The Effect of Environmental Risk on the Efficiency of Negotiated Transfer Prices Markus C. Arnolda Robert M. Gillenkirchb R. Lynn Hannanc November 1, 2012 This is an early version of the manuscript. Please do not circulate without permission. Authors’ note: a University of Hamburg; phone: +49 40 42838 7509; fax: +49 40 42838 8995; email: markus.arnold@wiso.uni-hamburg.de b University of Osnabrück; phone: +49 541 969 2730; fax: +49 541 969 3613; email: robert.gillenkirch@uni-osnabrueck.de c Georgia State University; phone: +01 404 413 7212; fax: +01 404 413 7203; email: rhannan@gsu.edu The Effect of Environmental Risk on the Efficiency of Negotiated Transfer Prices Abstract This study investigates whether and how environmental risk affects the efficiency (i.e., overall organizational profit) of negotiated transfer prices. We discuss three fairness-based sharing norms and the implications each would have for efficiency in our setting. We conduct an experiment in which a buying division and selling division negotiate over the transfer of a resource at six levels of environmental risk. Because the expected value of the transfer is positive, the transfer should be made from the risk neutral organization perspective. Results show that environmental risk decreases efficiency. That is, the frequency of agreement decreases as environmental risk increases. Supplemental analysis suggests that the cause of the decrease in agreements is differences in the focal points that buyers and sellers use for determining a fair transfer price. Specifically, buyers focus on the downside potential of the transfer and sellers focus on the expected value of the transfer. As environmental risk increases, the range between these focal points increases, resulting in failed negotiations (i.e., inefficiencies). Implications for practice and theory are discussed. Keywords: Transfer price, negotiations, risk, focal points Data availability: Data are available from the authors upon request. I. Introduction: This study investigates how environmental risk affects the efficiency (i.e., overall organizational profit) of negotiated transfer prices. Decentralized organizations use transfer prices when one division provides goods or services to another internal division. Because the transfer price affects the measured profitability of the divisions, the price can affect the incentives of the division managers as they decide whether to make the transfer, and if so, the appropriate quantity of the goods or service. From an organizational perspective, the transfer price does not matter (as it is just an internal mechanism for assigning profits) unless the price creates incentives for managers to transfer a sub-optimal quantity, thereby reducing organizational profits. Several mechanisms have been proposed for determining transfer prices, which can be loosely categorized as cost-based or negotiated methods. The relative advantage of these methods can be affected by various factors such as information asymmetry, communication difficulties, and outside market opportunities. Both methods are observed in practice, suggesting that neither clearly dominates (Dikolli and Vaysman 2006). Rather, organizations must weigh the costs and benefits of various transfer pricing policies and processes given their own unique environments (Eccles 1983). Our study focuses on the negotiation method, and investigates how one factor, environmental risk, affects the efficiency of this method. In our study, the buying division faces environmental risk because the profits earned from the transfer are affected by the ex post state of nature. Understanding how environmental risk affects the efficiency of negotiated transfer prices is important because transfer pricing is a component of the overall control system in decentralized organizations. An effective transfer pricing mechanism should provide incentives for managers to transfer a quantity that maximizes overall organization profit while also allowing central management to evaluate the performance of the managers without sacrificing their autonomy 1 (Dejong, Forsythe, Kim and Uecker, 1989). Since the degree of environmental risk is likely to vary across organizations or divisions within an organization, it is important to understand its effects so that organizations can weigh the costs and benefits of the negotiation method when determining the appropriate transfer pricing policies and procedures.1 We investigate how environmental risk affects the efficiency of negotiated transfer prices via two experiments. In experiment 1, participants play the role of division managers in which one manager (hereafter the buyer) has the potential to earn divisional profits if it can acquire a resource from another division (hereafter the seller). The buyer and seller negotiate whether to make the transfer, and, if so, the transfer price. Importantly, from the risk neutral organization’s perspective, the transfer should be made because the expected value of the buyer’s profit from the transfer is greater than the opportunity cost of the seller if the transfer is made. Both parties know the buyer’s expected value and range of profits earned from the transferred resource and the seller’s opportunity cost. The experiment is a 1x6 design in which buyer-seller dyads negotiate for six independent periods (with re-matching). The manipulated variable is the degree of environmental risk (i.e., range of potential profits from the transfer) across the six periods. Results of experiment 1 show that environmental risk decreases efficiency. Specifically, increases in environmental risk increase the frequency of failed negotiations and, because the transfer is not made, expected organizational profit is lower. Supplemental analysis suggests that the cause of the failed negotiations is differences in the focal points that buyers and sellers use for determining a fair transfer price: buyers focus on the downside potential of the transfer and sellers focus on the expected value of the transfer. As environmental risk increases, the range between these focal points increases, resulting in failed negotiations (i.e., inefficiencies). 1 In an experiment investigating the negotiation method, Chalos and Haka (1990) manipulate “market uncertainty” by manipulating whether the negotiating party may have an outside option in the next period. The purpose of this manipulation is to affect the relative bargaining power of the negotiating parties rather than to capture the construct of environmental risk investigated in our study. 2 We conduct experiment 2 to acquire additional evidence that experiment 1 results are caused by differences in focal points. An alternative explanation for experiment 1 results may be that the buyer’s risk aversion decreases the transfer’s value (i.e., the certainty equivalent) as environmental risk increases, and it is this decrease in value that makes it difficult to reach agreement. Therefore, experiment 2 investigates the effect of transfer value on the outcomes of transfer price negotiations. Procedures and the design are identical to experiment 1 except that the manipulated variable is the certain profit that the buyer would make from the transfer, i.e., the transfer value. If the decreasing transfer value underlies the results of experiment 1, then we would expect to observe an increase in failed negotiations as transfer values decrease. Results of experiment 2 indicate that this is not the case as negotiation failure is unaffected by transfer value. This result lends support for our interpretation of the experiment 1 data that the underlying process through which environmental risk leads to negotiation inefficiencies is via focal point differences. Results of our study have implications for both practice and theory. From a practical perspective, it informs organizations about the importance of considering environmental risk when determining the appropriate transfer pricing policies and procedures. From a theoretical perspective, our study contributes to the negotiated transfer prices stream of research, particularly those investigating the role of fairness (Luft and Libby, 1997; Kachelmeier and Towry, 2002; Chang, Cheng, and Trotman, 2008) rather than the intensity of negotiators’ financial incentives (e.g., Chalos and Haka 1990; Ghosh, 2000). Although Luft and Libby (1997) find that negotiators say that their fairness perceptions are affected by relative payoffs, Kachelmeier and Towry (2002) show that the willingness to make a fairness-based concessions when cash payment is affected depends on the negotiation medium (computer-based or face-to-face). Chang, Cheng, and Trotman (2008) show that perceptions of the negotiation context (due to differences in 3 gain/loss framing and partner’s objectives) affect fairness judgments, potentially leading to transfer price expectation gaps between buyers and sellers. Our study develops this stream further by showing that environmental risk can lead to differences in fairness-based focal points between buyers and sellers, thereby documenting a different source for transfer price expectation gaps, which ultimately may reduce efficiency. Our study also contributes to the stream of experimental research investigating the relative efficiency of negotiated compared to cost-based methods (e.g., Dejong, Forsythe, Kim and Uecker, 1989; Avila and Ronen, 1999; Ghosh, 1994), which have focused their investigations on stable environments. Our study suggests that it is important to consider environmental risk when comparing methods. II. Setting and theory development: We investigate whether environmental risk affects the efficiency of negotiated transfer prices in a simplified setting. Our division managers work in a decentralized organization where decision rights, such as transfer prices, are delegated to the managers and managers are held accountable for those decisions via incentive compensation based on divisional profits. Whereas negotiations typically involve both the transfer price and the transfer quantity, we simplify the decision as the transfer and price of a single resource. We also simplify the decision context by restricting the setting to one where there is no outside market from which the buying division could acquire the resource. The expected value of the transfer is always positive ensuring that, from a risk neutral organization perspective, the transfer should be made. Both managers have full information about each other’s expected profits, thereby eliminating any complications that might arise from information asymmetry between the negotiating managers. This also ensures that both managers know that the transfer maximizes expected organization profit. Given that the transfer maximizes expected profit, negotiations over the transfer price are, in effect, negotiations 4 over how to share the expected surplus. We investigate whether and how environmental risk complicates these negotiations. For simplicity, we use a setting where only one party, the buyer, is subject to risk. The buyer faces environmental risk because the profits earned from the transfer are affected by the ex post state of nature. Whereas a potential risk-sharing solution would be for the managers to agree to re-negotiate after the ex post realization of the state of nature, we assume that this is not feasible due to information asymmetry regarding realized profits or the unavailability of an enforcement mechanism to guarantee such re-negotiations. Therefore, the negotiated transfer price is binding on both parties. In order to systematically investigate the effects of environmental risk, we hold the expected value of the buyer’s profits from the transfer constant and vary the range of realized profits at two levels (low and high), each with 50% probability. Consistent with prior experimental findings (e.g., Luft and Libby, 1997; Kachelmeier and Towry, 2002; Chang, Cheng, and Trotman, 2008), we assume that managers have preferences for fairness and employ some type of fairness norm to determine whether a transfer price is acceptable. We propose that environmental risk creates the potential for the emergence of three different fairness norms, depending on whether managers focus on sharing based on expected values, certainty equivalents, or ex post outcomes. As we describe below, each of these fairness norms has different implications for the outcomes of the negotiation. Whereas our focus is on frequency of agreement because of its implications for efficiency, we also describe how each fairness norm would affect the transfer price. Because no theory or empirical evidence allows us to predict which fairness norm will obtain, we describe how environmental risk would affect the negotiation outcomes for each fairness norm, and then present a research question. We model the utility functions of the managers under each fairness norm by adapting Fehr and Schmidt’s (1999) fairness model. Fehr and Schmidt’s model makes two basic assumptions 5 about individuals’ utility for fairness. First, individuals derive disutility from inequitable outcomes, regardless of whether they are worse off or better off than others. Second, such disutility is greater when the inequity is to their disadvantage than to their advantage. The basic model for a two-person setting is: Ui ( x) xi i max{x j xi ;0} i max{xi x j ;0} Where, x is the payoff received by individuals i and j i i 0 i 1 The first term measures the utility from the payoff, the second term measures the utility loss from disadvantageous equity and the third term measures the utility loss from advantageous equity. Fair sharing based on expected values When faced with risk, one potential fairness norm would be to base negotiations on the fair sharing of expected values. Whereas it is beyond the scope of this paper to investigate the source of fairness norms, we note that such a fairness norm may emerge if, for example, the managers are risk neutral or use expected values as a tool to simplify negotiations. Building on the model of Fehr and Schmidt (1999), this would imply that a manager’s utility function can be represented by the following form (where E indicates expectation, s = state of nature and p = the probability of such state): EUi ( x) ps U ( xis ) i max{E ( x j ) E ( xi ),0} i max{E( xi ) E( x j ),0} s Such a utility function implies that the fairness terms only refer to the expected outcomes of both managers. Given that our setting holds expected values constant, if managers base their 6 fairness norms on expected values, there would be no effect of risk on negotiations. That is, neither the transfer price nor frequency of agreements would be affected by risk. Fair sharing based on certainty equivalents When faced with risk, one potential fairness norm would be to base negotiations on the fair sharing of certainty equivalents. Such a sharing rule would take into consideration the risk to be borne by the managers. In this case, a manager’s utility function can be represented by the following form (where CE refers to the certainty equivalent): EUi ( x) ps U ( xis ) i max{CE ( x j ) CE ( xi ),0} i max{CE ( xi ) CE ( x j ),0} s Such a utility function implies that the fairness terms only refer to the certainty equivalency of the outcomes of both managers. As the seller in our study is not exposed to risk, the seller’s certainty equivalent is not affected by changes in environmental risk. In contrast, assuming that the buyer is risk averse, the buyer’s certainty equivalent decreases if the expected value remains constant but the risk increases. Therefore, if both managers employ a fairness norm of certainty equivalents, the transfer price will decrease as risk increases. In other words, once the managers agree to a fair transfer price in the absence of risk, that price will decrease as risk increases in order to compensate the buyer for bearing risk. The impact on the frequency of agreements is less clear, however, because the certainty equivalent is a function of the buyer’s risk preferences. Therefore, even if the managers use a fairness norm of certainty equivalents, frictions may arise regarding the degree of the buyer’s degree of risk aversion. Such frictions may result in failed negotiations. In summary, if managers use a fairness norm of certainty equivalents, the transfer price will decrease as environmental risk increases, but the frequency of agreements may be either unaffected or decrease as risk increases. Fair sharing of ex post outcomes 7 Sharing based on expected values or certainty equivalents are fairness norms based on ex ante expectations about the realized state. In both cases, fairness is viewed from the perspective of the expected outcome. An alternative fairness norm could be based on the perspective of outcomes for each possible state, i.e., the sharing of ex post outcomes. In this case, their utility functions could take on the following form:2 EUi ( x) ps [U ( xis ) i max{ x js xis ,0} i max{ xis x js ,0}] s Such a utility function implies that the managers take into consideration their relative outcomes for all states of nature. Because the second and third terms measure the utility loss from disadvantageous equity and advantageous equity, respectively, utility is maximized if the transfer price determined ex ante minimizes both fairness terms ex post. Recall that in the Fehr and Schmidt model i i , i.e., the disutility from disadvantageous equity is greater than or equal to the disutility from advantageous equity. This implies that an ex post sharing norm will lead the buyer to prefer a low transfer price to avoid disadvantageous inequity in bad states (when the revenue the buying division receives from the transfer is low) and the seller to prefer a high transfer price to avoid disadvantageous inequity in good states (when the revenue the buying division receives from the transfer is high). Consequently, the buyer prefers an ex ante transfer price that is closer to the minimum outcome whereas the seller prefers an ex ante transfer price that is closer to the maximum outcome. Given that increases in environmental risk increase the spread between the minimum and maximum outcomes, conflict between the two managers increases. This implies that as environmental risk increases, the frequency of agreements decreases. 2 The following representation assumes that utility is separable into utility from wealth and disutility from inequity and that there is no risk aversion with regard to inequity. 8 The effect of environmental risk on agreement frequency may be exacerbated if managers approach ex post fairness with self-serving fairness biases (c.f., Messick and Sentis 1979). A selfserving fairness bias is the tendency of individuals to rely on fairness criteria which favor their own self interest. Such differences in fairness criteria could lead to conflict during negotiations. For example, Babcock and Loewenstein (1997) provide evidence that negotiation impasses in public school teacher contract negotiations are attributable to self-serving beliefs about which community represents the appropriate benchmark. Because environmental risk necessarily entails a range of outcomes, it may lead to self-serving notions of which outcome should be the focal point for determining the fair outcome. Differences in buyer and seller self-serving focal points could cause conflicts, resulting in failed negotiations. It is less clear how environmental risk would affect the transfer price itself. Although an ex post sharing norm would lead the managers to focus on different outcomes, how they resolve this in setting the transfer price would depend on the relative tenacity of the two managers during negotiations. In other words, we can make no prediction about how environmental risk systematically affects the transfer price. In summary, if managers use an ex post outcomes fairness sharing norm, the frequency of agreements will decrease as environmental risk increases, but the transfer price may increase, decrease or stay the same. Research question The discussion of the potential fairness sharing rules, above, suggests that the effect of environmental risk on the frequency of agreements depends on the sharing norm employed by the managers. Because theory and empirical evidence do not allow us to predict which sharing norm will dominate, we present a research question: RQ1: Will the frequency of agreement be affected by changes in environmental risk? 9 III. Experiment 1 – Experimental Design We conduct an experiment in which buyer-seller dyads negotiate over the price of a resource that the seller transfers to the buyer if the negotiations result in agreement. The experimental design is a 1x6, in which we manipulate environmental risk over six periods.3 Dyads are re-matched each period and each period is independent. Environmental risk is manipulated by varying the range of returns that the buyer earns from the transferred resource, depending upon the ex post state of nature. The expected value of the returns from the transferred resource is constant across all periods. Specifically, the six levels of environmental risk consist of the following returns for the low (high) state, with 50% probability of low or high state for each level: 160(160), 150(170), 140(180), 130(190), and 110(120). The primary dependent variable is agreement frequency. Because the expected value of the buyer’s profit from the transfer is greater than the opportunity cost of the seller if the transfer is made, a lower agreement frequency indicates decreased efficiency (i.e., expected organization profits). Participants and Procedures Participants are students recruited from the experimental economics database at a large public university in Germany. Two experimental sessions were conducted, with 24 participants (12 dyads) per session, resulting in a total of 48 participants (24 dyads) of whom 56% are business or economics majors. Fifty percent of the participants are female. The average age is 24.1 years, and the average full time working experience is 1.21 years. Participants interact over a computer network, with anonymity preserved both during and after the experimental sessions. The sessions last on average 70 minutes. At the beginning of the session, participants are provided with written instructions, which are read aloud by one of the experimenters. The instructions inform participants that they will 3 We use six orders of environmental risk, so order is a second factor. Since order has no effect on frequency of agreement (p > 0.10 for each order), we exclude it from further discussion. 10 play the role of a team manager, A or B, in a firm that has several teams. The firm wants managers to make decisions that will result in maximum profit for the firm, and, as an incentive to do so, team managers are paid based on their own team profit. Team profits are computed in points, which are converted to Euros at the end of the experiment for determining participant pay. Team B manager (i.e., the buyer) has the opportunity to work on an additional revenueproducing project which would require borrowing an employee from team A (i.e., the seller) for implementation. During the experiment, the two team managers negotiate the price that team B pays to team A for the borrowed employee. These negotiations take place over six independent periods during the experiment. Participants maintain the same role (team manager A or B) throughout. Participants are re-matched with a new manager each period such that they are never matched with the same manager more than once. Team profits are determined by whether the transfer is made, and if so, the negotiated transfer price. Specifically, in team A, where the employee currently works, the employee’s contribution to team profit is certain and equals 125 points. As the employee causes costs of 100 points, the certain net profit from the employee working in team A is 125 - 100 = 25 points. If team A transfers the employee to team B, then team A receives the transfer price, from which 100 points are deducted to determine team A’s profit. If no transfer is made, then team A keeps the employee and earns 25 points. Team B’s project has an expected contribution to team profit, or “return,” of 160 points. Although the expected return from the project is 160 points, in some periods there is uncertainty about the realized project return. When there is uncertainty, the realized project return is either high or low, with equal probability of 50%. Both returns have the same distance to 160 so that the average is always 160 points. If the transfer is made, then team B receives the realized returns from the project (high or low when there is uncertainty) from which the transfer price is deducted 11 to determine team B’s net profit. If no transfer is made, then team B earns nothing from the project. Team payoffs are summarized as follows: If the managers of teams A and B agree on a transfer price: Profit Team A = Transfer Price - 100 Profit Team B = Realized Project Return - Transfer Price If the managers of teams A and B do not agree on a transfer price: Profit Team A = 125 - 100 = 25 Profit Team B = 0 Participants are informed that, at the beginning of each negotiation period, both team managers will learn whether there is uncertainty about the amount of team B’s realized project return, and if so, the potential high and low amount. That is, there is no information asymmetry between the negotiating partners. Further, participants are clearly informed that, from an expectation perspective, firm profits are increased by 35 points (team B’s expected return of 160 - team A’s certain return of 125) if the transfer is made. Participants complete a computerized pre-experiment quiz to ensure that they understand the experiment before they are allowed to begin the first period. The same procedures are followed in each period. Each period begins with the computer forming a dyad via the matching protocol. Then the computer displays to both managers the range of project returns for team B for that period. We elicit information from the managers after they learn the range but before they start the negotiations. Specifically, team A managers input the minimum transfer price they are willing to accept and team B managers input the maximum transfer price they are willing to pay. This information is gathered for supplemental analyses purposes; no other participant is informed of the price and it is not binding on negotiations. 12 Negotiations begin with team manager B making the first offer. After the first offer is made, both managers are free to submit offers. Either manager may accept an offer or break off negotiations at any time during the allotted negotiation time, which is 150 seconds. Negotiation time is costly for both team members because points are deducted for time spent negotiating. Specifically, each team manager receives 5 points at the beginning of each negotiation period and loses one point after each 30 seconds. A clock is displayed on the computer showing the time remaining so that participants can keep track of the time. If no agreement is reached by the end of the allotted negotiation time, negotiations have failed and no transfer is made. At the end of the period, the computer displays the negotiation outcome (agreement or disagreement) and negotiation time. If the negotiations resulted in an agreement, the computer also displays to both managers the transfer price and randomly-determined realized project return for team B. Next, regardless of agreement status, the computer displays each manager’s earnings for the period (both profit and time). Then the next period begins. At the conclusion of the six periods, participants complete a post-experiment questionnaire. One period is selected at random to be the payment period. Participants receive cash payment based on the points earned in the experiment, and are dismissed. Experimental earnings Participants’ cash earnings are determined by converting their experimental points into Euros. The conversion rate is 1 Euro for each 4 points. The payment is determined by summing an initial endowment of 32 points and the points earned in the randomly-selected period (i.e., profit points and time points). If the realized return for team manager B results in negative profit for the payment period, the negative earnings are deducted from the 32 point initial endowment. Participants are informed of this procedure before negotiations begin. 13 IV. Experiment 1 – Results Our research question addresses whether environmental risk affects the frequency of agreements. Table 1 reports descriptive statistics for experiment 1. As reported in Table 1, and graphed in Figure 1, agreement frequency tends to decrease as environmental risk increases. Agreement frequency is 79.2% in the absence of environmental risk, and decreases to 37.5% for the highest level of risk (110/210), although the decrease is not strictly monotonic. This pattern of results suggests that participants did not use sharing based on expected value as their fairness norm, as this norm would have no effect on agreement frequency. Further, as reported in Table 1, for those dyads who were able reach an agreement, the time to reach an agreement (i.e., negotiation time) increases with environmental risk. This suggests that environmental risk increases conflict. Finally, the transfer price tends to decrease with environment risk, from 136.42 in the absence of risk to 130.00 for the highest level of risk (110/210). This suggests that there is some degree of compensation provided to the buyer for bearing risk. 4 To test whether environmental risk decreases agreement frequency, we conduct a logit regression, with environment risk as the independent variable and whether agreement is reached (0/1) as the dependent variable. 5 As reported in Table 2, the coefficient on environmental risk is negative and statistically significant (-0.733, p <0.001). This result shows that participants in our 4 A natural question is whether the results could be attributed to differences in risk preferences across the participants even though they were randomly assigned to be the buyers and sellers. To address this potential concern, we measured risk aversion on the post experiment questionnaire and find no differences in risk preferences. Specifically, we asked the following question: Imagine you own a gamble which offers a 50% chance of winning 20 Euro and a 50% chance of losing 5 Euro. Would you accept a sure payment of 7.50 Euro in exchange for this gamble? They could respond, yes, not or indifferent. If participants answered yes, they were asked to write an amount that would make them indifferent. If participants answered no, they were asked how high the sure payment would have to be to make them indifferent. This gamble corresponds to the payoffs from the transfer price negotiation for the project with the highest risk (110/210) if both parties had agreed upon a transfer price of 130. On average, buyers indicate a certainty equivalent of 8.50 Euro while sellers indicate a certainty equivalent of 9.64 Euro. The difference is not significant (t-test, t = 1.20, p = 0.235). 5 Due to multiple observations within subjects, we use fixed effects regressions with standard errors clustered at the seller level and include period fixed effects for all regressions reported in the paper. For the regression testing agreement frequency, the n is reduced from 144 to 138 dyads because we exclude one seller who always reached agreement, and, therefore predicts success perfectly in this fixed effects regression. Without fixed effects, the coefficient of environmental risk remains at the same significance level. 14 experiment were less likely to be able to reach an agreement as environmental risk increased. Because the transfer increases the expected value of overall organizational profits, negotiation failures lead to a loss of efficiency, i.e., environmental risk reduces efficiency. As reported in Table 1, it appears that it takes participants who reach an agreement a longer time to do as environmental risk increases. To test whether this is the case, we run a regression with environmental risk as the independent variable and time to agreement as the dependent variable. Untabulated results find that the coefficient on environmental risk is positive and statistically significant (4.51, p <0.02). This result shows that participants in our experiment took more time to reach an agreement as environmental risk increased, which is consistent with environmental risk increasing conflict. To gain insight into fairness sharing norms used by our participants, we also test whether environmental risk affects the transfer price. We run a regression with environmental risk as the independent variable and transfer price as the dependent variable for the 84 dyads who reached an agreement. Untabulated results find that the coefficient on environmental risk is negative and statistically significant (-0.81, p =0.052). These results show that, for dyads who reached agreement, the seller was willing to compensate the buyer somewhat for bearing risk. Overall, our results show that increases in environmental risk make it more difficult to reach an agreement, increasing the likelihood that negotiations fail, thereby reducing efficiency. We can reject the usage of a fairness sharing norm based on expected values because such a norm would not have reduced efficiency. We cannot distinguish between the other two sharing norms, though, because the decrease in agreements and decrease in transfer prices that we observe in response to increased environmental risk are consistent with both norms. Therefore, we conduct supplemental analysis in order to glean more insight into the fairness perspectives of the participants. 15 Supplemental analysis We gain insight into participants' fairness perspectives via data collected in the experiment before negotiations began each period. Recall that, after learning the project returns for the buyer (i.e., team B) each period, sellers input the minimum transfer price they are willing to accept and buyers input the maximum transfer price they are willing to pay. We use these data to explore how participants respond to environmental risk in determining their acceptable transfer prices. We are especially interested in how environmental risk affects the distance between acceptable transfer prices as the wider the distance the more difficult it would be to reach an agreement. Table 1 reports the buyers' maximum and the sellers' minimum transfer price for each level of environmental risk. These data suggest that the sellers' minimum transfer price is relatively flat, whereas the buyers' maximum transfer price decreases with environmental risk. To test whether this is the case, we run two regressions, one for sellers and one for buyers. The independent variable for the regressions is environmental risk and the dependent variable is minimum transfer price for the sellers and maximum transfer price for the buyers. Regressions include fixed round and player (seller or buyer) effects and use standard errors clustered at players’ level. As reported in Table 3, the sellers' minimum transfer price is not affected by environmental risk (p. = 0.965). However, the buyers' maximum transfer price decreases with environmental risk (-3.442, p. < 0.001). These results indicate that, when considering both negotiating parties, the fairness sharing norm was not based on certainty equivalents because the sellers' minimum transfer price is unaffected by the risk borne by the buyer. The results are consistent with environment risk leading to differing focal points, although not quite as we predicted under an ex post sharing norm. Specifically, the data suggest that sellers focus on the expected value of the return and desire a share of approximately 60% of the expected outcome independent of the level of environmental risk. That is, the seller’s minimum transfer 16 price is approximately 136 points for all levels of environmental risk. Given that the expected return is 160 points and the cost of the employee is 100 points, the overall surplus is expected to be 60 points. At a transfer price of 136 points the seller would earn 36 points (136-100) and the buyer would earn 24 points in expectation (160-136), representing a 60%-40% split. In contrast, the data suggest that buyers focus on the low outcome, reducing the acceptable transfer price as environmental risk increases in order to be compensated for the risk of a low return. V. Experiment 2 – Design and Results Experiment 1 shows that as environmental risk increases, the agreement frequency of transfer price negotiations decreases. Supplemental analysis suggests that buyers and sellers used different focal points. Since environmental risk increases the range of outcomes, and, therefore, the distance between focal points, it is reasonable to conclude that the results of experiment 1 are due to differences in focal points. We conduct experiment 2 to explore an alternative explanation for these results. As discussed previously, for a given expected value, the certainty equivalent for risk averse individuals decreases if risk increases. As the certainty equivalency decreases, the overall value to be shared decreases as well, making it potentially more difficult to come to an agreement over how to share a smaller surplus. Thus, it may be that this decrease in the surplus is behind the results of experiment 1. To explore whether the decrease in transfer value (from a certainty equivalent perspective) is responsible for the decrease in agreement frequency, we manipulate the value of the surplus while holding risk at a constant level of zero. That is, experiment 2 is identical to experiment 1 except that the known returns to the buyer (i.e., team B) are varied across the six periods. We use the following six levels of returns: 140, 150, 160, 170, 180, and 190 points. As in experiment 1, 17 the transfer is always value creating for the firm, so the transfer represents a Pareto improvement as long as the buyer and seller can reach an agreement on the transfer price. 6 If decreasing transfer value leads to increased difficulty in reaching an agreement, we would expect to see a decrease in agreement frequencies as the transfer value decreases. If we do not observe such a result, it bolsters our confidence that the results of experiment 1 are indeed due to differences in focal points between the buyers and sellers. As in experiment 1, we conducted two sessions with 24 participants (12 dyads) per session, resulting in a total of 48 participants (24 dyads). Participants were recruited from the same experimental economics database as in experiment 1. Descriptive statistics are reported in Table 4. As reported in Table 4, and graphed in Figure 2, there is an apparently slight decrease in the frequency of agreements as the transfer value decreases. However, as reported in Table 5, results of a logit regression show that this is not statistically significant. Specifically, a logit regression, with the value of the transfer as the independent variable and whether agreement is reached (0/1) as the dependent variable finds that the coefficient on transfer value is not significant (p = 0.676).7 Interestingly, we find that, for dyads who reach agreements, the negotiation time decreases as the transfer value decreases (untabulated, -5.18, p <0.02). This result is inconsistent with the notion of such decreases making negotiations more difficult. Notably, this result is the opposite of that observed in experiment 1, where decreases in transfer value (from a certainty equivalency perspective) increase negotiation time. 6 We chose 140 as the lowest value because the lowest certain amount indicated by the buyers to the risk question on the post experiment questionnaire of experiment 1 (see footnote 4) was 2.50 Euro. This corresponds to a certain gain of 10 points in the experiment. Thus, for a transfer price of 130 points, the most risk averse buyer would demand a certain project return of 140 points in order to exchange the certain project for the uncertain project with returns of 110 and 210 and an expected value of 160 points. 7 The regression uses seller and round fixed effects and clusters standard errors at seller level. Six sellers are excluded as the always reach agreement. Without fixed effects, the coefficient of transfer value remains insignificant at conventional levels (p > 0.4). 18 Finally, we explore the data gathered during the experiment from the buyers and sellers regarding the maximum transfer price and minimum transfer price, respectively, that they would be willing to accept for each transfer value (as reported in Table 4). Table 6 reports the results of regressions to test how transfer value affects the buyers and sellers maximum and minimum transfer prices. As in experiment 1, buyers’ acceptable transfer prices decrease with transfer value (-5.59, p <0.001). In contrast to experiment 1, sellers’ acceptable transfer price also decreases as the transfer value decreases (-4.29, p <0.001). Overall, the results of experiment 2 suggest that decreasing transfer value does not make negotiations more difficult as frequency of agreement is unaffected by transfer value. In fact, from a negotiation time perspective, decreases in transfer value make it easier to come to agreement. Further, both buyers and sellers decrease their acceptable transfer prices as the known transfer value decreases. This result suggests that, in the absence of the multiple focal points, as observed in risky environments, buyers and sellers can reach agreements. VI. Discussion and Conclusion This study investigates how environmental risk affects the efficiency (i.e., overall organizational profit) of negotiated transfer prices. In experiment 1, buyers and sellers negotiate the transfer price of a resource over six independent periods, with environmental risk manipulated at six levels (one level each period). There is no information asymmetry between the negotiating parties, and, from the risk neutral organization’s perspective, the transfer should be made. Results show that environmental risk decreases the efficiency (i.e., overall organization profit) of the negotiations because negotiations fail more frequently as risk increases. Supplemental analysis suggests that the cause of the failed negotiations is differences in the focal points that buyers and sellers use for determining a fair transfer price: buyers focus on the 19 downside potential of the transfer and sellers focus on the expected value of the transfer. As environmental risk increases, the range between these focal points increases, resulting in failed negotiations. Experiment 2 provides evidences that experiment 1 results cannot be attributed to a more difficult negotiation environment due to a decrease in the transfer’s value from a certainty equivalent perspective. Thus, it provides corroborating evidence that the experiment 1 results are due to differences in focal points caused by changes in environmental risk. The results of our study are important for both practical and theoretical reasons. From a practical perspective, it informs organizations about how environmental risk affects the efficiency of the negotiated method for setting transfer prices. As such, it provides useful insights for organizations as they weigh the costs and benefits of the negotiation method when determining the appropriate transfer pricing policies and procedures. From a theoretical perspective, it contributes to the stream of research investigating frictions that may arise in transfer pricing negotiations, especially as they relate to fairness perceptions. By showing that environmental risk may lead the buying and selling managers to use different focal points for determining a fair transfer price, our study helps to build this stream. Understanding factors that affect fairness perceptions in transfer pricing is especially important given that managers are likely to interact in other contexts and, therefore, such fairness perceptions may carry-over and have implications for these other interactions as well. 20 References Avila, M. and J. Ronen. 1999. Transfer-pricing mechanisms: An experimental investigation. International Journal of Industrial Organization 17(5): 689-715. Babcock, L. and G. Loewenstein. 1997. Explaining bargaining impasse: The role of self-serving biases. Journal of Economic Perspectives 11 (1): 109-126. Chang, L., M. Cheng and K. Trotman. 2008. The effect of framing and negotiation partner’s objective on judgments about negotiated transfer prices. Accounting, Organizations and Society 33, 704-717. Chalos, P. and S. Haka. 1990. Transfer pricing under bilateral bargaining, The Accounting Review 65(3): 624-641. Dejong, D. V., R. Forsythe, J. Kim, and W. C. Uecker. 1989. A laboratory investigation of alternative transfer pricing mechanisms. Accounting, Organizations and Society 14(1-2): 41-64. Dikolli, S. S. and I. Vaysman. 2006. Information technology, organizational design, and transfer pricing. Journal of Accounting and Economics 41(1-2): 201-234. Eccles, R. G. 1983. Control and fairness in transfer pricing. Harvard Business Review (November-December): 149-161. Fehr, E. and K. M. Schmidt. 1999. A theory of fairness, competition, and cooperation. The Quarterly Journal of Economics 114 (3): 817-868. Ghosh, D. 1994. Intra-firm pricing: Experimental evaluation of alternative mechanisms. Journal of Management Accounting Research 6: 78-92. Ghosh, D. 2000. Complementary arrangements of organizational factors and outcomes of negotiated transfer price. Accounting, Organizations and Society 25, 661-682. Kachelmeier, S. J. and K. L. Towry. 2002. Negotiated transfer pricing: Is fairness easier said than done? The Accounting Review 77(3): 571-593. Luft, J. L. and R. Libby. 1997. Profit Comparisons, Market prices and managers' judgments about negotiated transfer prices. The Accounting Review 72(2): 217-229. Messick, D., and K. Sentis. Fairness and preference. Journal of Experimental Social Psychology 15 (4):418-434. 21 Figure 1 Experiment 1 Agreement Frequency 22 Figure 2 Experiment 2 Agreement Frequency 23 Table 1 Experiment 1 Descriptive Statistics Mean Median (Standard deviation) Environmental risk 140/180 130/190 160 150/170 120/200 110/210 Agreement Frequency .792 1 (.415) .917 1 (.282) .542 1 (.509) .583 1 (.504) .292 0 (.464) .375 0 (.494) Transfer Price 136.42 135.00 (6.27) 135.50 137.50 (8.63) 133.15 134.00 (12.08) 133.43 133.50 (7.00) 133.57 133.00 (16.14) 130.00 132.00 (12.37) Negotiation Time if agreement 44.63 42.00 (31.82) 49.77 42.00 (36.20) 45.46 40.00 (30.04) 54.00 48.50 (34.84) 61.86 62.00 (28.52) 56.00 34.00 (40.74) Buyer’s Maximum Transfer Price 138.08 137.50 (13.58) 140.13 142.50 (13.10) 134.04 138.50 (12.25) 130.83 130.00 (11.97) 124.58 120.00 (11.36) 123.96 122.50 (16.01) Seller’s Minimum Transfer Price 135.42 135.00 (11.50) 136.75 135.00 (9.60) 136.00 134.00 (11.17) 134.67 130.00 (13.42) 138.29 132.50 (14.80) 134.92 130.00 (12.60) 24 Table 2 Experiment 1 The Effect of Environmental Risk on Agreement Frequency Dependent variable Agreement (0/1) Coefficient Standard error (p-values) Independent variables Environmental Risk -0.733 0.179 (<.001) Constant 4.856 1.070 (<.001) N Pseudo R2 138 .336 25 Table 3 Experiment 1 The Effect of Environmental Risk on Minimum and Maximum Acceptable Transfer Price Dependent variable Independent variables Seller’s Minimum Transfer Price Buyer’s Maximum Transfer Price Coefficient Standard error (p-values) Coefficient Standard error (p-values) 0.055 0.662 -3.442 .785 (.935) (<.001) 130.39 4.128 141.17 4.347 (<.001) (<.001) 144 .05 144 .19 Environmental risk Constant N R2 26 Table 4 Experiment 2 Descriptive Statistics Mean Median (Standard deviation) 190 .833 1 (.381) 180 .833 1 (.381) Transfer value 170 160 .667 .875 1 1 (.482) (.338) Transfer Price 154.50 154.00 (9.99) 145.45 149.00 (8.63) 144.63 145.00 (7.29) 136.71 137.00 (4.57) 132.41 132.00 (4.73) 129.33 130.00 (3.60) Negotiation Time if agreement 72.50 60.50 (37.50) 62.46 55.00 (40.52) 84.67 79.50 (54.06) 53.00 45.00 (36.77) 54.54 44.00 (38.81) 51.08 52.50 (34.77) Buyer’s Maximum Transfer Price 153.63 159.00 (21.03) 148.33 150.00 (19.86) 140.79 145.00 (18.39 135.46 140.00 (15.51) 130.08 132.50 (13.67) 126.83 130.00 (13.67) Seller’s Minimum Transfer Price 153.13 150.00 (9.92) 145.87 141.00 (12.86) 141.83 140.00 (15.78) 136.96 134.00 (15.27) 134.88 130.00 (15.00) 130.67 130.00 (15.57) Agreement Frequency 27 150 .708 1 (.464) 140 .750 1 (.442) Table 5 Experiment 2 The Effect of Transfer Value on Agreement Frequency Dependent variable Agreement (0/1) Coefficient Standard error Independent variables (p-values) Transfer Value -.075 .179 (.676) Constant 2.99 1.249 (.016) N Pseudo R2 108 .15 28 Table 6 Experiment 2 The Effect of Transfer Value on Minimum and Maximum Acceptable Transfer Price Dependent variable Independent variables Seller’s Minimum Transfer Price Buyer’s Maximum Transfer Price Coefficient Standard error (p-values) Coefficient Standard error (p-values) -4.290 .552 -5.587 .570 (<.001) (<.001) 149.27 1.685 158.47 4.347 (<.001) (<.001) 144 .24 144 .25 Transfer Value Constant N R2 29