Abstract We study a macroprudential regulation in the emerging market of Chile that raised loan-loss provisions for residential mortgages with loan-to-value (LTV) ratios above 80%. The policy reduced high-LTV borrowing and overall leverage but unintentionally affected households likely to borrow above 80% LTV based on preregulation characteristics. These borrowers liquidated term deposits to meet higher down payments, lowering liquidity and raising short-term delinquency, especially near the threshold. The results uncover a regulatory trade-off: systemic risk is curbed, but financially constrained households face short-term vulnerability. (JEL G21, G28, D40) Received: August 28, 2025; Editorial decision: January 16, 2026 Editor: Isil Erel Authors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online.
Abstract Student use of Artificial Intelligence (AI) in higher education is reshaping learning and redefining the skills of future workers. Using student-course data from a top Israeli university, we examine the impact of generative AI tools on academic performance. Comparisons across more and less AI-compatible courses before and after ChatGPT’s introduction show that AI availability raises grades, especially for lower-performing students, and compresses the grade distribution, eroding the signal value of grades for employers. Evidence suggests gains in AI-specific human capital but possible losses in traditional human capital, highlighting benefits and costs AI may impose on future workforce productivity.
Delegating high-stakes decisions creates a fundamental tension: incentivizing experts to acquire unobservable information inevitably distorts their final choices. In a principal-agent setting, we characterize the optimal compensation contract under hidden learning, showing it endogenously generates either contrarian or conformist bias. The direction of this bias depends on learning costs and the precision of public and private information. Our framework links information acquisition incentives to systematic biases in experts' choices and offers a unifying explanation for conflicting empirical evidence in financial advice: why analysts issue excessive contrarian recommendations, and why inexperienced analysts follow the consensus more than their experienced peers. (JEL D82, D83, G24)
As nonbanks' market share increases in a local residential mortgage market, the quality of their postorigination mortgage servicing improves. This finding is confirmed by two instrumental variable analyses exploiting (1) stress tests conducted by the Federal Reserve, and (2) mortgage industry surety bonds required by each state. Evidence suggests that improvements in service quality arise through two channels. First, expansion in nonbank market share leads to greater lender specialization: nonbanks increasingly service lower-income borrowers, while traditional banks increasingly concentrate on higher-income segments. Second, higher nonbank market share is associated with increased investment in technology by nonbanks. (JEL G21, G23, L13, L15)Received: May 22, 2025
Does AI make firms vulnerable or resilient to cyberrisk? We develop a firm-year measure of AI intensity for U.S. listed firms using patents and 10-K business descriptions. A 1-standard-deviation increase in cyberrisk reduces patenting by 25%-30% for non-AI firms, with larger declines in data-intensive technologies. Frontier AI firms are much less affected, and their valuations rise when cyberrisk is high. This resilience does not extend to firms that adopt external AI tools without internal AI innovation. The evidence fits two channels: cyberrisk raises the cost of data-intensive innovation, and internal AI development builds organizational capacity to sustain innovation under cyberrisk.
We examine how firms are affected by the political bargaining power of their headquarters' region. Exploiting variation in the strategic importance of swing states stemming from shifting partisan balance in the U.S. Senate, we find that corporate valuations and investments positively respond to increases in regional political influence. We verify the valuation findings using an event study based on the 2021 Georgia runoff election that unexpectedly produced a 50-50 balance in the Senate. We investigate potential policy mechanisms and find that tax incentives constitute the most likely channel through which firms benefit from the political bargaining power of their headquarters' region.
We present causal evidence on the effect of boardroom networks on firm value. We exploit a ban on interlocking directorates of Italian financial and insurance companies as exogenous variation and show that firms that lose centrality in the network experience negative abnormal returns around the announcement date. The key driver of our results is the role of boardroom connections in reducing asymmetric information. We find stronger effects for firms with fewer business partners and less experienced directors. Finally, we show that network centrality has a positive effect on directors' compensation, providing evidence of rent sharing, and liquidity. (JEL: D57, G14, G32, L14)Received: 25 January 2024Editor: Andrew EllulAuthors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online.
We study a nonlinear relationship between corporate investment and Tobin's q in the cross section of firms. After correcting for nonlinear errors using a repeated measurement of q derived from analysts' forecasts, we find evidence of varying investment-q sensitivity across firms. The investment-q sensitivity is low for firms with low q. It then becomes more pronounced at intermediate values before weakening at high values of q, resulting in an S-shaped pattern. In the cross section, the true investment-q relation is therefore not strictly linear. Firm investment is predicted to remain similar among firms with low q, suggesting that increases in q do not necessarily lead firms to increase investment significantly. (JEL C21, C26, E22, G31)
We examine the influence of consultants on the portfolio choices and performance of institutional asset owners' private equity (PE) investments. We find that asset owners using the same search consultant make similar investment choices. Asset owners advised by consultants that focus on a narrower list of PE managers perform better. Among asset owners that share a consultant, those with the largest PE mandates (top clients) end up with better-performing investments. For mechanisms underlying consultants' performance impact, we find evidence for both access and selection abilities. (JEL G11, G20, G23)
The presence of shadow banks in corporate term loan syndicates adversely affects credit lines' liquidity provision, despite shadow banks not directly funding credit lines. Within the same syndicated loan deal, shadow banks attract not only riskier borrowers but also fewer banks as co-lenders, both in the term loan and in the credit line. Furthermore, credit lines in deals funded by shadow banks, compared to those without shadow bank participation, are smaller, with shorter maturities, and lower drawdown rates. Overall, our results highlight that syndicated loan deals with a strong presence of shadow banks offer borrowers lower liquidity protection. JEL G21, G22, G23
We study how firms can design their organizational structures to overcome dynamic commitment problems when entering new markets. A manager exerts costly effort to first develop and subsequently manage an investment opportunity. Ex post, the firm underinvests in projects that generate high management rents. However, the prospect of those rents helps offset the manager's initial project development cost, making ex ante commitment to invest optimal. Levered subsidiaries mitigate this time-consistency problem by introducing risk-shifting incentives that counteract underinvestment. Subsidiaries are most valuable for projects that are costly to develop, have moderate management costs, and yield returns uncorrelated with existing business. (JEL G32, G34, L22)
Using granular data on global investment funds in difference-in-differences regressions around the announcement of the U.S. Inflation Reduction Act (IRA), we identify a novel international spillover channel of green industrial policies. Sustainable global investment funds received more inflows with the act announcement, in turn increasing their cross-border portfolio investments worldwide. Recipient economies better prepared to address climate change benefited most from sustainable global funds' additional investments. Our results are stronger for funds with a larger portfolio share invested in the United States and in IRA-targeted industries. Yet, we see strong international spillovers even for non-U.S.-domiciled sustainable funds investing entirely outside the United States. Thus, global investment funds have become an important conduit for the international spillover of climate policies. (JEL F3, G1, G2, Q5)Authors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online.
Downstream customer firms' bargaining power can lead to suboptimal diversification in upstream suppliers' innovation when customers cannot commit to a long-term relationship. After the revelation of financial fraud by a major customer, suppliers surprisingly outperform a control group in terms of sales growth, Tobin's q, and survival likelihood over a 10-year period. Our results suggest that, before a fraud revelation, supplier managers' short decision horizons and aversion to short-term risk enable influential customers to demand relation-specific innovation, leading to suboptimal diversification. When customer importance weakens, suppliers engage in riskier and novel innovation, thereby stimulating sales growth. (JEL G14, G3, L14, L24)
This study demonstrates that banking organizations with higher artificial intelligence (AI) investments are exposed to more operational risk. Using comprehensive supervisory data on operational losses from large U.S. bank holding companies (BHCs) combined with detailed company-level data on AI-skilled human capital, we show that BHCs with more AI investments suffer higher operational losses per dollar of total assets. The impact of AI investments on operational losses significantly varies by loss type and is driven by external fraud, client-related issues, and system failures. These losses stem not only from small, frequent incidents but also from severe, tail-risk events. The risk-enhancing effect of AI is more pronounced for BHCs with weaker risk management practices. Our findings have important implications for banking performance, risk, and supervision.
I study how firms adjust leverage, debt maturity, and cash to manage profitability shocks, and show that time variation in concentration of maturity dates arises endogenously. To avoid rollover risk, firms prefer long-term debt with dispersed maturity dates. However, severe negative shocks force firms to borrow above an optimal level. They issue short-term debt as a commitment to delever in the next period. This concentrates maturity dates in the next period. The calibrated version of the model matches empirical facts and makes novel predictions regarding dynamics of debt maturity dispersion.
We model credibility challenges financial regulators often face when disclosing bank stress test results. Since disclosures influence banks' risk-taking and depositors' withdrawal decisions, regulators may have incentives to misreport. We show that regulators can reveal results credibly through imprecise disclosures to both banks and depositors. The regulator reveals only the range or the interval in which the result lies. Crucially, our findings indicate that stress test results can be disclosed credibly without assuming that the regulator is committed to truthful disclosure.
Using a data set of public procurement auctions and registered shareholders of all bidding firms in Singapore, we study the effects of ownership networks on prices and efficiency in product markets. We find participating bidders with common owners or common owners' owners are more likely to submit identical bids, and the identical bids are associated with higher contract prices. Our structural estimates suggest removing ownership network effects improves a procurer's cost efficiency. Our findings are robust to falsification tests, bid rounding concerns, placebo tests using other common stakeholder relationships, and weighting based on a machine-learning prediction of the auction format.
Regulators condition bank capital on risk but struggle to measure risk accurately. Capital requirements thus rely on inputs from banks’ internal risk models and banks have discretion over modeling choices. Using novel hand-collected data we find systematic differences in reported risk for the chosen simulation method, holding period and historical data size. Hence, modeling choices can be a significant channel of underreporting of risk. Consistent with this presumption we find that less-capitalized banks tend to choose less conservative methods. Moreover, banks using a softer simulation method display higher actual market risk while reporting lower market risk to the regulator.
We study the long-run outcomes associated with hedge funds' compensation structure. Over a 22-year period, the aggregate effective incentive fee rate is 2.5 times the average contractual rate (i.e., around 50% instead of 20%). Overall, investors collected 36 cents for every dollar earned on their invested capital (over a risk-free hurdle rate and before adjusting for any risk). In the cross-section of funds, there is a substantial disconnect between lifetime performance and incentive fees earned. These poor outcomes stem from the asymmetry of the performance contract, investors' return-chasing behavior, and underwater fund closures. Institutional subscribers to the NBER working paper series, and residents of developing countries may download this paper without additional charge at www.nber.org.
