Content uploaded by Dan Immergluck
Author content
All content in this area was uploaded by Dan Immergluck on Mar 19, 2019
Content may be subject to copyright.
Urban Studies Institute
Multifamily Evictions, Large Owners, and Serial Filings:
Findings from Metropolitan Atlanta
A Working Paper
by
Dan Immergluck1
Jeff Ernsthausen2
Stephanie Earl3
and
Allison Powell3
March 19, 2019
1 Professor, Urban Studies Institute. Georgia State University, 55 Park Place, 8th floor, Atlanta,
30303. dimmergluck@gsu.edu.
2 Data Reporter, Propublica. 155 Avenue of the Americas, 13th Floor, New York, NY 10013.
3 Urban Studies Institute, Georgia State University.
1
Abstract
Evictions and eviction filings cause substantial harm to lower-income families and
neighborhoods. We examine multifamily eviction filings in the five-county metropolitan Atlanta
area with a rich data set of eviction filings, property characteristics, and ownership information.
We find that eviction filings include many “serial filings,” in which landlords file repeatedly on
the same tenant. The literature suggests that serial filings are aimed less at removing the tenant
and more at disciplining the tenant through state-sanctioned threat of removal. We analyze serial
and nonserial filing rates at the property level, and the share of a property’s filings that are serial
filings (serial share). Regressions on building, location, and neighborhood characteristics reveal
important factors associated with higher serial and nonserial filing rates and serial share. We then
examine the owners of the largest number of multifamily properties in the region and identify
those that exhibit high or low serial filing rates after controlling for building, neighborhood and
location characteristics. We find that the largest owners and larger buildings tend to have high
serial shares. We also find that, while properties in Black neighborhoods have higher nonserial
filing rates, their serial shares are also higher than otherwise similar properties. A key result is
that sales in the prior three years have a significant, nontrivial positive effect on the nonserial
filing rate, suggesting that building sales are significant predictors of rising evictions. Finally,
those few large owners that do have low serial filing shares tend to be located in neighborhoods
that are significantly less Black than the average building. We discuss implications for policy
and further research.
Acknowledgements
We thank the Atlanta Journal Constitution for sharing the eviction and property data used in this
study. All errors, omissions and opinions are solely the responsibility of the authors of this paper.
2
Introduction
Housing stability is critical to lower-income families and communities of color in the
U.S. One manifestation of housing instability and lack of affordability, as well as a long-term
determinant of such instability, are evictions. Evictions are common in many large metropolitan
areas, especially in minority low- or moderate-income neighborhoods Just under 15 percent of
children born in large U.S. cities from 1998 to 2000 experienced at least one eviction before age
15 (Lundberg & Donnelly, 2018). Beyond the formal, court-ordered removal of renters from
their homes, a larger set of eviction processes include informal evictions where tenants are
effectively forced from their homes before a formal eviction process is initiated as well as
departures that occur after the eviction filing but before a formal judgement occurs. What has
also become clear is that many landlords file evictions frequently but without a clear intention to
remove the tenant (Garboden and Rosen, 2018; Leung, Hepburn and Desmond, 2019; Raymond
et al. 2018; Rosen and Garboden, 2018).
Eviction and the threat of eviction are a critical part of the high-levels of housing
instability and the lack of affordability faced by many lower-income renters in the U.S. As the
number of lower-income renters has swelled, the supply of low-cost and subsidized rental
housing has failed to keep up. The supply of low-cost rental housing has been eroded both
through a process of upgrading, in which older units are converted into higher-end housing
through renovation or redevelopment, or through a process of filtering and disinvestment, in
which older properties are not maintained and eventually exit the functional rental stock
(Immergluck, Carpenter, and Lueders, 2018). Meanwhile, federal funding allocated to housing
support for poor households has flat-lined in recent decades as demand by lower-income
3
households has increased. From 2005 to 2015 the share of eligible low-income families receiving
some sort of housing assistance declined from 24 percent to 21 percent (Kingsley, 2017). At the
same time, rents steadily rose as incomes stagnated. In 1999, the US Department of Housing and
Urban Development estimated that 19 percent of renters were severely rent burdened, meaning
they spent more than 50 percent of their income on rent and utilities. By 2015, the proportion of
renters with a greater than 50 percent rent-burdened had risen to 25 percent (Kingsley, 2017).
With strong competition for affordable units, many landlords can afford to turn away prospective
tenants with blemished rental records, and tenants who have experienced evictions, or eviction
filings, may find it difficult to find decent, affordable housing.
In this article, we examine eviction filings in one high-eviction, large metropolitan area,
Atlanta, and seek to understand what factors are associated with buildings that exhibit high serial
eviction filing rates as compared to those that exhibit high nonserial filing rates. Due to our
ability to identify ownership relationships among buildings and to identify the largest owners of
larger multifamily apartment buildings, we are able to examine more closely those owners who
own more than five large buildings in the metropolitan area and more carefully scrutinize their
behavior. We identify 43 such owners in our study area, and together they own over 120,000
rental units, or about one-third of the multifamily units in the region.
After describing the social, health and economic harms that can be caused by evictions
and eviction filings, we review the limited literature on the nature of eviction filing processes and
why landlords may choose to file evictions repeatedly on the same tenant, in a process called
“serial filing.” We then review the literature on neighborhood, household and ownership
characteristics associated with evictions, as well as the one prior study that examined factors
associated with serial filing in particular. Next, we describe our approach to modeling eviction
4
filing behavior at the building level, including our identification of the ownership of properties
with over 50 units. In order to identify the neighborhood, building, and ownership characteristics
associated with eviction filing behavior, we estimate models that explain serial filing, nonserial
filing, and the share of filings that are serial filings. We discuss the results as well as their
implications for policy and further research.
The Harm of Evictions and Eviction Filings
The damaging effects of eviction are numerous and varied, ranging from immediate and
future housing instability, to homelessness, job loss, school turnover, deteriorated health, and
mental illness (Gold, 2016; Raymond et al. 2018). However, it is not only full eviction
judgments and forced removals that can harm tenants. Even when a tenant makes a payment
arrangement or leaves a unit voluntarily before the sheriff arrives, their record is marred by the
eviction (Desmond, 2015; Gold, 2016; Hartman & Robinson, 2003). Previously evicted tenants
that are lucky enough to find housing can be pushed down-market, forced to accept substandard
housing, often in a different and more precarious neighborhood. Renters that experienced a
forced move ended up in neighborhoods with a poverty rate 5.4 percentage points higher and a
crime rate 1.8 percentage points higher than renters that moved voluntarily (Desmond &
Shollenberger, 2015). The forced removal of residents is often endemic in the worst
neighborhoods, further contributing to neighborhood instability (Desmond, 2012). A variety of
research confirms that evictions are concentrated in distressed neighborhoods (Raymond, 2019;
Shelton, 2017; and Thomas, 2017). Worse yet, eviction filings can effectively bar tenants from
future eligibility for subsidized housing, making it even harder to escape the eviction cycle
(Desmond, 2012).
5
Job loss is one important potential outcome of eviction. Tenants need time to search for
new housing and low wage jobs have unforgiving schedules (Desmond, 2015). If new housing is
found, it may be in an area that makes it impossible to keep the same job (Hartman & Robinson,
2003). Using data from the Milwaukee Area Renters Survey (MARS), completed from 2009 to
2011, Desmond and Gershenson (2016) observed that 42% of low-wage workers who lost a job
in the previous two years had also experienced forced displacement. Renters who were forced to
move were between 11 and 25 percent more likely to lose their jobs than comparable workers
with stable housing.
Eviction can lead to further housing instability and bouts of homelessness, both short and
long term (Hartman & Robinson, 2003). In Desmond’s (2012) survey of renters in Milwaukee
eviction court, 53 percent of those who were being evicted didn’t know where they would live
next, while only 15 percent had found a new place to rent. The remaining evicted tenants were
relying on short term solutions such as friends, hotels, and shelters. After experiencing a forced
move, renters are twenty percentage points more likely to have stability issues in their next home
(Desmond, Gershenson, & Kiviat, 2015). Eviction substantially increases material hardship,
including homelessness and loss of possessions, for two to three years after the event, with the
effects diminishing slowly as time progresses (Desmond & Kimbro, 2015). Mothers who were
evicted are more likely to experience depression and stress from parenting. They are also more
than twice as likely to report their children being in poor health (Desmond & Kimbro, 2015).
Housing instability, of the sort caused by an eviction, has negative effects on children’s long
term health and wellbeing (Coulton et al., 2016; Gold, 2016).
Eviction filings, including serial filings, can cause substantial financial harm. (Serial
filings are those that occur repeatedly for the same tenant in the same rental unit and when the
6
landlord may not intend to have the tenant removed from the property.) When an eviction is
filed, the late rent is effectively transferred from being a rent payment into a recorded debt,
complete with state-enforced collection (Garboden & Rosen, 2018). The filing creates an
electronic trail that tenants are often unable to shake for a decade or longer. Serial filings can
create “psychological strain, social withdrawal, and legal cynicism toward the court system”
(Leung, Hepburn, and Desmond, 2019). Tenants in South Carolina, where serial filings are
common, also reported the shaming power of eviction notices that took the form of “pink slips
hung on doorknobs with rubber bands” (Leung, Hepburn, and Desmond, 2019).
Serial filings can be used by landlords as a tool to collect court fines and late fees that can
add substantially to the effective housing cost burden of a tenant, which in turn can worsen the
household’s residential instability. Leung, Hepburn and Desmond (2019) report that tenants
experiencing serial evictions experience fines and fees that increase their housing costs by an
average of 22 percent. Through the filing process, the courts enable landlords to collect fees
using a state-supported threat of removal, and with the filing on the tenant’s record, finding a
new place to live is not assured.
In their interviews in South Carolina, where serial filings are common, Leung, Hepburn,
and Desmond (2019) found that landlords typically assumed that tenants receiving a filing would
“pay to stay.” In most places, where eviction filings are easily accessible, perhaps the greatest
cost to the tenant is the scar that the filing leaves on the tenant in an already difficult search for
affordable housing. Tenants with serial filings are likely to be relegated to a segment of low-end
rental housing that is a form of housing of last resort, where they have even less power and their
new landlords have even more. Most landlords are wary of renting their properties to tenants
with evictions on their records, and many flat out refuse to consider it (Desmond, 2012; Leung,
7
Hepburn and Desmond, 2019). Quick and easy access to these eviction filing records further
shifts the balance of power toward landlords (Gold, 2016).
From Informal Evictions, to Filings, to Full Eviction
Eviction is a complex phenomenon. It is more than a singular event or even a discrete
series of events that always leads to the forced removal of a tenant. Informal evictions often
involve landlords threatening a tenant with a formal eviction filing or the nonrenewal of a lease.
Eviction filings can be the beginning of a forced removal process, but they are also frequently a
tool used to enforce the collection of rent and fees, including late payment fees (Garboden and
Rosen, 2018; Leung, Hepburn and Desmond, 2019). While the eviction filing is often the formal
initialization of the eviction process, state law is likely to have a strong influence on how and
when landlords file, and the extent to which landlords view filing eviction as a last resort or as a
common and almost automatic response to even an occasional delinquency. Other factors that
may influence the use of serial filings suggested in the literature include the size of the property,
the scale of the property owner, and the use of large property management firms (Garboden and
Rosen, 2018; Leung, Hepburn and Desmond, 2019).
Court records on both eviction filings and judgements are often the best available data to
measure eviction events and processes, but they do not provide complete insight into all forced
removals of renters, including those that happen outside of the legal system. In his work in
Milwaukee, Desmond (2016) estimated that, for every eviction executed through the courts,
there are two more that effectively occur outside formal judicial processes. Understanding that
eviction is on the table as an outcome, families often vacate their units after making agreements
with landlords to leave before he or she files with the court. Many others leave once the landlord
has made the initial filing with the court, but prior to a judgement or removal by law
8
enforcement. Further contributing to instability experienced in low-end rentals is the ability of
landlords to execute no-cause evictions with short notice when there is no long-term lease on the
property, or by refusing to renew a lease (Hartman & Robinson, 2003).
Landlord business practices are an integral part of the eviction equation. Rosen and
Garboden (2018) suggest two models of landlord business practices. The first is that of the
landlord who is quick to evict a delinquent or problematic tenant, and tends to file evictions
when they intend to see the process through to removal of the tenant. The second model, which
they refer to as “paternalistic” landlording, is one where the landlord is less interested in
removing the tenant, not necessarily out of altruism but because they see it as a costly process
that results in higher vacancies, turnover, and legal costs. These paternalistic landlords use a
“combination of incentives on the one hand and tactics of surveillance and control on the other”
(Rosen and Garboden, 2018). Part of the “sticks” of such an approach can be the use of eviction
filings, which bring in the power of the state. While many of these landlords work hard to avoid
actual full evictions, they are frequently less averse to the use of filings as a disciplinary
corrective that provides a clear signal that the state is on the side of the landlord.
Garboden and Rosen (2018), in their interviews of landlords in Baltimore, Cleveland and
Dallas, found that some landlords viewed the additional revenue from late fees, enforced by the
threat of an eviction filing, as a supplemental source of funding in addition to the regular rent
roll. However, they found that this was more likely to be the case among larger, professional
property managers, who were more common in Dallas than in the other two cities. Leung,
Hepburn and Desmond (2019), in their study of serial eviction filings, found that the average
tenant experiencing serial filings saw a 22 percent increase in their housing costs, corroborating
the potential profitability of serial filing for property managers or owners. Moreover, while they
9
are a single-family firm, Invitation Homes, a major national landlord, attributed a $2 million
increase in revenue to a standard lease that automated delinquency tracking and the charging of
various fees (Raymond et al., 2018).
In their study of 8 million court records in 28 states, Leung, Hepburn and Desmond
(2019) found that filings associated with serial cases comprised more than one half of filings in
2014. Moreover, serial filings varied greatly by state. In South Carolina, for example, where
serial filings are common, they comprised 73 percent of total filings. Like Garboden and Rosen
(2018), they also found that larger, professionalized management companies were more likely to
employ serial filings. Using eviction filing data in Fulton County, Georgia from 2015, Raymond
et al. (2108) found that single-family landlords that owned at least 15 homes were eight percent
more likely to file eviction than landlords that own fewer than 15, controlling for neighborhood
and property characteristics. This trend was especially pronounced among a handful of private
equity-backed landlords that filed evictions on over one-quarter of their tenants during the course
of the year.
Based on interviews with landlords in Mobile, Alabama and Charleston, South Carolina,
Leung, Hepburn, and Desmond (2019) suggest that larger landlords practice “eviction (filing) by
computer” finding that many such landlords were “at the mercy of their spreadsheets” and were
allowed little ability to work with tenants before initiating a filing. The distinction in the
literature between the more flexible smaller landlord and the more systematic, somewhat
automated, professionalized landlord is particularly important because the trend is toward larger,
professionally managed properties under large corporate ownership. Leung, Hepburn and
Desmond (2019) cite Census Bureau statistics suggesting that the top 50 rental firms went from
controlling 22 percent of the national rental market in 2002 to 32 percent by 2012. Given that
10
these are national statistics and that the larger regional or national firms are more prevalent in
large metropolitan areas, the concentration at the top of the industry is almost certainly greater in
large metros.
The shift from smaller-scale owner-operators to large, corporate owners using larger
property management operations is reminiscent of the transition in mortgage lending from
primarily portfolio lending (dominated by savings and loans) up through the 1970s to an
originate-to-distribute business model in which lenders sold off loans after originating them
(Immergluck, 2015). This transition led to the dominance of large, national servicing companies
which competed to automate the loan collection process and reduce costs by creating “light-
touch” interventions which led to initiating foreclosure rapidly when loans became delinquent.
The older, traditional portfolio lending model allowed for easier loan modifications when
homeowners encountered distress, and when the foreclosure crisis hit, the mass servicing
operations that had taken over were extremely slow to adapt (Immergluck, 2015).
What Might Explain Serial Filing?
The literature suggests at least five reasons that landlords may employ serial eviction
filings in their business models. While it is beyond the scope of this study to confirm or measure
the presence of such motivations, it is important to understand why some landlords may adopt
such practices. First, filings can be a disciplinary tool that clearly signals that the landlord is
prepared to resort to forced removal if need be. Visible notices might also provide a form of
shaming that makes it clear to tenants that rents are expected to be paid on time, or else (Leung,
Hepburn, and Desmond, 2019). Serial filings may be intended more “to squeeze money out of a
tenant” than to remove them from the premises.
11
Second, particularly for larger landlords, systematization, computerization and
routinization may be important drivers of serial filings (Raymond et al., 2018). Large firms may
face greater reluctance to vary in how they treat tenants and may be less flexible and less
“paternalistic” in how they deal with tenants (Rosen and Garboden, 2018; Leung, Hepburn, and
Desmond, 2019). Standardization and algorithmic routinization may be viewed by larger
landlords as tools to maintain certain profitability targets and to reduce risk and uncertainty,
while more customized procedures may be viewed as cost-ineffective or creating too much
potential uncertainty (Immergluck and Law, 2014). While their study focused on single-family
rentals, Raymond et al. (2018) found that large corporate owners were 68 percent more likely to
file evictions than smaller landlords, after controlling for property, household and neighborhood
factors.
A third potential motivation for serial filings is their ability to generate a supplemental
income stream for the property manager or owner. Serial filers may cater to tenants who are
economically fragile and, like banks charging overdraft fees, they may have identified a way to
capitalize on this fragility. Such a motivation also suggests that these landlords are more likely to
accept tenants with previous eviction filings. The effective premium that tenants pay through late
fees is a systematic penalty that the lightly regulated rental market inflicts on those who are
economically fragile, not dissimilar from the interest rate penalties that subprime lenders inflict
on those with previous credit challenges.
A fourth potential motivation for serial filings could be to punish tenants for certain
behaviors, such as complaining about building conditions (especially to local government).
While no current research suggests this is common practice, it’s conceivable that a well-known
practice of filing quickly and frequently on tenants who misbehave in some way could have a
12
deterrent effect on unwanted behavior, beyond simply being late with the rent. Many states have
habitability statutes that aim to prevent landlords from evicting tenants who complain to
authorities about unhealthy or unsafe housing conditions. Using eviction filings as a deterrent
would allow a landlord to sanction tenants without going to the costlier lengths of full eviction,
and then having to deal with vacancy and turnover costs.
Finally, the policy context of eviction law may create incentives (or the lack of
disincentives) for serial eviction filings. Leung, Hepburn, and Desmond (2019) interview
landlords in two cities with very different state eviction law contexts. In Charleston, South
Carolina law provides few impediments to serial filing. The filing fee charge in Alabama is over
$250, more than seven times the charge in South Carolina. In addition, in Alabama, landlords
have to provide tenants with a seven-day written notice before filing an eviction and, in the case
of corporate ownership (e.g., LLCs), an attorney must be retained for the filing. This slows the
process down, which can create additional costs for the landlord. As a result of these higher
costs, serial filings are a very small share (6%) of overall filings in Alabama. In a multivariate
analysis at the census tract level, Leung, Hepburn, and Desmond (2019) also found that states
with eviction filing fees of more than $200 had lower levels of serial filing, supporting the notion
that the ease and low cost of filing encourages serial filing. Hatch (2017) categorizes landlord-
tenant laws across states and classifies Georgia as a pro-business state with few landlord-tenant
laws and a high share of such laws favoring landlords over tenants.
Household and Neighborhood Characteristics and Eviction
A number of studies have identified household and owner characteristics associated with
higher eviction rates or eviction filing rates. Less research has focused on the relationship
13
between building or owner characteristics and eviction or eviction filing rates, and we are aware
of only one study that has looked at the share of eviction filings that are serial filings.
One common finding in the literature is the association between evictions and eviction
filings, on the one hand, and the racial composition of the neighborhood. Filings and evictions
are often clustered in predominantly Black neighborhoods (Desmond, 2012; Raymond et al.,
2018; Shelton, 2017). Using eviction filing data in Richmond, Virginia from 2006 through 2016,
Teresa (2018) found that the racial composition of a neighborhood had a significant influence on
eviction filings––he found that for every 10% increase in the population of Black residents, the
eviction filing rate increased by 1.2%. An important aspect of Teresa’s findings is that the
poverty rate and median income of the neighborhood were not statistically significant predictors
of the filing rate. Raymond et al. also found that the Black neighborhoods (modeled as imputed
Black race) were influential predictors of single-family rental eviction filings.
The spatial clustering of evictions in Black neighborhoods is not surprising given the
survey-based evidence that evictions disproportionately affect Blacks. In the 2011 Milwaukee
Eviction Court Study, 30 percent of evicted tenants were Black women, but Black women made
up only 9.6 percent of Milwaukee’s population (Desmond, 2012). In a similar examination of
2013 eviction court records from King County, Washington (Seattle), Thomas (2017) found that
eviction filings were made against Black renters at a rate of four times greater than against white
renters (Thomas, 2017).
Desmond and Gershenson’s (2017) analysis of Milwaukee survey data during the Great
Recession found the number of children in a household to be a significant predictor of eviction.
The presence of children in the neighborhood was correlated with higher levels of evictions in
the area, and this did not change significantly when controlling for race, single mothers, and
14
other socioeconomic indicators (Desmond et al. 2013). For every one percent increase in the
population of children, there was an associated 6.5% increase in a neighborhood’s eviction cases,
controlling for race, gender, household composition, and socioeconomic status.
Other research has found that the education level and, especially, sex of the head
householder are associated with eviction rates. Desmond (2016), for example, has suggested that
eviction disproportionately affects households headed by Black women, prominently comparing
the problem of eviction to that of mass incarceration by stating, “Poor black men were locked up.
Poor black women were locked out.”
Only one study thus far has examined the prevalence and patterns of serial eviction
filings using eviction data. Leung, Hepburn and Desmond (2019) estimated the share of filings
that were serial evictions at the census tract level across counties in 28 states. They found that
race, gender, and education were significant predictors of serial filing share. Majority-white
neighborhoods had serial filing shares that were more than 6 percentage points lower than
otherwise similar neighborhoods. The effect of education was somewhat counterintuitive, with
more highly educated tracts experiencing higher serial shares. The presence of female
householders was associated with higher serial shares, but the presence of children was not.
Importantly, neighborhoods in states with high filing fees (over $200) had serial shares that were
5.5 percentage points lower, other things equal.
Data and Methods
To examine factors that may be associated with the level of eviction filings – including
both serial and nonserial filings – we gathered address-level data on eviction filings for the five
core counties of the Atlanta metropolitan area for 2016 by scraping them from county court
15
websites. This work was part of a larger data gathering project by the Atlanta Journal
Constitution for a series the paper did on evictions in 2018 (Joyner, Ernsthausen, and Mariano,
2018). Data on parcel and building characteristics, including the owner’s name and address, were
obtained from tax assessor parcel files for each of the five counties. This enabled us to identify
the multifamily (5 units or more) residential buildings and link them to eviction filings by
address. We also obtained real estate sales data by parcel number and linked it to the tax assessor
parcel files. We used these data to identify recent sales history for the multifamily properties.
While the Atlanta metropolitan area as defined by the US Census Bureau is comprised of
29 counties, the five core counties account for 67 percent of the metropolitan area’s population
and 88% of the multifamily rental units in the region. When compared to other large
metropolitan areas in the U.S., the Atlanta metropolitan area suffers from very high eviction
filing rates and eviction rates. We identified comparable counties by population density in the
largest 20 metropolitan areas for which eviction filing rate and eviction rate data were available
from Princeton’s EvictionLab.i Twenty-two metropolitan counties across 13 states met our
selection criteria for comparison counties. The five core metro Atlanta counties had five of the
eight highest eviction filing rates among the total group of 22 counties, and had the five highest
eviction rates. The average eviction filing rate for the five core counties in 2016 was 25 percent,
compared to 9.3 percent for the remaining 22 counties (Eviction Lab, 2018). The average
eviction rate for the five core counties was 6 percent, compared to 1.8 percent for the other 22
counties. Thus, the Atlanta metro is a region that exhibits relatively high eviction levels and high
eviction filing rates.ii
We examine 1,861 multifamily properties in the five-county area. The average property
contained just under 200 rental units, ranging from 5 units to more than 1,700. The interquartile
16
range was from 60 to just under 300 units. Machine learning methods were used to assist in
identifying related ownership structures among entities listed as purchasers of the properties in
property tax and sales records. The software employed, called Dedupe, used active learning to
create ownership “clusters” referring to related entities based on similarities in purchaser names
and addresses listed on separate buildings. However, due to the complexity and time-consuming
nature of this work, we limited this analysis only to buildings with at least 50 units.
Table 1 illustrates the categorization of buildings by owner cluster sizes. Forty-three
owner clusters own at least six 50-plus unit multifamily properties, with the largest owner
holding 28 properties in the five-county area. These owners, together, own 431 (23%) of the
1,861 multifamily properties and 122,800 (33%) of the 369,629 multifamily rental units in the
data set. The next largest set of owner clusters are those owning from two to five 50-plus unit
multifamily properties. These owners own 448 (24%) of the multifamily properties and 112,521
(30%) of the multifamily units. The remaining 982 (53%) properties comprise 134,308 (36%) of
the multifamily units. Figure 1 illustrates the distribution of the 1,861 properties throughout the
five-county area.
Multifamily evictions in the five-county region accounted for more than 95,000 out of
more than 132,000 total eviction filings in 2016, with the difference being filings on single-
family (one-to-four unit) properties, so that multifamily eviction filings account for more than 70
percent of all filings in the area. The 95,000 filings occurred in 1,861 multifamily properties.
This equates to an average of 51 eviction filings per building in 2016. The average annual
eviction filing rate per building was 28 filings per 100 rental units.iii We were able to identify
when landlords filed evictions on the same tenant by examining the tenant names. When a tenant
in a building received more than one eviction filing within a one-year period, the second eviction
17
(or the third, etc.) was classified as a serial eviction filing. The average building had a nonserial
eviction rate (total filings – serial filings in a year, divided by units in a building) significantly
lower than the average total filing rate, at 16 nonserial filings per 100 units.
Buildings vary quite markedly in their total and nonserial filing rates. The interquartile
range for total filing rates for the 1861 buildings was 9 per 100 for the first quartile and 38 per
100 for the third quartile. For nonserial filing rates the interquartile range went from 6 per 100
units to 22 per 100 units.
Conceptually, it may be that the neighborhood characteristics, location, age, and size of a
building may explain away much of the difference in building-level eviction rates. A primary
goal here is to identify whether that is indeed the case, or whether some large, multi-property
owners appear to exhibit substantially higher (or lower) eviction filing rates than others, even
after controlling for building and neighborhood characteristics.
We begin with a model for estimating the eviction filing rate (eviction filings/rental units)
for a building, Y,
Y=+X+N+S+C+
where X is a set of building characteristics, including the number of units in the building, the
appraised value per unit in 2016,ivthe age of the building, and a set of dummy variables
indicating whether the building was sold in the same year, a prior year, or two to three years ago.
Our prior is that the sale of a building may result in a rise in eviction filings as some owners raise
rents, possibly after upgrading the property (Rosen and Garboden, 2018). N is a set of block
group-level characteristics, which are chosen based on the previous literature, including the share
of rental households who are housing-cost burdened, median gross rent, poverty rate, the share of
households that are female-headed families, the share of adults over 25 who are college-
18
educated, the share of households with children, the share of renters who are 65 or over, the
share of renters who are Black, the share of renters who are Latino, and the share of renters who
are Asian. S is a set of spatial dummy control variables indicating which of 36 different Atlanta
Regional Commission superdistricts the property is located within. These are included to account
for spatial clustering within the five-county metro. Each county contains from 4 (Clayton) to 12
(Fulton) superdistricts. C is a set of dummy variables that indicate which, if any, of the 43 large
owner clusters owns the property.
Diagnostics show that the estimations of equation 1 exhibit substantial heteroscedasticity,
so a robust standard error estimator is used. In addition, standard errors are clustered at the
block-group level due to the use of block-group-level variables. While the dependent variable is
substantially nonnormal (positively skewed), the residuals from the estimations are only slightly
skewed. For these reasons, a clustered-robust standard error is employed in estimating equation
1. No significant multicollinearity problems were detected. Only two variance inflation factors
exceeded 5, and those did not exceed 7. These were for the share of renters who are Black and
the share of residents who are college educated. The share Black variable was significant in all
estimations so multicollinearity does not cause any inference problems. Multicollinearity may be
inflating the standard error for the college-educated variable somewhat and so may create a
modest increase in the p-value for this variable. But the variable is found to be significant in the
literature and is retained.
Rather than simply estimating the aggregate eviction filing rate using equation 1, we
decompose the total eviction filing rate into serial and nonserial filing rates. The former is simply
the number of serial eviction filings divided by the number of rental units in the building, while
the nonserial filing rate is the number of nonserial evictions divided by the number of rental units
19
in the building. We estimate each of these separately and also estimate the share of total filings
that are serial filings.
Results
Table 2 provides descriptive statistics for the independent variables (excluding the spatial
and owner-cluster dummies) in equation 1, as well as the three dependent variables (serial filing
rate, nonserial filing rate, and serial filing share) and the cumulative filing rate. Table 3 describes
the results of estimating equation 1 when the dependent variable is the serial filing rate, which is
simply the number of serial eviction filings in a property divided by the number of rental units.
None of the building characteristics are significant in predicting the filing rate. Of the
neighborhood characteristics, the share of renters who are cost-burdened (pay over 30% of their
income on housing), the share of renters who are Black, and the share of renters who are Asian
are significant. For every 10 percentage-point increase in the share of renters who are cost
burdened, the serial filing rate increases by 0.5 percentage points. For every 10 percentage-point
increase in the share of renters who are Black, the serial filing rate increase by 1.1 percentage
point. Therefore a block group that is 90% Black would be expected to have a serial filing rate
that is 8.8 percentage points higher than an otherwise similar block group that is 10% Black. For
every 10 percentage-point increase in the share of renters who are Asian, the serial filing rate is
expected to decrease by 1.1 percentage points. Of the 43 large owners specified in the estimation,
15 have statistically significant differences in serial filing rates after controlling for the other
building and neighborhood characteristics. This suggests that owner characteristics account for
substantial variation in serial filing rates. In approximately half (8 out of 15) of these cases, the
20
serial filing rate is higher (coefficient is positive) than the regression-adjusted level of smaller
owners.
Table 4 provides the result of estimating equation 1 when the dependent variable is the
nonserial eviction filing rate. The nonserial filing rate is the number of nonserial filings (total
filings – serial filings) divided by the number of units in the building. While this variable
corrects somewhat for serial filing behavior it is not a direct measure of the number of evictions
that end in removal of the tenant, but it is expected to be better correlated with such activity than
the total filing rate and, certainly, than the serial filing rate. The results here are significantly
different than those in Table 3, with some exceptions. The three block group characteristics that
were significant in Table 4 are still significant with the same signs. In this regression however,
all but one of the building characteristics are significant as are more neighborhood
characteristics. The number of units in the building has a strong association with the nonserial
filing rate. For every 100 additional units in a building, the nonserial filing rate is expected to
decline by 1 percentage point. Given that the standard deviation for this variable is 165, this is a
substantial effect. Age of the building is also significantly associated with nonserial filing rate, so
that a building that is 45 years old is expected to have a nonserial filing rate that is 3.8 percentage
points higher than an otherwise similar building that is only five years old.
A critical finding here is that sales of the building in the three prior years are associated
with substantially higher nonserial filing rates. A sale in the same year (2016) is associated with
a nonserial filing rate that is 4.7 percentage points higher, other things equal. A sale in the
previous year is associated with a nonserial filing rate that is 4.5 percentage points higher, and a
sale two to three years prior is associated with a nonserial rate that is 4.2 percentage points
higher. Thus, ownership turnover is associated with higher nonserial filing rates that can last two
21
to three years, which is consistent with some of the qualitative literature suggesting that when
buildings are sold, evictions tent to follow. Reasons for this could include the upgrading or
conversion of the building to higher-end rentals, a trend that has been observed in metropolitan
Atlanta in recent years (Bhatia and Keller, 2018).
As for block group variables, each 10 percentage-point increase in the share of renters
who are cost-burdened is associated with a 0.6 percentage-point increase in nonserial filings.
Each 10 percentage-point increase in the share of renters who are Black is associated with a 0.8
percentage-point increase in the nonserial filing rate. Therefore, a block group that is 90% Black
would be expected to have a 6.4 percentage-point higher nonserial filing rate than an otherwise
similar block group that is 10% Black. Each 10 percentage-point increase in the share of renters
who are Asian is associated with a 0.6 percentage-point increase in the nonserial filing rate but
Asian share varies much less than Black share, with the standard deviation for Black share of
renters being more than three times that of the Asian share. Finally, the level of education is also
a significant predictor of the nonserial filing rate. Each 10 percentage-point increase in the share
of adults who are college educated is associated with a 0.7 percentage-point decline in the
nonserial filing rate, suggesting that education may have a modest suppressive effect on
nonserial filings.
Similar to the results in Table 3, of the 43 large owners specified in the estimation, 14
have statistically significant differences in nonserial filing rates after controlling for the other
building and neighborhood characteristics. This suggests that owner characteristics account for
substantial variation in nonserial filing rates. In six of the fourteen cases, the non-serial filing rate
is higher than the regression-adjusted level of smaller owners. This suggests that while there is
owner-level variation in the propensity to file nonserial evictions, there is not a strong systematic
22
tendency for the largest owners to issue nonserial filings more than smaller owners. While these
results may not appear consistent with those of Raymond et al. (2018) at first blush, it is
important to keep two differences in mind. First, these are multifamily properties, so even the
smaller owners tend to be quite different than the owners of small numbers of single-family
rentals. Second, we have isolated nonserial eviction filings here, which, as the differences
between Tables 3 and 4 show, exhibit different patterns than serial filings which are included in
any counts of total filings.
Finally, Table 5 shows the results for estimating equation 1 with the share of total filings
that are serial filings as the dependent variable. This is the “serial filing share” and is similar to
what Leung, Hepburn, and Desmond (2019) estimate, but we measure it at the building rather
than the census tract level. In this model, the number of units in the building is significantly and
positively associated with serial filing share, suggesting that larger, newer buildings have higher
serial eviction shares, other things equal. This is consistent with the qualitative work described
earlier suggesting that owners of larger properties may employ more mechanistic, routinized
filing behavior. Every100 unit increase in property size is associated with an increase of 2.1
percentage points in serial filing share.
Sales in the current year is also significant and the coefficient is negative, which is
consistent with the findings in Table 4 that recently sold properties have higher nonserial filing
rates. Age of the building is also significant with a negative sign, again consistent with the
findings in Table 4 showing that older buildings have higher nonserial filing rates, which will,
other things equal, result in lower serial filing shares. A building that is 45 years old is expected
to have a serial filing share that is 5.2 percentage points lower than an otherwise similar building
that is five years old. Share of adults with a college education is positively associated with the
23
serial filing share, which is consistent with lower nonserial filings in block groups with higher
college education rates.
The share of renters in a block group who are Black is strongly associated with serial
filing share. An 80 percentage-point increase in share of renters who are Black, other things held
constant, is associated with a 9.8 percentage-point increase in serial filing share. So even while
nonserial filings increase significantly as Black renter share increases, serial filings increase even
more, resulting in higher serial shares.
In the Table 5 results, 21 of the 43 large owners had significantly different serial filing
shares than the smaller owners, after controlling for building and neighborhood characteristics.
For 16 of these 21, the share was higher than the share among the smaller owners. This is
consistent with the qualitative literature suggesting that larger owners may be more likely to
employ serial eviction filings.
Figure 2 plots the locations of the 205 multifamily properties owned by the 21 large
owners whose serial filing share is significantly different from smaller owners after controlling
for neighborhood, locational, and building characteristics. Over 160 of these buildings are owned
by what we call the 16 “high serial filers,” owners who tend to have high serial filing shares that
are not explained by other independent variables. Forty properties are owned by the five “low
serial filers.” While properties owned by high serial filers are distributed widely across the
region, the 40 properties owned by low serial filers tend not to be located in Black
neighborhoods. Table 6 indicates that 10% of properties owned by low serial filers are located in
predominantly Black neighborhoods, while 34% of all properties and 30% of high serial filers
are located in such block groups. Moreover, 73% of properties owned by low serial filers are
located in block groups that are less than 50 percent Black, compared to 48% of multifamily
24
properties and 50% of those owned by high serial filers. Thus, properties owned by landlords
who file fewer serial eviction notices are disproportionately located in neighborhoods that are
less Black compared to other multifamily properties.
Conclusion
A fundamental finding of this research is that, while patterns of serial and nonserial
evciton filings share some similarities, including being disproportionately concentrated in Black
neighborhoods, there are important differences between them. Researchers, policymakers, and
housing advocates should be careful in distinguishing between these two types of activity when
measuring eviction activity. While serial and nonserial filings are important and can harm
tenants, they indicate different phenomena and different landlord intentions. Nonserial filings are
more likely to indicate landlords’ strong intentions to remove tenants, while serial filings are
more likely to represent the use of the filing as a tool to coerce rent payment with the assistance
of the state.
Nonserial filing rates are associated with relatively smaller buildings, a sale in the last
three years, high neighborhood rental burdens, lower neighborhood education levels, and the
share of renters who are Black. Serial filing rates are associated with high neighborhood rent
burdens and race. Finally, higher shares of filings that are serial filings is associated with larger
properties, newer properties, higher education levels, and the share of renters who are Black.
While most of these findings corroborate findings in other literature, the finding on recent
property sales is a new contribution to the quantitative literature. The finding corroborates the
perhaps expected notion that property turnover, which may be followed by upgrading, is
associated with physical displacement of tenants. While purchase and redevelopment projects
25
may occasionally involve compensation or inducements to affected tenants, these findings
suggest a sizeable eviction filing effect, which is of significant policy concern, especially given
the damage that an eviction filing can have on a tenant’s future prospects in the rental market.
Even after controlling for building, location and neighborhood characteristics, a
substantial share (16 out of 43) of the largest owners, have high serial filing shares, while only
five have low serial filing shares. (The remaining 22 owners’ serial filing shares are not
significantly different from those of smaller owners.) This, together with the finding that larger
buildings tend to have larger serial filing shares, suggests that larger owners and owners of larger
properties are more likely to be serial filers. Moreover, serial filers tend to be located in a diverse
variety of neighborhoods, while large owners who do not file serially tend to be located
disproportionately in neighborhoods with lower Black populations.
This article points to several implications for further research and housing policy. First,
serial filing patterns in a high serial-filing state such as Georgia may not be highly generalizable
to other states, especially those with low levels of serial filing. Therefore, it is important to
conduct similar research across states with different state legal contexts. Second, while we
employ block-group level demographic data, we do not have direct household-level data, so
work is needed to corroborate the notion that Black renters may be more likely to not only
experience higher eviction rates, but also higher serial filing rates.
One implication of these findings is the need for policies to reduce evictions affecting
tenants in buildings after they have been sold. Tenant advocates and legal assistance
organizations could target tenants in larger buildings that have been put up for sale to ensure that
tenants are advised of the likely sale, are aware of their rights, and can potentially negotiate with
their landlords before an eviction is filed. Local and state government could provide relocation
26
assistance to lower-income tenants in these properties and encourage landlords not to file
evictions. Moreover, building owners, especially any receiving public subsidy, could be asked
(or required) to adopt a list of best practices aimed at reducing eviction filings and full evictions.
Another implication concerns the social costs of state policy that facilitates low-cost, easy
filing, which may foster serial filing (Leung, Hepburn and Desmond, 2019). In states where
filing is inexpensive and quick, serial filing is effectively a state-favored practice, and becomes
an easy tool to discipline tenants and, potentially, a complementary way of generating additional
revenue out of economically fragile households. However, merely raising the cost of evictions,
without somehow limiting how much can be passed on to the tenant, could backfire and leave
tenants with even greater financial penalties. One way to make filing somewhat less easy for
landlords is to adopt and lengthen notice periods that precede the landlord filing the eviction. In
addition, requiring or expanding legal representation to renters prior to their receiving the
eviction filing, during a pre-filing notice period, could both deter serial filing and provide critical
assistance to tenants.
Given the large variation in serial filing and eviction behavior among landlords, it may be
useful to tenants if local governments or housing advocates published easily accessible data on
the eviction and serial filing behavior of multifamily properties in a city, or other helpful
information such as whether a building has been recently sold or is on the market, both indicators
of future eviction risk. There are examples of online tools and apps such as
antievictionmappingproject.net (2019), which allows users in San Francisco to look up the
eviction history of a building by simply typing in its address. Another is the Association for
Neighborhood and Housing Development’s (2019) Displacement Alert Project, which provides
alternative parcel-level map layers, one of which indicates the number of evictions per unit for a
27
building. In states were serial filings are common, it is important for such technology to
disaggregate serial versus nonserial filings. Civic technology might also be used to develop cell
phone-accessible applications that provide limited legal information to tenants who are not able
to obtain legal representation, or to guide them through the process (Greater Boston Legal
Services, 2019; Phonesis Technologies, 2019).
Finally, any policies that increase the supply of safe and secure affordable housing to
low-income tenants, either through demand- or supply-side subsidies, is likely to reduce eviction
rates and serial filings. Moreover, efforts to protect tenants from retaliatory evictions, to provide
legal counsel for tenants, and to slow down the eviction process are also likely to improve
housing stability for low-income tenants.
28
References
Antieviction Mapping Project. (2019). http://www.antievictionmappingproject.net/pledge/.
Armstrong Jr, G. M., & LaMaster, J. C. (1986). Retaliatory eviction as abuse of rights: A civilian
approach to landlord-tenant disputes. La. L. Rev., 47, 1.
Association for Neighborhood and Housing Development. (2019). Displacement alert project.
https://www.displacementalert.org/.
Bhatia, A. & Keller, M. (2018). Preserving naturally-occurring housing affordability in metro
Atlanta neighborhoods: Findings and recommendations for policymakers, foundations,
developers, and nonprofits. Enterprise Community Partners. May.
https://www.enterprisecommunity.org/download?fid=9645&nid=7096.
Coulton, C.J., Richter, F., Kim, S.J., Fischer, R. & Cho, Y., (2016). Temporal effects of
distressed housing on early childhood risk factors and kindergarten readiness. Children
and Youth Services Review, 68, pp.59-72.
Desmond, M. (2012). Eviction and the reproduction of urban poverty. American Journal of
Sociology, 118(1), 88-133.
Desmond, M. (2016). Evicted: Poverty and profit in the American city. Broadway Books.
Desmond, M., An, W., Winkler, R., & Ferriss, T. (2013). Evicting children. Social Forces, 92(1),
303-327.
Desmond, M., & Gershenson, C. (2016). Housing and employment insecurity among the
working poor. Social Problems, 63(1), 46-67.
Desmond, M., & Gershenson, C. (2017). Who gets evicted? Assessing individual, neighborhood,
and network factors. Social Science Research, 62, 362-377.
Desmond, Matthew, Gershenson, Carl, and Kiviat, Barbara. (2015). Forced relocations and
residential instability among urban renters. Social Service Review 89(2): 227-262.
Desmond, M., & Kimbro, R. T. (2015). Eviction's fallout: housing, hardship, and health. Social
forces, 94(1), 295-324.
Desmond, M., & Shollenberger, T. (2015). Forced displacement from rental housing: Prevalence
and neighborhood consequences. Demography, 52(5), 1751-1772.
Donnelly, L., Brooks-Gunn, J., McLanahan, S., & Garfinkel, I. (2017). The protective effects of
housing assistance programs on eviction. Fragile Families Working Paper WP17-07-FF.
https://fragilefamilies.princeton.edu/sites/fragilefamilies/files/wp17-07-ff.pdf
Eviction Lab. (2018). County level data on evictions and eviction filings. https://evictionlab.org/.
29
Garboden, P. & Rosen, E. (2018). The threat of eviction: How landlords shape a contingent
tenure. City and Community, forthcoming.
Gold, A. E. (2016). No home for justice: How eviction perpetuates health inequity among low-
income and minority tenants. Georgetown Journal on Poverty Law Policy, 24, 59-87.
Greater Boston Legal Services. (2019). Massachusetts defense for eviction: Self-guided eviction
help. https://www.gbls.org/MADE.
Hartman, C., & Robinson, D. (2003). Evictions: The hidden housing problem. Housing Policy
Debate, 14(4), 461-501.
Hatch, M. E. (2017). Statutory protection for renters: Classification of state landlord–tenant
policy approaches. Housing Policy Debate, 27(1), 98-11
Immergluck, D. (2015). Preventing the next mortgage crisis: The meltdown, the federal
response, and the future of housing in America. Rowman & Littlefield.
Immergluck, D., Carpenter, A., & Lueders, A. (2018). Hot city, cool city: explaining
neighborhood-level losses in low-cost rental housing in southern US cities. International
Journal of Housing Policy, 18(3), 454-478.
Immergluck, D. & Law, J. (2014). Investing in crisis: The methods, strategies, and expectations
of investors in single-family foreclosed homes in distressed neighborhoods. Housing
Policy Debate, 24(3), 568-593.
Joyner, C., Ernsthausen, J. & Mariano, W. (2018). Eviction tactics squeeze renters: AJC analysis
shows landlords increasingly use filings to collect late rent. Atlanta Journal Constitution.
June 22. https://www.ajc.com/news/local/stressful-way-live-how-eviction-threats-get-
used-collect-rent/vDYSUw49bdNqtNPIht9K1O/
Kingsley, G. T. (2017). Trends in housing problems and federal housing assistance. Washington,
DC: Urban Institute.
Leung, L., Hepburn, P, and Desmond, M. Serial evictions: Property managers, tenants, and civil
court sanctions. Eviction lab. Princeton University. Unpublished manuscript. January 16.
Lundberg, I., & Donnelly, L. (2018). A research note on the prevalence of housing eviction
among children born in US cities. Demography, 1-14.
Phonesis Technologies. (2019). More about our work. https://phonesistech.com/more-about-our-
work#_ftn1.
30
Shelton, T. (2017). Locked out: Foreclosure, eviction and housing instability in Lexington, 2005-
2016. Lexington Fair Housing Council. https://taylorshelton.info/publication/lfhc-locked-
out/
Raymond, E.L., Duckworth, R., Miller, B., Lucas, M. & Pokharel, S. (2018). From foreclosure to
eviction: Housing insecurity in corporate-owned single-family rentals. Cityscape, 20(3),
159-188.
Rosen, E. & Garboden, P. (2018) Housing the poor: Paternalism and profit in urban landlording.
Unpublished manuscript.
Teresa, B. (2018). The geography of eviction in Richmond: Beyond poverty. RVA Eviction Lab.
Virginia Commonwealth University. https://cura.vcu.edu/media/cura/pdfs/cura-
documents/GeographiesofEviction.pdf.
Thomas, T. A. (2017). Forced out: Race, market, and neighborhood dynamics of evictions.
Doctoral dissertation. University of Washington.
https://digital.lib.washington.edu/researchworks/handle/1773/40705.
31
Table 1. Distribution of Multifamily Rental Properties with Eviction Filings in 5-County Atlanta Metropolitan Area by Owner
Cluster Groups*
Owner Clusters
Buildings Units Buildings/Cluster
Units/Cluster
Properties Owned by Ownership Clusters with 6+ 50+ unit buildings 43 431 122,800 10.0 2,856
Properties Owned by Ownership Clusters with 2 to 5 50+ unit buildings 161 448 112,521 2.8 699
Properties Owned by Owners with 1 building, or multiple <50 unit buildings NA 982 134,308 NA NA
Total NA 1,861 369,629 NA NA
*Includes rental buildings with 5 or more units
32
Table 2. Descriptive Statistics
Variable Mean Std. Dev.
Serial Eviction Filings / Unit 0.1120 0.1453
Nonserial Eviction Filings / Unit 0.1630 0.1536
Share of Filings that are Serial Filings 0.3483 0.2261
Units in Building 198 165
Value/Unit ($) $67,050 $66,899
Sale in Same Year (0,1) 0.0862 0.2808
Sale in Previous Year (0,1) 0.1529 0.3600
Sale 2-3 Years Prior (0,1) 0.2430 0.4290
Age of Building (years) 35.9 19.0
BG Share, Renters Who are Rent Burdened 0.5064 0.1539
BG Median Gross Rent ($) $983 $213
BG Share in Poverty 0.2124 0.1692
BG Share, Female Headed Families as Share of Households 0.1795 0.1223
BG Share, Over 25 w/College Education 0.3511 0.2331
BG Share, Households w/Children <17 0.2956 0.1495
BG Share, Renters >16 Who are 65+ 0.0883 0.1002
BG Share, Renters Who are Black 0.5461 0.3098
BG Share, Renters Who are Latino 0.1280 0.1744
BG Share, Renters Who are Asian 0.0510 0.0920
N = 1,844
33
Table 3. Serial Eviction Filing Rate 2016, Estimation by Multifamily Building*
Coefficient Std. Error**
Beta t Sig.
Constant -0.0195
0.0470
.
-0.42
0.678
Units in Building 1.10E-05 1.93E-05 0.0125 0.57 0.569
Value/Unit ($) 1.47E-07 1.54E-07 0.0678 0.96 0.338
Sale in Same Year -0.0081 0.0094 -0.0156 -0.86 0.391
Sale in Prior Year 0.0052 0.0089 0.0128 0.58 0.559
Sale 2-3 Year Prior 0.0124 0.0083 0.0367 1.49 0.136
Age of Building (years) -1.54E-04 2.54E-04 -0.0201 -0.60 0.546
BG Share, Renters Who are Rent Burdened 0.0515 0.0240 0.0546 2.14 0.032
BG Median Gross Rent ($) -9.64E-06 2.22E-05 -0.0141 -0.43 0.664
BG Share in Poverty -0.0141 0.0244 -0.0164 -0.58 0.564
BG Share, Female Headed Households -0.0333 0.0478 -0.0281 -0.70 0.486
BG Share, Over 25 w/College Education 0.0535 0.0383 0.0858 1.40 0.163
BG Share, Households w/Children <17 0.0659 0.0411 0.0678 1.60 0.109
BG Share, Renters >16 Who are 65+ -0.0482 0.0333 -0.0333 -1.45 0.148
BG Share, Renters Who are Black 0.1069 0.0320 0.2279 3.34 0.001
BG Share, Renters Who are Latino 0.0216 0.0385 0.0260 0.56 0.574
BG Share, Renters Who are Asian -0.1134 0.0330 -0.0718 -3.44 0.001
Owner Clusters (total 43) where p-value <0.10
Owner 3 0.0769 0.0217 0.0548 3.55 0.000
Owner 4 0.1043 0.0233 0.0725 4.48 0.000
Owner 7 -0.0168 0.0100 -0.0104 -1.69 0.092
Owner 8 -0.0473 0.0114 -0.0283 -4.16 0.000
Owner 9 -0.0303 0.0181 -0.0181 -1.67 0.095
Owner 10 0.1220 0.0462 0.0703 2.64 0.008
Owner 12 0.0634 0.0336 0.0351 1.89 0.060
Owner 13 0.0726 0.0274 0.0385 2.65 0.008
Owner 17 0.1119 0.0590 0.0566 1.90 0.058
Owner 22 -0.0401 0.0161 -0.0182 -2.49 0.013
Owner 36 -0.0474 0.0197 -0.0186 -2.41 0.016
Owner 37 -0.0295 0.0135 -0.0116 -2.18 0.029
Owner 38 -0.0607 0.0162 -0.0238 -3.75 0.000
Owner 39 0.6742 0.2874 0.2414 2.35 0.019
Owner 40 0.0786 0.0301 0.0308 2.61 0.009
N = 1,844
R-square = 0.26
*Also included as independent variables (not shown here) are spatial dummies for 36 “superdistricts” within the 5-county region
and an additional 28 owner cluster dummy variables where p-value >= 0.10.
**Clustered-Robust standard errors. Clustered at the block group level.
Underlined, p<0.10; bold, p<0.05; 0.0
34
Table 4. Nonserial Eviction Filing Rate, Estimation by Multifamily Building*
Coefficient Std. Error**
Beta t Sig.
Constant 0.0174
0.0526
.
0.33
0.741
Units in Building -1.00E-04 2.33E-05 -0.1080 -4.31 0.000
Value/Unit ($) 2.10E-07 1.54E-07 0.0914 1.36 0.173
Sale in Same Year 0.0467 0.0111 0.0853 4.19 0.000
Sale in Prior Year 0.0452 0.0108 0.1060 4.20 0.000
Sale 2-3 Year Prior 0.0415 0.0087 0.1159 4.79 0.000
Age of Building (years) 9.48E-04 2.82E-04 0.1173 3.36 0.001
BG Share, Renters Who are Rent Burdened 0.0583 0.0264 0.0585 2.21 0.027
BG Median Gross Rent ($) 7.85E-06 2.99E-05 0.0109 0.26 0.793
BG Share in Poverty 0.0352 0.0442 0.0388 0.80 0.426
BG Share, Female Headed Households -0.0264 0.0495 -0.0210 -0.53 0.594
BG Share, Over 25 w/College Education -0.0668 0.0342 -0.1014 -1.95 0.051
BG Share, Households w/Children <17 -0.0128 0.0366 -0.0125 -0.35 0.726
BG Share, Renters >16 Who are 65+ -0.0412 0.0405 -0.0269 -1.02 0.309
BG Share, Renters Who are Black 0.0791 0.0308 0.1596 2.57 0.010
BG Share, Renters Who are Latino -0.0380 0.0324 -0.0432 -1.17 0.242
BG Share, Renters Who are Asian -0.0612 0.0332 -0.0367 -1.84 0.065
Owner Clusters (total 43) where p-value <0.10
Owner 3 0.1355 0.0231 0.0914 5.87 0.000
Owner 5 -0.0872 0.0214 -0.0527 -4.08 0.000
Owner 8 -0.0575 0.0171 -0.0325 -3.36 0.001
Owner 10 0.0643 0.0225 0.0351 2.86 0.004
Owner 13 0.0738 0.0349 0.0370 2.12 0.035
Owner 16 0.0541 0.0307 0.0259 1.76 0.078
Owner 17 -0.0303 0.0179 -0.0145 -1.69 0.092
Owner 19 0.1005 0.0348 0.0430 2.89 0.004
Owner 22 -0.0297 0.0151 -0.0127 -1.97 0.049
Owner 24 -0.0409 0.0217 -0.0175 -1.89 0.059
Owner 30 0.0253 0.0129 0.0094 1.96 0.051
Owner 34 -0.0513 0.0214 -0.0190 -2.40 0.017
Owner 37 -0.0461 0.0167 -0.0171 -2.76 0.006
Owner 41 -0.0394 0.0207 -0.0146 -1.90 0.058
N = 1,844
R-square = 0.33
*Also included as independent variables (not shown here) are spatial dummies for 36 “superdistricts” within the 5-county region
and an additional 29 owner cluster dummy variables.
**Clustered-Robust standard errors. Clustered at the block group level
Underlined, p<0.10; bold, p<0.05; 0.01
35
Table 5. Serial Filing Share, Estimation by Multifamily Building*
Coefficient
Std Erro
r
**
Beta t
Si
g
.
Constant 0.1496
0.0767
. 1.95
0.051
Units in Building 2.09E-04
3.87E-05
0.1531 5.41
0.000
Value/Unit ($) 9.23E-08
9.09E-08
0.0273 1.01
0.311
Sale in Same Year -0.0416
0.0161
-0.0517 -2.58
0.010
Sale in Prior Year -0.0128
0.0135
-0.0204 -0.95
0.341
Sale 2-3 Year Prior -0.0103
0.0128
-0.0195 -0.81
0.421
Age of Building (years) -0.0012 0.0004 -0.1003 -3.15 0.002
BG Share, Renters Who are Rent Burdened 0.0549
0.0394
0.0374 1.39
0.164
BG Median Gross Rent ($) 6.06E-05
4.09E-05
0.0571 1.48
0.139
BG Share in Poverty -0.0168
0.0408
-0.0125 -0.41
0.681
BG Share, Female Headed Households 0.0377
0.0744
0.0204 0.51
0.612
BG Share, Over 25 w/College Education 0.1226
0.0549
0.1264 2.23
0.026
BG Share, Households w/Children <17 -0.0102
0.0605
-0.0067 -0.17
0.866
BG Share, Renters >16 Who are 65+ -0.1083
0.0578
-0.0480 -1.88
0.061
BG Share, Renters Who are Black 0.1395
0.0443
0.1911 3.15
0.002
BG Share, Renters Who are Latino 0.0056
0.0580
0.0043 0.10
0.923
BG Share, Renters Who are Asian -0.0297
0.0735
-0.0121 -0.40
0.687
Owner Clusters
(
total 43
)
where
p
-value <0.10
Owner 2 0.0903
0.0345
0.0414 2.61
0.009
Owner 4 0.1740
0.0342
0.0777 5.09
0.000
Owner 7 0.1165
0.0506
0.0463 2.31
0.021
Owner 8 -0.1516 0.0428 -0.0582 -3.54 0.000
Owner 11 0.1399 0.0486 0.0498 2.88 0.004
Owner 12 0.0987 0.0241 0.0351 4.09 0.000
Owner13 0.0669 0.0350 0.0228 1.91 0.057
Owner 17 0.2177 0.0597 0.0707 3.65 0.000
Owner 18 0.0971 0.0568 0.0299 1.71 0.088
Owner 19 -0.1197 0.0385 -0.0348 -3.10 0.002
Owner 20 0.0882 0.0437 0.0256 2.02 0.044
Owner 21 0.1614 0.0522 0.0469 3.09 0.002
Owner 23 0.0998 0.0338 0.0290 2.96 0.003
Owner 28 0.1499 0.0646 0.0408 2.32 0.020
Owner 31 0.0755
0.0308
0.0205 2.45
0.014
Owner 36 -0.2280
0.0684
-0.0574 -3.33
0.001
Owner 38 -0.1169 0.0234 -0.0294 -5.00 0.000
Owner 39 0.2269 0.0565 0.0522 4.02 0.000
Owner 40 0.3260 0.0685 0.0821 4.76 0.000
Owner 41 0.1081 0.0594 0.0272 1.82 0.069
Owner 42 -0.1324 0.0615 -0.0333 -2.15 0.032
N = 1,844
R-square = 0.22
*Also included as independent variables (not shown here) are spatial dummies for 36 “superdistricts” within the 5-county
region and an additional 22 owner cluster dummy variables.
**Clustered-Robust standard errors. Clustered at the block group level.
Underlined, p<0.10; bold, p<0.05; bold and underline p<0.01
36
Table 6. Neighborhood Racial Composition of Properties Owned by Low and High Serial
Filers
Share of Renters
who are Black*
Properties of
Owners with Low
Serial Filing Share
Properties of
Owners with High
Serial Filing Share
All Multifamily
Properties
0-24.9% Black 45.0% 24.2% 22.0%
25-49.9% Black 27.5% 26.1% 25.9%
50-74.9% Black 17.5% 20.0% 18.7%
75%-100% Black 10.0% 29.7% 33.5%
Properties (owners) 40 (5 owners) 165 (16 owners) 1,861
*Block group, per 2011-2016 American Community Survey
37
Figure 1. Multifamily Rental Buildings in the Atlanta 5-County Metro
38
Figure 2. High Serial-Filing Share and Low Serial-Filing Share Owners
39
Notes
i To identify comparison counties from the 20 largest metropolitan areas, we identified
counties in these metros for which the Eviction Lab had eviction filing rate and eviction rate
data. We then screened out counties that had population densities greater than 4,000 people per
square mile or less than 1,400 per square mile. (The five core counties range from 1,748 to
2,586 persons per square mile.) This left us with the 22 comparison counties.
ii For a detailed description of the eviction process in Georgia, see Raymond et al. (2018).
iii Raymond et al. (2018) found that, in 2015 in Fulton County (the largest of the 5 metro
counties examined here), eviction filing rates were 28 per 100 units, compared to 7 per 100 for
single-family rentals.
iv For Fulton County, instead of using 2016 appraised values, 2017 values were used.
Fulton County did not reassess properties from 2011 through 2016, so 2017 appraised values
were deemed more accurate estimates of 2016 market values than 2016 appraised values.
