THE EFFECT OF THE LOW INCOME HOUSING TAX CREDIT ON NEIGHBORHOOD CHANGE

Transcription

1 THE EFFECT OF THE LOW INCOME HOUSING TAX CREDIT ON NEIGHBORHOOD CHANGE A Thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy in Public Policy By Melanie A. Jaskolka, B.A Washington, DC April 10, 2015

2 Copyright 2015 by Melanie A. Jaskolka All Rights Reserved ii

3 THE EFFECT OF THE LOW INCOME HOUSING TAX CREDIT ON NEIGHBORHOOD CHANGE Melanie A. Jaskolka, B.A Thesis Advisor: Donna Ruane Morrison, Ph.D. ABSTRACT The Low-Income Housing Tax Credit (LIHTC) program awards tax credit to developers who allocate at least a certain percentage of their development to low-income units. The program creates affordable housing and encouraging investment and development. Most projects funded with the tax credit are located in primarily minority, low-income neighborhoods. An important question is whether in addition to adding to the housing stock, these projects have spillover effects on the economic standing of the neighborhood. There are at least two possible answers. If the newly constructed or renovated dwellings make the location more attractive, the presence of LIHTC developments could spur neighborhood revitalization and perhaps eventually, gentrification. Alternatively, the presence of affordable housing in a neighborhood may dissuade potential newcomers and investors and eventually drive out middle class residents concerned about falling housing values. Using OLS and looking at the effect of the tax credit on median household income over a thirty year period, I find no relationship between the number of LIHTC projects and household income at any point throughout the period. My results do reveal, however, that the percentage of affordable units in LIHTC-funded projects is negatively correlated with median income. I argue that this relationship may be due to developers differing goals: some developers build projects that are either primarily or entirely targeted towards lowincome residents, whereas other developers put in the minimum number of affordable units iii

4 simply as a means to qualify for the tax credit. Future research might examine projects built by different types of developers, such as non-profit versus for-profit, and the differences between them. iv

5 Many thanks to my thesis advisor, Donna Morrison. Her patience, support, guidance, and advice were invaluable. Many thanks, Melanie A. Jaskolka v

6 TABLE OF CONTENTS Introduction... 1 Policy Background... 4 Literature Review Conceptual Framework Data and Methods Results Discussion and Conclusion Appendix References vi

7 LIST OF TABLES Table 1: Descriptive Characteristics for Census Tracts in Table 2: Descriptive Characteristics for Census Tracts in Table 3: Descriptive Characteristics for Census Tracts in Table 4: Descriptive Characteristics by Census Tract Based on Number of LIHTC Projects by Tract for Table 5: Descriptive Characteristics by Census Tract Based on Number of LIHTC Projects by Tract for Table 6: Descriptive Characteristics by Census Tract Based on Number of LIHTC Projects by Tract for Table 7: OLS Regression Coefficients for Models Predicting Median Household Income in Central Cities with Low-Income Housing Tax Credit Projects (Tract Level), by Period Put Into Service and Duration of Effect Table 8: OLS Regression Coefficients for Models Predicting Median Household Income in Central Cities Both With and Without Low-Income Housing Tax Credit Projects (Tract Level), by Period Put Into Service and Duration of Effect vii

8 INTRODUCTION The Low-Income Housing Tax Credit (LIHTC) program, established in 1987, is a popular program that enjoys widespread bipartisan support. The purpose of the program is to increase the construction and rehabilitation of affordable rental housing for low- and moderateincome families by providing a financial incentive to private developers. The primary eligibility requirement for receiving federal housing tax credits is that a certain percentage of units are set aside as affordable housing. The program serves several goals simultaneously: it provides affordable housing for low-income people, it saves the government money by shifting the provision of housing to the private market, and it provides economic incentives for real estate developers and investors (National Housing Law Project, 2014). LIHTC is currently the largest creator of affordable housing. The credit can be used in any region or community. Special consideration is given to applications for developments in one of two areas: Qualified Census Tracts, which are neighborhoods with high poverty levels that are depressed economically, and Difficult Development Areas, which are neighborhoods where building and land costs are high a (Novogradac Affordable Housing Resource Center, 2015). LIHTC is currently the largest creator of affordable housing in the United States. In 2010, one half of new multi-family units constructed were located in LIHTC developments. LIHTC is the latest iteration in affordable housing policy. As public housing has fallen out of favor, privately-provided rental housing has replaced it as the preferable solution to providing affordable housing to low-income families (Schwartz, 2015, p. 135). a Developments in these neighborhoods also receive a higher tax credit 130% of what the tax credit would be normally (Nolden, Buron, Heintz, and Stewart, 2000). 1

9 Encouraging the growth of mixed-income developments is seen as one of the best ways to alleviate historical racial segregation and concentrated poverty in urban areas. Along with increasing socioeconomic diversity, mixed-income housing can also encourage the economic revitalization of depressed and disinvested neighborhoods (Urban Land Institute, 2003, p. 6). In recent years, many American cities have been experiencing economic growth and development and an influx of new residents, but at the same time continue to suffer from racial segregation and concentrated poverty. Numerous researchers have demonstrated the negative effects of socioeconomic segregation, including lower education level, lower incomes, higher poverty rates, and higher crime rates (Massey et. al., 2009, p. 87). Programs like LIHTC provide the potential to alleviate socioeconomic segregation by encouraging mixed-income developments in neighborhoods where investors may not ordinarily choose to build. But on the other hand, LIHTC projects could have a more harmful outcome if they encourage gentrification. New, affordable apartments could attract new residents who usher in a wave of corresponding changes to the detriment of the original residents the program was intended to benefit. Journalist Megan McArdle, who often writes about gentrification in Washington, DC, talks about such cases with affordable housing units at places like City Vista going to college grads in entry level jobs rather than anyone who could legitimately be considered in need of a housing subsidy (McArdle, 2010). Is McArdle s concern well founded? The purpose of the present study is to examine the impact of LIHTC developments on the neighborhoods within which they are built. Specifically, the study addresses two questions: does the presence of LIHTC-funded properties lead to a 2

10 neighborhood becoming gentrified, or does it have the opposite effect, discouraging market-rate investment and enforcing concentrated poverty? I examine the relationship between LIHTC-funded projects and change in neighborhood median income between 1990 and LIHTC s role in providing affordable housing to lowincome residents is clear, but analyzing whether it has any role in neighborhood socioeconomic change can provide direction as to its role and usefulness in promoting mixed-income housing and economic development. As many previously depressed urban areas continue to grow and change, promoting the development of heterogeneous neighborhoods, encouraging investment and development, and providing affordable housing so that residents of many income levels, both new and old, can thrive will be vital objectives to promoting successful cities. 3

11 POLICY BACKGROUND The Low-Income Housing Tax Credit, created by the Tax Reform Act of 1986, is currently the largest government subsidy for the creation of low-income housing, While its predecessors were administered by the Department of Housing and Urban Development, the Low-Income Housing Tax Credit program falls under the Internal Revenue Service b (Schwartz, 2015, p. 135). Program History The Low-Income Housing Tax Credit, currently the largest government subsidy for the creation of low-income housing, was created by the Tax Reform Act of Unlike previous affordable housing programs administered by the Department of Housing and Urban Development, the Low-Income Housing Tax Credit program is administered by the Internal Revenue Service c (Schwartz, 2015, p. 135). The program was created in part as a reaction to previous housing policy. Affordable housing policy in the United States began in 1937 with the creation of publically constructed housing (what most people think of as public housing) (Schwartz, 2015, p. 7-8). High-rise style housing complexes, owned and operated by government, such as Cabrini Green in Chicago were constructed (Vale, 2012). b However, HUD (along with the Department of Justice) is responsible for enforcing the Act. HUD also conducts assessments and monitors the program (Low-Income Housing Tax Credit Memorandum, 2000). c However, HUD (along with the Department of Justice) is responsible for enforcing the Act. HUD also conducts assessments and monitors the program (Low-Income Housing Tax Credit Memorandum, 2000). 4

12 By the late 1960s and 1970s, many of these of these projects had come to be seen as failures for a number of reasons. Public housing was originally created for poor, working class families. Over time, public housing residents became increasingly poor as policies prioritized extremely low-income residents. By the 1990s, the median household income of public housing residents was less than 20% of the national median income (Schwartz, 2015, p ). Public housing was almost exclusively located in very poor, racially segregated neighborhoods (Schwartz, 2015, p ). The physical design of public housing was usually very simple and basic, and projects were often isolated from the surrounding neighborhood, making public housing easily recognizable. Public housing was infamous for having long, dark, isolated hallways and other interior public spaces, and exterior courtyards isolated from the surrounding buildings, leading, to crime, vandalism, and the absence of a community atmosphere. Construction and materials were usually cheap and poor quality, which meant that public housing complexes became run-down and in need of repairs and maintenance fairly quickly (Schwartz, 2015, p ). Furthermore, because tenants paid extremely low rents (whether through their own income, subsidies, or some combination), rents were not enough to pay for building operation and maintenance, and operating subsidies from HUD were not enough to make up the difference. This led to a constant backlog in maintenance, repair, and modernization needs (Schwartz, 2015, p ). While many of the high-rise public housing complexes have since been demolished and in many cases have been replaced by low-rise garden-style apartment complexes, there has been a definite shift away from the construction of new publically-funded affordable housing (Schwartz, 2015, ). The supply of public housing has shrunk due to the demolition and 5

13 redevelopment of many public housing complexes, and almost no new construction has been built since the mid-1980s (Schwartz, 2015, p ). LIHTC came about in part as a successor to these previous housing policies. Besides LIHTC representing a paradigm shift in the way low-income families should be housed, the Low-Income Housing Tax Credit was also intended to encourage developers to invest in neighborhoods they might not otherwise, such as low-income and economically depressed areas. While affordable housing policies have shifted, there is still a clear need for affordable housing. A comprehensive study of rental housing supply and demand by Harvard University s Joint Center for Housing Study found that there is a significant shortage of affordable rental housing. There are only 36 units for every 100 low-income renters in central city neighborhoods (Ferland (ed.), 2013, p ). Furthermore, affordable units tend to be in older and more rundown buildings. Because of this, the number of affordable units is decreasing as older units are demolished or taken off the market (Ferland (ed.), 2013, p. 19). LIHTC is a valuable tool in adding new and newly renovated affordable units to the rental market. Program Guidelines The LIHTC program has remained essentially the same since its inception in The program is outlined in Section 42 of the IRS Code, and it works by awarding tax credit to developers whose rental housing developments meet certain criteria (Tax Reform Act of 1986). The actual distribution of credits happens at the state level: credits are allocated to state housing finance agencies, which then distribute them to developers. States may award $2.20 per capita per year in tax credits, or a minimum of $2.5 million, whichever is larger. Credits are good for 10 years after they are awarded. The amount of the tax credit depends on the size and cost of the 6

14 project; state housing finance agencies have formulas to use to calculate the exact amount of the credit. Once awarded credits, developers either use the credits to offset the cost of their project or to raise capital for the project by selling the credits to investors. They can also syndicate credits; that is, they can sell the right to future credits to investors. Developers may seek other financing assistance along with the tax credit, such as tax-exempt bonds or Community Development Block Grants (Schwartz, 2015, p ). Credits are only available for rental housing developments. The project can either be new construction or can be the rehabilitation of an already existing building (National Housing Law Project, 2014). Projects must provide a certain percentage of affordable housing. There are two options from which developers may choose: 20-50, where at least 20% of units are reserved for households making 50% or less of the area median income, or 40-60, where at least 40% of units are reserved for households making 60% or less of the area median income. The amount of the tax credit depends on the percentage of affordable units. Many developers choose to allocate a higher percentage of units to affordable housing to increase both the tax credit amount and their chance of being awarded credits (Schwartz, 2015, p 145). Rent for affordable units must be set at 30% of income of the top of the income limit selected, and rent limits must include utilities and all other fees. Families receiving Section 8 subsidies for their rent are eligible to apply for affordable units (National Housing Law Project, 2014). The application process for the credits is competitive. Priority is given to projects serving the most low-income families and that are planned to remain affordable for the longest period of time. By law, projects must remain affordable for at least 30 years: through a 15-year compliance 7

15 period, followed by a 15-year extended use period d. Additionally, 10% of tax credits must go to non-profit projects (National Housing Law Project, 2014). States have some discretion over the program. They can set longer time requirements for affordability or set stricter priority levels for application to the program, for example, but they must follow all federal laws (LIHTC Basics, 2014). Challenges and Criticisms Although the LIHTC program is popular and generally well-received, it is not without criticism. The program faces several challenges. There is concern over what will happen to the affordability of developments after the compliance period ends. One study found that only 6% of projects whose compliance period ended reverted to market rate housing. The remainder of the projects either renewed their tax credits or remained affordable without tax credits (Schwartz, 2015, p. 151). The program is also criticized for being extremely complex and inefficient. The program has many rules and guidelines and can be difficult to interpret. Another concern is that the program is sensitive to shifts in the economy; during the financial crisis, for example, developers had difficulty finding investors to buy credits. The program has since recovered, but it is something that could be a problem again in the future (Schwartz, 2015, p ). Lastly, there is some criticism that the program is best for relatively moderate-income families and doesn t serve extremely low-income families well. Very low-income families can face a significant rent burden (Moelis Center for Affordable Housing, 2012). d There are some exceptions by which developers can exit the LIHTC program after the 15-year compliance period, but this is rare (National Housing Law Project, 2014). 8

16 LIHTC Project Statistics Since 1987, the LIHTC program has created 2.4 million rental units in almost 40,000 developments (as of 2013) (Fernald (ed), 2013, p. 36). Developments are spread throughout the United States: about 20% of projects are located in the Northeast and West each, 35% are in the South, and 28% are in the Midwest. Slightly less than half, 46%, of developments are located in central city neighborhoods, and 30% are in the suburbs and 24% are in non-metropolitan areas. 61% of developments are new construction, and 36% are rehabilitated buildings. (Schwartz, 2015, p. 144). 9

17 LITERATURE REVIEW To place this study within the proper scholarly context requires a look at the literature on socioeconomically segregated neighborhoods and gentrification, housing policy trends, and previous research on the LIHTC program. Socioeconomic Segregation, Gentrification, and Solutions The relationship between concentrated poverty and racial segregation has been well established in the literature. Scholars such as Douglas Massey and Robert Sampson have shown that cities tend to be heavily racially segregated, with white areas being affluent compared to high poverty minority neighborhoods. This segregation has a host of ill effects, including lower education levels, higher poverty levels, higher crime rates, and fewer opportunities for jobs and economic advancement and mobility. Massey finds that cities are becoming increasingly economically segregated, with segregation at the micro neighborhood level (2009, p. 81). Alex Schwartz and Kian Tajbakhsh point out that numerous scholars, such as William Julius Wilson, have written extensively about the harms of segregation (1997, p. 72). Sampson and Hwang find a troubling relationship between segregation and gentrification, which they define as the process by which upper and middle class newcomers move into poorer, generally minority neighborhoods (2014, p. 728). They report that, as predicted, gentrification often leads to the uprooting of the original residents. They also find that neighborhoods that are already whiter are more likely to gentrify. While some resident may be displaced, low-income residents who remain in the neighborhood benefit from the economic development and growth. Minority neighborhoods are much less likely to gentrify and to get economic development, leading to a continuation of isolated and disadvantaged poor, minority neighborhoods. Neither 10

18 gentrification nor segregation is desirable. Some other solution is needed. Sampson and Hwang recommend promoting affordable housing and socioeconomic integration (2014, ). The commonly proposed remedy to avoid gentrification while still encouraging growth and development is promoting socioeconomically diverse neighborhoods. Mixed-income housing is one of the key components, and much of the current literature on housing policy focuses on mixed-income housing. Current affordable housing policy supports privately-owned rental units housing a variety of income levels. This can be achieved through a number of methods, such as the HOME Investment Program, Community Development Block Grants, and of course the LIHTC program (Urban Land Institute, 2003, p. 16). Many old public housing high rises, now seen as failures, are being demolished and converted into new housing. For example, Lawrence Vale (2012) writes about the transformation of Cabrini Green into mixed-use developments, which was partially paid for through the LIHTC. These changes have ushered in upscale businesses, like Starbucks and yoga studios. He sees Cabrini Green as an example of a mixed income development that is a promising solution for modern urban neighborhoods. Scholars support mixed income housing for a variety of reasons. According to Alex Schwartz and Kian Tajbakhsh, heterogeneous neighborhoods are better because they tend to be more physically attractive, safer, and are more likely to encourage interactions among neighbors (1997, p. 77). Another study by Diane Levy, Zach McDade, and Kassie Dumalo at the Urban Institute finds that mixed-income neighborhoods usually have higher incomes, less racial segregation, and more economic development and revitalization. The authors describe four major benefits that mixed-income communities provide for low-income residents: social networks (i.e. 11

19 expanded and better networks for job searching), social controls (i.e. the presence of people from different income groups improves accountability), behavioral effects (i.e. the example of people from higher income backgrounds who are educated, employed, etc. has a positive effect on people from lower income groups), and political economy of place (i.e. the presence of multiple income groups creates market demand and expanded opportunities) (2010, p. 8). Interestingly, their study also finds that despite all these promising outcomes, mixed income neighborhoods perform best when the income disparity is relatively small (2010, p. 4). Evaluations of LIHTC Much of the research and literature on the LIHTC program focuses on the financial and tax credit side. Malepzzi and Vandell (2002) believe the program has been under-studied, and they cite the need for research into how effective the program has been at meeting the need for affordable housing. Smith and Wiliamson (2008) also suggest that there has been too little empirical research into the program s operations and outcomes (p. 130). Of the research that has been done on LIHTC, much of it focuses on the program s descriptives: where LIHTC developments are located and the tenant makeup. HUD completed an assessment of the program after the first 10 years, and they concluded that the majority of units were built in low-income neighborhoods, and many were built in gentrifying neighborhoods (2012). Casey Dawkins (2011) similarly finds that there is a large percentage of developments in high-poverty, minority neighborhoods in metropolitan areas. She believes that because of this, the program hasn t been successful in enabling opportunities in lower poverty neighborhoods. Finally, an evaluation by Abt Associates looks at low-income tenant makeup. They find that many Section 8 voucher recipients reside in LIHTC units. Many residents are elderly. Many of 12

20 the residents are moderately low-income and working families. The racial makeup of tenants is about equally divided between black and white residents (Nolden, Buron, Heintz, and Stewart, 2000). Several scholars have look at the question of whether LIHTC affordable units really are affordable. A report by the Moelis Center for Affordable Housing Policy at New York University (2012) finds that 40% of units serve extremely low-income households, but more assistance is likely needed for this group. The rent for extremely low-income households is burdensome. In terms of affordability, the program best serves moderately low-income residents. Anne Williamson (2011) comes to similar conclusions. She examines the rent burden for extremely low-income residents in LIHTC projects more closely. She finds that the burden is equal across races; no one group is suffering more than others. Most tenants of affordable units face some sort of burden, but the burden is especially great for those at the very low end of the income spectrum. Those that are within 50-60% of the area median income (i.e. the upper limit of the income limits) face the smallest burden and do best. There has been less research in terms of evaluating the impact of the LIHTC program on neighborhoods. Lance Freeman (2004), who has written extensively on both affordable housing and gentrification, closely examines the types of neighborhoods where LIHTC developments are located. He finds that as compared to other affordable housing programs, there is a large number of LIHTC developments in the suburbs. LIHTC neighborhoods (his term for neighborhoods that contain developments funded by the LIHTC) have higher poverty rates, lower median incomes, and lower median home values as compared to the national average. Freeman concludes that residents are better off than those in traditional affordable housing programs but worse-off than 13

21 the average resident in a non-lihtc neighborhood. Nathaniel Baum-Snow and Justin Marion (2008) find that neighborhoods with LIHTC developments have higher homeowner turnover, higher property values in declining areas, and lower incomes in gentrifying neighborhoods. In gentrifying neighborhoods, LIHTC developments crowd out other new construction, but this does not happen in stable or declining neighborhoods. Kirk McClure s research diverges somewhat from the other authors mentioned so far; he looks at affordable housing in suburbs. He concludes that suburban neighborhoods with LIHTC developments have lower poverty rates than urban neighborhoods with LIHTC developments (2006). Only one study so far has looked at the impact of LIHTC projects on racial segregation. Horn and O Regan (2008) note that there is no evidence that the presence of LIHTC projects increase racial segregation. In fact, they find some evidence that areas with LIHTC projects have lower racial segregation. Residents of LIHTC projects in high-minority neighborhoods either have similar racial makeup to the rest of the neighborhood or are slightly less likely to be minorities. There has also been limited research on LIHTC s relationship with neighborhood revitalization. While noting that there is a dearth of research, Williamson and Smith (2008) suggest that the difference between type of developer i.e. non-profit versus for-profit is worth examining in terms of effects on neighborhood revitalization (p ). They also posit that the extreme level of complexity of the program may be limiting in attracting new investors and developers, and it may also be inhibiting the participation of community-level partners (p. 139). Lan Deng (2009) makes a similar suggestion that it is worth looking into the difference between the actions and effects of non-profit versus for-profit developers (p ). This suggestion 14

22 comes out of his research into LIHTC projects in Miami-Dade County. He uses a cluster analysis to examine and compare neighborhoods with projects, and he finds that LIHTC projects have a positive effect in high-poverty neighborhoods, but mixed effects in working and middle-class neighborhoods (p ). Additionally, he finds that projects seem to be the most successful at spurring neighborhood revitalization when they are targeted to neighborhoods with specific and well-organized plans and strategies (p. 37). Jill Khadduri s report (2013) proposing policy recommendations regarding LIHTC and neighborhood revitalization makes similar suggestions. She proposes either limiting or prioritizing the allocation of LIHTC credits in low-income neighborhoods to those neighborhoods with specific economic development and revitalization plans (p ). The allocation of credits in other low-income neighborhoods should be limited to avoid perpetuating the traditional locational segregation of low-income and affordable housing. While there has been some research done on the LIHTC program, overall it has been limited, and a more in-depth, nuanced look at the outcomes and effects of the program is lacking. Relatively little research focuses on these issues, and there is a particular lack of research regarding effects on the neighborhoods in which the projects are located. The relationship between LIHTC projects and neighborhood income has clear policy interest and relevance both in terms of attempting to successfully provide affordable housing and contributing to neighborhood growth and development. My study attempts to fill in this gap. 15

23 CONCEPTUAL FRAMEWORK Housing policy in recent years has trended away from socioeconomically and racially segregated public housing towards privately owned rental housing and mixed-income developments. The LIHTC program creates affordable housing, while also encouraging private developers to build in neighborhoods they may not otherwise do so. The purpose of the present study is to examine the relationship between the Low-Income Housing Tax Credit and neighborhood change in urban areas. Given the fact that the basis of the LIHTC program is to both provide affordable housing for low-income families and to encourage private economic development, it seems likely that the projects funded by the program would have an effect on the demographic characteristics of the neighborhoods in which they are located. Although the credit can be used for developments in any type of neighborhood urban, suburban, or rural I focus exclusively on urban neighborhoods, as I am interested in seeing whether the presence of LIHTC-funded developments leads to gentrification. I hypothesize that the relationship between the LIHTC program and gentrification could go in one of several directions. LIHTC developments are usually built in lower income neighborhoods, which also tend to have higher concentrations of minorities as well as more poverty, more unemployment; lower median incomes, and less educated residents. LIHTC projects could have a positive effect: the tax credit is used to build housing in economically depressed neighborhoods, successfully helping to spur revitalization in the area. Moderately lowincome tenants may be attracted to the affordable units. Economic revitalization could bring in additional new residents with higher incomes. This could lead to gentrification of the 16

24 neighborhood, with the socioeconomic makeup of the neighborhood completely changing. Median household income would increase. On the other hand, the program could have a negative relationship with change in neighborhood socioeconomic demographics. If middle and upper class residents are displeased with the presence of affordable housing, they may move farther away from neighborhoods with LIHTC projects or be less likely to move into a neighborhood. (In some respects, this would defeat one of the purposes of the program, as it would not be encouraging mixed income housing.) Median household income would decrease. Lastly, the program could have no effect on the makeup of neighborhoods projects are built in. Since LIHTC projects are usually built in lower-income neighborhoods, this may discourage higher-income residents from moving in. Because middle or higher income people don t want to live in the neighborhood, the tenants who move into the new buildings have incomes similar to the neighborhood median, leading to no change. 17

25 DATA AND METHODS Data I use data on the Low-Income Housing Tax Credit from the Department of Housing and Urban Development. The dataset is a comprehensive collection of data on all projects financed by the tax credit between 1987 and HUD compiles information on project characteristics and financing. My demographic data comes from the US Census. Since the LIHTC dataset presents data as the census tract level, I use decennial Census data for corresponding neighborhood characteristics. The data comes from the long form questionnaire, which is answered by the household head. I used data from 1990, 2000, and Because the LIHTC data is at the annual level, I combine data from ten-year periods e to create averages to use with the decennial Census data 1990, 2000, and Some Census tracts changed boundaries over those periods, so I weighted tracts from 1990 and 2000 based on 2010 borders f. The data includes all 50 states as well as the District of Columbia. I limited my sample exclusively to neighborhoods designated as central cities. There are a few limitations that must be noted. The LIHTC data is available at the annual level, with detailed information on each project. However, because the Census data is only available at the tract level every ten years, the LIHTC data loses some level of detail since I am e 1990 LIHTC data includes projects from 1987 through LIHTC data includes projects from LIHTC data includes projects from LIHTC data from was not included. f I used weights created by researchers at the American Communities Project at Brown University. 18

26 working with aggregates. Additionally, about 2,000 projects were dropped due to missing or inaccurate data. Ideally, the information from these projects would have been included. 2,252 Census tracts were also dropped due to inaccurate information, leaving 6,442 tracts. Dependent Variable I use the log of median household income as my dependent variable. This measures the annual income of all family members residing together. A family is defined by the Census as persons related by marriage, birth, or adoption who reside together. Income is reported for the year the Census is taken. I chose to look at income over other measures of socioeconomic characteristics, such as race or poverty levels, because income is generally the key indicator of neighborhood change according to most definitions of gentrification. Jackelyn Hwang and Robert Sampson cite Neil Smith s definition of gentrification as widely accepted: the process by which central urban neighborhoods that have undergone disinvestments and economic decline experience a reversal, reinvestment, and the in-migration [emphasis added] of a relatively well-off middle- and upper middle-class population. This definition does not require that displacement or racial turnover occur, which are still widely debated empirical questions (Hwang and Sampson, 2014, p. 727). Independent Variables My key explanatory variables come from the LIHTC data. The first is the number of LIHTC projects in a Census tract. This is simply an aggregation of all projects receiving the tax credit during each of the ten-year periods corresponding to the decennial Census. I also look at the accumulation of tax credits over time. For example, what is the effect of the number of tax credits over a twenty-year period (such as ) on median household income? Whether 19

27 the effect is positive or negative, I expect the presence of LIHTC developments in a neighborhood to affect median households. My second key explanatory variable is the average percent of affordable units in LIHTC projects. This variable is simply the ratio of the average number of affordable units to all units in LIHTC projects in a Census tract. While developers must have a certain percentage of affordable units in their project to qualify for the tax credit, projects vary widely in the percentage of affordable units, anywhere from the minimum percent to 100%. Similarly to the number of LIHTC projects, I also look at the cumulative effect of the percentage of affordable units. Since developers can choose the percentage of affordable units to put in, the percentage may influence the type of residents the developer is attracting. I also control for a number of factors associated with neighborhood makeup, including demographic and housing stock characteristics. Demographic control variables include total population, the percent of families, the percent of Black residents, the percent of Hispanic residents, the percent of adults with a high school degree or more education, the percent of unemployed adults, the percent of households under the poverty line, and the percent of residents living in the neighborhood ten years or less. Housing characteristics include the percent of dwellings that are vacant, the percent of housing units that are in multiunit buildings, the percent of dwellings that are 30 years or older, the median home value, and the median rent. Descriptive statistics for the dependent and independent variables, as well as a number of other neighborhood demographic characteristics, for each of the three periods, are in Tables 1 through 3 in the Appendix. 20

28 Methods I estimate OLS regression models to analyze the relationship between changes in the number of LIHTC projects, the percent of affordable units in LIHTC projects, and median income. I use the three time periods for which I have Census data as my points of comparison for measuring changes over time. I first estimate bivariate models using income and number of projects and income and percent of affordable units to establish relationships between each of the key explanatory variables and the dependent variable. I then estimate multivariate models with both key explanatory variables and incorporating the control variables discussed above. I examine relationships over the short, medium, and long-term using each of the three time periods: 1990, 2000, and I also analyze the relationship between the cumulative number of projects and the percent of affordable units and changes in the median income over time. 21

29 RESULTS Before estimating any regressions, I first divided Census tracts into six categories based on the number of projects in each tract and compiled summary statistics for each of the three periods, found in Tables 4 through 6 in the Appendix. Given that my study examines the relationship between the number of LIHTC projects in a Census tract and median income, I wanted to first see if neighborhood characteristics varied according to concentration of projects. As expected given the prioritization of projects that serve the most low-income residents, tracts with more projects tend to have lower median incomes, higher poverty rates, higher percentages of female-headed households, and higher concentrations of minorities. Neighborhoods with more LIHTC projects also have more renters, a higher percentage of dwellings in multiunit buildings, lower median home values, lower median rent, and a higher percentage of residents who have lived in the neighborhood ten years or less. These numbers all increase or decrease accordingly as the number of LIHTC projects in a tract increases. Interestingly, the percentages and measures are relatively similar across the three time periods. I begin by estimating a bivariate regression of each of the key independent variables, number of projects and percent of affordable units, on the log of median household income for each of the three periods, 1990, 2000, and 2010, to establish a relationship between the explanatory variables and the dependent variable. When I regress the number of projects in 1990 on the log of 1990 median household income, the coefficient on the number of projects is -0.05, and it is statistically significant at the 1% level. For the separate regression using the percent of affordable units in 1990, the coefficient on the percent of affordable units is -0.16, and it is statistically significant at the 1% level. Last, I estimated a regression using both the number of 22

30 projects in 1990 and the percent of affordable units in The coefficient on the number of projects is and is significant at the 1% level, and the coefficient on the percent of affordable units is and is significant at the 1% level. When I regress the number of projects in 2000 on the log of 2000 median household income, the coefficient on the number of projects is -0.03, and it is statistically significant at the 1% level. A regression of the percent of affordable units in 2000 produces a coefficient on the percent of affordable units of -0.06, which is significant at the 1% level. When I estimated a regression using both variables, the coefficient on the number of units is -0.02, which is significant at the 1% level, and the coefficient on the percent of affordable units is -0.03, which is significant at the 5% level. For 2010, for the regression of the number of projects on the log of median household income, the coefficient on the number of projects is -0.05, and it is statistically significant at the 1% level. For the separate regression using the percent of affordable units in 2010, the coefficient on the percent of affordable units is -0.05, and it is statistically significant at the 1% level. Last, I estimated a regression using both variables. The coefficient on the number of projects in 2010 is -0.05, and it is significant at the 1% level. The coefficient on the percent of affordable units is 0.03, and it is significant at the 10% level. It is interesting that the relationship between both number of projects and percent of affordable units and household income is negative over all three periods. I had anticipated that the presence of LIHTC developments may contribute to gentrification, one of the key aspects of which is higher income. Preliminarily, at least, this does not seem to be the case. 23

31 After looking at the bivariate models, I add in the control variables to see whether the relationships between the explanatory variables and the dependent variable still hold and to see how the control variables affect the model. Table 7 represents the results of the regression of the number of LIHTC projects and percent of affordable units on the log of median household income for all tracts with LIHTC projects. For all of the models, the dependent variable is the log of median household income. For the first set of models, the key independent variables are the number of projects and the percent of affordable units for that period For the second set of models, the key independent variables are the cumulative number of projects and percent of affordable units through the period specified. The first set of models looks at the relationship between the number of projects and the percent of affordable units and the log of median household income in the short, medium, and long-term for each of the three periods. I take projects and the percent of affordable units for 1990 and estimate effects on income in 1990 (short-term), 2000 (medium), and 2010 (long-term). I estimate the effect of projects and percent affordability in 2000 on 2000 income (short-term) and 2010 (medium). Lastly, I estimate effects of 2010 projects and percent affordability on 2010 income (short-term). I could imagine projects having different effects over time. For example, in the case of a positive relationship between LIHTC projects and income, it may take time for the neighborhood to attract additional business, investment, and higher-income residents, so effects might not be seen for ten years or even more. In all three of the models for 1990 (short, medium, and long-term), neither the number of projects nor the percent affordable is statistically significant. The 1990 variables only cover four years 1987, 1988, 1989, and 1990 and represent the beginning years of the program. There 24

32 may have been no effect during this range due to the shorter time period and smaller number of projects. Alternatively, investors and developers may have been hesitant or slow to apply to participate in the program, or the program may have had a slow start up period. Percent affordability is statistically significant in both models for The number of projects is statistically significant for short-term effects (i.e. as a predictor for income in 2000) but not for medium-term effects (i.e. as a predictor of income in 2010). The number of projects built between 1991 and 2000 is associated with a 0.24% decrease in median household income and is significant at the 10% level. The percent of affordable units created between 1991 and 2000 is associated with a 5.28% decrease in median household income in 2000 and is significant at the 1% level. The percent of affordable units created between 1991 and 2000 is associated with a 9.3% decrease in median household income in 2010 and is significant at the 1% level. The percent of affordable units created between 2001 and 2010 is associated with a 3.87% decrease in median household in 2010 and is significant at the 1% level. The number of projects is also statistically significant for this period, although the coefficient is quite low. The number of projects is associated with a 0.06% decrease in household income. It is surprising that the number of projects, in most cases, did not affect household income. I had expected to see some relationship, whether positive or negative. Furthermore, I had expected to see a parallel relationship between number of projects and percent of affordable units, whereas that did not happen. In all of these models, other factors that are strong predictors of household income in the short, medium, and long term include percent poverty, percent of renters, median rent, and median home value. Poverty levels are directly related to income, so this relationship makes 25

33 sense. Similarly, it is logical that median rent and home values in the neighborhood would be positively correlated with income neighborhoods with higher rent and home costs would likely attract residents with higher incomes. The percent of renters is negatively correlated with income. Again, it is not surprising that neighborhoods with higher proportions of renters would have lower median incomes, as people at the lower end of the income distribution are more likely to rent their home (Fernald (ed), 2013, p. 4). My second series of models analyzes the relationship between the cumulative number of units and percent of affordable units and the log of median household income over short and medium ranges of time (I cannot look at long-term effects as I only have a 20 year range with the cumulative data). The purpose of the cumulative variables is to try to approximate a way to measure change over time by seeing the effect of the aggregate number of projects and percent of affordable units. In looking only at one time period, the effect of all the projects put in place before that time period is ignored. The cumulative variables attempt to remedy that. First, I look at the cumulative projects and percent of affordable units created between 1987 and I analyze the relationship between projects and units created during this time period and income in the short term (2000) and medium term (2010). Cumulative number of projects is not statistically significant for either period, but cumulative percent affordable is. The cumulative percent of affordable units created between 1987 and 2000 is associated with a 1.79% decrease in median household income and is statistically significant at the 1% level. Using this same range to look at 2010 income shows a 3.94% decrease in median household income, which is statistically significant at the 1% level. 26

34 I next take the cumulative projects and percent of affordable units created between 2000 and Again, projects are not statistically significant, but the cumulative percent of affordable units is associated with a 4.83% decrease in median household income, which is statistically significant at the 1% level. Lastly, I take the whole range of time and look at the cumulative number of projects and percent of affordable units created between 1987 and 2010 and regress this on 2010 income. The cumulative number of projects is statistically significant at the 10% level, but the coefficient is small: -0.23, or a 0.23% decrease in median household income. The cumulative percent of affordable units in associated with a 1.01% decrease in median household income and is statistically significant at the 10% level. There are a number of other strong predictors of household income. Like the first set of models, percent poverty, percent families, percent of renters, median rent, and median home value are all strong predictors of median household income. In these models, the percent of people with a high school diploma or higher degree is positively correlated with income. There is a well-established, strong relationship between education and earnings, so this relationship makes sense. I compare the results of the regression for tracts with LIHTC projects to all census tracts. The results are in Table 8 in the Appendix. Here, I look only at the number of projects since all tracts didn t have projects, so it does not make sense to look at the percent of affordable units. Similarly to the previous models, the coefficient on number of projects is not statistically significant in the short, medium, or long term for For 2000, 2010, and cumulative effects, the results are statistically significant at either the 1% of 5% level depending on the period, but in 27

35 all cases, the coefficient on the number of projects is very low (less than a 1% increase in median household income). Contrary to my expectations, the number of projects alone seems to have relatively little effect on a neighborhood. The models show that the number of LIHTC projects, whether for a single period or cumulatively, has little to no effect on median household income, either in the short or longer term. The percentage of affordable units, somewhat surprisingly, has a negative impact on income, both for individual periods and cumulatively. I did not anticipate the percentage of affordable units alone to have an effect on income; if anything, I thought the number of projects would be the main driver for income effects. 28