Knowledge that Transforms

Do commercial property taxes affect business microlocation decisions? Using precise geolocations for 2.8 million U.S. establishments, we employ a spatial discontinuity density design, comparing for-profit establishments to nonprofits, which are generally exempt from property taxes. Relative to nonprofit establishments, for-profit establishments disproportionately locate on the low-tax side of county borders. This effect strengthens with the size of the tax differential. At borders with differentials in the top decile (quartile), we estimate a 6–7 (5–6) percent shift in establishment location toward the low-tax side over a 10-km band. Businesses move from as far as 10 km away on the high-tax side but concentrate within 1 km on the low-tax side. Firms trade off tax savings with customer access: low-foot-traffic establishments respond more to tax differentials than high-foot-traffic ones. A survey of U.S. mayors finds that perceptions of property taxes’ importance for business location rise with cross-border tax differentials. Data Availability: Survey data are available upon request. All other data were obtained from publicly and commercially available sources described in the manuscript.

We examine the impact of loan officer changes on debt contract design. Loan officers face pressure to originate new loans, incentivizing them to offer loan terms that are more favorable than warranted by borrowers’ creditworthiness. Because a loan officer who replaces the originating loan officer is not subject to this loan prospecting pressure, we predict that she reverses the favorable discretion exercised at loan origination. Consistent with this prediction, we find that the first renegotiation following the replacement of the originating loan officer is associated with more unfavorable renegotiation outcomes. This effect is more pronounced when loan prospecting pressure at loan origination is greater—when the originating loan officer is less experienced, when the lead arranger faces tight performance benchmarks or lower income growth, and when credit market competition is more intense. Overall, we provide novel insights into how loan officer changes restore contracting efficiency through renegotiations. Data Availability: Syndicated loan data are from DealScan; borrower financial data are from Compustat and CRSP; bank financial data are from the Call Report and Compustat; loan agreements are from Morningstar, FactSet, and SEC filings; analyst data are from I/B/E/S; credit rating data are from S&P Global Ratings and the Mergent Fixed Income Securities Database (FISD); media coverage data are from RavenPack; covenant violation data are from Nini, Smith, and Sufi (2012); and loan officer employment histories were hand collected from LinkedIn.

Large language models are increasingly used in social sciences, but their training data can introduce lookahead bias and training leakage. A good chronologically consistent language model requires efficient use of training data to maintain accuracy despite time-restricted data. Here, we overcome this challenge by training a suite of chronologically consistent large language models, ChronoBERT and ChronoGPT, which incorporate only the text data that would have been available at each point in time. Despite this strict temporal constraint, our models achieve strong performance on natural language processing benchmarks, outperforming or matching widely used models (e.g., BERT), and remain competitive with larger open-weight models. Lookahead bias is model and application-specific because even if a chronologically consistent language model has poorer language comprehension, a regression or prediction model applied on top of the language model can compensate. In an asset pricing application predicting next-day stock returns from financial news embeddings, we find that ChronoBERT and ChronoGPT’s real-time outputs achieve Sharpe ratios comparable to a much larger and potentially biased Llama model. Our results demonstrate a scalable, practical framework to mitigate training leakage, ensuring more credible backtests and predictions across finance and other social science domains.

We study branch openings and closures in U.S. banking. Both are concentrated where depositors are rate-sensitive and place less value on physical access, patterns driven by financially sophisticated households with greater digital adoption. The effects are strongest for large banks, with lending playing a lesser role. Incumbents retain branches where depositors are less sensitive and more reliant on proximity; entrants avoid such markets because sticky customers are difficult to attract. The Pandemic accelerated closures while allowing banks to retain deposits through online channels. Overall, depositor sophistication — through rate sensitivity and reduced branch reliance — drives modern branch restructuring.

Price jump intensity is predictable using simple order book depth statistics. We examine the relationship between liquidity fragmentation, consolidated depth, and stock price jumps for S&P 500 stocks. Using a stylized model, we posit that liquidity suppliers fragment offered liquidity and reduce depth to manage adverse selection risk on days with high expected jump intensity. Consistent with our model’s predictions, we find 80% (19%) more jumps when fragmentation (depth) increases (decreases). The findings suggest that market makers anticipate volatility and the associated adverse selection cost, providing traders with advance warning.

We study a large and unanticipated bank cleanup in India after prolonged zombie lending under regulatory forbearance. The cleanup forced widespread zombie exits through bankruptcy. In contrast to a frictionless world where entry automatically follows exit, we show that with financial frictions, local banking conditions shape creation after destruction. New firm entry occurs primarily in regions where banks are present, well-capitalized, competitive, and possess sectoral expertise. We further find a decline in the dispersion of total factor productivity in regions with stronger financial intermediation, consistent with a reduction in misallocation in such areas.

This paper examines whether and how the peer information environment – generated by peer firms’ disclosures and by the information intermediaries covering them – affects the CEO’s delegation of decision rights to division managers (DMs). Using an international survey dataset, I measure within-firm delegation of decision-making over capital investment, human capital investment, investment in new products, and sales. I find that CEOs delegate more to managers of divisions in country-industries with richer peer information environments. Cross-sectional tests highlight three underlying channels. Specifically, the results suggest that peer information (i) helps the CEO acquire information to reduce potential information disadvantages relative to DMs (CEO learning), (ii) mitigates CEO-DM agency frictions (agency mitigation), and (iii) provides DMs with relevant information to make better decisions (DM learning). While CEO learning reduces delegation, agency mitigation and DM learning increase it. These findings provide evidence that information about peers influences the internal organization of firms.

Audits of complex fair value estimates involve evaluating and challenging key client assumptions. Auditors generally lack the necessary expertise to perform this work, and thus, they often engage valuation specialists to perform this audit work. However, valuation specialists primarily perform advisory work and may not put enough care into their audit work. We examine whether valuation specialists' psychological ownership of audit work and auditors' infringement on that work affect specialists' judgment quality. Using a survey and an experiment with practicing valuation specialists, we find that higher psychological ownership improves specialists' evidence evaluation and judgments, but only when auditors do not behave intrusively by infringing on their work. When auditors behave intrusively, the benefits of specialists' higher psychological ownership disappear. Our findings show that audit quality depends not only on specialist behavior but also on auditor interactions with specialists. Our study offers practical guidance to audit firms on how to organize audit work and manage internal working relationships between auditors and valuation specialists.

Concerns about climate risk suggest it should affect risk assessment and pricing of corporate securities, particularly for firms facing potential regulatory restrictions. Employing a shock to expected climate regulations, we find support for this hypothesis given our evidence that climate regulatory risks causally affect bond credit ratings and yield spreads. Moreover, a structural credit model indicates the increased spreads for high carbon issuers, especially those located in stricter regulatory environments, derive from changes in firms’ asset volatilities rather than asset values, highlighting that regulatory uncertainty affects security pricing. The results have important implications for corporate decisions, portfolio management, and policymaking.

We examine how bonus deferral and bonus recovery affect performance in an effort-sensitive task, where improving overall performance requires effort towards output quantity and output quality. Deferred bonuses and bonus recovery are important elements in contemporary incentive schemes designed to motivate employees to act in the best interest of the firm. We propose that bonus deferral improves performance by encouraging employees to adopt an abstract mindset that focuses on behavior for the good of the firm (“to do the right thing”) and thus to exert more effort. In contrast, bonus recovery serves an effort-directing role by indicating undesirable behavior that should be avoided. Bonus recovery activates a mindset that focuses on the monetary implications of their behavior and thus supersedes the abstract mindset of bonus deferral and its positive effects. Our experimental results show that bonus deferral leads to higher performance quantity; in contrast, bonus recovery increases performance quality, the performance domain related to bonus recovery, but at the expense of performance quantity. Furthermore, we find that bonus recovery counteracts the positive effects of bonus deferral and leads to a reduction in overall performance. Two supplemental experiments provide direct process evidence supporting the proposed theoretical mechanism. Our study contributes to the debate on effective compensation by showing that combining bonus deferral and bonus recovery may have undesirable consequences on effort provision and allocation.