Principals and Their Car Dealers: What Do Targets Tell about Their Relation?

Transcription

1 Principals and Their Car Dealers: What Do Targets Tell about Their Relation? Presented by Dr Eddy Cardinaels Professor University of Leuven #2016/17-03 The views and opinions expressed in this working paper are those of the author(s) and not necessarily those of the School of Accountancy, Singapore Management University.

2 Principals and Their Car Dealers: What Do Targets Tell about Their Relation? Jan Bouwens* University of Amsterdam, Amsterdam Business School Eddy Cardinaels KU Leuven, Department of Accountancy, Finance and Insurance, Tilburg University, Department of Accountancy, Jingwen Zhang Tilburg University, Department of Accountancy, March 2016 * Corresponding author. This study received generous support from the Thomas Henry Carroll-Ford Foundation at Harvard Business School. We thank our colleagues at the accounting department of Tilburg, seminar participants at Frankfurt School of Finance & Management, Michigan State University, Ludwig- Maximilian-University of Munich and IÉSEG School of Management, participants of the 2015 MAS conference and 2015 EAA conference, and Karen Sedatole, Will Demere, and Michael Matějka for their comments on previous versions of the paper.

3 Principals and Their Car Dealers: What Do Targets Tell about Their Relation? Abstract The paper tries to establish for a large European car dealership whether target ratcheting and target achievements are associated with the type of contract offered to the franchisee in the dealership. We find that the franchisor sets different targets conditional on whether or not the car dealer (franchisee) exclusively acquire the cars they sell from the franchisor. We also find that compared to non-exclusive dealers exclusive dealers work harder to achieve their targets and are less likely to mute their effort when their target achievement is relatively high. We argue that the evidence we present is consistent with the ideas put forward in relational-contracts theory. Keywords: Target setting; Target ratcheting; Ratchet effect; Target achievement; Relationalcontracts theory.

4 1. Introduction This paper examines how choices in target setting affect relations within a franchise organization, where the franchisor contracts with two types of franchisees. The first type of franchisee exclusively sells the franchisor s brand of cars (exclusive franchisee), while the second type of franchisee also sells brands of other manufacturers (non-exclusive franchisee). We test whether the franchisor treats the two types of franchisees differently in terms of target updating and whether franchisees respond differently to the targets conditional on their type. Our inquiry is motivated by relational-contracts theory. According to this theory, cooperation between parties is a function of the de facto contracts these parties are engaged in (Gibbons and Henderson, 2012a). These contracts arise from a chain of interactions between the parties through which these parties show their commitment to meet each other s expectations. These expectations are contingent on circumstances and may differ over time, and thus a formal contract can capture only some of the effort necessary to meet each party s expectations. To the extent one party meets the other s expectations time after time, the second party will increase its faith in the first s commitment. In return, the second party will then be more likely to show commitment to the first party. Over time, mutual trust arises, with both parties believing that implicit obligations will be honored at each engagement. By their very nature, these sorts of contracts preclude the parties from upholding their implicit agreements in court. As trust grows through repeated interactions, it is difficult for all parties involved to build and refine their relational contract. Since it takes time for signals to become mutually credible within an organizational setting, relations are difficult to copy by other parties once trust is instilled (Gibbons and Henderson, 2012a). As a consequence, two competing firms may look the same on surface and face the same conditions except that one firm has built a 1

5 sophisticated implicit understanding as to what constitutes the contract within the firm, while the other is waiting for these particulars to emerge. The latter faces potentially a bigger challenge in assuring that coordination occurs between all parties included in the relational contract. We propose that this idea from relational-contracts theory can help us explain why seemingly similar firms may perform in markedly different ways (Gibbons and Henderson, 2012a). We argue that as long as trust has not grown in the relational contract, parties will refrain from sharing the benefits that accrue from their cooperation. 1 An empirical challenge for studying relational contracts is the difficulty of finding firms that lend themselves to comparison and allow for the identification of relational contracting. Firms in the same industry may look superficially similar but often differ as far as relations and formal contracts between principals and agents comprising the firm are concerned (Gibbons and Henderson, 2012a). Using the data from a car dealer network, we study the target setting and updating process in which the principal (franchisor) offers the same formal contract to its agents (franchisees). Regardless of dealership type, all dealers face a common stair-step incentive scheme, where they receive larger discounts on the list price of cars, conditional on the dealer achieving the target, The observable difference between the agents in our franchise organization is that they either exclusively sell only the brand of cars provided by the franchisor (exclusive dealers) or they sell brands they acquire via the franchisor and sell in addition other brands offered by other suppliers than the franchisor (non-exclusive dealers). Because the franchisees are similar in many other aspects, the setting offers anopportunity to identify how these two types of agents relate to their principal and vice versa. We measure whether our treatment variable dealership type (exclusive/nonexclusive dealer) 1 Standard agency theory would predict that agents exposed to the same incentive scheme should respond the same. One distinguishing feature of relational contract theory entails that it can explain why agents subject to the same incentive system respond differently given that trust has yet to arise or already has arisen in the organization. 2

6 is associated with a notable difference in how the franchisor and franchisees treat each other in terms of target setting (franchisor) and target achievement (franchisee). We propose that, in terms of establishing a relational contract, the franchisor may treat non-exclusive dealers differently compared to exclusive dealers, because the franchisor is more confident about the loyalty of their exclusive dealers. Non-exclusive dealers can direct effort away from the franchisor s brand towards other brands; that is, they can direct it to where they expect the highest profit. That may be the brand of the franchisor or the brand of a competitor. A second stream of literature that motivates our study focuses on target setting. While the targetsetting literature has examined target ratcheting (e.g., Leone and Rock, 2002; Bouwens and Kroos, 2011), only recently have studies start to investigate the conditions that lead firms to ease targets (e.g., Indjejikian et. al., 2014b). Indjejikian et al. (2014b) argue that firms may decide to mute target ratcheting to signal to managers that they can trust the principal to share the proceeds from high performance. That is, firms may still ratchet targets but not to the extent that would give rise to describe as the perverse incentive effect where managers mute their effort or misreport their performance to have a lower future target (Aranda et al., 2014). Indjejikian et al. (2014a) demonstrate that targets of high performers are ratcheted less than those of low performers. This comports with the idea that firms try to offer high performers a fair share of the proceeds they realize for the company. These actions would also discourage high performing agents from shirking in attempts to lower their future targets. Bol and Lill (2015) also show that the principal does not fully revise targets for agents with whom she has an implicit agreement. We find that ratcheting levels differ between exclusive and nonexclusive dealers. Results indicate that the franchisor uses past performance realizations to a lesser extent when updating the sales target for the exclusive dealer. Hence, the exclusive agent has to fear less that the principal will 3

7 seize all of the benefits of the additional effort by setting very high targets (Baron and Besanko, 1984; Indjejikian et al., 2014b). This result suggests that the principal, who expects an ongoing relationship with agents, is more likely to commit to an informal contract in which the agents can trust her to allow them to reap benefits that accrue from performance improvements. We also show that non-exclusive dealers mute their effort more once they achieve their target and are less likely to meet or exceed an equally difficult target compared to exclusive dealers. The latter results show that franchisees may direct their effort to other (more profitable) brands, despite the fact that the franchisor communicates to franchisees the importance of achieving sales targets. We argue that the lower level of engagement from non-exclusive dealers signals to the franchisor that these dealers may not reciprocate an easy target by actually meeting the target. Exclusive dealers, in contrast, can more credibly communicate to the franchisor that they will do their best to achieve or exceed the target, because they have fewer opportunities to redirect effort. This, in turn, can motivate the franchisor to honor this signal, by abstaining from excessive target ratcheting and thus allowing the agent to receive a larger cut of the sum of the discounts made available by the franchisor. To our knowledge this is the first study that provides empirical evidence to show that companies untighten controls once both parties have come to a mutual (implicit) understanding of what is expected within their relationship. Previous analytical research has shown that imposing controls can be costly in terms of effort choice (e.g., Falk and Kosfeld, 2006). The next section reviews the literature and motivates our hypotheses. Section 3 describes the research setting and empirical measures. Section 4 examines our hypotheses and presents some additional analyses. Finally, the last section concludes with remarks on limitations of our study and suggestions for future research. 4

8 2. Literature review and hypotheses We study a firm where the agents are employed via a franchise contract. We examine two types of agents: agents who exclusively sell the product (cars) from the principal (exclusive agents) and agents who sell both the product of the principal and other brands provided by other firms (nonexclusive agents). In our setting, agents can increase their income by selling more cars. In addition, they are subject to a common incentive scheme in which they receive a volume discount on the cars they purchase from the principal. To earn these discounts, the agent needs to achieve sales targets set by the principal (i.e., the franchisor). Our inquiry focuses on how target setting and target achievement play out. Setting targets The target setting literature shows that principals ratchet targets to assure that agents continue to try to increase profitability. However, because agents know that the principal will use current performance to set the next-period target, agents are often reluctant to show performance improvements. Agents thus must balance how much they benefit from current performance improvements and how much target updating their current performance entails (e.g., trading off a high bonus now against the prospect of a lower bonus in the future or vice versa). To ensure a larger bonus later, the agent may reduce his effort in the current period. This perverse effect is empirically demonstrated by Leone and Rock (2002) and Bouwens and Kroos (2011). Indjejikian et al. (2014b) argue that, to mute the impact of perverse effects associated with target ratcheting, principals can commit to a target-updating scheme that allows the agents to keep some of the rents created through their (high) effort. Their evidence suggests that principals commit to lower levels 5

9 of target ratcheting for the high performing agents, in comparison to agents who have yet to achieve high performance, to assure that the high performers continue to work hard. Relational-contracts theory suggests that principals may not exclusively lower target ratcheting for their high performers. The theory predicts that principals are likely to commit to lower levels of target ratcheting when they expect their agents to strive to achieve their target (Gibbons and Henderson, 2012a). Based on this literature, we argue that principals are more inclined to impose difficult targets on non-exclusive agents than on exclusive agents. The reason is that the firm may have fewer incentives to invest into the relationship with non-exclusive agents. Baron and Besanko (1984) and Levin (2003) argue that agency costs in a principal-agent relation can decrease if the principal is willing to commit to a long-term stationary contract. In the extreme, such a contract, under risk neutral conditions, would entail that the principal commits to a fixed target regardless of the performance improvements an agent achieves. This raises a question: to whom is the firm more likely to offer such a contract, the non-exclusive agent or the exclusive one? If a non-exclusive agent receives such a contract, the firm runs the risk of investing in an agent who may resort to selling other brands or may even decide to end the relationship with the firm altogether. The reason is that the firm cannot work out in advance how much better her deal is compared to the reservation income (of complete or partly switching to the other brand) of the agent. Gibbons and Henderson (2012b) show analytically that, under such conditions, the principal will not trust the agent to cooperate in period 2. Therefore the non-exclusive agent will not be offered such a contract. The costs for the exclusive agent to discontinue his relationship with the principal are arguably higher (Fein and Anderson, 1997). The fact that an exclusive agent limits his portfolio to the brand 6

10 of the franchisor strongly signals that he is committed to meeting the principal s expectations (Fein and Anderson, 1997). Given these signals, the principal considers how much she wants to invest in the relationship (Fein and Anderson, 1997). In expectation she may expect that investments in in relations with exclusive agents yield higher returns than investments in relations with nonexclusive agents. From the perspective of the exclusive agent, their stronger commitment may prompt them to work harder, which, in turn, would allow the principal to set more difficult targets to reap the full benefits of the relationship. However, doing so would provide the exclusive agent with the incentive to become a non-exclusive agent because this would open the opportunity to redirect effort from one brand to the other when it was more profitable for the agent to do so. In addition, such a decision may counter the idea of the principal s willingness to honor her implicit promise to share benefits that accrue from consistently meeting high expectations. As long as the principal can commit to keep compensating the agent for additional sales, the mutual understanding will lead the agent to keep working hard to achieve high levels of performance (Baron and Besanko, 1984). Relational contracts vest themselves in the extent that principals refrain from exploiting the other party (Morgan and Hunt, 1994). The above forces would suggest that the principal would commit to easier target updates for the exclusive agent, to ensure that the exclusive agent maintains his commitment to the brand. The principal can achieve this by allowing the agent to earn some rents when he is performing well (Indjejikian et al., 2014b). When dealing with the non-exclusive agents, principals have more questions to work through when the agent fails to meet or exceed the target: is it because the agent could not meet the target, because the agent is trying to avoid harder targets in the next period, or because the agent redirects effort to another brand? Based on this increased complexity, we predict that higher levels of mutual 7

11 understanding is harder to attain when the franchisor deals with non-exclusive agents. That is, for the principal it is easier (more difficult) to correctly attribute achievement to effort for exclusive (for nonexclusive) agents. As a consequence the principal will be more inclined to lock in an achievement in the next-period target for the nonexclusive agents than for exclusive agents. To summarize, we argue that it is more difficult for the principal to work out whether performance levels of non-exclusive agents are due to effort directed to pursue the franchisor s objective. As a consequence it is more likely that the franchisor mistakenly attributes good performance in the current period as a demonstration of low effort choice in the past. She is therefore more likely to ratchet the target using past (proven) performance for nonexclusive agents than for exclusive agents. Given our discussion, we propose to test the following hypothesis in the null form: Hypothesis 1: Type of dealer is not associated with the extent to which past performance is used for future target updates. The effort response to targets According to the target setting literature, agents will protect themselves against target ratcheting (Milgrom and Roberts, 1992, p ). Empirical evidence speaks to this idea by demonstrating that agents who perform well either manage earnings (e.g., Leone and Rock, 2002) or manage real earnings even downwards to reduce target ratcheting (Bouwens and Kroos, 2011). However, what will agents do if they have the opportunity to redirect their effort? Will they do so even at the cost of their relationship with the principal? This subsection aims at the identification of conditions that would prompt the agent to cease the redirection of effort. Agents must decide how much effort they are willing to exert to increase their income. When the target is met, further exceeding the target entails two income sources to the agent: (1) sales minus 8

12 the gross purchase price and (2) the discount on the purchase price. Hence, once the agent meets the target, he is incentivized to keep increasing the number of products sold. At first glance, there seem to be few reasons to assume that exclusive agents will behave differently than non-exclusive dealers in terms of target achievement. However, as the market for the franchisor s brand approaches its level of saturation, it may become attractive for the non-exclusive agent to switch to another brand because the effort required from the agent to continue selling the product may be larger than the effort required by another (nonsaturated) brand. Non-exclusive agents can relatively inexpensively resort to alternative income sources and are therefore more likely to give up on meeting or exceeding a difficult target than are exclusive agents. Exclusive agents have no alternative profit source; their income is bound to the principal s brand. If we assume that it is equally difficult for both types of agents to achieve the target, nonlinearity in the effort-sales function will likely feature an inflection point where it becomes more attractive for the nonexclusive agent to switch to another brand. Such an inflection point does not exist for the exclusive agent, who might as well continue to exert effort. In addition, there is the target ratcheting argument. Non-exclusive agents are in a better position than exclusive agents to taper their performance to reduce target ratcheting because the nonexclusive agents can switch to the other brand. The exclusive agent, however, may still fear that his next-period target will be too high. To keep the future target achievable, he can reduce his effort in the current period, too. Bouwens and Kroos (2011) demonstrate that very well performing store managers mute effort by putting a hold on how much they exceed their target over an accounting period. In doing so, these agents protect themselves against a principal setting unachievable targets for the next period. Indjejikian et al. (2014a), on the other hand, find that high-performing agents do not need to fear unachievable target updates. This suggests that 9

13 companies commit themselves by allowing the agents to keep a larger share of the rents gained with achieving higher levels of performance (Baron and Besanko, 1984; Indjejikian et al., 2014b). If the exclusive agent and the principal develop a mutual informal understanding, whereby the principal commits to ratchet targets to a lesser extent, she implicitly offers a higher compensation level to agents who are loyal to the brand. In that case the exclusive agent may be more willing to continue to work hard for the principal (Indjejikian et al., 2014b) and is less likely to mute effort in an attempt to protect themselves against target ratcheting. This leads to the following hypothesis (in the null form): Hypothesis 2: Dealer type is unrelated to target achievement. 3. Research setting To test our hypotheses, we use data from a franchise organization, which is responsible for a network of car dealerships in a Western European country. The franchisor is a national subsidiary of a car manufacturer with worldwide operations. The franchisor specifies wholesale and retail prices, sets incentive schemes for the franchisees and arranges marketing promotions for dealers in the network. The sales network consists of franchisees (i.e., car dealers) who acquire their cars from the franchisor. The dealers are separate private companies whose formal relations with the franchisor are contractual. The dealers have both exclusive and non-exclusive sales relationships with the franchisor. Exclusive dealers sell only the brand of cars provided by the franchisor. Non-exclusive 10

14 dealers also sell other brands, which can be supplied by other franchisors. 2 The decision about being an exclusive or a non-exclusive dealer can be affected by factors such as the number of dealers in a sales area, market conditions and strategies of the franchisor and the franchisees. It is a mutual choice made by both the franchisor and the franchisees. 3 All the dealers are represented in a council through which they can express their needs and advise the franchisor on business operations and strategies. Franchisees make their own pricing decisions within a range set by the franchisor. Hence dealers can quote different prices for the same car. Franchisees have full decision rights over their internal organization, staffing and how much they wish to invest. They can also make dealer-specific advertisement decisions. Consistent with the policy of a sufficiently dense network, the franchisees (i.e., dealerships) covered all the sales areas of the domestic market over the sample period During the period, sales and revenues slightly increased from 2004 to 2008, and achieved the highest levels in 2011, but decreased dramatically in Panel A of Table 2 reproduces the yearly composition of dealer network. The composition of these franchisees slightly changes over time, with the percentage of exclusive dealerships declining and the percentage of non-exclusive ones increasing. Panel A of Table 2 further shows that non-exclusive dealers compared to exclusive dealers more often enter into the franchise network or end their relations with the franchisor. 2 The manufacturer is a subsidiary of a holding company, which sells other car brands, too. Hence the holding company offers a family of brands. We focus on one brand offered by the franchisor. However, some dealers also sell other brands in the family. We define exclusive dealers as the ones who sell the focal brand offered by the franchisor and consider dealers who also sell other family brands offered by the franchisor as non-exclusive sales. The reason is that the franchisor does not set targets for the other cars that belong to the family. We also run the data where we consider the family dealers also as exclusive dealers. The results, however, are largely consistent. Each dealer also receives a unique identifier. Based on these identifiers we do not observe switches of dealerships. However, the franchisor may have assigned new identifiers to a dealer after the dealer switched. This may also indicate that the franchisor will treat them differently if a franchisee changes his type. 3 Franchisees do not need to pay franchise fees to the franchisor. They do have to invest in franchisor-specific show rooms, displays, services, etc. 11

15 We conduct our empirical analyses using dealer-level performance data of all the dealers in the network. This data comprises almost 100 percent of the sales the network realized in the national market because all the sales of the brand are realized through this network. The incentive system The franchisor offers three main incentives to their dealers: the difference between the advised consumer price and the purchase price, a bonus for meeting qualitative standards based on qualitative measures and a bonus for the volume based on quantitative measures. The data we study is the volume bonus, which car dealers receive conditional on them achieving the targeted sales volume. In our sample, the volume bonus accounts for 25% 30% of the dealer s total margin. For dealers to receive a quarterly bonus, they need to meet their quarterly targets. The bonus takes the form of a discount on the invoice price of the cars that franchisees acquire from the franchisor. As shown in Panel B of Table 2, this target setting system does feature a steps schedule, which is common in the industry. The stair-steps schedule purportedly incentivizes dealers to work harder to achieve a higher discount. The discount system does feature a minimum threshold, while there is no cap on how much bonus a car dealer can earn. The specific bonus scheme of the franchisor features different steps where each step leads, conditional on the target achievement of a dealer, to higher bonus percentages (i.e., higher discounts). The firm defines bonus areas based on the percentage of sales realized relative to the target. For the at target area, sales are at least equal to the target (target achievement 100%). In general, the dealer gets no additional cash payment or discount for sales levels much below the sales target. The bonus increases from the Threshold level till the Maximum level. The highest bonuses are granted to dealers who convincingly exceed their sales target. 12

16 Target-setting and revision process The franchisor sets the annual sales target for each dealer in December for the next budget year, which coincides with the calendar year. Sales targets are denominated in number of cars, not in euros. The franchisor decides on the annual target for each dealer according to an estimation method developed by the national dealer organization. This method includes a set of factors deemed relevant by the franchisor: historic sales performance of the dealers, the firm s development strategies, expected local sales conditions and nationwide market conditions. To gauge these sales conditions, the franchisor uses its discretion to account for specific market conditions facing a dealer. Once the annual target for each dealer is established, the firm sets quarterly sales targets. Bonuses are awarded quarterly. A simulated example of target setting is given in Appendix A. While these targets are set for the year, they can still be adjusted if market conditions change dramatically. That is, the franchisor may alter the quarterly targets in the course of the current quarter. We do not have detailed information about these intra-period changes. What we have is the final quarterly and annual targets for each dealer over our sample period. Sample composition We collect daily sales information of each individual dealer over the period from 2004 to We also gather quarterly sales targets and annual sales targets for each dealer. However, 21 dealeryear observations and 44 dealer-quarter observations are excluded due to the missing information about quarterly targets, resulting in 538 dealer-year observations and 2,090 dealer-quarter observations with both sales and target data. Some dealer-year observations are incomplete as the dealers are not actively selling the franchisor s brand during the full year. This happens when a 13

17 new dealer enters into a contract with the franchisor or when a dealer decides to discontinue his contract with the franchisor during the year. For our analyses, however, we only use the data of dealers for which we have complete information over the four quarters of each year. This reduces the sample size to 497 dealer-year observations and 1,988 dealer-quarter observations. Since we need past performance as controls in our model, only dealers with more than one year data are included in the analysis. This results in a final sample of 410 dealer-year observations and 1,640 dealer-quarter observations for our empirical tests. Variable measurement Measure of target update: The franchisor considers target setting an important control instrument to motivate its dealers to achieve desired sales levels. Sales targets are based on the dealers past sales performance and on local and national market conditions. How these conditions translate into expected sales levels is subject to the franchisor s assessment. We test in hypothesis 1 whether these assessments result in different target revisions for non-exclusive dealers compared to the exclusive dealers. The resulting target update is captured by the variable TARGET_UPDATE i,t, which represents the difference between last year s target and the current target and comprises both objective and subjective factors. A larger percentage of target update is perceived as a more difficult target. Dealer performance: Effort choice of a dealer is reflected in his sales performance. Sales achievements proxy for the dealer s willingness to invest in the relationship with the franchisor. We capture dealership performance in three variables: DEV_TARGET i,t, ACHIEVED i,t and NUMQT_ACHIEVED i,t. DEV_TARGET i,t represents the sales performance of dealer i compared to his sales target in a particular period t. ACHIEVED i,t is a dummy variable, which indicates 14

18 whether the dealer meets his sales target in period t. NUMQT_ACHIEVED i,t represents the frequency with which a dealer meets his quarterly target in a full year (0 4). Exclusive or non-exclusive dealer: To capture the differences between two types of dealers in how targets are set and how much effort dealers exert, we create a dummy variable EXCLUSIVE i to indicate dealer who exclusively sells the specific brand of the franchisor (1) or a non-exclusive dealer who also deals in other brands (0). Good performers: Indjejikian et al. (2014a) show that a principal may display more commitment to well-performing agents by assigning them easier targets or by protecting them from ratcheting. We focus on three different variables to classify a dealer as a good performer: a) if a dealer s target deviation is among the top 25 percent of all the dealers (GPM_DEV i,t = 1; 0 otherwise), b) if his sales are among the top 25 percent (GPM_SALES i,t = 1 and 0 otherwise), and c) if he achieves all the four quarterly targets in the last period (GPM_ALLQT i,t = 1 and 0 otherwise). With the analyses, we examine whether the good performers are treated differently in terms of target setting by the franchisor, depending on their dealer type. Difficulty: To examine whether nonexclusive dealers differ from exclusive dealers in terms of effort levels, we must control for target difficulty. The dummy variables DIFFICULT_MEDIAN i,t and DIFFICULT_MEAN i,t capture whether the target is difficult (1) or easy (0) for a dealer, in the comparison with the median and mean of target updates of the dealers located in the same area. A target is difficult if the target update for a dealer is larger than his peers coping with similar market situations. To conduct this analysis, we need a sufficient number of dealers to make a meaningful comparison. We require therefore that a particular dealer has at least three dealers in its vicinity in 15

19 any particular year. This procedure reduces our final sample to 291 observations for regressions where target difficulty enters into the model. Control variables: As the study aims at identifying whether type of dealer affects the franchisor s decision and a dealer s effort, we include several controls in our model that may also affect the outcomes: target levels and target achievement. We use the natural logarithm of the target, i.e., LNTARGET i,t to control for the size of the dealership. 4 Sales growth of the brand in the domestic market can also affect each dealer s sales potential. We therefore measure GROWTH_BRAND t to represent the year-to-year changes in the market sales of the brand. In addition, we use GROWTH_ALLBRAND t to trace year-to-year sales changes of all the car brands in the domestic market. Past performance of each dealer is also included in our main tests. Some dealers represent the collective dealers in a dealer committee. They more often communicate directly with the franchisor than other dealers. To control for this effect, we create the dummy variable COMM_MEM i,t, indicating whether the dealer is a delegate to the dealers committee. In addition, some dealers have an individual agreement with the franchisor. This agreement may result in benefits or privileges that can affect dealer s effort. We therefore create the dummy variable IND_AGREE i,t, which equals one if an individual agreement applies and zero otherwise. Table 1 contains an overview of the variables used in this study. Descriptive statistics Panel A of Table 3 summarizes the statistics of the main variables we use for our empirical tests. As shown in the table, 40.7% (59.3%) of the dealers are exclusive (nonexclusive) dealers. The 4 We also control for size effects by using the logarithm of sales, and results do not change significantly. 16

20 actual sales at the dealer level are on average lower than the sales target. The mean (median) of sales-target deviation is about 2.24% ( 6.18%), i.e., dealers show an adverse deviation from the annual sales target. Less than half (40.5%) of the dealers achieve a sales level that exceeds the sales target. The mean (median) target update of the sales target is 7.898% (6.217%), and sales targets are adjusted downward in 166 out of 410 cases (not tabulated). Panel B of Table 3 offers a comparison between exclusive dealers and non-exclusive dealers on some of the key variables we use in our analyses. Target levels of exclusive dealers are lower in terms of volume than the target levels of non-exclusive dealers. On average, targets are adjusted upward to a smaller extent for exclusive dealers (3.6%) compared to the non-exclusive dealers (10.85%), indicating that non-exclusive dealers are treated differently. The table also shows that more than half of the exclusive dealers (53.3%) achieve their sales targets, while only 31.7% of the non-exclusive dealers do. 4. Empirical Results Test of hypothesis 1: Target updates and the type of dealership Hypothesis 1 states that a franchisor will use past performance to an equal extent for non-exclusive and exclusive dealers when considering target revisions to set the target for the next period. To test H1, we estimate the following regression model: TARGET_UPDATE i,t = β 0 + β 1 EXCLUSIVE i + β 2 DEV_TARGET i,t 1 + β 3 EXCLUSIVE i DEV_TARGET i,t 1 + control variables + ε i,t (1) 17

21 where TARGET_UPDATE i,t denotes the difference between the new target and the current target, which is decided by the franchisor for dealer i in year t based on a mixture of objective and subjective factors; EXCLUSIVE i is a dummy variable equal to one if the dealer signs the exclusive contract with the franchisor and zero otherwise; DEV_TARGET i,t 1 indicates the sales performance of dealer i in period t compared to the sales target. The variable of interest in this model is the interaction term between EXCLUSIVE i and DEV_TARGET i,t 1 indicated by the coefficient β 3. A significant loading for β 3 on target update suggests that the franchisor tends to use past sales data for target updating differently conditional on dealership type. A negative sign on β 3 indicates that exclusive dealers are given easier targets than non-exclusive dealers. Since the firm sets yearly targets, we use the annual data for this estimation. 5 Our estimates are based on our regression equation using OLS over the sample period of 2004 to The t-statistics are based on clustered standard errors to adjust for heteroskedasticity and autocorrelation. In Table 4, we reproduce the results of our tests by using model (1). 6 As shown in column 1 of Table 4, the coefficients β 1 of EXCLUSIVE i is negative and significant, meaning that overall target updating is easier for exclusive dealers. Past performance (DEV_TARGET i,t 1 ) is positively related with target updates (coefficient=0.482, p<0.01). This is consistent with the concept of target ratcheting, documented previously (e.g., Leone and Rock, 2002; Bouwens and Kroos, 2011). The negative interaction term between EXCLUSIVE i and DEV_TARGET i,t 1 (coefficient= 0.233, 5 Quarterly targets are decided after the yearly targets have been set. Since we do not have information about how the yearly targets are further divided into four quarters, we only use yearly data for the test of hypothesis 1. 6 We also run ours tests by winsorizing TARGET_UPDATE i,t at 1 percent and 99 percent (not tabulated) to control for the effects of large target updates, and we find the same results (not tabulated). The large values of TARGET_UPDATE i,t shown in Table 3 are probably caused by mergers between dealers or expansion of a particular dealer. These large values are usually applied to non-exclusive dealers. 18

22 p=0.067) provides weak evidence to suggest that franchisors use past sales information to a lesser extent to update next period targets for exclusive dealers. The effect is that non-exclusive dealers face more demanding targets than do exclusive dealers and hence that the latter stand a better chance to qualify for a (higher) discount on the purchase price of cars that the non-exclusive dealer. We propose that this difference in target setting reflects how much the franchisor is willing to invest in building a relation with its franchisees, conditional on them being exclusive or nonexclusive dealers. In the second column of Table 4, we follow the target-setting process as it was described to us by the firm. Representatives of the franchisor said that they also considered a dealer s performance in t 2 and t 3. 7 The results in column 2 again provide evidence to suggest that target ratcheting occurs to a lesser extent for exclusive dealers than for non-exclusive dealers (coefficient= 0.242, p=0.065), when using t 2 and t 3 next to t 1 sales data for updating targets. We primarily find evidence that performance in the last period affects the target setting for the next period. This suggests that, to the extent that sales achievements of t 2 and t 3 affect the target updates, this data is impounded only to the extent it is related to t 1 sales. Based on these results, we reject hypothesis 1: past performance is not used to the same extent for exclusive and non-exclusive dealers. Our results suggest that the exclusive dealers face easier targets because past sales realizations are impounded to a lesser extent in next year targets. Target update, dealership and target achievement 7 As indicated by the example in Appendix A, we are told that the franchisor considers the previous three years performance to set the new target. However, we find that only the last year s performance plays a significant role. Therefore, for the rest of our tests, we only include last year s performance. By doing so, we also keep a reasonable sample size for analyses. 19

23 A second test for hypothesis 1 is to examine how target achievement affects target updates. Previous literature finds that principals set targets differently depending on whether agents meet their target (e.g., Leone and Rock, 2002; Bouwens and Kroos, 2011). 8 In this section, we examine whether the franchisor sets different targets for dealers who achieve their targets conditional on dealer type, exclusive or non-exclusive. We rerun our basic model (1) but now create two subsamples: one is comprised of observations where dealers exceed their targets in the last year, while the other is comprised of observations where dealers fail to meet their target last year. The results reported in Table 5 resemble those reported in Table 4. The coefficient on the interaction term is negative (coefficient= 0.408, p<0.05). This result indicates that, compared to nonexclusive dealers, exclusive dealers get a better deal (less tough target updating) when they reach their targets. However, as shown in the second column of Table 5, if targets of the previous year are not met, the franchisor does not differentiate between the two types of dealers. These results suggest that exclusive dealers who show loyalty to the franchisor by achieving the target are given easier access to discounts on the purchase price of cars compared to non-exclusive dealers who express the same level of loyalty. This evidence again leads us to reject H1. More specifically, the results suggest that the franchisor puts less weight on past performance to set a target for the exclusive dealers, i.e., target ratcheting occurs to a greater extent for the non-exclusive dealers. Target update, dealership and good performer 8 We also test whether targets are updated asymmetrically, i.e., where positive target deviations lead to a higher targets while adverse deviations are ignored or only come through in the target to a very limited extent. However, we do not find evidence for asymmetric target updating in our sample. Table 5 also shows that the main effect of DEV_TARGET i,t 1 is significant for target updating, indicating a more symmetric updating scheme. Nevertheless, exclusive dealers relative to non-exclusive dealers are treated differently. A possible explanation for symmetric updating may lie in the fact that we study franchisee-franchisor relations, rather than employee-supervisor relations. 20

24 Indjejikian et al. (2014a) find that principals show their commitment to good performers by assigning them easier targets. They argue that principals do this because these agents have limited scope to increase their performance much further. It stands to reason that, should further improvements occur, management will grant to the franchisees a fair share of the proceeds incurred from that achievement. This action taken by the principal is also supposed to mitigate target ratcheting effects such that agents are less likely to mute their efforts. In our setting, we expect exclusive dealers to be more likely to be rewarded for good performance without having to fear a stark target update. A dealer qualifies as a good performer if (1) his sales deviation, compared to the target level, is among the top 25 percent of the dealers (GPM_DEV i,t ), (2) if his sales volume belongs to the top 25 percent of all the dealers (GPM_SALES i,t ), or (3) if he achieved all the quarterly targets in the last period (GPM_ALLQT i,t ). To test whether the data supports our expectations, we focus on the interaction term between these three indicators for good performance and the variable EXCLUSIVE i because our interest lies in different treatments among better-performing dealers depending on their dealership type. Results are reproduced in Table 6. The interaction terms in Table 6 all have negative and significant coefficients ([1] coefficient= , p<0.05; [2] coefficient= , p<0.01); [3] coefficient= ; p<0.01). In sum, our theory seems to be supported by the data in that different treatments surface for good performers conditional on the type of dealership, i.e., the franchisor shows his commitment to exclusive dealers by refraining from target ratcheting if they perform well, while such commitment is shown to a lesser extent to non-exclusive dealers. Relationship building 21

25 Gibbons and Henderson (2012a) argue that trust between parties can grow over time. We construct a test that refers to the history of the relation between the franchisor and the franchisee. Our variable of interest is the percentage of times that a franchisee achieved the target in the past (NUMACHIEVED_YRPCT i,t ). We argue that this measure proxies for the relation because it measures the extent to which the franchisee has strived to meet the target and the extent to which the franchisor saw to it that the target was achievable and hence whether a history of mutual understanding ensued. We expect that mutual understanding is more likely to develop between the franchisor and exclusive franchisees, since the franchisor is more confident that exclusive dealers will follow his guidance in sales. Thus the history of past achievements loads negatively for exclusive dealers on how much their targets will be updated in the next period. Such a muted update signals to the franchisee that the franchisor is sharing a part of the gain that accrues from the (increased) effort the franchisee exerts to meet the target. The results in Table 7 are consistent with our prediction. We find negative and significant coefficients for the interaction terms of NUMACHIEVED_YRPCT i,t and EXCLUSIVE i in both column 1 (coefficient= , p<0.05) and column 2 (coefficient= , p<0.05). These results suggest that, compared to exclusive dealers, non-exclusive dealers must have a longer history of achieving their targets before the franchisor rewards the dealers loyalty shown by achieving the target. In sum, the four tests indicate that the franchisor uses past performance to a higher extent to set targets for non-exclusive dealers. Based on this evidence, we reject hypothesis 1 and conclude that franchisors does distinguish between types of dealers when updating targets. The evidence would also suggest that it takes longer for the non-exclusive dealer to convince the franchisor of his good behavior. As a result, the franchisor ratchets targets to a lesser extent for the exclusive dealers, which could reflect a willingness to invest in the relationship with the exclusive dealer. 22

26 Mutual understanding in relation building: the side of the dealer For hypothesis 1, we examine the franchisor-franchisee relation mainly from the franchisor s perspective. This section examines relationship building from the franchisee s point of view (H2). We expect that exclusive dealers are more inclined to invest in their relation with the franchisor. Hence we expect that exclusive dealers try harder than non-exclusive dealers to increase sales, with the aim of building a long-term relationship where both parties commit to act in a predictable fashion. We expect that exclusive dealers express their commitment to a greater extent than nonexclusive dealers. We measure commitment by examining whether we observe that target ratcheting occurs. We expect that ratcheting will be present to a lesser extent for exclusive dealers, i.e., non-exclusive dealers are more likely to mute their effort and are less likely to reach their sales targets. The net effect of target levels Before we turn to the how franchisees respond to their target, we want to estimate the net effect of the decision of the franchisor to require non-exclusive dealers to achieve larger sales improvements. The purpose of this test is to estimate whether exclusive and non-exclusive dealers differ in their target achievement, given that the franchisor has decided that the non-exclusive dealers must meet higher targets to access discount levels. We run the following logit regression to test whether the likelihood of meeting the targets differs across exclusive and non-exclusive dealers. We run these tests both for quarterly (target achievements per quarter) and yearly data (achievements per year). ACHIEVED i,t = β 0 + β 1 EXCLUSIVE i + β 2 DEV_TARGET i,t 1 + control variables + ε i,t (2) 23

27 where, ACHIEVED i,t is a dummy variable capturing whether dealer i exceeds the sales target in a given year t (quarter t). We examine whether the type of dealer relates to target achievement, i.e., whether the coefficient of β 1 (EXCLUSIVE i ) loads on dealer performance measured in terms of target achievement. We reproduce our results in Table 8, Column 1 (annual data) and Column 2 (quarterly data). The coefficients of EXCLUSIVE i for the two samples are both positive. However, for the year data, the coefficient is close to the 10% level of significance (i.e., p=10.9). When analyzing the quarterly data, the effect of EXCLUSIVE i is significant at the 5% level of significance. The different significance levels of our tests would suggest that power issues affect our results. From the evidence, we conclude that the non-exclusive dealers are less likely to achieve their targets. To acquire a first idea of whether non-exclusive dealers can achieve higher targets, we test in panel B of Table 8 whether the sales progress (percentages) differs for exclusive dealers. The results in panel B suggest that non-exclusive dealers realize more sales progress. However, these differences are not significant. Taken together the latter result suggests that the non-exclusive dealers are not in a position to meet the higher target requirements the franchisor imposes on them. 9 Overall, the analysis summarized in panel A suggests that non-exclusive dealers are less likely to achieve their targets. Tests of hypothesis 2: dealer type is unrelated target achievement To test our second hypothesis, we first examine whether types of dealers are equally likely to push their sales levels beyond their targets. Our identification strategy aims at establishing whether non- 9 We also ran a regression controlling for other factors to test differences in sales progress of exclusive and nonexclusive dealers. However, this analysis did not change the conclusion we drew from the univariate analysis. 24