This article develops the uniform asymptotic theory for local projection (LP) regression when the true lag order of the model is unknown and potentially infinite. The theory allows for varying degrees of persistence in the data, growing response horizons, and general conditionally heteroskedastic martingale-difference shocks. Based on the theory, we make two main contributions. First, we show that LPs can achieve semiparametric efficiency at a given horizon under classical assumptions on the data, provided that the controlled lag order diverges. Thus, the commonly perceived efficiency loss of LPs can become asymptotically negligible with many controls. Second, we propose LP-based inference procedures for (level and cumulated) impulse responses that possess robustness properties not shared by existing methods. Inference methods using two distinct standard errors are considered. The uniform validity for the first method depends on a zero fourth-order cumulant condition on shocks, while that of the second holds more generally for conditionally heteroskedastic martingale-difference shocks. We propose a bootstrap procedure that improves finite-sample performance and extend the standard error construction to structural responses.

This article views hiring as a contextual bandit problem: to find the best workers over time, firms must balance “exploitation” (selecting from groups with proven track records) with “exploration” (selecting from under-represented groups to learn about quality). Yet modern hiring algorithms, based on supervised learning approaches, are designed solely for exploitation. Instead, we build a resume screening algorithm that values exploration by evaluating candidates according to their statistical upside potential. Using data from professional services recruiting within a Fortune 500 firm, we show that this approach improves the quality (as measured by eventual hiring rates) of candidates selected for an interview, while also increasing demographic diversity, relative to the firm’s existing practices. The same is not true for traditional supervised learning-based algorithms, which improve hiring rates but select far fewer Black and Hispanic applicants. Together, our results highlight the importance of incorporating exploration in developing decision-making algorithms that are potentially both more efficient and equitable.

Experts often collect and report information over time. What reporting protocol elicits the most information? Here, a principal receives reports sequentially from an agent with privately known ability, who observes two signals about the state of the world. The signals differ in initial quality and, unlike previous work, differ in quality improvement. The paper finds that “mind changes” (inconsistent reports) can signal talent if a smart agent improves faster. Also, sequential reports dominate when the principal's decision is very sensitive to information; a single report dominates if the mediocre agent's signals improve faster or the agent is likely mediocre.

This paper studies the long-term effects of technology and know-how transfers on structural transformations. In the 1950s, the Soviet Union supported the construction of the 156 Projects, which were large-scale, capital-intensive industrial clusters in China. These projects included a technology transfer, consisting of state-of-the-art Soviet machinery and equipment, and a know-how transfer, via the training of Chinese engineers, production supervisors, and high-skilled technicians by Soviet experts. We use newly assembled data that follow steel plants for over four decades, and we exploit natural variation in the transfers they eventually received. We find that, while production advantages stemming from Soviet technology faded away if not complemented with training, the know-how transfer had a long-lasting impact on plant performance, stimulated technology upgrade when China was a closed economy, and increased exports to the Western world when China engaged in international trade. The know-how transfer also generated productivity and technology spillovers onto complementary establishments.

An agent samples projects over time, observing quality for each, while a principal can select at most one. The principal values quality, whereas the agent only wants a project chosen. Transfers are unavailable, but the principal can verify quality by paying a cost. We fully characterize the dynamics of verification by determining optimal mechanisms for this problem. With a low verification cost and a long horizon, the optimal mechanism involves a deterministic selection rule that initially discriminates on quality but chooses a project irrespective of its quality at a deadline. Verification occurs with an intermediate probability before the deadline, declining over time. We show how these conclusions change if the verification cost is high or the horizon is short, and under certain forms of imperfect commitment. Our analysis provides guidelines on how dynamics interact with the benefits of verification.

The majority of innovations are developed by multi-sector firms. The knowledge needed to invent new products is more easily adapted from some sectors than from others. We study this network of knowledge linkages between sectors and its impact on firm innovation and aggregate growth. We first document a set of sectoral-level and firm-level observations on knowledge applicability and firms’ multi-sector patenting behaviour. We then develop a general equilibrium model of firm innovation in which inter-sectoral knowledge linkages determine the set of sectors a firm chooses to innovate in and how much R&D to invest in each sector. It captures how firms evolve in the technology space, accounts for cross-sector differences in R&D intensity, and describes an aggregate model of technological change. The model matches new observations as demonstrated by simulation. It also yields new insights regarding the mechanism through which sectoral fixed costs of R&D affect growth.

This article proposes a test for the conditional superior predictive ability (CSPA) of a family of forecasting methods with respect to a benchmark. The test is functional in nature: under the null hypothesis, the benchmark’s conditional expected loss is no more than those of the competitors, uniformly across all conditioning states. By inverting the CSPA tests for a set of benchmarks, we obtain confidence sets for the uniformly most superior method. The econometric inference pertains to testing conditional moment inequalities for time series data with general serial dependence, and we justify its asymptotic validity using a uniform non-parametric inference method based on a new strong approximation theory for mixingales. The usefulness of the method is demonstrated in empirical applications on volatility and inflation forecasting.

Observed choices between risky lotteries are difficult to reconcile with expected utility maximization, both because subjects appear to be too risk averse with regard to small gambles for this to be explained by diminishing marginal utility of wealth, as stressed by Rabin (2000), and because subjects’ responses involve a random element. We propose a unified explanation for both anomalies, similar to the explanation given for related phenomena in the case of perceptual judgments: they result from judgments based on imprecise (and noisy) mental representations of the decision situation. In this model, risk aversion results from a sort of perceptual bias—but one that represents an optimal decision rule, given the limitations of the mental representation of the situation. We propose a quantitative model of the noisy mental representation of simple lotteries, based on other evidence regarding numerical cognition, and test its ability to explain the choice frequencies that we observe in a laboratory experiment.

We explore the power that precise information about consumers’ preferences grants an intermediary in shaping competition. We think of an intermediary as an information designer who chooses what information to reveal to firms, which then compete à la Bertrand in a differentiated product market. We analyse the information designs that maximize consumer and producer surplus and uncover systematic differences in how information can be used to intensify or soften competition. Our analysis demonstrates the power that users’ data can endow intermediaries with and speaks directly to current regulatory debates regarding digital marketplaces.

We study collusion among firms against imperfectly monitored environmental regulation. Firms increase variable profits by violating regulation and reduce expected noncompliance penalties by violating jointly. We consider a case of three German automakers colluding to reduce the effectiveness of emissions control technology. By estimating a structural model of the European automobile industry from 2007 to 2018, we find that collusion lowers expected noncompliance penalties substantially and increases buyer and producer surplus. Due to increased pollution, welfare decreases by € 1.57–5.57 billion. We show how environmental policy design and antitrust play complementary roles in preventing noncompliance.