Abstract
This paper examines the principal’s preferences over reporting systems in multi-agent settings. In multi-agent settings, the principal’s contract offer depends on the credibility of the agents’ commitment not to collude on the terms of the contract. If the agents can credibly rule out collusion, then the principal prefers that the agents observe all performance measures and she prefers an accounting system that releases detailed instead of aggregated information. To the contrary, when the principal cannot preclude the agents from writing side contracts, it can be efficient to suppress certain information signals for contracting purposes. Specifically, restricting the information dissemination within organizations or releasing aggregated instead of detailed information are two efficient avenues to suppress information.
Similar content being viewed by others
Notes
Such barriers (Chinese walls) are widespread in investment banks and accountancy firms. Typically, they are complemented by nondisclosure agreements that aim at restricting the dissemination of sensitive information.
Simons (1992, p. 4), describes the reporting system at Asea Brown Boveri as follows:
Access to reports and the ad hoc database was restricted by authority and responsibility levels. Company managers, for example, had access to data for their own company and the profit centers within the company; they did not have access to data for other companies. Business Area managers had access to data on all the Business Area Units within their Business Area, but did not have access to other Business Areas. Executive vice presidents had access in detail to the Business Units, Business Areas, and countries (i.e., companies) for which they were responsible; in addition, each had access to broader corporate results.
Analysis of multi-agent settings using the LEN-framework is also performed by Holmström and Milgrom (1990), Itoh (1992), Ziv (2000), Baldenius et al. (2002), Huddart and Liang (2005), and Rajan and Reichelstein (2006). In general, optimal contracts are not linear. We do not explore optimal contracts in our setting but assume the results are relatively robust for the setting we are considering. However, as Hemmer (2004) notes, to prove the robustness of results one needs to analyze the unrestricted program.
As Feltham and Hofmann (2007) point out, in addition to the usual simplifications provided by the LEN-model in single-agent settings, it provides another significant simplification in the multi-agent setting. As we note later, in the LEN-model, there is a unique Nash equilibrium in the subgame played by the agents after the principal has offered them their contracts.
Inter-agent negotiations can be beneficial or costly to the principal. While such negotiations are often referred to as collusion if there is a negative effect, Holmström and Milgrom (1990), Itoh (1992), and Villadsen (1995), among others, refer to them as cooperation if it has a positive effect on the principal’s expected payoff. Our analysis differs from these studies by assuming that the agents do not mutually observe each others’ effort, which implies that they act non-cooperatively. Consequently, inter-agent negotiations are (weakly) costly to the principal. We refer to agent collusion also as limited commitment in the form of the agents’ inability to commit to not write side contracts.
A comprehensive approach explicitly models the long-term relationship between the agents to capture the reputational phenomena needed such that side contracts are enforceable (Tirole 1992; Martimort 1997). According to Laffont and Martimort (1997b), the consequences of these long-term relations can be described within a framework allowing side contracts to be enforceable. To simplify the analysis, similar to Tirole (1986) and Holmström and Milgrom (1990), among others, we assume that the agents can sign an enforceable side contract.
Firms such as Manco Inc. (Aggarwal and Simkins 2001) or Springfield ReManufacturing Corporation (Case 1997) employ such a concept of “open-book management.” According to Case (1995, p. 37), “Every employee in an open-book company sees—and learns to understand—the company’s financials, along with all the other numbers that are critical to tracking the business’s performance.”
While the agents have little incentives to change these contracts under partial information dissemination, such incentive schemes are not negotiation-proof under global reporting. There, the agents also share information on the performance of the agent with a high-powered incentive scheme and side-contracting on both measures is beneficial to the agents.
Our argument is similar to Demski et al. (1999) and Laffont and Martimort (1999). They argue that introducing bureaucratic frictions like organizational barriers or separated communication channels induces transaction costs that can keep agents from breaching information security and from writing side contracts on information not shared by the agents.
A common example for negative correlation refers to the allocation of expenses to two divisions that are subject to measurement error. Aggregation of divisional profits allows the measurement error to cancel out.
Additionally, restricting an agent’s access to information can be motivated, for example, by the adverse effect of information on his decisions (Christensen 1981), by information overload considerations (Galbraith 1973; Geanakoplos and Milgrom 1991), or by the firm’s communication cost (Marschak and Reichelstein 1998).
For simplicity, and for emphasis, we assume that the agents are identical with respect to their payoff productivity.
The scales of the performance measures are arbitrary. To simplify notation, we scale y j so that they have unit variance. For emphasis, agents are identical with respect to the effect of their actions on the performance measures.
Superscript “†” denotes results under full commitment.
Contract negotiations after the agents have taken their actions are similar to ex post contract negotiations between the principal and an agent (Fudenberg and Tirole 1990). Then, the agents will merely focus on efficiently sharing the incentive risk, and the fixed payment will reflect the conjectured actions.
Expression (8) indicates a Pareto improvement among the coalition of two agents, ignoring the principal’s payoff.
The second condition implies, e.g., that even for uncorrelated performance measures, if the principal were to offer the pair of optimal full-commitment contracts, with limited commitment, the agents will still want to collude to better allocate risk among each other.
In our setting, v sj trades off benefits from risk sharing and incentive allocation. This is different from, e.g., Holmström and Milgrom (1990) and Ramakrishnan and Thakor (1991), where the agents mutually observe each others’ efforts. There, the agents can implement any effort allocation using, e.g., a forcing contract, and select incentive rates for the side contract that minimize their joint cost of risk, taking the principal’s pair of contracts as given.
The superscript “‡” denotes results under limited commitment.
If the principal selects v jl according to (10), then v e sj = − v ‡ lj and v r sj = ½ m 2/r v ‡ lj , implying sgn[v e sj ] ≠ sgn[v r sj ], j,l = 1,2, and j ≠ l.
Some intuition for this result is provided below after having compared global reporting with partial information dissemination in Proposition 1.
Interestingly, even with perfect correlation, agent collusion results in an efficiency loss to the principal. Given perfect correlation, ρ = ±1, while the principal can fully insure the agents against incentive risk (by setting v jl = − sgn[ρ] v ll ), these incentive rates are not negotiation-proof. Providing insurance with negative (positive) correlation calls for a positive (negative) incentive rate for the other agent’s performance measure, which induces a positive (negative) externality in the agent’s compensation and creates incentives for the other agent to supply more (less) effort.
For notational brevity, we do not explicitly emphasize that incentive rates relate to a specific reporting system.
In the context of information dissemination in hierarchies, we may refer to agent 1 as the senior agent and agent 2 as the junior agent. However, besides the assumed information dissemination, neither agent is endowed with any particular skill or additional information that would induce a strict preference of the principal regarding the agents’ hierarchy.
Note that the lower and upper bounds are not symmetric (specifically, |ρ l sp | < ρ u sp ). While the benefit from providing insurance is symmetric, the cost to collusion is not. Given ρ > 0, agent 1’s incentive rate for agent 2’s performance measure is negative, creating a negative externality, such that agent 2 is induced to supply less effort. This reverses for ρ < 0.
In our setting, unit weights are not a simplifying assumption. Specifically, endogenizing the choice of (relative) weights reveals that unit weights are optimal for the principal. Following Feltham and Hofmann (2007), Proposition 3, with identical agents, limited commitment is not detrimental to the principal compared with full commitment. Specifically, unit weights imply that the agents’ efforts have the same influence on the aggregate performance measure. To the contrary, non-unit weights create incentives for side-contracting and, while they may yield a more precise aggregate performance measure as compared with unit weights, the impact on each agent’s signal-to-noise ratio is ambiguous. Thus, unit weights remove any incentives for agent side-contracting, and the resulting benefits exceed any benefits that may accrue from improvements in the agents’ signal-to-noise ratios.
Conceptually, while segment reporting requires revenues and expenses to be allocated to divisional performance measures, this is not necessary with aggregate reporting. For simplicity, we do not explicitly consider measurement errors (e.g., Arya, Fellingham, and Schroeder 2004). Such errors are included in the analysis, e.g., by considering performance measures y 1 = m a 1 + ε 1—ε a and y 2 = m a 2 + ε 2 + ε a , where ε a ~ N(0,σ 2 a ) represents measurement error. While ε a increases the relative advantage of aggregate reporting over segment reporting, our main inferences remain unchanged.
More specifically, under limited commitment, the principal’s decision problem reduces to specifying incentive rates for an aggregate performance measure. With aggregate reporting, this measure is y a = y 1 + y 2, and with segment reporting, agent j’s “total” performance measure is \(y_{ja}^\prime\) = y j + v ‡ jl /v ‡ jj y l , j,l = 1,2, and j ≠ l, where v ‡ jl /v ‡ jj follows from collusion-proofness constraint (10). To the contrary, under full commitment, agent j’s total performance measure is \( y_{ja}^{\prime\prime}\) = y j + v † jl /v † jj y l , j,l = 1,2 and j ≠ l, where v † jl /v † jj = − ρ follows from (7a). For very negative correlation, the relative weights for y 1 and y 2 implied by aggregate reporting (i.e., “1”) are “closer” to the optimal relative weights under full commitment than the relative weights that are implied by segment reporting (i.e., v ‡ jl /v ‡ jj = [m 2 + r (1 − ρ 2)]−1 r (1 − ρ) according to (10)). This relation reverses if the correlation coefficient is not sufficiently negative.
Corollary 2 is related to Holmström and Milgrom (1990), Proposition 5, Ramakrishnan and Thakor (1991), Theorem 7, and Itoh (1992), Proposition 6. They derive conditions for when competition (via RPE, our global reporting under full commitment) is preferred to cooperation (where agents mutually observe each others’ efforts and may side-contract on this information). In their settings, competition is preferred if the correlation coefficient is sufficiently large.
References
Aggarwal, R., & Simkins, B. J. (2001). Open book management - optimizing human capital. Business Horizons, 44, 5–13.
Antle, R., & Demski, J. S. (1988). The controllability principle in responsibility accounting. The Accounting Review, 63, 700–718.
Arya, A., Fellingham, J. C., & Schroeder, D. A. (2004a). Aggregation and measurement errors in performance evaluation. Journal of Management Accounting Research, 16, 93–105.
Arya, A., Glover, J., & Liang, P. J. (2004b). Intertemporal aggregation and incentives. European Accounting Review, 13, 643–657.
Arya, A., Glover, J., & Mittendorf, B. (2006). Hierarchical reporting, aggregation, and information cascades. Managerial and Decision Economics, 27, 355–362.
Baldenius, T., Melumad, N. D., & Ziv, A. (2002). Monitoring in multiagent organizations. Contemporary Accounting Research, 19, 483–511.
Case, J. (1995). Open-book management. New York: HarperCollins.
Case, J. (1997). Opening the books. Harvard Business Review, 75, 118–127.
Christensen, J. (1981). Communication in agencies. Bell Journal of Economics, 12, 661–674.
Demski, J. S. (2008). Managerial uses of accounting information (2nd ed.). New York: Springer.
Demski, J. S., Lewis, T. R., Yao, D., & Yildirim, H. (1999). Practices for managing information flows within organizations. The Journal of Law, Economics, & Organization, 15, 107–131.
Dequiedt, V., & Martimort, D. (2009). Mechanism design with bilateral contracting. Working paper, CERDI. http://idei.fr/display.php?a=6274.
Feltham, G. A., & Hofmann, C. (2007). Limited commitment in multi-agent contracting. Contemporary Accounting Research, 24, 345–375.
Fudenberg, D., & Tirole, J. (1990). Moral hazard and renegotiation in agency contracts. Econometrica, 58, 1279–1319.
Galbraith, J. (1973). Designing complex organizations. Boston: Addison-Wesley.
Geanakoplos, J., & Milgrom, P. (1991). A theory of hierarchies based on limited managerial attention. Journal of the Japanese and International Economics, 5, 205–225.
Gigler, F., & Hemmer, T. (2002). Informational costs and benefits of creating separately identifiable operating segments. Journal of Accounting and Economics, 33, 69–90.
Hemmer, T. (2004). Lessons lost in linearity: A critical assessment of the general usefulness of LEN models in compensation research. Journal of Management Accounting Research, 16, 149–162.
Hodge, B. J., Anthony, W. P., & Gales, L. M. (2003). Organization theory (6th ed.). Upper Saddle River: Pearson.
Holmström, B., & Milgrom, P. (1990). Regulating trade among agents. Journal of Institutional and Theoretical Economics, 146, 85–105.
Huddart, S., & Liang, P. J. (2005). Profit sharing and monitoring in partnerships. Journal of Accounting and Economics, 40, 153–187.
Indjejikian, R., & Nanda, D. (1999). Dynamic incentives and responsibility accounting. Journal of Accounting and Economics, 27, 177–201.
Ishiguro, S. (2004). Collusion and discrimination in organizations. Journal of Economic Theory, 116, 357–369.
Itoh, H. (1992). Cooperation in hierarchical organizations: An incentive perspective. Journal of Law Economics and Organization, 8, 321–345.
Laffont, J. J., & Martimort, D. (1997a). Collusion under asymmetric information. Econometrica, 65, 875–911.
Laffont, J. J., & Martimort, D. (1997b). The firm as a multicontract organization. Journal of Economics & Management Strategy, 6, 201–234.
Laffont, J. J., & Martimort, D. (1999). Separation of regulators against collusive behavior. RAND Journal of Economics, 30, 232–262.
Marschak, T., & Reichelstein, S. (1998). Network mechanisms, informational efficiency, and hierarchies. Journal of Economic Theory, 79, 106–141.
Martimort, D. (1997). A theory of bureaucratization based on reciprocity and collusive behavior. Scandinavian Journal of Economics, 99, 555–579.
Mookherjee, D. (1984). Optimal incentive schemes with many agents. Review of Economic Studies, 51, 433–446.
Mookherjee, D. (2006). Decentralization, hierarchies, and incentives: A mechanism design perspective. Journal of Economic Literature, 44, 367–390.
Nikias, A. D., Schwartz, S., & Young, R. A. (2005). Optimal performance measures with task complementarity. Journal of Management Accounting Research, 17, 53–73.
Rajan, M. V., & Reichelstein, S. (2006). Subjective performance indicators and discretionary bonus pools. Journal of Accounting Research, 44, 585–618.
Ramakrishnan, R. T. S., & Thakor, A. V. (1991). Cooperation versus competition in agency. Journal of Law Economics and Organization, 7, 248–283.
Simons, R. (1992). Asea Brown Boveri: The ABACUS system. Harvard Business School Case, 9-192-140.
Simons, R. (1995). Levers of control. Boston: Harvard Business School Press.
Tirole, J. (1986). Hierarchies and bureaucracies: On the role of collusion in organizations. Journal of Law Economics and Organization, 2, 181–214.
Tirole, J. (1992). Collusion and the theory of organizations. In J.J. Laffont (Ed.), Advances in economic theory 2 (pp. 151–206). Cambridge: Cambridge University Press.
Varian, H. R. (1990). Monitoring agents with other agents. Journal of Institutional and Theoretical Economics, 146, 153–174.
Villadsen, B. (1995). Communication and delegation in collusive agencies. Journal of Accounting and Economics, 19, 315–344.
Wilson, R. (1968). The theory of syndicates. Econometrica, 36, 119–132.
Ziv, A. (2000). Information technology and optimal firm structure. Journal of Accounting Research, 38, 297–328.
Acknowledgements
We are grateful to Shane Dikolli, Jon Glover, Stefan Reichelstein (the editor), Stefan Wielenberg, two anonymous reviewers, and workshop participants at the University of Tübingen, the 2006 MAS Conference, the 2006 Workshop on Accounting and Economics, Operations Research 2007, and the 2007 German Economic Association of Business Administration symposium for many helpful suggestions. Financial support has been provided by the Social Sciences and Humanities Research Council of Canada. An earlier version of this paper was entitled: “Impact of Alternative Reporting Systems in Multi-agent Contracting.”
Author information
Authors and Affiliations
Corresponding author
Appendix A
Appendix A
1.1 Proof of propositions
Lemma 1
Substituting (7b) and (11b) into Π †(η g) − Π ‡(η g) and simplifying, yields
Differentiating (11b) with respect to ρ and setting ρ = 0 yields
Proposition 1
Substituting (11b) and (13b) into Π ‡(η g) > Π ‡(η s), rearranging, and simplifying yields
The left hand side of (A.1) is equal to zero for ρ ∈ {1, ρ gs , (2 m 2 + r)/(4r) + [½ + (2 m 2 + r)2/(4r)2]½}, with ρ gs = (2 m 2 + r)/(4r) − [½ + (2 m 2 + r)2/(4r)2]½, where the latter two elements are the roots to −r (1 − ρ)(1 + 2ρ) − 2 m 2 ρ = 0. The upper root is larger than 1. Also, Π ‡(η g) = Π ‡(η s) if ρ = 1, ∂(Π ‡(η g) − Π ‡(η s))/∂ρ| ρ=1 > 0, and \(\partial({\varPi ^{\ddagger}}(\eta^{g}) - {\varPi ^{\ddagger}} (\eta^{s}))/\partial\rho\vert_{\rho = \rho_{gs}} < 0.\) Since ρ ∈ [−1,1], strict preference for η g requires ρ < ρ gs . Inspection of ρ gs reveals that ρ gs ∈ (−½,0). The preference reverses for ρ > ρ gs .
Lemma 2
Substituting (17a) and (17c) into v ‡11 − v ‡22 and simplifying, yields
since D > 0.
Using (17b), v ‡12 < 0 if
Observe that, with ρ ∈ [−1,1], the expression in brackets of the left hand side in (A.2) is positive. Thus, (A.2) holds only if ρ > 0. For ρ > 0, the left hand side of (A.2) increases monotonically in ρ, and the condition is surely satisfied for ρ = 1. Hence, there exists a single ρ ∈ (0,1) such that (A.2) is true for ρ > ρ.
Proposition 2
Substituting (13b) and (17d) into Π ‡(η p) > Π ‡(η s) and simplifying, yields
Note that for (A.3) to hold true, the absolute value of ρ must exceed a certain threshold, which is given by ρ t sp , t = l,u, where ρ t sp are the roots that solve (A.3) as an equality. In particular, (A.3) does not hold if ρ = 0. Given uncorrelated reports, the principal strictly prefers single-agent reporting to partial information dissemination. Inspection of ρ t sp , t = l,u reveals that ρ l sp ∈ (−1,0) and ρ u sp ∈ (0,1).
Corollary 1
Using Propositions 1 and 2, the principal strictly prefers partial information dissemination for ρ ∈ (ρ u sp ,1], and she strictly prefers single-agent reporting for ρ ∈ (max{ρ gs ,ρ l sp },ρ u sp ). Comparing ρ l sp with ρ gs reveals ρ l sp < ρ gs . Finally, \( \Pi ^{\ddagger}(\eta^{g})\vert_{ \rho = \rho_{gs}}> \Pi^{\ddagger}(\eta^{p})\vert_{ \rho = \rho_{gs}}\,and\,\partial\varPi ^{\ddagger}(\eta^{g}\vert \rho)/\partial\rho > \partial\varPi^{\ddagger}(\eta^{p}\vert \rho)/\partial\rho\) for ρ < 0. Hence, for ρ < ρ gs the principal has strict preferences for global reporting.
Proposition 3
Substituting (11b) and (18b) into Π ‡(η a) > Π ‡(η g) and simplifying, yields
(A.4) is satisfied as an equality for
Inspection of the two roots reveals that the upper root exceeds 1. While (A.4) does not hold for ρ ∈ [0,1], for ρ = −1, (A.4) simplifies to
which is always true. Hence, Π ‡(η a) > Π ‡(η g) if ρ < ρ ag , where ρ ag is the lower root in (A.5). Inspection of ρ ag reveals that ρ ag ∈ (−1, − ½). Finally, the preference reverses for ρ > ρ ag .
Corollary 2
Extending Corollary 1, it is straightforward to show that ρ ag < ρ gs . Substituting the two closed-form solutions for ρ ag and ρ gs and simplifying completes the proof.
Rights and permissions
About this article
Cite this article
Feltham, G.A., Hofmann, C. Information suppression in multi-agent contracting. Rev Account Stud 17, 254–278 (2012). https://doi.org/10.1007/s11142-011-9175-2
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11142-011-9175-2
Keywords
- Incentives
- Inter-agent negotiations
- Multi-agent contracting
- Performance evaluation
- Relative performance evaluation