Golnaz Bahrami

The rise in passively managed corporate debt funds has resulted in an increasingly inelastic demand for corporate debt. In this study, we quantify how passive debt ownership affects firms' financial policy. Using fund-specific flows to capture firm-level changes in passive debt ownership that are exogenous to firm fundamentals, we find that firms respond to higher levels of passive debt ownership by increasing leverage. The borrower-friendly terms provided by passive debtholders could also theoretically lead to several potential changes in investment or payout policy. We show that passive debt holding does not affect investment policy. Instead, higher passive ownership predicts increased dividend payouts — even for firms far from index thresholds — exacerbating shareholder-debtholder conflicts. Passive debtholders enable these effects by reducing aggregate ex-ante and ex-post monitoring. The presence of a bank monitor moderates the relationship between passive debt ownership and increased payout, reinforcing the importance of this monitoring channel.

We provide new evidence on how U.S. biodiversity protections affect private land markets using a comprehensive, nationwide parcel-level panel of conservation easements, land values, and land use. Parcels near future easement borders show no preexisting differences or trends in values or land cover. After protection, newly protected parcels experience 40–50% declines in value relative to comparable unprotected parcels, driven almost entirely by losses on protected parcels rather than gains on unprotected ones. These effects are concentrated where development option values are highest, reflecting value losses from foregone development. Protections also substantially reduce observed development rates. The results underscore nuanced trade-offs between conservation, land value, and development.

We study how resolving local land-use uncertainty affects household reinvestment. We assemble a nationwide panel linking single-family properties in Cotality to newly established conservation easements in PAD-US and compare homes very near an easement boundary to otherwise similar homes slightly farther away. Using property fixed effects and easement-by-year fixed effects in an event-study design, we find that homes within 100 meters of a newly formed easement experience an approximately 1.5–2% increase in assessed improvement value relative to homes 100–500 meters away, with no evidence of differential pre-trends and effects that attenuate with distance. The valuation response is mirrored by real investment behavior: by three years after formation, proximate homes are about 1.6 percentage points more likely to exhibit a change in total living area and about 1.6 percentage points more likely to change bedroom or bathroom configurations, with larger effects under a stricter adjacency definition. Heterogeneity by proxies for homeowners' relative focus on consumption versus resale value is consistent with a consumption mechanism rather than local amenity level or resale-driven effects. Together, the results indicate that reducing development uncertainty can trigger highly localized, gradual increases in reinvestment in the existing housing stock.

As climate-driven wildfires intensify, their economic impact on real estate markets is crucial. This paper investigates how the disclosure and regulatory aspects of California's 2007 Fire Hazard Severity Zone (FHSZ) maps affected housing prices and mortgage outcomes. We model the policy through two competing channels — the "hardening mechanism" (positive price effect from safety codes) and the "information shock mechanism" (negative price effect from risk disclosure). Using Difference-in-Differences and Spatial Regression Discontinuity, we find the net price effect is generally negative for Very High Risk areas, as the information shock dominates. In the credit market, the probability of mortgaged purchases rises in Moderate and High Risk zones, while financing tightens in Very High Risk areas. These results show that climate-risk disclosure is not neutral, causing dynamic and locally different changes in asset values and credit conditions.

We study whether local air pollution is priced in municipal bond markets. Exploiting quasi-experimental variation in county-level fine particulate matter (PM2.5) from (i) Clean Air Act nonattainment designations beginning in 2005 and (ii) staggered coal-plant retirements, we find that municipalities in more polluted counties issue debt at higher offering yields and yield spreads. A one-standard deviation increase in log PM2.5 raises municipal bond yield spreads by 1.1 basis points, adding about $259 million in annual interest costs across U.S. counties in our sample. The pollution premium is larger for bonds with greater repayment risk and for longer maturities, consistent with investors pricing persistent credit risk. We provide evidence for three reinforcing fiscal mechanisms — higher local hospital expenditures, lower intergovernmental transfers, and weaker tax revenues — linking pollution to tighter fiscal conditions. Overall, the findings indicate that air quality is capitalized in municipal borrowing costs and meaningfully constrains local governments' fiscal capacity.

Millions now ask AI for financial advice. We study the advice they choose on Chatbot Arena, where two randomly assigned, anonymous AI models answer the same question and the user votes for the answer they prefer. Across 3,060 votes on financial questions, users select advice that is confident and actionable. The more accurate advice wins only 52.2% of decisive votes. When one answer contains a serious error while the other is factually correct, users choose the flawed answer in 47.8% of decisive votes. When an answer is verifiable, user choices favor accurate advice. The more accurate advice wins 57.6% of decisive votes when it is possible to verify the advice but only 46.9% of decisive votes if advice depends on the user's circumstances or on the future. When LLM developers train on similar preference data, they can be expected to supply more confident and less accurate advice. In our sample, we find evidence of such catering responses for Meta's Maverick model, which was tuned to perform better on Arena's leaderboard.