Wenzhen Lin

Economist at the Office of Policy Development and Research (PD&R) in the U.S. Department of Housing and Urban Development

Welcome! I am an economist within the Housing Finance Analysis Division at the U.S. Department of Housing and Urban Development (HUD). My primary focus is on researching and analyzing mortgage performance. Currently, I am actively involved in projects related to FHA TOTAL scorecard algorithm modeling and loss mitigation. My research primarily revolves around applying econometric models to urban and real estate economic issues, with a particular emphasis on the mortgage market.

Contact: wenzhenlin001@gmail.com

My research explores the dynamic interplay of economic, environmental, and policy factors in mortgage and housing markets, aiming to enhance understanding and inform effective housing strategies.

Mortgage and Housing Market Dynamics

Mortgage Curtailment Pre and Post-HARP: Income and Substitution Effects

Wenzhen Lin and Yingqi Xu (2025)

(Under Revise at The Journal of Real Estate Finance and Economics)

Analyzed mortgage curtailment behavior among HARP refinance participants
Utilized quasi-experimental methods, leveraging HARP eligibility criteria to address endogeneity concerns related to borrower financial literacy.
Employed instrumental variable (IV) estimation and matching techniques to measure the causal impact of HARP on mortgage curtailment behavior.
Identified income and substitution effects, finding that curtailment behavior is more responsive to changes in cash flow than to mortgage rate adjustments.

Post-Forbearance Prepayments

Robert Dunsky, Wenzhen Lin, and Fan Yang (2025)

(Forthcoming as a working paper on the FHFA website)

Forthcoming as a working paper on the FHFA website

Investigated the effect of the CARES Act Mortgage Forbearance program on borrowers’ prepayment behavior
Applied the p-score to pair forbearance loans with non-forbearance loans
Analyzed prepayment benefits using the Heckman two-step model, showing the program's generally negative impact on refinance benefits but varying by borrowers' exit types.

A Dirichlet Nested Logit Model of Household Mortgage Payment in the CARES Act Forbearance Program

Robert Dunsky, Wenzhen Lin, Jan Ondrich (2024)

Analyzed mortgage payment behavior under the CARES Act Mortgage Forbearance program to develop predictive models for borrowers' exit strategies using extended beta-logistic and Dirichlet nested logit models.
Found a strong correlation between payment patterns and exit strategies; borrowers making additional payments were more likely to conclude forbearance through prepayment or reinstatement, while those deferring payments tended towards modification or deferral plans.

A Transformed Mixed Multinomial Logit Model with an Application to Residential Mortgage Terminations

An Ran, Wenzhen Lin, Jan Ondrich (2021)

Developed and applied a transformed mixed multinomial logit model to analyze mortgage terminations, including 90-day delinquency and prepayment risks, enhancing the accuracy of predicting early mortgage defaults.
Conducted comprehensive empirical tests using Fannie Mae public use data, demonstrating that the transformed models consistently outperformed traditional multinomial logit models both in-sample and out-of-sample validations.

The Influence of Crime, Disasters, and Climate on FHA Mortgage Behavior

Stephanie Hawke and Wenzhen Lin (2024)

(HUD PD&R Edge, 2024)

Analyzed the impact of local context factors such as crime rates, natural disasters, and climate changes on FHA mortgage behavior, integrating multiple public datasets with FHA loan performance data.
Developed predictive models using logit regression to assess mortgage termination risks, highlighting significant predictors of 90-day delinquency in FHA-backed mortgages.
Demonstrated through empirical analysis that local stressors significantly influence mortgage delinquency.

Housing Trend and Rental Markets

Positive Rental History

Wenzhen Lin and Jeffrey Perry

(HUD PD&R Edge, 2024)

Positive rental history has been incorporated into the Federal Housing Administration’s scoring system to improve mortgage approval rates for first-time homebuyers and those without extensive credit histories.
Initial analysis shows that including rental history helps lower-income, younger, and minority applicants to secure mortgages, potentially increasing homeownership among these groups.

What makes a landlord in 2018? Updating the shared understanding of rental property ownership by US households

Mark Shroder, Stephanie Hawke, and Wenzhen Lin (2024)

(under review at Cityscape)

Analyzed small landlords' characteristics and motivations using 2018 Health and Retirement Study data.
Applied advanced econometric techniques, including logistic regression, Cragg Hurdle models, and machine learning, to assess landlord decision-making.
Found risk aversion and higher education levels to be negatively correlated with landlord status, while individuals over 65 were significantly less likely to own rental properties.

Applying Seasonal Adjustments to Housing Markets

William M. Doerner and Wenzhen Lin

(Cityscape: A Journal of Policy Development and Research, 2022)

Examined house price seasonality trends, highlighting inconsistencies in public data adjustments despite increasing market seasonality.
Analyzed the impact of weather, social, and industry characteristics on the differences between adjusted and non-adjusted House Price Indices (HPIs).
Demonstrated geographic variations in housing market seasonality and identified areas where seasonal adjustments are most necessary

Sentiment Analysis and Machine Learning Applications

The Price of Short-term Housing: A study of Airbnb on 26 regions in the United States

Wenzhen Lin and Fan Yang

(Journal of Housing Economics, Volume 65, 2024)

Investigated the influence of customer review content and other attributes on Airbnb listing prices across 26 U.S. regions
Combined sentiment analysis and bi-term topic modeling techniques with the hedonic model and quantile regression to detect the effect of review content on Airbnb listing prices
Revealed negative reviews impact listing prices twice as much as positive reviews, particularly for lower-priced rentals
Uncovered that review influence diminishes with distance from the city center and a lag effect on prices

Apply Machine Learning in FHA TOTAL Scorecard

Stephanie Hawke and Wenzhen Lin

(Work in Process, 2025)

Developed a comprehensive study applying six machine learning techniques—Logit, Stepwise Regression, XGBoost, LightGBM, Random Forest, and SVM—to predict early default probabilities using FHA mortgage data.
Enhanced predictive accuracy and risk assessment in mortgage lending, contributing to more informed decision-making in financial institutions.

Page updated

Google Sites

Report abuse