literature review
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=rhpd20
Download by: [University of Utah] Date: 21 July 2017, At: 14:21
Housing Policy Debate
ISSN: 1051-1482 (Print) 2152-050X (Online) Journal homepage: http://www.tandfonline.com/loi/rhpd20
How Location Efficient Is LIHTC? Measuring and
Explaining State-Level Achievement
Arlie Adkins, Andrew Sanderford & Gary Pivo
To cite this article: Arlie Adkins, Andrew Sanderford & Gary Pivo (2017) How Location Efficient
Is LIHTC? Measuring and Explaining State-Level Achievement, Housing Policy Debate, 27:3,
335-355, DOI: 10.1080/10511482.2016.1245208
To link to this article: http://dx.doi.org/10.1080/10511482.2016.1245208
Published online: 06 Jan 2017.Submit your article to this journal Article views: 200View related articles View Crossmark data Housing Policy Debate, 2017
Vol. 27, no . 3, 335– 355
http://dx.doi.org/10.1080/10511482.2016.1245208
How Location Efficient Is LIHTC? Measuring and Explaining
State-Level Achievement
Arlie Adkins , Andrew Sanderford and Gary Pivo
school of landscape architecture and Planning, university of arizona, tucson, usa
ABSTRACTA growing recognition that the cost of transportation should be included
in calculations of housing affordability has led to efforts to promote
location efficiency (LE) in affordable housing policy. Because the program
is responsible for most new affordable housing in the United States, the Low Income Housing Tax Credit (LIHTC) program has the potential to be a link
between housing affordability and LE. This research analyzes the extent to
which LIHTC units built between 2007 and 2011 were in location-efficient
places. Ordinary least squares regression analysis was used to test the role
of market, policy, developer, and urban form factors in determining state-
level LIHTC LE. We find that for the nation as a whole, from a quarter to half
of LIHTC units added during this period were in location-efficient places,
depending on the LE criteria applied. State-by-state comparisons showed
wide variation in both our absolute measures of LIHTC LE and our relative
measures of LIHTC LE compared with overall housing in each state. State
policy and nonprofit developers were associated with higher LIHTC LE and
had a positive effect on a state’s ability to outperform its underlying urban
form.
The primary aim of affordable housing policy is to make housing available at a price affordable to
low-income households, either directly or through subsidies to tenants or developers. Whether a house -
hold can afford to live in a particular place, however, depends on more than just the cost of renting
or buying a housing unit (Haas, Makarewicz, Benedict, & Bernstein, 2008). Transportation costs are
typically a household’s second largest expense after housing, so transportation-related characteristics
of housing locations such as the accessibility of employment, services, and affordable transportation
modes should also be considered when determining affordability.
Location affordability is an umbrella term for this broader set of characteristics that together more
holistically capture the costs and benefits of living in a particular location. Location efficiency (LE),
the accessibility of a site or neighborhood to everyday destinations and a determinant of transpor -
tation costs, is a key component of location affordability. When applied to affordable housing, LE has
the potential to be beneficial for low-income households while also producing broader societal gains
(EPA, 2011; HUD, n.d.). Given this potential, as well as recent evidence that in some cities low-income
households are increasingly being priced out of location-efficient neighborhoods (Adkins, 2014), we
see value in exploring LE within affordable housing programs. In this article we investigate the extent
to which the Low Income Housing Tax Credit (LIHTC), one of the nation’s largest affordable housing
programs, is producing location-efficient housing.
ARTICLE HISTORYReceived 8 July 2015
a ccepted 3 o ctober 2016
KEYWORDSlow-income housing;
location efficiency; location
affordability; l ow-income
Housing tax credit; urban
form
© 2017 Virginia Polytechnic i nstitute and state university
CONTACT arlie adkins [email protected]
336 A. ADKINS ET AL.
We aim to answer two related research questions. First, what share of LIHTC, both nationally and
in each state, has been built in location-efficient locations? And second, what policy and other factors
explain differences between states in their share of LIHTC that is location efficient? Before proceeding, we offer two important caveats. First, insofar as this is the first nationwide look
at the LE of LIHTC, we recognize that there is currently no established standard for the LE of affordable
housing. We recommend that further work be done to standardize and validate appropriate measures.
Furthermore, although our work is based on the assumption that location-efficient affordable housing has
potential benefits for low-income households, we recognize that there are potential tradeoffs involved
in developing and siting affordable housing in location-efficient areas (for a discussion of these tensions,
see Goetz, 2015). Our research findings are discussed within the context of these potential tradeoffs. Foreshadowing our results, our analysis indicates that, nationally, LIHTC units built during the study
period were in more location-efficient census block groups (CBGs) than overall housing stock was.
Nonetheless, most of the LIHTC units built during the study period were not in location-efficient CBGs,
suggesting a role for further policy intervention. Moreover, our multivariate analysis suggests that
although LIHTC LE is determined by underlying urban form and market conditions, state housing policy
and an active nonprofit sector can help a state achieve greater LE in its LIHTC portfolio than would be
expected given its underlying urban form. We begin with a summary of the relevant literature that helped inform our study. Next we present
our data, methods, and findings, before concluding with a discussion of policy implications and rec -
ommendations for future research.
Literature
What Is Location Efficiency?
Definitions of LE describe locations with a variety of characteristics that reduce both the social and
household costs of transportation by increasing accessibility.
According to the U.S. Department of Housing and Urban Development (HUD):
Location efficiency describes how accessible everyday destinations—like jobs, shopping, entertainment, parks and other amenities—are to a neighborhood or community. Location-efficient neighborhoods are often characterized by compact, mixed-use development, easy access to public transit, good walking and biking conditions, and nearby commercial and retail hubs. (HUD, n.d.)
Similarly, according to the U.S. Environmental Protection Agency (EPA):
Location efficient sites are located near transit and use compact design to facilitate pedestrian access to transit,
linking people to a range of services, amenities, and employment centers. They include a mix of uses, and offer
comfortable and convenient transit service, thereby increasing the number of viable transportation options available to residents to commute to work, school, or other destinations. (EPA, 2011)
Ewing and Cervero’s meta-analysis of the travel and built environment literature (2010) shows that
many of the features used to determine LE are indeed associated with lower vehicle miles traveled,
more walking, and transit use. The features they found to make the most difference include accessi-
bility (i.e., time-distance to desired destinations), street network design (e.g., urban grids), land use
diversity, intersection density, the number of destinations within walking distance, and proximity to
transit. Similarly, Haas et al. (2008) suggest that LE is the product of accessibility, street network design,
density, and transit connectivity.
Benefits of Location Efficiency
Previous researchers have found that LE may be beneficial to general social and environmental welfare.
The social benefits of LE could include better public health as a result of more physical activity (Berrigan
et al., 2012; Lee & Buchner, 2008; World Cancer Research Fund/American Institute for Cancer Research,
2007) and increased social capital including more community cohesion, political participation, trust,
HOUSING POLICY DEBATE 337
and social activity (Du Toit, Cerin, Leslie, & Owen, 2007; Leyden, 2003; Rogers, Halstead, Gardner, &
Carlson, 2011; Wood, Frank, & Giles-Corti, 2010). Social capital has in turn been linked to the capacity
of cities to transition toward greater sustainability (Geels, 2012; Portney, 2005). Environmental benefits
may include less air pollution, auto use, and gasoline consumption (Cao, Handy, & Mokhtarian, 2006;
Ewing & Cervero, 2001; Frank & Engelke, 2005; Frank, Stone, & Bachman, 2000; Handy, Cao, & Mokhtarian,
2005). LE may also be a strategy for mitigating climate impacts from the transportation sector (Bosch
& Metz, 2011; Chapman, 2007).
For lower income households, LE may be particularly beneficial for several reasons. First, low-income
households often spend a disproportionately higher share of household income on transportation
compared with higher-income households (Litman, 2003), so reduced transportation costs have the
potential to have a greater positive impact on the financial well-being of lower income households.
Similarly, for those who cannot afford to own or operate a car, having nearby services and opportu-
nities and the means to safely access them by transit, walking or bicycling is a necessity. And, second,
many chronic health conditions linked to physical inactivity (e.g., obesity and cardiovascular disease)
are more prevalent in low-income populations (Korda, Paige, Yiengprugsawan, Latz, & Friel, 2014), so
locations supportive of active transportation could contribute to positive individual health outcomes
and greater health equity.
The association between opportunity structures facing lower income households and their neigh-
borhood conditions is an important concern to affordable housing scholars and advocates (Squires &
Kubrin, 2005). Transportation is one factor that affects the opportunity set available to lower income
households. Spatial mismatch between the locations of jobs available to low-wage workers and afforda -
ble housing has long been discussed as contributing to either reduced opportunities or higher trans-
portation costs (Ihlanfeldt & Sjoquist, 1998; Kain, 1968, 1992). More recently, Fisher, Pollakowski, and
Zabel (2009) argued that the “spatial opportunity set facing households” can vary by location within a
metropolitan area because of differences in accessibility to jobs, services, education, and other locational
amenities. If access can shape opportunity, then it is reasonable to ask whether location-efficient places
with lower transport costs could make it easier for lower income households to access opportunities. Haas et al. (2008) looked specifically at the estimation of location-specific transportation costs so
they could be included in calculations of housing affordability. They found that household transpor -
tation costs could be predicted by considering certain dimensions of urban form where a household
is located by using variables including residential density, job density, transit connectivity, block size,
and distance to the nearest large employment center. The urban form variables they used are similar to
those that Ewing and Cervero (2010) found to be good predictors of travel behavior. Thus, urban form
can be said to be a common driver of both travel behavior and related household transportation costs. So far no studies have established a clear connection between LE and the employment outcomes
for lower income households. Indeed, in one study, Sanchez, Shen, and Peng (2004) found that being
near transit did not improve employment prospects for low-income workers (Sanchez et al., 2004).
More recently, Blumenberg and Pierce (2014) found that “keeping or gaining access to an automobile
is positively related to the likelihood of employment” whereas “improved access to public transit is
positively associated with maintaining employment, but not with job gains” (p. 52). Meanwhile, Klein
and Smart (2015) found that automobile access was associated with a decreased probability of future
unemployment, but those costs of owning and maintaining a car, they argued, may be greater than
the income gains associated with increased car ownership. They also found a less clear relationship
between public transit and improved economic outcomes. Their results showed that high-quality public
transportation had no effect on future earnings for lower income households, but transit was important
in providing for those who cannot or choose not to own a car. This work suggests that both transit and
auto accessibility may be significant elements when considering the LE of affordable housing, and we
have included both transit- and car-related accessibility indicators of LE in our analysis. Because location-efficient housing locations tend to be in more urban places, there is potential for
efforts to promote LE affordable housing may be at odds with policy efforts aimed at reducing concen -
trated urban poverty by facilitating access to neighborhoods with greater opportunity (Goetz, 2015).
338 A. ADKINS ET AL.
Potential tradeoffs for low-income households include neighborhood safety (Katz, Kling, & Liebman,
2000), access to quality schools (Makarewicz, 2013), racial concentration and segregation (Horn &
O’Regan, 2011), and environmental quality (Katz et al., 2000). Bernstein (2011), however, points out
that there are many low-poverty (i.e., high-opportunity), location-efficient places with shortages of
affordable housing.
Location, Market, and Policy Factors
Three studies have looked at the location of affordable housing in relationship to certain dimensions
of LE. Welch (2013) found that LIHTC properties in Baltimore had less transit access than would be
expected from a random distribution of housing, Talen and Koschinsky (2014b ) found that 72% of
the HUD-subsidized units (public housing and vouchers) in six large cities were in areas with poor
walkability, and Pivo (2014) found that affordable multifamily properties financed by Fannie Mae were
in locations with less transit service and less walkability compared with other multifamily buildings. Work by Lang (2012), Newman and Schnare (1997), and Rohe and Freeman (2001) suggests why
subsidized housing may not be prominent in LE neighborhoods. They found that developers are more
likely to build LIHTC housing in locations with low rent. The explanation for this is that locations with
higher rents have a higher opportunity cost of developing subsidized housing. Increased land costs in
walkable, transit-rich areas observed by Bartholomew and Ewing (2011), Cortright (2009), Leinberger
and Alfonzo (2012) and others likely contribute to the higher rents and thus make it more difficult to
create location-efficient affordable housing. Other studies concerned with the location of LIHTC housing would seem to support these findings.
Dawkins (2013), in his review of LIHTC location studies, concluded that LIHTC properties are concen-
trated in central cities, and although the suburban share is increasing, that is occurring in locations
with high poverty rates, in tracts with high minority concentrations, and in neighborhoods that feed
into lower performing schools. Although LIHTC projects may allow people to live in less disadvantaged
places than public housing (Galster, 2013) they are not helping lower income renters locate in low-pov -
erty areas (McClure, 2008).
Meanwhile, LIHTC locations can also be influenced by policy or political factors linked to state housing
policies, so housing policies could channel LIHTC into location-efficient places. Most but not all LIHTC
are allocated using state-established criteria (exceptions are New York City, New York, and Chicago,
Illinois). Deng (2011) and Dawkins (2013) show that state Qualified Allocation Plans (QAP) can affect
LIHTC locations, while Gay (2015) found modest evidence that LIHTC allocations favor locations that
voted for the state’s governing party. Nedwick and Burnett (2015), in a study directly relevant to ours,
show that QAPs that award points for transit proximity are associated with a slight increase in the
likelihood of LIHTC being located near fixed-guideway transit. Meanwhile, Ellen, Horn, Kuai, Pazuniak,
and Williams (2015) found no statistical evidence that developers have responded to the variation over
time in the point allocation schemes used by state agencies in their QAPs. Most of the evidence seems
to support the possibility that QAPs can alter LIHTC location patterns independent of market forces.
Methodology
Study Overview
We examined the LE of affordable housing units built under the LIHTC program from 2007 to 2011, which
was the most recent 5-year period for which there was national data when the study was undertaken.
We focused on this period because we were interested in characterizing contemporary development
practice. Our analysis was at the state level because while the federal government distributes LIHTC
credits, the criteria for their allocation to developers and projects are typically established by states
through their QAPs. Because of this, any effort to increase the LE of affordable housing via LIHTC will
almost certainly involve state housing policy.
HOUSING POLICY DEBATE 339
Our analysis was done in several stages. First, we determined the share of LIHTC housing units that
were in CBGs meeting our established LE thresholds for each LE criterion: residential density, network
density, retail sufficiency, regional employment accessibility, transit ridership, proximity to rail transit,
and transportation costs. Each LE criterion and its LE threshold are described in detail below.
Next, we averaged the share of LIHTC units for each state, and nationally, that were in CBGs meeting
each LE criterion to have national- and state-level indicators of overall LIHTC LE. Additionally, because
the underlying built environment of each state may mask the effect of other factors, we created two
relative LIHTC LE variables. One benchmarked our LIHTC LE variable for each state against the LE of all
housing units in the state and one against the LE of all rental housing units in the state. The LE for all
housing and all rentals was determined using the same method that produced our overall LIHTC LE
score (an average of proportions). The two relative LIHTC LE variables were produced by computing
the difference between our LIHTC LE indicator and the LE of all housing and all rental housing in each
state. These steps provided us with three state-level dependent variables: absolute LIHTC LE; LIHTC LE
relative to all housing, and LIHTC LE relative to all rental housing. The relative measures illustrate the
degree to which a state’s LIHTC is outperforming its underlying LE. We used these three dependent LE variables to describe LIHTC LE nationally and in each state.
Additionally we used ordinary least squares (OLS) regression models to test whether state policy and
developer characteristics could explain observed variation between states in these dependent variables
after controlling for statewide market and built environment conditions.
Below, we describe the two data sets that we created to carry out our analysis. The first characterizes
the LE of each CBG as well as the total number of housing units, rental housing units, and LIHTC housing
units in each CBG. The second is a state-level data set that includes the dependent variables derived
from the first data set as well as other explanatory and control variables used in our regression models.
CBG Location Efficiency Data Set
To determine the share of LIHTC units that are location efficient, we first had to determine whether each
CBG was location efficient. A key challenge we faced was that although the dimensions of LE (such as
density or transit service) may be known, the thresholds of each dimension necessary for a place to be
considered location efficient are not. Some recommended standards are found in the literature, and
where possible we used those. For example, many have cited seven housing units per net acre as the
minimum density for good transit service; and transportation costs are thought to be affordable when
they remain below 20% of household income (Calthorpe, 1993; Duany, Plater-Zyberk, & Speck, 2010).
However, for variables where there was no established standard for LE, we had to produce our own
using the methods described below. These are not based on detailed studies looking for cut points at
which significant benefits emerge for low-income households. We acknowledge that these thresholds
should be sought in future studies to find the best possible standards for the LE of affordable housing. Our effort to define LE standards was based on two key assumptions. The first was that the CBG is
the correct level of analysis. Although some LE metrics may be more strongly correlated with house -
hold well-being when described at other scales, we focused on CBGs for several reasons. As a matter
of practicality, CBGs are the geographic units used in both EPA’s Smart Location Database (SLD) and
HUD’s Location Affordability Index (LAI) data sets, which were primary sources for our work. But more
importantly, CBGs are appropriate geographic units for understanding conditions near a person’s home
that are likely to be associated with accessibility. The second assumption was that it is appropriate to describe LE as a categorical or dichotomous as
opposed to continuous metric. We understand there may be utility in using continuous variables where
we might report, for example, the average score on an LE metric for block groups with and without
LIHTC in a state. However, we feel that the LE metrics we used are often discussed as threshold-based
standards and guidelines in practice, where, for example, an area does or does not achieve the minimum
density to support public transit. In an applied setting, thresholds or cut-points can be more useful
340 A. ADKINS ET AL.
than continuous indicators because they allow a simple classification of cases into qualified and not
qualified, and they communicate clearly the threshold above (or below) which LE will consistently be
above (or below) some given norm or minimum requirement. We feel that, in practice, policymakers
and planners would find it easier to work with data that show the degree to which they are or are not
placing LIHTC in LE places. This use of binary metrics in this field, however, is no doubt just a starting
point for this sort of analysis. There could be multiple thresholds that should be considered for any
given metric, or relative levels of achievement described along a continuum. Future studies on the
relationship between different levels of LE metrics and various outcomes for household, society, and
the environment could help untangle what we admit is a complicated definitional and policy problem
(see Pivo 2014 as an example of such work on walkability which found two distinct thresholds that
affect apartment building default risk).
To link the LIHTC property list to the LE data we obtained a 2010 CBG geographic identifier code
(GEOID) for each LIHTC property. This was done by mapping the LIHTC properties using latitudes and
longitudes given in HUD’s Low Income Housing Tax Credit Database (LIHTCD) and then overlaying that
map with a U.S. Census Bureau CBG Cartographic Boundary file that included file-based metadata with
the GEOID for each CBG. This gave us a CBG GEOID for each LIHTC property. We then imported data
from the HUD LAI, EPA SLD, and Census Bureau American Community Survey (ACS) data sets about the
LE and other characteristics of the CBGs where each LIHTC unit was located, matching the data sets
using the CBG GEOIDs. The result was an LIHTC property data set that included information about each
property (e.g., number of dwelling units) and its CBG (e.g., housing unit density). We then used CBG data in our LIHTC property data set to determine whether each property was in
a location-efficient CBG. These CBG data, as noted, came from two sources. The first was EPA’s SLD. The
SLD includes information describing housing and employment density, land use diversity, design of
the built environment, destination accessibility, and transit service measures for all 2010 census CBGs.
Built as a research resource by the EPA and other agency members of the Sustainable Communities
Partnership, the SLD assists those investigating the relationships between location and other social
and market phenomena. Measures in the SLD are largely drawn from information gathered by the
U.S. Census Bureau as part of the Decennial Census, the ACS, the Longitudinal Employer-Household
Dynamics survey, the U.S. Geological Survey’s Protected Area Database, and a range of proprietary
data licensed to the EPA (e.g., NAVSTREETS). In many cases, the SLD combines measures from these
individual data streams to create aggregate or meta-variables describing a more complex phenomenon.
The second source of LE data was HUD’s LAI database. The LAI uses regression models to generate
estimates of the expenditures families make on both housing and transportation at the CBG level
for all populated CBGs in all 50 states. As the estimates are generated for several different household
types, the LAI provides insight into the cost of living in a specific geography, controlling for household
size and location. LAI also includes other information about CBGs including one of our transit service
variables, discussed below. We now describe each of the variables in our LIHTC property data set and the thresholds adopted
for determining the LE of each CBG.
LIHTC Units. This variable indicated the number of LIHTC housing units built during our study period
in each CBG. Our LIHTC properties data set was built by first obtaining a list of all LIHTC properties put
into service from 2007 through 2011. We obtained the list from HUD’s LIHTCD, which is a complete
national data set on the location and other characteristics of LIHTC properties. It gives information on
more than 39,000 LIHTC housing units placed into service from 1987 through 2011. We did not extract
information on the properties in U.S. territories, protectorates, or other administrative units because
those locations were not covered by the other data sets that are described below which we used to
characterize LE. Further, we constrained the LIHTC property data to those properties placed into service
from 2007 to 2011, for reasons given above. This resulted in a final data set containing 4,768 projects
and 325,292 low-income units. The LIHTC database has some missing data, but missing data are more
likely in projects built before our study period (Climaco, Finkel, Nolden, & Vandawalker, 2006; McClure,
HOUSING POLICY DEBATE 341
2006) and there is no reason to believe that the distribution of missing LIHTC data would be patterned
in a way that would produce bias with regard to LE.
Residential Density. This is a dummy variable indicating whether the housing density in the CBG was
(1) or was not (0) at least seven units per acre. This was derived from variable D1a (gross residential
density, HU/acre on unprotected land in the CBG) in the SLD. Seven housing units per net acre has been
cited by Calthorpe (1993) as the minimum density for neighborhood transit-oriented development, and
by many others as the minimum density to support moderate-level transit service (e.g., Cervero, 2004;
Downs, 2005; Duany et al., 2010). Because we used the SLD’s measure of gross residential density, we
explored using a higher threshold for gross residential density to correspond with seven units per net
acre. However, given the national distribution of gross residential density, even increasing the threshold
to 10 resulted in some states having very few qualifying CBGs. Network Density. This is a dummy variable indicating whether the network density in the CBG was
greater than 23 miles of street length per square mile based on SLD variable D3a. Measures of connec -
tivity such as network density have been widely incorporated into research on the impacts of urban
form on transportation outcomes (Dill, 2004) and were one of the variables that Ewing and Cervero
( 2010) found to be associated with travel outcomes. Higher connectivity makes travel by nonmotorized
modes more feasible while lower density requires more circuitous, longer routes more suitable to car
travel. In the absence of an established standard we used the top quintile of network density in the
distribution for the CBGs in all 50 states, which was 23 miles of street per square mile.
Retail Sufficiency. This is a dummy variable indicating whether there were 0.043 retail jobs per
household in the CBG. There is no explicit measure of local retail accessibility in the SLD, so we used
retail employment as a proxy. Based on the simplifying assumption that the number of retail jobs per
household in the United States is equal to the number needed to provide sufficient retail services to
households in a CBG, we used that ratio (0.043 jobs per household) as the standard for sufficient retail
services at the CBG level.
Regional Auto Accessibility. This is a dummy variable for whether the Regional Centrality Index for the
CBG, given by the SLD (D5cri), was in the top quintile for the national distribution (or a value of 0.72).
Again, because there is no national standard for defining the level of regional accessibility that is location
efficient, we used the top quintile for the distribution as the standard. The Regional Centrality Index (SLD
variable D5cri) is a score that ranges from 0 to 1 and provides the relative employment accessibility by
auto for each block group relative to the maximum in the core based area. The variable is specific to
each CBG but is calculated using regional employment data. Further details about its calculation are
provided in Ramsey and Bell (2014).
Transit Ridership. This is a dummy variable for whether at least 5% of the working-age population in
the CBG commuted via any form of transit according to ACS 2010–2014 estimates (U.S. Census Bureau,
2015). This was used as a proxy for there being at least a moderate level of transit service. Based on an
examination of the national distribution of transit ridership, we used a 5% threshold as the amount of
transit commuting that suggests a place is supportive of transit. We recognize that transit ridership is a
rough proxy for transit service, but, lacking national CBG-level transit service data, we think that using
this metric in concert with the rail transit variable provides a meaningful indication of transit-related LE.
Half-mile to Rail Transit. This is a dummy variable indicating whether any portion of the CBG is within
a half-mile of a fixed-rail station. It is derived from SLD variable D4b050, proportion of CBG employ -
ment within a half-mile of fixed-guideway transit stop. For states with no rail transit, this variable was
coded as missing and was not included in averages of state LIHTC LE. Half-mile buffers are frequently
used to define the area in which a rail transit station will have the most influence over travel behavior,
and recent empirical evidence suggests that despite some variation in different contexts, this distance
remains a useful standard (Guerra, Cervero, & Tischler, 2011). Transportation Costs. This is a dummy variable derived from the LAI database indicating whether
transportation costs in the CBG are modeled to be 20% or less of household income for a single-parent
household earning 50% of the median area income (LAI household type 6). Twenty percent was selected
as the cutoff based on previous national research indicating that 20% was the mean transportation
342 A. ADKINS ET AL.
expenditure across 28 metro areas (Haas, Makarewicz, Benedict, Sanchez & Dawkins, 2006). This standard
has also been incorporated into Center for Neighborhood Technology’s (2010) Housing + Transportation
(H+T ) Index definition of housing affordability as 30% of household income and combined housing
and transportation affordability as 50% of household income.
A reliability test performed on our seven LE indicator variables indicated good internal consistency
(α = .77).
State-Level Data Set
We needed a second data set at the state level to answer our question about why some states have
more location-efficient LIHTC units than others. The following describes the eight state-level variables
we used as independent variables in the OLS regression models. A summary of each state level variable
can be found in Table 1. LE_Housing and LE_Rental are the percentage of all housing or all rental housing units in a state that
are in CBGs that meet three of our seven LE thresholds criteria. They are intended as control variables for
the overall LE of urban form for each state. We chose a cutoff of three of our seven LE criteria based on
the distribution observed in Figure 1 to establish a relatively strict test for designating location-efficient
CBGs that did not have the effect of excluding any states from having at least some location-efficient
CBGs. As already noted, however, this standard deserves further research-based investigation to find the
best combination of location-efficient variables and thresholds that produce the best possible benefits
for the occupants of affordable housing. We expected both variables would have a positive association
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80
0.90
01+2+3+4 +5+6 +
Share of Units
LE Criteria Met
All housing
Rental housing
LIHTC
Figure 1. national share of each housing type meeting location efficiency criteria.Note. liHtc = l ow-income Housing tax credit; le = location efficiency.
Table 1. summary of state-level data.
Note. liHtc = l ow-income Housing tax credit; le = location efficiency; Qct = qualified census tract; Q aP = qualified allocation plan.
NMin MaxMean SD
liHtc le Relative to all housing 500 0.34 0.120.06
absolute 500.17 0.59 0.300.09
Relative to rental housing 500 0.14 0.070.03
Qct_le 500.09 0.85 0.430.21
Housing_le 500.64 2.9 1.140.43
le_rent_premium 50−0.67 0.69−0.04 0.18
nonprofit_share 500 1 0.280.26
Q aP_services 450 1 0.620.49
HOUSING POLICY DEBATE 343
with our three dependent variables (absolute LIHTC LE; LIHTC LE relative to rental all housing, and LIHTC
LE relative to all rental housing). However, because of multicollinearity, only LE housing was included
in the final models, although both variables produced similar results.
LE_Rent_Premium is the difference between the median rent in all CBGs in the state that meet three
of the seven LE criteria and the median rent in all other (i.e., nonlocation efficient) CBGs. It was calculated
from median rent data in the 2009–2014 American Community Survey 5-Year Estimates (U.S. Census
Bureau, 2015). Based on findings in Lang (2012), our expectation was that this variable would be neg-
atively associated with LIHTC LE. Like Lang, we expected that developers are more likely to build LIHTC
projects in areas where they can maximize both program subsidies and profit incentives. QCT_LE is the percentage of qualified census tracts (QCTs) meeting three of our seven LE criteria.
QCTs are census tracts designated by HUD in which at least 50% of the households earned less than
60% of the area median income. QCT_LE was derived by dividing the number of QCTs that achieved
at least three of the LE metrics by the total number of QCTs in the state. To calculate the variable, we
counted which of the QCTs had at least one CBG that achieved at least three of the LE metrics. Then, we
divided the result by the total number of QCTs in the state. We expected this variable to have a positive
association, indicating that the greater the intensity of poverty (as indicated by QCTs) in LE places, the
higher the proportion of LIHTC that would occur in LE CBGs. This is partly due to market drivers, but
also because some QAPs prioritize QCTs. Nonprofit_share is a ratio variable that describes the percentage of LIHTC housing units built in a
state that have a nonprofit sponsor. This variable is drawn from the LIHTC database. During our study
period, the nonprofit Enterprise Community Partners initiated their Green Communities program, a
$500 million investment in green affordable housing. The initiative included LE attributes as a part of
their definition of a green community. We expected that nonprofits might be better positioned to prior -
itize attributes such as LE that may serve a particular nonprofit’s mission, but not necessarily maximize
profit. We found no previous research telling us whether this expectation is valid in terms of LE, but
there is evidence that nonprofit affordable housing developers are more likely to consider social and
community aspects of their projects and are more willing to develop in areas and for populations not
well served by private developers’ projects (Bratt, 2008; Ellen & Voicu, 2006). We expected the share of
LIHTC units with nonprofit sponsors to be positively associated with LIHTC LE. QAP_Services, QAP_Transit and QAP_Active are dummy variables that describe whether a state pri-
oritized proximity to services, transit accessibility, or being located in a place supportive of walking or
bicycling in their 2006 QAP (the year before our study period). QAPs contain state policies that guide their LIHTC credit allocation decisions. State housing finance
agencies submit a QAP to HUD each year detailing their LIHTC credit allocation system. Data on all state
QAPs came from Novogradac & Co., LLP. We used the QAPs to develop a binary measure indicating
whether the QAP included scoring systems or language that prioritized LIHTC projects with access to
key services, transit, or supportive infrastructure for active transportation. Because of the co-occurrence
of these variables in many QAPs, however, only QAP_services was included in the final regression
models because it had the largest effect when tested in models with other QAP variables. After Ellen
et al. (2015) and Nedwick and Burnett (2015), we expected this variable to have a positive but weak
association with LIHTC LE and both relative LIHTC LE variables.
State QAPs are changed and updated periodically, so our snapshot of 2006 does not perfectly reflect
the policy context for every LIHTC project at the time of its tax credit allocation. Some LIHTC units
included in our study were allocated before 2006, and some after. The year 2006 was chosen for the
QAP variables because it is the year that best reflects LIHTC allocations within our study period. Ninety
percent of LIHTC projects placed into service during our study period were allocated in 2006 or later.
Less than 1% were allocated prior to 2004. As a test of whether 2006 QAPs were appropriate for our
analyses, we looked at change in QAPs between 2002 and 2006 and between 2006 and 2010. Although
several states’ QAPs changed with respect to our QAP variables during this time, only about 2% of units
in our study period were allocated in states and at times where and when the 2006 QAP would not
accurately reflect the policy when the credits were allocated. Removing those 2% of units from our
344 A. ADKINS ET AL.
analysis did not result in any meaningful difference in model R 2 or the size, direction, or significance of
any coefficients. As a result, rather than removing these data, we chose to keep all LIHTC data during
our study period of 2007 to 2011.
Data Reliability
For our analysis we used data from several sources (and those sources calculated variables from several
additional sources). Although we are confident that our analysis paints a valid picture of state-level LE,
we do have some reservations about data reliability for those variables derived from the ACS 5-year
estimates and the Longitudinal Employer–Household Dynamics (LEHD) program. For example, transit
use data from ACS generally has large margins of error because of low ridership and the small geography
of block groups. Similarly, data derived from the LEHD are imperfect. Data are not available for all states
for all years—most notably, Massachusetts prior to 2011—and the Census Bureau has acknowledged
difficulties in assigning individual employees to the correct worksite if the company has more than one
location (Graham & Ong, 2007). This diminishes the accuracy of our retail sufficiency indicator, which
was derived from LEHD. We have chosen to include these indicators in the calculation of our dependent variables despite
concerns about reliability, for several reasons. First, we are using these variables in an admittedly rough
manner to create binary indicators of LE. This does not eliminate, but rather lessens, the risk that we are
misidentifying LE places because we are not making fine-grained distinctions. And, second, as pointed
out by Spielman and Singleton’s (2015) justification for using ACS data, using multiple variables to
create typologies of small geographies, as we have done, lessens data reliability concerns related to
high margins of error. Furthermore, four of our seven LE indicators are derived from sources that do
not share the sampling and reliability concerns of ACS or LEHD. Using a combination of these seven
indicators lessens the risk that even systematic error in any one variable would undermine our results.
These strategies and evidence of high internal consistency among our indicators (α = .77) give us
confidence that our dependent variables adequately and appropriately represent variations in state
LE, as we have defined it.
Methods of Analysis
Descriptive Analysis of Absolute and Relative LIHTC Location Efficiency
As previously described, we used our CBG-level LE data to calculate the proportion of each housing
type meeting each of the LE criteria, both nationally and in each state (see Table 2). We then averaged
these proportions across each of the LE criteria to determine our three dependent variables: absolute
LIHTC LE, and LIHTC LE relative to both all housing and rental housing (see Table 3). In addition to being
the dependent variables for our regression analyses, these variables are also useful for describing the
LE of LIHTC in the United States and for making comparisons between states.
OLS Regression
To expand the analysis of the seven LE metrics and to develop an explanatory model of state achieve -
ment answering our second research question, we developed simple ordinary least squares (OLS) regres -
sions using the dependent variables described above predicted by policy, market, poverty, developer,
and urban form factors. The dependent variables were normally distributed, making OLS regression
appropriate. However, because our dependent variables were calculated from a series of proportions,
we also ruled out sigmoidal (S-shaped) relationships between our independent and dependent varia-
bles as a precaution. Sigmoidal relationships, which may occur with proportional dependent variables,
would have made OLS regression inappropriate (Cohen, 2003).
We tested three models. In Model 1, the goal was to measure the influence of the five independent
variables on states’ proportion of LIHTC in location-efficient CBGs. The dependent variable is LIHTC LE,
HOUSING POLICY DEBATE 345
Table 2. share of each housing type meeting each location efficiency criterion.
Res. density (7+ HU/acre) Connectivity (network density > 23) Retail sufficiency (> .043 jobs/HH)Relative regional accessibility (> .72)
State All
housing RentalLIHTC LIHTC
diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTC LIHTC
diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTC LIHTC
diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTC LIHTC
diff.
(housing) LIHTC
diff.
(rental)
all states 0.140.25 0.30 0.16 0.050.170.240.28 0.100.040.390.420.52 0.13 0.100.170.250.32 0.150.07
aK 0.050.10 0.00−0.05 0.050.100.160.15 0.050.060.330.420.73 0.40 0.100.230.340.46 0.230.11
al 0.010.03 0.00−0.01 0.020.030.060.03 −0.01 0.030.430.520.56 0.14 0.090.210.340.41 0.190.13
aR 0.0040.010 0.00 0.00 0.010.050.080.11 0.060.030.410.490.58 0.17 0.080.260.380.39 0.130.12
aZ 0.070.16 0.10 0.03 0.080.190.240.17 −0.02 0.050.340.400.50 0.16 0.060.140.240.53 0.390.10
ca 0.250.40 0.36 0.11 0.150.280.320.28 0.000.050.360.400.53 0.17 0.040.160.220.32 0.170.07
co 0.100.20 0.29 0.19 0.100.170.220.22 0.050.050.370.440.49 0.12 0.060.220.340.46 0.240.11
ct 0.110.26 0.48 0.37 0.150.070.150.26 0.190.070.390.430.54 0.16 0.040.150.270.41 0.260.12
De 0.080.18 0.25 0.17 0.090.090.150.14 0.050.060.460.460.68 0.22 0.000.130.170.26 0.130.05
Fl 0.110.18 0.14 0.03 0.070.190.240.25 0.060.050.420.480.47 0.06 0.060.130.200.29 0.160.07
ga 0.030.06 0.05 0.02 0.030.030.050.21 0.180.020.420.480.45 0.02 0.060.170.270.41 0.240.09
Hi 0.260.36 0.34 0.08 0.100.230.290.09 −0.15 0.060.300.340.40 0.10 0.040.170.260.11 −0.06 0.08
ia 0.020.05 0.04 0.02 0.030.090.140.26 0.170.050.490.560.54 0.06 0.070.350.500.47 0.120.15
iD 0.010.01 0.00−0.01 0.010.070.120.04 −0.04 0.040.450.530.65 0.20 0.080.320.470.45 0.130.14
il 0.230.41 0.36 0.13 0.180.190.280.26 0.070.090.390.390.38−0.01 0.010.100.150.19 0.090.04
in 0.020.05 0.08 0.06 0.030.080.140.13 0.050.060.420.490.60 0.17 0.070.210.300.25 0.050.09
Ks 0.020.04 0.00−0.02 0.030.090.120.05 −0.04 0.030.430.490.67 0.24 0.070.32
0.480.75 0.430.16
Ky 0.030.06 0.19 0.16 0.030.050.090.31 0.260.040.430.500.38−0.05 0.070.220.360.45 0.230.13
la 0.060.12 0.12 0.07 0.060.160.240.28 0.130.080.440.490.36−0.08 0.050.230.350.52 0.290.12
Ma 0.260.45 0.58 0.32 0.200.270.430.45 0.180.160.460.470.63 0.17 0.020.080.140.11 0.030.06
MD 0.180.35 0.35 0.17 0.170.140.210.22 0.080.070.400.420.45 0.05 0.020.060.110.18 0.120.05
Me 0.050.12 0.02−0.03 0.080.070.160.20 0.130.090.510.580.82 0.31 0.070.160.320.28 0.120.16
Mi 0.030.09 0.15 0.11 0.050.120.170.23 0.110.050.380.420.51 0.14 0.050.260.390.50 0.240.13
Mn 0.070.18 0.12 0.05 0.110.150.260.10 −0.05 0.100.380.440.73 0.34 0.060.140.280.25 0.110.14
Mo 0.050.10 0.15 0.10 0.050.090.150.28 0.190.060.430.490.57 0.14 0.060.200.340.54 0.340.14
Ms 0.0020.004 0.00 0.00 0.000.030.050.03 0.000.020.430.510.61 0.18 0.080.230.390.54 0.310.15
Mt 0.010.02 0.00−0.01 0.010.120.190.39 0.260.070.450.520.67 0.22 0.070.360.560.78 0.420.19
nc 0.010.02 0.01 0.00 0.010.020.030.05 0.030.020.410.490.65 0.24 0.070.200.310.45 0.260.11
nD 0.040.09 0.00−0.04 0.050.110.120.26 0.150.020.550.650.67 0.12 0.100.500.670.45 −0.05 0.17
ne 0.040.11 0.00−0.04 0.070.180.250.00 −0.18 0.070.440.480.97 0.53 0.040.510.660.60 0.090.15
nH 0.040.11 0.05 0.01 0.070.070.150.11 0.040.080.470.530.75 0.28 0.050.130.250.29 0.160.12
nJ 0.250.48 0.49 0.24 0.230.240.380.48 0.230.130.400.400.49 0.08 0.000.040.050.05 0.010.01
nM 0.030.09 0.27 0.23 0.050.180.240.40 0.230.060.350.390.86 0.51 0.050.270.390.00 −0.27 0.12
nV 0.200.31 0.38 0.18 0.110.260.240.05 −0.22 −0.02 0.320.370.71 0.39 0.050.200.300.42 0.220.10
ny 0.490.71 0.79 0.31 0.230.420.550.59 0.170.130.380.360.39 0.01−0.01 0.140.180.18 0.040.04
oH 0.060.12 0.12 0.06 0.060.090.130.21 0.120.050.390.440.52 0.13 0.050.210.30
0.41 0.200.09
(Continued)
346 A. ADKINS ET AL.
Res. density (7+ HU/acre) Connectivity (network density > 23) Retail sufficiency (> .043 jobs/HH)Relative regional accessibility (> .72)
State All
housing RentalLIHTC LIHTC
diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTC LIHTC
diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTC LIHTC
diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTC LIHTC
diff.
(housing) LIHTC
diff.
(rental)
oK 0.020.04 0.00−0.02 0.030.090.140.13 0.040.050.410.470.43 0.02 0.060.260.380.35 0.080.12
oR 0.060.12 0.34 0.28 0.060.170.240.39 0.220.070.410.500.65 0.24 0.090.220.310.56 0.340.09
Pa 0.180.31 0.40 0.22 0.130.230.350.53 0.310.130.420.440.43 0.01 0.020.160.240.32 0.160.08
Ri 0.200.34 0.21 0.01 0.140.310.440.34 0.030.130.420.440.74 0.32 0.010.130.270.26 0.130.14
sc 0.010.02 0.00−0.01 0.010.020.030.24 0.220.010.430.510.85 0.42 0.080.150.240.60 0.450.09
sD 0.010.01 0.02 0.02 0.010.090.120.12 0.040.030.510.610.39−0.12 0.100.410.580.40 0.000.17
tn 0.010.03 0.00−0.01 0.020.030.060.07 0.040.030.390.470.55 0.15 0.080.200.320.36 0.160.11
tX 0.070.16 0.07 0.00 0.090.140.180.10 −0.04 0.050.400.450.61 0.21 0.050.190.280.16 −0.03 0.09
ut 0.050.12 0.00−0.05 0.070.070.090.00 −0.07 0.020.440.530.89 0.45 0.090.310.490.82 0.510.18
Va 0.100.21 0.26 0.15 0.100.110.180.25 0.140.070.410.460.54 0.13 0.050.140.230.23 0.090.09
Vt 0.030.09 0.00−0.03 0.060.010.050.00 −0.01 0.030.560.630.66 0.10 0.070.230.430.54 0.310.20
Wa 0.100.20 0.36 0.26 0.100.180.260.21 0.040.080.350.440.61 0.26 0.090.120.190.16 0.040.07
Wi 0.080.16 0.28 0.20 0.090.140.210.25 0.120.070.380.410.39 0.02 0.040.240.350.43 0.190.11
WV 0.020.05 0.00−0.02 0.030.090.150.06 −0.03 0.050.420.490.82 0.40 0.070.210.330.06 −0.15 0.13
Wy 0.010.02 0.00−0.01 0.010.130.190.00 −0.13 0.060.470.510.77 0.30 0.040.370.530.44 0.070.17
Transit commute share (> 5%) Half mile to rail transit
a
Transportation costs < 20%
State All
housing RentalLIHTCLIHTC diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTCLIHTC diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTCLIHTC diff.
(housing) LIHTC
diff.
(rental)
u .s. 0.240.330.43 0.20 0.100.100.160.24 0.14 0.060.090.17 0.24 0.15 0.07
aK 0.130.220.21 0.08 0.09--- - -0.13 0.23 0.05−0.08 0.11
al 0.030.060.07 0.04 0.02--- - -0.002 0.005 0.00 0.00 0.00
aR 0.030.050.13 0.10 0.020.0050.0100.04 0.03 0.010.000.00 0.00 0.00 0.00
aZ 0.140.230.38 0.24 0.090.020.050.07 0.05 0.020.010.02 0.00−0.01 0.01
ca 0.300.400.43 0.13 0.100.150.200.31 0.17 0.060.170.28 0.34 0.17 0.12
co 0.240.350.45 0.21 0.110.050.080.14 0.09 0.030.080.18 0.25 0.17 0.10
ct 0.310.470.72 0.41 0.160.110.150.25 0.14 0.040.120.28 0.55 0.43 0.16
De 0.200.320.39 0.18 0.120.030.070.11 0.08 0.030.040.09 0.11 0.07 0.06
Fl 0.130.220.37 0.24 0.080.030.040.09 0.06 0.020.010.02 0.01 0.01 0.01
ga 0.140.220.54 0.41 0.080.040.060.31 0.27 0.020.010.02 0.08 0.07 0.01
Hi 0.880.490.33−0.55 −0.39 --- - -0.21 0.31 0.24 0.03 0.11
ia 0.080.160.10 0.01 0.08--- - -0.01 0.02 0.08 0.08 0.01
iD 0.040.040.13 0.09 0.00--- - -0.00 0.00 0.00 0.00 0.00
il 0.450.570.57 0.11 0.120.280.390.41 0.13 0.110.130.25 0.29 0.16 0.12
in 0.080.140.29 0.22 0.060.010.010.00−0.01 0.000.0030.008 0.00 0.00 0.01
Ks 0.030.050.00−0.03 0.02--- - -0.0004 0.0011 0.00 0.00 0.00
Ky 0.090.160.48 0.39 0.070.00 -0.00 0.00 -0.002 0.007 0.05 0.05 0.00
Table 2. (Continued).
(Continued)
HOUSING POLICY DEBATE 347
Transit commute share (> 5%) Half mile to rail transit
a
Transportation costs < 20%
State All
housing RentalLIHTCLIHTC diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTCLIHTC diff.
(housing) LIHTC
diff.
(rental) All
housing RentalLIHTCLIHTC diff.
(housing) LIHTC
diff.
(rental)
la 0.090.170.00−0.09 0.070.030.060.08 0.05 0.030.0010.005 0.00 0.00 0.00
Ma 0.440.560.71 0.27 0.120.290.390.38 0.09 0.100.240.43 0.58 0.34 0.19
MD 0.500.620.59 0.09 0.120.160.240.38 0.23 0.080.280.52 0.60 0.32 0.24
Me 0.040.070.04 0.01 0.030.040.070.11 0.07 0.030.020.04 0.00−0.02 0.03
Mi 0.100.190.26 0.16 0.090.0030.0080.01 0.00 0.010.010.02 0.02 0.01 0.01
Mn 0.260.410.30 0.03 0.150.030.040.01−0.01 0.020.080.21 0.23 0.15 0.13
Mo 0.120.200.43 0.31 0.080.050.080.28 0.23 0.030.010.03 0.09 0.08 0.02
Ms 0.030.030.05 0.02 0.01--- - -0.000 0.000 0.00 0.00 0.00
Mt 0.040.070.19 0.14 0.03--- - -0.00 0.00 0.00 0.00 0.00
nc 0.070.140.19 0.12 0.060.010.010.01 0.00 0.010.0040.009 0.01 0.00 0.00
nD 0.030.050.00−0.03 0.03--- - -0.000 0.000 0.00 0.00 0.00
ne 0.050.090.34 0.29 0.04--- - -0.02 0.05 0.00−0.02 0.03
nH 0.040.080.10 0.06 0.040.020.040.11 0.09 0.020.070.18 0.15 0.08 0.11
nJ 0.560.690.63 0.07 0.130.250.360.14−0.11 0.110.180.38 0.32 0.14 0.20
nM 0.070.090.00−0.07 0.030.040.040.00−0.04 0.000.010.01 0.00−0.01 0.00
nV 0.250.350.68 0.43 0.110.080.110.04−0.04 0.030.010.03 0.00−0.01 0.02
ny 0.650.790.86 0.21 0.150.440.630.69 0.25 0.190.380.62 0.74 0.36 0.24
oH 0.120.210.35 0.23 0.090.020.040.08 0.06 0.020.010.01 0.04 0.03 0.01
oK 0.030.050.52 0.49 0.02--- - -0.001 0.002 0.00 0.00 0.00
oR 0.290.390.51 0.22 0.100.150.220.38 0.23 0.070.020.05 0.22 0.20 0.03
Pa 0.300.410.21−0.09 0.120.180.250.27 0.09 0.070.080.15 0.31 0.23 0.07
Ri 0.200.300.28 0.08 0.100.030.030.00−0.03 0.010.040.08 0.09 0.05 0.04
sc 0.040.070.00−0.04 0.03--- - -0.0004 0.0012 0.00 0.00 0.00
sD 0.030.050.10 0.07 0.02--- - -0.00 0.00 0.00 0.00 0.00
tn 0.060.110.27 0.21 0.050.020.030.09 0.08 0.010.0010.003 0.00 0.00 0.00
tX 0.110.180.22 0.11 0.070.030.050.06 0.03 0.020.050.13 0.01−0.04 0.07
ut 0.150.240.75 0.59 0.090.120.180.74 0.62 0.060.030.08 0.00−0.03 0.05
Va 0.270.360.47 0.20 0.090.060.080.07 0.01 0.020.160.28 0.29 0.13 0.11
Vt 0.080.150.24 0.16 0.080.030.050.02 0.00 0.020.020.05 0.00−0.02 0.03
Wa 0.390.480.61 0.22 0.100.060.100.13 0.07 0.030.060.13 0.08 0.02 0.07
Wi 0.140.240.30 0.16 0.110.010.010.04 0.03 0.000.020.05 0.15 0.13 0.03
WV 0.060.090.07 0.01 0.030.010.010.00−0.01 0.000.0020.002 0.00 0.00 0.00
Wy 0.110.130.10−0.02 0.02--- - -0.002 0.003 0.00 0.00 0.00
Note. Re = Residential; Hu = housing units; liHtc = l ow-income Housing tax credit; HH = households. aFor states with no rail transit, this column is not included in averages. Table 2. (Continued).
348 A. ADKINS ET AL.
which is the average of the proportion of LIHTC units in CBG meeting each LE criteria. In Model 2, the
goal was to test the influence of the same five independent variables on LIHTC LE relative to all hous-
ing, an indicator of out-performance in the LE of the LIHTC units relative to the LE of all housing in the
state. In Model 3, the goal was the same as for Model 2 except we used the LE of LIHTC units relative to
rental housing, rather than all housing, as the dependent variable. For clarity, only the first two models
are presented. Model 3 was inconclusive, with nonsignificant target variables and nonsignificant or
unexpected signs for control variables. We think this may be because either the smaller range of the
dependent variable or additional nuances related to the geographic distribution of rental housing
versus all housing.
Findings
Location Efficiency of LIHTC in the United States
For most LE variables, LIHTC was about twice as likely to be in a CBG meeting the LE standard than
housing generally, with 30% of LIHTC units and 14% of all housing units meeting our density criteria and,
respectively, 43% versus 24% for our transit service criteria; 28% versus 17% for network density; 52%
versus 39% for retail sufficiency; 32% versus 17% for regional access; 24% versus 9% for transportation
affordability; and 24% versus 10% for rail transit proximity. The average proportion across all seven LE
variables was 33% and the average difference between LIHTC and all housing was 15%, whereas the
average difference between LIHTC and rental housing was 7%.
Examining the distribution of the share of each housing type that met different numbers of LE criteria,
similar patterns emerge (see Figure 1). Thirty-five percent of LIHTC units built during the study period
were in CBGs meeting three or more LE criteria, versus 28% of rental units and 16% of all units. Notably,
rental units were slightly more likely to be in CBGs meeting five or more and six or more LE criteria
than LIHTC units. We are not able to say why this was the case, but it is possibly due to the preponder -
ance of highly location-efficient CBGs in New York, giving that market outsized influence at the high
end of the LE scale. Of all CBGs meeting five or more LE criteria, 40% are in New York, whereas only
10% of CBGs meeting three LE criteria are concentrated in New York (vs. a population that is roughly
7% of the U.S. population).
State-by-State Comparisons
We observed considerable variation in both absolute and relative LIHTC LE achievement at the state
level (see Table 3). As anticipated, many denser states were in the top quartile of absolute LIHTC LE. In
New York, for example, an average of 60% of the LIHTC units met each of the LE criteria. Similarly, the
denser states of Massachusetts (.49), Connecticut (.46), Maryland (.40), California (.37), New Jersey (.37),
and Pennsylvania (.35) were all in the top quartile of absolute LIHTC LE. More surprisingly, however,
several lower density western states were also in the top quartile of absolute LIHTC LE: Utah (.46), Oregon
(.44), Colorado (.33), and Nevada (.32). Looking at LIHTC LE relative to all housing resulted in a different list of high-performing states (see
Table 3). The states with the highest relative LIHTC LE were Utah (+.29), Oregon (+.25), Connecticut
(+.28), Missouri (+.20), Massachusetts (+.20), and New York (+.19). These are the states that have a higher
degree of LE within their LIHTC portfolios than would be expected based on the overall statewide LE
for all housing units.
We observed a narrower range of LIHTC LE relative to rental housing (see Table 3). Like absolute
LIHTC LE, this list was dominated by denser states, with New York (+.14), Massachusetts (+.12), New
Jersey (+.11), Connecticut (+.11), Maryland (+.11), Minnesota (+.10), and Illinois (+.10) all exceeding
0.10 LIHTC LE relative to rental housing.
HOUSING POLICY DEBATE 349
Explanations for State Variations
To begin explaining why some states achieved a higher concentration of LIHTC in location-efficient
places, we specified two OLS regression models to test the influence policy, market, and other factors
on states’ absolute and relative LIHTC LE. The first regression model predicted absolute LIHTC LE, and the
second predicted LIHTC LE relative to all housing. In both models, Hawaii was a significant outlier. This
Table 3. location efficiency summary statistics and dependent variables.
Note. le = location efficiency; liHtc = l ow-income Housing tax credit.
States Average across LE criteria
Calculated from averages
Absolute LIHTC LE Rental LEAll housing LE Relative LIHTC LE
(vs. all housing) Relative LIHTC LE
(vs. rental housing)
all states 0.330.260.19 0.15 0.07
aK 0.270.250.16 0.11 0.09
al 0.180.170.12 0.06 0.05
aR 0.180.150.11 0.07 0.04
aZ 0.250.190.13 0.12 0.06
ca 0.370.320.24 0.13 0.08
co 0.330.260.18 0.15 0.08
ct 0.460.290.18 0.28 0.11
De 0.280.210.15 0.13 0.06
Fl 0.230.200.15 0.09 0.05
ga 0.290.170.12 0.17 0.05
Hi 0.250.340.34 −0.09 0.00
ia 0.250.240.17 0.08 0.07
iD 0.210.190.15 0.06 0.05
il 0.350.350.25 0.10 0.10
in 0.190.160.12 0.08 0.05
Ks 0.240.200.15 0.10 0.05
Ky 0.260.200.12 0.15 0.06
la 0.200.200.15 0.05 0.06
Ma 0.490.410.29 0.20 0.12
MD 0.400.350.25 0.15 0.11
Me 0.210.200.13 0.08 0.07
Mi 0.240.180.13 0.11 0.05
Mn 0.250.260.16 0.09 0.10
Mo 0.330.200.14 0.20 0.06
Ms 0.210.160.12 0.09 0.04
Mt 0.340.230.17 0.17 0.06
nc 0.200.140.10 0.09 0.04
nD 0.230.270.20 0.03 0.06
ne 0.320.270.21 0.11 0.07
nH 0.220.190.12 0.10 0.07
nJ 0.370.390.27 0.10 0.12
nM 0.220.180.14 0.08 0.04
nV 0.320.240.19 0.14 0.06
ny 0.600.550.41 0.19 0.14
oH 0.250.180.13 0.12 0.05
oK 0.240.180.13 0.10 0.05
oR 0.440.260.19 0.25 0.07
Pa 0.350.310.22 0.13 0.09
Ri 0.270.270.19 0.08 0.08
sc 0.280.140.11 0.17 0.04
sD 0.170.230.17 0.00 0.06
tn 0.190.150.10 0.09 0.04
tX 0.180.200.14 0.03 0.06
ut 0.460.250.17 0.29 0.08
Va 0.300.260.18 0.12 0.08
Vt 0.210.210.14 0.07 0.07
Wa 0.310.260.18 0.13 0.08
Wi 0.260.210.14 0.12 0.06
WV 0.150.160.12 0.03 0.05
Wy 0.220.230.18 0.03 0.05
350 A. ADKINS ET AL.
is not surprising considering that Hawaii has the highest share of LIHTC units developed by nonprofits
and is the only state where LIHTC is less location-efficient than that of all housing. Removing Hawaii
increased model fit and effect sizes. Versions of the model with and without Hawaii are displayed side
by side. The interpretation of the results below is of the models without Hawaii.
Model 1 (see Table 4), predicting absolute LIHTC LE, indicated that after controlling for housing_
LE, QCT_LE, and LE_rent_premium, nonprofit_share had the expected positive association (B = .03,
p. = .02). QAP_services had a positive, but only marginally significant, association (B = .03, p. = .09). This
means that a 1-percentage-point increase in the share of LIHTC in a state developed by nonprofits is
associated with a .03 increase in absolute LIHTC LE. The controls were all significant; however, the sign
for LE_rent_premium was reversed relative to our literature-based expectations.
Model 2 (see Table 5) predicted LIHTC LE relative to all housing using the same predictor variables as
Model 1 (see Table 5). As expected, nonprofit_share had a significant positive association on LIHTC LE
relative to all housing (B = .08, p . = .02) while QAP_services had a positive, but only marginally significant,
association (B = .03, p. = .06). Here again, the controls were significant and the sign for LE_rent_premium
was positive, rather than negative as we expected. In addition, the sign for housing_LE was negative,
reversing the relationship found in Model 1. This makes sense, however, because it would be more
difficult for LIHTC to outperform all housing in terms of LE in states where all housing was already more
commonly found in location-efficient block groups.
The models indicated that a 1-percentage-point increase in the share of LIHTC in a state developed
by nonprofits was associated with a .03 increase in absolute LIHTC LE and a .08 increase in relative LIHTC
LE. These are small effects, but not inconsequential given that the highest relative LIHTC LE score was
.29. Likewise, having a QAP that prioritized proximity to services was associated with a .03 increase in
both absolute and relative LIHTC LE.Although these associations are all quite small, they do suggest a role for state housing
policy and nonprofit developers in helping to boost a state’s LIHTC LE portfolios and outperform
underlying LE.
Table 4. ols regression model predicting absolute l ow-income Housing tax credit (liHtc ) location efficiency (with and without
Hawaii).
Note. Qct = qualified census tract; le = location efficiency; Q aP = qualified allocation plan.
* = p < .10; ** = p < .05.
Model with all states Model without Hawaii
B T B T
Qct_le 0.132.53** 0.091.94*
le_rent_premium 0.122.72** 0.092.19**
Housing_le 0.103.29** 0.144.71**
Q aP_services 0.021.35 0.032.19*
nonprofit_share 0.030.86 0.032.08**
R
2 .61 .70
Table 5. ols regression model predicting l ow-income Housing tax credit (liHtc ) location efficiency relative to all housing (with
and without Hawaii).
Note. Qct = qualified census tract; le = location efficiency; Q aP = qualified allocation plan.
* = p < .10; ** = p < .05.
Model with all statesModel without Hawaii
B T B T
Qct_le 0.122.60** 0.102.14**
le_rent_premium 0.112.79** 0.12.36**
Housing_le −0.062.06**−0.04−1.34
Q aP_services 0.031.76* 0.031.93*
nonprofit_share 0.061.91* 0.082.50**
R
2 .32 .32
HOUSING POLICY DEBATE 351
Discussion
Nationally and in nearly all states, LIHTC is more concentrated in location-efficient CBGs than other
housing is, yet barely one third of LIHTC units built between 2007 and 2011 meet three or more of
our seven LE criteria. This suggests that there is room to increase the LE of LIHTC housing. Given the
important connections between affordability and LE, it follows that increased LE within the LIHTC port -
folio is a reasonable policy goal at both the state and federal levels. However, we did not endeavor to
provide a specific target for the proportion of LIHTC that should meet our LE standard, and suggest that
empirically establishing such targets would be a useful exercise for state housing authorities and HUD.
Our findings also suggest that state QAPs that prioritize LIHTC projects with proximity to necessary
services are associated with higher absolute and relative LIHTC LE. Consistent with previous findings,
however, the impact is quite small when compared with the role of the underlying built environment.
This suggests that policy and nonprofits can help to increase the LE of LIHTC, but that it is largely driven
by market and urban form factors. From a policy perspective, QAPs and nonprofit developers can help,
but meaningful change in the absolute level of LE LIHTC will require broader efforts to increase overall
LE through land use, housing, and transportation policy. Taken together, these findings are positive news as they suggest that state housing authorities,
finance agencies, and developers have begun to leverage existing attributes of their urban form to
create more location-efficient affordable housing within their jurisdictions. However, there is still sig-
nificant opportunity for state policymakers to continue to refine their policies and programs to create
more vibrant and livable places for households across the income spectrum. As more research emerges
on this topic, affordable housing policymakers and developers should continue to work together to
iteratively develop LIHTC credit distribution systems to recognize the potential benefits of location-
efficient housing locations. An important caveat to our findings is that there are reasons to be wary of policy objectives that
prioritize LE without considering other characteristics of an area and possible tradeoffs of locational
decisions. Some possible tradeoffs that have been identified in the affordable housing literature are
access to quality schools (Makarewicz, 2013), neighborhood safety (Katz et al., 2000), concentrated
poverty (Goetz, 2003), racial segregation (Horn & O’Regan, 2011), and environmental quality (Katz et
al., 2000). Indicative of these concerns, a follow-up analysis of our data showed that in most states LE
CBGs were more likely than non-LE CBGs to have concentrations of poverty higher than 40%. Policy
and planning approaches that consider LE but do not account for these potential tradeoffs could
potentially make low-income households worse off. But, as Bernstein (2011) points out with an example
from Chicago, there are many areas lacking affordable housing that have both high opportunity and
high LE. By our own calculations using our measure of LE, about 16% of CBGs in the United States are
both location efficient and have poverty rates lower than 40%. We therefore see value in including LE
as one among several allocation criteria.
Limitations and Future Research Needs
There are several limitations that should be considered along with our findings and should help to guide
future inquiry on this topic. First, though we examined the literature, searching for existing cut points
on LE metrics, there is little empirical validation for the thresholds we have used. Second, as previously
mentioned, we recognize limitations inherent in using ACS- and LEHD-derived data at the CBG level.
But we feel that we have structured our variable creation and analysis in a manner that adequately
mitigates risk to our findings. Third, our binary coding system for analyzing state QAPs for emphasis
on LE factors greatly diminished our ability to assess important nuances in policy and practice. Fourth,
the study is cross-sectional and not longitudinal, so it was impossible to determine a causal effect of
policy on LE LIHTC development. Fourth, while analysis at the state level has significant benefits as a
foundation for future analysis, it is coarse and tends to mask finer-grained variation in the findings.
For example, it is possible that LIHTC LE in some states is largely driven by a single metro area. Each of
these limitations, however, marks an opportunity for future research.
352 A. ADKINS ET AL.
We are confident that our LE criteria adequately capture the overall LE of LIHTC projects, although
the lack of national standards for LE targets may hinder efforts to establish goals for LE within the LIHTC
program and for affordable housing generally. We hope that our efforts to establish LE thresholds from
existing data sets will be explored by other researchers as a step toward finding consensus on how LE
can be operationalized by housing policymakers and advocates. As we have acknowledged, however,
there are shortcomings of the data we have chosen to use. Several SLD and LAI variables are derived
from ACS data, which, given sampling changes over the last decade, can be an imperfect accounting of
populations in small geographies. However, as previously stated, our approach of using multiple vari-
ables to categorize CBGs into broad categories of location efficient and not location efficient increases
our confidence in the reliability of our analysis.
Given the limitations noted above and the data available, there several key opportunities for research
on this topic: validation of LE indicators, particularly those derived from large national data sets that
are prone to error at small geographic scales; a pretest/posttest research design to examine whether
the adoption of allocation criteria within a QAP changed the location characteristics of LIHTC; an anal-
ysis of the LIHTC development dynamics at the CBG scale to further test whether LE is emerging as an
independent factor that is attracting developers; an analysis of LE at the metro area level to expand
investigation of policy and market effects at a smaller scale; and a more explicit exploration of tradeoffs
between LE and other measures of opportunity.
Conclusion
In this article we have shown that, as expected, LIHTC in the United States and in most states tends
to be more concentrated in location-efficient places than other housing is. However, most LIHTC built
between 2007 and 2011 is not in location efficient CBGs, with just 36% of LIHTC units having been built
in places meeting three or more of the LE criteria. States vary widely in the LE of LIHTC units. And it
appears that factors of market, poverty, underlying urban form, state policy, and the nonprofit housing
sector all play a role in predicting LIHTC LE. More work is needed to refine LE thresholds and set state -
wide targets for LIHTC LE, and such efforts should include careful consideration of possible tradeoffs
with other characteristics of place that might counteract improvements in well-being. We hope that
this article will be a first step toward that larger goal.
Disclosure Statement
No potential conflict of interest was reported by the authors.
Notes on Contributors
Arlie Adkins, PhD, is an assistant professor in the College of Architecture, Planning, and Landscape Architecture at the University of Arizona. His research focuses on understanding relationships between active transportation, housing, equity, and health.
Andrew Sanderford, PhD, is an assistant professor in the College of Architecture, Planning, and Landscape Architecture
at the University of Arizona. His research focuses on housing, the property development process, various aspects of real
estate finance, and the diffusion of innovation across property markets.
Gary Pivo, PhD, is a professor in the College of Architecture, Planning, and Landscape Architecture at the University of
Arizona. His research focuses on responsible property investing, urban form, and sustainable cities.
References
Adkins, A. (2014). Determinants of recent mover non-work travel mode choice. Portland State Dissertations and Theses. Paper 1919. Retrieved from http://pdxscholar.library.pdx.edu/open_access_etds/1919
Bartholomew, K., & Ewing, R. (2011). Hedonic price effects of pedestrian- and transit-oriented development. Journal of Planning Literature, 26, 18–34.
HOUSING POLICY DEBATE 353
Bernstein, S. (2011). Transportation and fair housing part 2: Consider transportation cost to make fair housing practical. Shelterforce, Spring 2011. Retrieved from http://www.shelterforce.org/article/2175/transportation_and_fair_housing_part_2_consider_transportation_cost_to_make/Berrigan, D., Carroll, D., Fulton, J., Galuska, D., Brown, D., Dorn, J. … Paul, P. ( 2012). Vital signs: walking among adults - United States, 2005 and 2010. Morbidity and Mortality Weekly Report, 61, 595–601.
Blumenberg, E., & Pierce, G. (2014). A driving factor in mobility? Transportation’s role in connecting subsidized housing
and employment outcomes in the moving to opportunity (MTO) program. Journal of the American Planning Association, 80, 52–66.
Bosch, P., & Metz, B. (2011). Options for mitigating climate change results of working group III of the fourth assessment report of the IPCC. Coping with Global Environmental Change, Disasters and Security, 5, 1283–1291.
Bratt, R. G. (2008). Nonprofit and for-profit developers of subsidized rental housing: Comparative attributes and collaborative opportunities. Housing Policy Debate, 19, 323–365. doi:10.1080/10511482.2008.9521638.
Calthorpe, P. (1993). The next American metropolis: Ecology, community, and the American dream. New York, NY: Princeton
Architectural Press.
Cao, X., Handy, S. L., & Mokhtarian, P. L. (2006). The influences of the built environment and residential self-selection on pedestrian behavior: Evidence from Austin, TX. Transportation, 33, 1–20.
Center for Neighborhood Technology. (2010, March). Penny wise, pound fuelish: New measures of housing + transportation affordability. CNT. Retrieved from http://www.cnt.org/publications/penny-wise-pound-fuelish-new-measures-of-
housing-and-transportation-affordability
Cervero, R. (2004). Transit-oriented development in the United States: Experiences, challenges, and prospects (Transit Cooperative Research Report 102). Washington, DC: Transportation Research Board.
Chapman, L. (2007). Transport and climate change: A review. Journal of Transport Geography, 15, 354–367.
Climaco, C., Finkel, M., Nolden, S., & Vandawalker, M. (2006). Updating the low income housing tax credit (LIHTC) database:
Projects placed in service through 2003. Washington, DC: Abt Associates for the U.S. Department of Housing and Urban
Development.
Cohen, J., & Cohen, P. (2003). Applied multiple regression/correlation analysis for the behavioral sciences. New York, NY:
Psychology Press.
Cortright, J. (2009). Walking the walk: How walkability raises housing values in US cities. Prepared for CEOs for Cities.
Dawkins, C. (2013). The spatial pattern of low income housing tax credit properties: Implications for fair housing and poverty deconcentration policies. Journal of the American Planning Association, 79, 222–234.
Deng, L. (2011). The external neighborhood effects of low-income housing tax credit projects built by three sectors. Journal of Urban Affairs, 33, 143–166.Dill, J. (2004). Estimating emissions reductions from accelerated vehicle retirement programs. Transportation Research Part D: Transport and Environment, 9, 87–106.
Downs, A. (2005). Still stuck in traffic: Coping with peak-hour traffic congestion. Washington, DC: Brookings Institution Press.
Du Toit, L., Cerin, E., Leslie, E., & Owen, N. (2007). Does walking in the neighbourhood enhance local sociability? Urban Studies, 44, 1677–1695.
Duany, A., Plater-Zyberk, E., and Speck, J. (2010). Suburban nation: The rise of sprawl and the decline of the American dream.
New York, NY: North Point Press.
Ellen, I. G., Horn, K., Kuai, Y., Pazuniak, R., & Williams, M. D. (2015). Effect of QAP incentives on the location of LIHTC properties.
Washington, DC: U.S. Department of Housing and Urban Development.
Ellen, I. G., & Voicu, I. (2006). Nonprofit housing and neighborhood spillovers. Journal of Policy Analysis and Management, 25, 31–52.
Ewing, R., & Cervero, R. (2001). Travel and the built environment: A synthesis. Journal of the Transportation Research Board, 87–114.
Ewing, R., & Cervero, R. (2010). Travel and the built environment: A meta-analysis. Journal of the American Planning Association, 76, 265–294.
Fisher, L. M., Pollakowski, H. O., & Zabel, J. (2009). Amenity-based housing affordability indexes. Real Estate Economics, 37,
705–746.
Frank, L. D., & Engelke, P. (2005). Multiple impacts of the built environment on public health: Walkable places and the exposure to air pollution. International Regional Science Review, 28, 193–216.
Frank, L. D., Stone, B., & Bachman, W. (2000). Linking land use with household vehicle emissions in the central Puget
Sound: Methodological framework and findings. Transportation Research Part D: Transport and Environment, 5, 173–196. doi:10.1016/S1361-9209(99)00032-2.Galster, G. C. (2013). U.S. assisted housing programs and poverty deconcentration: A critical geographic review. In D. Manley, M. van Ham, N. Bailey, L. Simpson, & D. Maclennan (Eds.), Neighbourhood effects or neighbourhood based problems? (pp.
215–249). Netherlands: Springer.
Gay, C. (2015). A room for one’s own: The partisan allocation of affordable housing (unpublished working paper).
Geels, F. W. (2012). A socio-technical analysis of low-carbon transitions: Introducing the multi-level perspective into transport studies. Journal of Transport Geography, 24, 471–482.
Goetz, E. G. (2003). Clearing the way: Deconcentrating the poor in urban America. Washington, DC: The Urban Insitute.
354 A. ADKINS ET AL.
Goetz, E. G. (2015). The fair housing tightrope in the Obama administration: Balancing competing policy objectives of
fair housing and locational efficiency in assisted housing. Journal of Urban Affairs, 37, 53–56. doi:10.1111/juaf.12162.
Graham, M. R., & Ong, P. (2007). Social, economic, spatial and commuting patterns of informal jobholders (Technical Paper No. TP-2007-02). Retrieved from http://lehd.dsd.census.gov/led/library/techpapers/tp-2007-02.pdf
Guerra, E., Cervero, R., & Tischler, D. (2011). The half-mile circle: Does it best represent transit station catchments? (Working Paper UCB-ITS-VWP-2011-5). Berkeley, CA: University of California Transportation Center.
Haas, P., Makarewicz, C., Benedict, A., & Bernstein, S. (2008). Estimating transportation costs by characteristics of neighborhood and household. Transportation Research Record, 62–70.
Haas, P. M., Makarewicz, C., Benedict, A., Sanchez, T. W., & Dawkins, C. J. (2006). Housing & transportation cost trade-offs and
burdens of working households in 28 metros. Chicago, IL: Center for Neighborhood Technology.
Handy, S., Cao, X., & Mokhtarian, P. (2005). Correlation or causality between the built environment and travel behavior? Evidence from Northern California. Transportation Research Part D: Transport and Environment, 10, 427–444.
Horn, K. M., & O’Regan, K. M. (2011). The low income housing tax credit and racial segregation. Housing Policy Debate, 21,
443–473.
Ihlanfeldt, K. R., & Sjoquist, D. L. (1998). The spatial mismatch hypothesis: A review of recent studies and their implications for welfare reform. Housing Policy Debate, 9, 849–892.
Kain, J. F. (1968). Housing segregation, negro employment, and metropolitan decentralization. The Quarterly Journal of Economics, 175–197.
Kain, J. F. (1992). The spatial mismatch hypothesis: Three decades later. Housing Policy Debate, 3, 371–460.
Katz, L. F., Kling, J. R., & Liebman, J. B. (2000). Moving to opportunity in Boston: Early results of a randomized mobility experiment (Working Paper No. 7973). Cambridge, MA: National Bureau of Economic Research. Retrieved from http://
www.nber.org/papers/w7973
Klein, N. J., & Smart, M. J. (2015). Car today, gone tomorrow: The ephemeral car in low-income, immigrant and minority families. Transportation, 1–16.
Korda, R. J., Paige, E., Yiengprugsawan, V., Latz, I., & Friel, S. (2014). Income-related inequalities in chronic conditions, physical functioning and psychological distress among older people in Australia: Cross-sectional findings from the 45 and up
study. BMC Public Health, 14, 1438.
Lang, B. J. (2012). Location incentives in the low-income housing tax credit: Are qualified census tracts necessary? Journal of Housing Economics, 21, 142–150.
Lee, I.-M., & Buchner, D. M. (2008). The importance of walking to public health. Medicine & Science in Sports & Exercise, 40,
S512–S518.
Leinberger, C. B., & Alfonzo, M. (2012). Walk this way: The economic promise of walkable places in metropolitan Washington,
DC. Washington, DC: The Brookings Institution.
Leyden, K. M. (2003). Social capital and the built environment: The importance of walkable neighborhoods. American Journal of Public Health, 93, 1546–1551.
Litman, T. (2003). Economic value of walkability. Transportation Research Record, 1828, 3–11.
Makarewicz, C. (2013). Vouchers, magnet schools, charter schools, and options. Transportation Research Record, 2327, 1–8.
doi:10.3141/2327-01.
McClure, K. (2006). The low‐income housing tax credit program goes mainstream and moves to the suburbs. Housing Policy Debate, 17, 419–446.
McClure, K. (2008). Deconcentrating poverty with housing programs. Journal of the American Planning Association, 74, 90–99.Nedwick, T., & Burnett, K. (2015). How can the LIHTC program most effectively be used to provide affordable rental housing near transit? Cityscape, 17, 113–137.
Newman, S. J., & Schnare, A. B. (1997). “… And a suitable living environment”: The failure of housing programs to deliver on neighborhood quality. Housing Policy Debate, 8, 703–741.
Pivo, G. (2014). Walk score: The significance of 8 and 80 for mortgage default risk in multifamily properties. Journal of Sustainable Real Estate, 6, 187–210.
Portney, K. (2005). Civic engagement and sustainable cities in the United States. Public Administration Review, 65, 579–591. Ramsey, K., & Bell, A. (2014). The smart location database: A nationwide data resource characterizing the built environment and destination accessibility at the neighborhood fiscalement and destination accessibility at the neighborhood scale. Cityscape, 16, 145–162.
Rogers, S. H., Halstead, J. M., Gardner, K. H., & Carlson, C. H. (2011). Examining walkability and social capital as indicators of quality of life at the municipal and neighborhood scales. Applied Research in Quality of Life, 6, 201–213.
Rohe, W. M., & Freeman, L. (2001). Assisted housing and residential segregation: The role of race and ethnicity in the siting of assisted housing developments. Journal of the American Planning Association, 67, 279–292.
Sanchez, T.W., Shen, Q., & Peng, Z.-R. (2004). Transit mobility, jobs access and low-income labour participation in us metropolitan areas. Urban Studies, 41, 1313–1331.
Squires, G. D., & Kubrin, C. E. (2005). Privileged places: Race, uneven development and the geography of opportunity in urban America. Urban Studies, 42, 47–68.
Talen, E., & Koschinsky, J. (2014). The neighborhood quality of subsidized housing. Journal of the American Planning Association, 80, 67–82.
HOUSING POLICY DEBATE 355
United States Census Bureau. (2015). Means of Transportation to Work, 2010–2014 American Community Survey 5-year
estimates. Retrieved from: http://www.socialexplorer.com/tables/ACS2014_5yr/R11261801
United States Department of Housing and Urban Development (HUD). (n.d.). Location affordability portal. Retrieved from http://www.locationaffordability.info .
United States Environmental Protection Agency (EPA). (2011). Location efficiency and housing type. Retrieved from http://
www.epa.gov/sites/production/files/2014-03/documents/location_efficiency_btu.pdf .
Welch, T. F. (2013). Equity in transport: The distribution of transit access and connectivity among affordable housing units. Transport Policy, 30, 283–293.
Wood, L., Frank, L. D., & Giles-Corti, B. (2010). Sense of community and its relationship with walking and neighborhood design. Social Science & Medicine, 70, 1381–1390.
World Cancer Research Fund/American Institute for Cancer Research. (2007). Physical activity, and the prevention of cancer:
A global perspective. Washington, DC: Author.