by Brian Y. An, Andrew Jakabovics, Jing Liu, Anthony W. Orlando, Seva Rodnyansky, Richard
Voith, Sean Zielenbach, and Raphael W. Bostic
309
Cityscape: A Journal of Policy Development and Research • Volume 25, Number 2 • 2023
U.S. Department of Housing and Urban Development • Office of Policy Development and Research
Cityscape
Factors Affecting Spillover
Impacts of Low-Income Housing
Tax Credit Developments: An
Analysis of Los Angeles
Brian Y. An
Georgia Institute of Technology
Andrew Jakabovics
Enterprise Community Partners
Jing Liu
Econsult Solutions
Anthony W. Orlando
California State Polytechnic University, Pomona
Seva Rodnyansky
Occidental College
Richard Voith
Econsult Solutions
Sean Zielenbach
SZ Consulting
Raphael W. Bostic
Federal Reserve Bank of Atlanta
Andrew Jakabovics is employed by Enterprise Community Partners, whose subsidiary, Enterprise
Community Investments (ECI), syndicates low-income housing tax credits. Although ECI may have
syndicated tax credits attached to properties analyzed in this study, all the data came from public datasets
and independent proprietary sources. No ECI employees or resources participated in the research.
JPMorgan Chase & Co. supported this work. The funders were not involved in conducting this research or
in the preparation of this article. The views in this paper do not necessarily reflect the views of the Federal
Reserve Bank of Atlanta or the Federal Reserve System.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
310
Refereed Papers
Abstract
The Low-Income Housing Tax Credit (LIHTC) program is one of the largest sources of financing for
aordable housing in the United States. Contrary to many residents’ fears, research typically shows
that LIHTC-financed properties generate positive spillover impacts in their surrounding communities
in the form of increased housing prices. Some critics yet suspect that the overall positive eects obscure
the properties’ negative impacts for a significant subset of neighborhoods. This article examines these
concerns by assessing the housing price eects of LIHTC properties in Los Angeles. The authors explore
how the eects dier based on various characteristics of the LIHTC property and of the surrounding
neighborhood. The authors supplement these statistical analyses with interviews of key aordable
housing developers to understand their decisionmaking process regarding the siting and structuring of
LIHTC properties. Regardless of the property or neighborhood characteristics, LIHTC developments in
the region have positive spillover price eects. These findings can help inform policymakers who strive to
maximize the secondary benefits of aordable housing developments.
Introduction
The country continues to suffer from a significant shortage of affordable rental housing, a problem
that has worsened since the onset of the COVID-19 pandemic. Households that struggle to pay the
rent tend to be more likely to suffer from poor health and chronic illness. They are more likely to
experience food insecurity, and their children are more likely to struggle academically. Perhaps not
surprisingly, many developers view the creation and rehabilitation of affordable housing not only
as an end in and of itself, but also as a central component of a strategy to stabilize and revitalize
lower-income communities.
At the same time, many homeowners continue to have a negative perception of affordable rental
housing properties. Influenced in part by demagogic politicians and well-publicized concerns
about crime, these homeowners fear that the presence of a publicly subsidized rental housing
development in their community will have negative effects on local property values and public
safety. This “not in my backyard” sentiment has been most evident in middle- to upper-income
neighborhoods, and it frequently has racial or ethnic overtones. (Interestingly, the sentiment
typically represents itself simply as opposition to a proposed project, not necessarily as a preference
for another use of the property.)
A growing body of research offers evidence that rebuts the negative perception of affordable
housing properties. The largest public subsidy in the country, the federal Low-Income Housing
Tax Credit (LIHTC) program, has facilitated equity investments in properties that collectively
have created or rehabilitated, or both, more than 3 million affordable rental units since 1986.
Researchers have found that LIHTC-financed developments generally have neutral to positive
effects on surrounding property values. Several studies have documented the relatively long-lasting
nature of these spillover effects.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
311
Cityscape
The research also has documented considerable variation in the extent and duration of the property
value effects—not only across metropolitan areas, but also within individual cities. Neighborhoods
are inherently dynamic environments, with multiple internal and external factors that affect local
real estate values and quality of life indicators. Some of those factors affect multiple communities;
some are idiosyncratic in nature. It is not surprising, therefore, that developments in different
neighborhoods could or would have different spillover effects. Unfortunately, not enough is known
about the causes of these variations.
A better understanding of the factors that most influence the spillover effects of affordable
housing developments is important for several reasons. It can help policymakers better allocate
and target comparatively scarce housing and neighborhood development resources. It can help
developers focus activity in areas where the local dynamics create a more favorable environment
for positive project spillover. A greater understanding of the interplay between affordable housing
developments and local dynamics also can help inform—and ideally alleviate—the persisting
opposition to subsidized rental housing.
All these outcomes are in addition to the primary focus and benefit of the LIHTC and other
affordable housing programs: The creation of quality homes that do not impose cost burdens on
their residents. The intent is not to change the LIHTC program into something for which it was
not intended. Rather, this research seeks to determine how various actors can create and preserve
affordable housing in a way that best contributes to the stabilization, enhancement, or revitalization
of the surrounding community.
This study begins to tease out the factors that influence the type and extent of a LIHTC property’s
effects on its neighborhood. It uses changes in residential home values as a proxy for improvements
in the area. If a community becomes more attractive, it should increase the willingness and
desire of people to live there. That increased interest should translate into increased demand
for local property, which will bid up local real estate values. In the model used here, therefore,
the “treatment effect” is the percent change in nearby prices that occurs after a new LIHTC
development is completed, relative to the change in comparable non-LIHTC neighborhoods.
This study examines LIHTC price effects while controlling for a range of project-specific
and neighborhood-level factors. On the project side, it considers the role that the size of the
development (that is, its number of units) plays in changing local values. It assesses whether
spillover effects are greater around entirely subsidized LIHTC properties or around those
developments with a mix of subsidized and market-rate units. It also assesses whether the
corporate structure of the developer has an effect: Do for-profit sponsored projects have a different
effect on the surrounding area’s property values than those sponsored by nonprofit organizations?
With respect to neighborhood dynamics, this study examines whether LIHTC developments have
greater spillover effects in low-income, high-income, or more moderate-income communities. It
examines property effects in communities with higher and lower proportions of people of color and
in predominantly Black, Latino, and Asian neighborhoods. It also analyzes whether the concentration
of LIHTC properties in a community enhances or limits local property value appreciation.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
312
Refereed Papers
This study focus on Los Angeles, a city with an extreme shortage of affordable housing and very
strong competition for relatively scarce LIHTC allocations. It finds that, across each of the different
property- and neighborhood-level dimensions, the average LIHTC property contributes to a
meaningful increase in surrounding property values. Not surprisingly, some factors lead to more
significant property value increases than others. Although the market and political dynamics in
Los Angeles may limit the generalizability of some of these findings to other markets, several of the
findings are actionable for both developers and policymakers.
Contribution to the Literature
This analysis focuses on the spillover effects of LIHTC-financed properties that involve either
new construction or rehabilitation (and potential expansion) of existing properties. Substantial
evidence shows that both types of residential projects positively influence local property values.
In Cleveland, for instance, new construction in the 1980s and 1990s increased the sale price of
nearby homes by about $5,000, whereas significant rehabilitation had a positive $4,000 effect
(Ding and Knaap, 2002; Ding, Simons, and Baku, 2000; Simons, Quercia, and Maric, 1998). One,
therefore, would expect similar outcomes for LIHTC developments, as they typically involve either
new construction on a previously vacant lot or the often-significant enhancement of one or more
existing residential, commercial, or mixed-use properties. Because a typical LIHTC development
contains 60 to 80 units, it has a large enough physical footprint to have a noticeable impact on its
surrounding area.
Indeed, a growing body of research has documented that LIHTC properties have neutral to positive
effects on surrounding real estate prices. Exhibit 1 summarizes several of these analyses.
Exhibit 1
Selected Analyses of Low-Income Housing Tax Credit Properties’ Effects on Home Prices (1 of 2)
Study Market Basic Spillover Findings
Green, Malpezzie,
and Seah (2002)
Madison and
Milwaukee, WI
No evidence that LIHTC properties depressed surrounding
home sale prices; some evidence that properties near LIHTC
developments in Madison appreciated more rapidly than
those elsewhere in the city.
Johnson and
Bednarz (2002)
Cleveland, OH;
Portland, OR;
Seattle, WA
Property values increased within a few blocks of LIHTC
developments after the developments had been placed
in service.
Furman Center (2006)
Ellen and Voicu (2007)
Ellen et al. (2007)
New York City, NY Property values surrounding LIHTC buildings increased by as
much as 9% in the 5 years after the LIHTC property’s opening.
Ezzet-Lofstrom and
Murdoch (2006)
Dallas, TX LIHTC developments had a small, positive effect on
surrounding single-family house prices.
Baum-Snow and
Marion (2009)
National Home prices increased by an average 14.9% in census block
groups within 1 kilometer of a LIHTC property.
Woo, Joh, and
Van Zandt (2016)
Cleveland, OH Home values near LIHTC developments increased 15.4%
relative to price trends elsewhere in the city.
Young (2016) National (20
highest cost
markets)
Proximity to LIHTC property had no signicant effect on
home values.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
313
Cityscape
Exhibit 1
Selected Analyses of Low-Income Housing Tax Credit Properties’ Effects on Home Prices (2 of 2)
Study Market Basic Spillover Findings
Edmiston (2018) Kansas City, MO LIHTC properties had little positive or negative effect on
surrounding property conditions.
Bostic et al. (2020) Cook County, IL Home values within 1/8 of a mile of a LIHTC development
experienced a 10.8 percentage point increase relative to the
countywide average.
LIHTC = low-income housing tax credit.
Although the average spillover price effects of LIHTC developments are generally positive,
the averages mask considerable variation within and across regions. Such variations are not
surprising, given the often-substantial differences in amenities and real estate trends affecting
even adjacent neighborhoods within the same city. Several researchers have taken specific local
factors into account in their analyses of affordable housing developments’ spillover effects. Others
have examined some of the factors specific to the affordable housing properties. The following
subsections highlight some of their key findings.
Project Size
Intuitively, one would expect larger residential developments to have commensurately larger
effects—either positive or negative—on surrounding neighborhood conditions than smaller
developments. Larger developments occupy more physical space and are consequently more
visible within an area. They also house more people and, therefore, increase the community’s
population density.
Multiple studies of affordable housing developments found that larger projects tend to have greater
spillover effects, as exhibit 2 summarizes.
Exhibit 2
Selected Analyses of the Size Effects of Low-Income Housing Tax Credit Properties
Study Market Key Project Size Findings
Ellen (2007) New York City, NY Affordable housing properties with more units generally
generated more positive spillover price effects, although the
marginal benet decreased with project size.
Deng (2011a) Santa Clara County, CA LIHTC properties with 50 or more units boosted surrounding
home values 5 to 6%; smaller properties had no signicant
spillover effect.
Dillman, Horn,
and Verilli (2017)
Review of 24 separate
studies across
multiple markets
Larger, well-managed affordable housing properties tend to
generate more signicant spillover price effects, and they also
contribute to reductions in local violent crimes.
LIHTC = low-income housing tax credit.
At the same time, larger projects can be problematic in certain markets. Several analyses
documented the potential for poorly managed developments of scale to exacerbate local crime
issues and contribute to neighborhood decline (Dillman, Horn, and Verilli, 2017). Mid- to
large-sized multifamily properties placed in service in low-density areas can negatively affect
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
314
Refereed Papers
surrounding home prices, especially in more affluent communities (Ericksen and Yang, 2022).
Even in areas with relatively high population densities such as New York City, spreading affordable
units across multiple properties instead of concentrating them in one or two developments can
result in greater and more positive overall property value effects (Ellen and Voicu, 2007). Density
concerns drive much of the “not in my backyard” opposition to affordable housing. Local residents
fear that the increased population density associated with a larger development will irrevocably
alter the existing community dynamics.
Extent of Project Subsidization
The LIHTC program does not require all units within a tax credit financed property to be
affordable to low-income households. In fact, the income mandates can appear comparatively
modest. Statutorily, developers must commit to creating properties that meet one of three income
thresholds: (1) Households earning 50 percent or less of the area median income (AMI) occupy at
least 20 percent of the units; (2) households earning 60 percent or less of AMI occupy at least 40
percent of the units; (3) households earning an average of 60 percent or less of AMI occupy at least
40 percent of the units, and no household in the property makes more than 80 percent of AMI.
The regulations, therefore, give developers flexibility. A developer can opt to create a facility in
which a portion of the units rent at market rates. Alternatively, the developer can elect to have
all the units be affordable to households making 60 percent or less of AMI—in which case the
property becomes eligible for the maximum LIHTC subsidy. LIHTC developers have tended
to focus primarily on creating income-restricted units, in part, as a way of increasing their
competitiveness in the tax credit allocation process. An analysis of 12,228 LIHTC properties
containing more than 760,000 units in 16 different states found that households earning 60
percent or less of the prevailing AMI occupied 93 percent of the units (Furman Center, 2012).
Similarly, an examination of LIHTC properties in 18 states found that 81 percent of the properties’
tenants made 50 percent or less of AMI (O’Regan and Horn, 2013).
Still, certain areas contain a fair number of LIHTC properties with market-rate units. In Chicago,
for instance, 19.3 percent of the non-senior LIHTC properties placed in service between 1987 and
2016 contained at least five market-rate units. In those properties, unsubsidized units accounted
for an average 27 percent of all units (Bostic et al., 2020).
Both fully subsidized LIHTC properties and those with a mix of subsidized and market-rate units
have generated positive spillover property value effects. Exhibit 3 summarizes the key findings of
several studies of “mixed-income” and fully subsidized properties financed in part with LIHTC or
federal HOPE VI monies, or both.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
315
Cityscape
Exhibit 3
Selected Analyses of Fully Subsidized Versus Partially Subsidized Affordable Housing
Developments’ Spillover Property Value Effects
Study Market Key Subsidy-Related Findings
Turbov and
Piper (2005)
Atlanta, GA;
Louisville, KY;
Pittsburgh, PA;
St. Louis, MO
Home values in the areas surrounding the mixed-income HOPE
VI developments increased more quickly than elsewhere in the
respective cities.
Castells (2010) Baltimore, MD Of three HOPE VI communities analyzed, only the more mixed-
income community demonstrated positive and signicant spillover
property value increases; no spillover effects were observed
surrounding the fully subsidized HOPE VI developments.
Funderberg and
MacDonald (2010)
Polk County, IA Property value appreciation near fully subsidized family LIHTC
developments was 2 to 4% less than elsewhere in the county;
partially subsidized LIHTC properties had no signicant effects on
price trends.
Zielenbach, Voith,
and Mariano
(2010)
Boston, MA;
Washington, DC
Both partially and fully subsidized HOPE VI developments had
positive property value effects, with the greatest values in areas
already experiencing development pressures.
Cloud and
Roll (2011)
Denver, CO The ¼-mile area around the downtown mixed-income HOPE VI site
had a greater increase in property values and homebuying, a greater
reduction in blight, and a greater increase in other investments than
other similar areas in city.
Bostic et al.
(2020)
Cook County, IL Spillover price effects for LIHTC properties with at least ve market-
rate units were higher than the effects of properties consisting entirely
of subsidized units; the price effects were positive in both cases.
LIHTC = low-income housing tax credit.
The relatively limited literature exploring differences in property value spillover effects of partially
versus fully subsidized affordable housing developments suggests that complexes that include
market-rate units have more positive effects on local home prices. It is important not to draw hard
conclusions at this stage, however. The most in-depth examination of the issue, by Bostic et al.
(2020) in Chicago, found that LIHTC properties with market-rate units had a disproportionately
high effect on nearby home prices in higher-income areas. In lower-income communities, home
values appreciated more near fully subsidized LIHTC properties than near partially subsidized ones.
For-Profit Versus Nonprofit Developer
The LIHTC statute requires that at least 10 percent of all tax credit allocations go to projects
sponsored by nonprofit developers. Several states and localities have allocated higher proportions
of credits to these organizations. Overall, nonprofits were responsible for about 22 percent of the
LIHTC properties placed in service between 1987 and 2004, although that proportion may have
declined since (Bratt, 2007). A 2015 national survey of 100 affordable housing developers found
that, among the 52 most active entities, for-profits were responsible for starting 89 percent and
completing 86 percent of the affordable units produced during the year (Bratt and Lew, 2016).
Several analyses have documented the differences between nonprofit and for-profit developers.
Not surprisingly, the fundamentally disparate goals of the two types of entities help explain much
of the variation. The quest for financial returns drives most for-profit activity, as the developers
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
316
Refereed Papers
need to generate profits for their shareholders. In contrast, nonprofit developers tend to focus more
on neighborhood improvements and affordable housing provision. Nonprofits, consequently, are
more likely to develop properties in poorer areas. Their properties frequently target lower-income
households, and they also are more likely to target people with disabilities, the homeless, seniors,
and other “special” populations (Johnson, 2012; Silverman and Patterson, 2011). Because they
serve more disadvantaged populations, nonprofit properties often charge lower rents. Bratt and
Lew (2016), for instance, found that nonprofit-sponsored LIHTC developments had a higher
proportion of units with a low rent-fair market rent ratio than for-profit developments.
Overall, the cost of developing a LIHTC project tends to be higher for a nonprofit developer than
for its for-profit counterparts. Some of that difference can be attributed to nonprofits being more
likely to engage in the rehabilitation of existing properties and offering more services to tenants.
Whereas for-profit developers are more likely to engage in new construction, with models that
can be replicated across sites, nonprofits often need to develop project-specific designs for existing
properties (Silverman and Patterson, 2011). At least in metropolitan Richmond, Virginia, nonprofit
developers of LIHTC properties have been more likely to incorporate rehabilitation, certified
property management, and standard-use terms in their properties than their for-profit counterparts
(Johnson, 2012). These factors can contribute to greater operating costs. Many nonprofit
developers also struggle to obtain capital from conventional lenders and, therefore, are forced to
piece together different subsidies, particularly if they try to serve very low-income households. This
process can take time and ultimately drive up overall project costs.
Development costs notwithstanding, some evidence exists that nonprofit-developed affordable
housing complexes have at least similar, and potentially more positive, effects on surrounding
home values than properties developed by for-profit firms. Exhibit 4 summarizes the relatively
sparse literature on the issue.
Exhibit 4
Selected Analyses of Spillover Price Effects from Nonprot and For-Prot Affordable
Housing Developments
Study Market Key Developer-Related Findings
Goetz, Lam,
and Heitlinger
(1996)
Minneapolis, MN Subsidized multifamily properties developed by nonprot CDCs
enhanced the value of surrounding market-rate homes by 86 cents per
square foot. Publicly subsidized housing owned by private for-prots had
a negative 82 cent per square-foot effect on surrounding home prices.
Smith (2003) Indianapolis, IN For 13 years, home prices in areas with signicant CDC activity increased
7.14% relative to homes in non-CDC neighborhoods.
Ellen and
Voicu (2007)
New York City, NY Nonprot-developed, smaller affordable housing properties had larger
home price spillover effects than similar for-prot developed properties.
The price-value effects associated with nonprot-developed projects
were more stable over time.
Deng (2011a) Santa Clara
County, CA
LIHTC projects developed by HPN-member* nonprots had 4 to 6
percentage points higher effects on surrounding values than properties
developed by for-prots or by non-HPN member nonprots.
Edmiston
(2018)
Kansas City, MO CDC investments in owner-occupied, single-family homes contributed to an
11.8% increase in home prices within 500 feet of the targeted properties.
CDC = Community Development Corporation. HPN = Housing Partnership Network. LIHTC = low-income housing tax credit.
* HPN is a national network of high-capacity nonprofit developers.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
317
Cityscape
Neighborhood Income Level
Residential developments do not take place in a vacuum. Their ability to attract and retain tenants,
generate revenue streams, and improve (or worsen) conditions in the surrounding area depends
not only on their size, management, structure, and other characteristics, but also on the dynamics
of the neighborhood. A development’s capacity to generate spillover effects can be augmented or
constrained by, among other factors, the physical geography of its surroundings, the strength or
weakness of the local economy, the extent of real estate development activity, local public safety
issues, and various local demographic and socioeconomic characteristics. Not surprisingly, studies
of LIHTC and other affordable housing developments’ spillover found significant differences in the
direction and extent of property value effects across communities.
In general, research has documented more positive spillover price effects from LIHTC
developments in lower-income areas than in middle- to upper-income communities. Exhibit 5
summarizes the findings of several studies that examined impacts across different communities.
Exhibit 5
Selected Analyses of Low-Income Housing Tax Credit Developments’ Spillover Price Effects by
Neighborhood Income Level
Study Market Key Findings by Income Level
Baum-Snow and
Marion (2009)
National Median home prices increased 14.9% within 1 kilometer of a LIHTC
property, but the price increases were noticeably lower in stable
(10.6%) and gentrifying (5.6%) communities.
Deng (2011a) Santa Clara
County, CA
LIHTC developments in low-income neighborhoods had positive
price effects, but the effects for the county overall were statistically
insignicant.
Deng (2011b) Miami-Dade
County, FL
Price effects were most positive around LIHTC developments in high-
poverty neighborhoods and most negative around developments in
middle-class communities.
Woo, Joh, and
Van Zandt (2016)
Charlotte, NC LIHTC properties had negative effects on surrounding values, but the
effects were much more noticeable in moderate- and upper-income
areas than in lower-income ones.
Woo, Joh, and
Van Zandt (2016)
Cleveland, OH The home price effects of LIHTC properties were much lower in lower-
income areas than in more moderate- and upper-income communities.
Dillman, Horn,
and Verilli (2017)
Summary of 24
studies spanning
the country
LIHTC and other affordable housing properties generally boosted
values in low-income areas but had more mixed effects in moderate-
and high-opportunity areas.
Diamond and
McQuade (2019)
Multistate Home prices within 1/10 of a mile of a LIHTC property increased
6.5% during 10 years in low-income neighborhoods but declined
nearly 2.5% in higher-income areas.
LIHTC = low-income housing tax credit.
LIHTC properties tend to be developed in relatively distressed areas. Nationally, 32 percent of
LIHTC units placed in service prior to 2011 were in census tracts with poverty rates of at least 30
percent in 2010, and another 23 percent of the units were in tracts with poverty rates between
20 and 30 percent. In 12 sampled states, the average LIHTC unit sat in a tract where the poverty
rate was 6 percentage points higher than that of a tract housing a typical unsubsidized rental unit
in the same metropolitan area (Ellen, Horn, and Kuai, 2018). The concentration of properties in
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
318
Refereed Papers
weaker markets reflects both LIHTC allocation criteria—which frequently give projects greater
points for being in more distressed communities—and some developers’ desire to use the LIHTC
developments to help catalyze other investments in the area.
The relatively greater spillover benefits of LIHTC properties in lower-income communities should
not mask the positive effects that these developments often have in more affluent areas, however.
For instance, an analysis of Cleveland in the 1990s and early 2000s found positive price effects
of LIHTC developments in moderate- and upper-income communities. In fact, the effects were
greater in those areas than in the city’s more distressed markets (Woo, Joh, and Van Zandt, 2016).
An evaluation of Chicago trends encompassing the same period found strong and enduring LIHTC
price effects in both lower- and upper-income neighborhoods (Voith et al., 2022).
It is also possible that the observed negative effects of LIHTC properties in some higher-income
neighborhoods may result less from the introduction of affordable housing per se and more
from the introduction of comparatively dense multifamily properties in lower-density areas
with a preponderance of single-family homes. A recent study reran the nonparametric models
that Diamond and McQuade used in their 2019 analysis (see exhibit 5), adding unsubsidized
multifamily developments and LIHTC properties to the analysis (Eriksen and Yang, 2022). The
new study found that all types of multifamily developments depressed surrounding property values
in higher-income areas. Once they controlled for population density, the authors found that the
negative effects of LIHTC properties dissipated and even became moderately positive.
Neighborhood Racial and Ethnic Composition
Most research on LIHTC spillover price effects has taken the subject neighborhoods’ racial and
ethnic composition into account as part of the evaluation structure. Because of the strong inverse
correlation between neighborhood incomes and the communities’ proportions of individuals of color,
the observed effects of racial or ethnic composition largely have tracked the observed income-related
effects. Spillover price effects tend to be greater in predominantly Black and Latino communities.
Comparatively, little analysis has been done on different effects in higher- versus lower-income
communities of color—the Bostic et al. (2020) analysis of mixed-income properties in Chicago
being a notable exception. Similarly, little in-depth examination has been done on the spillover
effects in different types of majority-minority neighborhoods. It is not clear, for instance, whether
LIHTC properties have different price effects in predominantly Black, predominantly Latino, or
predominantly Asian-American neighborhoods.
LIHTC Project Concentration
One of the challenges in assessing the spillover impact of LIHTC developments is that the properties
tend to be geographically concentrated. Quite often, some overlap is among the distance bands
surrounding individual LIHTC properties. A home sale transaction, therefore, may be included
in multiple analyses, which can complicate the assessment of any single LIHTC development’s
true impact. Such an issue affects evaluations of many affordable housing properties, but LIHTC
developments especially, as they tend to be much more concentrated than other subsidized housing
properties (Oakley, 2008). In New York City, for example, 71 percent of LIHTC properties were
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
319
Cityscape
clustered; in Boston, the proportion was 50 percent (Dawkins, 2013). More than 90 percent of the
non-senior LIHTC properties placed in service in Cook County, Illinois, between 1987 and 2016
were within one-half a mile of at least one other LIHTC development (Voith et al., 2022).
To date, few analyses have addressed the effects of this concentration directly. Deng’s (2011b)
analysis of LIHTC-related effects in south Florida found mixed effects of concentrated development
in Miami-Dade County. Some areas with multiple LIHTC properties showed improvement, but
the presence of multiple LIHTC developments was potentially worsening conditions in certain
suburbs. Two reviews of the affordable housing assessment literature raised concerns about the
property value implications of geographically concentrated subsidized housing and the low-
income households such complexes support (Dillman, Horn, and Verilli, 2017; Nguyen, 2005).
Those concerns were not based on LIHTC-specific findings. Moreover, some evidence exists that
clustering affordable housing properties can have more beneficial effects on a community than
introducing a single property, based on an analysis of a scattered-site public housing program in
Denver (Santiago, Galster, and Tatian, 2001).
The one study to date that deliberately addressed the impacts of LIHTC project clustering focused
on Chicago. Voith and his colleagues (2022) found that the introduction of a single LIHTC
property to a community had positive and sustainable impacts on surrounding home prices. They
did not find any evidence that placing subsequent LIHTC developments in the neighborhood
detracted from the positive benefits associated with the initial property. In some cases, the
subsequent LIHTC properties had positive and additive effects on surrounding values.
Working Hypotheses
As described previously, a range of studies has documented the generally positive (or at least
neutral) overall effects of LIHTC developments on surrounding home prices—findings that rebut
the perception that such developments have inherently negative effects on communities. At the
same time, they have demonstrated the range of project-specific and neighborhood-level factors
that can influence such developments’ spillover impacts. Most analyses have incorporated only
a few of these independent variables, generally ignoring the complexities associated with the
tendency of LIHTC properties to be geographically concentrated. This study represents an initial
attempt to account for this wider range of factors in a single analysis.
Based on previous findings, it is expected that the introduction of a LIHTC property in a
community typically will have a positive and lasting effect on surrounding home prices. That
positive effect is likely to be more pronounced in low-income communities than in more affluent
areas, and the introduction of one or more subsequent nearby LIHTC properties is likely to
augment it.
This study posits that larger LIHTC properties and those nonprofit entities develop are likely to
have somewhat greater spillover price effects than smaller properties and those for-profit firms
develop or rehabilitate. (These relationships should hold even after controlling for the fact that
larger, nonprofit-sponsored properties are more prevalent in lower-income communities.) It is also
anticipated that partially subsidized LIHTC properties—those containing a mix of market-rate
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
320
Refereed Papers
and income-restricted units—will have somewhat greater spillover effects on area prices than fully
subsidized properties, in part, because the higher-income residents’ additional purchasing power
will contribute to the attraction and retention of a wider range of local retail and other amenities.
It is not expected that this analysis will find any meaningful difference in price effects across
neighborhoods that are predominantly Black, Latino, or Asian-American (again controlling for
neighborhood income level).
LIHTC Developments in Los Angeles
To understand better the variations in the spillover effects of LIHTC properties, this study examines
both property and neighborhood characteristics in Los Angeles County, California. Los Angeles
is the country’s largest county and contains 833 LIHTC properties. It has a widespread and
widely acknowledged need for affordable housing, with several public, private, and philanthropic
initiatives working to alleviate the shortage. It has considerable demographic and socioeconomic
diversity, and it continues to be one of the country’s strongest real estate markets. It also has strong
political support for creating and preserving affordable housing. A Los Angeles-based analysis,
therefore, can be beneficial for developers and policymakers looking to address affordable housing
needs in other large cities with strong real estate markets, diversity of population and income, and
a political commitment to helping address residents’ housing cost burdens.
Data
This study analyzes the spillover effects of LIHTC developments placed in service in Los
Angeles County between 1987 and 2015. Pre-development and post-development prices in the
neighborhoods with one or more LIHTC developments are compared to price trends during the
period in neighborhoods with no LIHTC properties.
Data was obtained from HUD for each of the 833 LIHTC properties placed in service during that
period. The information includes the property’s street address, the year it was placed in service,
and its total number of units. Data on all Los Angeles residential property sales from 1987 to
2015 (more than 1.8 million arm’s length transactions) was obtained from DataQuick Information
Systems, Inc. and CoreLogic, Inc., the transactions were geocoded, then the distance between each
sold home and nearby LIHTC developments was calculated. During the 28-year period, 145,056
transactions were within one-fourth of a mile of a LIHTC property, and 362,811 transactions were
within one-fourth to one-half of a mile.
Exhibit 6 highlights the differences between Los Angeles County census tracts that contain at least
one LIHTC development and those without any such properties during the study period. As the
exhibit shows, Los Angeles’s LIHTC properties tend to be in disproportionately low-income areas.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
321
Cityscape
Exhibit 6
Characteristics of Los Angeles Median Low-Income Housing Tax Credit and Non-Low-Income
Housing Tax Credit Census Tracts
Variable Median LIHTC Tract Median Non-LIHTC Tracts
Household income $46,883 $71,750
Population 4,362 4,182
White non-Hispanic residents 9.6% 19.5%
Black non-Hispanic residents 5.6% 3.1%
Asian non-Hispanic residents 7.9% 9.8%
Hispanic residents 58.9% 41.5%
Poverty rate 21.1% 11.0%
Unemployment rate 6.4% 5.5%
Residential vacancy rate 5.3% 4.9%
Median gross rent $1,259 $1,544
Median home value $473,250 $545,400
LIHTC = low-income housing tax credit.
Source: 2016 American Community Survey 5-year estimates
These characteristics are similar to those Basolo, Huarita, and Won (2022) identified in their
recent analysis of LIHTC developments in the county. They found that, relative to residential
properties generally, LIHTC properties tend to be in neighborhoods that have more economic
hardship, higher population density, a higher proportion of renter-occupied units, and more racial
and ethnic diversity.
LIHTC properties in Los Angeles also tend to be clustered geographically. Of the county’s 833
properties, 679 are within one-half a mile of at least 1 other LIHTC property. The greatest
concentration of these developments is in south-central Los Angeles, as exhibit 7 illustrates. The
different dots indicate both non-overlapping properties in the city and the overlapping ones.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
322
Refereed Papers
Exhibit 7
Map of Sampled Low-Income Housing Tax Credit Properties and Surrounding One-Half-Mile Radii
LA = Los Angeles. LIHTC = low-income housing tax credit.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
323
Cityscape
Methodology
The aim here is to illuminate the roles that different project- and neighborhood-level characteristics
play in influencing the spillover property value effects of LIHTC development. This analysis
focuses on six different characteristics: (1) development size (small, medium, and large); (2)
proportion of subsidized units (all versus some); (3) developer type (for-profit versus nonprofit);
(4) neighborhood income level; (5) neighborhood racial and ethnic composition; and (6) the
number of existing LIHTC developments within the neighborhood. These differences are examined
both quantitatively and qualitatively to understand how the various factors influence both the
developers’ decisionmaking and the ultimate spillover impacts.
Quantitative Approach
For the quantitative methods, this study builds on the standard difference-in-differences regression
models typically used in program evaluation studies of this kind (for example, Butts, 2022;
Chen, Glaeser, and Wessel, 2022; Keeler and Stephens, 2022; Voith et al., 2022). Initially, the
typical model used in the literature is created, focusing on the difference in residential prices after
constructing a LIHTC project between houses near the completed project and houses farther away.
This model considers two distance bands: One within one-fourth a mile of the LIHTC project
and the second in the area between one-fourth and one-half a mile from the LIHTC project.
1
This
model is illustrated in the following equation 1.
(1) ln(P
itk
) = ∑
d∈D
α
0d
Pre
idt
+ ∑
d∈D
α
1d
Post
idt
+
β
X
it
+
ε
k
+ τ
t
+ μ
itk
,
Where—
ln (P
itk
) is the natural log of the price of house I at time t in census tract k;
D is a set of distance bands d, where D = {0–¼ miles, ¼–½ miles}
Pre
idt
is a dummy variable equal to 1 if the transaction of house I in distance band d at
time t is prior to the construction of a LIHTC project;
Post
idt
is a dummy variable equal to 1 if the transaction of house i in distance band d at
time t is after the construction of a LIHTC project;
X
it
is a vector of hedonic characteristics of house i at time t;
2
ε
k
is a vector of k tract-specific fixed effects;
τ
t
is a vector of t year-specific fixed effects;
3
and
μ
itk
is a random error variable.
1
These distance bands are common in the literature (for example, Diamond and McQuade, 2019; Orlando and Welke,
2022). Other distance bands were tested, yielding very similar results.
2
These hedonic characteristics include living area square footage, lot size square footage, floor-area ratio, age at sale,
number of stories, distance to central business district, seller type, and seasonal dummies (spring, summer, fall).
3
Time-specific fixed eects control for marketwide inflation, allowing the model to use nominal house prices as the
outcome variable.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
324
Refereed Papers
All transactions are included—those within and outside of the distance bands—to provide a better
estimate of the price trends within the overall market and to provide more data on the value of
individual housing and neighborhood traits. The “average treatment effect”
4
is the difference between
the pre-treatment and post-treatment coefficients (
α
1d
—
α
0d
) for a given distance band d.
5
The pre-
treatment baseline is included explicitly to show how the LIHTC neighborhoods were different from
the control neighborhoods before the introduction of the affordable housing property.
The resulting treatment effect compares the average change in property values of properties within
the distance band with the average change in values of homes more than one-half a mile away,
controlling for overall marketwide changes in prices. This control group offers a counterfactual for
the average pre-post LIHTC change in property values of homes outside the distance bands. This
distance distinction is a typical counterfactual for spatial difference-in-differences studies, as these
neighborhoods are within the same metropolitan area and experience many of the same supply and
demand shocks, especially after controlling for hedonic characteristics, tract-specific fixed effects,
and year-specific fixed effects. This approach focuses on the difference in levels before and after
development rather than the difference in trends, because previous research (for example, Voith
et al., 2022) has demonstrated that level changes are the dominant impact. Assessing changes in
levels also allows for a less complicated exposition of each model described in the following.
Having mirrored the standard model, the analysis then expands on it by incorporating different
factors associated with the treatment effect. One by one, each of the six subcategorizations are
considered, across which one might expect to find different LIHTC effects.
When assessing the importance of neighborhood income and race and ethnicity composition, the
sample is subdivided, and separate regressions are run for the different groups of neighborhoods.
For income, the analysis looks at trends in low-income communities (those census tracts whose
median household incomes fell in the bottom one-third of all Los Angeles tracts per the 2016
American Community Survey’s 5-year estimates) and medium- to high-income communities (the
remaining tracts). For race and ethnicity, the analysis uses the same census tract tercile approach
to analyze communities with high proportions of Black, Latino, Asian, and non-White residents,
respectively. In each case, the remaining two-thirds of the tracts serve as our “control group” for the
analysis. Equation 1 is then estimated separately for each subsample.
Note that a single LIHTC development can have both high- and low-income neighborhoods within
one-half of a mile of the development. The split-sample subgroup approach is a clean way of
estimating different LIHTC impacts in communities with different incomes and demographics.
Exhibit 8 shows the differences in sample sizes for each category of analysis. Of the 1.8 million
transactions in the whole sample, approximately 26 percent are within one-half a mile of a
LIHTC development, as shown in column 2 (487,453). This ratio ranges from 10 percent for the
low-income subsample to 45 percent for the high-Hispanic subsample. Even where it is lowest,
4
In other words, the typical price eect associated with a LIHTC development being placed in service.
5
As with any dierence-in-dierences analysis with treatments in multiple periods, the amount of “pre” and “post” years
available for each treatment depends on the timing of the treatment (that is, the introduction of the LIHTC property).
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
325
Cityscape
however, there are still more than 75,000 transactions, ensuring enough statistical power to detect
significant effects.
Exhibit 8
Number of Transactions Near Low-Income Housing Tax Credit by Racial and Ethnic Subsample
Sample
Transactions Within
½ Mile of LIHTC
Transactions More
Than ½ Mile Away
Total Transactions
Whole sample 487,453 1,355,272 1,842,725
Low % non-White 282,395 1,059,632 1,342,027
High % non-White 205,058 295,640 500,698
High % Black 252,599 342,078 594,677
Low % Black 234,854 1,013,194 1,248,048
High % Asian 111,384 473,298 584,682
Low % Asian 376,069 881,974 1,258,043
High % Hispanic 200,613 250,017 450,630
Low % Hispanic 286,840 1,105,255 1,392,095
Low income 75,459 663,251 738,710
High income 411,994 692,021 1,104,015
LIHTC = low-income housing tax credit.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database; U.S. Census Bureau
It is not easy to subdivide the sample to analyze the property value effects associated with different
property-specific characteristics or for neighborhoods with multiple LIHTC developments.
Too many transactions involve properties that fall within distance bands of different properties
developed at different times. For these factors, a regression is run with the entire sample, each
time focusing on the subcategory under consideration by adding a new, factor-specific variable and
two different dummy variables. Exhibit 9 shows that the sample size for each of these interaction
variables is large enough to estimate these coefficients. This approach allows for the examination of
effects associated with each characteristic while controlling for nearby properties. See the following
equation 2.
(2) ln P
itk
= ∑
d∈D
α
0d
Pre
idt
+ ∑
s∈S
∑
d∈D
α
1ds
Post
idst
+
β
X
it
+
ε
k
+ τ
t
+ μ
idstk
,
Where—
Pre
idt
is a dummy variable equal to 1 if the transaction of house i in distance band d at time
t is prior to the construction of a LIHTC project; and
Post
idst
is a dummy variable equal to 1 if the transaction of house i in distance band d at time
t is after the construction of a LIHTC project with either neighborhood or property
characteristic s.
S is defined as one of the following sets of neighborhood or property characteristics—
1. S = {1 LIHTC project nearby, 2 LIHTCs projects nearby, 3+ LIHTC projects
nearby};
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
326
Refereed Papers
2. S = {small LIHTC project, medium LIHTC project, large LIHTC project};
3. S = {mixed-income LIHTC project, fully subsidized LIHTC project}; or
4. S = {for-profit LIHTC developer; nonprofit LIHTC developer}.
Exhibit 9
Number of Transactions Near Low-Income Housing Tax Credit by Interaction Variable
Interaction
Variable
For-Profit Versus
Nonprofit
Subsidy Project Size
Project
Concentration
½+ mile away* 1,628,449 1,628,449 1,628,449 1,628,449
For-prot 130,312
Nonprot 83,964
Partially subsidized 56,845
Fully subsidized 196,431
Small property 93,601
Medium property 70,935
Large property 49,740
1 LIHTC project 214,276
2 LIHTC projects** 55,791
3+ LIHTC projects** 18,058
LIHTC = low-income housing tax credit.
* Includes transactions of homes more than one-half a mile from a LIHTC property at the time of sale.
** Figures represent subsets of transactions within one-half a mile of a single property.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Potential Endogeneity Issues
These findings could reflect some implicit site selection bias if developers have chosen to locate the
LIHTC properties in neighborhoods where values are already trending upward. The model could
potentially be revealing existing appreciation trends, not changes associated with the introduction
of the LIHTC property.
Two factors lend credence to the selection bias concern. First, developers are inherently more likely
to locate properties in areas where they can obtain the greatest tax credit benefit. In their national
analysis, Baum-Snow and Marion (2009) found that LIHTC properties in program-designated
qualified census tracts (QCTs) had an average of six more units than properties in tracts that
fall just below the QCT eligibility threshold. Basolo, Huarita, and Won (2022) found a positive,
statistically significant association between LIHTC neighborhoods and QCTs in Los Angeles
County. Second, private developers are more likely to select properties in gentrifying—or at least
appreciating—neighborhoods than in stable or declining ones (Baum-Snow and Marion, 2009;
Ellen and Voicu, 2007).
To the extent that developers seek to maximize profits from rents or property appreciation, an
incentive is for them to build or rehabilitate properties in improving neighborhoods. Although they
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
327
Cityscape
may earn additional points in the LIHTC allocation process for targeting properties in QCTs or
difficult development areas, they could deliberately target properties in distressed areas displaying
clear signs of improvement. If property values are already trending upward in these areas, it
becomes harder to demonstrate convincingly that the LIHTC development is responsible for the
observed appreciation of the local market. Well-crafted statistical models may be able to document a
post-development trend in values that is steeper than the pre-development trend, but such findings
prove inherently less noteworthy than those that document a distinct change in trends. Without
knowledge of the developers’ particular location decisions, it is hard to determine the extent to
which observed neighborhood effects should be attributed to the initial selection of the site.
Following standard difference-in-differences methodology, a test is run for any observable evidence
of such behavior by including a linear “pre-trend” (that is, price trends prior to the LIHTC
development) in the model. Within both the ¼- and ½-mile distance bands, the coefficient on this
pre-trend is statistically insignificant—and within the ½-mile band, it is even negative. Thus, no
empirical evidence exists indicating that price trends in the areas around LIHTC developments were
any different from trends elsewhere in the market prior to the LIHTC development completion.
Qualitative Approach
The quantitative analysis described above is supplemented with interviews with LIHTC developers
active in the Los Angeles market. Using the authors’ collective network of developers, lenders,
public officials, and affordable housing advocates, a list of individuals with extensive experience
developing LIHTC properties in the region was identified. The authors specifically sought
individuals who had experience with both for-profit and nonprofit developers—either by virtue
of their work in both types of firms or through their interactions, or both, and joint ventures
on particular projects. The authors ultimately were able to schedule interviews with six separate
developers. Although it cannot be claimed that these individuals speak for all developers in the
market, the authors’ conversations with individuals throughout their various networks give them
confidence that the interviewees are generally representative of Los Angeles area LIHTC developers.
The quantitative analyses were conducted prior to interviewing the developers. This approach
gave the authors the opportunity to obtain context and some interpretation of the findings. Each
developer was asked standard questions about the six subcategorizations identified previously,
using the questionnaire included in appendix A. As a way of teasing out the extent to which
the quantitative findings merely captured preexisting price trends (and, thus, were skewed by
endogenous factors), each interviewee was specifically asked about the factors underlying different
developers’ site selection decisions. The results of those interviews were incorporated into the
discussion of the findings.
Findings
This section is divided into subparts, each of which contains our analysis of one of the specific
project- or neighborhood-level factors described previously. Within each subsection, the quantitative
findings are summarized first, then key insights from the developer interviews are incorporated.
Doing so provides a more nuanced understanding of the mechanisms underlying the observations.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
328
Refereed Papers
LIHTC Spillover Eects in Los Angeles
To create a baseline set of housing price spillover effects, the simplest difference-in-differences
model is used as a baseline (that is, equation 1). This model identifies the average effect of
all LIHTC developments on surrounding home values. The key results are shown in the
“Neighborhoods with Any LIHTC Properties” columns of exhibit 10. (The full regression results—
with all hedonic coefficients—are available in appendix B.) Here, the model does not account for
the implications of having LIHTC projects geographically concentrated and individual home sales
falling within multiple distance bands. Instead, a home sale is designated as a “pre” transaction if
it occurs before the first LIHTC project is built in the area and a “post” transaction if it occurs after
that initial project is placed in service.
Exhibit 10
Baseline Model for Low-Income Housing Tax Credit Price Effects in Los Angeles County
Variable
Distance from
LIHTC Property
Neighborhoods with Any LIHTC Properties
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Pre
0–¼ mile
– 0.037*** – 5.37
Post – 0.004 – 0.66
Effect 0.034*** 15.96
Pre
¼–½ mile
– 0.033*** – 5.68
Post – 0.003 – 0.51
Effect 0.030*** 11.23
Observations 1,842,725
R
2
0.7242
LIHTC = low-income housing tax credit.
*** p < 0.001.
Notes: Regressions control for census tract fixed effects, year fixed effects, and the following property traits—living area square footage, lot size square footage,
floor-area ratio, age at sale, number of stories, distance to central business district, seller type, and seasonal dummies (spring, summer, fall). Full results (with all
hedonic coefficients) are available in appendix B. Treatment effect is calculated manually from the differences in the regression coefficients, as the Methodology
section describes. T-statistics are used for regression coefficients, and F-statistics are used for treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Reading the exhibit from top to bottom, the first set of estimates focuses on the transactions within
a ¼ mile of the LIHTC development. The “pre” coefficient, -0.037, indicates that average home
sale prices in the LIHTC neighborhoods were 3.7 percent less than comparable transactions in
non-LIHTC neighborhoods before the LIHTC development was completed. After a given LIHTC
project was built, the “post” coefficient, -0.004, indicates that average sale prices near the LIHTC
development were only 0.4 percent less than comparable sale prices in non-LIHTC neighborhoods.
Thus, the treatment “effect” is 0.034, the difference between pre and post coefficients, indicating
that prices rose 3.4 percent more in LIHTC neighborhoods.
This positive, statistically significant effect is consistent with many of the studies cited previously—
and it is similar in magnitude to the most recent estimates, such as Diamond and McQuade
(2019) and Voith et al. (2022). If a negative, supply-driven effect exists as Eriksen and Yang
(2022) suggest, it is significantly outweighed by the positive spillover effect of the high-quality
LIHTC investment. Although it is not possible to disentangle these two competing effects, it is
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
329
Cityscape
possible to conclude that these results represent a lower bound on the positive effects of the LIHTC
investment, as it may or may not be attenuated by unobservable supply effects.
Moving farther down the table, the second set of estimates focuses on transactions within the ¼–½-
mile band, where prices increased by 3.0 percent after the LIHTC development. Again, this result
reflects the difference between the post (-0.003) and pre (-0.033) estimates, indicating that LIHTC
neighborhoods had 3.3 percent lower prices before development and only 0.3 percent lower prices
after development. In other words, once the LIHTC development was in service, the relative price
differences nearly disappeared.
In the following discussion, the regression tables have a similar format. They incorporate more
post and effect categories to document the estimates of LIHTC effects associated with each
category of factors.
Project Size
Regardless of their size, LIHTC properties in Los Angeles County have generated positive effects
on surrounding home values. In fact, the effects progressively increased with the size of the LIHTC
property, at least within the smallest distance band. Exhibit 11 presents the findings. “Small”
developments are those with 50 or fewer units, “medium” developments have between 51 and 100
units, and “large” developments have 101 or more units. Among the 833 Los Angeles properties
in the sample, 381 qualify as small, 266 qualify as medium, and 186 qualify as large. Specific post
variables are indicated for “small” and “large,” because “medium” is the reference category. In other
words, the standard “post” coefficient captures the “medium” project size, and the “small post” or
“large post” coefficient must be added to “post” to calculate the effect of small or large properties.
Exhibit 11
Property Value Effects of Different Size Low-Income Housing Tax Credit Developments (1 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Pre
0–¼ mile
– 0.037*** – 5.37
Post 0.002 0.19
Small post – 0.015 – 1.38
Large post 0.005 0.42
Small Property Effect 0.024* 5.75
Medium Property Effect 0.039*** 13.34
Large Property Effect 0.044*** 13.32
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
330
Refereed Papers
Exhibit 11
Property Value Effects of Different Size Low-Income Housing Tax Credit Developments (2 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Pre
¼–½ mile
– 0.033*** – 5.67
Post 0.005 0.80
Small post – 0.018 – 1.56
Large post – 0.003 – 0.22
Small Property Effect 0.021+ 3.23
Medium Property Effect 0.038*** 28.56
Large Property Effect 0.036** 7.34
Observations 1,842,725
R
2
0.7242
LIHTC = low-income housing tax credit.
+ p < 0.1. * p < 0.05. ** p < 0.01. *** p < 0.001.
Notes: The regression controls for census tract fixed effects, year fixed effects, and the following property traits—living area square footage, lot size square
footage, floor-area ratio, age at sale, number of stories, distance to central business district, seller type, and seasonal dummies (spring, summer, fall). Full results
(with all hedonic coefficients) are available in appendix B. Treatment effect is calculated manually from the differences in the regression coefficients, as the
Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Regression results indicate a greater LIHTC spillover effect from medium and large properties
than from small ones. Within one-fourth of a mile of the LIHTC project, the largest properties
have the highest spillovers, but in the one-fourth to one-half of a mile distance band, the medium
properties’ effect surpasses that of large ones. Within both distance bands, the difference in
spillover effects between small- and medium-sized LIHTC properties is greater than the difference
between medium and large developments. That suggests that, although larger projects generally
have a larger effect, the marginal benefit decreases—and potentially even stops or reverses—once
the project reaches a certain size. Because the analysis did not include a continuous unit number
variable, it is not possible to comment on what that threshold might be.
Importantly, no negative price effects associated with introducing larger LIHTC properties into
a neighborhood were found. This finding refutes the perception—identified in a few previous
studies—that larger properties could have deleterious neighborhood effects. That said, the
“success” of a larger project cannot be taken for granted. Several developers with whom the
authors spoke emphasized the importance of addressing community concerns about larger LIHTC
developments early in the planning process. The developers frequently encountered resistance
to larger planned projects from area residents concerned about the additional traffic and parking
difficulties that increased population density could bring. Some residents also had concerns about
increased crime and other negative stereotypes associated with “those people,” the low-income
people of color that tend to occupy many of the region’s LIHTC properties.
Alleviating the concerns often required conscious and concerted efforts on the part of the
developers to address and ameliorate local residents’ reservations. Contending that a “thoughtful
LIHTC project only enhances a neighborhood,” several interviewees described their emphasis
on extensive community programming when designing and carrying out a development. They
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
331
Cityscape
engage regularly with local residents to help them understand and (ideally) benefit from the new
housing. The developers believe that a project’s success depends in large part on the quality of its
construction and management, its architectural design, and the supportive services provided to
its tenants. Another interviewee noted that “we’ve only ever had positive impacts” for his firm’s
projects, attributing the positivity to the considerable time spent during the development process
educating nearby residents about “them” (the likely tenants) and the steps the developer takes to
help the tenants and the development integrate seamlessly within the community.
Fully Versus Partially Subsidized Developments
Both fully subsidized and partially subsidized LIHTC developments have positive and often
significant effects on surrounding property values.
6
Partially subsidized or “mixed” developments
are defined as those with six or more market-rate units. In Los Angeles, 6.72 percent of the LIHTC
developments placed in service between 1987 and 2016 meet this “mixed” criterion.
7
These
mixed developments are treated as the default “post” variable in the analysis and add another
dichotomous variable for properties with fewer than six market-rate units (that is, fully subsidized
developments). Again, the effect of fully subsidized properties on surrounding property values can
be determined by adding the coefficient of this dichotomous variable to the “post” variable. For
example, reading the coefficients from top to bottom in exhibit 12, the pre variable indicates that
property prices were 3.7 percent lower in LIHTC neighborhoods than non-LIHTC neighborhoods
before the LIHTC development was completed, the post variable indicates that they were 1.6
percent higher after a partially subsidized development, and therefore, the partially subsidized
property effect was an increase of 5.4 percent. By comparison, property prices were 2.2 percent
lower after a fully subsidized development (“fully subsidized post” coefficient), and therefore, the
fully subsidized property effect was an increase of 3.2 percent.
Exhibit 12
Property Value Effects of Partially Versus Fully Subsidized Low-Income Housing Tax
Credit Developments (1 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Pre
0–¼ mile
– 0.037*** – 5.37
Post 0.016 0.57
Fully subsidized post – 0.022 – 0.74
Partially Subsidized Property Effect 0.054+ 3.17
Fully Subsidized Property Effect 0.032*** 15.36
6
One cannot rule out the possibility that the low significance for the partially subsidized properties is a result of the limited
number of such properties.
7
As discussed in more detail to follow, the severe shortage of aordable housing in southern California contributes to
the relatively low proportion of partially subsidized properties in the Los Angeles market. Recent changes to the LIHTC
allocation process in the state have created additional incentives for developers to maximize the number of aordable units
in their properties.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
332
Refereed Papers
Exhibit 12
Property Value Effects of Partially Versus Fully Subsidized Low-Income Housing Tax
Credit Developments (2 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Pre
¼–½ mile
– 0.033*** – 5.70
Post 0.005 0.22
Primarily subsidized post – 0.008 – 0.40
Partially Subsidized Property Effect 0.038 2.24
Fully Subsidized Property Effect 0.029*** 12.82
Observations 1,842,725
R
2
0.7242
LIHTC = low-income housing tax credit.
+ p < 0.1.*** p < 0.001.
Notes: Regression controls for census tract fixed effects, year fixed effects, and the following property traits—living area square footage, lot size square footage,
floor-area ratio, age at sale, number of stories, distance to central business district, seller type, and seasonal dummies (spring, summer, fall). Full results (with all
hedonic coefficients) are available in appendix B. Treatment effect is calculated manually from the differences in the regression coefficients, as the Methodology
section describes. T-statistics are used for regression coefficients, and F-statistics are used for treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Partially subsidized developments have a larger, but not necessarily statistically significant, effect in
both distance bands, which suggests that including some market-rate units within a development
is likely to generate greater spillover effects within the surrounding neighborhood. In Los Angeles,
the competition for tax credits has led developers to move away from such “mixed” properties.
Each developer with whom the authors spoke now focuses primarily on fully subsidized properties
serving very low-income households. Several nonprofit developers always have focused on
providing housing to tenants well down the income ladder. Two of the interviewees’ current firms
have in their portfolios substantial numbers of supportive housing units affordable to tenants
earning 40 percent or less of area median income. One of the region’s larger nonprofit developers
typically serves households earning less than 50 percent of AMI in developments it has financed
with 9-percent LIHTCs, whereas primarily housing tenants with incomes closer to 60 percent of
AMI in properties financed with the shallower 4-percent credits.
One of the for-profit interviewees emphasizes that the competitiveness of the LIHTC program
drives developers’ decisions around unit affordability. His firm prefers to develop properties that
primarily serve households earning between 50 and 60 percent of AMI, because the profit margins
are much tighter when units are set aside for tenants earning closer to 30 percent of AMI (especially
if those tenants do not have vouchers to help subsidize their rents). To receive maximum points
on a tax credit application, the firm needs to commit to serving households earning as little as 30
percent of AMI.
The high costs of land and construction in the Los Angeles area make it difficult to finance LIHTC
properties containing market-rate units. Non-LIHTC public funds generally cannot be used for
market-rate units, so developing mixed property inevitably involves separating the market-rate and
affordable units into distinct condominium-like entities. It becomes necessary to attract private,
non-LIHTC-related equity to finance the market-rate properties.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
333
Cityscape
The California Housing Finance Authority operates a Mixed Income Program that helps support
properties serving renters earning between 30 and 120 percent of AMI. For all practical purposes,
the program is useful only for properties with a relatively small proportion of affordable units.
Many municipalities in the state now have inclusionary zoning ordinances that require market-
rate apartment properties to set aside at least 15 percent of their units for low-income households.
According to one interviewee, including much more than the minimum requirement subjects the
developer to financing constraints. “There’s a real sensitivity among [conventional] lenders and
investors once a project has more than 20 percent affordable units,” and that sensitivity leads to a
reluctance to commit capital.
Another interviewee explains that “the economics don’t really support a mixed-income approach.”
In Los Angeles, the costs of development exceed the rents that are affordable to low- and moderate-
income households—even for households making as much as 140 percent of AMI. As a result,
all units in a development effectively need to be subsidized to be affordable. Given the limited
amounts of public subsidy available, it makes more sense financially to maximize the number of
units that can receive LIHTC-related capital. Moreover, property owners generally are exempt from
property taxes on units designated as affordable to households making 80 percent or less of AMI;
that exemption disappears for units renting to households above the 80-percent threshold. Not
surprisingly, LIHTC developers in the region now tend to undertake partially subsidized projects
only if they are large, part of a broader development, and present an opportunity for a substantial
financial return.
For-Profit Versus Nonprofit Developer
Although both for-profit and nonprofit-sponsored LIHTC developments have positive effects on
surrounding home values, the effects of the for-profit projects appear to be greater. Within one-
fourth of a mile of a for-profit LIHTC property, the observed increase in home values is nearly twice
as large as the effect on homes near a nonprofit development (4.0 versus 2.1 percent).
8
Similarly,
for-profit properties have a greater effect on properties between one-fourth and one-half of a
mile from the LIHTC site. Exhibit 13 presents the results, with the post coefficient representing
nonprofit sponsored developments.
8
Throughout this article, the coecients are interpreted as percentages, which is the common protocol in the literature
when the outcome variable is a natural logarithm. It is possible to be slightly more precise by converting all coecients
using exponential functions, but readers often find this approach more confusing when they try to compare the exhibit with
the text.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
334
Refereed Papers
Exhibit 13
Property Value Effects of Nonprot Versus For-Prot Sponsored Low-Income Housing Tax Credit
Developments
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/F
Stat (Treatment Effects)
Pre
0–¼ mile
– 0.038*** – 5.43
Post – 0.017 – 1.35
For-prot post 0.019 1.42
Nonprot Treatment 0.021+ 2.84
For-Prot Treatment 0.040*** 15.51
Pre
¼–½ mile
– 0.033*** – 5.72
Post – 0.014 – 1.31
For-prot post 0.013 1.36
Nonprot Treatment 0.019 2.42
For-Prot Treatment 0.032*** 11.84
Observations 1,842,725
R
2
0.7242
LIHTC = low-income housing tax credit.
+ p < 0.1. *** p < 0.001.
Notes: Regression controls for census tract fixed effects, year fixed effects, and the following property traits—living area square footage, lot size square footage,
floor-area ratio, age at sale, number of stories, distance to central business district, seller type, and seasonal dummies (spring, summer, fall). Full results (with all
hedonic coefficients) are available in appendix B. Treatment effect is calculated manually from the differences in the regression coefficients, as the Methodology
section describes. T-statistics are used for regression coefficients, and F-statistics are used for treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
What accounts for the observed difference between for-profit and nonprofit-developed properties?
One of the interviewees, who has worked for both nonprofit and for-profit development firms,
contends that a “decent LIHTC deal is no different from any other multifamily property,” at least
not architecturally. He believes that a well-designed and well-managed LIHTC property should
have the same effect as any other residential development on the surrounding community.
Developers have yet different goals when building or rehabilitating a property, and those differing
motivations likely influence both the extent and type of their spillover potential.
According to the nonprofit developers with whom the authors spoke, stabilizing or revitalizing the
surrounding community often is, at best, a secondary goal for a project. Their overriding interest
lies in ensuring that cost-burdened households have an affordable and safe place to live. One
organization, for example, focuses primarily on alleviating and preventing homelessness. It looks
for sites that can support both affordable housing units and a range of ancillary human services
for its targeted very low-income population; its principal (or even sole) concern is its clientele,
not the broader neighborhood. Given space needs, its projects frequently are in less residential
neighborhoods, areas where less obvious opportunities exist for influencing single-family home
prices. Another nonprofit development organization focuses chiefly on properties that can help
alleviate the region’s affordable housing shortage. Although the organization aspires to help
facilitate community development, it realizes that many of its projects are unlikely to have much
catalytic spillover effect. “Some developments are just developments—most, in fact—while others
have more possibility for catalyzing neighborhood revitalization,” explains the firm’s president
and chief executive officer. Some properties are inherently more self-contained by virtue of their
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
335
Cityscape
location or population (those serving senior citizens or people with disabilities, for instance),
whereas others are more clearly part of a neighborhood.
In contrast, the economic considerations underlying for-profit developments appear to lead such
developers to focus more consistently on the ramifications of their properties on the local market.
A typical for-profit firm frequently looks to secure properties—especially vacant sites—whose
development can help catalyze investment in the surrounding area. Development team members
join local community crime watch groups, erect fencing around the site, hire security, and generally
work to ensure a safe environment. Post-construction, the firm imposes very strict rules on who
can live in or visit the property, employs national property management companies, and offers
extensive programming for tenants’ children. These steps help ensure that the development is well
received within the community and contributes to its overall improvement. Ideally, that positive
experience can help translate into political support for subsequent developments by the firm—
either in that community or in others nearby. For-profit LIHTC developers often are engaged in
non-LIHTC development, as well, and are routinely seeking sites for their next projects. In the
competitive real estate environment that is Los Angeles, strong community support can make the
difference in bids for desirable sites.
Neighborhood Income Level
LIHTC developments have had positive price effects across both lower- and higher-income
neighborhoods throughout Los Angeles. “Low-income” communities are defined as those census
tracts whose median household incomes were in the bottom one-third of all census tracts
throughout Los Angeles. “Medium- and high-income” tracts are those in the top two-thirds. The
incomes are based on the 2016 American Community Survey 5-year estimates. Exhibit 14 presents
the findings for both sets of neighborhoods.
Exhibit 14
Neighborhood Income Models
Variable
Distance from
LIHTC Property
All Neighborhoods Low Income Medium to High Income
Coefficient T/F Stat Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.037*** – 5.37 – 0.099*** – 6.20 – 0.032*** – 4.91
Post – 0.004 – 0.66 – 0.021 – 1.35 0.012* 2.05
Effect 0.034*** 15.96 0.078*** 18.91 0.044*** 29.08
Pre
¼–½ mile
– 0.033*** – 5.68 – 0.077*** – 5.42 – 0.030*** – 5.61
Post – 0.003 – 0.51 0.004 0.22 0.009+ 1.77
Effect 0.030*** 11.23 0.080*** 21.25 0.039*** 23.47
Observations 1,842,725 738,710 1,104,015
R
2
0.7242 0.7033 0.6237
LIHTC = low-income housing tax credit.
+ p < 0.1. * p < 0.05. *** p < 0.001.
Notes: Regressions control for census tract fixed effects, year fixed effects, and the following property traits—living area square footage, lot size square footage,
floor-area ratio, age at sale, number of stories, distance to central business district, seller type, and seasonal dummies (spring, summer, fall). Full results (with all
hedonic coefficients) are available in appendix B. Treatment effect is calculated manually from the differences in the regression coefficients, as the Methodology
section describes. T-statistics are used for regression coefficients, and F-statistics are used for treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database; U.S. Census Bureau
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
336
Refereed Papers
The effects are greater in low-income communities, where values have increased between 7.7 and
8.0 percent relative to similar neighborhoods with no LIHTC developments. These effects are
roughly twice the size of those in more affluent communities. Even in these medium- and high-
income neighborhoods, the presence of a LIHTC development increases surrounding home values
by about 4 percent. In Los Angeles, fears that LIHTC properties will depress local home values do
not conform to the data.
Neighborhood Race and Ethnicity
The regressions find little difference in the direction or size of LIHTC price effects in predominantly
White and predominantly non-White neighborhoods.
9
Drawing on the 2016 American Community
Survey data, “high non-White” tracts are defined as those whose proportion of minorities is among
the top one-third of all Los Angeles census tracts. Conversely, “low and medium non-White” tracts
are those in the bottom two-thirds of the distribution. In both types of neighborhoods, LIHTC
properties have positive effects on surrounding house prices, with the effects dissipating slightly as
the distance from the LIHTC site increases. Although the model shows slightly higher price effects
in communities with higher proportions of White residents, those differences are not statistically
significant. Exhibit 15 presents the findings.
Exhibit 15
Neighborhood Race and Ethnicity Models (1)
Variable
Distance
from LIHTC
Property
All Neighborhoods High Non-White
Low to Medium
Non-White
Coefficient T/F Stat Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.037*** – 5.37 – 0.024** – 3.21 – 0.048*** – 4.86
Post – 0.004 – 0.66 0.008 1.12 – 0.010 – 1.15
Treatment 0.034*** 15.96 0.032*** 15.56 0.038** 7.33
Pre
¼–½ mile
– 0.033*** – 5.68 – 0.026*** – 4.96 – 0.038*** – 3.85
Post – 0.003 – 0.51 0.004 0.76 – 0.005 – 0.56
Treatment 0.030*** 11.23 0.030*** 29.68 0.033* 3.91
Observations 1,842,725 500,698 1,342,027
R
2
0.7242 0.6872 0.7273
LIHTC = low-income housing tax credit.
* p < 0.05. ** p < 0.01. ** p < 0.001.
Notes: Regressions control for census tract fixed effects, year fixed effects, and the following property traits—living area square footage, lot size square footage,
floor-area ratio, age at sale, number of stories, distance to central business district, seller type, and seasonal dummies (spring, summer, fall). Full results (with all
hedonic coefficients) are available in appendix B. Treatment effect is calculated manually from the differences in the regression coefficients, as the Methodology
section describes. T-statistics are used for regression coefficients, and F-statistics are used for treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database; U.S. Census Bureau
A similar methodological approach is taken to identify any differences in price effects across
communities with high proportions of Asian, Black, and Hispanic residents. Again, census data is
used to determine the proportion of each population group within a census tract, then the model
is run with the top one-third and bottom two-thirds (by proportion) of tracts within the county.
Exhibit 16 presents these findings.
9
White is defined as those who identify as White regardless of ethnicity; non-White is everyone else.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
337
Cityscape
Exhibit 16
Neighborhood Race and Ethnicity Models (2)
Variable
Distance from
LIHTC Property
High Asian Low to Medium Asian
Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.050*** – 4.07 – 0.035*** – 4.44
Post 0.011 1.16 – 0.006 – 0.88
Treatment 0.061*** 20.36 0.028** 9.56
Pre
¼–½ mile
– 0.033*** – 4.31 – 0.034*** – 5.00
Post 0.006 0.81 – 0.005 – 0.69
Treatment 0.039*** 11.63 0.029** 7.91
Observations 584,682 1,258,043
R
2
0.7249 0.7194
Variable
Distance from
LIHTC Property
High Black Low to Medium Black
Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ Mile
– 0.032** – 3.16 – 0.049*** – 4.75
Post 0.008 0.98 – 0.002 – 0.27
Treatment 0.039*** 13.10 0.047** 10.43
Pre
¼–½ mile
– 0.042*** – 6.32 – 0.032*** – 4.03
Post 0.001 0.11 0.005 0.60
Treatment 0.043*** 18.10 0.037** 10.22
Observations 594,677 1,248,048
R
2
0.6476 0.7356
Variable
Distance from
LIHTC Property
High Hispanic Low to Medium Hispanic
Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.022* – 2.43 – 0.055*** – 5.91
Post 0.019** 2.94 – 0.007 – 0.80
Treatment 0.040*** 11.74 0.047*** 12.45
Pre
¼–½ mile
– 0.023*** – 4.55 – 0.043*** – 4.33
Post 0.017* 2.48 – 0.006 – 0.60
Treatment 0.040*** 18.26 0.037* 5.13
Observations 450,630 1,392,095
R
2
0.5930 0.7221
LIHTC = low-income housing tax credit.
* p < 0.05. ** p < 0.01. *** p < 0.001.
Notes: Regressions control for census tract fixed effects, year fixed effects, and the following property traits—living area square footage, lot size square footage,
floor-area ratio, age at sale, number of stories, distance to central business district, seller type, and seasonal dummies (spring, summer, fall). Full results (with all
hedonic coefficients) are available in appendix B. Treatment effect is calculated manually from the differences in the regression coefficients, as the Methodology
section describes. T-statistics are used for regression coefficients, and F-statistics are used for treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database; U.S. Census Bureau
The more race- and ethnicity-specific models follow the same pattern as the initial non-White
model. Regardless of race or ethnicity to classify census tracts, the model shows that LIHTC
projects have a significant positive effect on surrounding house prices within both the 0- to ¼-mile
band and the ¼- to ½-mile band. In the narrowest band where LIHTC investment is most likely
to affect houses, within one-fourth of a mile of a development, the LIHTC price effect is largest
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
338
Refereed Papers
for neighborhoods with a high proportion of Asian residents (6.1 percent) and second largest for
areas with a low to medium proportion of Black and Hispanic residents (4.7 percent in both cases).
The latter statistic is particularly important, as it contradicts the common concern that LIHTC
investment will be less advantageous to neighborhoods with predominantly White residents.
LIHTC Project Concentration
The introduction of subsequent LIHTC properties in a neighborhood tends to build on the
positive price effects associated with the initial LIHTC development. Multiple “post” variables are
incorporated for each distance band, with each such variable representing whether one, two, three,
or more LIHTC projects are present nearby when a given transaction occurs. The coefficients in
the “neighborhoods with one, two, or three LIHTC properties” columns in exhibit 17 reflect the
marginal effect of each successive LIHTC project on homes within the overlapping distance band
areas. (The “neighborhoods with any LIHTC properties” column presents the original price effect
model findings as points of reference.)
Exhibit 17
Baseline Model Versus Neighborhood Low-Income Housing Tax Credit Concentration
(Overlap) Model
Variable
Distance from
LIHTC Property
Neighborhoods with
Any LIHTC Properties
Neighborhoods with
1, 2, or 3 LIHTC Properties
Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.037*** – 5.37 – 0.035*** – 5.16
Post1 – 0.004 – 0.66 – 0.006 – 0.95
Post2 – 0.002 – 0.24
Post3 0.074** 2.70
Effect1 0.034*** 15.96 0.030*** 13.31
Effect2 0.027* 5.30
Effect3 0.101*** 12.07
Pre
¼–½ mile
– 0.033*** – 5.68 – 0.032*** – 5.64
Post1 – 0.003 – 0.51 – 0.006 – 1.12
Post2 0.001 0.07
Post3 0.049*** 4.70
Effect1 0.030*** 11.23 0.026*** 11.11
Effect2 0.027+ 3.69
Effect3 0.076*** 20.86
Observations 1,842,725 1,842,725
R
2
0.7242 0.7242
LIHTC = low-income housing tax credit.
+ p < 0.1. * p < 0.05. ** p < 0.01. *** p < 0.001.
Notes: The regressions control for census tract fixed effects, year fixed effects, and the following property traits—living area square footage, lot size square
footage, floor-area ratio, age at sale, number of stories, distance to central business district, seller type, and seasonal dummies (spring, summer, fall). Full results
(with all hedonic coefficients) are available in appendix B. Treatment effect is calculated manually from the differences in the regression coefficients, as the
Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
339
Cityscape
The coefficient for the pre variable in the ¼-mile band indicates that home prices are 3.5 percent
lower in the LIHTC neighborhoods relative to non-LIHTC neighborhoods, prior to completing any
LIHTC projects. After the first LIHTC property is placed in service (post1), average prices in the
LIHTC communities are only about 0.6 percent less than those in the non-LIHTC areas. Therefore,
the first LIHTC project leads to a 3.0-percent (with rounding) increase in home prices. Adding a
second project to the neighborhood does not significantly change the impact observed from the
first project. The addition of a third LIHTC development in the area yet significantly increases the
overall spillover price effect. It is unclear why the introduction of a third LIHTC property has a
greater (and more positive) effect than the introduction of a second such property, but these results
are consistent with the recent study of Chicago LIHTC price effects, which took a similar modeling
approach (Voith et al., 2022).
The positive price effects from adding subsequent LIHTC properties to a neighborhood apply up
to one-half of a mile from the LIHTC sites. Not surprisingly, the price effects within one-fourth of
a mile from the LIHTC sites are greater than those between one-fourth and one-half of a mile from
the properties.
10
Project Siting
Positive price effects are consistently associated with LIHTC developments in Los Angeles—
regardless of the characteristics of the properties or the surrounding neighborhoods. What remains
somewhat unclear are the underlying factors that help to bring about these improvements.
As noted previously, one possible explanation for some of the observed effects is simple
endogeneity: Developers are choosing to build or rehabilitate properties in neighborhoods where
prices already are trending upward. The model might simply be measuring baked-in effects. Each
developer downplayed the role local real estate market factors play in the selection of LIHTC
project sites. One interviewee has spent multiple decades in the affordable housing industry as an
investor, developer, and advocate. To him, the local market is “irrelevant” when considering sites
for prospective LIHTC development, saying “it doesn’t help the project in any way.” Another agrees
that “local price trends don’t come into play, because [LIHTC unit] rents are too deeply subsidized.”
Because rents in LIHTC-subsidized units are tied to AMIs, the only way a developer can realize
additional revenues from those units is through an overall increase in AMI. An increase in local
market rents has no effect on the economic returns from subsidized units. Furthermore, the rent
restrictions last for at least 15 years, well beyond the point at which current trends can predict
future rents and land values. LIHTC developments receiving certain state subsidies are subject to
California’s 55-year affordability requirements.
In theory, developers could undertake LIHTC projects with the expectation of selling their
interest after 15 years and realizing a significant capital gain from the property’s appreciated value.
Nonprofit developers frequently have no intention of ever selling their LIHTC properties; some
10
The actual price eects could be greater than those reported. It is possible that prices may begin trending upward once
plans for the development are announced, when the developer receives a formal allocation of tax credits, or when ground
is broken on the project. Thus, the actual pre-development, pre-announcement home values in the LIHTC neighborhoods
may be less than the reported average.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
340
Refereed Papers
impose ground leases that ensure the properties’ affordability for up to 99 years. Many for-profits
developers take a similar “long-term hold” approach, meeting their economic return thresholds
from developer fees and ongoing rents.
Local market conditions factor into site selection decisions for projects with a mix of subsidized
and market-rate units. Such projects tend to involve for-profit developers (either alone or in a joint
venture with a nonprofit), and those firms certainly look for properties that can command higher
rents for their market-rate units. Because of the intensity of the competition for LIHTC allocations
and related public bond financing and the reality that applicants generally receive more points in
the allocation process for promising higher proportions of affordable units, new mixed-income
affordable housing developments are relatively rare now in Los Angeles. (They were more prevalent
8 to 10 years ago, when competition for allocations was less severe.) In the current environment,
such projects tend to occur only when necessary to satisfy local zoning regulations. In effect,
a mixed-income property is “really a market-rate deal with a small amount of affordability for
political or financing reasons,” according to one interviewee.
Nonprofit developers also may deliberately target properties in gentrifying areas, such as the Boyle
Heights neighborhood and low-income communities near the University of Southern California.
They do so not for the property’s appreciation potential, but rather as a way of preserving existing
affordable housing and preventing the displacement of lower-income residents. LIHTC financing
becomes a tool to help residents afford to continue living in their communities.
If developers are not basing site decisions on local market trends, what are their primary
considerations? Four key factors influence the location of LIHTC developments in greater Los Angeles.
1. Site Availability
Although Los Angeles does not have the development density of some other markets, it has
relatively few sites available for multifamily rental properties. Some developers undertake projects
primarily in response to specific requests for proposals issued by local housing agencies. The
agency typically has control of one or more specific parcels of land and searches for the best
strategy for developing it as affordable housing. Requests for proposals respondents, therefore, have
limited, if any, flexibility in the location of their proposed development.
Other developers proactively seek properties for construction or rehabilitation, often in partnership
with a local community organization. The challenge is that many desirable properties either are not
for sale or are too costly for an affordable housing development. Few developers have the financial
luxury of waiting indefinitely for a favored site to come on line. Some for-profit firms will pay a
premium for a desirable property, but the success of that approach still depends on the willingness
of the existing landowner to work with the developer. For every property a typical firm acquires, it
analyzes between 25 and 50 potential sites.
2. Project’s Economic Feasibility
A potential site and the desired development must be both physically and economically feasible.
Each of the developers emphasized that “the deal has to pencil out financially” to be considered.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
341
Cityscape
Among the factors developers must account for are the shape and contours of the site, which help
determine the potential size of the building or buildings and the difficulty of the construction or
rehabilitation process. They also need to weigh the costs of gaining control of the property, carrying
out any necessary environmental remediation and (potentially) relocating tenants during the
construction period.
Separate from the actual development costs are zoning considerations. Local land use plans must
already allow for multifamily development or be flexible enough to incorporate such activities.
Moreover, developers need to identify any existing entitlements or other constraints that could
affect the proposed project. Not surprisingly, many potential properties are not suitable or feasible
for affordable multifamily housing.
3. Extent of Local Political Support
Perhaps the most salient factor when considering sites for LIHTC projects is the extent of local
support for the proposed development. Developers noted that they “almost always” encounter
some community opposition to a proposed affordable housing project. Developments targeting
seniors and working adults generate less concern than those designed for the homeless and families
with children, but virtually every proposed project generates some local opposition. For nonprofit
organizations developing “special needs housing,” achieving 70 percent or greater local support
represents the targeted benchmark. They consequently devote considerable time and energy to
local outreach during a project’s pre-development stage to help assuage resident concerns and
alleviate local opposition.
The support—or at least neutrality—of local public officials is critical for a project to move forward
in Los Angeles, particularly if the developer is pursuing any public funding for the project. This
principle effectively applies to all projects, because a commitment of local funding enhances
a project’s competitiveness for tax credits. As one interviewee attested, “political support is
mandatory” for a LIHTC development to be successful. “You don’t want to drive a square peg into
a round hole . . . and you’re looking for the least amount of resistance to complete a project in a
reasonable timeframe.”
Local political support is not always forthcoming, however, even in communities with an objective
need for more affordable housing. One of the nonprofit developers interviewed contends that
Angelenos generally are “very aware of the lack of affordable housing” in the city but have limited
knowledge of the steps that need to be taken to address the problem. Educating them about the
importance of taking advantage of favorable properties can be “difficult.” For an organization that
focuses primarily on housing for the homeless, the fate of its developments depends almost entirely
on the support of a local city council member. These individuals are not always supportive, and
if they are not, their colleagues will not overrule them. Various local dynamics play into council
members’ decisions. One current council member, for instance, has been supportive of affordable
housing development but has imposed an unofficial moratorium on new such developments in his
neighborhood, because he feels that it has too much affordable housing right now.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
342
Refereed Papers
4. Competitiveness for LIHTCs and Other Public Resources
In California, competition for LIHTC allocations, state affordable housing bonds, and various other
public subsidies has become hypercompetitive due to the broad acknowledgment of the imperative
to address the state’s growing homeless population and its expanding deficit of affordable
housing. The increased competition has changed the dynamics surrounding project selection and
prioritization, with developers increasingly focused on structuring planned developments in ways
that can maximize their likelihood of scoring well in the application review. With finite resources
available and a limited number of application periods, developers must be ready to make their best
presentations during those application windows.
Applicants receive additional points for projects in designated areas such as qualified census
tracts and difficult-to-develop areas—communities with high land, construction, and utility costs
relative to the median income. Allocators also look more favorably on proposed developments near
amenities such as public transit and grocery stores. Developments whose financing limits the use
of state bond proceeds to the portion of the property serving the lowest income renters also tend to
score higher.
Implications for Policy and Future Research
This analysis demonstrates the widespread, positive spillover price effects associated with LIHTC
properties in Los Angeles. It explicitly refutes the pervasive perception among certain politicians
and policymakers that such developments somehow worsen neighborhood economic conditions.
Even in predominantly White, middle- to upper-income neighborhoods, LIHTC developments
have positive effects on local home values. Moreover, residents should not be concerned about
introducing a subsequent LIHTC property in the community; the concentration of such properties
typically has an additive effect on values. This study finds that larger scale LIHTC projects and
fully subsidized developments tend to bring about greater spillover benefits to the surrounding
neighborhoods, and such positive effects are not only found in projects sponsored by nonprofit
developers, but also by for-profit developers.
From a policy perspective, the key takeaway is that LIHTC developments, in addition to creating
and preserving badly needed housing that is affordable to low-income households, consistently
have positive effects on surrounding property values. A “bad” place for such properties to be
developed does not exist, nor does a “bad” type of LIHTC development exist. Regardless of the
development’s size or neighborhood in which it is placed into service, a LIHTC property is likely to
have a positive spillover effect on its neighborhood.
Are there types of properties or types of neighborhoods that are likely to produce more positive
spillover effects than others? Perhaps, and this study identified some differences in spillover
price effects associated with some project- and neighborhood-level factors. It is important to note
that these differences, although potentially significant statistically, are not meaningfully different
economically. At most, they may reflect a percentage point or two difference. Although not
insignificant, the variation is hardly enough to spend considerable time and energy searching for
the “best” fit of development and neighborhood. After all, the property value effects ultimately are a
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
343
Cityscape
secondary benefit of the LIHTC development; the primary benefit remains the affordable housing it
supplies for low-income people.
Moreover, trying to identify the ideal project and neighborhood rarely is realistic, given the
inherent political and economic constraints developers must negotiate. In a city such as Los
Angeles, with relatively little land available for development, finding a suitable property in an area
where residents are supportive of affordable housing is its own challenge. Developers often have
to take advantage of whatever opportunities are available; they do not have the luxury of waiting
for the highest impact scenario, particularly because no guarantee is given that a development will
obtain a LIHTC allocation and other subsidies.
Some of these findings and accompanying conclusions could be specific to Los Angeles. The city’s
well-publicized problem with homelessness and its severe—and widely acknowledged—shortage
of affordable housing resulted in the passage of several public ordinances to encourage more
LIHTC and other affordable housing development. Los Angeles continues to be one of the country’s
strongest real estate markets, with many of its neighborhoods experiencing substantial home price
appreciation in the past few years. These and other factors create an environment that is conducive
to positive LIHTC spillover price effects. It is important to see if these findings can be replicated in
weaker and smaller urban markets throughout the country.
This study presented rising property values as inherently beneficial for a community. They certainly
benefit local property owners, but they simultaneously can disadvantage local renters. Rising values
typically translate into higher rents; like many other cities throughout the country, Los Angeles has
experienced double-digit average annual rent increases in the past few years. Ironically, introducing
a LIHTC property in a community could conceivably reduce the housing affordability for other
renters in the area. To date, little research—in Los Angeles or elsewhere—has been on the spillover
rental ramifications of creating affordable housing in a neighborhood.
More generally, it is important to understand the precise mechanisms that contribute to the
observed price appreciation around LIHTC properties. If site selection is not contributing
significantly to the observed changes—as this quantitative and qualitative research suggests—
then researchers need to identify the factors that are driving the change. How much is a result of
additional population density in the community—density that can shape investors’ perception
of the community’s appeal? To what extent is the improvement driven by the replacement of a
vacant or underutilized, potentially deteriorating, property into a more positive community asset?
How much of the effect results from active and capable property management? Answering these
questions provides fruitful avenues for future research.
Appendix A. Los Angeles Low-Income Housing Tax Credit
Developer Questions
1. First, how did you choose the particular location for your LIHTC development?
a. How did the LIHTC allocation criteria affect your decision?
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
344
Refereed Papers
b. How difficult was it to find available land or properties?
c. What local market dynamics affected your decision? For instance, did you consider local
crime rates? Did you focus more on areas with appreciating property values? Did you
focus on areas where community organizations were actively encouraging affordable
housing development?
d. How supportive was the community of your planned development? Did their support
or opposition affect your decision to develop the property—or the characteristics of the
development?
2. What was your targeted mix of tenant incomes in the property?
a. What was the financial and mission rationale behind that goal?
b. Were you successful in achieving the desired mix? Why or why not?
c. Since the building has been operational, how has the tenant mix changed? What has been
the rate of tenant turnover?
d. Has the turnover rate been about what you expected? What factors have you found to be
most important in attracting and keeping tenants?
3. How (if at all) has the property affected the dynamics of the surrounding neighborhood?
a. How has the community’s opinion of the project evolved since the property was placed in
service?
b. How has the neighborhood changed since you broke ground? Has it become more or less
appealing for investment?
c. Do you believe that the LIHTC property has had a significant effect on the surrounding
community? If so, what kind of effect? Why?
d. Have you contemplated or undertaken subsequent LIHTC developments in this
neighborhood? If so, are you considering other or different factors now than you did prior
to the first LIHTC investment in the area?
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
345
Cityscape
Appendix B. Full Regression Results
Exhibit B-1
Baseline Model for Low-Income Housing Tax Credit Price Effects in Los Angeles County (1 of 2)
Variable
Distance from
LIHTC Property
Neighborhoods with Any LIHTC Properties
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Pre
0–¼ mile
– 0.037*** – 5.37
Post – 0.004 – 0.66
Effect 0.034*** 15.96
Pre
¼–½ mile
– 0.033*** – 5.68
Post – 0.003 – 0.51
Effect 0.030*** 11.23
Lot size 0.000*** 4.04
Lot size
2
– 0.000*** – 4.40
Living area 0.000*** 35.86
Living area
2
– 0.000*** – 24.06
Floor-area ratio – 0.004 – 1.35
Age – 0.001 – 0.54
Age
2
0.000 0.60
2 Stories 0.019** 2.84
3 Stories 0.067** 3.17
Spring 0.004 1.63
Summer 0.037*** 14.88
Fall 0.045*** 19.79
Distance to central business district – 0.010 – 1.06
Government seller – 0.176*** – 4.21
Bank seller – 0.141*** – 9.95
1989 0.168*** 21.45
1990 0.193*** 28.68
1991 0.202*** 17.39
1992 0.208*** 7.66
1993 0.124*** 8.42
1994 0.103*** 5.49
1995 0.039*** 4.05
1996 0.041*** 4.47
1997 0.078*** 7.45
1998 0.174*** 19.86
1999 0.270*** 28.36
2000 0.347*** 37.47
2001 0.446*** 46.47
2002 0.585*** 59.70
2003 0.764*** 72.21
2004 1.009*** 88.21
2005 1.198*** 100.99
2006 1.267*** 96.35
2007 1.012*** 84.16
2008 0.801*** 38.30
2009 0.801*** 19.29
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
346
Refereed Papers
Exhibit B-1
Baseline Model for Low-Income Housing Tax Credit Price Effects in Los Angeles County (2 of 2)
Variable
Distance from
LIHTC Property
Neighborhoods with Any LIHTC Properties
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
2010 0.846*** 26.53
2011 0.811*** 24.84
2012 0.850*** 26.20
2013 1.017*** 35.33
2014 1.135*** 40.76
2015 1.217*** 40.76
2016 1.291*** 45.13
Constant 11.439*** 71.77
Observations 1,842,725
R
2
0.7242
LIHTC = low-income housing tax credit.
** p < 0.01. *** p < 0.001.
Notes: Regressions also control for census tract fixed effects, which are not listed due to the large number of tracts. Treatment effect is calculated manually from
the differences in the regression coefficients, as the Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for
treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Exhibit B-2
Property Value Effects of Different Size Low-Income Housing Tax Credit Developments (1 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Pre
0–¼ mile
– 0.037*** – 5.37
Post 0.002 0.19
Small post – 0.015 – 1.38
Large post 0.005 0.42
Small Property Effect 0.024* 5.75
Medium Property Effect 0.039*** 13.34
Large Property Effect 0.044*** 13.32
Pre
¼–½ mile
– 0.033*** – 5.67
Post 0.005 0.80
Small post – 0.018 – 1.56
Large post – 0.003 – 0.22
Small Property Effect 0.021+ 3.23
Medium Property Effect 0.038*** 28.56
Large Property Effect 0.036** 7.34
Lot size 0.000*** 4.04
Lot size
2
– 0.000*** – 4.40
Living area 0.000*** 35.86
Living area
2
– 0.000*** – 24.06
Floor-area ratio – 0.004 – 1.35
Age – 0.001 – 0.54
Age
2
0.000 0.60
2 Stories 0.019** 2.84
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
347
Cityscape
Exhibit B-2
Property Value Effects of Different Size Low-Income Housing Tax Credit Developments (2 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
3 Stories 0.067** 3.17
Spring 0.004 1.63
Summer 0.037*** 14.88
Fall 0.045*** 19.79
Distance to central business district – 0.010 – 1.06
Government seller – 0.176*** – 4.21
Bank seller – 0.141*** – 9.95
1989 0.168*** 21.45
1990 0.193*** 28.68
1991 0.202*** 17.38
1992 0.208*** 7.66
1993 0.124*** 8.42
1994 0.103*** 5.49
1995 0.039*** 4.06
1996 0.041*** 4.47
1997 0.078*** 7.45
1998 0.174*** 19.85
1999 0.270*** 28.36
2000 0.347*** 37.46
2001 0.446*** 46.48
2002 0.585*** 59.69
2003 0.764*** 72.22
2004 1.009*** 88.23
2005 1.198*** 100.96
2006 1.267*** 96.33
2007 1.012*** 84.16
2008 0.801*** 38.30
2009 0.801*** 19.28
2010 0.846*** 26.53
2011 0.811*** 24.82
2012 0.850*** 26.19
2013 1.017*** 35.35
2014 1.135*** 40.78
2015 1.217*** 40.76
2016 1.291*** 45.13
Constant 11.439*** 71.86
Observations 1,842,725
R
2
0.7242
LIHTC = low-income housing tax credit.
+ p < 0.1. * p < 0.05. ** p < 0.01. *** p < 0.001.
Notes: Regressions also control for census tract fixed effects, which are not listed due to the large number of tracts. Treatment effect is calculated manually from
the differences in the regression coefficients, as the Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for
treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
348
Refereed Papers
Exhibit B-3
Property Value Effects of Partially Versus Fully Subsidized Low-Income Housing Tax
Credit Developments (1 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Pre
0–¼ mile
– 0.037*** – 5.37
Post 0.016 0.57
Fully subsidized post – 0.022 – 0.74
Partially Subsidized Property Effect 0.054+ 3.17
Fully Subsidized Property Effect 0.032*** 15.36
Pre
¼–½ mile
– 0.033*** – 5.70
Post 0.005 0.22
Primarily subsidized post – 0.008 – 0.40
Partially Subsidized Property Effect 0.038 2.24
Fully Subsidized Property Effect 0.029*** 12.82
Lot size 0.000*** 4.04
Lot size
2
– 0.000*** – 4.40
Living area 0.000*** 35.86
Living area
2
– 0.000*** – 24.06
Floor-area ratio – 0.004 – 1.35
Age – 0.001 – 0.54
Age
2
0.000 0.60
2 Stories 0.019** 2.85
3 Stories 0.067** 3.17
Spring 0.004 1.63
Summer 0.037*** 14.88
Fall 0.045*** 19.79
Distance to central business district – 0.010 – 1.05
Government seller – 0.176*** – 4.21
Bank seller – 0.141*** – 9.95
1989 0.168*** 21.45
1990 0.193*** 28.67
1991 0.202*** 17.37
1992 0.208*** 7.66
1993 0.124*** 8.42
1994 0.103*** 5.49
1995 0.039*** 4.05
1996 0.041*** 4.46
1997 0.078*** 7.44
1998 0.174*** 19.84
1999 0.270*** 28.36
2000 0.347*** 37.47
2001 0.446*** 46.46
2002 0.585*** 59.67
2003 0.764*** 72.18
2004 1.009*** 88.16
2005 1.198*** 100.95
2006 1.267*** 96.29
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
349
Cityscape
Exhibit B-3
Property Value Effects of Partially Versus Fully Subsidized Low-Income Housing Tax
Credit Developments (2 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
2007 1.012*** 84.13
2008 0.801*** 38.30
2009 0.801*** 19.29
2010 0.846*** 26.53
2011 0.811*** 24.83
2012 0.850*** 26.20
2013 1.017*** 35.33
2014 1.135*** 40.77
2015 1.217*** 40.77
2016 1.291*** 45.15
Constant 11.439*** 71.74
Observations 1,842,725
R
2
0.7242
LIHTC = low-income housing tax credit.
+ p < 0.1. ** p < 0.01. *** p < 0.001.
Notes: Regressions also control for census tract fixed effects, which are not listed due to the large number of tracts. Treatment effect is calculated manually from
the differences in the regression coefficients, as the Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for
treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Exhibit B-4
Property Value Effects of Nonprot Versus For-Prot Sponsored Low-Income Housing Tax
Credit Developments (1 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Pre
0–¼ mile
– 0.038*** – 5.43
Post – 0.017 – 1.35
For-prot post 0.019 1.42
Nonprofit Treatment 0.021+ 2.84
For-Profit Treatment 0.040*** 15.51
Pre
¼–½ mile
– 0.033*** – 5.72
Post – 0.014 – 1.31
For-prot post 0.013 1.36
Nonprofit Treatment 0.019 2.42
For-Profit Treatment 0.032*** 11.84
Lot size 0.000*** 4.04
Lot size
2
– 0.000*** – 4.40
Living area 0.000*** 35.85
Living area
2
– 0.000*** – 24.06
Floor-area ratio – 0.004 – 1.35
Age – 0.001 – 0.55
Age
2
0.000 0.61
2 Stories 0.019** 2.84
3 Stories 0.067** 3.17
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
350
Refereed Papers
Exhibit B-4
Property Value Effects of Nonprot Versus For-Prot Sponsored Low-Income Housing Tax
Credit Developments (2 of 2)
Variable
Distance from
LIHTC Property
Coefficient
T Stat (Coefficients)/
F Stat (Treatment Effects)
Spring 0.004 1.63
Summer 0.037*** 14.91
Fall 0.045*** 19.80
Distance to central business district – 0.010 – 1.06
Government seller – 0.176*** – 4.21
Bank seller – 0.141*** – 9.95
1989 0.168*** 21.46
1990 0.193*** 28.71
1991 0.202*** 17.40
1992 0.208*** 7.67
1993 0.124*** 8.42
1994 0.103*** 5.49
1995 0.039*** 4.06
1996 0.041*** 4.48
1997 0.078*** 7.45
1998 0.174*** 19.82
1999 0.270*** 28.31
2000 0.347*** 37.33
2001 0.446*** 46.36
2002 0.585*** 59.54
2003 0.764*** 72.12
2004 1.009*** 88.16
2005 1.198*** 100.93
2006 1.267*** 96.25
2007 1.012*** 84.11
2008 0.801*** 38.31
2009 0.801*** 19.29
2010 0.846*** 26.54
2011 0.811*** 24.83
2012 0.850*** 26.20
2013 1.017*** 35.33
2014 1.135*** 40.79
2015 1.217*** 40.81
2016 1.291*** 45.19
Constant 11.439*** 71.69
Observations 1,842,725
R
2
0.7242
LIHTC = low-income housing tax credit.
+ p < 0.1. ** p < 0.01. *** p < 0.001.
Notes: Regressions also control for census tract fixed effects, which are not listed due to the large number of tracts. Treatment effect is calculated manually from
the differences in the regression coefficients, as the Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for
treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
351
Cityscape
Exhibit B-5
Neighborhood Income Models (1 of 2)
Variable
Distance
from LIHTC
Property
All Neighborhoods Low Income Medium to High Income
Coefficient T/F Stat Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.037*** – 5.37 – 0.099*** – 6.20 – 0.032*** – 4.91
Post – 0.004 – 0.66 – 0.021 – 1.35 0.012* 2.05
Effect 0.034*** 15.96 0.078*** 18.91 0.044*** 29.08
Pre
¼–½ mile
– 0.033*** – 5.68 – 0.077*** – 5.42 – 0.030*** – 5.61
Post – 0.003 – 0.51 0.004 0.22 0.009+ 1.77
Effect 0.030*** 11.23 0.080*** 21.25 0.039*** 23.47
Lot size 0.000*** 4.04 0.000*** 4.82 – 0.000 – 0.36
Lot size
2
– 0.000*** – 4.40 – 0.000*** – 4.63 0.000 0.09
Living area 0.000*** 35.86 0.000*** 26.27 0.000*** 28.95
Living area
2
– 0.000*** – 24.06 – 0.000*** – 19.10 – 0.000*** – 17.65
Floor-area ratio – 0.004 – 1.35 – 0.002 – 1.27 – 0.356*** – 10.35
Age – 0.001 – 0.54 0.002 1.51 – 0.004 – 1.40
Age
2
0.000 0.60 – 0.000 – 0.66 0.000 1.16
2 Stories 0.019** 2.84 0.004 0.56 0.029*** 3.56
3 Stories 0.067** 3.17 0.052* 2.30 0.102+ 1.86
Spring 0.004 1.63 0.015*** 4.35 – 0.002 – 0.62
Summer 0.037*** 14.88 0.047*** 13.01 0.029*** 10.00
Fall 0.045*** 19.79 0.051*** 15.07 0.041*** 10.80
Distance to central
business district
– 0.010 – 1.06 – 0.006 – 0.51 – 0.013 – 1.04
Government seller – 0.176*** – 4.21 – 0.317** – 2.93 – 0.128** – 2.94
Bank seller – 0.141*** – 9.95 – 0.123** – 3.07 – 0.126*** – 18.49
1989 0.168*** 21.45 0.179*** 18.20 0.160*** 22.52
1990 0.193*** 28.68 0.176*** 17.87 0.202*** 29.48
1991 0.202*** 17.39 0.178*** 11.86 0.220*** 15.49
1992 0.208*** 7.66 0.146*** 9.82 0.251*** 7.22
1993 0.124*** 8.42 0.072*** 5.08 0.165*** 8.50
1994 0.103*** 5.49 0.054*** 3.84 0.144*** 5.15
1995 0.039***
4.05 0.008 0.79 0.064*** 5.05
1996 0.041*** 4.47 0.022* 1.99 0.059*** 4.82
1997 0.078*** 7.45 0.087*** 6.81 0.073*** 5.03
1998 0.174*** 19.86 0.201*** 16.68 0.151*** 13.90
1999 0.270*** 28.36 0.298*** 21.01 0.249*** 23.95
2000 0.347*** 37.47 0.381*** 25.65 0.324*** 32.82
2001 0.446*** 46.47 0.464*** 31.37 0.435*** 37.03
2002 0.585*** 59.70 0.586*** 40.22 0.585*** 49.01
2003 0.764*** 72.21 0.749*** 51.41 0.774*** 57.59
2004 1.009*** 88.21 0.965*** 62.05 1.036*** 73.62
2005 1.198*** 100.99 1.135*** 69.04 1.242*** 91.63
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
352
Refereed Papers
Exhibit B-5
Neighborhood Income Models (2 of 2)
Variable
Distance
from LIHTC
Property
All Neighborhoods Low Income Medium to High Income
Coefficient T/F Stat Coefficient T/F Stat Coefficient T/F Stat
2006 1.267*** 96.35 1.205*** 72.48 1.320*** 79.23
2007 1.012*** 84.16 1.204*** 71.51 1.338*** 60.39
2008 0.801*** 38.30 1.056*** 51.46 0.985*** 27.57
2009 0.801*** 19.29 0.942*** 48.34 0.736*** 14.08
2010 0.846*** 26.53 0.944*** 48.29 0.795*** 19.25
2011 0.811*** 24.84 0.902*** 39.93 0.761*** 18.01
2012 0.850*** 26.20 0.936*** 38.18 0.800*** 18.68
2013 1.017*** 35.33 1.078*** 44.68 0.975*** 25.51
2014 1.135*** 40.76 1.186*** 45.80 1.100*** 30.16
2015 1.217*** 40.76 1.258*** 44.97 1.187*** 30.19
2016 1.291*** 45.13 1.316*** 47.34 1.273*** 33.79
Constant 11.439*** 71.77 11.425*** 62.24 11.649*** 54.78
Observations 1,842,725 738,710 1,104,015
R
2
0.7242 0.7033 0.6237
LIHTC = low-income housing tax credit.
+ p < 0.1. * p < 0.05. ** p < 0.01. *** p < 0.001.
Notes: Regressions also control for census tract fixed effects, which are not listed due to the large number of tracts. Treatment effect is calculated manually from
the differences in the regression coefficients, as the Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for
treatment effects.
Sources: DataQuick Information Systems and CoreLogic; LIHTC HUD User database; U.S. Census Bureau
Exhibit B-6
Neighborhood Race and Ethnicity Models (1) (1 of 2)
Variable
Distance
from LIHTC
Property
All Neighborhoods High Non-White
Low to Medium
Non-White
Coefficient T/F Stat Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.037*** – 5.37 – 0.024** – 3.21 – 0.048*** – 4.86
Post – 0.004 – 0.66 0.008 1.12 – 0.010 – 1.15
Treatment 0.034*** 15.96 0.032*** 15.56 0.038** 7.33
Pre
¼–½ mile
– 0.033*** – 5.68 – 0.026*** – 4.96 – 0.038*** – 3.85
Post – 0.003 – 0.51 0.004 0.76 – 0.005 – 0.56
Treatment 0.030*** 11.23 0.030*** 29.68 0.033* 3.91
Lot size 0.000*** 4.04 0.000*** 3.04 0.000*** 3.44
Lot size
2
– 0.000*** – 4.40 – 0.000* – 2.38 – 0.000*** – 3.80
Living area 0.000*** 35.86 0.000*** 31.61 0.000*** 29.85
Living area
2
– 0.000*** – 24.06 – 0.000*** – 18.00 – 0.000*** – 20.85
Floor-area ratio – 0.004 – 1.35 – 0.206** – 3.19 – 0.003 – 1.33
Age – 0.001 – 0.54 – 0.001 – 0.48 – 0.001 – 0.53
Age
2
0.000 0.60 0.000 0.36 0.000 0.65
2 Stories 0.019** 2.84 0.049*** 6.43 0.010 1.16
3 Stories 0.067** 3.17 0.048 0.87 0.062** 2.78
Spring 0.004 1.63 – 0.009 – 1.31 0.010*** 3.98
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
353
Cityscape
Exhibit B-6
Neighborhood Race and Ethnicity Models (1) (2 of 2)
Variable
Distance
from LIHTC
Property
All Neighborhoods High Non-White
Low to Medium
Non-White
Coefficient T/F Stat Coefficient T/F Stat Coefficient T/F Stat
Summer 0.037*** 14.88 0.025*** 3.70 0.041*** 16.37
Fall 0.045*** 19.79 0.036*** 4.20 0.049*** 23.84
Distance to central
business district
– 0.010 – 1.06 0.021 1.25 – 0.014 – 1.39
Government seller – 0.176*** – 4.21 – 0.054 – 0.58 – 0.234*** – 4.88
Bank seller – 0.141*** – 9.95 – 0.141*** – 9.27 – 0.141*** – 7.88
1989 0.168*** 21.45 0.173*** 8.86 0.166*** 24.13
1990 0.193*** 28.68 0.188*** 20.93 0.195*** 23.56
1991 0.202*** 17.39 0.206*** 17.94 0.201*** 12.88
1992 0.208*** 7.66 0.198*** 13.92 0.212*** 5.95
1993 0.124*** 8.42 0.149*** 9.91 0.115*** 6.16
1994 0.103*** 5.49 0.138*** 5.16 0.091*** 5.10
1995 0.039*** 4.05 0.067*** 5.18 0.029** 2.63
1996 0.041*** 4.47 0.063*** 4.24 0.034*** 3.44
1997 0.078*** 7.45 0.087*** 4.16 0.075*** 7.36
1998 0.174*** 19.86 0.164*** 13.37 0.178*** 17.76
1999 0.270*** 28.36 0.240*** 15.43 0.281*** 25.76
2000 0.347*** 37.47 0.326*** 27.68 0.355*** 32.96
2001 0.446*** 46.47 0.430*** 31.06 0.453*** 43.15
2002 0.585*** 59.70 0.569*** 44.58 0.592*** 53.60
2003 0.764*** 72.21 0.751*** 54.70 0.770*** 65.36
2004 1.009*** 88.21 1.016*** 67.68 1.007*** 76.51
2005 1.198*** 100.99 1.226*** 67.86 1.188*** 86.60
2006 1.267*** 96.35 1.299*** 56.02 1.257*** 90.50
2007 1.012*** 84.16 1.324*** 50.12 1.268*** 80.84
2008 0.801*** 38.30 1.013*** 25.62 1.013*** 39.27
2009 0.801*** 19.29 0.778*** 12.15 0.812*** 19.36
2010 0.846*** 26.53 0.831*** 17.42 0.854*** 26.62
2011 0.811*** 24.84 0.810*** 16.57 0.812*** 24.66
2012 0.850*** 26.20 0.853*** 17.37 0.851*** 26.21
2013 1.017*** 35.33 1.021*** 23.45 1.017*** 35.59
2014 1.135*** 40.76 1.145*** 27.03 1.133*** 40.29
2015 1.217*** 40.76 1.237*** 28.09 1.211*** 40.29
2016 1.291*** 45.13 1.317*** 31.51 1.283*** 44.69
Constant 11.439*** 71.77 11.428*** 25.64 11.496*** 65.75
Observations 1,842,725 500,698 1,342,027
R
2
0.7242 0.6872 0.7273
LIHTC = low-income housing tax credit.
* p < 0.05. ** p < 0.01. *** p < 0.001.
Notes: Regressions also control for census tract fixed effects, which are not listed due to the large number of tracts. Treatment effect is calculated manually from
the differences in the regression coefficients, as the Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for
treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database; U.S. Census Bureau
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
354
Refereed Papers
Exhibit B-7
Neighborhood Race and Ethnicity Models (2) (1 of 4)
Variable
Distance
from LIHTC
Property
High Asian Low to Medium Asian
Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.050*** – 4.07 – 0.035*** – 4.44
Post 0.011 1.16 – 0.006 – 0.88
Treatment 0.061*** 20.36 0.028** 9.56
Pre
¼–½ mile
– 0.033*** – 4.31 – 0.034*** – 5.00
Post 0.006 0.81 – 0.005 – 0.69
Treatment 0.039*** 11.63 0.029** 7.91
Lot size 0.000*** 7.67 0.000* 2.49
Lot size
2
– 0.000*** – 7.00 – 0.000** – 2.86
Living area 0.000*** 27.40 0.000*** 30.76
Living area
2
– 0.000*** – 17.92 – 0.000*** – 19.76
Floor-area ratio – 0.001 – 0.81 – 0.006 – 0.97
Age 0.000 0.14 – 0.002 – 0.60
Age
2
– 0.000 – 0.84 0.000 0.705
2 Stories 0.031*** 4.53 0.014 1.59
3 Stories 0.048 1.44 0.071* 2.54
Spring 0.010*** 3.84 0.002 0.65
Summer 0.042*** 16.13 0.034*** 10.49
Fall 0.050*** 22.11 0.043*** 13.72
Distance to CBD 0.022* 2.54 – 0.017 – 1.61
Government seller – 0.147 – 1.32 – 0.175*** – 3.83
Bank seller – 0.106*** – 15.11 – 0.144*** – 9.56
1989 0.178*** 18.87 0.164*** 18.17
1990 0.189*** 18.70 0.197*** 24.75
1991 0.170*** 17.58 0.219*** 14.17
1992 0.133*** 10.45 0.245*** 6.78
1993 0.068*** 5.34 0.153*** 7.98
1994 0.055*** 3.54 0.127*** 5.11
1995 0.001 0.07 0.057*** 4.58
1996 0.012 0.94 0.057*** 4.68
1997 0.051*** 3.50 0.092*** 6.67
1998 0.150*** 14.44 0.186*** 15.42
1999 0.231*** 19.48 0.290*** 22.98
2000 0.316*** 29.34 0.363*** 27.72
2001 0.418*** 37.97 0.461*** 33.39
2002 0.555*** 49.28 0.600*** 43.29
2003 0.735*** 61.72 0.779*** 52.84
2004 0.969*** 70.78 1.027*** 68.71
2005 1.145*** 73.37 1.222*** 81.35
2006 1.203*** 77.87 1.298*** 74.19
2007 1.214*** 78.68 1.313*** 58.69
2008 1.017*** 79.89 1.011*** 26.17
2009 0.903*** 65.34 0.763*** 14.10
2010 0.919*** 68.98 0.819*** 18.93
2011 0.871*** 60.49 0.789*** 17.55
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
355
Cityscape
Exhibit B-7
Neighborhood Race and Ethnicity Models (2) (2 of 4)
Variable
Distance
from LIHTC
Property
High Asian Low to Medium Asian
Coefficient T/F Stat Coefficient T/F Stat
2012 0.905*** 57.87 0.829*** 18.37
2013 1.055*** 64.63 1.001*** 24.76
2014 1.165*** 64.53 1.123*** 28.77
2015 1.234*** 66.48 1.211*** 28.85
2016 1.293*** 70.44 1.292*** 31.82
Constant 10.952*** 82.04 11.548*** 60.85
Observations 584,682 1,258,043
R
2
0.7249 0.7194
Variable
Distance
from LIHTC
Property
High Black Low to Medium Black
Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.032** – 3.16 – 0.049*** – 4.75
Post 0.008 0.98 – 0.002 – 0.27
Treatment 0.039*** 13.10 0.047** 10.43
Pre
¼–½ mile
– 0.042*** – 6.32 – 0.032*** – 4.03
Post 0.001 0.11 0.005 0.60
Treatment 0.043*** 18.10 0.037** 10.22
Lot size – 0.000 – 1.19 0.000*** 6.54
Lot size
2
0.000 1.17 – 0.000*** 6.64
Living area 0.000*** 22.66 0.000*** 36.78
Living area
2
– 0.000*** – 12.30 – 0.000*** – 24.06
Floor-area ratio – 0.314*** – 4.88 – 0.002 – 1.29
Age – 0.007+ – 1.69 0.002* 2.21
Age
2
0.000+ 1.69 – 0.000* – 2.31
2 Stories 0.018 1.36 0.018*** 3.41
3 Stories 0.197** 2.67 0.057** 2.68
Spring – 0.006 – 0.88 0.010*** 4.62
Summer 0.026*** 4.48 0.042*** 18.25
Fall 0.039*** 5.78 0.048*** 22.77
Distance to CBD – 0.023+ – 1.69 – 0.003 – 0.28
Government seller – 0.104* – 2.19 – 0.208** – 2.90
Bank seller – 0.143*** – 17.75 – 0.113*** – 5.84
1989 0.154*** 13.88 0.172*** 30.50
1990 0.184*** 21.93 0.196*** 26.67
1991 0.233*** 9.46 0.188*** 19.04
1992 0.300*** 5.23 0.164*** 16.07
1993 0.193*** 5.83 0.093*** 9.42
1994 0.162*** 3.34 0.077*** 6.96
1995 0.053** 2.75 0.031*** 3.30
1996 0.047* 2.41 0.039*** 4.26
1997 0.077** 3.24 0.078*** 7.97
1998 0.166*** 9.57 0.176*** 19.55
1999 0.261*** 17.48 0.273*** 25.47
2000 0.338*** 20.09 0.351*** 34.61
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
356
Refereed Papers
Exhibit B-7
Neighborhood Race and Ethnicity Models (2) (3 of 4)
Variable
Distance
from LIHTC
Property
High Black Low to Medium Black
Coefficient T/F Stat Coefficient T/F Stat
2001 0.438*** 20.37 0.448*** 46.00
2002 0.585*** 28.26 0.583*** 58.18
2003 0.770*** 33.15 0.759*** 75.53
2004 1.025*** 43.69 0.996*** 89.97
2005 1.233*** 50.46 1.175*** 96.74
2006 1.308*** 44.63 1.243*** 95.68
2007 1.322*** 33.05 1.256*** 92.46
2008 0.949*** 15.39 1.041*** 74.19
2009 0.654*** 8.62 0.890*** 61.10
2010 0.731*** 11.38 0.910*** 63.92
2011 0.705*** 9.98 0.866*** 55.86
2012 0.741*** 10.46 0.905*** 53.48
2013 0.923*** 15.02 1.061*** 62.01
2014 1.048*** 18.19 1.177*** 64.78
2015 1.141*** 18.30 1.253*** 64.98
2016 1.234*** 20.83 1.318*** 67.33
Constant 11.812*** 36.64 11.271*** 72.44
Observations 594,677 1,248,048
R
2
0.6476 0.7356
Variable
Distance
from LIHTC
Property
High Hispanic Low to Medium Hispanic
Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.022* – 2.43 – 0.055*** – 5.91
Post 0.019** 2.94 – 0.007 – 0.80
Treatment 0.040*** 11.74 0.047*** 12.45
Pre
¼–½ mile
– 0.023*** – 4.55 – 0.043*** – 4.33
Post 0.017* 2.48 – 0.006 – 0.60
Treatment 0.040*** 18.26 0.037* 5.13
Lot size 0.000* 2.19 0.000*** 3.69
Lot size
2
– 0.000* – 2.00 – 0.000*** – 4.03
Living area 0.000*** 31.05 0.000*** 28.99
Living area
2
– 0.000*** – 18.11 – 0.000 – 21.06
Floor-area ratio – 0.282*** – 9.31 – 0.003 – 1.33
Age – 0.001 – 0.25 – 0.001 – 0.78
Age
2
0.000 0.15 0.000 0.94
2 Stories 0.043*** 5.28 0.012 1.34
3 Stories 0.113 0.91 0.061** 2.80
Spring – 0.015* – 2.30 0.011*** 4.67
Summer 0.018* 2.52 0.043*** 17.99
Fall 0.032*** 5.16 0.050*** 27.21
Distance to CBD – 0.010 – 0.50 – 0.009 – 0.94
Government seller – 0.106 – 1.45 – 0.228*** – 4.90
Bank seller – 0.119*** – 11.37 – 0.143*** – 7.39
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
357
Cityscape
Exhibit B-7
Neighborhood Race and Ethnicity Models (2) (4 of 4)
Variable
Distance
from LIHTC
Property
High Hispanic Low to Medium Hispanic
Coefficient T/F Stat Coefficient T/F Stat
1989 0.168*** 9.88 0.168*** 22.38
1990 0.219*** 25.21 0.183*** 23.43
1991 0.238*** 23.32 0.190*** 12.79
1992 0.269*** 8.67 0.188*** 6.27
1993 0.180*** 15.95 0.107*** 5.70
1994 0.182*** 8.49 0.080*** 4.33
1995 0.114*** 6.07 0.016+ 1.70
1996 0.098*** 5.26 0.025** 2.72
1997 0.103*** 4.87 0.071*** 6.88
1998 0.170*** 12.78 0.174*** 18.46
1999 0.253*** 13.61 0.274*** 24.33
2000 0.327*** 24.72 0.352*** 34.52
2001 0.435*** 25.76 0.449*** 44.72
2002 0.581*** 32.76 0.585*** 56.80
2003 0.767*** 39.74 0.763*** 70.08
2004 1.044*** 51.96 0.995*** 79.87
2005 1.277*** 69.63 1.172*** 86.28
2006 1.376*** 51.35 1.236*** 91.58
2007 1.397*** 37.55 1.246*** 87.65
2008 0.991*** 19.42 1.017*** 43.60
2009 0.713*** 10.54 0.838*** 19.87
2010 0.780*** 14.31 0.873*** 28.11
2011 0.760*** 13.42 0.830*** 26.12
2012 0.800*** 13.54 0.867*** 28.39
2013 0.963*** 17.78 1.031*** 38.73
2014 1.101*** 20.80 1.143*** 43.51
2015 1.197*** 21.11 1.219*** 43.55
2016 1.287*** 23.34 1.289*** 48.65
Constant 11.474*** 32.36 11.434*** 65.86
Observations 450,630 1,392,095
R
2
0.5930 0.7221
CBD = central business district. LIHTC = low-income housing tax credit.
+ p < 0.1. * p < 0.05. ** p < 0.01. *** p < 0.001.
Notes: Regressions also control for census tract fixed effects, which are not listed due to the large number of tracts. Treatment effect is calculated manually from
the differences in the regression coefficients, as the Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for
treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database; U.S. Census Bureau
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
358
Refereed Papers
Exhibit B-8
Baseline Model Versus Neighborhood Low-Income Housing Tax Credit Concentration
(Overlap) Model (1 of 2)
Variable
Distance
from LIHTC
Property
Neighborhoods with
Any LIHTC Properties
Neighborhoods with
1, 2, or 3 LIHTC Properties
Coefficient T/F Stat Coefficient T/F Stat
Pre
0–¼ mile
– 0.037*** – 5.37 – 0.035*** – 5.16
Post1 – 0.004 – 0.66 – 0.006 – 0.95
Post2 – 0.002 – 0.24
Post3 0.074** 2.70
Effect1 0.034*** 15.96 0.030*** 13.31
Effect2 0.027* 5.30
Effect3 0.101*** 12.07
Pre
¼–½ mile
– 0.033*** – 5.68 – 0.032*** – 5.64
Post1 – 0.003 – 0.51 – 0.006 – 1.12
Post2 0.001 0.07
Post3 0.049*** 4.70
Effect1 0.030*** 11.23 0.026*** 11.11
Effect2 0.027+ 3.69
Effect3 0.076*** 20.86
Lot size 0.000*** 4.04 0.000*** 4.04
Lot size
2
– 0.000*** – 4.40 – 0.000*** – 4.40
Living area 0.000*** 35.86 0.000*** 35.87
Living area
2
– 0.000*** – 24.06 – 0.000*** – 24.06
Floor-area ratio – 0.004 – 1.35 – 0.004 – 1.35
Age – 0.001 – 0.54 – 0.001 – 0.53
Age
2
0.000 0.60 0.000 0.58
2 Stories 0.019** 2.84 0.019** 2.85
3 Stories 0.067** 3.17 0.067** 3.16
Spring 0.004 1.63 0.004 1.63
Summer 0.037*** 14.88 0.037*** 14.88
Fall 0.045*** 19.79 0.045*** 19.79
Distance to central
business district
– 0.010 – 1.06 – 0.009 – 1.05
Government seller – 0.176*** – 4.21 – 0.177*** – 4.21
Bank seller – 0.141*** – 9.95 – 0.141*** – 9.95
1989 0.168*** 21.45 0.168*** 21.45
1990 0.193*** 28.68 0.193*** 28.69
1991 0.202*** 17.39 0.202*** 17.39
1992 0.208*** 7.66 0.208*** 7.66
1993 0.124*** 8.42 0.124*** 8.42
1994 0.103*** 5.49 0.103*** 5.49
1995 0.039*** 4.05 0.039*** 4.06
1996 0.041*** 4.47 0.041*** 4.48
1997 0.078*** 7.45 0.078*** 7.48
1998 0.174*** 19.86 0.174*** 19.90
1999 0.270*** 28.36 0.270*** 28.39
2000 0.347*** 37.47 0.347*** 37.51
2001 0.446*** 46.47 0.446*** 46.51
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
359
Cityscape
Exhibit B-8
Baseline Model Versus Neighborhood Low-Income Housing Tax Credit Concentration
(Overlap) Model (2 of 2)
Variable
Distance
from LIHTC
Property
Neighborhoods with
Any LIHTC Properties
Neighborhoods with
1, 2, or 3 LIHTC Properties
Coefficient T/F Stat Coefficient T/F Stat
2002 0.585*** 59.70 0.585*** 59.71
2003 0.764*** 72.21 0.764*** 72.22
2004 1.009*** 88.21 1.009*** 88.19
2005 1.198*** 100.99 1.198*** 100.90
2006 1.267*** 96.35 1.267*** 96.24
2007 1.012*** 84.16 1.282*** 84.09
2008 0.801*** 38.30 1.012*** 38.30
2009 0.801*** 19.29 0.801*** 19.29
2010 0.846*** 26.53 0.846*** 26.52
2011 0.811*** 24.84 0.810*** 24.81
2012 0.850*** 26.20 0.850*** 26.17
2013 1.017*** 35.33 1.017*** 35.29
2014 1.135*** 40.76 1.135*** 40.74
2015 1.217*** 40.76 1.216*** 40.74
2016 1.291*** 45.13 1.290*** 45.10
Constant 11.439*** 71.77 11.437*** 71.70
Observations 1,842,725 1,842,725
R
2
0.7242 0.7242
LIHTC = low-income housing tax credit.
+ p < 0.1. * p < 0.05. ** p < 0.01. *** p < 0.001.
Notes: Regressions also control for census tract fixed effects, which are not listed due to the large number of tracts. Treatment effect is calculated manually from
the differences in the regression coefficients, as the Methodology section describes. T-statistics are used for regression coefficients, and F-statistics are used for
treatment effects.
Sources: DataQuick Information Systems, Inc. and CoreLogic, Inc.; LIHTC HUD User database
Authors
Brian Y. An is an assistant professor of public policy and finance at the Georgia Institute of
Technology. Andrew Jakabovics is vice president for policy development at Enterprise Community
Investments. Jing Liu is an associate director at Econsult Solutions. Anthony W. Orlando is an
associate professor of finance, real estate, and law at California State Polytechnic University,
Pomona. Seva Rodnyansky is an assistant professor of urban and environmental policy at
Occidental College. Richard Voith is president and founding principal at Econsult Solutions. Sean
Zielenbach is president at SZ Consulting. Raphael W. Bostic is president and CEO at the Federal
Reserve Bank of Atlanta.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
360
Refereed Papers
References
Basolo, Victoria, Edith Huarita, and Jongho Won. 2022. “A Neighborhood-Level Analysis of Low-
Income Housing Tax Credit Developments in the State of California and Los Angeles County,”
Urban Science 6 (2): 39 https://doi.org/10.3390/urbansci6020039.
Baum-Snow, Nathaniel, and Justin Marion. 2009. “The Effects of Low-Income Housing Tax Credit
Developments on Neighborhoods,” Journal of Public Economics 93 (5–6): 654–666.
Bostic, Raphael, Andrew Jakabovics, Richard Voith, and Sean Zielenbach. 2020. “Mixed-Income
LIHTC Developments in Chicago: A First Look at Their Income Characteristics and Spillover
Effects.” In What Works to Promote Inclusive, Equitable Mixed-Income Communities, edited by Mark L.
Joseph and Amy T. Khare. San Francisco, CA: Federal Reserve Bank.
Bratt, Rachel G. 2007. Should We Foster the Nonprofit Housing Sector as Developers and Owners
of Subsidized Rental Housing? Discussion paper RR07-12. Cambridge, MA: Harvard Joint Center
for Housing Studies.
Bratt, Rachel G., and Irene Lew. 2016. “Affordable Rental Housing Development in the For-Profit
Sector: A Review of the Literature,” Cityscape 18 (3): 229–262.
Butts, Kyle. 2022. “Difference-in-Differences with Geocoded Microdata,” Journal of Urban
Economics 103493.
Castells, Nina. 2010. “HOPE VI Neighborhood Spillover Effects in Baltimore,” Cityscape 12 (1): 65–98.
Chen, Jiafeng, Edward Glaeser, and David Wessel. 2022. “The (Non-)Effect of Opportunity Zones
on Housing Prices,” Journal of Urban Economics 103451.
Cloud, William, and Susan Roll. 2011. “Denver Housing Authority’s Park Avenue HOPE VI
Revitalization Project: Community Impact Results,” Housing Policy Debate 21 (2): 191–214.
Dawkins, Casey J. 2013. “Exploring the Spatial Distribution of Low-Income Housing Tax Credit
Properties,” Journal of the American Planning Association 73 (3): 222–234.
Deng, Lan. 2011a. “The External Neighborhood Effects of Low-Income Housing Tax Credit
Projects Built by Three Sectors,” Journal of Urban Affairs 33 (2): 143–165.
———. 2011b. “Low-Income Housing Tax Credit Developments and Neighborhood Change: A
Case Study of Miami-Dade County,” Housing Studies 26 (6): 867–895.
Diamond, Rebecca, and Timothy McQuade. 2019. “Who Wants Affordable Housing in Their
Backyard? An Equilibrium Analysis of Low-Income Property Development,” Journal of Political
Economy 127 (3): 1063–1117.
Dillman, Keri-Nicole, Keren Mertens Horn, and Ann Verilli. 2017. “The What, Where, and When
of Housing Policy’s Neighborhood Effects,” Housing Policy Debate 27 (2): 282–305.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
361
Cityscape
Ding, Chengri, and Gerrit-Jan Knaap. 2002. “Property Values in Inner-City Neighborhoods: The
Effects of Homeownership, Housing Investment, and Economic Development,” Housing Policy
Debate 13 (4): 701–727.
Ding, Chengri, Robert Simons, and Esmail Baku. 2000. “The Effect of Residential Investment on
Nearby Property Values: Evidence from Cleveland, Ohio,” Journal of Real Estate Research 19 (1): 23–48.
Edmiston, Kelly D. 2018. Low-Income Housing Tax Credit Developments and Neighborhood
Property Conditions. Working paper 11-10. Federal Reserve Bank of Kansas City.
Ellen, Ingrid Gould. 2007. Spillovers and Subsidized Housing: The Impact of Subsidized Rental
Housing on Neighborhoods. Working paper RR07-3. Cambridge, MA: Harvard Joint Center for
Housing Studies.
Ellen, Ingrid Gould, Keren Mertens Horn, and Yiwen Kuai. 2018. “Gateway to Opportunity?
Disparities in Neighborhood Conditions Among Low-Income Housing Tax Credit Residents,”
Housing Policy Debate 28 (4): 572–591.
Ellen, Ingrid Gould, Amy Ellen Schwartz, Ioan Voicu, and Michael H. Schill. 2007. “Does Federally
Subsidized Rental Housing Depress Neighborhood Property Values?” Journal of Policy Analysis and
Management 26 (2): 257–280.
Ellen, Ingrid Gould, and Ioan Voicu. 2007. The Impact of Low Income Tax Credit Housing on
Surrounding Neighborhoods: Evidence from New York City. Working paper 07-02. NYU Furman
Center for Real Estate and Urban Policy.
Eriksen, Michael D., and Guoyong Yang. 2022. “Does Affordability Status Matter in Who Wants
Multifamily Housing in Their Backyards?” Presentation at the Urban Economics Association
meeting in Washington, DC.
Ezzet-Lofstrom, Roxanne, and James Murdoch. 2006. “The Effect of Low Income Housing Tax
Credit Units on Residential Property Values in Dallas County,” Williams Review 1 (1): 107–124.
Funderberg, Richard, and Heather MacDonald. 2010. “Neighbourhood Valuation Effects from New
Construction of Low-Income Housing Tax Credit Projects in Iowa: A Natural Experiment,” Urban
Studies 47 (8): 1745–1771.
Furman Center for Real Estate and Urban Policy (Furman Center). 2012. “What Can We Learn
About the Low-Income Housing Tax Credit Program by Looking at the Tenants?” New York: Moelis
Institute for Affordable Housing Policy Brief. https://furmancenter.org/files/publications/LIHTC_
Final_Policy_Brief_v2.pdf.
———. 2006. The Impact of Subsidized Housing Investment on New York City’s
Neighborhoods. Working paper 06-02 (July 2006). https://furmancenter.org/files/publications/
Impactofsubsidizedhousingcombined0602_001.pdf.
An, Jakabovics, Liu, Orlando, Rodnyansky, Voith, Zielenbach, and Bostic
362
Refereed Papers
Goetz, Edward G., Hin Kin Lam, and Anne Heitlinger. 1996. There Goes the Neighborhood?
Subsidized Housing in Urban Neighborhoods. Minneapolis: University of Minnesota Center for Urban
and Regional Affairs.
Green, Richard K., Stephen Malpezzi, and Kiat-Ying Seah. 2002. Low-Income Housing Tax Credit
Housing Developments and Property Values. Madison: University of Wisconsin Center for Urban Land
Economics Research. https://medinamn.us/wp-content/uploads/2014/04/Low-Income-Housing-
Tax-Credit-Housing-Developments-and-Property-Values-UW-Study.pdf.
Johnson, Jennifer, and Beata Bednarz. 2002. Neighborhood Effects of the Low-Income Housing Tax Credit
Program: Final Report. Washington, DC: U.S. Department of Housing and Urban Development.
Johnson, Monique S. 2012. “The Neighborhood Divide: Poverty, Place, and Low-Income Housing
in Metropolitan Richmond, Virginia (USA).” In Living on the Boundaries: Urban Marginality in
National and International Contexts, Vol. 8, edited by Carol Camp Yeakey. St. Louis, MO: Emerald
Group Publishing.
Keeler, Zachary T., and Heather M. Stephens. 2022. “The Capitalization of Metro Rail Access in
Urban Housing Markets,” Real Estate Economics.
Nguyen, Mai Thi. 2005. “Does Affordable Housing Detrimentally Affect Property Values? A Review
of the Literature,” Journal of Planning Literature 20 (1): 15–26.
Oakley, Deidre. 2008. “Locational Patterns of Low-Income Housing Tax Credit Developments: A
Sociospatial Analysis of Four Metropolitan Areas,” Urban Affairs Review 43 (5): 599–628.
O’Regan, Katherine M., and Keren M. Horn. 2013. “What Can We Learn About the Low-Income
Housing Tax Credit Program by Looking at the Tenants?” Housing Policy Debate 23 (3): 597–613.
Orlando, Anthony W., and Gerd Welke. 2022. Borrowing on the Wrong Side of the Tracks:
Evidence from Mortgage Loan Discontinuities. Working paper.
Santiago, Anna M., George C. Galster, and Peter Tatian. 2001. “Assessing the Property Value
Impacts of the Dispersed Housing Subsidy Program in Denver,” Journal of Policy Analysis and
Management 20 (1): 65–88.
Silverman, Robert Mark, and Kelly L. Patterson. 2011. “A Case for Expanding Nonprofit Activities
in Affordable Housing: An Analysis of Low Income Housing Tax Credit Outcomes 1987-2006,”
Journal of Public Management and Social Policy: 33–48.
Simons, Robert A., Roberto G. Quercia, and Ivan Maric. 1998. “The Value Impact of New Residential
Construction and Neighborhood Disinvestment on Residential Sales Price,” Journal of Real Estate
Research 15 (2): 147–161.
Smith, Brent C. 2003. “The Impact of Community Development Corporations on Neighborhood
Housing Markets,” Urban Affairs Review 39 (2): 181–204.
Factors Affecting Spillover Impacts of Low-Income
Housing Tax Credit Developments: An Analysis of Los Angeles
363
Cityscape
Turbov, Mindy, and Valerie Piper. 2005. HOPE VI and Mixed-Finance Redevelopments: A Catalyst for
Neighborhood Renewal. Washington, DC: The Brookings Institution Metropolitan Policy Program.
Voith, Richard, Jing Liu, Sean Zielenbach, Andrew Jakabovics, Brian An, Seva Rodnyansky,
Anthony W. Orlando, and Raphael W. Bostic. 2022 “Effects of Concentrated LIHTC Development
on Surrounding House Prices,” Journal of Housing Economics 56: 101838.
Woo, Ayoung, Kenneth Joh, and Shannon Van Zandt. 2016. “Unpacking the Impacts of the Low-
Income Housing Tax Credit Program on Nearby Property Values,” Urban Studies 53 (12): 2488–2510.
Young, Cheryl. 2016. “There Doesn’t Go the Neighborhood: Low-Income Housing Has No Impact
on Nearby Home Values.” https://www.trulia.com/research/low-income-housing/.
Zielenbach, Sean, Richard Voith, and Michael Mariano. 2010. “Estimating the Local Economic
Impacts of HOPE VI,” Housing Policy Debate 20 (3): 485–522.
