Are Inclusionary Housing Programs Color-blind? The Case of Montgomery County MPDU Program∗

Adji Fatou Diagne †

Benoˆıt Schmutz ‡

January 31, 2017 - Very Preliminary -

Abstract Relying on exhaustive administrative data spanned over four decades, this paper studies the treatment of African-American applicants by the Moderately-Priced Dwelling Unit (MPDU) program in Montgomery County, MD. We show that this program was equally accessible to African-American applicants, except between 1995 and 2000, when African-Americans’ conditional probability of being selected into the program was lowered by 10% compared to that of Caucasian applicants, maybe as a temporary response to the sudden surge in AfricanAmerican applicants that occurred at that time. Turning to the outcome of the selection process, we show that even if the spatial allocation of beneficiaries does reflect preference-based sorting patterns observed on the private housing market at the neighborhood level, the program seems to induce some scattering of different ethnic groups at the most local level. When comparing beneficiaries living in the same housing development, but at different addresses, we find that African-American beneficiaries have a conditional share of African-American neighbors between 0.4 and 1 standard deviation lower than other beneficiaries.

JEL codes: R31, R38, J15. Keywords: Housing Market Discrimination; Public Housing; Spatial Sorting; Propensity Score Matching.

∗

We thank the MPDU past and present agents, and in particular Stephanie Killian and Gael Le Guellec, for access and help on the data. † Howard University, EEA, and NABE; email: [email protected] ‡ Howard University and CREST; email: [email protected]

Introduction African Americans have historically faced many barriers that limit access to and choice of housing. Aside continued discrimination in the housing market documented through use of fair housing audits (Yinger, 1986; Turner et al., 2002; Turner et al., 2012), rising housing prices in many U.S. metropolitan markets have only made things worse. Many regions have suffered from a perennial shortage in affordable housing due acute economic conditions, rising rent prices, difficulties in obtaining mortgages (HUD, 2012) and in several cases strict zoning regulations (Rosen & Katz, 1981; Ihlanfeldt, 2004; Mills, 2009). Regions with least affordable rental and ownership housing are clustered on the East and West Coasts (Glaeser and Gyourko, 2002). Nonmarket sources attributed to spatially differentiated patterns of income and race have been linked to local government regulations as well as discriminatory practices engaged in by various actors in the housing market (Schill & Wachter, 1995). Those factors including local government land use regulations in turn hamper housing affordability. Inclusionary zoning (IZ) is a housing policy that requires developers to set aside a percentage of units in housing developments for low and moderate income residents (Schuetz et al., 2009; Meltzer & Schuetz, 2010). IZ policy allows for the creation of mixed-income communities promoting access to low-poverty neighborhoods (Schwartz et al., 2012). Density bonuses or other cost-reducing incentives are provided to compensate developers for providing affordable units to offset the potential reduction in profit margins associated with this supply. One of the main goals of the program is to promote income integration at the project level through the dispersion of housing units. Such integration would then spillover to surrounding neighborhoods avoiding ethnic and income clustering traditionally created by other subsidized housing programs (Calavita & Grimes, 1998). In addition, the program would also help counter residential racial segregation and discrimination often caused by exclusionary effects of restrictive land use practices (Rosen & Katz, 1981; Ihlanfeldt, 2004; Mills, 2009). This paper investigates African Americans’ access to housing in the Montgomery County, Maryland IZ program enacted in 1974. Our analysis is based on over 22,000 original participants or applicants of the MPDU program from 1980-2015. Using propensity score matching techniques, we look specifically into the selection of African Americans among homeownership applicants using eligibility requirements set by the program. We then employ sorting indices and hedonic price techniques to examine the success of the MPDU program at integrating participants at the development level and to investigate whether there exists racial price differentials. We show that while the program has not always ensured equal access to African-American applicants, it does seem to have a positive impact on integration at the very local level. To our knowledge, this is the first paper to study discrimination using exhaustive data from a government-regulated housing program as well as looking specifically into participants of any inclusionary zoning program. This is also the first study to empirically model racial integration analysis at an even smaller scale than the building level. Studying potential ethnic biases in the selection process of such program not only adds to the literature on housing market discrimination and segregation but also provides guidance for fair housing policy.

1

Related literature Discrimination in Access to Housing Previous work documenting the experiences of minorities in housing access has relied heavily on experimental studies using matched-pair techniques or audits such as that of Yinger’s (1986) real estate agent experiment in Boston and those conducted by the Department of Housing and Urban Development (HUD) in 1977, 1989, 2000, and 2012 (Turner et. al, 2002; and Turner et. al, 2012). The method also inspired numerous organizations in conducting their own audits, 71 of these were reviewed by Galster (1990). Significant level discrimination was found in all studies although gap in access between whites and African Americans shrunk in the last two HUD studies. Although praised as a premier method in detecting discriminatory practices in the housing market, audits have become increasingly critiqued due to likelihood of bias in results if actors are not identical along all dimensions except race (Heckman 1998; Heckman and Siegelman 1993). Such actor bias may be minimized if actors are carefully chosen and trained, yet Hanson and Hawley (2011) suggest that actors may either misreport or unintentionally prompt a discriminatory response. In attempts to avoid actor bias, their study along with many others began to experiment with audit methods using the Internet through email and online correspondence. Hanson et al.’s (2016) study using electronic correspondence with mortgage loan originators (MLOs) is the most recent and only study looking into specifically home buying and mortgage market discrimination using online audits. Results revealed prejudice in MLO response between whites and African Americans by 1.8 percent. Credit scores were seen to affect much of the response rates, an 8.5% response difference between high credit score groups versus low credit score groups. Besides being highly experimental, paired testing studies only focus on the initial encounter between a homeseeker and a rental or sales agent (Hanson et al., 2016). Discrimination may occur later in the housing transaction when a homebuyer makes an offer on a particular unit or applies for financing. Our study on access differs greatly from the aforementioned studies in such that it is non-experimental, provides a broad sample of home seekers where we can see the characteristics of selected individuals and avoid biases caused by experimental data mentioned above. The pool of applicants is identical at all levels thus provides us to isolate race as a determinant of selection. Racial Price Differentials, Sorting, and Segregation Empirical measures of racial differentials in housing prices have also yielded mixed results due to data limitations in order capture intraneighborhood racial price differentials. Often these studies have had to rely on the immediate block surrounding a household (Knowles Myers, 2004). The first attempt to look into price differentials at the smallest geographical area possible is the study of King and Mieszkowski (1973) using a special survey data from 1968-1969 for approximately 220 rental units in New Haven, Connecticut and found while controlling for renter’s race as well as the racial composition of a neighborhood and whether the neighborhood is in what they call black ghetto, white interior, or boundary areas. They find that blacks pay about 7 percent more than whites in the boundary areas. Although the study controlled for neighborhood racial composition, its data size and 2

dated results sparked further research. To address the former, Follain and Malpezzi (1981) followed a similar path of study but expanded data to 39 SMSAs from the Annual Housing Survey (AHS). They found contrary price premiums, showing black owners paying 15 percent less and black renters paying 6 percent less than whites. However, their data only allowed for controls using the central city level dummy, which could?ve omitted the neighborhood racial composition impact. Later studies using census tracts as the lowest geography found blacks paying less while another using special block level data found blacks paying a premium. Chambers (1992) and Kiel and Zabel (1996) both conclude that census tracts were too broad of a geographical area to capture reliable price discounts. After controlling for neighborhood composition, the former study finds the racial discounts to disappear for renters while declines to12 percent for owners while the latter study finds the discount for owners to decrease to 5 percent. Contrary to those results, Knowles Myers (2004) finds black homeowners paying 10 percent more for housing after controlling for neighborhood fixed effects. Our address level data allows us to counter geographical limitations of the studies mentioned above. In addition, our nearly 40 year timespan also allows us to control for year fixed effects in order to test whether an expansion in the housing supply available to blacks eventually lower, eliminate, or even reverse price differential in the long run as suggested by Bailey (1959). We are able to control for area fixed effects at the zip code and building level, smallest geographies ever used to study price discrimination. Our distinctive unit level data hence allows us to investigate whether or not African Americans pay a premium after controlling for housing quality variables. The remainder of the paper is organized as follows. Section 1 presents the study area, features of the MPDU program, data, and variables. Our selection analyses including robustness checks are presented in Section 2, while section 3 presents our racial sorting analysis. We then conclude by discussing results and recommendations.

1 1.1

Study Area and Data History and context

Inclusionary zoning (IZ) housing policy began in the Washington, D.C. metropolitan areas in the early 1970s. Among the first regions to experiment with IZ policy was Montgomery County, Maryland in 1974. The county named this program the “Moderately Priced Dwelling Unit” (MPDU) and it was designed to address three housing policy concerns: land use and density, fair housing and desegregation, and workforce housing and economic development. In exchange, developers were granted ability to build more units than zoning codes typically permit or offered tax abatements and other incentives. Affordability became a major issue in the county due to rapid appreciation of housing prices as a result of slow residential and commercial building caused by a sewer moratorium in the county for most part of the 1970s. Once resolved, building arose and the program benefited from a surge in affordable units built in the 1980s. By 2013, the program had created over 14,000 units including rental and owner-occupied 3

housing (DHCA 2016). Currently, the MPDU law stipulates that 12.5 to 15% of all units constructed in subdivisions of 30 units or more must be set below market price, as low as 30 percent lower than market prices. In return developments are allowed a density bonus for up to 22 percent. Ever since its adoption, the program has been through several changes, with most centered on ways to expand the affordable housing stock. Most of these changes were extensions to unit control periods, participating developers’ building size limits, and percentage of housing to be allocated. Table 1 below provides summary of those changes. Table 1: MPDU Ordinance Timeline

Project Size Affordable Units Required Density Bonus Allowed Control period - rental units (years) Control period - Ownership units (years)

1974

1981

1988

2002

2005-Present

50 15% 20% 5 5

50 12.5% 20% 10 10

50 12.5-15% up to 22% 20 10

35 12.5-15% up to 22% 20 10

20 12.5-15% up to 22% 99 30

Notes: A sliding scale requirement was inacted in 1988, with the minimum set 12.5%, whether the devleoper uses the density bonus or not. The bonus increases up to 22% by providing more affordable units up to a maximum of 15%. Source: HUD 2012.

Since this paper focuses specifically on access, we will only elaborate on changes pertaining to participants. In order to increase resident participation in the MPDU program there were several historical changes. In the late 1970s for example, high mortgage rates were disqualifying many families from purchasing; as a result the County Council modified the ordinance to include the cost of financing when calculating the income limit. Although not a formal ordinance change, in 1995, as a result of an increase in the number of housing units built, the county’s Department of Housing and Community Affairs (DHCA) engaged in providing more information about the program by sending information to county regional service centers and libraries (MC Confidential Interview 2016). This advertising effort let to a major evolution in the number of applicants and especially those of minorities as shown in figure 1. In addition, the county’s demographic profile dramatically changed since the 1980s, becoming more diverse at a variety of levels, part of this change widely attributed to aging of existing population and a surge in foreign immigration (see table 2). Total non-Hispanic white population declined by 18% between 1987 and 1997. Today, an estimated 55% of the county’s residents are minorities. Table 2: Evolution of Montgomery County Population

Population % African American % Hispanic % Asian % Non-Hispanic White

1980

1990

2000

2015

579,053 8.8 3.9 3.9 82.5

757,027 12.2 7.4 8.1 72.4

873,341 15.1 11.5 11.3 59.5

1,040,116 19.1 19.0 15.4 45.2

Source: US Census Bureau.

4

1.2

Program Participants

The county restricts participation to households earning 70% or less of than the area’s median income, adjusted for family size, tenure, and unit size. Income limits for both purchase and rental units are set by the County Executive and updated on a yearly basis. Currently, they range from $53,500 for a single-person household to $82,500 for a family of five for purchasing households. We restrict our sample to individuals who participated in the ownership program between 1980 and 2015; these individuals make up more than half of total participants. Applicants seeking rental units were excluded. To be eligible for purchase, applicants were also restricted to a minimum income level to serve as a guarantee that they could afford costs associated with the purchase. As of September 2016, the purchaserss minimum annual household income requirement is $35,000; the purchaser must not already own a house or have owned a house anywhere in the past five years. The participants are required to be able to make a down payment, as well as absorb settlement fees and other closing costs. In terms of financing, applicants must provide a pre-qualification letter from a Maryland Housing Opportunities Commission approved lender in an amount of at least $120,000 (DHCA 2016). It is important to note that the income requirements have changed since the program’s inception, for example, in 1987, the minimum income for purchase was set at $20,000. However, if the HOC lender deemed the applicant able to afford their mortgage payments, the lender could pre-approve the application (MC Confidential Interview 2016). Figure 1: Evolution of the Number of Applicants

Source: MPDU

The county conducts a random selection drawing to determine who will receive an opportunity to purchase MPDU home. Separately, a developer or the current owner of a MPDU notifies the county of unit availability and in turn the county notifies residents of availability. These 5

units are then sold to approved-purchase participants. In order be eligible for a listing, the participant must be pre-approved for a mortgage loan equal to or greater than the sales price of the available property. The participant household size must also meet the deemed appropriate household size determined by the county for the number of bedrooms in the property, unless an insufficient number of households of eligible size enters the drawing. It is important to note that the county only assures that participants meet the requirements listed above but does not participate in the lending or purchase process. That said, “priority points” are given to those who live (1 point), work (1 point) in the county and have frequently applied and been approved MPDU purchase program participant on a yearly basis (1 point per year for a maximum of 3 years). The household ranked first has the first opportunity to purchase a home. If they decline, the household ranked second has the opportunity to purchase, followed by the household ranked third, etc. Declining to purchase a home does not in any way penalize a household in future selections (DHCA, 2016). The builder’s sales agent will contact the highest ranked participants, beginning with number 1 in the group with 5 priority points. The number of participants that will be contacted depends on the number of homes available in that selection process. For example, if 5 homes are available, the 5 highest ranked participants will be contacted. The most important factor for the purpose of our study is that the county only assures that participants meet the requirements listed above but do not participate in the lending or purchase process. Thus, discrimination can arise after the county conducts random selection and provides list to sales agents.

1.3

Samples

Tables 3 and 4 report the average characteristics of participants used in Section 2 and in Section 3 respectively, using two different ethnic-based partitions of the population. Table 3: The population of applicants: individual and eligibility characteristics Total

Non-Cauc.

Cauc.

Afr.-Am.

Non-Afr.-Am.

Missing

Income Household size Disable Number of applications Only lives in MC Only works in MC Both lives and works in MC

43,034 2.42 0.003 1.21 0.26 0.09 0.65

42,960 2.70 0.004 1.25 0.27 0.09 0.64

43,032 1.85 0.002 1.12 0.23 0.11 0.66

42,578 2.57 0.005 1.19 0.28 0.12 0.60

43,193 2.35 0.003 1.22 0.25 0.08 0.67

44,400 2.47 0.002 1.42 0.21 0.06 0.68

Number of observations

11,324

7,330

3,596

3,604

10,497

398

Notes: (i) The sample is restricted to purchase applicants who have applied less than 9 times to the program, observed between 1982 (no minority applicants before) and 2014 (incomplete information on 2015); (ii) Income is in 2015 $; (iii) For applicants with multiple applications, income and household size are averaged over all observations Source: MPDU.

They show that the different groups do not differ much in terms of observable characteristics. We choose to focus on the second partition, which isolates African-American households from 6

the rest of the population, and we drop the observations for which information on ethnic group is missing. Table 4: The population of beneficiaries: individual and dwelling characteristics

Income Household size Purchase price Townhouse City Share Zipcode Share Building Share Number of observations

Total

Non-Cauc.

Cauc.

Afr.-Am.

Non-Afr.-Am.

Missing

46,397 2.48 138,127 0.61 0.13 0.13 0.16

46,218 2,95 143,744 0.66 0.14 0.14 0.18

46,643 1.90 130,632 0.55 0.13 0.13 0.12

47,851 2.78 141,353 0.65 0.15 0.15 0.20

46,059 2.41 137,062 0.60 0.13 0.13 0.14

46,127 2.44 145,064 0.60 0.14 0.14 0.17

3,874

2,082

1,659

726

3,010

133

Notes: (i) The sample is restricted to beneficiary households applicants who have applied less than 9 times to the program, observed between 1982 (no minority applicants before) and 2014 (incomplete information on 2015); (ii) Income and price are in 2015 $; (iii) For applicants with multiple applications, income and household size are averaged over all observations; (iv) Shares variables are the share of African-American neighbors within the population of neighbors also registered in the program at the city, zipcode and development levels. Source: MPDU.

2 2.1

Access: Modelling Selection Raw statistics

For initial measures of access to housing units by ethnic group, we focused on the quantity rationing and allocation of the raw shares of total applicants versus units purchased by race. Supposing that the information we have on participants are the same available to the MPDU program administrators, we proceed to testing whether there evidence of prejudice in African American Access to housing. Figure 2 below depicts these shares and shows that once the program increased advertisement to attract more applicants, the surge in African American showed potential periods of racial discrimination in the housing market. We will empirically test for the significance of such difference next.

2.2

Propensity Score Matching

To test whether the observed probabilities of selection during the years where the gap between applicant numbers and African Americans for robustness of the selection process, we use a propensity score matching technique. This methodology allows us to control for selection requirements, participant characteristics, and time fixed effects while comparing households within each category. We compute the average treatment effect of being African American over the sample of observable characteristics described in Table 3. Our results shown in Figure 3 confirm our previous diagram providing indication of differential treatment in the years 1996-2000 where African American applications increased substantially. During those years, their probability of purchasing a home through the MPDU program 7

Figure 2: Share of African Americans Applicants and Purchasers vs. Caucasian Applicants and Purchasers

Source: MPDU

Figure 3: The impact of being African-American on the probability of being selected into the program

Notes: (i) Average treatment effect of being African-American on the probability of being selected into the program (ii) Estimations are performed year by year or over the period specified on the x-axis (iii) Kernel propensity score matching with a 10% bandwidth, using a logit estimation in the first stage (iv) 95% confidence intervals obtained with bootstrapped standard using 100 replications Source: MPDU

8

went down tremendously ranging from 7 to 15 percent. Figure 3 below depicts a plot of our predicted coefficients.

3 3.1

Outcomes: Price differentials and Spatial Integration There are no observed differences in purchase prices

An additional check of the relevance of the quantity approach followed in section 2 is to examine how various factors affect prices using a hedonic price model. In doing so we also employ finer locational fixed effects to control for broader locations of the dwelling unit. Since demand and supply are not separately identified, the hedonic price model is the only economic theory technique that leads to conclusions about the different types of differentials (Knowles Myers, 2004). Our specification includes occupant characteristics (ethnic group, income and household size) and unit characteristics (number of bedrooms and bathrooms) as well as location controls. Our regression findings on Table 5 below leads us to the conclusion that there is no racial premium to be paid in this market, even after controlling for the smallest level of geography, units located within development. Table 5: Hedonic regressions on purchase price

African-American Income Household size

(1)

(2)

(3)

(4)

-4,557*** (1,236) 0.246*** (0.0522) 4,125*** (434.3)

-3,139** (1,245) 0.249*** (0.0518) 3,936*** (436.2)

-2,953** (1,225) 0.241*** (0.0511) 3,312*** (437.6)

-1,452 (1,080) 0.139*** (0.0453) 772.0* (415.4)

City fixed effects Zipcode fixed effects Building fixed effects Observations R-squared

X X X 4,251 0.278

4,251 0.295

4,251 0.321

4,186 0.548

Notes: (i) Ordinary-least-square regressions of purchase price as a function of individual and dwelling characteristics, with *, ** and *** respectively denoting 10%, 5% and 1% significance levels; (ii) All specifications include a set of dummy variables coding for the number of bathrooms and bedrooms as well as time dummies; (iii) “City” is the municipality and “Building” is the development Source: MPDU

3.2 Spatial correlation results suggest demand-driven sorting of MPDU beneficiaries Besides differential access and prices, the other important feature of the program is its impact on the spatial distribution of its beneficiaries. Controlling for prices, we seek to investigate whether 9

there are residual differences in the places people end up living. Our choice of model allows us for the possibility to detect residential location choices of households that makes explicit the way individual locational decisions aggregate to form a housing market. Like in our previous model, we incorporate location-specific unobservables and use these as indicators of household preference over choice of characteristics, including those that depend on household sorting such as city, neighborhood, and even building ethnic composition. Table 6: Raw sorting: the share of African-American neighbors at various spatial levels (1) Afr.-Am.

Municipality (2)

(4)

Zipcode (5)

(6)

(7)

Building (8)

(9)

0.026*** (0.0028)

0.025*** (0.0028) -8.4e-08 (1.2e-07) 0.000017 (0.00095)

0.024*** (0.0028) -5.0e-08 (1.2e-07) -0.00011 (0.0010) -1.8e-07*** (3.5e-08)

0.029*** (0.0031)

0.028*** (0.0031) -6.3e-08 (1.3e-07) 0.0012 (0.0010)

0.027*** (0.0031) -2.5e-08 (1.3e-07) 0.00094 (0.0011) -1.3e-07*** (3.9e-08)

0.056*** (0.0050)

0.053*** (0.0051) 2.8e-07 (2.1e-07) 0.0050*** (0.0017)

0.050*** (0.0050) 3.0e-07 (2.1e-07) 0.0054*** (0.0018) -3.0e-07*** (6.2e-08)

4,380 0.138

4,313 0.135

4,250 0.147

4,381 0.126

4,314 0.124

4,251 0.133

4,370 0.287

4,304 0.288

4,243 0.295

Income HH size Price Obs. R-sq.

(3)

Notes: (i) Ordinary-least-square regressions of the share of African-American MPDU neighbors at the city (columns 1 to 3), zipcode (columns 4 to 6) and development (columns 7 to 9) levels, as a function of individual and dwelling characteristics, with *, ** and *** respectively denoting 10%, 5% and 1% significance levels; (ii) Columns 3, 6 and 9 include a set of dummy variables coding for the number of bathrooms and bedrooms and all specifications include time dummies Source: MPDU

The results are shown in Table 6. It is striking to see how stable the raw sorting coefficients are at the municipality and zipcode levels: on average, African-American beneficiaries live in cities and neighborhoods with 2% more other African-American beneficiaries. The effect is twice as large at the building level. However, this result is likely to be driven by demand factors, leading households to apply to neighborhoods or cities that they already know of, or like. One way to mitigate this concern is to include area fixed effects at a higher level in the specification. This is what is shown in Table 7, which reproduces column 9 in Table 6 but studies the impact of including city (column 2) and zipcode (column 3) fixed effects. As expected, the sorting coefficient drops dramatically, even if it remains highly significant. This impact of greater location on the outcome of the allocation process may also be seen through the increase in Rsquared, which increases from 30% to 50%.

3.3

At the most local level: spatial integration of MPDU beneficiaries

However, these results should still not be interpreted as evidence of ethnic-based sorting through the program, because applicants do choose which development (or building) to apply to. Therefore, the only plausible test of the sorting impact of the program has to take place within the development. This analysis can be performed over a subsample of MPDU beneficiaries, who live in developments that feature several addresses, and for which each address hosts at least 10

Table 7: Sorting conditional on greater location: the share of African-American neighbors at the building level

African-American Income Household size Price

(1)

(2)

(3)

0.050*** (0.0050) 3.0e-07 (2.1e-07) 0.0054*** (0.0018) -3.0e-07*** (6.2e-08)

0.020*** (0.0044) 2.3e-07 (1.8e-07) 0.0063*** (0.0016) -1.1e-07** (5.4e-08) X

0.016*** (0.0043) 2.2e-07 (1.8e-07) 0.0046*** (0.0015) -1.1e-07** (5.4e-08)

City fixed effects Zipcode fixed effects Observations R-squared

X 4,243 0.295

4,243 0.479

4,243 0.510

Notes: (i) Ordinary-least-square regressions of the share of African-American MPDU neighbors at the development level, as a function of individual and dwelling characteristics, with *, ** and *** respectively denoting 10%, 5% and 1% significance levels; (ii) All columns include a set of dummy variables coding for the number of bathrooms and bedrooms as well as time dummies Source: MPDU

two MPDU households. The identifying assumption is that while applicants do choose which developments to apply to, they do not choose which part of the development (which staircase, for example). This seems very plausible. Table 8 shows that the sorting coefficient is actually reversed when comparing households that live within the same development building. Regardless of whether one controls for other individual and dwelling characteristics or not, AfricanAmerican beneficiaries are less likely to share the same address with other African-American beneficiaries, than their non-African-American building neighbors.

4

Conclusion

Recognized as a preferred tool in local affordable housing provision, inclusionary zoning policy has served low and moderate income households in various ways, including the increase of the housing stock in below market price tiers and expansion of housing options. As a form of mixed-housing policy, it also emerged to counter discrimination and segregation often caused by traditional subsidized housing programs (Calavita and Grimes, 1998). Using participant and unit level data from the leading inclusionary zoning policy in the country, the Montgomery County, Maryland?s Moderately Priced Dwelling Unit (MPDU), we examine the experiences of minorities, particularly African Americans in the ownership selection process. Additionally, we test for one of the program?s original goals, integration. The results of the analysis indicate that between 1996-2000, a period of a surge in applications from this group as a result of a 1995 outreach to local residents, African Americans were up to 15% less likely to purchase an MPDU unit. The results indicate possible prejudice practices in the 11

Table 8: Sorting conditional on applicants’ choice: the share of African-American neighbors at the address level

African-American

(1)

(2)

(3)

(4)

0.0763** (0.0320)

0.0434 (0.0319) 1.84e-06 (1.43e-06) 0.0382** (0.0161) -5.19e-07** (2.50e-07) X

-0.127*** (0.0313)

X

-0.126*** (0.0318) 5.78e-07 (1.39e-06) 0.0496*** (0.0189) -2.66e-07 (3.03e-07) X X

605 0.380

582 0.396

Income Household size Price Dwelling characteristics Building fixed effects Observations R-squared

608 0.089

585 0.152

Notes: (i) Ordinary-least-square regressions of the share of African-American MPDU neighbors at the address level, as a function of individual and dwelling characteristics, with *, ** and *** respectively denoting 10%, 5% and 1% significance levels; (ii) Dwelling characteristics are a set of dummy variables coding for the number of bathrooms and bedrooms; all speficiations include time dummies Source: MPDU

housing market during those years, including lending market discrimination. However, the selection process converged back to a non-discriminatory, high-minority, equilibrium after then. Furthermore, results imply that applicant preference of certain neighborhoods may partly overcome program efforts. In other words, participants are shown to sort into neighborhoods of similar racial composition, African American beneficiaries are seen to reside in cities and neighborhoods with 2% more other African Americans receivers, with a twice as large effect at the building level made up of mostly townhomes and condos. However, if we really control for applicant preferences, by comparing households sharing the same development building, there is suggestive evidence that the program acts, voluntarily or not, as an integration device at the most local (the postal address) level. Although this finding calls for further research to be confirmed, it certainly suggests that the integration goal of this example of an inclusionary zoning program does not only help mitigate income segregation, but ethnic-based segregation as well.

12

REFERENCES     Bailey,  Martin  J.  "Note  on  the  Economics  of  Residential  Zoning  and  Urban  Renewal."  Land   Economics  35,  no.  3  (1959):  288-­‐292.   Calavita,  Nico,  and  Kenneth  Grimes.  "Inclusionary  Housing  in  California:  The  Experience  of   Two  Decades."  Journal  of  the  American  Planning  Association  64,  no.  2  (1998):  150-­‐ 169.   Chambers,  Daniel  N.  "The  Racial  Housing  Price  Differential  and  Racially  Transitional   Neighborhoods."  Journal  of  urban  Economics  32,  no.  2  (1992):  214-­‐232.   Follain,  James  R  and  Stephen  Malpezzi.  "Another  Look  at  Racial  Differences  in  Housing   Prices."  Urban  Studies  18,  no.  2  (1981):  195-­‐203.   Galster,  George.  "Racial  Discrimination  in  Housing  Markets  During  the  1980s:  A  Review  of   the  Audit  Evidence."  Journal  of  Planning  Education  and  Research  9,  no.  3  (1990):  165-­‐ 75.   Glaeser,  Edward  and  Joseph  Gyourko.  "Zoning's  Steep  Price."  Regulation  25  (2002):  24.   Hanson,  Andrew  and  Zackary  Hawley.  "Do  Landlords  Discriminate  in  the  Rental  Housing   Market?  Evidence  from  an  Internet  Field  Experiment  in  Us  Cities."  Journal  of  Urban   Economics  70,  no.  2  (2011):  99-­‐114.   Hanson,  Andrew,  Zackary  Hawley,  Hal  Martin,  and  Bo  Liu.  "Discrimination  in  mortgage   lending:  Evidence  from  a  correspondence  experiment."  Journal  of  Urban  Economics   92  (2016):  48-­‐65.   Heckman,  James  J.  "Detecting  Discrimination."  The  Journal  of  Economic  Perspectives  12,  no.   2  (1998):  101-­‐116.   Heckman,  James  J,  and  Peter  Siegelman.  1993.  “The  Urban  Institute  Audit  Studies:  Their   Methods  and  Findings.”  In  Clear  and  Convincing  Evidence:  Measurement  of   Discrimination  in  America,  edited  by  Michael  E  Fix  and  Raymond  J  Struyk,  187–258.   Washington,  DC:  Urban  Institute  Press.     Ihlanfeldt,  Keith  R.  "Exclusionary  Land-­‐Use  Regulations  within  Suburban  Communities:  A   Review  of  the  Evidence  and  Policy  Prescriptions."  Urban  Studies  41,  no.  2  (2004):  261-­‐283.     Kiel,  Katherine  A.,  and  Jeffrey  E.  Zabel.  "House  Price  Differentials  in  US  Cities:  Household  and   Neighborhood  Racial  effects."  Journal  of  housing  economics  5,  no.  2  (1996):  143-­‐165.     King,  A  Thomas  and  Peter  Mieszkowski.  "Racial  Discrimination,  Segregation,  and  the  Price  of   Housing."  The  journal  of  political  economy    (1973):  590-­‐606.  

Ladd,  Helen  F.  1998.  “Evidence  on  Discrimination  in  Mortgage  Lending.”  Journal  of  Economic   Perspectives  12  (2):  41–62.       Meltzer,  Rachel,  and  Jenny  Schuetz.  "What  Drives  the  Diffusion  of  Inclusionary  Zoning?."   Journal  of  Policy  Analysis  and  Management  29,  no.  3  (2010):  578-­‐602.     Munnell,  Alicia  H.,  Geoffrey  MB  Tootell,  Lynn  E.  Browne,  and  James  McEneaney.  "Mortgage   lending  in  Boston:  Interpreting  HMDA  data."  The  American  Economic  Review  (1996):   25-­‐53.   Myers,  Caitlin  Knowles.  "Discrimination  and  Neighborhood  Effects:  Understanding  Racial   Differentials  in  US  Housing  Prices."  Journal  of  urban  economics  56,  no.  2  (2004):  279-­‐ 302.   Rosen,  Kenneth  T.,  and  Lawrence  F.  Katz.  "Growth  management  and  land  use  controls:  The   San  Francisco  bay  area  experience."  Real  Estate  Economics  9,  no.  4  (1981):  321-­‐344.   Straszheim,  Mahlon  R.  "An  Econometric  Analysis  of  the  Urban  Housing  Market."  NBER  Books     (1975).   Ross,  Stephen  L.,  Margery  Austin  Turner,  Erin  Godfrey,  and  Robin  R.  Smith.  2008.  “Mortgage   Lending  in  Chicago  and  Los  Angeles:  A  Paired  Testing  Study  of  the  Pre-­‐Application   Process.”  Journal  of  Urban  Economics  63  (3):  902–19.     Ross,  Stephen  L.,  and  John  Yinger.  2002.  The  Color  of  Credit:  Mortgage  Discrimination,   Research  Methodology,  and  Fair-­‐Lending  Enforcement.  MIT  Press.     Schill,  Michael  H.,  and  Susan  M.  Wachter.  "Housing  Market  Constraints  and  Spatial   Stratification  by  Income  and  Race."  Housing  Policy  Debate  6,  no.  1  (1995):  141-­‐167.   Schuetz,  Jenny,  Rachel  Meltzer,  and  Vicki  Been.  "31  Flavors  of  Inclusionary  Zoning:   Comparing  Policies  from  San  Francisco,  Washington,  DC,  and  Suburban  Boston."   Journal  of  the  American  Planning  Association  75,  no.  4  (2009):  441-­‐456.   Schwartz,  Heather  L.,  Liisa  Ecola,  Kristin  J.  Leuschner,  and  Aaron  Kofner.  "Is  Inclusionary   Zoning  Inclusionary?  A  Guide  for  Practitioners.  Technical  Report."  RAND  Corporation   (2012).   The  Urban  Institute.  "Expanding  Housing  Opportunities  Through  Inclusionary  Zoning:  Lessons   From  Two  Counties."  Washington  DC:  US  Department  of  Housing  and  Urban   Development  (2012).   Turner,  Margery  Austin,  Stephen  L.  Ross,  George  C.  Galster,  and  John  Yinger.  "Discrimination   in  Metropolitan  Housing  Markets:  National  Results  from  Phase  I  HDS  2000."   Washington,  DC:  US  Department  of  Housing  and  Urban  Development  (2002).    

Turner,  Margery  Austin,  Claudia  Aranda,  Diane  K.  Levy,  Robert  Pitingolo,  Robert  Santos,  and   Douglas  Wissoker.  "Housing  Discrimination  Against  Racial  and  Ethnic  Minorities   2012."  Washington,  DC:  US  Department  of  Housing  and  Urban  Development  (2012).     Yinger,  John.  "Measuring  Racial  Discrimination  with  Fair  Housing  Audits:  Caught  in  the  Act."   The  American  Economic  Review    (1986):  881-­‐93.