The Racially Disparate Influence of Filing Fees on Eviction Rates

Abstract Eviction is a common and consequential event in the lives of tenants and is shaped by the legal environments in which it takes place. In this study, we show that eviction filing fees, or the amounts of money it costs landlords to begin formal evictions, have a large effect on eviction practices. Specifically, fees that are higher by $76 (one standard deviation) lead to lower eviction filing rates by 1.71 percentage points (0.26 standard deviations) and lower eviction judgment rates by 0.49 percentage points (0.19 standard deviation). Filing fees affect not only the rate but also the purpose of filing, as lower fees make landlords more likely to file serially against the same tenants as a form of rent collection. Each of these effects appears to be disproportionately large in majority-Black tracts, suggesting that low filing fees have disparate impacts on Black renters. These findings contribute to our understanding of the legal basis of housing insecurity and the racialization of eviction practices in the United States.

Reflecting the racialization of eviction practices, some of these policies appear to have disparate effects for tenants of different races (Lempert & Monsma, 1994;Merritt & Farnworth, 2021).Despite this wealth of research, important policies remain unexplored, in particular how differences between jurisdictions in the legal eviction process shape landlords' eviction practices (Nelson, Garboden, et al., 2021).Demonstrating the importance of such procedural differences, Gromis et al. (2022) find that the simple requirement that landlords notify tenants before filing evictions has a pronounced impact on filing rates.Another key aspect of eviction procedure, which studies have discussed (Leung et al., 2021;Nelson, Garboden, et al., 2021) but never fully examined, is the cost of filing an eviction in court.
This study examines how filing fees affect the prevalence and uses of eviction filings.Drawing on a data set containing county-level filing fees collected online from county legal codes, we find that the amount of the filing fee is one of the strongest predictors of eviction, and fees higher by $76 (one standard deviation) lead to eviction filing and judgment rates that are 1.71 and 0.49 percentage points lower (0.26 and 0.19 standard deviations), respectively.Differences in filing fees also affect the prevalence of serial eviction filings, with $76 higher fees leading to lower serial filing rates by 3.11 percentage points (0.28 standard deviations).These effects are particularly pronounced in majority-Black neighborhoods, where they are two to four times larger than those in majority-white tracts.These findings make two contributions to our understanding of housing insecurity, urban governance, and racial inequality.First, they demonstrate that eviction filing fees, and procedural eviction policies more generally, play a key role in determining eviction practices.Second, they show how a lack of renter protections can exacerbate the racial inequality in rental market outcomes already present in racially segmented rental markets.

Eviction Trends, Causes, and Consequences
In the last 30 years, rental housing has become increasingly unaffordable and precarious.Between 1990 and 2010, average rent increased by 15% in real dollars while average tenant income did not increase at all (Collinson, 2011;Desmond, 2018).During the same period, the American renting population grew considerably, from 34.0 to 43.9 million households between 1999 and 2015 (Kingsley, 2017).As the need for rental assistance has skyrocketed, government subsidies have not kept pace (Kingsley, 2017), with the yearly number of new subsidies falling from over 160,000 in the 1980s to less than 3,000 in the late 1990s and early 2000s (Goetz, 2013;Schwartz, 2014).As a result, rent burden, or the percentage of income a household spends on rent, has increased dramatically.Although 30% of income has long been considered the threshold above which rent becomes fiscally burdensome, nearly half of all renters today pay more than this share, and around a quarter pay more than 50% of their income (Joint Center for Housing Studies, 2020).
Unsurprisingly, eviction has become a common occurrence in the lives of poor renters.On average, more than 3.6 million evictions are filed each year (Gromis et al., 2022), and more than one in seven children born in a city between 1998 and 2000 was evicted by age 15 (Lundberg & Donnelly, 2019).Eviction has negative consequences for those tenants who are forcibly displaced.Evicted households experience worsened material hardship, mental and physical health problems, increased job loss, and greater housing insecurity in subsequent months (Collinson & Reed, 2018;Desmond & Kimbro, 2015).For these reasons, Desmond (2016) calls eviction a cause as well as a consequence of poverty.Eviction may also have consequences that extend beyond the targeted household, affecting entire communities.Recent studies suggest that high-eviction communities have higher rates of crime (Alm, 2018;Alm & B€ ackman, 2022) and lower rates of community involvement (Holm & Monaghan, 2021), compared to otherwise-similar neighborhoods, although the causal link is not entirely clear.
Eviction filings are also used for tenant discipline and rent collection, rather than removing tenants from their homes.This practice, often called "serial filing" (Garboden & Rosen, 2019) or "serial eviction" (Leung et al., 2021), entails landlords repeatedly filing for eviction against latepaying tenants, in order to procure legally binding judgments that compel tenants to pay late rent and court fees, but which do not remove them from the property.Landlords do this because filing for eviction alters the legal relationship between them and their tenants from one of equal parties in a transaction to one of creditor and debtor (Garboden & Rosen, 2019).The resulting asymmetry provides the landlord with considerable legal benefits, such as the ability to garnish a tenant's wages, as well as an expedited path to eviction if the tenant does not eventually pay the back rent.Repeated filings also burden tenants with additional late fees and court fees, increasing their housing costs by an average of 20% (Leung et al., 2021), and mark them with an eviction record (So, 2022).
Individual-, neighborhood-, and landlord-level factors shape a tenant's risk of eviction.Black and Hispanic tenants, in particular women, experience dramatically higher rates of eviction and serial eviction filings than white and Asian renters do (Desmond, 2012;Hepburn et al., 2020), and neighborhoods containing larger shares of Black and poor residents have much higher rates of eviction as well (Lens et al., 2020;Nelson, Gromis, et al., 2021).Beyond demographic factors, the number of children in a household and recent job loss predict eviction at the household level (Desmond & Gershenson, 2017), and large-scale, corporate landlords evict at higher rates (Decker, 2023;Raymond et al., 2018) and are more likely to file serially (Gomory, 2022;Immergluck et al., 2020) even after controlling for property characteristics.
Reflecting the combination of individual, neighborhood, and landlord factors, segmented rental submarkets with disparate eviction practices often exist within the same city (Teresa & Howell, 2021).Many high-poverty, majority-Black or Hispanic areas function as high-evicting submarkets that house renters whose residential options are limited by a combination of racial discrimination, low incomes, eviction histories, and low credit scores.Landlords take advantage of these tenants' limited options to rent them undermaintained, dilapidated dwellings, using eviction as a first resort for late-paying tenants and a key part of their high-turnover business models (Gomory & Desmond, 2023;Teresa & Howell, 2021).The disparities in eviction practices between these and other areas reflect one way in which distinct norms of economic behavior develop within racialized, segmented submarkets (Fields & Raymond, 2021;Taylor, 2019).

The Policies Shaping Eviction Practices
Although tenant, landlord, and neighborhood characteristics are important proximate determinants of eviction patterns, they can only go so far in explaining the eviction rates in different areas.Often, communities with similar characteristics have highly disparate eviction rates, due to legal structures and housing policies that put tenants at different levels of risk (Gromis et al., 2022;Nelson, Garboden, et al., 2021).In their 2015 review of legal studies of housing, Desmond and Bell argue that studies of private rental market regulations represent an important gap in the literature, and in the last several years, a number of studies have begun to fill this gap.
Unsurprisingly, public policies that address the affordability of renting also affect eviction prevalence.Rental subsidies, in particular those that offer deep subsidies that make rent payments affordable to very low-income tenants, such as public housing and project-based Section 8, have been shown to reduce eviction rates for subsidized tenants (Harrison et al., 2021;Preston & Reina, 2021).Dawkins (2022) argues that restrictive land-use zoning increases eviction rates by limiting housing supply and increasing rents, and Miller et al. (2021) show that Medicaid expansion under the Affordable Care Act reduced eviction rates for those who enrolled.
A large number of recent studies have examined the effectiveness of different eviction regulations during the COVID-19 crisis, finding that the multiple moratoria (Cowin et al., 2020;Haas, 2021), as well as unemployment insurance and stimulus payments (Manville et al., 2022), were effective in reducing eviction filings in 2020 and 2021.Hepburn et al. (2021) estimate that these policies resulted in 1.55 million or 65% fewer eviction cases between March 15 and December 31, 2020, than would be expected in a typical year.Benfer et al. (2022) examine differences in the specific provisions of state-level eviction moratoria, finding that those that halted evictions earlier in the process were more effective in reducing eviction case counts.
Focusing on the regulatory environments that shape eviction filings and landlord-tenant relationships more broadly, Hatch (2017) collected data on a wide range of landlord-tenant laws, such as those pertaining to rental default periods, the warranty of habitability, and retaliation against tenants, categorizing states as probusiness, protectionist, or contradictory.Merritt and Farnworth (2021) examine whether these categorizations predict differences in eviction filing rates, finding that protectionist states have the lowest rates, net of controls, whereas, perhaps surprisingly, contradictory regimes have the highest rates.They also find that the associations between state-level regimes and eviction rates vary for areas with different racial compositions, arguing that the effects of regulatory policy differ by racial group.
Particular scholarly attention has been devoted to courtroom dynamics and legal representation.Housing courts operate with a presumptive judgment in favor of landlords and it is very difficult for tenants, particularly those from marginalized backgrounds, to make their cases heard, even when they have a legitimate legal standing (Bezdek, 1991;Lempert & Monsma, 1994).One key aspect of this unequal power dynamic is that although the vast majority of landlords are represented by lawyers, only a very small percentage of tenants have legal representation (Petersen, 2020), because the right to an attorney only applies to criminal cases.Studies show that legal representation has a substantial positive effect on tenant outcomes, increasing the chance that tenants are allowed to stay in their homes and that landlords are compelled to make repairs or allow rent abatements (Ellen et al., 2021;Seron et al., 2001).Even simple legal advice educating tenants on self-advocacy can increase tenants' legal success (Golio et al., 2022).
Most directly related to the focus of this article, two studies show that differences in the specific procedures governing eviction cases affect eviction practices.Nelson, Garboden, et al. (2021) use ethnographic data to contrast the legal eviction processes in four areas, arguing that differences, such as the length of the eviction process from start to finish, play decisive roles in determining how often and for what reasons landlords file evictions.In line with this argument, Gromis et al. (2022) show that policies requiring landlords to notify tenants before beginning eviction filings decrease filing rates in those areas substantially.One key policy factor that Nelson, Garboden, et al. (2021) identify is the cost of filing an eviction, noting in an exploratory analysis that filing fees are negatively associated with eviction rates in the 50 largest American cities but not formally estimating the effects of these fees.In this paper, we empirically investigate how filing fees change landlords' eviction practices.

The Potential Impacts of Eviction Filing Fees
Policy analysts have long shown that behavior can be incentivized or disincentivized with fees or taxes that affect the cost of that behavior (Brooks, 2007;Wu, 2005).Accordingly, higher filing fees might alter landlords' eviction practices in several ways, depending on how landlords make eviction decisions.For some landlords, eviction is a last resort that they avail themselves of only after other methods have failed and a tenant has missed multiple payments (Balzarini & Boyd, 2021;Garboden et al., 2018).For this group, differences in filing fees may be insignificant compared to the back rent owed by tenants and may not have a large impact on their eviction practices.However, many landlords file over small sums of money, often only a few hundred dollars (Hepburn et al., forthcoming;Teresa & Howell, 2021).These landlords use eviction as a first resort, filing as soon as tenants fall behind, whether to compel late rent payments, gain legal power over their tenants, or actually remove them from the property (Decker, 2023;Garboden & Rosen, 2019).For these landlords, even small increases in filing fees may make it unprofitable to file as readily as they otherwise would, particularly if large numbers of tenants are late each month.For this reason, higher filing fees may particularly disincentivize serial filing, because the practice involves filing repeatedly against the same tenants and even a small increase in the cost will add up over time and may outweigh the potential benefits.
However, it is important to remember that the decision to evict is not made on purely economic grounds.It is very difficult for landlords to accurately weigh the costs of a vacant unit, renovations, and possible court costs against lost income from missed rent (Balzarini & Boyd, 2021;Garboden et al., 2018).Accordingly, the tangible indicator of an upfront filing fee may have an outsized impact on eviction behavior, and even in instances where a difference in filing fees would not substantially change the economic logic of evicting, it may lead landlords not to file or to defer filing longer than they would otherwise.
Because of the distinct patterns of evicting in majority-Black and Hispanic neighborhoods, differences in filing fees may have disproportionate impacts in these areas.As discussed above, landlords in racially segmented housing markets more often adopt a strategy of frequent eviction filings, in which they make few efforts to work with tenants who are behind on rent (Teresa & Howell, 2021).Eviction filings, whether to compel rent payment or remove tenants, become a part of everyday property management, rather than an aberrant event (Desmond, 2016).Accordingly, low filing fees may have magnified effects in these neighborhoods, where the existing rental market structures already make landlords more likely to evict.Such a finding would provide evidence of how formally race-blind public policies can have racially disparate effects (Pager & Shepherd, 2008).
Although we hypothesize that filing fees will alter landlords' eviction behaviors, it is unclear how large the effect will be.In our sample, the standard deviation for filing fees is only $76, and such a small sum may be insignificant compared to other economic considerations, such as the amount of back rent tenants often owe and the costs of finding new tenants.Alternatively, if filing fees have a large effect on eviction practices, changing filing fees could be a relatively easy policy change for local governments to implement in order to reduce housing insecurity in their jurisdictions.However, it is important to note that insofar as increasing filing fees may reduce formal evictions, it may also incentivize informal evictions (Nelson, Garboden, et al., 2021).Informal evictions are any means through which landlords remove tenants from their properties without using the court system, including threatening the tenant with formal eviction if they do not leave or simply telling the tenant that they must leave the property (Desmond, 2016;Desmond & Shollenberger, 2015).Because, by definition, informal evictions do not entail filing fees, higher costs for formal evictions may indirectly incentivize such tactics.

Data and Measures
We obtained information on the costs and requirements for filing evictions, which vary county by county, from online legal records and correspondence with local court officials during the summer of 2018.In this way, we were able to obtain the dollar filing fee for 98.7% of all US census tracts.Although we collected the filing fee data in mid-2018, these fees change rarely, and we assume that they were in place throughout 2018, when the eviction filings we analyze were occurring.We validated this assumption by re-collecting filing fees for a random 5% of counties in our sample in late 2022, more than 4 years after they were originally collected.In this followup, more than 90% of counties had filing fees that were either the same as they were in 2018 or had increased by a small sum (less than $10).This follow-up suggests that in the vast majority of years, filing fees either do not change or increase by small amounts to keep pace with inflation.
We obtained rates of eviction filings and judgments at the census tract level from Princeton's Eviction Lab, which purchased them from LexisNexis Risk Solutions, which in turn collected them from local housing courts.An eviction filing is typically the first recorded step in the legal process of eviction, and a judgment indicates a legal closure to the case.Although the precise legal significance of an eviction judgment varies from jurisdiction to jurisdiction, it is a strong indicator that the tenant was removed from the rental property.Lab personnel prepared these data for analysis by cleaning and geocoding (for details see Gromis et al., 2022).The data are national in scope, covering the years 2000 to 2018, and we use the 2018 data because our filing fee measurements are from that year.The data quality varies, and the Eviction Lab provides variables indicating how similar the counts are to an external comparison data set, where such external data sets are available, and to a predicted filing count, where external comparisons are unavailable.
We consider a tract to have valid eviction data if its observed count is at least half the count in the comparison data set.This threshold was chosen to ensure both adequate data quality and sufficient sample size, after testing other more stringent data quality thresholds and finding that they show similar results.Analyses with more stringent thresholds are included in the Supplementary Appendix as robustness checks.We include all tracts that have valid eviction filing data in 2018.We calculate the rate of serial filing by subtracting the number of households that were ever filed against from the total number of filings, which leaves the number of filings that were made against households that had already received a filing in 2018.We then divide that by the total number of filings and multiply by 100, creating the percentage of filings that were against households that had been filed against previously in 2018.
We also collected measures of federal rental subsidies from the Housing and Urban Development Data Portal.These describe the number of rental units in each tract that are subsidized via public housing, Housing Choice Vouchers, or another program, in 2018.To operationalize government regulations other than filing fees that might affect eviction rates, we draw on Hatch's (2017) typology of state-level renter protections (for other uses of this data see Dawkins, 2022;Merritt & Farnworth, 2021).Hatch collects information on 22 types of renter protections and uses cluster analysis to develop three categories of rental market policies, labeled protectionist, contradictory, and probusiness.These data include a wide range of policies, many of which directly or indirectly affect eviction practices, such as those pertaining to rent default time and rent withholding, but they do not include filing fees.This allows us to control for policies other than filing fees that might affect eviction practices, while estimating the effects of filing fees.We obtained these state-level measures from figure 5 in Hatch's study.
We collected a range of tract-, county-, and state-level control variables from the American Community Survey 5-year estimates centered around 2015 that measure racial composition, socioeconomic status, household composition, housing characteristics, and other demographic characteristics.To examine whether filing fee levels vary by political climate, we downloaded data on the number of votes cast for each candidate in the 2016 presidential election, by county, from the Harvard Dataverse (MIT Election Data and Science Lab, 2018).
Our data set includes all census tracts in the United States with valid filing fee and demographic data, as well as valid eviction filing rates in 2018.This results in a data set of 34,727 census tracts in 1,317 counties and 42 states, covering about 49.7% of the US population.We chose to use census tracts as our unit of analysis, although filing fees are set at the county level, to better control for other determinants of eviction practices, discussed below.Figure 1 shows a map of counties in our sample, shaded to reflect filing fees, as well as a histogram describing the distribution of filing fees at the tract level.Filing fees in our sample vary from $15 to $350, with a standard deviation of $76.The breadth of this data set provides considerable variation in filing fees, allowing us to examine the legal predictors of eviction in a range of contexts.For an analysis of missing data, see Supplementary Appendix E.

Analytic Strategy
Identification Strategy -Determinants of and Variation in Filing Fee Levels The goal of this study is to estimate the effect that filing fees have on the rate at which landlords file evictions, reach eviction judgments, and file serially against the same tenants, as well as the extent to which these effects vary by neighborhood racial composition.Despite having only 1 year of filing fee data, we argue that we are able to identify these effects because filing fee levels are not associated with any known determinants of eviction practices or even with the broader social conditions of the areas to which they pertain.Although such randomness is not strictly necessary for identification (assuming we are able to properly control for relevant covariates), the arbitrariness of filing fee levels supports our identification assumptions by suggesting that there are few unobserved confounding processes and aids in estimation by easing the task of controlling for relevant covariates.
Because eviction filing fees are often determined at the state level, the majority of variation in filing fee levels, 87%, exists between rather than within states. 1 At the state level, eviction filing fees do not follow a strong regional or political pattern, as shown by State-level demographic characteristics are not associated with filing fee levels either.Supplementary Figure A.1.2shows scatter plots of average filing fees with state-level proportions of white, Black, and Hispanic residents, household median incomes, poverty rates, and population densities.The maximum correlation with filing fees is 0.20, and none of the variables are significant predictors in univariate linear regression models.
Although state-level filing fees may be uncorrelated with demographics and political contexts, it is possible that within states, counties' demographic, socioeconomic, or political differences Note.County-level filing fees are for all counties in our sample.Red indicates high filing fees whereas white indicates low fees.
shape their filing fees levels.We analyze the 12 states for which there is substantial filing fee variation (defined as a standard deviation of $10 or more at the county level): Alabama, Georgia, Illinois, Indiana, Kentucky, Massachusetts, Missouri, Nevada, Ohio, Pennsylvania, Tennessee, and Texas.Supplementary Figures A.2.1 through A.2.7 show scatter plots between county-level filing fees and measures of counties' racial demographics, socioeconomic statuses, population densities, and 2016 presidential election Democratic vote shares.Separate regression lines are drawn for each state to examine whether there are any characteristics that consistently predict filing fees within states.We find that for each variable there are states in which the association with filing fees is positive and others where the association is negative, although in most states the association is small or nonexistent.This demonstrates that in those cases where filing fees are correlated with a particular social characteristic within one state, there is another state in which the opposite relationship exists.For example, the urban counties containing Chicago and St. Louis have the highest filing fees in Illinois and Missouri, but the urban counties containing Philadelphia and Pittsburgh have among the lowest fees in Pennsylvania.Similarly, although heavily Democratic counties in Massachusetts have higher filing fees, in Ohio Democratic counties have slightly lower fees.
This lack of correlation between social characteristics and eviction filing fee levels is due to the often-arbitrary ways in which filing fees are determined.In many instances, eviction filing fees are not distinguished from fees for other civil suits.For example, in Minnesota the same civil filing fee applies to filing an eviction as applies to requesting a name change or filing for adoption.Additionally, civil court fees are typically made up of multiple subcomponents, which are often unrelated to the judicial processes to which they pertain and vary haphazardly between jurisdictions.For example, a report on court fees in Illinois notes that these subcomponents often fund "programs completely unrelated to the administration of justice like roadside memorials and after-school programs," and a report on civil fees in Texas notes that there are many fees for which the evaluators could not determine a purpose (Statutory Court Fee Task Force, 2016, p. 7; see also Office of Court Administration, 2014). 2 For these reasons, eviction filing fees do not appear to be determined by local housing conditions or sociodemographic indicators.
Because filing fees do not reflect the characteristics of the areas to which they pertain, it is unlikely that there are unmeasured social processes confounding the relationship between filing fees and eviction practices, allowing us to assume conditional ignorability between filing fees and our potential outcomes.Equation (1) shows this assumption, where F i is the filing fee for tract i, Y i (f) is the outcome (e.g., the eviction filing rate) for the tract under different filing fees f, and X i is an array of predictors described below.

Control Variables
Although the arbitrariness of filing fee levels suggests that controlling for covariates may be unnecessary, we do so to ensure that we identify the causal effects of filing fee differences and to improve the precision of our estimates.At the tract level, we control for a range of demographic and rental market characteristics that could affect eviction practices.First, we control for the racial composition of each tract, specifically the proportion of Black residents and the proportion of Hispanic residents.Racial composition is one of the key predictors of eviction rates, with minority, particularly Black, neighborhoods experiencing much higher rates of eviction (Hepburn et al., 2020;Lens et al., 2020).Second, we control for tract socioeconomic status, namely the proportion of residents in poverty, the proportion with bachelors' degrees or higher, and the household median income, because poorer tenants are at greater risk of eviction (Desmond, 2016;Lens et al., 2020).Third, we control for household composition, namely the proportion of family households, the proportion of single-parent households, the proportion of residents that are minors, and the average household size of renter households.Households with children are at increased risk of eviction, and landlords often evict women more readily than men (Desmond, 2012;Desmond & Gershenson, 2017).Fourth, we control for several rental market characteristics, including the rental vacancy rate, the median proportion of household income spent on rent, and the median home value.These operationalize rental market affordability and the degree of demand relative to supply, both of which affect eviction rates.We also control for the proportion of homeowners, the proportion of rental units in single-family properties, and the proportion in multifamily units with two to four units.Areas with fewer homeowners and more large rental buildings are more likely to have large-scale, professional landlords, who evict at higher rates than their small-scale counterparts (Gomory, 2022;Raymond et al., 2018).Finally, we control for the proportion of rental units that are covered by federal rental subsidies, distinguishing between public housing, Housing Choice Vouchers, and other subsidies.Renters in subsidized housing, particularly those in public housing, are less likely to be evicted (Lundberg et al., 2021;Preston & Reina, 2021).
At the county level, we control for socioeconomic status, family composition, rental market affordability, and the built environment using composite variables, each of which is the first factor from a principal component analysis.Socioeconomic status is calculated using county-level household median income, poverty rate, unemployment rate, use of public assistance, and proportion of residents with bachelors' degrees, with higher values indicating higher SES.Family composition is calculated from the proportion of owner-occupied households, proportion of family households, and proportion of residents under 18, with higher values indicating more families and children.Rental market affordability is calculated using median rent burden, proportion of units that rent for under 30% of median household income, and median rent, with higher values indicating more affordability.The built environment measure is calculated using the proportion of units in single-family, two-to four-unit, and five-plus-unit properties, and the average renters' household size, with larger values indicating larger properties.
At the state level, we control for Hatch's categorical measures of rental market protections.These are well suited for our purposes because they use a range of policies that could affect eviction practices to produce their measures, but they do not include eviction filing fees themselves.Accordingly, by controlling for these other types of rental protections, we can better isolate the particular effects of filing fees.
Table 1 shows the values of our covariates for tracts with high and low filing fees, demonstrating a high degree of balance in the sample.High and low filing fee areas have similar racial demographics (16% Black vs. 15%, 60% white vs. 65%, and 17% Hispanic vs. 13%), the same poverty rates (17% vs. 17%), the same household compositions (65% family households vs. 65%), and similar housing market conditions (32% median rent burden vs. 31%, 43% of rental units in single-family properties vs. 45%, 2% in public housing vs. 2%).Accordingly, we assume common support for the treatment among our covariates (see Equation 2), meaning that for all observed combinations of covariates, the probability of a filing fee is not zero for any filing fee value (see Supplementary Appendices B.2 and C.4 for results with different covariate pruning techniques).

Estimation Strategy
We estimate the effects of eviction filing fees by reweighting our sample using coviaratebalanced propensity score weighting for a continuous treatment (Fong et al., 2018), as implemented by Imai and Ratkovic's (2014) R package, and estimating a weighted regression.We do this to ensure that, after applying the weights (W i ), filing fees are ignorable with respect to our potential outcomes (Equation 3).Furthermore, reweighting our data using these weights ensures that each covariate is uncorrelated with the filing fee (Equation 4) (see Supplementary Appendices B.1 and C.3 for details on covariate balance in the weighted sample).Using those weights, we then estimate a weighted OLS model (Equation 5) predicting each outcome (Y i ), using the same array of covariates (X i ) and filing fees (F i ).Using a regression to estimate the conditional association between filing fees and outcomes ensures that any covariate imbalance that remains after reweighting is controlled for.The weights are calculated such that the mean values of covariates in the weighted sample match those in the unweighted sample, meaning our estimand reflects the average treatment effect.
We use these methods to estimate the effects of filing fees on three outcomes.First, we estimate the effect on tract-level filing and judgment rates.Estimating each of these outcomes is important because it allows us to examine whether filing fees affect not only the rate at which landlords use eviction, but also the rate at which they remove tenants.Next, we estimate the effect of filing fees on the serial filing rate.This allows us to see whether low filing fees make landlords more likely to file serially against the same households, which is often a rent collection technique (Garboden & Rosen, 2019;Leung et al., 2021).
In addition to the main effects of filing fees on our outcomes (average treatment effects), we are interested in whether the effects vary for different types of neighborhoods (conditional average treatment effects), in particular whether the effects are larger for minority neighborhoods.To do so, we reestimate our effects for tracts with more than 50% white residents, more than 50% Black residents, and more than 50% Hispanic residents.We do the same for tracts with more than 20% of households in poverty as a comparison.

Alternate Estimation Strategies and Robustness Checks
Although we use inverse covariate-balanced propensity weighting (ICBPW) for a continuous treatment to estimate our main results, we also reestimate them using ICBPW for a binary treatment (see Supplementary Appendix C).This ensures the robustness of our results because the binary method has been used more extensively in past studies (Gormley et al., 2018;Scott, 2015), whereas the continuous version was developed only recently (Fong et al., 2018).We also estimate our results using multilevel regression models (see Supplementary Appendix D), which is useful for two reasons.First, it is possible that disparate results from these racial subsamples could be driven by other factors in those areas, such as higher rates of poverty in majority-Black neighborhoods.Multilevel regression models allow us to ensure that sociodemographic differences are not responsible for any disparate effects we see in racial subsamples, by allowing us to control for interactions between filing fees and a range of other indicators.Multilevel regression models also allow us to estimate the association between filing fees and our results separately for within-state and between-state variation in filing fees.This reduces the chance that an unmeasured confounder is biasing our results because, given that different demographic and political characteristics are associated with filing fee levels within and between states, it is unlikely that a single confounding variable would create spurious correlations between our outcomes and both sources of filing fee variation.
Finally, we conduct several robustness checks for each of our three estimation methods to ensure that they are not the result of arbitrary analytic decisions.We reestimate each with different and more extensive collections of covariates, with different pruning thresholds, and with different eviction data quality thresholds (see Supplementary Appendices B.2, C.4, and D.3).

Modeling the Social Process of Eviction
The Effect of Filing Fees on Eviction Filing Rates We begin by investigating the effect that filing fees have on eviction filing rates.Figure 2 shows that the estimated effect of having $76 (one standard deviation) higher filing fees is 1.71 percentage point lower eviction filing rates (0.26 standard deviations), for the full sample.Figure 2 also reports estimated effects for different subsamples.In majority-white areas, the effect of having a $76 higher filing fee is only a 1.17 percentage point lower filing rate.In contrast, the estimated effect size in majority-Black neighborhoods is more than four times larger: À4.85 percentage points.The estimated effect in majority-Hispanic neighborhoods is À1.47,only slightly larger than that in majority-white neighborhoods.The estimate in high-poverty tracts (more than 20% of households in poverty) is À2.99 percentage points.These results suggest that filing fees have an inordinate effect on eviction filing rates in majority-Black neighborhoods.

The Effect of Filing Fees on Eviction Judgment Rates
Having seen that filing fees affect the rate at which landlords file evictions, next we examine whether they affect the rate of eviction judgments.In contrast to filings, eviction judgments suggest that tenants were removed from their properties, allowing us to see whether filing fees increase renters' rates of legal forced displacement.Figure 3 shows that the estimated causal effect of having a $76 higher filing fee is a 0.49 percentage point lower eviction judgment rate (0.19 standard deviations).Figure 3 also reports estimated effects for different subsamples.In majority-white neighborhoods, a $76 higher filing fee predicts only a 0.32 percentage point lower judgment rate, whereas in majority-Black neighborhoods, the estimated effect is more than four times larger, À1.41 percentage points.In majority-Hispanic neighborhoods, the estimated effect is À0.49 percentage points.In high-poverty areas the estimated effect is À0.88 percentage points.Note.Each point represents the estimated effect of having filing fees that are $76 higher on the tract-level eviction rate, for different subsamples indicated on the x-axis."High-poverty" indicates tracts with a poverty rate of more than 20%.Error bars indicate the 95% confidence interval of each estimate.

The Effect of Filing Fees on Serial Filing
Finally, we estimate the effect that filing fees have on the percentage of filings that are serial, or filed against tenants who already received a filing in the same year.We hypothesize that higher filing fees will discourage serial filing, resulting in a lower serial filing rate.Figure 4 shows that a $76 higher filing fee predicts a 3.11 percentage point lower share of serial filings (0.28 standard deviations).Figure 4 also reports the effects of a $76 higher filing fee on serial filings for tracts with different characteristics.Again, we find a diminished effect in majority-white neighborhoods, only À2.70 percentage points, whereas in majority-Black areas, the estimated effect is À5.32 percentage points.In majority-Hispanic areas the estimated effect is À2.93 points.In high-poverty areas the estimated effect on serial proportion is À3.88 percentage points.

Evidence Concerning Racial Disparities
Although for each of our three outcomes, filing fees have a disproportionately large effect in majority-Black census tracts, these results could be driven by other characteristics of these neighborhoods, such as their heightened poverty rates.However, two results suggest that the large effect sizes are driven by racial composition specifically.First, in the results presented above, high-poverty tracts show substantially smaller effect sizes than majority-Black tracts.Second, in multilevel regression models that control more explicitly for interactions between filing fees and tract characteristics including socioeconomic status, the disparate results for majority-Black areas remain large and significant (see Supplementary Appendix D.2).In fact, the coefficients for interaction terms between the tract-level proportion of Black residents and the filing fee decrease by less than 20% with the addition of controls for interactions with other tract characteristics.Although the results above do not show disproportionate effect sizes in Hispanic tracts, the results from binary ICBPW and multilevel regressions do (see Supplementary Appendices C.2 and D.2).In those analyses, filing fees have much larger effects in majority-Hispanic tracts than in majority-white areas, although not as large as in majority-Black areas.

Robustness Checks
We demonstrate the robustness of our results in several ways.First, we show that the results are robust to different estimation strategies, including binary ICBPW and multilevel regression.Second, we show that the estimates from each technique are not the results of the particular covariates chosen, by reestimating each using different covariates.Third, we show that the results are robust to requiring different levels of covariate balance and common support and to choosing different split points when converting filing fees to binary treatments.Fourth, we show that the results are robust to more stringent data quality standards for the eviction data.All robustness checks are shown in Supplementary Appendices B.2, C.4, and D.3.

Discussion
In the past three decades, rental housing has become increasingly expensive (Collinson, 2011;Kingsley, 2017), leading to more tenants missing rental payments each month.Given the negative consequences of eviction, scholars have investigated policies that can reduce the frequency with which landlords use eviction filings (Gromis et al., 2022;Nelson, Garboden, et al., 2021).Drawing on a national data set containing county-level filing fees, this study examines the effects that filing fees have on landlord eviction behaviors and how those effects vary by neighborhood racial composition.
We find evidence that a $76 higher filing fee (one standard deviation) leads to eviction filing and judgment rates that are lower by approximately 1.71 and 0.49 percentage points, respectively.We also find that a $76 higher filing fee reduces the percentage of filings that are serial by about 3.11 percentage points.Each of these findings is highly robust, showing similar results from three different estimation techniques and under a range of different analytic decisions.These findings suggest that filing fees exert a powerful influence both on how often and for what reasons evictions are filed.When filing fees are low, landlords make frequent use of eviction filings and often file serially against the same tenants to collect rents.However, when fees are high, landlords make use of the court system less often, and when they do, they are more likely to file in order remove tenants than file repeatedly against them to collect rents.
Although the finding that filing fees affect landlord behavior is not entirely surprising, the size of the effects is remarkable.A one standard deviation difference in filing fees results in a 0.19 to 0.28 standard deviation difference in each outcome, suggesting that filing fees are among the most important determinants of landlords' eviction practices (compared to associations from other predictors in regression models in Supplementary Appendix D.2).That a relatively small difference in fees, such as only $76, can have such large effects, is owed to the readiness with which many landlords use eviction filings.Although for some landlords eviction is a last resort, for many others it is their first reaction when tenants are late in paying rent (Decker, 2023;Leung et al., 2021).In some cases, these quick-filing landlords do so to compel rent payments or gain legal power over tenants (Garboden & Rosen, 2019;Leung et al., 2021), whereas for others, frequent filing is part of a high-turnover business strategy (Desmond, 2016;Teresa & Howell, 2021).The prevalence of this style of evicting is evident in the fact that more than 3.6 million evictions are filed each year (Gromis et al., 2022) and approximately one in nine is filed for less than $500 (Hepburn et al., forthcoming).
Higher filing fees appear to disrupt the eviction practices of these frequent-filing landlords.For landlords who file serially for rent collection, an additional cost of $76 may outweigh the benefit of filing.For example, if a landlord has 100 occupied units and 15% are late each month, a $60 filing fee means $900 in fees each month, whereas a $136 fee is over $2,000 per month.Studies of serial filings echo this logic, noting that these serial-filing strategies appear to be more viable in low-fee areas (Garboden & Rosen, 2019;Leung et al., 2021).Similarly, for landlords who file frequently as part of an extractive, high-turnover strategy, additional filing fees may constitute one of their larger expenses.These landlords are often called "bleeders" or "milkers" (Mallach, 2014;Stegman, 1972) because they remain profitable by neglecting their properties, spending very little money or time on maintenance and other aspects of property management (Desmond, 2016;Teresa & Howell, 2021).Accordingly, increases in filing fees may constitute an outsized portion of their total expenditures. 3 We also find evidence that filing fees have larger effects in areas with larger Black populations.Specifically, the effects of filing fees on eviction filing, eviction judgment, and serial filing rates appear to be 357%, 361%, and 107% larger, respectively, in tracts with a majority of Black residents, compared to tracts with a majority of white residents.As discussed above, the effect sizes in high-poverty areas and results from multilevel regression suggest that this is not a result of higher poverty rates in those areas, but instead a direct consequence of racial composition.In majority-Hispanic neighborhoods, the main results showed similar effect sizes to those in majority-white neighborhoods, but alternate estimation strategies showed disproportionate effects (see Supplementary Appendices C.2 and D.2).
The disparate effect sizes in majority-Black neighborhoods are likely a consequence of the more eviction-prone strategies that landlords adopt in racially segmented housing markets (Teresa & Howell, 2021).Regardless of filing fee, landlords file serially against Black tenants more often (Hepburn et al., 2020), and landlords in majority-Black neighborhoods more often employ high-turnover milking strategies (Desmond, 2016;Taylor, 2019).Accordingly, in areas where these racialized norms of property management predominate, low filing fees have a particularly pernicious effect.
Although we have discussed filing to collect rent and filing to evict as separate strategies, in reality they cannot be so easily disentangled, and by discouraging rent collection filings, higher filing fees may also reduce the frequency with which tenants are removed from their homes.Even when landlords file evictions as a means of collecting rent or gaining legal power over their tenants, often the ultimate result is the removal of the tenant from the property, because tenants often do not appear in court (Seron et al., 2001), leading to a default judgment against them, and, in other cases, the mere act of filing can sour the landlord-tenant relationship in an irrevocable way (Garboden et al., 2018).
Nevertheless, this study has several limitations.Most importantly, because we lack longitudinal data or instances of exogenous changes in filing fees, we are limited to a cross-sectional analysis of a single year of data.Although our analysis cannot definitively ascribe causality, we believe the evidence for a causal effect of filing fees on landlord behaviors is strong.First, filing fees show no large correlations with known predictors of eviction practices or broader demographic conditions, either between or within states.Second, this lack of correlation is explained by our analysis of the determinants of filing fee levels, which suggests that filing fees are determined in complex and bureaucratic ways that do not reflect social conditions in their areas (Office of Court Administration, 2014;Statutory Court Fee Task Force, 2016).Third, a formal analysis of possible unobserved confounding (see Supplementary Appendix F) shows that including covariates in our analyses actually increases the magnitude of our estimated effects, suggesting unobserved confounding is not producing our results.Relatedly, the associations we identify are substantively large, suggesting any confounding influence would have to be large as well.Fourth, analyses that draw on variation in filing fees between states show similar results to analyses using variation within states (see Supplementary Appendix D.2), suggesting any confounding variable would have to be operating in both contexts.Nevertheless, further research leveraging longitudinal or exogenous changes in filing fees would contribute to our understanding of how filing fees shape eviction patterns.
A further limitation of our analyses is that, because we do not have data on informal evictions outside of the courts, we are unable to determine whether decreases in formal eviction are compensated for by increases in informal eviction.Further research could examine whether, in high filing fee environments, landlords are more likely to evict tenants informally, which can entail shutting off tenants' utilities or refusing to make needed repairs, rather than using the court system.However, even if that is the case, differences in the use of formal evictions are important because eviction filings alter the landlord-tenant relationship (Garboden & Rosen, 2019), entail additional fees for the tenant (Leung et al., 2021), and mark the tenant with an eviction record (So, 2022).Furthermore, because informal evictions are more oriented toward removing tenants than collecting rents, decreases in rent collection filings due to heightened filing fees are unlikely to be fully compensated for by increases in informal evictions.
Bearing these caveats in mind, this study makes two contributions to our understanding of the relationship between the regulatory environment, residential insecurity, and the racialization of rental markets.First, our findings demonstrate that the specific regulations concerning eviction procedures are important determinants of renters' residential insecurity and housing instability.Filing fee is easily one of the strongest predictors of eviction filing, eviction judgment, and serial filing rates, similar in magnitude to many indicators of socioeconomic status, some of the largest predictors in the literature (Lens et al., 2020;Nelson, Gromis, et al., 2021).This complements work by Gromis et al. (2022) that shows the impact of regulations requiring landlords to notify tenants before filing and answers a call by Nelson, Garboden, et al. (2021) for more research into the regulations governing eviction procedures.
Second, our results showing pronounced effects in Black neighborhoods demonstrate the way in which formally race-blind policies can have racially disparate effects.The racial segmentation of American housing markets creates structural incentives, complemented by racialized ideologies, that lead landlords to manage their properties differently in neighborhoods with different racial compositions (Fields & Raymond, 2021;Taylor, 2019).The results of this study suggest that racialized differences in landlord practices are magnified in rental market contexts where there are few barriers to evicting.Thus, a policy that on its face does not mention race has highly discriminatory impacts (Pager & Shepherd, 2008).The current findings complement prior work on eviction dynamics in racialized housing markets (Teresa & Howell, 2021) by demonstrating how legal environments can magnify or diminish differences in landlord behavior between submarkets.
Finally, this study has clear policy implications, suggesting that if county officials were to increase their filing fees, they would decrease the rate of eviction filings and judgments in their areas, improving residential stability for renters.The findings suggest that such a legislative change would have its largest deterrent effect on serial filings, discouraging a pernicious practice that marks tenants with an eviction filing, increases their housing costs, and increases the power that landlords have over their tenants (Garboden & Rosen, 2019;Leung et al., 2021).Furthermore, these increases in residential stability would have their largest effects on Black renters, helping to remedy the disproportionate levels of eviction experienced by this group.Finally, filing fees are a strategic point at which to intervene, because among other predictors of eviction rates with similar magnitudes, it is by far the easiest for policymakers to change and existing fees are often set in arbitrary, bureaucratic ways.Although local politicians cannot easily improve the socioeconomic statuses of their city's renters, they can achieve similar gains in residential stability by simply voting in their city council or state legislature to increase eviction filing fees.
Nevertheless, such policy changes should consider the potential indirect impacts of increasing filing fees on informal filings and take steps to ensure that landlords do not compensate for reductions in formal filings with increases in informal evictions.For example, municipalities could increase their oversight of utility shutoffs to ensure such tactics are not being used to pressure Figure A.1.1 in the Supplementary Appendix, which presents scatter plots of state-level filing fees by region and by the proportion of Democratic voters in the 2016 presidential election.Alabama and Mississippi, Southern Republican states, have respective average filing fees of $276 and $65, and Vermont and New York, Northeastern, Democratic states, have average filing fees of $295 and $45, respectively.Filing fee levels are not correlated with broader renter protections, either, as demonstrated by Supplementary Figure A.1.1, which shows a scatter plot between Hatch's (2017) rental protection categories and filing fee levels.Maryland and California, both of which are protectionist, have average filing fees of $46 and $241, respectively, and Michigan and Illinois, both of which are probusiness, have filing fees of $55 and $249, respectively.

Figure 1 .
Figure 1.County-level map and tract-level histogram of filing fees.

Figure 2 .
Figure 2. Estimates of the effect of filing fee difference on eviction filing rate.Note.Each point represents the estimated effect of having filing fees higher by $76 on the tract-level eviction filing rate, for different subsamples indicated on the x-axis."High-poverty" indicates tracts with a poverty rate of more than 20%.Error bars indicate the 95% confidence interval of each estimate.

Figure 3 .
Figure 3. Estimates of the effect of filing fee difference on eviction rate.

Figure 4 .
Figure 4. Estimates of effect heterogeneity on serial proportion from subsamples.Note.Each point represents the estimated effect of having filing fees that are $76 higher on the tract-level serial filing rate, for different subsamples indicated on the x-axis."High-poverty" indicates tracts with a poverty rate of more than 20%.Error bars indicate the 95% confidence interval of each estimate.

Table 1 .
Descriptive statistics for high-and low-filing fee tracts.

Table 1 .
Continued.The table shows the mean and standard deviation (SD) of key variables for tracts below and above the median filing fee of $121, as well as the difference between the means, in standard deviations.