نگهداری در مدیریت پرتفولیو واگذار شده: چه زمانی اطلاعات عملکرد مقایسه ای مطلوب است؟

We address the issue of investors’ asset allocation decisions when portfolio management is delegated to an agent. Contrary to predictions from traditional financial theory, it is shown that investors may not induce their manager to allocate funds to the asset with the highest return. Instead they may herd in their asset allocation decision and induce trade in a particular asset, because another manager is trading in it and despite the presence of a more profitable alternative. Doing so allows investors to write an efficiency-improving relative-performance contract. On the other hand, herding leads investors to design wage contracts strategically, resulting in more aggressive and thus less profitable trade in equilibrium. We show that herding occurs, when the cost of information is high, information precision is low and when managers are sufficiently risk averse. Moreover, when investors can decide whether or not to disclose information about their manager's performance, they will not do so.

A large fraction of individuals’ savings is managed and invested by professionals working for financial institutions and acting on behalf of those individuals. The type of institution ranges from pension and mutual funds over hedge funds to insurance companies. Allen and Gale (2000) report that pension funds and other financial institutions constitute 61% of all shareholders in the UK in 1995. In the US institutional investors held US$ 8.0 trillion in US equities at the end of 2000, which amounts to 45.8% of the total. This compares to 28.2% at the end of 1970 and 7.2% at the end of 1950.1 Delegating portfolio management to an agent creates conflicts of interest between the investor and his agent. This paper explores the effects of these conflicts on the asset allocation decision, trading behaviour and price formation in financial markets. The main finding is that the asset allocation decision affects the way in which agency problems can be resolved and that, as a result, investors may be inclined to herd on a particular asset or asset class, foregoing investment opportunities that would yield higher returns. It is well known from contract theory (Holmstrom, 1982; Mookherjee, 1984) that contracting on relative rather than absolute performance may mitigate agency problems. But such contracts require comparative performance information. Investors may therefore prefer to allocate funds to assets or markets in which other active managers are also present. The resulting herding of informed traders in some assets while neglecting others, leads to highly informative prices in some assets and uninformative prices in others. We discuss the optimal design of compensation contracts taking into account that they serve not only to mitigate agency problems, but also affect financial market competition. Accounting for this provides conditions under which it is desirable to tie the manager's compensation to relative performance. The setting is one of double moral hazard: it is costly for managers to learn about the future value of an asset, and managers maintain discretion over the trading strategy used to exploit this private information. Principals choose whether to invest in an environment where comparative performance information is available, by assigning their manager to an asset class dependent on whether another manager also trades in that class. Relative performance-based contracts may be desirable, because they provide incentives to acquire information at lower risk exposure to the manager compared to absolute performance contracts. Moreover, the manager tends to compensate for risk exposure by taking less aggressive trading positions. This further agency cost can be reduced by relative performance contracts. In our model these benefits of relative performance contracts carry two costs.2 Firstly, relative performance contracts can only mitigate the agency problem if managers who are compared to one another also trade in the same asset, so that their returns are correlated. Therefore, managers lose market power over their private information, and thus active management is less profitable. Secondly, it is known from the Industrial Organisation literature that firms engaging in Cournot competition have an incentive to design managerial compensation contracts strategically. When one firm rewards aggressive behaviour of its manager by basing wage on profits and sales, the manager acts like a Stackelberg leader over the other manager whose best response is to reduce output ( Vickers, 1985; Fershtman and Judd, 1987; Sklivas, 1987). We show that in a similar way relative performance contracts can be used to induce more aggressive trading behaviour, which gives rise to strategic interaction in financial markets. Investors face a prisoner's dilemma in that they would be better off inducing less aggressive trade, but in equilibrium they provide relative performance contracts that induce overly aggressive trade compared to the profit maximising level. 3 Contracts are then designed to meet two objectives: mitigate the agency problem (the agency motive) and affect the nature of competition in financial markets (the strategic motive). In spite of the costs, relative performance contracts are desirable when managers’ private information is relatively poor, costly to produce or when managers are more strongly averse to risk. Moreover, when investors can choose whether or not to disclose information regarding their manager's performance, they choose not to do so in order to deny the competing fund manager the possibility of a relative performance-based contract. Doing so reduces the degree of competition between managers in equilibrium. The merit of relative performance-based contracts in delegated portfolio management and their impact on asset prices remains a poorly understood issue. Admati and Pfleiderer (1997) explore the use of benchmark portfolios in the compensation of privately informed fund managers and show that it is generally harmful and at best useless. In contrast to their treatment we consider compensation relative to the performance of another active manager, rather than a predetermined benchmark portfolio. This explains why relative performance compensation may be efficiency improving in our setting. Moreover, we consider a setting in which managers’ trades are sufficiently large to affect prices, which has important implications for performance contracts. Most of the literature on wage contracts in delegated portfolio management (Bhattacharya and Pfleiderer, 1985; Stoughton, 1993; Heinkel and Stoughton, 1994; Eichberger et al., 1999) assumes that asset prices follow an exogenous process, i.e. portfolio managers are modelled as price takers in a partial equilibrium setting. Admati and Pfleiderer (1997) identify a manager's ability to undo incentive contracts by adjusting portfolio weights as one reason for the ineffectiveness of such contracts. We show that when a fund manager's portfolio choice affects prices, incentive contracts are not undone completely and thus remain effective. Palomino (2000) provides a model of equilibrium in a financial market where traders are compensated on the basis of relative performance of the average competitor. Similar to our results, he finds that if traders have some market power, relative performance compensation strengthens competition among traders and trading intensities are higher compared to an equilibrium with absolute performance-based payment. In our paper the agency problem that may lead to relative performance compensation is modelled explicitly and wage contracts are derived, taking into account the competition effect. Palomino (2000) also addresses the possibility of herding, although he focuses on herding as a result of unobservable asset allocation by managers. In our setting asset allocation is contractible. Maug and Naik (1996) consider the question of herding in the asset allocation decision in the context of relative performance contracts. In their setting trading intensities are exogenous. As shown in this paper, endogenous trading intensities are particularly relevant once strategic interaction between several principals is taken into account. Other authors (e.g. Scharfstein and Stein, 1990; Trueman, 1994; Zwiebel, 1995) have shown that herding among agents who are evaluated relative to their peers might result due to reputational concerns. In this paper we neglect reputational concerns and focus instead on explicit incentives. Herding, however, remains an important issue, as agents’ explicit incentives are based on relative performance and hence one agent's actions do affect another agent's incentives. Herding in our treatment occurs in the sense that one principal induces acquisition of a piece of information, because another agent acquires the same piece of information (see also Brennan, 1990; Froot et al., 1992; Dow and Gorton, 1994). This is due to payoff externalities and can be modelled in a simultaneous moves setting. This contrasts with sequential herding where learning effects can generate informational cascades (see Banerjee, 1992; Bikhchandani et al., 1992; Welch, 1992; Vives, 1996, among others). In contrast to other treatments of herding, our results suggest that herding might not be such a bad thing after all. Herding is induced by investors in order to mitigate the inefficiencies associated with delegation, rather than an instance of inefficiency, as for example in Scharfstein and Stein (1990). Moreover, herding increases the efficiency of prices of the asset in which agents herd, rather than reducing it as in Froot et al. (1992). The paper proceeds as follows. In Section 2 the basic framework is developed. Section 3 derives equilibrium trading and price setting strategies as a function of the contracting parameters and the assets chosen by the principals. Section 4 derives the optimal linear incentive scheme under non-herding and herding, and illustrates the impact on equilibrium trading strategies. Section 5 contains the main results concerning herding. Section 6 endogenises the choice of disclosure of performance information. Section 7 concludes. The appendix provides the proofs.

This paper explores the contracting problem between an investor and a risk averse fund manager, showing that it can give rise to herding in investors’ asset allocation decisions. Wage contracts serve to mitigate two moral hazard problems: a fund manager needs to be given incentives to acquire costly information and he has discretion over exploiting this information by trading in a financial market. A contract that resolves the first problem must expose the manager to risk. The latter reacts by trading less aggressively and therefore does not fully exploit his private information. Assigning two managers to the same market makes comparative performance information available, which can improve the insurance–efficiency trade-off of contracts. The gain from writing better contracts may outweigh two disadvantages of using comparative performance information. Firstly, it is only available when managers trade in the same market, which means that they cease to be monopolists on their information. This reduces trading profits. Secondly, investors write relative performance contracts so as to induce aggressive trading behaviour for strategic reasons. This further stiffens competition and reduces trading profits. Investors nonetheless herd in their asset allocation decision, when the degree of managerial risk aversion is high, when the quality of private information is low or its cost is high. Furthermore, when investors can choose whether or not to disclose information about their manager's performance they choose not to do so. Compensation can then only be based on absolute performance, which reduces competition in financial markets. Future research may be able to develop further the regulatory implications of information disclosure.