Patrick Power

Shomik Ghosh

Markus Schwedeler

Do Not Cite

The Right to Counsel at scale

The economics of Housing & Homelessness

Overview

Research Question: Impact of legal aid in eviction cases on housing instability

Motivation

Slide on Affordability

Slide on Noted Costs of Eviction

  • 30 day notice of Evictions for tenants receiving housing subsidies
  • Five day grace periods on late rent

Increased Housing Legislation

  • " 'Band-aid' for bad substantive law"
  • Targeted "type of antipoverty program"

Motivations

(0) Do tenants have a strong defense?

The Essential Questions have been asked for decades

(1) Does legal aid help tenants win in housing court?

(2) Does legal aid reduce the likelihood of homelessness, and improve housing stability?

(3) What are the costs to landlords (and are they passed on to other individuals?)

(4) Does the effectiveness of this policy diminish if tenants anticipate support?

This Note will examine the impact of one Legal Services Program, the New Haven Legal Assistance Association (LAA) on landlord-tenant disputes. The Note's findings on LAA's impact on this one area of law are in no way conclusive as to its general performance or to the overall value of Legal Services in the nation. Nevertheless, they may suggest questions for further study in other programs and other types of litigation.

- Legal Services and Landlord-Tenant Litigation: A Critical Analysis (1973)

Gunn 1995

Do tenants really have a defense?

Challenges of Empirical Work in this Setting

  • Housing Court Outcomes are not always informative

What does it mean when a case is withdrawn?

  • Housing court files contain little information on tenant defenses

Data

LANDLORD COMPLAINTS

Background

The (Eviction) Rental Market

Notice to Quit

Summons

Complaint

Appearance

Answer

Landlord

Tenant

Do tenants have a strong defense?

We don't observe an Answer for the Majority of Cases

Have Second LLM explain scoring

Implementation

Evans et al. (2024)

Literature Review

"The primary factor leading to this differential appears to be LAA's use of the procedural complexities available in summary process litigation. However, despite such efforts, the landlord almost inevitably obtains judgment of possession."

Interpretation

These are what Gelman and Imbens (2013) would refer to as "Reverse Causal Questions"

We don't know "why" tenants with substantive defenses took the least amount of time to resolve. It could be because they didn't want to stay in their units or it could be for some unobserved reason

Gunn 1995

Credibility

"No two eviction stories are quite alike. Each has its own set of challenges, and families cope in different ways."

notation

\textrm{Instrument} := \ \textrm{Tenant Covered by Right to Counsel}
\textrm{Treatment} := \ \textrm{Legal Aid Lawyer}
Z_i
D_i
\textrm{Outcome} := \begin{cases} \textrm{Judgment of Possession}\\ \textrm{Observed Move}\\ \textrm{Change in Poverty Rate} \\ \textrm{Emergency Shelter} \end{cases}
Y_i
\textrm{Controls} := \ \textrm{Details of the Case}
X_i
\textrm{Case} := \ \textrm{Eviction Case Filed}
C_i

Not Difference-in-Difference

"Selecting a pre-treatment period for the analysis is a challenge of this approach due to the substantial volatility in eviction during the COVID pandemic period, when various state and federal policies were implemented and sunsetted, often in an overlapping fashion."

  • A common question we get when discussing our context is -- under what assumptions are you identified? 
  • This we cannot answer in a satisfactory way, but we feel that it's the right thing to focus on nonetheless.
  • We can answer this (and we will answer this), but the better question is to consider why are the estimates that we report credible.

Motivation

Identification Assumption

Controls

  • Month
  • Housing Court
  • Plaintiff has lawyer
  • Gender
  • Race
  • Monthly Rent
  • Zip Code Eviction Filings
  • Landlord Reason for Evictions
  • Tenant Defense
  • Length of Lease
  • Poverty Rate
\tilde{Y}_i \perp Z_i \vert X_i
\tilde{D}_i \perp Z_i \vert X_i

Concerns

Selection Bias

Sufficient Set of Controls?

Information

Is there meaningful variation across Landlords?

Context

To what extent are we picking up the residual effects of the Pandemic?

Attrition

Withdrawn Cases are less likely to have text files

Clusters

Noisy Estimates

Housing Court Outcomes

Focus

Legal Aid Representation

Right to Counsel 

\beta_1 = \int _X \mathbb{E}[D_i \vert Z_i=1, X_i] - \mathbb{E}[D_i \vert Z_i=0, X_i]d\mathbb{P}_X

Nonparametric Model

Approximate Conditional Expectation Functions with Linear Models / LLM / MLP

Models

Residualized Model

D_i = \beta_1 (D_i - \mathbb{E}[D_i \vert X_i]) + \eta_i

Approximate Conditional Expectation Functions with Linear Models / LLM / MLP

\beta_1 = \int _X \frac{{\mathbb{E}[Y_i \vert Z_i=1, X_i] - \mathbb{E}[Y_i \vert Z_i=0, X_i]}}{\mathbb{E}[D_i \vert Z_i=1, X_i] - \mathbb{E}[D_i \vert Z_i=0, X_i]}d\mathbb{P}_X

Nonparametric Model

Focus

Housing Court Outcomes

Legal Aid Representation

Y_i = \beta_1 (\mathbb{E}[D_i \vert X_i, Z_i] - \mathbb{E}[D_i \vert X_i]) + \eta_i

Residualized Model

Model

Housing Stability Outcomes

"After an August eviction, she and Dexter have bounced around. They’ve stayed in motels, at a shelter and in the back of a U-Haul. In her last days at the shelter, one of the other women staying there got an apartment. Austin has been crashing on her couch the past few weeks."

Experiencing Homelessness

Add slide that compares how other papers make this match

Address Changes

Emergency Shelters

Results