales for the second tournament (June to August 2013) are higher for retailers competing for tangible rewards versus cash rewards (respectively $795 and $329). The overall average reward payouts in tournament two increased in the tangible condition compared to tournament one ($181.09 versus $141.89) while the payouts decreased in the cash condition ($221.26 versus $271.33).

Table 5 (Panels A – C) summarizes the results of our regression analyses for the second tournament. [20] Given our results in support of H2, we include an additional binary variable indicating whether the retailer won or lost the first tournament (WinLoseT1, 0 = lose, 1 = win) and we interact it with RewardType. As shown in Table 5, RewardType is significant in all models (all p-values < 0.05) with retailers competing for tangible rewards outperforming their counterparts competing for cash rewards. However, the RewardType x WinLoseT1 interaction is also marginally significant (p = 0.09) in Panel B (robust regression) requiring that we interpret the main effect of Reward Type in the context of this interaction. Results for all regression models (not tabulated) show that in the second tournament, losers of the first tournament competing for tangible rewards significantly outperformed losers of the first tournament competing for cash rewards (all p-values < 0.05). Conversely, for tournament one winners, performance in tournament two does not differ between reward type conditions (all two-tailed values > 0.94). Thus the overall performance advantage related to tangible rewards in the second tournament is entirely attributable to the positive effects of those rewards on sustaining the motivation of tournament one losers.

Table 5: Tournament Two Analysis (June to August 2013)1


Panel A: OLS Regression (n = 54) 

 

Model: SalesJunetoAugust2013i = b0 + b1RewardTypei + b2WinLoseT12i + b3RewardType x WinLoseT1 + b4Languagei + b5VideoViewsMarchtoAugust2013i + b6MarchtoMaySales2013i + ei

 

 

Coefficient

t-statistic

p-value3

Constant

56.81

0.31

0.761

   Reward Type

405.06

2.70

0.031 

   WinLoseT1

196.17

0.54

0.604

   RewardType x WinLoseT1

-394.71 

-1.42 

0.200

   Language

-115.26 

-0.49 

0.637

   VideoViewsMarchtoAugust2013

-13.12 

-0.24 

0.820 

   MarchtoMaySales2013

0.668 

2.05

0.079

Adjusted R2

46.8% 

 

 

 

Panel B: Robust Regression (n = 54) 

 

Model: SalesJunetoAugust2013i = b0 + b1RewardTypei + b2WinLoseT1i + b3RewardType x WinLoseT1+ b4Languagei + b5VideoViewsMarchtoAugust2013i + b6 MarchtoMaySales2013i + ei

 

 

Coefficient

t-statistic

p-value

Constant

175.18

1.28

0.206 

   Reward Type

478.77

2.50 

0.016 

   WinLoseT1

404.80

1.59

0.119

   RewardType x WinLoseT1

-522.81 

-1.75 

0.087 

   Language

-280.29 

-1.76 

0.085 

   VideoViewsMarchtoAugust2013

35.02

0.77 

0.445 

   MarchtoMaySales2013

0.17 

1.79

0.079

 

Panel C: Rank Regression (n = 54) 

Model: SalesRankJunetoAugust2013i = b0 + b1RewardTypei + b2WinLoseT1i b3RewardType x WinLoseT1 + b4Languagei + b5VideoViewsMarchtoAugust2013i + b6 MarchtoMaySalesRank2013i + ei

 

 

Coefficient 

t-statistic 

p-value 

Constant  

19.22 

3.57 

0.009 

   Reward Type

-10.26 

-2.61 

0.035 

   WinLoseT1

-7.35 

-0.83 

0.432 

   RewardType x WinLoseT1

11.48

1.35 

0.220 

   Language

4.43

1.32 

0.228 

   VideoViewsMarchtoAugust2013

-0.63 

-0.68 

0.519 

   MarchtoMaySalesRank2013

0.45 

3.38 

0.012 

Adjusted R2 

36.2% 

 

 

 

1See Table 3 for variable definitions. All p-values are two-tailed. OLS regressions (Panels A and C) robust standard errors clustered on tournament.

WinLoseT1: 0 = lose Tournament One; 1 = win Tournament One.

            We also examine monthly performance within the second tournament to evaluate whether in keeping with H2, retailers show smaller performance decreases in the month after receiving feedback that they are trailing the leaders (i.e., ranked 4th or worse) when competing for tangible rather than cash rewards. Examining retailers who lost the first tournament, performance in June (not tabulated) improves significantly over May for retailers competing for tangible versus cash rewards using all three regression analysis approaches (all p-values < 0.05). In both July and August the results are directionally consistent with H2 in that weaker performing retailers (i.e., ranked 4th or worse) eligible for tangible rewards show larger performance increases compared to retailers eligible for cash rewards. However, RewardType is only significant (not tabulated) for the change in August sales (August – July) using robust regression (coefficient 55.86, p = 0.03).  Collectively, these results provide some additional support for the theory underlying H2.

Additional Analysis

            As described earlier, we conducted a survey to gather information related to retailers’ perceptions of the sales competitions and the related rewards. At the Company’s request we designed the survey to be answered in five minutes or less, which limited the number of questions we could include. We asked a representative of the sales team at each retail location to indicate the extent to which sales staff: (1) found competing with other stores fun; (2) were motivated to be one of the top three stores in the competition; (3) found the incentives attractive; and (4) would have rather received cash rewards (tangible rewards condition only). Each question used an 11-point scale with endpoints labeled “not at all” (1) and “extremely” (11). In total we received 16 usable responses for a response rate of about 30%.[21] This is a reasonable response rate given that we did not have direct contact with representatives of the sales team at any point during the study and because retailer representatives had no incentive to complete the survey. Results for these measures are summarized in Table 6.

Table 6: Survey Responses1

 

 

Means (Standard Deviations)

Questions

Cash (n = 7)

Tangible (n = 9)

Fun competing with other stores 

4.86 (2.97)

7.22 (2.39)

Motivated to be one of the top three stores 

4.71 (3.04)

7.33 (2.54)

Incentives attractive 

6.29 (2.92)

7.66 (2.69)

Rather have received cash rewards 

NA

8.11 (1.17)

            Inter-items correlations among the first three questions are all highly significant (all p-values < 0.001). Results from an exploratory factor analysis (not tabulated) indicate the three items load highly (all loadings ≥ 0.81) on one construct, the eigenvalue is 2.50, the explained variance is 83%, and the Cronbach’s Alpha is 0.96 (Stevens 1996). Accordingly we treat the three items as a single construct, which we label Reward Attractiveness. We compare the results between the two reward type conditions using both the simple average of the three items and the factor scores for Reward Attractiveness. Results of t-tests (not tabulated) show that both the simple average and the factor scores are higher in the tangible rewards condition than the cash rewards condition (both one-tailed p-values < 0.065). This is consistent with differences in the mental accounting for the two types of rewards rendering gift card rewards more attractive than cash rewards, even though the monetary value of the rewards was relatively low. Moreover, these results are consistent with prior research showing that when separately evaluating cash versus tangible rewards, retailers anticipate greater enjoyment from tangible rewards (Shaffer and Arkes 2009).

            Responses for the fourth item in Table 7, Rather Receive Cash Rewards, are consistent with prior research indicating that when given the choice between cash and tangible rewards, retailers pursuing gift cards (n=9) consistently indicate they would prefer cash (Jeffrey 2009; Shaffer and Arkes 2009). Analysis (not tabulated) shows that the mean of 8.11 is significantly above (p < 0.01) the scale mid-point (6) indicating a stronger preference for cash. Thus our results in support of H2 do not appear attributable to retailers in the gift card rewards condition having a preference for tangible rewards.

            The Company also provided us with sales data for the 54 retailers included in our study for the three months immediately following the conclusion of the second tournament (September – November 2013). We use this additional data to further evaluate the similarity of the retailers in our two reward type conditions. Results (not tabulated) based on the approach used to test our hypotheses show no significant differences in sales between conditions during this post-experiment period (all p-values > 0.50). We interpret these results as further evidence that the observed effects of reward type on performance seem unlikely to be attributable to unobserved differences between the reward conditions.  

V. DISCUSSION

            Recent surveys of compensation practices indicate a growing use of tangible rewards to motivate and reward performance in organizations (Incentive Federation Inc. 2007, 2013). However, empirical evidence regarding the effects of tangible rewards relative to cash on performance in tournaments is limited and the results are equivocal (Jeffrey 2009; Shaffer and Arkes 2009). In an effort to further our understanding as to when tangible rewards may induce better performance than cash rewards, we conduct a field experiment at 54 retail outlets eligible for either cash or tangible (gift card) rewards based on sales of rugs supplied by the Company that provided the sales incentives. Supporting our null hypothesis, we observe no performance effects of reward type in the first tournament. However, in keeping with our second hypothesis, losers competing for gift card rewards significantly increased their performance more from the first tournament to the second tournament compared to the cash rewards retailers. Finally, the strength of the effects found in support of our second hypothesis lead to significantly higher performance in the second tournament by retailers who lost in the first tournament eligible for tangible versus cash rewards.

            Our findings make three primary contributions to the academic and practitioner literatures related to reward-type and tournament incentives. First, to the best of our knowledge, we are the first to demonstrate in a field setting that over time, tangible rewards can result in better performance than cash rewards in a tournament incentive scheme. In particular, prior research examining the performance effects of reward type in tournament settings have used single-shot tournaments and find either no differences between reward conditions or weak results (Jeffrey 2009; Shaffer and Arkes 2009). While similar to these studies we find no effects for reward type in the first tournament, our repeated tournament design reveals that significant performance advantages for tangible rewards emerged over time as weaker performers were more likely to sustain and even increase effort in response to losing an initial tournament. We observe these positive performance effects of tangible rewards even though the value of the rewards used in our study was relatively low, and despite the well documented finding that tournament incentive schemes more generally have positive effects on effort and performance (Hannan et al. 2008; Harbring and Irlensbusch 2008). As such we believe our findings provide convincing evidence that tangible rewards can induce better performance in repeated tournament settings or by extension tournaments that are longer in duration, incremental to the positive social comparison effects inherent to the design of tournament schemes.

Second, our evidence that tangible rewards may be an effective means of sustaining effort in repeated tournament settings for individuals who performed poorly in initial competitions is important from a theoretical perspective as it sheds new light on how the problem of effort reduction by losers, often observed in tournament settings, can be moderated by the type of rewards individuals are competing for. Building on Rottenstreich and Hsee (2001) our results suggest affect-rich tangible rewards may attenuate the effort reductions by weaker performers in tournaments documented in previous field research by encouraging them to over-estimate their likelihood of success (e.g., Berger et al. 2013; Casa-Arces and Martinez-Jerez 2009).

Finally, our results are of practical importance to designers of compensation schemes given the relative ease with which tangible rewards can be used in many organizations. Further, while many tournament features can be implemented to sustain the effort of better performers such as graduated payouts or allowing a larger percentage of the competitors to win, our results suggest an effective means of motivating weaker performers to continue to exert effort, even when their expectancy of succeeding is low. However, of equal importance to practice, our findings also suggest that the effectiveness of tangible rewards may be contingent upon the nature of the incentive scheme with which they are coupled. For example Presslee et al. (2013) find that tangible rewards lead to lower performance than cash rewards when used with a bonus for goal attainment scheme because they induced individuals to self select easier performance goals. Our results, in conjunction with Presslee et al. (2013) indicate tangible rewards may be better suited for use in incentive schemes where there is a direct link between rewards and effort.

Our study’s limitations provide several opportunities for further research examining the effects of reward type on motivation and performance. First, we apply an individual level theory to explain retailer (group) level performance differences. Although we do not believe that there is a theoretical basis for expecting that reward type would differentially affect behaviors at the group level than the individual level, future research that captures individual level tournament performance would be helpful in testing the legitimacy of our application of theory. Second, although our evidence in support of H2 is consistent with weaker performers working harder for tangible rewards relative to cash, it is an empirical question as to whether or not this result would hold for rewards of a larger value. It could be that when the monetary value of rewards is relatively low, the memorability and attractiveness of hedonic tangible rewards are more salient since the cash reward offers limited opportunities even for utilitarian spending. However, when rewards are sufficiently large in value, the utilitarian purposes for which cash is often spent (e.g., education, health care, housing) may become more motivating because there is more cash available to make these important expenditures or investments.[22] Future research in field settings with larger reward values would be helpful in addressing this possibility. Third, we examine performance over just two, relatively short duration tournaments, leaving open the possibility that longer term, the effectiveness of tangible rewards in mitigating giving up effects by tournament losers could be limited. Further research in settings with multiple repeated tournaments or longer duration tournaments is needed to examine the longevity of the effects we observe across the two tournaments. Finally, despite our ability to randomly assign tournament groups to the reward type conditions, we cannot rule out the possibility that subsequent to the beginning of the study, unobserved changes occurred at the retailer level that influenced our results either within or between the two tournaments (e.g., the hiring of more capable sales staff, new competitors entered the market, etc.). Our analysis of post-study sales suggests this is unlikely but research using lab experiments with tightly controlled task environments would still be useful as a means of building on our key findings. Despite the foregoing limitations, overall we believe our study makes an important contribution to the burgeoning literature on the performance effects of tangible rewards.

REFERENCES

 

Alonzo, V. 1996. The trouble with money. Incentive, 170 (2): 26–31.

 

Becker, B.E., and M.A. Huselid. 1992. The incentive effects of tournament compensation systems. Administrative Science Quarterly, 37. 336-350.

 

Berger, L., K. Klassen, T. Libby and A. Webb. 2013. Complacency and giving up across repeated tournaments: Evidence from the field. Journal of Management Accounting Research, 25(1): 143-167.

 

Casas-Arce, P. and F. Martinez-Jerez. 2009. Relative performance compensation, contests, and dynamic incentives. Management Science, 55(8):1306-1320.

 

Cheema, A., and D. Soman. 2006. Malleable mental accounting: The effect of flexibility on the justification of attractive spending and consumption decisions. Journal of Consumer Psychology 16 (1): 33-44

 

Hannan, R.L., R. Krishnan and A.H. Newman. 2008. The effects of disseminating relative performance feedback in tournament and individual performance compensation plans. The Accounting Review 83(4): 893-913.

 

Harbring, C. and B. Irlenbusch. 2008. How many winners are good to have? On tournaments with sabotage. Journal of Economic Behavior & Organizations, 65: 682-702.

 

Heath, C. and J. Soll. 1996. Mental budgeting and consumer decisions. Journal of Consumer Research, 23: 40-52.

 

Hein, K. and V. Alonzo. 1998. This is what we want! Incentive, 172(10): 40-47.

 

Helion, C. and T. Gilovich. 2014. Gift cards and mental accounting: Green-lighting hedonic spending. Journal of Behavioral Decision-Making,  27(4): 386-393.

 

Incentive Federation Inc. 2007. Federation Study 2007: A Study of the Incentive Merchandise and Travel Marketplace. Westfield, NJ: Author.

 

Incentive Federation Inc. 2013. Incentive Market Study: October 2013. Westfield, NJ: Author

 

Incentive Research Foundation and Incentive Gift Card Council. 2012. B2B Gift Card Market Study.

 

Jeffrey, S. 2009. Justifiability and the motivational power of tangible noncash incentives. Human Performance, 22(2): 143-155.

 

Jeffrey, S., and V. Shaffer. 2007. The motivational properties of tangible incentives. Compensation & Benefits Review, 39(3): 44–50.  

 

 

Lazear, E.P., and S. Rosen. 1981. Rank-order tournaments as optimum labor contracts. The Journal of Political Economy 89(5): 841-864.

 

Long, R., and J. Shields. 2010. From pay to praise? Non-cash employee recognition in Canadian and Australian firms. International Journal of Human Resource Management, 21(8): 1145-1172.

 

Loewenstein, G., E. Weber, C. Hsee, and N. Welch. 2001. Risk as feelings. Psychological Bulletin 127(2): 267-286.

 

Lynch, J. 2005. The effort effects of prizes in the second half of tournaments. Journal of Economic Behavior in Organizations, 57: 115-129.

 

Matsumura, E.M., and J.Y. Shin. 2006. An empirical analysis of an incentive plan with relative performance measures: Evidence from a postal service. The Accounting Review 81(3): 533-566.

 

McGraw, A.P, E. Shafir, and A. Todorov. 2010. Valuing money and things: Why a $20 item can be worth more and less than $20. Management Science, 56(5): 816-830.

 

Offenberg, J. 2007. Markets: Gift cards. Journal of Economic Perspectives. 21(2): 227-238.

 

Orrison, A., A. Schotter and K. Weiglt. 2004. Multiperson tournaments: An experimental examination. Management Science, 50(2): 268-279.

 

Peterson, M. 2008. Estimating standard errors in finance panel data sets: Comparing approaches. The Review of Financial Studies, 22(1): 435-480.

 

Peterson, S. and F. Luthans. 2006. The impact of financial and nonfinancial incentives on business unit outcomes over time. Journal of Applied Psychology, 91(1): 156-65.

 

Presslee, A., T. Vance and A. Webb. 2013. The Effects of Reward Type on Employee Goal Setting, Goal Commitment and Performance. The Accounting Review, 88(5): 1805-1832.

 

Rottenstreich, Y. and C. Hsee. 2001. Money, kisses, and electric shocks: On the affective psychology of risk. Psychological Science 12(3): 185-190.

 

Shaffer, V., and H. Arkes. 2009. Preference reversals in evaluations of cash versus non-cash incentives. Journal of Economic Psychology, 30: 859-872.

 

Stevens, J. 1996. Applied multivariate statistics for the social sciences (3rd Ed.). Hillsdale, N.J.: Erlbaum Associates.

 

 

Tafkov, I. 2013. Private and public relative performance information under different compensation contracts. The Accounting Review, 88(1): 327-350.

 

Thaler, R. 1985. Mental accounting and consumer choice. Marketing Science, 4: 199-214.

 

Thaler, R. H. 1999. Mental accounting matters. Journal of Behavioral Decision Making, 12: 183-206.

 

Thaler, R. and H. Shefrin. 1981. An economic theory of self-control. The Journal of Political Economy, 89(2): 193-205.

 

Waldfogel, J. 1993. The deadweight loss of Christmas.  American Economic Review, 83: 1328-1336.

 

1 All questions used an 11-point response scale with endpoints labeled “not at all” (1) and “extremely” (11). Only those in the tangible rewards condition responded to the question about “Rather have cash rewards”.

Appendix 1: Sample email providing feedback on sales and sales ranking performance

Interim monthly feedback (provided twice during each 3-month competition period)

Dear [retailer contact name], 

[Retailer name] is competing anonymously with 6 other stores with similar levels of [company name] product sales. [Cash condition has the following additional sentences here: The top 3 of the 7 stores in your competition group will receive a cash prize equal to 15% of the total invoice dollar value of [company name] products sold during each 3-month competition period. The cash prize should be equally distributed among the sales staff who are responsible for [company name] product sales.] [Giftcard condition has the following additional sentences here: The top 3 of the 7 stores in your competition group will receive a prize in the form of gift cards worth 15% of the total invoice dollar value of [company name] products sold during each 3-month competition period. The prize should be equally distributed among the sales staff who are responsible for [company name] product sales. Sales staff would be able to choose their gift cards from the following locations: Starbucks, Tim Hortons, Chapters, Cineplex, Marble Slab, Future Shop, Best Buy, EB Games, The Keg, Bath and Body Works, The Body Shop, or Bon Appetit (usable at Kelsey’s, Montana’s, Swiss Chalet, Harvey’s, and Milestones).] As at the end of Month 2 (April) of the first competition period (March – May 2013), your store is ranked 3rd out of the 7 stores in your competition group. Your store’s sales performance and ranking to date are as follows.

Interim Sales Competition Ranking: Month 1/Month 2

 

Rank: End of Month 1 (March)

Rank: End of Month 2 (April)

Retailer A

1st

1st

Retailer B

2nd

2nd

Your store (insert name)

3rd [Cumulative total invoice dollars of [company name] products sold to date = $xxx]

3rd [Cumulative total invoice dollars of [company name] products sold to date = $xxx]

Retailer C

4th

4th

Retailer D

5th

5th

Retailer E

6th

6th

Retailer F

7th

7th

Please forward this information about the sales competition to all sales staff who are responsible for [company name] product sales. As a reminder, our product knowledge video can be viewed at [insert link].

Thank you,

[name of company representative]

Appendix 1 continued

End of competition feedback (provided at the end each 3-month competition period)

 

Dear [retailer contact name], 

[Retailer name] is competing anonymously with 6 other stores with similar levels of [company name] product sales. The first competition period (March – May 2013) has ended. Congratulations, for the first competition period, your store finished 3rd out of the 7 stores in your competition group! Your store’s sales performance and final ranking for the first competition period are as follows. [Cash condition has the following additional sentences here: As a result, your store has won a cash prize equal to 15% of the total invoice dollar value of [company name] products sold in your store during the competition period. Your store’s cash prize, that should be divided equally among the sales staff who are responsible for [company name] product sales, is $_______. Please email us at [email address] the with the names/emails/contact telephone number of sales staff who are eligible for the cash prize and we will provide each sales staff with information on how they can collect the cash prize.] [Giftcard condition has the following additional sentences here: As a result, your store has won a prize in the form of gift cards worth 15% of the total invoice dollar value of [company name] products sold in your store during the competition period. Your store’s prize, that should be divided equally among the sales staff who are responsible for [company name] product sales, is $_______. Please email us at [email address] with the names/emails/contact telephone number of sales staff who are eligible for the prize and we will provide each sales staff with information on how they can collect the prize. Sales staff would be able to choose their gift cards from the following locations: Starbucks, Tim Hortons, Chapters, Cineplex, Marble Slab, Future Shop, Best Buy, EB Games, The Keg, Bath and Body Works, The Body Shop, or Bon Appetit (usable at Kelsey’s, Montana’s, Swiss Chalet, Harvey’s, and Milestones).]

Final Sales Competition Ranking

 

Rank: End of Month 1 (March)

Rank: End of Month 2 (April)

Final Rank: (March – May)

Retailer A

1st

1st

1st

Retailer B

2nd

2nd

2nd

Your store (insert name)

3rd [Cumulative total invoice dollars of [company name] products sold to date = $xxx]

3rd [Cumulative total invoice dollars of [company name] products sold to date = $xxx]

3rd [Cumulative total invoice dollars of [company name] products sold to date = $xxx]

Retailer C

4th

4th

4th

Retailer D

5th

5th

5th

Retailer E

6th

6th

6th

Retailer F

7th

7th

7th

Please forward this information about the sales competition to all sales staff who are responsible for [company name] product sales. The second competition period runs from June to August 2013. As a reminder, our product knowledge video can be viewed at [insert link].

Thank you,

[name of company representative]



[1] In comparison, a similar but earlier survey conducted in 2007 indicated 34% of respondents used tangible rewards with estimated annual spending of $46 billion (Incentive Federation Inc. 2007).

[2] The extent to which a reward will be considered hedonic will likely vary by individual but research indicates that this variation is not necessarily a function of the value of the reward. For example Helion and Gilovich (2014) classify candy, music CDs, novels, and multi-colored pens as all being hedonic in nature, despite their relatively low monetary value, because their anticipated consumption or use is pleasurable.

 

[3] Recognition programs involve an acknowledgement, often public, of good performance (e.g., employee of the month) in the form of thank-you notes, plaques, company newsletter articles, token gifts, etc., but typically do not involve rewards with significant monetary value (Peterson and Luthans 2005).

[4]Moreover, Alonzo (1996) does not report descriptive details that would allow a comparison of the similarity of retailers in the two conditions (e.g., past sales, geographic location, etc.), which makes it impossible to rule out the alternative explanation that the results are attributable to unobserved differences between retailers in the two reward type conditions.

[5] Employees could choose their non-cash rewards from an extensive catalogue of hedonic choices including home electronics, barbeques, coffee makers, bicycles, etc.

[6] Gift cards are cited as a means of providing rewards that are more likely to be attractive to a larger proportion of employees given the choice inherent to this type of reward (Jeffrey and Shaffer 2007).

[7] When people separately evaluate cash versus tangible rewards, they tend to focus on the affective characteristics of the tangible rewards and rate them as significantly more attractive (Shaffer and Arkes 2009).

[8] This result holds for each of the three reward levels ($100, $350, $1,000) in Presslee et al. (2013).

[9] Rottenstreich and Hsee (2001) also find that individuals have a fear of failing to receive affect-rich outcomes and hence underweight high probability affect-rich outcomes to reduce the potential disappointment of failing to receive the outcome. We think this underweighting is unlikely in our setting because it seems implausible that there would be a sufficiently strong ‘fear of failure’ for the relatively small dollar value tangible rewards available to our competitors.

[10] The Company provided retailers with a small rack of about 50 mini-rug samples for free to assist them in selling and marketing the rugs. Each rack is also equipped with a banner, catalogues and brochures that provide information about sizing of rugs, the materials used in various rugs, and cleaning options

[11] All p-values are two-tailed unless otherwise stated.

[12] Performance contingent cash and gift cards rewards are both subject to personal income tax in Canada. Both require the recipient to claim the reward as taxable earnings. Thus, the tax treatment does not differ across the two reward type conditions.  

[13] The gift card locations and the proportion selected by winners include: Starbucks (4%), Tim Horton’s (39%), Chapters (4%), Cineplex (7%), Marble Slab (0%), Future Shop (10%), Best Buy (0%), EB Games (4%), The Keg (14%), Bath and Body Works (14%), The Body Shop (0%), or Bon Appetit (usable at Kelsey’s, Montana’s, Swiss Chalet, Harvey’s, and Milestones restaurants) (4%).

[14] All but one winning retailer in the cash rewards condition and all but one winning retailer in the gift card rewards condition distributed rewards equally to multiple sales staff. Winning retailers in the cash (tangible) rewards condition distributed the rewards to an average of 3.0 (3.2) salespeople.

[15] Because results within a tournament are not independent, we also use robust standard errors clustered on tournament group (n = 8) (Peterson 2008).

[16] Parametric and non-parametric analysis (not tabulated) shows that March to May 2012 sales do not differ between the two reward type conditions (all p-values > 0.190).

[17] We also examine alternative models (not tabulated) in which the dependent variable is either the dollar change between sales during the first tournament period minus sales during March to May in the prior year or the percentage change in sales between the two periods. Similar to our main results, we fail to reject the H1 null hypothesis in any of these alternative models.

[18] Models in which the control variable is prior year change in sales (instead of first tournament sales) show similar results, where RewardType is significant in the OLS regression, the robust regression, and the rank regression for first tournament losers (all one-tailed p values < 0.10).

[19] We also examine monthly performance changes in both tournaments using only retailers who are leading the competition (i.e., ranked 1st, 2nd, or 3rd) within a tournament; we find no consistent significant effects of RewardType on the change in monthly performance in either tournament.

[20] We also examine models in which the dependent variable is the difference between sales during the second tournament minus sales during June to August in the prior year. RewardType is not significant in the OLS regression (not tabulated: B=470.96, one-tailed p=0.25) and the rank regression (not tabulated: B=-7.74, one-tailed p=0.18), but it is significant in the robust regression (not tabulated: B=481.71, one-tailed p=0.06). Models in which the dependent variable is the percentage change in sales in the second tournament period from June to August in the prior year show similar results, where RewardType is significant in the OLS regression, the robust regression, and the rank regression (all one-tailed p values < 0.05).

[21] Of the 16 responses, one was received from a retailer that lost both tournaments, six were from retailers that won one of the two tournaments, and 9 were from retailers who won both tournaments. Given this distribution of respondents it is possible that the results for these questions may not generalize to retailers who were unsuccessful in the tournaments. However, our analysis of H2 is consistent with losers of the first tournament eligible for gift card rewards finding them more attractive and working harder to attain them in the second tournament relative to retailers eligible for cash rewards.

[22] It may also be that as cash rewards become larger, individuals may be more likely to spend some of the earnings on non-utilitarian items thereby increasing the attractiveness of the total cash rewards.