Eviction Reduction via Court Policy?
The Difficulty of Reducing Default Evictions in LA County
This post summarizes material from a talk on my law job market paper Eviction Reduction Policies, presented recently at the Canadian Law & Economics Association 2025 Conference.
The public and lawmakers are increasingly concerned with housing policy. According to recent Pew polling, a majority of households across the political spectrum are “very concerned” about the cost of housing. The only issue more concerning is the price of eggs!
In California, policymakers are especially interested in protecting renters at-risk of eviction, which has been linked to numerous harmful effects: substance abuse, adverse physical and mental health outcomes, increased hospital visits and homelessness, disrupted education for children, and long-term economic and social instability.
Given such concern, how should policymakers design legal policies to reduce or prevent evictions?1 The paper looks at one possible reform: expanding the number of eviction courts. The intuition is that we can decrease (default) evictions by lowering the tenant cost of getting to court.
The paper is inspired by a real-world legal challenge to a 2013 change in Los Angeles County court policy. After budget cuts in 2013, the Los Angeles Superior Court system changed the way it assigns eviction cases to courthouses. Moving from a “neighborhood court” model to a “hub” model, the budget cuts reduced the number of eviction courthouses from 26 to only 5. Plaintiff-tenants challenged the hub model on various legal grounds. As explained by the 9th Circuit in Miles v. Wesley (2015), the crux of the claims is that changing to the hub model hurt tenants because it increased the cost of getting to court:
[The neighborhood] model, while costly, provided convenient access to justice for the residents of Los Angeles County because, for most types of cases—whether small claims, traffic, unlawful detainers, or criminal in nature—no one had to travel very far to attend court proceedings.
[B]ecause individuals with disabilities and minorities are disproportionately renters ... the closure of these courtrooms would have a disparate impact on these communities. [T]he importance of neighborhood court access is heightened in light of the expedited timeline of unlawful detainer actions, the fact that most low-income tenants are not represented by counsel, and the prospect that a default judgment could render a tenant homeless.
The idea is that court assignment policy generates localized differences in the tenant cost of getting to court. Consider, for example, the current eviction case assignment rule, shown in the table below for the first few LA County zip codes.
In each of the first three zip codes, cases are sent to different courthouses depending on a tenant’s location within the zip code: different neighborhoods, or even different sides of a street (in 90003), determine the eviction courthouse. This generates different costs to get to court. Florence 90001 cases, for example, face about 29 min by car and 46 min by public transit to the Stanley Mosk courthouse. By contrast, nearby tenants in Huntington Park 90001 may have to travel 41 min by car or 1 hour 15 mins by public transit to reach the Norwalk courthouse.
The plaintiff claim in Miles is that the 2013 court “hub” model exacerbated the tenant cost of getting to court. Unfortunately, data availability issues preclude empirically examining the dramatic 2013 court policy change, so the paper instead looks at the most recent August 2017 change in court policy, when the number of eviction courts increased from 8 to 11. The changes in court assignment are shown below for regions with available data.
The August 2017 policy shock is significant because, like the earlier 2013 assignment reform, it changed the tenant cost of getting to court. Using distance-to-court as a measure of the “cost of getting to court,” the August 2017 shift increased distance-to-court for some tenants, decreased distance-to-court for others, and left many tenant court assignments unchanged.
The idea here is to leverage this policy shock to do causal inference. Essentially, the changes in distance-to-court over time can be used to craft comparison cohorts, which are then used for counterfactual exercises. There are many possible ways to define the cohorts, but perhaps the simplest is to group addresses2 by whether their distance-to-court after August 2017 is the same, greater than, or less than their distance-to-court before August 2017. The cohorts—which I call Control, Increased, and Decreased—are defined in the table below.
On the tenant cost theory, we expect that the August 2017 change increased evictions for tenants with increased distance-to-court and decreased evictions for tenants with decreased distance-to-court.3 This is summarized in the “Expectation” column of the table.
So what happened to the three cohorts after the August 2017 policy shock? Below, I show a before-after comparison of the average number of default evictions for each cohort in bold colors, with more detailed trends shown transparently. The August 2017 policy shock is depicted as the vertical dashed line.
The Increased cohort of addresses, for instance, saw an increase in the average number of default evictions post-August 2017. This is shown by the higher solid green line after the policy shock (i.e. to the right of the dashed vertical line). But care is needed for causal interpretation: the Control cohort also saw an increase in the average number of defaults, even though the Control addresses saw no change in tenant costs (as measured by distance-to-court).4
The paper therefore uses a causal framework (the difference-in-differences strategy) to estimate the effect of the August 2017 reform. The intuition is to use the Control cohort to impute counterfactual outcomes for our two treated cohorts (Increased/Decreased). The key assumption underlying the strategy is that counterfactual outcomes for the treated cohort would’ve evolved “in parallel” to the Controls.
Results from this empirical exercise are shown below.5 The different models listed on the y-axis are further explained in the paper, but they are different ways to define the treated cohorts using different criteria or additional data. For example, the “Commute Times” model defines the treated cohorts using changes in commuting time rather than changes in distance-to-court, whereas the “Above-Median” model only considers addresses with “large” changes in distance-to-court.
Most of the estimates are statistically insignificant, although additional data could help reduce noise. The direction of the estimated effects are also incorrect in some cases (e.g. the baseline Decreased estimate is positive, when we expected it to be negative). But more importantly, in terms of practical legal significance, the CIs imply small effects from this particular intervention: the maximum standardized mean differences (SMDs) are shown next to each point estimate and are considered “small” effects under conventional interpretations.6
Discussion
Does the empirical study imply distance-to-court, tenant costs, court procedural policy, case assignment, etc. are not important? Of course not. This is a reduced-form econometric exercise in one large city from a moderate eviction court reform. And, importantly, the CIs in the three models above do not rule out true effects that are directionally consistent with the tenant cost theory.7
Instead, the results imply that limited reforms may have limited effects. A more drastic change in court policy (e.g. the 2013 fall from 26 courts to only 5) may have a more dramatic effect on the eviction system. Because this study is part of an emerging research agenda by law and economics scholars to understand the impact of various factors—of which court assignment policy is but one factor—on eviction outcomes, I view observational studies as a helpful guide for further research. In particular, lawmakers can use observational causal work to determine the policies with largest potential upsides to investigate further, including via randomized controlled trials (which are underutilized in legal policymaking).
Another possibility is leaving many court policies—the number of courts, the zip-neighborhood assignment rule, etc.—fixed and instead providing tenants guaranteed legal representation. Indeed, many scholars and much eviction advocacy is focused on leveling the playing field by securing a “civil Gideon” right for tenants in housing court. In LA County, representation rates differ substantially between landlords and tenants. In 2019, 87.50% of landlords but only 16.04% of tenants had legal representation among all eviction cases.8 Guaranteed representation can make an important difference, although it may be too expensive for some cities.
A final possibility, though, is targeting some other mechanism that affects eviction rates. An increasingly recognized culprit in worsening the housing crisis is zoning regulation, which in many large U.S. cities may constrain the supply of housing. Lower supply leads to increased rent. Something that is underdiscussed is how zoning reform might lessen eviction problems. Eviction is, after all, primarily a rent problem: in LA City, 94% of eviction notices are issued for non-payment of rent. Lowering rents by increasing housing supply should, ceteris paribus, lower eviction rates.
Especially default evictions, which occur when a tenant fails to answer or does not show up to court and thereby loses by default. Defaults make up a large share of eviction case outcomes.
Which addresses? The paper explores several possible ways to define the address population-of-interest. Once an address population is fixed, a panel of eviction outcomes at each address may be constructed.
Increased or decreased in the “average treatment effect on the treated” sense.
In other words, the simple before-after comparison may not yield causal conclusions, absent support for certain counterfactual assumptions.
These are ATT estimates in address-month units. See paper for further models and additional details.
The maximum SMD implied by the endpoints of the CIs.
In particular, I do not think this work is necessarily inconsistent with prior research on long trips to eviction court in other cities, or the importance of neighborhoods and spatial clustering to eviction proceedings. Again, in some cases the aggregate ATT point estimates are consistent with expectations (even if they are not statistically significant), even though many models cannot pin down the direction of the ATT(g,t) in most periods.
In 2023, 91.79% of landlords and 13.62% of tenants are represented at any time during the case according to the same data. Among defaulting tenants, the representation rates are, as expected, much lower: only 0.80—2.83% of defaulting tenants have legal representation at any point during their case across courthouses. See Eviction Reduction Policies paper.