Knowledge that Transforms

The advent of large language models (LLMs) creates new opportunities to conduct qualitative interviews at scale, with thousands of respondents, thereby bridging qualitative and quantitative methods. We develop a simple, versatile approach for researchers to run AI-led qualitative interviews, including voice interviews. We assess its robustness by drawing comparisons to human experts and with several respondents-based quality metrics. The versatility of the approach is illustrated through four broad classes of applications: eliciting key factors in decision making, political views, subjective mental states, and mental models of the effects of public policies. High performance ratings are obtained in all of these domains. Our applications highlight the potential of AI-led interviews as a tool for measurement, hypothesis generation, and discovering mechanisms.

We examine whether PCAOB inspection deficiencies related to management review controls (MRCs) affect financial reporting decisions. Using the goodwill impairment setting, we find that firms are more likely to record goodwill impairments when their auditors have received more PCAOB inspection deficiencies related to goodwill MRCs. The association is stronger when inspection deficiencies are more likely to lead to changes to the design or implementation of an internal control (repeat deficiencies and deficiencies related to MRC precision). We observe some evidence that MRC inspection deficiencies increase the likelihood of goodwill impairment even when unrelated to goodwill, indicating that improving the audit of MRCs can broadly affect financial reporting. Our findings provide evidence that PCAOB inspections affect not only auditor behavior but also the financial reporting behavior of audit clients. Specifically, inspection feedback appears to influence reporting outcomes through changes in the audit and evaluation of MRCs, identifying a channel through which regulatory oversight affects issuers. The results have implications for audit firms seeking to remediate inspection deficiencies, for management and audit committees responsible for overseeing significant estimates and impairment analyses, and for regulators evaluating the broader effectiveness of the PCAOB inspection process.

I study how an activist shareholder organizes a cost-sharing coalition with active shareholders while seeking support from passive shareholders outside the coalition. A larger coalition shifts more of the campaign cost to members, allowing the activist to pursue lower-value proposals. Observing a larger coalition therefore lowers passive shareholders’ posterior expectation of proposal value. The activist chooses the largest coalition that preserves support from pivotal passive shareholders, endogenizing coalition size. Consequently, coalition size need not be a monotone measure of activist influence, and activism frequency depends on passive shareholders’ preferences. Coalition building strengthens the proxy-fight threat and can increase prevote settlements.

We investigate a dynamic oligopoly model with two phases: a normal phase where an essential input is available, and a disruption phase where firms must rely on inventories. Disruptions create an overcutting pricing incentive arising from the option to forgo sales and monopolize the market once rivals stock out. This monopoly option causes the price to jump to a supracompetitive level at the advent of the disruption and to remain above the competitive price path throughout. Price caps raise welfare, but reduce inventory investment. Government reserves reduce prices and raise buffer stocks, as do reserve mandates when financially feasible.

We propose a novel approach to synthesize presumably information-driven insider trading signals for the cross section of stocks. We find that the resulting straightforward composite strategy can often forecast returns in a global sample. This predictability is strongest in equal-weighted portfolios, short holding periods, and insider-buying signals, and it improves further with global diversification. Cross-country analysis reveals that varying insider trading restrictions between countries have limited explanatory power for the performance of the composite strategy. Overall, the results are consistent with a short-term informational advantage of insiders, which might be due to the skillful interpretation of non-private news.

This study examines whether programmable spreads, defined as the difference between decentralized finance smart-contract lending rates and U.S. Treasury-Bill benchmarks, contain forward-looking information about liquidity stress and risk aversion. Using data from major stablecoin lending markets, it finds that programmable spreads are associated with subsequent financial tightening and contain statistically significant forward-looking information regarding future changes in U.S. financial conditions. Wider spreads also forecast higher future excess returns in equity and bond markets, consistent with the predictive relationships implied by time-varying risk premia. These relationships hold both in sample and out of sample and are strongest at short horizons. Moreover, these findings suggest that programmable spreads may serve as high-frequency, complementary market-based indicators of funding stress and information associated with time-varying risk aversion.

This paper studies how the accuracy of a short‐term performance measure affects its value when a principal decides whether to retain or fire an agent, who can manipulate that measure. Compared with ignoring performance and never firing the agent, the performance measure is valuable when the savings in the expected cost of the retention error is more than the expected cost of the firing error. A more accurate performance measure can have a non‐monotonic effect on its value: Greater accuracy improves the principal's ability to correctly identify the agent, which increases efficient firing, but it causes the incompetent agent to increase manipulation to avoid being fired, which increases inefficient retention. With a less accurate performance measure, the incompetent agent has weaker incentives for manipulation, and an increase in accuracy increases its value due to the identification effect. With a more accurate performance measure, the manipulation effect dominates, and an increase in accuracy decreases its value. The results imply that the principal prefers an imperfect performance measure when the agent can choose the level of manipulation.