# Research by Lily Shen

Canonical: https://lilyshenai.github.io/research/ . BibTeX: https://lilyshenai.github.io/papers.bib

## Published articles

### Tenant Rights, Eviction, and Rent Affordability
Coulson, N. E., Le, T., Ortego-Marti, V., & Shen, L. (2025). Tenant Rights, Eviction, and Rent Affordability. Journal of Urban Economics, 147, 103762. https://doi.org/10.1016/j.jue.2025.103762 [ABDC A*]

We use state-level differences in the legal relationship between landlords and tenants to estimate the impact of these differences on housing markets. We construct a search-theoretic model of landlord and tenant search and matching, which predicts that an increase in the cost of eviction reduces the number of evictions, but raises rents and homeless rates, and lowers housing supply and vacancy rates. To test these predictions, we construct an annual index to measure the level of the legal protection of tenant rights in each state. Our instrumental variable results indicate that while a one-unit increase in the Tenant-Right Index reduces the eviction rate by 0.32 percentage points, rental housing is 1.8 percent more expensive in areas where tenants have more protections against landlords. A higher Tenant-Right Index is also associated with a decrease in housing supply and an increase in the homeless rate. Taken together, our findings highlight a significant trade-off between tenant protections and rent affordability.

### The More, the Merrier: Agent Connectedness in Real Estate Trading Networks
Shen, L., & Sun, X. (2025). The More, the Merrier: Agent Connectedness in Real Estate Trading Networks. Journal of Real Estate Research, 47(2), 149-169. https://doi.org/10.1080/08965803.2025.2466957 [ABDC B] American Real Estate Society Best Article Award.

This study applies network theory to examine how real estate agents' connections influence transaction outcomes, using data from over 40,000 sales in Atlanta, GA. It finds that agents with stronger networks achieve higher sales prices, with a one standard deviation increase in connectedness leading to a 2.61% price boost. These well-connected agents sell unique properties faster and more accurately price them, suggesting that their networks enhance market efficiency through improved information flow.

### Information in Financial Contracts: Evidence from Securitization Agreements
Ambrose, B. W., Han, Y., Korgaonkar, S., & Shen, L. (2024). Information in Financial Contracts: Evidence from Securitization Agreements. Journal of Financial and Quantitative Analysis, 59(4), 1692-1725. https://doi.org/10.1017/S0022109023000509 [ABDC A*]

We introduce a novel application of machine learning to compare Pooling and Servicing Agreements (PSAs) that govern commercial mortgage-backed securities (CMBS). In contrast to the view that the PSA is largely boilerplate text, we document substantial variation across PSAs, both within- and across-underwriters and over time. A part of this variation is driven by differences in loan collateral across deals. Additionally, we find that differences in PSAs are correlated with ex-post loan and bond performance. Collectively, our analysis suggests the importance of examining the entire governing document, rather than specific components, when analyzing complex financial securities.

### Past Experiences and Investment Decisions: Evidence from Real Estate Markets
Ambrose, B. W., & Shen, L. (2023). Past Experiences and Investment Decisions: Evidence from Real Estate Markets. Journal of Real Estate Finance and Economics, 66(2), 300-326. https://doi.org/10.1007/s11146-021-09844-2 [ABDC A]

This paper investigates how market participants form their risk perspectives about the real estate market through a sequence of information shocks. Guided by a theoretical Bayesian learning model, we exploit a natural experiment afforded by the fracking boom in Pennsylvania in the late 2000s. We empirically examine whether familiarity with conventional gas exploration affects home-buyers' willingness to pay for houses near fracking wells.

### Cleanliness is Next to Income: The Impact of COVID-19 on Short-Term Rentals
Shen, L., & Wilkoff, S. (2022). Cleanliness is Next to Income: The Impact of COVID-19 on Short-Term Rentals. Journal of Regional Science, 62(3), 799-829. https://doi.org/10.1111/jors.12581 [ABDC A]

We use Airbnb data from Austin, Texas to empirically investigate the impact of the onset of COVID-19 on the short-term rental market. Specifically, we employ a machine learning algorithm to create an extensive cleanliness dictionary to detect whether an Airbnb unit is clean. We use a difference-in-difference specification to value the change in income related to reviewer-perceived cleanliness during the COVID-19 pandemic. We find the following results: First, available listings declined by 25% once the pandemic hit and those that remained lost 22% of their income and had occupancy decrease by 20%. Second, properties that were perceived to be clean increased their income by 17.5% and their occupancy by 16.5%, mitigating the negative shock due to COVID-19. Third, rental prices for clean Airbnb listings did not increase after COVID-19. In addition, we study the interaction of Airbnb supply on the long-term rental market during a market decline.

### The Odd One Out? The Impact of Property Uniqueness on Selling Time and Selling Price
Shen, L., & Springer, T. M. (2022). The Odd One Out? The Impact of Property Uniqueness on Selling Time and Selling Price. Journal of Housing Research, 31(2), 220-240. https://doi.org/10.1080/10527001.2022.2078530 [ABDC B] Best Paper Award, Journal of Housing Research, 2022 volume.

We employ machine learning to develop measures of residential real estate uniqueness from written advertisements. These measures are exogenous from sales prices. We distinguish the effect of market uniqueness (compared to houses for sale at the same time) from the effect of universal uniqueness (compared to houses in the same sub-market) on sale prices and time-on-the-market (TOM). The hedonic models show that a one standard deviation increase in market uniqueness leads to a 13% ($48,490) increase in sale prices at the cost of delaying the transaction for 1.7 days, whereas a one standard deviation increase in universal uniqueness only leads to an 11% ($41,030) increase in sale prices at the cost of delaying the transaction for 3 days. We validated the impact of uniqueness on TOM using two hazard models. Our results highlight the importance of uniqueness and market timing in real estate.

### Information Value of Property Description: A Machine Learning Approach
Shen, L., & Ross, S. L. (2021). Information Value of Property Description: A Machine Learning Approach. Journal of Urban Economics, 121, 103299. https://doi.org/10.1016/j.jue.2020.103299 [ABDC A*]

This paper employs machine learning to quantify the value of "soft" information contained in real estate property descriptions. Textual descriptions contain information that traditional hedonic attributes cannot capture. A one standard deviation increase in the uniqueness of a property based on this "soft" information leads to a 15% increase in property sale price in a hedonic price model and a 10% increase in a repeat sales price model. An annual hedonic price index ignoring our measure of unobserved quality overstates real estate prices by between 10% to 23% and mistimes the stabilization of housing prices following the Great Recession. Similar, but smaller effects, are observed for the repeat sales price index.

### The Double Trigger for Mortgage Default: Evidence from the Fracking Boom
Cunningham, C., Gerardi, K., & Shen, L. (2021). The Double Trigger for Mortgage Default: Evidence from the Fracking Boom. Management Science, 67(6), 3943-3964. https://doi.org/10.1287/mnsc.2020.3647 [ABDC A*]

Our results show that homeowners exposed to fracking were significantly less likely to default on their loans. Instrumenting for fracking increases the magnitude of the effect appreciably, consistent with underlying endogeneity in the location of drilling activity. The fracking measures most associated with the labor-intensive act of creating wells have the largest net effects. Furthermore, we show that fracking activity significantly increased employment growth at the county level. These results are consistent with the double-trigger theory of mortgage default, which contends that liquidity shocks in addition to negative equity drive borrowers to default due to their inability to afford payments or sell their homes.

### Economies of Scale and the Operating Efficiency of REITs: A Revisit
Highfield, M. J., Shen, L., & Springer, T. M. (2021). Economies of Scale and the Operating Efficiency of REITs: A Revisit. Journal of Real Estate Finance and Economics, 62(1), 108-138. https://doi.org/10.1007/s11146-019-09741-9 [ABDC A]

Building on past research in the economies of scale debate, and recognizing the substantial changes in this industry since the turn of the millennium, we use a time-varying stochastic frontier approach to monitor changes in REIT efficiency over the 2001-2015 period. While the stochastic frontier model suggests that the overall level of operating efficiency for US equity REITs is stable, using a series of linear models we document evidence that economies of scale still exist in this industry.

## Under revision

### The Good, the Bad and the Ordinary: Estimating Agent Value-Added using Real Estate Transactions
Cunningham, C., Gerardi, K., & Shen, L. (2024). The Good, the Bad and the Ordinary: Estimating Agent Value-Added using Real Estate Transactions. Revise and resubmit, Management Science. https://www.atlantafed.org/research/publications/wp/2022/09/29/11--estimating-value-added-by-real-estate-agents

We recover measures of real estate agents' value added in the housing transaction process using a detailed Multiple Listing Service (MLS) property transactions database covering three large metro areas. Controlling for property and location characteristics, we find considerable heterogeneity in the sale prices obtained by realtors as well as the time that it takes to complete a sale. Focusing on a sample of repeat sales, we estimate an inter-quartile price range for both listing and buying agents to be approximately 4 to 5 percent. The inter-quartile range for days-on-the-market is between 15 and 25 days. Listing agents that tend to sell for a premium do not take any longer to sell, indicating that agents do not have fixed price posting/acceptance strategies. Agents that sell homes at a premium also tend to pay more when acting as a buyer's agent, which is inconsistent with realtors being good negotiators. Finally, homes sold via a "flat-fee" agency on the MLS obtain prices that are 1 to 5 percent higher compared to those obtained by a traditional listing agent and take no longer to sell.

## Working papers

### Cheap Talk or Priced Information? Listing Language and Housing Transaction Outcomes
Shen, L. (2026). Cheap Talk or Priced Information? Listing Language and Housing Transaction Outcomes. Working paper, SSRN, June 2026. https://papers.ssrn.com/abstract=7097938

Using 63,067 Atlanta MLS transactions from 2007 to 2017, we examine what information listing remarks reveal and where it appears in housing transaction outcomes. An LLM scans the corpus and organizes broad listing language into auditable property and transaction categories. Within search-area-year markets and conditional on a saturated hedonic specification, a one-standard-deviation increase in residual information is associated with 7.7 percent higher sale prices, 7.6 percent higher list prices, and 2.5 fewer days on market. Adverse disclosure is associated with 17.4 percent lower sale prices, 16.7 percent lower list prices, and a 0.7 percent lower sale-to-list ratio. Urgency pressure is associated with 4.7 percent lower sale prices and 4.4 percent lower list prices, but not with a distinct sale-to-list adjustment or shorter marketing time. Market-stress language rises during the housing crisis and is more prevalent when local prices are lower, marketing time is longer, and discounts are larger. Repeat-sale and listing-agent specifications imply more conservative but economically meaningful price differences, and residual information and adverse disclosure retain pricing content after LASSO-selected token controls. The findings show that remarks reveal priced property and transaction information beyond structured MLS fields while identifying the economic categories in which that information appears.

## Grants and research awards

- 2026: American Real Estate Society Kinnard Young Scholar Award (cumulative research contribution)
- 2026: James R. Webb ARES Foundation Global Ambassador Grant
- 2026: Clemson College of Business Senior Scholar Research Excellence Award (2025–26)
- 2025: ARES Best Article Award, Journal of Real Estate Research (agent connectedness paper, with Xiaojin Sun)
- 2025: Homer Hoyt Institute / Weimer School Post-Doctoral Honoree
- 2025: Red Pen Award (best referee), Journal of Real Estate Literature
- 2024: Cocoziello Institute Seed Grant, $85,000 (reshoring and real estate markets, with Penn State coauthors)
- 2024: Clemson University Mid-Career Faculty Grant; featured faculty member at the Clemson Board of Trustees meeting
- 2022: Best Paper Award, Journal of Housing Research (2022 volume); Best Paper Award (3rd place), AREUEA International Conference, Dublin
- 2021: Real Estate Research Institute (RERI) Research Grant
- 2020: Clemson College of Business Dean's Award for Student Engagement
- 2019: American Real Estate Society Manuscript Prize in Real Estate Valuation (single author)
- 2018: AEA CSWEP Economics Fellowship (also 2015, 2014)