Innovation, endogenous overinvestment, and incentive pay.

We analyze how two key managerial tasks interact: that of growing the business through creating new investment opportunities and that of providing accurate information about these opportunities in the corporate budgeting process. We show how this interaction endogenously biases managers toward overinvesting in their own projects. This bias is exacerbated if managers compete for limited resources in an internal capital market, which provides us with a novel theory of the boundaries of the firm. Finally, managers of more risky and less profitable divisions should obtain steeper incentives to facilitate efficient investment decisions.

1. Introduction

* That managers are excessively "hungry for capital" is a common notion found among both practitioners and scholars working on the capital budgeting process. (See, for instance, "Curing Capital Addiction," The McKinsey Quarterly, 1993 Number 4.) We show how managers who are expected to generate new growth opportunities will become endogenously biased toward overinvesting in their own projects. The source of this distortion is that managers are expected to both generate new projects and to, subsequently, feed information into the corporate budgeting process.

The tension between the two tasks creates a dilemma for corporations. The more managers are incentivized to grow their business, the more they become biased toward overspending. Reducing this bias by dampening incentives may be too costly for businesses that crucially depend on innovation and growth. (1) As illustrated in Jensen (2003), the inefficiencies caused by "lying" in the capital budgeting process may, however, also be substantial. Our analysis shows that the use of high-powered incentives may be key to mitigating this tension--in particular in corporations where multiple divisions must compete for limited resources.

In our model, the role of high-powered incentives is not to tease out incrementally more effort. Instead, their purpose is to reduce corporate overspending. The argument unfolds as follows. The division manager will only work hard at generating new investment opportunities if this increases his expected compensation. Once a new opportunity has been created, however, the "reward" that he was promised for growing the business makes the manager want to undertake even some unprofitable investments. The optimal compensation scheme seeks to minimize this bias. By tying the manager's compensation more closely to the division's profits, the manager has less to gain from convincing headquarters to invest in a relatively unpromising opportunity. As steep incentives also allow the manager to receive a larger share of the profits from a very promising investment, the total reward that he can expect from working on new investment opportunities remains unchanged. Hence, while preserving the manager's incentives to grow the business, steep incentive schemes reduce the potential for (over)investing in negative net present value (NPV) projects.

Some of our predictions link the steepness of managers' compensation to firm characteristics and investment decisions. We find that incentives should be steeper in divisions that require more capital injection, in divisions that look less promising to headquarters, and also in divisions that are more risky. (2)

If managers compete for scarce resources, this increases the tension between providing incentives to generate new investment opportunities and providing incentives to reveal accurate information about their profitability. If this is still feasible, the optimal response is then to further tighten the link between managers' pay and divisions' performance. The insight that competition in an internal capital market can lead to more biased information is shared with Ozbas (2005). This complements the analysis of Stein (2002), Brusco and Panunzi (2005), and Inderst and Laux (2005), who show how competition can adversely affect the incentives to generate information, cash flow, or investment opportunities. We also analyze the decision of when to create competition in an internal capital market, including through the integration of previously stand-alone businesses. Here, one of our results is that on average more investment is made in an internal capital market, although some of it may prove to be less profitable than the investment made in comparable stand-alone businesses.

The analysis of competition is also a key difference to a related paper by Levitt and Snyder (1997), which we discuss in more detail below. There, an agent can both increase a project's likelihood of success and provide information that may allow to prematurely cancel unprofitable projects. (3)

The way we endogenize managers' bias toward overinvesting in their own projects may also prove useful in different strands of the literature. That managers derive benefits from building larger empires is a central notion of numerous theories of corporate control and financial contracting that build on Jensen's (1986) free cash flow problem. Here, our approach provides an alternative to the use of nonpecuniary benefits.

The rest of this article is organized as follows. Section 2 introduces the model. Section 3 analyzes the case where divisions do not compete for scarce resources, and Sections 4 and 5 introduce competition. Section 6 concludes.

2. The model

*** The firm and its investment opportunities. We consider a firm that is run by headquarters in the interest of its risk-neutral owners. The firm must employ (specialized) division managers to run its individual businesses. There are three time periods: t = 0, 1, and 2. In t = 0, headquarters hires division managers. Once hired, managers can exert effort to generate new investment opportunities. Although the decision whether to undertake a new project lies with headquarters, when generating the project the respective division manager becomes better informed about its prospects. Hence, a division manager will have to perform the twin tasks of creating new investment opportunities in t = 0 and of subsequently guiding headquarters' investment decision in t = 1. In the final period, t = 2, payoffs are realized. We first analyze the case where divisions do not compete with each other. Section 4 considers the opposite case where only one project can be undertaken at a given time, which may follow as (organizational or financial) resources are scarce or as projects are close substitutes.

Division managers' effort in t = 0 involves private disutility c > 0. If headquarters wants to realize the generated new investment opportunity, it must invest capital k > 0. (4) We normalize the return from alternative investments to zero. If undertaken, the new project realizes positive cash flows of x > k with probability 0 < p [less than or equal to] 1 if it is of the "good type" [theta] = g, which a priori is the case with probability 0 < q < 1. In this case, the expected cash flow equals [mu] : = xp > k. If the project has a "bad type" [theta] = b, it realizes zero cash flows. After generating the project, in t = 1 only the respective manager can observe a noisy signal about the project's type. The signal s [member of] S = [[s.bar], [bar.s]] is generated from the distribution functions [F.sub.[theta]](s), which has no atoms and an everywhere continuous and strictly positive density [f.sub.[theta]](s). Because [f.sub.g](s)/[f.sub.b](s) is strictly increasing and satisfies the monotone likelihood ratio property, the posterior belief that the project is of the good type,

q(s) := q[f.sub.g](s)/q[f.sub.g](s) + (1 - q) [f.sub.b](s)],

is strictly increasing in s. We denote the conditional success probability by p(s) := q(s)p and the conditional expected cash flow by [mu](s) := xp(s). We assume that [mu](s) < k < [mu]([bar.s]), implying that there exists a unique cutoff [S.sub.FB] [member of] ([s.bar], [bar.s]) that satisfies [mu]([S.sub.FB]) = k. (5) Hence, it is first-best efficient to undertake the investment only if s [greater than or equal to] [S.sub.FB]. Finally, it will prove convenient to work with the ex ante distribution over signals, G(s), which is defined by its density g(s) := q[f.sub.g](s) + (1 - q)f [sub.b](s).

[] Contracting. The manager's alternative to working for the firm has value R > 0. It turns out that without affecting results, we can suppose that a division's value with a new investment is also equal to R. (6)

Besides satisfying the participation constraint, the contract must incentivize managers to both create new investment opportunities and to assist headquarters in making a more informed investment decision. As we stipulate that the generation of a new project is itself not verifiable, these two tasks cannot be perfectly disentangled. (7) If a division's new project is undertaken and funds k are invested, the manager's pay can, however, be made contingent on the project's success. Precisely, in this case the manager receives a base wage [alpha] and, in case of success, a bonus [beta]. We require that [alpha] [greater than or equal to] [[alpha].bar], to which we simply refer as the "base-wage onstraint." For [[alpha].bar] = 0, the latter represents a standard limited-liability constraint. Finally, if no new investment was made, the manager only receives a wage equal to R.

[] Discussion of contracts. In the rest of this section, we bring out and discuss two restrictions on the set of feasible contracts. (8)

Assumption 1. In case no new investment opportunity is realized in a given division, the respective manager's wage is equal to R.