Knowledge that Transforms

The editors of Management Science acknowledge all our 2022 reviewers and guest associate editors who generously provided expert counsel and guidance on a voluntary basis. We are grateful for their contributions; without them, the journal could not function. Here, we list those reviewers who contributed four or more reviews in 2022.

Artificial intelligence systems are increasingly demonstrating their capacity to make better predictions than human experts. Yet recent studies suggest that professionals sometimes doubt the quality of these systems and overrule machine-based prescriptions. This paper explores the extent to which a decision maker (DM) supervising a machine to make high-stakes decisions can properly assess whether the machine produces better recommendations. To that end, we study a setup in which a machine performs repeated decision tasks (e.g., whether to perform a biopsy) under the DM’s supervision. Because stakes are high, the DM primarily focuses on making the best choice for the task at hand. Nonetheless, as the DM observes the correctness of the machine’s prescriptions across tasks, the DM updates the DM’s belief about the machine. However, the DM is subject to a so-called verification bias such that the DM verifies the machine’s correctness and updates the DM’s belief accordingly only if the DM ultimately decides to act on the task. In this setup, we characterize the evolution of the DM’s belief and overruling decisions over time. We identify situations under which the DM hesitates forever whether the machine is better; that is, the DM never fully ignores but regularly overrules it. Moreover, the DM sometimes wrongly believes with positive probability that the machine is better. We fully characterize the conditions under which these learning failures occur and explore how mistrusting the machine affects them. These findings provide a novel explanation for human–machine complementarity and suggest guidelines on the decision to fully adopt or reject a machine.

Start-ups are increasingly using social media to signal quality and provide information to potential investors. However, the effectiveness of social media on venture capital (VC) financing is likely to be heterogeneous, differing by demographic and network characteristics of the founders. In this paper, we examine whether social media use can improve funding outcomes for firms founded by women and by other people also lacking connections to the investor network, two groups that face greater difficulties in securing VC financing. Using Twitter data and data on VC investment in start-ups from Crunchbase, we explore the interaction effect between Twitter usage and gender and between Twitter usage and the network constraint measure. Overall, we show that social media can mitigate some disparities in financing experienced by these firms through improving information access. We find that this effect is stronger for first-time entrepreneurs than for experienced ones, stronger for attracting new investors than repeat ones, and stronger in more competitive markets. Collectively, these results suggest that social media could primarily help women and less connected individuals obtain financing by alleviating information asymmetry between founders and investors.

Weather risk affects the economy, agricultural production in particular. Index insurance is a promising tool to hedge against weather risk, but current piecewise-linear index insurance contracts face large basis risk and low demand. We propose embedding a neural network-based optimization scheme into an expected utility maximization problem to design the index insurance contract. Neural networks capture a highly nonlinear relationship between the high-dimensional weather variables and production losses. We endogenously solve for the optimal insurance premium and demand. This approach reduces basis risk, lowers insurance premiums, and improves farmers’ utility.

Recent studies have proposed a large set of powerful anomaly-based factors in the stock market. This study examines the role of investor inattention in the corresponding anomalies underlying these factors and other underreaction-related anomalies. Using media coverage as a proxy for investor attention, we show that the anomalies underlying many recently proposed prominent factors are much more pronounced among firms with low media coverage in portfolio-formation periods. In addition, we find many other prominent anomalies that previous literature has attributed to underreaction also tend to perform much better among firms with low media coverage. The average Fama-French five-factor alpha spread of these anomalies is about 0.97% per month among firms with low news coverage and only 0.24% per month among firms with high news coverage. Moreover, most of the alpha spread comes from the short leg of the anomalies and from the firms that are more difficult to arbitrage. Overall, our evidence indicates that investor inattention at least partially drives many of the recently proposed factors.

We develop a model of private equity capturing two fundamental features of this market: the fund structure and illiquidity. A fund structure with sequential capital calls arises as an optimal solution to fund managers’ (GPs) moral hazard problem but exposes investors (LPs) to illiquidity risk. Funds with more illiquidity-tolerant LPs realize higher returns, leading to different expected returns across both funds and LPs in equilibrium. GPs may inefficiently accelerate drawdowns to avoid default by LPs on capital commitments. With a secondary market for LP claims, differences in fund returns are attenuated but differences in LP returns remain. The model can rationalize several empirical findings on primary and secondary private equity markets.

This paper studies a pricing problem for a single-server queue where customers arrive according to a Poisson process. For each arriving customer, the service provider announces a price rate and system wait time. In response, the customer decides whether to join the queue, and, if so, the duration of the service time. The objective is to maximize either the long-run average revenue or social welfare. We formulate this problem as a continuous-time control model whose optimality conditions involve solving a set of delay differential equations. We develop an innovative method to obtain the optimal control policy, whose structure reveals interesting insights. The optimal dynamic price rate policy is not monotone in the wait time. That is, in addition to the congestion effect (the optimal price rate increases in the wait time), we find a compensation effect, meaning that the service provider should lower the price rate when the wait time is longer than a threshold. Compared with the prevalent static pricing policy, our optimal dynamic pricing policy improves the objective value through admission control, which, in turn, increases the utilization of the server. In a numerical study, we find that our revenue-maximizing pricing policy outperforms the best static pricing policy, especially when the arrival rate is low, and customers are impatient. Interestingly, the revenue-maximizing policy also improves social welfare over the static pricing policy in most of the tested cases. We extend our model to consider nonlinear pricing and heterogeneous customers. Nonlinear pricing may improve the revenue significantly, although linear pricing is easier to implement. For the hetergeneous customer case, we obtain similar policy insights as our base model.

Policy learning using historical observational data are an important problem that has widespread applications. Examples include selecting offers, prices, or advertisements for consumers; choosing bids in contextual first-price auctions; and selecting medication based on patients’ characteristics. However, existing literature rests on the crucial assumption that the future environment where the learned policy will be deployed is the same as the past environment that has generated the data: an assumption that is often false or too coarse an approximation. In this paper, we lift this assumption and aim to learn a distributionally robust policy with incomplete observational data. We first present a policy evaluation procedure that allows us to assess how well the policy does under worst-case environment shift. We then establish a central limit theorem type guarantee for this proposed policy evaluation scheme. Leveraging this evaluation scheme, we further propose a novel learning algorithm that is able to learn a policy that is robust to adversarial perturbations and unknown covariate shifts with a performance guarantee based on the theory of uniform convergence. Finally, we empirically test the effectiveness of our proposed algorithm in synthetic datasets and demonstrate that it provides the robustness that is missing using standard policy learning algorithms. We conclude the paper by providing a comprehensive application of our methods in the context of a real-world voting data set.