Slack and participative budgeting – always an evil?

Michael Kopel (University of Graz) Christian Riegler (Vienna University of Economics and Business)

November 2014

Abstract: Participative budgeting is widely used in corporate practice despite empirical research finds only mixed evidence of its benefits. This may be due to neglected drivers in the theoretical grounding of related research. So far analytical models focus on the benefits and costs of participative budgeting within a firm. They find that participative budgeting may improve the knowledge of a superior or increase the motivation of subordinates. However, this often comes at the costs of padding budgets and slack building by subordinates. Most studies find that slack only drives the costs of participating budgeting decreasing its benefits. By additionally considering market effects caused by participative budgeting, we find that slack may also be beneficial for a firm. In an analytical model, we characterize the basic trade-off between slack building and the decisions of market participants outside the firm. The total effect may be positive or negative. Therefore, focusing only on aspects of budgeting within a firm may miss an important effect of participative budgeting. It seems fruitful to consider this effect in future empirical research.

1. Introduction Budgets are one of the most widely used management accounting tools all over the world. We can find extensive discussions of budgeting techniques and problems in almost every management accounting textbook.1 Research also documents a wide use in corporate practice in medium and large firms and a high perceived usefulness.2 The high usefulness is also a result of the multiple functions of budgeting, ranging from planning and controlling annual operations, communicating information, coordinating activities, motivating managers and evaluating their performance. It is obvious that these functions face increasing importance in decentralized large companies, where managers of different hierarchical layers may have more decision making authority. It has a long tradition in the budgeting literature to discuss whether the participation of managers in the process of preparing a budget is beneficial for the firm. In the past, this important question was often addressed based on different research traditions.3 Psychology based research focuses on motivation and behavior. It is argued that involvement in the budgeting process may increase job satisfaction and morale. This leads to greater commitment and acceptance of goals fixed in the budget. As a consequence, performance of managers and of the company may increase. Empirical research addressing these aspects finds mixed evidence. Several studies find a significant positive relationship of participation and performance, others observe no relationship and some even find a negative association.4 The role of asymmetric information and its impact on firm profit was especially addressed in economic based research. The question if and how the better information of the subordinate manager should be used in the budgeting process was addressed in a series of papers in the 1980s and 1990s. These studies focus on the game, in which managers try to bias communicated information. By padding the budget and building budgetary slack the subordinate manager communicates underestimated revenues or overestimated costs to the superior. This results in more easily achievable targets and in inefficient resource consumption. Participation in the budgeting process has therefore the advantage of improving the knowledge of the superior, which he may use for planning and controlling activities on the firm level. As a result, the process may also provide incentives for the subordinate to work harder. However, participation comes at the costs of slack diminishing the benefits of this budgeting regime compared to authoritative top down budgeting. As a consequence, participation in the budgeting process may be beneficial or not. A common view of slack in the budgeting literature is that slack is not beneficial for the firm. Slack leads to biased information causing inefficient decisions based on this information like biased resource allocation, biased targets for subordinates or inefficient effort choices and resource consumption. Participation per se could be beneficial. However, slack building behavior of the subordinate limits its benefits. To make better use of the beneficial aspects, firms try to find mechanisms to decrease the amount of slack.5 Related literature finds that benefits and costs of using budgets and the optimal design of the budgeting process is highly context specific. The optimal solution depends on the importance of the (sometimes conflicting) functions of budgeting as well as on the characteristics 1

See e.g. Drury (2012), Bhimani/Horngren/Datar/Foster (2011) or Atkinson/Kaplan/Matsumura/Young (2007). See e.g. Shastri/Stout (2008). 3 See for a survey Covaleski/Evans/Luft/Shields (2007). 4 See Maiga/Nilsson/Jacobs (2014), p. 6. 5 See Atkinson/Kaplan/Matsumura/Young (2007), p. 490. 2

2

of the situation analyzed. The advantage from using budgeting as a mechanism itself and the best way how to design the mechanism may differ from setting to setting. Thus, it is not surprising that empirical research also observes mixed results when searching for evidence for the findings of analytical research. The mixed results also motivate our following analysis. Due to the complexity of budgeting it does not seem possible to develop a closed theory considering all functions and characteristics of the participative budgeting setting. Therefore, it is important to discuss aspects, which have been neglected so far in the related literature. This may be helpful to further explain why participative budgeting is widely used in corporate practice. We also want to emphasize the danger of misleading findings if the focus of an analysis is set too narrow. Our aim is to add an additional piece to the budgeting puzzle. We address the question if slack is always only a disadvantage or if slack could also be beneficial for a firm increasing the value of participative budgeting. This is contrary to the common view of the budgeting literature. To do this we see participative budgeting as being part of the culture and leadership style of a company. Especially for large firms, corporate culture is often well known also outside the firm. Some firms are known for a participative, employee oriented culture whereas others are known for being authoritative. As culture does not change rapidly, a firm shows a strong sustainable commitment of being participative or authoritative. This is also reflected in the way the budgeting process is designed. When other market participants observe the commitment, this may also have an impact on their decisions, whenever the decisions of the firms are in some way interrelated. For an economic analysis, we model such a setting. We analyze the strategic impact of the choice between two alternative designs of the budgeting process (top down versus participative) on the market behavior of firms. The interrelated decisions of firms on the market are captured by the pricing decision of a supplier. He delivers a critical resource for the production process of the firm choosing the budgeting mechanism. The decentralized firm consists of two divisions, both requiring the critical input resource for the production in their respective unit. Based on the budgeting process, the cost budget for production for each division is fixed. The design of the budgeting process impacts the cost budget of each division, but also the pricing decision of the supplier. Participative budgeting may be beneficial despite slack building and increasing real production costs of the firm, because of its strategic effect on the input market. Central headquarters can derive a threshold for the preferred budgeting design based on its expected production costs (building on its ex ante expectation without getting information of a better informed divisional manager). We find that c.p. headquarters prefers participative budgeting if the ex ante expected production costs are in an intermediate range. If expected production costs are low or high, top down budgeting is preferred. Our contribution to literature is to show that slack is not always only a cost for participative budgeting. Slack may even increase its benefits in some settings. This finding adds to the literature searching for contingencies driving the benefits of the design of the budgeting process. It also highlights that models focusing only on the effects of the budgeting process within the firm may come at the risk of missing important effects on the market behavior. As a consequence, recommendations may also be misleading. The analysis offers also room for empirical research. The setting is (as in most of the models analyzing budgeting issues) very specific. However, the important drivers of the results are (to some extent) observable outside the firm, which makes an empirical investigation possible. 3

The paper proceeds as follows. The next section discusses related literature. Section 3 provides the model setup, analysis and presents the findings. Section 4 concludes.

2. Related Literature Very recently, participative budgeting again gains increasing interest in analytical management accounting research. Heinle/Ross/Saouma (2014) compares top down and participative budgeting in a different setting from ours. They assume that either the principal (top down budgeting) or the agent (manager, bottom up budgeting) receives and communicates private information. In each of the two budgeting mechanisms the better informed player has an incentive to misreport. For participative budgeting, the manager has an incentive to bias the report pessimistically to allow for reducing effort building slack. For top down budgeting, the principal has an incentive to bias the report optimistically to motivate the manager to exert higher effort. The authors find that the level of information asymmetry drives the choice of the preferred budgeting mechanism. There is a switch from top down budgeting to participative budgeting as the amount of private information increases. In this setting slack is a cost negatively affecting the benefits of participative budgeting. However, slack is also a commitment device of the principal to signal truthful reporting when top down budgeting is implemented. Our analysis differs from Heinle/Ross/Saouma (2014) as we do not only consider budgeting as a communication tool within the firm. Budgeting may also serve to convey information to market participants outside the firm. In this context, slack may not only have a negative impact on the superiority of participative budgeting for the principal. To the contrary, it may also increase its benefits. Weiskirchner-Merten (2014) provides an analysis for a budgeting setting with one principal and two agents (managers). The divisions of the managers are interdependent and jointly generate the firm’s profit. For both divisions, cost budgets for production have to be determined either based on top down or participative budgeting. Weiskirchner-Merten finds that if the profit potential is low, top down budgeting is preferred. As the profit potential increases participative budgeting becomes more favorable. The level of cooperation between the two divisions and the effectiveness of the principal’s coordination of the two divisions also has an impact on the choice of the preferred budgeting regime. Differing from her analysis we also consider budgeting as a tool to convey information to other market participants. In this regard, slack does not necessarily decrease the benefits of participative budgeting and may even be advantageous. In our setting we find that top down budgeting is preferred for low expected ex ante production costs (which could be seen as high profit potential). This is contrary to the findings of Weiskirchner-Merten and highlights the importance of considering different contextual factors and settings to get a more complete picture of the drivers of the benefits of participative budgeting. In the 1980s and 1990s, several analytical papers addressed the role of communication and target setting on the effort choice of an agent.6 The findings emphasize that in most cases the principal has to offer rents (i.e. slack) to the agent to induce truthful reporting. This is also in line with the findings of the revelation principle. Again, slack has only a negative impact on the benefits of participative budgeting. This is true for some functions of budgeting discussed in the introduction. However, it is also possible to think of settings where it is not in the best interest of the principal to become in6

See e.g. Demski/Feltham (1978), Christensen (1982) or Baiman/Evans III (1983).

4

formed truthfully by the agent, although participative budgeting is implemented. By this, the principal may credibly commit to a behavior that an informed principal would never show. This may have an impact on the decisions of market participants outside the firm. The total effect may be beneficial for the principal, even if slack induces real costs and decreases the real profit of the firm.7 The economic forces driving the result are comparable to the findings of Sappington/Weisman (2005) that increasing the cost of an upstream product may disadvantage downstream rivals. It is fruitful to consider their basic idea developed for the setting of a vertically-integrated producer for the budgeting context as well. The model also relates to the strategic use of transfer pricing on input markets analyzed by Arya and Mittendorf (2010). Using transfer prices above marginal costs can have an impact on the pricing decision of a supplier of an input resource for the firm. An increase in the transfer price may induce the supplier to lower the price for the input to stimulate demand for his resource. Depending on the characteristics of the firm, this may be a profitable strategy for the firm. Contrary to these models we focus on the design and choice of the budgeting regime. The corporate culture of a firm (authoritative or participative) is reflected in the design of its budgeting process. Additionally, corporate culture is often well known (observable) by other market participants. Thus, we can expect comparable economic forces to be in place. However, in order to represent the budgeting process, the basic model has to be altered significantly, especially by adding a stage representing the communication between principal and divisional managers and the choice of the budgeting regime. On the one hand, this provides a significant extension of this model class. On the other hand, this enables us to gain new insights into the context of budgeting. By this we want to contribute to the puzzle of mixed empirical results with respect to questions of participative budgeting.8 By modeling aspects not included in analytical research in the past we extend the theory. The findings can also be investigated by empirical research. Divisions of different relative profitability, who have to rely on a common input resource provided by a supplier with high market power, represent important drivers of the results derived. This information is (to some extent) also observable for empirical researchers outside the firm.

3. The Model We consider a principal, who employs two managers each running one of the two separate divisions of the firm. We assume that the principal and the managers are risk-neutral. Both divisions serve two separate monopolistic output markets. The demands for the respective product of each division are described by pi=ai – xi, i=1,2. The demand functions are common knowledge. An interdependence of the two divisions arises, because both of them require the same type of input resource for their production process. The input comes from a single supplier from outside the firm. There is no alternative source offering the resource needed. The firm buys the quantity needed in one single order. So the price of the resource depends on the quantity ordered by both divisions. For 7

E.g. thinking of strategic transfer pricing, delegation of decisions and fixing transfer prices above marginal costs may soften competition and also have an impact on other market participants. But here, only the costs driving the delegated decision of a manager are biased, not the real costs of production within the firm. See e.g. Göx (2000). 8 See for an overview Heinle/Ross/Saouma (2014), p. 1026f.

5

simplicity, we assume that each unit of the final products requires one unit of the input resource. The supplier sets the price q for the resource to maximize his profit. For simplicity, we normalize the production costs of the supplier to zero. To produce one output unit causes costs ci, i=1,2, in each division. The production costs of division 1

are stochastic and uniformly distributed in the interval [ c1 ,c1 ] . The distribution is assumed to be

common knowledge. The production costs c2 are deterministic and known by all players. The production costs ci include all manufacturing costs of one output unit except the costs of the input resource delivered by the supplier. So the total product costs per unit are ci+q. Each divisional manager receives a cost budget for production. The budget is fixed at the beginning of the period and is specified per unit. The budgeting process may be either designed top down or participative. When implementing top down budgeting, the principal only uses his ex ante knowledge of the production costs without asking the managers to report. Therefore, the principal fixes the cost budget based on his expectation of the stochastic costs in division 1, that is E(c1), and his knowledge of the costs of division 2, c2. When implementing participative budgeting, the principal asks the divisional managers to report and fixes the cost budgets based on their respective cost reports cˆ i ( ci ) . To introduce asymmetric information, we assume that the manager of division 1 may acquire costless additional information and observes the true costs c1. This information is personal to manager 1. Neither manager 2 nor the principal may acquire the information ex ante. The following illustrates this idea: Given his expertise, manager 1 can thoroughly analyze the production process of division 1 generating the personal information. Manager 2 has no access to this information, because he has no detailed knowledge of division 1 and its production site. The principal cannot acquire the information due to a lack of technical expertise to analyze the production facility. Also ex post, the true costs of division 1 are not observable by the principal and bymanager 2. Manager 1 may consume resources not needed for production on the job. It is not possible to separate real costs of production and slack consumed when observing actual production costs. This represents the general idea of slack in literature.9 Manager 1 has an incentive to overstate costs, to build slack and to consume slack privately. As true costs are never observable by all other players, there is no risk of ex post detection. The consumption of slack increases the utility of the manager. We assume that the amount of slack equals the utility arising from consuming slack. To build slack, the manager has to overstate real costs observed and to bias the cost report to the principal. We assume that the biasing behavior follows the linear function cˆ1( c1 ) = b ⋅ c1 ,10 with b > 1. If the report is unbiased, then b=1. If the manager reports truthfully, he receives no slack and his corresponding utility is zero. This represents the lower bound of a report of a rational manager. Whenever b>1, the manager overstates real costs. He builds slack and his utility is positive. The objective of manager 1 is assumed to be maximizing slack.

9

See e.g. Schiff/Lewin (1970). Linear biasing functions are widely accepted in the literature. See e.g. Ewert/Wagenhofer (2012), 11ff, for important arguments in favour of this assumption. A linear function, which additionally considers a fixed term, e.g. cˆ1 ( c1 ) = β + b ⋅ c1 , would not change the qualitative results of the analysis. 10

6

Following this objective, manager 1 fixes b in a way that given his observation of the true costs c1 the difference between the biased and unbiased cost budget is maximized. Hereby, manager 1 has also to consider his production choice of output. As the demand function is common knowledge, the quantity x1 has to be in line with the reported costs. Otherwise, the biased cost report is revealed. Additionally, the manager has also to consider the impact of his decision on the price q of the input resource set by the supplier. So division manager 1 maximizes slack s by maximizing the following expression with respect to b: s = cˆ1( b,c1 ) ⋅ x1( q( cˆ1( b,c1 )),cˆ1( b,c1 )) − c1 ⋅ x1( q( cˆ1( b,c1 )),cˆ1( b,c1 )) It is common knowledge that the reporting strategy of the manager is linear in b. However, the manager’s choice of b itself is assumed to be not observable. The manager observes the true costs c1 before making the report. So he can use this information when fixing the optimal bias level b to maximize slack. To simplify the analysis, we assume that the costs of division 2 are deterministic. Its manager has no opportunity of building up slack by biasing reports and cˆ 2 ( c2 ) = c2 . The principal faces the typical benefits and costs of participative budgeting described in the literature:11 The budget may build on the better information of managers involved in the budgeting process. Though, this comes at the costs of a possible biased information and slack building behavior. Hence, the principal has to decide whether to use top down budgeting not making use of the information of the better informed managers or implementing participative budgeting. Here, the principal runs the risk of bearing the costs of slack. The budgeting literature sees slack as the major obstacle of using participative budgeting. The time line of events of the setting analyzed is as follows: t=1: The principal decides whether to use top down or participative budgeting. t=2: If top down budgeting is used: the principal fixes the cost budgets per unit for each division based on his expectation of the costs. If participative budgeting is implemented: The manager of division 1 observes the true production costs c1 and reports costs per unit of cˆ1( c1 ) . Manager 2 reports costs cˆ 2 ( c2 ) = c2 . The principal provides a cost budget based on these reports. t=3: The supplier fixes the price q for the input resource. t=4: If top down budgeting is used: The principal fixes production quantity and prices for both divisions. If participative budgeting is implemented: The divisional managers fix production quantities and prices for their respective product. The products are sold and profits realize for both budgeting regimes.

11

See e.g. Atkinson/Kaplan/Matsumura/Young (2007), p. 489f.

7

The problem to find the most beneficial budgeting regime for the principal has to be solved by backward induction. The solution concept is SPNE. For convenience, we separate the analysis of the budgeting regimes to give a better illustration of each case. At the end, we compare the expected profits of each regime and argue which one is more beneficial for the principal.

Top down budgeting: At the final stage (t=4) the principal fixes production quantities of both divisions to maximize total expected firm profit π, which is the sum of the expected divisional profits π1 and π2. This decision is based on the expected production costs of division 1, E(c1) and contingent upon the price of the input resource to be set by the supplier at stage t=3. As the distribution of c1 is common knowledge, the supplier sets the price for the resource to be delivered to the firm based on E(c1) and the reaction functions of stage 4 (quantity decisions of the principal). Stage 2 has only to be considered in the participative budgeting setting. So we proceed to stage 1. Given the price of the input resource, the equilibrium output quantities and corresponding expected profits can be calculated if top down budgeting is used.

Observation 1: If the principal implements top down budgeting, the equilibrium output quantities are12 • •

1 * x1*,TD ( qTD ) = ⋅ ( 3a1 − a2 − 3E( c1 ) + c2 ) for division 1 and 8 1 * x*2 ,TD ( qTD ) = ⋅ ( 3a2 − a1 − 3c2 + E( c1 )) for division 2. 8

The supplier fixes the price for the input resource to be delivered to the firm with •

* qTD =

1 ⋅ ( a1 + a2 − E( c1 ) − c2 ) . 4

The firm expects to earn a profit of •

* * * * * * π TD ( x1*,TD ( qTD ),x*2 ,TD ( qTD ),qTD ) = x1*,TD ( qTD )2 + x*2 ,TD ( qTD )2 ,

where the first term in the sum is the expected profit provided by division 1 and the second term the expected profit earned by division 2.13

Observation 1 describes the benchmark solution for the principal. It states the expected profit of the firm achievable without participation of the divisional managers in the budgeting process. Participa-

12

The index “*” characterizes the equilibrium solution, the index TD the top down budgeting regime. It is highlighted that this is the expected profit of the principal at stage 1. To improve readability, the operator E(.) is suppressed here and later on. 13

8

tive budgeting is therefore only beneficial for the principal if the expected profit is at least as high as when top down budgeting is used. Participative Budgeting: Compared to top down budgeting, we have to consider additionally stage t=2 of the time line. Now the principal asks the managers for reports. Manager 1 observes the true costs c1 of producing one unit of output x1 in division 1 before reporting. He uses this information for preparing the cost report

cˆ1( c1 ) . Given the cost information, he decides whether and to what extent to overstate the true costs in the report. This is reflected by his objective of maximizing slack when choosing the bias level b. Contrary to the top down budgeting setting, the divisional managers make the production quantity decision at stage 4. Again the equilibrium solution is derived by backward induction. At the last stage, the divisional managers make their output quantity decisions conditional upon the price of the input resource set by the supplier at stage 3 and upon the cost reports cˆ1 and cˆ 2 made by the managers at stage 2. Given the reaction functions of stage 4, the supplier fixes the price for the input resource at stage 3 conditional depending on the cost reports cˆ1 and cˆ 2 of stage 2. Given the reaction functions of stage 3 and 4 the managers chose the cost reports at stage 2. As the true costs of division 2 are common knowledge, we simplify the following notation by setting cˆ 2 =c2. Manager 1 privately observes the true costs c1 and selects the bias level b to maximize slack. Based on b he reports costs cˆ1 = b ⋅ c1 . Given all decisions of stages 2 to 4, the equilibrium quantities, equilibrium price of the input factor, the equilibrium bias level and the resulting expected profit for the principal can be derived. The resulting equilibrium solution for the participative budgeting mechanism is described in observation 2.

Observation 2: If the principal implements participative budgeting the equilibrium output quantities are14 • •

1 ⋅ ( 3a1 − a2 − 3c1 + c2 ) in division 1 and 16 1 x*2 ,PB ( q*PB ( cˆ1* )) = ⋅ ( 17 a2 − 3a1 − 17c2 + 3c1 ) in division 2. 48 x1*,PB ( q*PB ( cˆ1* )) =

The supplier charges an equilibrium price for the input resource of •

q*PB ( cˆ1* ) =

1 ⋅ ( 3a1 + 7 a2 − 3c1 − 7c2 ) . 24

The slack maximizing bias for the manager and the resulting reported costs in equilibrium are •

14

b* ( c1 ) =

3a1 − a2 + 3c1 + c2 1 and cˆ1* ( b* ,c1 ) = ⋅ ( 3a1 − a2 + 3c1 + c2 ) . 6 6c1

The index “PB” characterizes the participative budgeting regime.

9

The resulting expected profit of the firm is •

π *PB ( x1*,PB ( q*PB ( cˆ1* ),x*2 ,PB ( q*PB ( cˆ1* )) = x1*,PB ( q*PB ( cˆ1* ))2 + x*2 ,PB ( q*PB ( cˆ1* ))2 , where the first term in the sum is the expected profit provided by division 1 and the second term the expected profit earned by division 2.15

A closer look on the characteristics of b*(c1) at stage 2 shows the following: For every feasible solution the manager overstates true costs, that is b*(c1)>1 is true for all x1,PB >0.

1 3 the function b*(c1) shows that b*(c1)=1 if x1,PB =0 and b*(c1)>1 if x1,PB >0. So the manager always

For x1,PB >0 to be true, the condition c1 < ( 3a1 − a2 + c2 ) must hold. Inserting the condition into

overstates the true costs in his cost report cˆ1 if production in division 1 is profitable and output is nonnegative. Inspection of the first derivative of b*(c1) with respect to c1 shows that b*(c1) is a concave function. b*(c1) is decreasing in c1 if x1,PB >0. This demonstrates that the lower the true costs of production are the more the manager overstates the reported costs. The resulting cost report cˆ1 is linear increasing in the costs c1: b* ( c1 ) ⋅ c1 =

1 ⋅ ( 3a1 − a2 + 3c1 + c2 ) . 6

The following graphs illustrate these findings based on a numerical example (for a1=110, a2=170 and c2=50). Given this parameter setting, x1 is positive if c1 π TD ( x1*,TD ( qTD ),x*2 ,TD ( qTD ),qTD )

and vice versa.

Proposition 1 Expected profit of participative budgeting exceeds expected profit of top down budgeting if the following condition holds: 11

1 1 ⋅ ( 45a1 − 31a2 + 31c2 ) ≤ E( c1 ) ≤ ( 3a1 − a2 + c2 ) 16 45 3 Otherwise the principal prefers top down budgeting. Proposition 1 states that the principal prefers bottom up budgeting if the expected production costs E(c1) are in an intermediate range. For low and high expected production costs, the principal prefers top down budgeting. The total difference of expected profits between participative budgeting and top down budgeting is given by: * π *PB − π TD =−

1 ⋅ [( 45 ⋅ ( a1 − E( c1 )) − 31 ⋅ ( a2 − c2 )) ⋅ ( 3 ⋅ (( a1 − E( c1 )) − ( a2 − c2 ))] . 1152

This difference is positive if the two multipliers in the brackets have different signs. This is only the case if E(c1) is between the bounds stated in proposition 1. If E(c1) is smaller than a1 −

31 ⋅ ( a2 − c2 ) , 45

then the first and the second multiplier in brackets are negative. As there is a negative sign at the beginning of the term, the entire difference is negative. If E(c1) is larger than a1 −

31 ⋅ ( a2 − c2 ) , but 45

1 3

smaller than a1 − ⋅ ( a2 − c2 ) , then the first term in brackets is positive, the second negative. In total, due to the negative sign at the beginning, the difference is positive. If E(c1) is larger than

1 a1 − ⋅ ( a2 − c2 ) , both terms in brackets become positive and the total difference becomes nega3 tive. Thus, the difference in expected profits is only positive, whenever E(c1) is between the bounds shown in proposition 1. In this case, participative budgeting is more beneficial than top down budgeting. The economic intuition of this finding is as follows: To switch from top town to participative budgeting, the profitability of the market, in which division 2 operates, must reach a minimum level represented by the lower bound. The lower bound can be rewritten as

31 ( a2 − c2 ) > ( a1 − E( c1 )) . The 45

benefit of participative budgeting is to increase product costs in division 1 to influence the output of division 2. Inspection of the equilibrium output functions stated in observation 2 shows that increasing costs of product 1 lead to decreasing output of product 1, but also to an increasing output of product 2. This effect is only beneficial for the principal if the relative profitability of product market 2 compared to product market 1 is large enough. However, there is also a floor on inflating the costs of product 1. If the relative profitability of product market 1 gets too small [i.e.

a2 − c2 > 3( a1 − E( c1 )) ], then it is not profitable to produce product 1 anymore. In this case, the shift in production quantities stops and there is no further benefit of increasing the product costs of division 1. This defines the upper bound stated in proposition 1. The selling price of the input resource is a major driving force of the trade-off described above. This is discussed later. Exhibit 3 highlights the finding of proposition 1. It shows that for low values of E(c1), top town budgeting is preferred. Reaching the lower bound the decision of the principal switches to participative 16

The lower bound is smaller than the upper bound if c2