Knowledge that Transforms

Modeling and leveraging consumers’ dynamic search behaviors presents significant business opportunities. Although deep learning methods excel at processing vast consumer data for predictive tasks, their opaque nature limits interpretability and fails to explicitly model consumer decision making. In contrast, economic theory suggests that consumers follow a sequential search strategy, evaluating alternatives until they find the best match for their preferences. To bridge this gap, we propose the sequential search transformer (SST), a deep structural econometric model that integrates deep learning with sequential search theory to model search and purchase decisions. SST unifies these two approaches into an end-to-end trainable model, improving both predictive accuracy and policy evaluation capabilities. Unlike conventional deep learning models, SST explicitly models consumer decision making, and unlike existing sequential search models, it enables consumer behavior modeling across sessions and sequentially resolves utility uncertainty for searched items. We provide a theoretical analysis of the identification strategy for the SST model and show that all parameters can be identified under the proposed framework. Then, we apply SST to a data set with detailed clickstream data collected from a U.S. e-commerce website. Empirical evaluations show that SST outperforms state-of-the-art deep learning and structural models in predicting consumer searches and purchases. Moreover, policy experiments demonstrate SST’s effectiveness in optimizing product recommendations and new product promotion strategies, ultimately enhancing consumers experience and driving revenue growth.

This paper examines the effect of crossborder carbon taxes on shareholder wealth. Using stock price reactions to key announcements of the European Union (EU) carbon border adjustment mechanism (CBAM), we find that EU purchasers of CBAM-covered products experience significantly lower returns than non-European producers of such products. The effect is strongest for EU purchasers with non-EU supply chains. Further cross-sectional analyses show that these negative reactions are more pronounced when CBAM-related costs are higher and when firms have a lower ability to pass them on to their trading partners. Overall, the evidence suggests that equity markets expect crossborder carbon pricing on imports to be costly for EU firms. This paper was accepted by Caroline Flammer, sustainability. Funding: G. Ormazabal thanks the Cátedra de Dirección de Instituciones Financieras y Gobierno Corporativo del Grupo Santander, the R + D + I Project [Reference PID2022-143016NB-I00 funded by MCIN/AEI/10.13039/501100011033], and ERDF “A way of making Europe” [Grant TED2021-132531B-I00 funded by MCIN/AEI/10.13039/501100011033, the European Union NextGeneration EU/PRTR, IESE’s High Impact Projects Initiative–2023, and the Social Trends Institute]. R. Raney acknowledges financial support from the Spanish Ministry of Science and Innovation [Grant PID2019-111143GB-C31 funded by MICIU/AEI/10.13039/501100011033] and [Grant PID2023-150744NB-C41 funded by MICIU/AEI/ 10.13039/501100011033].

Do job seekers consider the race or gender of an employer when applying for a job? Although we have extensive research on employer-side discrimination, we know less about employee-side biases and their consequences. In this study, we examine how the gender and race of the employer shape the willingness of prospective employees to apply for a job. To examine this, we conducted a field experiment where we randomized real jobseekers into three conditions according to employer demographics. We find that job candidates were less likely to apply to a job when they learn that the founders are Black, and, when they did apply, they requested 10% higher salary. In addition, the more qualified a candidate, the less likely they were to apply to Black founders, leaving Black founders with a pool of candidates that is smaller, worse, and more expensive than their White peers. We find no gender penalty for White female founders. Findings from two survey experiments suggest that the penalty is unique to White applicants evaluating Black founders and reflects a concern among White applicants that they will be less likely to fit within a firm and that the firm is less likely to be successful in the long run. We find no evidence of a widespread applicant homophily where all applicants favor founders of their own ethnic group, nor do we find evidence of widespread statistical discrimination whereby all applicants penalize Black founders for being atypical members of the entrepreneurial class.

This paper investigates how environmental regulatory uncertainty affects green innovation in polluting firms. The findings suggest that, instead of adopting a passive “wait-and-see” strategy, polluting firms proactively engage in green innovation and R&D activities. To address endogeneity, I employ an instrumental variable approach using political polarization in roll-call votes on environmental and climate issues in the U.S. House of Representatives. I also exploit the proposal of the Affordable Clean Energy rule in 2018 as an exogenous shock in a difference-in-differences framework. The results are consistent with the growth options view of green innovation in polluting firms. Additional evidence suggests that environmental regulatory uncertainty leads to reductions in toxic emissions.

We examine the value of skilled immigrants in finance by exploiting evidence from visa lotteries. We find that hedge fund management companies that secure more H-1B visas in random lotteries deliver higher alphas, Sharpe ratios, and information ratios. The superior performance of funds with high H-1B visa allocations can be attributed to well-paid and highly educated H-1B workers with quantitative skills. H-1B workers add value by helping hedge funds develop distinctive investment strategies, arbitrage prominent stock anomalies, and overcome capacity constraints. Hedge funds appear to exploit labor market frictions as alpha generation is greatest by workers from countries with the longest wait times for U.S. permanent residency.

Gender-anonymous hiring practices have been widely advocated as a means to reduce labor market inequalities, such as the gender wage gap and the underrepresentation of women in leadership. However, their effectiveness remains debated. This paper studies an experimental labor market where employee candidates set their own performance targets for a real-effort task and employers hire based on these self-set targets. In many professional settings, such targets serve as performance indicators and influence hiring and promotion decisions. In an online experiment with 4,674 participants, we vary in a 2 × 2 design (1) whether employers know the candidates’ genders and (2) the severity of the payoff consequences if the hired employee misses their target. This allows us to examine the interaction effects of gender anonymity in hiring and the performance target’s payoff relevance. We find that given equal ability, women set lower targets than men. Higher targets increase the likelihood of being hired, whereas a larger expected target-performance gap reduces hiring chances, particularly when missing the target has severe consequences for the employer. Importantly, our findings suggest that gender-anonymous applications may have unintended consequences. When gender is revealed, employers appear to adjust for gender differences in target setting, expecting a smaller target-performance gap for women than for men. As a result, women are more likely to be hired and receive higher payoffs when their gender is known. These results indicate that gender-anonymous hiring may backfire for women by preventing employers from accounting for behavioral gender differences (e.g., in self-promotion), ultimately reducing women’s hiring prospects.

Using China’s randomized pre-initial public offering on-site inspection program, we examine the causal effects of proactive public enforcement on the financial reporting quality of initial public offering applicants. Our analysis shows that the program enhances the reporting quality of selected applicants by screening out firms with questionable financial disclosures. More importantly, it generates a significant deterrence effect, discouraging future applicants from submitting low-quality financial reports. However, we also find that public enforcement can lead to the rejection of some firms that may otherwise merit approval. These findings highlight both the benefits and costs of employing public enforcement to improve financial reporting in settings with weak institutional oversight.

The future architecture of financial systems is a subject of contention, with centralized and decentralized governance proponents. Here, we ask the following question. Would the architecture affect the quality of decision making? We propose a game where financial network participants demarcate the ownership of claims to income. This governance task can be decentralized (shared authority), centralized (single authority), or hybrid (alternating authority). Without communication, all architectures supported poor outcomes. With communication, decentralization ensured good governance and maximum profits, whereas centralization did not—lowering communication’s potency in promoting socially optimal decisions. This indicates that there is scope for decentralization in innovating financial institutions. This paper has been accepted by Camelia Kuhnen for the Virtual Special Issue on Digital Finance.

Wisdom of crowds estimates can be compromised when some agents’ predictions are systematically biased. A natural remedy is to aggregate predictions from a subset of more accurate agents rather than the entire crowd. I propose cluster weight on advice (CWOA), a novel “two-shot” algorithm to identify a more accurate subgroup in a single-prediction-problem context. CWOA first applies kernel density estimation to identify clusters of similar initial predictions. If multiple clusters emerge—indicating potential heterogeneity in agents’ information—the algorithm proceeds to present a piece of numerical advice (e.g., the group mean) and elicit updated predictions. This enables the calculation of the weight on advice (WOA)—the scaled magnitude of each agent’s belief revision. CWOA then averages the updated predictions within the cluster with the lowest mean WOA. A behavioral model and simulations explain both why and when cluster-level WOA signals accuracy. Better-informed agents—having already incorporated higher-quality information—perceive less corrective value in the advice and therefore, exhibit lower WOA. CWOA does not require agents to know the true biases or the composition of the crowd; a modest relative advantage in perceived estimation bias by better-informed agents may be sufficient, even under misperceptions of variance, advice quality, and psychological biases in advice taking. Empirically, I first test and confirm the model’s key insight in a controlled experimental setting. I then validate CWOA’s performance across multiple preregistered and archival data sets, including a study in which numerical advice comes from artificial intelligence. CWOA consistently outperforms benchmarks, including the state-of-the-art metaprediction-based methods.

Research on discrimination has predominantly relied on audit studies, which provide clean causal estimates through randomized profile testing but offer only static snapshots of bias. We adopt a dynamic perspective to examine how racial discrimination evolves over time. Using observational panel data from an online educational platform, we first find that African-American teachers receive 30.2% fewer bookings, and their opened classes are 3.3% less likely to be booked compared with their White counterparts during their first week on the platform, a gap that may indicate initial discrimination. Employing individual fixed effects models, we estimate how this gap evolves over time. The results show a striking 1,007% widening of the initial gap over the 12-week period. Our mechanism analysis shows that new students are significantly less likely to book classes with African-American teachers than with White teachers, even when African-American and White teachers have minimal or comparable reputation metrics. As a result, African-American teachers accumulate reputation metrics and customers more slowly than White teachers. Over time, the growing gaps in reputation and customer base further exacerbate the booking gap, driven by both new and repeat students. Our findings suggest that traditional audit studies may significantly underestimate the long-term consequences of early-stage discrimination and highlight how reputation and repeat customer accumulation serve as bias amplifiers on online platforms. We also show that this amplification mechanism operates regardless of whether the initial gap stems from racial discrimination and thus generalizes to other contexts.