Managerial career concerns and investments in information.

We extend the "implicit incentives" literature by analyzing how career concerns affect a risk-neutral manager's decision about how much to learn about a project before investing in it. The manager has unknown ability that determines the probability with which a good project is available, so the market updates ability assessments from project outcomes. While project choice is efficient in equilibrium, an unobservable investment in the precision of project evaluation allows the manager to control the probabilities of future reputational states. This distorts his investment in precision above first best when project payoffs can be observed only on accepted projects.

1. Introduction

* Effort-aversion and project-choice moral hazard have been extensively studied in the literature, including the manner in which agents' career concerns affect effort choices (see Fama, 1980 and Holmstrom, 1982, 1999) and project selection (see Holmstrom and Ricart i Costa, 1986 and Hirshleifer and Thakor, 1992). In this article we focus on a fundamentally different incentive problem, namely one that arises from the manager's desire to distort corporate investments in information in order to influence the way the labor market learns about his ability. More specifically, the manager's investment in information precision allows him to alter the likelihoods with which the market forms inferences about managerial ability, thereby diminishing the probability of undesirable reputational outcomes.

Our focus is motivated by the observation that in real-world resource allocation decisions, the incentive problems that have to be resolved extend beyond effort and project choices to sophisticated games that managers play with corporate resources. One of these games has to do with how much of a firm's resources the manager devotes to acquiring information about investment opportunities, such as the amount of credit analysis a loan officer does before deciding on a loan application. More precise information requires more resources but reduces the likelihood of decision errors. Thus, the firm faces a tradeoff between the cost and the benefit of information precision. But a manager may face a different tradeoff due to career concerns if the precision of information influences the firm's perception of his ability. Our objective is to study these tradeoffs by addressing three questions: How do the career concerns of managers affect their incentives to invest in information? How are a manager's information-investment incentives affected by greater labor market uncertainty about his ability? What are the implications of the interaction between career concerns and information-investment incentives for regulatory policy and observed organizational practices?

As for the first question, we show that career concerns can cause the manager to overinvest in information. In doing so, he influences what the market can observe to draw inferences about his ability. To fix ideas and get to the heart of the intuition, consider a risk-neutral manager who generates new product ideas for the firm. Ideas may be good or bad, and a manager of higher ability is more likely to generate a good idea. Information is symmetric--a priori nobody knows either the manager's ability or whether the idea is good or bad.

The manager can invest in research and development, solicit consultants' opinions, engage in competitive intelligence, and so on to "test" his idea before proposing or abandoning it. These investments in information generate a (noisy) signal about the true quality of the idea, and the precision of the signal is increasing in the resources expended on such information. Everyone observes the signal the manager generates about the project; the critical asymmetry is that the manager's investment in signal precision is privately observed by him.(1)

The labor market learns about the manager's ability based on its prior assessment, the signal of project quality that is publicly observed, its equilibrium belief about the manager's choice of information precision, and the ultimate outcome of accepted projects. Ex ante the manager's choice of information precision does not influence the labor market's ex post beliefs about ability for any observed outcome. In fact, for any observable outcome, everybody agrees on the posterior assessment of the manager's ability. Moreover, the risk-neutral manager always follows the firm's desired (first-best) project-selection strategy. However, the manager's preference for investment in information (signal precision) does not agree with the firm's. More precise signals increase the probability with which bad projects are weeded out and good projects are accepted, implying that the manager's investment in signal precision influences the likelihood of a good reputation forming ex post. By manipulating his investment in information, the career-conscious manager can tilt the probabilities toward outcomes that are reputationally more favorable.

We consider a de novo project, where success or failure is observed only if the project is accepted. The manager gets the highest reputation if an accepted project turns out to be good, the next-highest reputation if a project is rejected, and the lowest reputation if an accepted project turns out to be bad. This tempts the manager to overinvest in information relative to the first best, because doing so increases the likelihood of the highest reputation forming and decreases the likelihood of the lowest reputation forming.

To develop the intuition further, we consider two alternative structures. First, suppose that project outcomes are observed on rejected projects in addition to accepted projects. In this case, the information-investment inefficiency disappears, since all the relevant outcomes are observed; there is no longer the "safe haven" of rejected projects that exists when only the payoffs on accepted projects are observed. With all possible outcomes observable, the manager cannot benefit from distorting investments in information to influence outcome probabilities. Second, suppose that project outcomes are observed only on rejected projects. Now projects that are accepted generate no additional information, while rejected projects reveal the project's type perfectly. In this case, the manager prefers the signal to be wrong, so that a rejected project ultimately turns out to be good. The equilibrium for this information structure entails the manager underinvesting in information.

As for the second question, we find that when the manager's information investment is unobservable and the payoff can be observed only on an accepted project, the manager's private returns to investment in information are increasing in the market's uncertainty about his ability. This suggests that younger or less-seasoned managers will overinvest in information to a greater extent. Moreover, since increased information expenditures lead to fewer errors in project selection, we predict that the ex ante variance of firm value is decreasing in the market's ex ante (presignal) uncertainty about managerial ability, as long as the prior beliefs about managerial ability aren't too low. By contrast, the more weight the manager's utility function attaches to firm value, the smaller the distortion away from first best and the higher the variance of firm value, again as long as prior beliefs about managerial ability aren't very low.

As for the third question, our analysis has a host of implications for regulatory policy and organizational practice. Our analysis (i) implies that the self-interested behavior of career-conscious bank managers may actually reduce the risk exposure of federal deposit insurance, (ii) explains why younger mutual fund managers may choose portfolios with less unsystematic risk, (iii) indicates that firms with more career-conscious managers will have longer product-development cycle times, and (iv) shows that, in the context of divesting unproductive assets, the manager's overinvestment in information could actually ameliorate another distortion arising from his career concerns, namely his propensity to delay asset divestiture relative to the shareholders' optimum.

Our model extends the seminal work of Holmstrom (1982, 1999), which demonstrates that career concerns create a divergence between the project-choice preferences of a firm and a manager who invests for the firm. In the Holmstrom model, a manager observes projects and recommends promising ones for investment. No one (including the manager) knows the manager's ability, but project outcomes are related to ability. Holmstrom's main result is that if the manager is risk neutral and his payoffs are linear in the posterior belief about his ability, then there is no divergence between the manager's choice of project and the first best, as long as the manager can communicate project risk to the market. When the manager's signal about project quality is also observed by the market,(2) career concerns distort the manager's decision away from first best only when he is risk averse or the output technology is nonlinear in managerial ability so that the process of learning about ability is nonlinear.