Measuring the Consequences of Promoting Inner City Homeownership Jean L. Cummings and Denise DiPasquale City Research Matthew E. Kahn Fletcher School, Tufts University October 2001 Contact Author Matthew Kahn Fletcher School Tufts University Medford MA 02155. E-mail: email@example.com 617-627-4167 Keywords: Home Ownership, Community Development This research was funded, in part, by The Pew Charitable Trusts. The authors are solely responsible for the analysis and conclusions contained in this paper. We thank seminar participants at Syracuse and the January 2001 ASSA meetings for helpful comments. Measuring the Consequences of Promoting Inner City Homeownership Abstract This paper examines low- and moderate-income households in the city of Philadelphia who are becoming homeowners for the first time. We examine two Nehemiah developments subsidized by the City of Philadelphia that offer newly constructed homes at well below cost. This paper uses a unique survey of these new owners to measure what Nehemiah residents gain in terms of structure and community attributes as they make the transition from renting to owning. The new owners in the Nehemiah complex significantly improve their housing structures while raising their exposure to crime and weak local public schools. As part of the City’s community development strategy, these developments were expected to increase economic activity near these sites. We document that there is no evidence of “local benefit spillovers” for census tracts where the Nehemiah was built. Our survey results suggest that the new housing complex represents an “oasis” where there are few interactions between the new home owners and the incumbent residents of the greater community. 2 I. INTRODUCTION Homeownership has long been the centerpiece of U.S. housing policy. As a society, we have created many policies to promote homeownership, such as the mortgage interest deduction for federal income tax, mortgage insurance, and direct subsidies to lower income households. A major rationale for these policies is the conviction by many that homeownership builds better citizens and communities by giving residents a stake in those communities.1 This paper examines low- and moderate-income households in the city of Philadelphia who are becoming homeowners for the first time. The City of Philadelphia has made a strong commitment to promoting homeownership as part of its overall community development strategy (see OHCD 1997b). The City provides people-based subsidies for homeownership through its Settlement Grant program, which offers qualifying households up to $1,000 to help cover closing costs to purchase a home. In addition, the City provides place-based subsidies, which we focus on in this paper. We examine two Nehemiah developments subsidized by the City that offer newly constructed homes at well below cost—West Philadelphia Nehemiah and West Poplar Nehemiah. The Nehemiah name is taken from a federal housing program, Nehemiah Housing Opportunity Grant Program (NHOP), which promotes homeownership. While both projects share the Nehemiah name, only the West Poplar project actually received federal Nehemiah funds. With these place-based programs, the hope is that the housing will 1 There is little empirical evidence of the connection between homeownership and “good citizenship.” DiPasquale and Glaeser (1999) find some connection between homeownership and good citizenship as measured by involvement in local politics and nonprofit organizations. Green and White (1997) report evidence of greater educational attainment among children of homeowners relative to children of renters. 3 provide not only quality residences for its occupants, but also stimulate new investment in the area. Both Nehemiah developments are located in very distressed neighborhoods. Encouraging middle class households to move to high poverty areas is an alternative anti- poverty strategy to HUD’s Moving to Opportunity (MTO) program that encourages public housing residents to move to better neighborhoods.2 This paper focuses on the private and social gains from moving to these new housing complexes. For the minority middle class households who move into these two Nehemiah developments, what do they gain in terms of structure and community attributes? How do these attributes differ from what they consumed as renters? Most Nehemiah residents also receive a Settlement Grant to cover closing costs. For each Nehemiah participant, the alternative to moving into the complex would have been to use that grant to purchase a house elsewhere in the city of Philadelphia. We examine the community and structure consumption choices of black non-Nehemiah households. They represent a “control group” to help us measure the net gains from participating in the place-based program. The data assembled for this study permit a unique opportunity to examine the progress made by new homeowners in terms of structure and community quality when they make the transition from owner to renter. We are able to measure this progress because the City provided us with a database that includes geocodes for each Settlement grant participant’s origin and destination census tract. This information allowed us to assign local public goods consumption before and after the move. Unfortunately, the 2 For an analysis of the Boston MTO program see Katz, Kling and Liebman (2000). 4 City’s database includes little information about the structure attributes that these program participants consumed as renters or now own. To measure the structure gains, we have conducted a survey of 400 Settlement Grant recipients, as well as 76 Nehemiah residents, that asked detailed questions on the characteristics of the house purchased and the quality of the neighborhood. By choosing the Nehemiah developments, residents move to considerably worse communities than their renter neighborhoods. While our results indicate that the gains in structure are larger than the losses in community quality, the estimated dollar increase in housing consumption provided by the city to Nehemiah residents is substantially lower than the public subsidies allocated to these projects. As part of the City’s community development strategy, these developments were expected to increase economic activity near these sites. We study whether there were “local spillovers” for census tracts where the Nehemiah was built. If these projects generate spillovers, we would expect to see an increase in real estate prices near these sites. We estimate hedonic price regressions to measure whether the Nehemiah’s construction is capitalized into home prices using data on every real estate transaction record in the city from 1986 to 1997. We found no evidence of benefits to the surrounding community. Unlike previous hedonic evaluation studies, we use our new survey’s findings to further explore why we observe little local spillovers. Our survey measures respondent attitudes and perceptions of their new community. We find evidence consistent with the 5 hypothesis that the Nehemiah complex is an “oasis”. The middle class households who have moved to these high poverty census tracts view the new complex as their community. If such households have few interactions with their census tract neighbors who live out side the Nehemiah complex, and do not spend their disposable income at local stores, then there is no possible channel through which there can be spillover effects. In the next section of this paper, we give an overview of the Philadelphia housing market. In Section III, we provide a detailed description of the Nehemiah developments followed by an analysis of community choice in Philadelphia. In Section IV, we examine the characteristics of Nehemiah residents and their communities. In Section V, we measure how moving to the Nehemiah complex affects a household’s consumption of structure and community attributes. In Section VI, we test the hypothesis that the Nehemiah complex has improved the local communities that are adjacent to it. We conclude with a discussion of the merits of placed-based housing subsidies as a tool in community development. II. TRENDS IN THE PHILADELPHIA HOUSING MARKET 1986 to 1997 Households who moved into the new Nehemiah houses could have purchased a home in the housing market. To understand the opportunities in this market at that time, we look at time trends in the real price of housing throughout the City. Housing prices in the City of Philadelphia are low relative to other large American cities. Using data from 6 the Philadelphia Board of Revision of Taxation on every real estate transaction from 1986 to 1997, we estimate quality-controlled house prices. As shown in Figure 1, real house price appreciation in both the metropolitan area and the city of Philadelphia lagged behind the national average during the upswing of the 1990s. From 1989 to 1997, the national average annual real house price appreciation was 4%, metropolitan Philadelphia house price appreciation was 3%, and in the city of Philadelphia, the average was -2% per year. Figure 1 shows that Philadelphia center city ownership has been an awful financial investment relative to investing in the stock market as measured by the Dow Jones Industrials. Figure 2 shows that the real average home price in center city Philadelphia fluctuated between $40,000 and $55,000 between 1986 and 1997. The real price of Philadelphia housing has been falling from 1989 until 1997. Figure 2 breaks out the city’s house prices by sub-market. Each of the city’s 350 tracts is assigned to an income group according to its income level as given by the 1990 Census of Population and Housing. Figure 2 shows that the greatest price declines have taken place in the richest census tracts. Note that between 1989 and 1997, the price differential between homes in the richest tracts and the poorest tracts narrowed. III. NEHEMIAH DEVELOPMENTS IN THE CITY OF PHILADELPHIA The City of Philadelphia has long encouraged homeownership as part of its overall community development strategy. The first goal stated in the strategic plan developed by the Office of Housing and Community Development (OHCD) of the City of Philadelphia is “promoting homeownership and housing preservation.” The plan states that, “to more effectively support economic development and reinvestment in 7 Philadelphia, the City will continue to emphasize homeownership and preservation of the existing occupied housing stock” (OHCD 1997b, p. 9).3 Place-based programs such as the Nehemiah developments are designed to encourage reinvestment in inner city communities by making investments in specific developments. By revitalizing formerly declining central city areas, such programs may act as a magnet, attracting households and other investments that might have left the neighborhood in the absence of the incentives. In this paper, we focus on two housing developments supported, in part, by the City: West Poplar Nehemiah and West Philadelphia. Nehemiah. The Nehemiah developments offer newly constructed units to qualified households at prices substantially lower than construction costs. The Nehemiah Housing Opportunity Grants Program (NHOP) is a national program created under Title VI of the National Housing and Community Development Act of 1987. Under the program, HUD is authorized to make grants to non-profit organizations to provide loans to families purchasing homes that are constructed or substantially rehabilitated in accordance with a HUD-approved program. The non-profits sponsoring the developments are responsible for marketing and allocating the units, with federal guidelines regulating such things as eligibility and fair housing rules. By 1998, 1,874 units had been completed nationwide; an additional 57 units were to be completed in 1999.4 3 For an extensive discussion of the city’s homeownership and community development strategies, see Kromer (2000). 4 The loans may not exceed $15,000, must be interest free, must be secured by a second mortgage held by the HUD Secretary, and are repayable to the Secretary upon the sale, lease or transfer of the property. HUD funds must be applied to the purchase price of the home and HUD funds may not be used to provide the downpayment. The statute and regulations require that the eligible homebuyer make a downpayment equal to 10% of the purchase price, unless a local or state government will hold the first mortgage under a home- 8 The West Poplar Nehemiah project received funding from the NHOP program. The West Philadelphia project borrows the Nehemiah name, but did not receive any NHOP funding. Both of these new-construction projects received subsidies from the City of Philadelphia, including direct expenditures for some land acquisition and site improvements. West Philadelphia Nehemiah is at the corner of 46th and Market Streets, built on a 10 ¼-acre site at the edge of University City. It is across the street from a medical building, formerly the Institute of Pennsylvania Hospital; a tall fence separates it from nearby public housing high-rises. The developer is the Philadelphia Interfaith Action (PIA), a private non-profit coalition of 43 religious institutions from across the metropolitan area. The West Poplar project is in North Philadelphia, a few blocks north of the Convention Center and downtown area, at N. 12th, N. 13th, Poplar and Ogden Streets. The site is also near Yorktown, a community in North Philadelphia of over 600 single family homes built in the 1960s with substantial federal funds. The current development covers over 10 acres. Related City improvements included adding and reconfiguring some streets. The development includes a new village green and borders an existing church. Construction on the West Philadelphia project began in 1994 and was completed in the fall of 1997. Households began moving into the development in 1995. As shown in Table 1, the project consists of 135 units; 116 of the units have three bedrooms and 19 have four bedrooms. Homebuyers paid $49,500 to $52,000 for the three-bedroom units loan program of that unit of government. If such a program provides the mortgage, the homebuyer must pay whatever downpayment is required under that government's program (this may be less than 10%). The grantee would need HUD approval to allow the use of the state or local government program. (Source: Author’s correspondence with staff in the Office of Policy, Development and Research at the U.S. Department of Housing and Urban Development.) 9 and $55,000-$56,000 for the four-bedroom units. As shown in the Table, homebuyers paid significantly less for these homes than the cost of producing these units. The Redevelopment Authority estimates the “bricks and mortar” costs at $75,853 per unit, with total land costs and site improvements adding another $31,883 in costs. The total costs of these units averaged $107,736 per unit, with sales proceeds averaging $50,878 and public subsidies averaging $56,858 per unit. The West Poplar project has received funding under the NHOP program as well as additional funding from the City. The project was developed in two major phases; construction of the 75 units in the first phase began in 1996 with households beginning to move into these units that year. The second phase consists of 101 units with construction expected to be completed in the fall of 2000. All the units have three bedrooms. The sales price averages $59,881; prices for these units range from $57,000 in Phase I to $63,000 for the last units in Phase II. As shown in Table 1, the costs of producing these units are also considerably higher than these purchase prices. The Redevelopment Authority estimates the “bricks and mortar” costs at $121,061 per unit, with land and site improvement costs adding $42,346 per unit (although Redevelopment Authority staff indicate that total site improvement costs for Phase II are not finalized). The total costs of these units averaged $163,407 per unit, with sales proceeds averaging $59,881 and public subsidies averaging $103,526 per unit. As shown in Table 1, 14.5% of these subsidies came from NHOP. Total costs per unit of the West Poplar development are over 50% higher than for West Philadelphia. The public subsidies per unit are almost twice those used in the West Philadelphia project. 10 IV. COMMUNITY CHOICE IN THE CITY OF PHILADELPHIA Nehemiah housing units are brand new but the complex is located in some of the highest poverty census tracts in the City. To compare community quality in these tracts relative to the tracts that non-Nehemiah black first time home buyers choose, we have collected information for each of the 350 census tracts in the City of Philadelphia. Each census tract contains roughly 4000 people. We constructed data on each tract’s average school quality, exposure to crime, distance to the Central Business District and characteristics of tract residents. Table 2 reports Philadelphia residents’ exposure to different measures of community attributes. Using 1990 Census data, we calculate the weighted average of tract attributes using household shares as the weights. The table reports average exposure for all people, people who live in the Nehemiah tracts, and blacks and whites separately. We measure school quality by the average percentage of 8th graders who scored above the state median on the state standardized math test, using the public schools serving students in each census tract.5 We also include the average class size, using the number of students per teacher. The average City of Philadelphia household lives in a census tract where 12% of eight graders, on average, scored above the state median math score. In the census tracts 5 These test scores are available from the Pennsylvania Department of Education website (www.pde.psu.edu/esscores.html). The Philadelphia School Districts has 274 public schools; 90 include eighth grade. Scores for individual schools range from 0 to 99 (three have scores above 33). Private and parochial schools do not provide scores for these tests. Using maps provided by the school district, we assigned each public school with an eighth grade to the census tracts they served. If a tract was served by more than one school we assigned that tract the average of the test scores for those schools weighted by enrollment. Regional or district-wide schools where admission is not determined by a student’s home address, such as exam schools or charter schools, were excluded from these calculations. 11 that house the Nehemiah projects only 2.7% of eighth graders score above the state median in math. The average white lives in a census tract where 17.3% of children scored above the median while the average black lives in a census tract where 5.5% of children scored above the median. Across the city, on average, there are 19.1 students per teacher; in the schools serving Nehemiah tracts, the average class size is 19.8 students. Crime is measured by the average murders for 1994 and 1995 that occurred within each census tract per thousand in population. We average the 1994 and 1995 murder rates in an attempt to mitigate the effect of a fluke tragedy in a generally safe community. As shown Table 2, the average census tract experienced 0.28 murders per thousand of population.6 The data indicate that households living in Nehemiah tracts are exposed to considerably higher murder rates. The average black household lives in a tract that has a murder rate that is four times greater than the average white but the average black household lives in a tract that features 50% less murder than the Nehemiah tracts.7 A growing literature in economics and sociology has documented the positive 6 Data are from the Philadelphia Police Department Homicide Division and are available at www.philly.com; the information provides the address for each murder that occurred in the city. We geocoded these addresses to get murder rates in each census tract per thousand in population. 7 Murder may be viewed as a narrow measure of crime. We use murder because murder data are available by address of the crime. We were able to geocode these data to construct murder rates by census tract. These data provide a rare opportunity to examine differences in crime rates across very small geographic areas that closely mimic neighborhood boundaries. Broader measures of crime such as violent crime or property crime, as defined by the FBI’s Uniform Crime Reports, are available by police districts. However, there are only 26 police districts in the city, which would force us to aggregate 365 census tracts to the district level, losing a great deal of neighborhood detail. We used the district level data to test how well murder works compared with broader measures of crime. Those results suggest that murder is a good proxy for crime. 12 benefits from living near highly educated people (Case and Katz (1991), Cutler and Glaeser (1997), Rauch (1993)). We measure the socio-economic status of neighbors by using the tract’s percent of adults who have college degrees. As shown in Table 2, for the average Philadelphia census tract, 18% of its population over age 24 has a college degree. In neighborhoods chosen by Settlement grant recipients and housing the Nehemiah developments, far fewer adults have college degrees. As would be expected, the poverty rate in the Nehemiah tracts is much higher. The average black lives in a tract with a poverty rate of 27.8% while the average Nehemiah tract resident is exposed to a poverty rate of 57.9%. The average white lives in a tract featuring a poverty rate of 11.7%. Community quality is not just a function of service provision and neighbor quality. Shopping opportunities and access to jobs are other relevant criteria. The Philadelphia Board of Revision of Taxes provided an inventory of all properties in the city (Inventory). From these data, we calculate the percent of building area in each census tract used for commercial purposes (not including industrial uses).8 On average, 11% of real estate in a Philadelphia census tract is commercial; in the Nehemiah neighborhoods, 29% of real estate is commercial. The impact of commercial space on a community can be positive or negative. We expect a positive impact on community quality when basic commercial services, such as dry cleaning, a pharmacy, a grocery store, or a local restaurant are available. However, commercial activity not designed to serve residents could be viewed as detrimental to community quality. We measure accessibility to the city center as the distance in miles from the center of each census tract to the center of 8 Some other variables that would be useful in evaluating the impact of commercial space on community quality include such measures as vacant land and abandoned buildings as well as retail sales. Unfortunately, these data were not available by census tract. 13 census tract 5, which houses city hall. The average census tract is 5.2 miles from city hall. Nehemiah tracts are rather close to the city center. Each census tract in the City of Philadelphia differs with respect to their attributes along the five dimensions of crime, school quality, human capital, commercial real estate, and Central Business District accessibility. To form an index of how census tracts rank in terms of quality, we need index weights. Such weights can be estimated from hedonic housing regressions that yield estimates of the marginal price of purchasing a housing attribute such as a safer community. Following our earlier work (see DiPasquale and Kahn 1999), we estimate a City of Philadelphia hedonic home price regression as in equation (1). log(Priceijt) = B * Xijt + γ * Ζ jt + ε ijt (1) In equation (1), the dependent variable is the log of home i’s price in census tract j in year t. The vector X includes calendar year dummy variables and structure characteristics. The vector Z includes census tract level characteristics. Our data covers every real estate transaction that took place in the city of Philadelphia between 1986 and 1997 (Sales Journal) provided by the Philadelphia Board of Revision of Taxation. We have full information on 146,053 arm’s length home sales.9 These data are quite limited on structural attributes. Unfortunately, they do not include 9 Arm’s length excludes any transactions identified as those that are priced outside of the market—for example, sales between family members, or government sales of properties seized for nonpayment of taxes. 14 information on important standard measures such as the number of rooms, bedrooms, and baths. We know the total area of the lot, the height of the structure measured by number of stories, and whether or not the unit has a garage. In this data, the median price of homes purchased between 1986 and 1997 was $54,604 (in 1998 dollars). Table 3 reports the hedonic regression estimates. Our focus is on the ordinary least squares estimates of γ in equation (1).10 The results for the tract-level community attributes all impact home prices in the expected way, and all but distance to the city center are statistically significant. A ten-percentage point increase in the percent of the population with college degrees increases prices by 21.3%. Increasing the portion of 8th graders scoring above the state median in math by 10 percentage points increases home prices by 12.6%. Increasing class size by one student decreases house prices by 2.4%.11 As expected, an increase in crime, measured by murders, results in a decrease in house prices; an increase of one murder per 1,000 in population decreases house prices by 60.6%. This may seem like a very large impact, but one murder per thousand in population is a large increase in murders. Given that the average census tract has 4,000 households, our results imply that an increase of one murder in this average tract would decrease home prices by 15%. A 10 percentage point increase in the portion of real estate that is commercial in the tract raises house prices by 2.9%. A one-mile increase in the distance from the center of the city increases home prices by 1.3%, but this estimate is not statistically significant. 10 For a more detailed analysis of the structure hedonic coefficient estimates see the longer report (Cummings, DiPasquale and Kahn 2001). This hedonic approach for ranking community quality is quite similar to the approach used by Gyourko and Tracy (1991). 11 Black (1999) also finds a positive impact of schools on home prices, although her estimates are smaller than ours. Her models are based on data on suburban school districts in Massachusetts. 15 To form our measure of community quality, we take the OLS estimates of equation (1) and log-linearize the coefficients. Defining p as the marginal price of a community attribute (such as schools) and the characteristics of community as Z, we calculate the sum of p*Z. Based on this method, Philadelphia’s three best communities are Society Hill, Washington Square and Olde City while the three lowest ranked communities are Ludlow, Mantua and 5th and Lehigh. In Map 1, we map the community quality index and the location of the two Nehemiah complexes. The map highlights two points. First, there is enormous heterogeneity in this city’s quality of life. Second, the Nehemiah complexes have been built in two of the lowest ranked communities in the city. V. WHO ARE NEHEMIAH HOME BUYERS AND HOW DO THE NEHEMIAH AREAS COMPARE TO THE REST OF THE CITY? Up to this point, we have used traditional data sources to measure how Nehemiah tracts compare to other areas in the city and to estimate the marginal price of purchasing a better community. We are especially interested in how a family’s quality of life is affected by switching from renting to owning. This “before/after” comparison requires special data. We have data on over 8,000 households that participated in the Settlement Grant program. For each of these households we observe some demographic attributes 16 and the census tract the household chose. For roughly 4000 of these households, we also know their origin census tract.12 Table 4 presents some basic demographic information on over 8000 new Philadelphia home owners who participated in the Settlement Grant program. This includes 86 grant recipients who bought homes in either the West Philadelphia or West Poplar Nehemiah projects. All 86 grant recipients living in these projects are households headed by African Americans. As shown in the Table, females head 75.6% of these households. When they purchased their homes, these households had a median household income of $25,379 (in 1998 dollars), which is 30.5% higher than the median income of Settlement Grant recipients. They paid $49,129 for the homes and made a downpayment of $1,280, on average. Table 5 reports community average attributes for Nehemiah and non-Nehemiah households (broken out by race) in the origin and destination areas. Over 90% of Nehemiah households changed census tracts when they moved to these developments; on average, they moved 2.2 miles from their previous homes, farther than the average Settlement Grant household. The 76 households that changed census tracts when they moved came from 54 census tracts. As shown in Table 5, Nehemiah residents, on average, came from significantly better neighborhoods than their new owner neighbor- hoods. The census tracts that house the Nehemiah developments have populations with considerably lower incomes, lower house values, less education, and lower homeownership rates than the origin tracts for these households. The Nehemiah 12 Newberger (1999) worked with a subset of this data set to explore search method differences between whites and blacks and to document the income and racial attributes of the census tracts that program participants were entering and exiting. 17 households moved to census tracts that were much more racially segregated and in which no more than 3.5% of the population was white. Black non-Nehemiah households move to much whiter tracts, that are further from the city center. Table 6 repeats the results presented in Table 2 but now focuses on the local public goods consumption for Settlement Grant participants. Note that Table 2 is based on the population of the entire City in1990. Relative to other black Settlement Grant participants, new owners in the Nehemiah complex are exposed to poor schools, a three times greater murder rate, 50% less college graduates as neighbors and are significantly closer to the center city. Nehemiah tracts do feature significantly more commercial real estate than the tracts where the average non-Nehemiah black resides. To study how the transition from renting to owning affects black versus white consumption of local public goods, in Table 7 we report mean consumption before and after the move. Through switching tenure and community, blacks significantly increase the quality of their local schools and reduce their exposure to murder. White households also move to areas with slightly better schools and lower crime. Despite the fact that settlement grant recipients have roughly equal income, new white owners live in census tracts featuring better schools, and much lower murder rates than blacks. Tables 5 and 6 showed that Nehemiah tracts feature low quality community attributes. Constructing the Nehemiah complex would have little value added if the households who chose to move to the complex would have moved to the Nehemiah area census tracts even without the subsidy. If new homebuyers do not want to live in these tracts, then the only way to recruit middle class households to live there is to provide deep subsidies. In 18 1990 there were 25 tracts in the city of Philadelphia where the population was over 25% black and over 25% of the population has a college degree. This suggests that middle income black households could chose to migrate away from high poverty minority areas. To measure the probability that a Settlement Grant participant would choose to move to the Nehemiah tracts (without a place-based subsidy), we estimate a 204 dimensional discrete choice model of household choice of census tract.13 We estimate a conditional logit model to explain the probability that a black household chose a particular census tract. We model tract choice as a function of its racial composition (percent of residents who are black), income (the percent of its residents living in poverty) and educational attainment (the percent of residents who are college graduates). In Table 8, we report two sets of estimates of this conditional logit model.14 In the second set of estimates, we interact the census tract attributes with the household’s income. Based on the results in the right columns of Table 8, we predict that a black household with an income of $15,000 has only a 1.1 percent probability of moving to a Nehemiah census tract while the black household with an income of $40,000 is predicted to have an even smaller (0.11 percent) likelihood of moving to these tracts. This shows that black middle class households would not chose to move to the Nehemiah general area without a deep subsidy. 13 Since the households in the Settlement Grant program are lower-middle class, we restrict the choice of the 350 tracts in the City of Philadelphia to the set of 204 census tracts whose median home price in 1990 was less than $60,000 ($1990). 19 VI. SURVEY OF NEHEMIAH RESIDENTS AND SETTLEMENT GRANT RECIPIENTS The previous section used the City of Philadelphia’s database to examine the community choices made by Nehemiah households and by Settlement grant households. Unfortunately, the database includes no information on the home’s physical attributes. Using information from the Settlement Grant program and the Nehemiah housing developments, we surveyed 476 households (400 Settlement Grant recipients and 76 Nehemiah residents). This survey provides more detailed information on structure and community characteristics than standard census data. Unlike census-based or American Housing Survey studies, our survey provides a detailed picture of the quality of life for new homeowners before and after the move, identifying how much households upgrade their housing structures and their communities. Between November 1998 and January 1999, The Response Center, a Philadelphia based market research firm, called both Settlement Grant and Nehemiah households to conduct the survey. This firm followed standard procedures in sampling from a list of potential survey participants provided by City Research (Response Center 1999). The list included all Settlement Grant recipients between 1993 and 1997, as provided by the City program, for whom phone numbers could be found. From this list, households were randomly selected to be called with equal weights placed on each household’s probability of being called; surveys needed to be completed by a respondent who identified him/herself as a primary decision-maker in buying the home. Households were called 14 The conditional logit model is a standard limited dependent variable model estimated using stata’s clogit command based on maximum likelihood. 20 until we reached a target goal of 400 completed Settlement Grant surveys. In addition, we had a list of 140 Nehemiah residents for whom phone numbers could be found, taken from a 1998 end-of-year list of homeowners, provided by the two Nehemiah projects; the Response Center made at least one attempt (usually 3 to 5) to reach each of these households. The survey took an average 26.8 minutes to complete. Of the Settlement Grant recipients who answered the phone, 39% completed the survey. Of Nehemiah households who answered the phone, 74% completed the survey.15 This survey participation rate was higher than the usual rate for other studies conducted by The Response Center.16 In order to get a reasonable sample size for our analysis, we over- sampled the Nehemiah residents because the total number of project residents was small. In Table 9, we report basic demographic summary statistics for Nehemiah and non-Nehemiah households. In our survey, 23% of Settlement Grant households are white, 57% are black and 12% are Hispanics, as shown in Table 9. Another 8% are other race or did not identify race.17 The average household has 3.4 members; 79% have children under age 18. Settlement Grant recipients in our survey have a median income of $27,500; median income for white households is $32,500, $27,500 for black households and $22,500 for Hispanic households. As shown in Table 9, these households had a median household income of $37,500, 36% higher than the incomes for the average black Settlement Grant recipient. The median house prices and downpayment amounts are also 15 476 households completed the survey. An additional 243 households refused at the onset to participate, another 191 refused due to language difficulties, 75 refused claiming that they were not the household decision maker, and 30 more hung up during the survey. An additional 100 refused because they claimed that they did not receive a Settlement Grant, even though the OHCD data indicate they did receive a grant. 16 We had the opportunity to listen in on over ten phone calls to potential respondents. We were struck by the willingness of these new homeowners to discuss their tenure transition. 17 Household race is defined as the race of the person taking the survey, who by definition was a key decision-maker in buying the home. 21 significantly higher than those for grant recipients. Nearly 94% of Nehemiah households paid for the downpayment with savings; none reported assistance from friends or family. On average, 28% of adults in these households have completed at least some college. (This compares with 12% for all Settlement Grant recipients, and 19% for black Settlement Grant recipients.) The average household size is 2.8, with 58% of these households having children under age 18; 45% are female-headed households. Of the 76 survey respondents living in the Nehemiahs, 75 indicated that they are black and 1 refused to provide her race. Table 10 reports the means for housing structure attributes that Nehemiah households consumed as renters versus as owners. These households increased the number of rooms, bedrooms, and bathrooms, and most gained off street parking and central air conditioning by moving into the Nehemiah complexes. To compare to the Nehemiah household gains relative to the gains of black Settlement grant households, we turn to Table 11. Table 11 provides statistics describing the housing structures purchased by Settlement Grant recipients surveyed. These households paid a median price of $39,493 for their homes, though white owners paid nearly $10,000 more than black or Hispanic owners. The average house has 7.6 rooms, 3.0 bedrooms, and 1.3 bathrooms. Black owners have slightly larger homes by these measures than whites or Hispanics. The vast majority of homes are older than 20 years and are single-family attached houses for all races. Only 11% of these units have air conditioning and 42% have a garage; the figures are very similar across races. 21% of those surveyed complained about structural defects such as leaks; whites were more likely to report problems with leaks or electrical 22 problems than blacks or Hispanics. Virtually all Settlement Grant recipients and Nehemiah households surveyed are satisfied with their new homes. While Nehemiah households enjoy structure gains, they have made community sacrifices. Evidence that they have moved to worse communities is presented in Table 12. Nehemiah residents surveyed certainly indicate that they have problems with the location. Only 26% of residents rated the local schools as good; 66% have taken precautions against crime and 84% indicate that litter and abandoned buildings are a problem. Even with these problems, 65% of Nehemiah households said that they were satisfied with their neighborhood, the same percentage as that for Settlement Grant recipients. This high rate of neighborhood satisfaction, given the low satisfaction with local schools, neighborhood aesthetics and crime control, may be due to several factors. A smaller portion of Nehemiah households have school-aged children than Settlement Grant households, which may decrease concern over school quality. In addition, over half of Nehemiah residents indicated that it was very important to them that their new home was in a development with lots of new homes. These households may view their neighborhood primarily as the Nehemiah development and see the surrounding area included in the census tract as less important. Evidence that black Settlement grant households are improving their communities is presented in Table 13. Households in our survey chose neighborhoods with higher average household incomes, higher house values, and lower poverty rates than their previous communities (although white households saw little increase in neighborhood income). On average, households moved to communities with whites 23 representing a larger fraction of the total population than in their previous communities. While Nehemiah households clearly gained more housing structure amenities from becoming homeowners than their Settlement Grant counterparts, they moved to communities with far fewer neighborhood amenities. Comparisons of Nehemiah and Settlement Grant neighborhoods indicate that Nehemiah neighborhoods have much greater poverty rates and considerably lower average house prices, income, and homeownership rates. Their new neighborhoods are more segregated; unlike the neighborhoods of Settlement Grant recipients, the Nehemiah communities are virtually all black. Nehemiah offers a mixed opportunity; excellent structure in a low quality of life community. Given the large subsidy provided to Nehemiah residents, is purchasing a home in these developments a good deal? What is the market value of these homes? We can estimate the market value using the statistical model from our survey similar to the model we estimated in Table 3 using the sales transaction data. While the sales transaction data provides a larger sample, they have limited structure characteristics. One advantage of our survey sample is that we collected considerably more information on structure characteristics. Table 14 presents this survey based hedonic which we use to estimate the price the Nehemiah homes. Our price estimates indicate that, all else being equal, the average estimated market value for the Nehemiah homes is 23% less than the average purchase price of these homes. These results imply that even with the deep subsidies provided, the portion of costs paid by Nehemiah residents are considerably more than the market value of the homes. 24 Why would these residents pay a premium to live in these homes? These homes offer a quite unique housing opportunity in the city of Philadelphia. There are very few opportunities to live in a new home in a new development. Because these homes are so unique our model may underestimate the value of new construction in a new development. In addition, new residents may have an expectation that the community is on the rise and thus current crime levels and school quality levels are not reflective of where the community will be in a few years. Residents may expect that the placement of the Nehemiah complex will upgrade the community through increasing local purchasing power and this in turn will attract new commercial activity. While our data indicate that it is costly to choose to live in Nehemiah, as measured by the possible gains from living in other communities, it is possible that Nehemiah households have not engaged in an intensive search and thus “do not know what they are missing”. Our survey asked households how many homes were considered, in how many neighborhoods, and whether they looked in the suburbs and what information sources they used in their search. Table 15 shows some summary statistics on where households searched for new homes. In our survey, 15% of Settlement Grant households looked at homes in the suburbs. While about 14% of blacks and whites searched in the suburbs, only 8% of Hispanic households looked in the suburbs. 26% of Nehemiah households, though, looked in suburbs. There are differences across households in how many homes they saw during their housing search. The average Settlement Grant household looked at 7.1 homes, with white, black, and Hispanic households looking at 8.0, 7.1 and 6 homes, respectively. Nehemiah residents looked at only 4.3 homes. On average, Settlement Grant households, 25 regardless of race, looked at homes in only one or two neighborhoods. In our survey, 70% of Settlement Grant households used realtors, 26% used friends and relatives, 24% used newspapers, 9% used neighborhood organizations, and 9% used the Settlement Grant program’s housing counselors. There is some variation in these search methods across races. Black and Hispanic households were more than twice as likely as whites to use a counselor (around 10% versus 4%). Nehemiah households were considerably less likely to use realtors. They were much more likely to use friends/relatives, newspapers, neighborhoods organizations, and churches in their search. This is not surprising given that both Nehemiah projects were developed by nonprofit community organizations that relied heavily on church networks to market the developments. The differences in search may, at least in part, explain why Nehemiah households chose to their homes despite the obvious problems with their locations. The presence of discrimination in the housing market could also explain their locational choice. If minority households feel constrained in terms of their housing options, the Nehemiah developments would be more attractive. Our data provide no direct test of housing market discrimination. In our data, we found no evidence of the presence of price discrimination. Since housing discrimination can take many forms, we cannot determine the extent to which discrimination influenced the housing choices of Nehemiah residents. 26 VII. MEASURING THE SOCIAL BENEFITS OF NEHEMIAH As shown in Tables 10 and 12, Nehemiah households sharply changed their housing consumption upon making the transition from renter owner. Based on the hedonic price regression reported in Table 14, we calculate that through subsidizing the Nehemiah development, the City has increased housing consumption of Nehemiah residents by $10,496.18 This increase is quite small relative to the per-unit subsidies reported in Table One. This means that the City could have achieved the same increase in Nehemiah participant housing consumption at much lower cost if it had simply given the program participants a cash rebate of $10,496 to spend in the Philadelphia housing market. The City did not chose this option because it views Nehemiahs as playing an important role in redeveloping low quality of life communities. Thus, an important policy question is whether these Nehemiah projects have had a positive impact on the areas surrounding these developments. For the typical household who lives just outside the Nehemiah areas, has their quality of life been improved by the presence of these projects? A possible benefit for local residents near these developments is the impact of having new residents with incomes significantly higher than the average incomes in the neighborhood. Nehemiah residents have incomes that are on average close to three times that of the census tract median household income. Potentially, the presence of these residents could attract better commercial and shopping opportunities. There has been little evidence of this. To document this, we have used City of Philadelphia deeds data on 27 the price and quantity of real estate transactions by census tract by year. If more properties changed hands at a higher price in the Nehemiah tracts after the construction of Nehemiah than in the pre-Nehemiah period (1987-1994 in our data), we would conclude that the Nehemiah development provided external benefits to nearby residents. Through our analysis of the real estate transaction data, we have found no evidence that as the Nehemiah projects were built and sold that there was a growth in commercial real estate transactions or values in or adjacent to the Nehemiah census tracts.19 In fact, we find that the level of real estate activity and the trends in real estate prices in the Nehemiah tracts resemble those in other poor tracts in the city. To more carefully document this fact, we estimate hedonic home price regressions to compare price appreciation in Nehemiah census tracts versus two sets of control tracts. The first control group is the set of census tracts that have poverty rates greater than 50% and are located more than 1.2 miles away from the nearest Nehemiah census tract. We chose this group in order to compare trends in the Nehemiah area to other poor tracts in the city. These tracts are similar to the Nehemiah tracts in terms of poverty rates but far enough away that they are unlikely to be influenced by the Nehemiah developments. The second control group is the set of census tracts that share a border with at least one Nehemiah tract. By definition, these tracts are very close to the Nehemiah tracts and may 18 This calculation is based on calculating the market price of each housing attribute and multiplying this by the change in each housing attribute upon moving to the Nehemiah complex. Defining p as the marginal price and ∆q as the change in housing consumption, we calculate p*∆q and sum over the housing attributes. 19 Specifically, there have been virtually no private market commercial transactions in the data for property in the census tracts of the Nehemiah developments, or in adjoining and nearby tracts. There were a handful of non-market commercial transactions in the adjacent area (e.g., a $1 transaction), but the level of these kinds of transactions remained constant. During a tour of the Nehemiah projects and their surrounding areas we observed two recent commercial developments near the West Philadelphia project that included a grocery store and a drug store. In the area surrounding the West Poplar development there was a new drug 28 represent the best control group. We estimate two hedonic regressions using the Sales Journal micro data that have the following form; log(Priceijt) = B * Xijt + γ * Ζ jt + Nehemiah + control + U (2) In equation (2), the dependent variable is the log of home i’s price in census tract j at time t. In this equation, X represents a set of year dummies, community attributes, and structure attributes identical to the set discussed in Table 3 using the same data. The results are presented in Table 16. The major difference between the results in Table 3 and Table 16 is that we have added four dummy variables. One dummy variable equals one if the property lies in a Nehemiah tract and zero otherwise (called Nehemiah). The other dummy variable is an indicator of whether the property lies is a control tract. Both the variable “Nehemiah” and the variable “control” are interacted with a time dummy “post” that indicates whether the Nehemiah complex has been built. We test the hypothesis that the Nehemiah tracts experienced greater appreciation than other high poverty tracts during the same time period. As shown in Table 16, a quality-adjusted home in the Nehemiah tracts has increased in price by 12% after the Nehemiah complex was constructed. The coefficients on control group #1 (the distant high poverty tracts) shows an increase in price of 22.3% relative to the same tracts before the Nehemiah was built. This indicates that relative to other high poverty tracts that the Nehemiah poverty tracts appreciated by less than similar tracts once the Nehemiah complex had been built. Turning to the regression on the right side of Table 16, Nehemiah tracts homes increased by 11.7% post-Nehemiah while bordering census tract houses increased in price by 9%. store. Still, these commercial activities were too limited to have a noticeable impact on census tract real 29 We cannot reject the hypothesis that these two estimates are the same. This price test provides no evidence of a residential price effect. As a robustness test, we have estimated these same regressions using census tract fixed effects rather than our community measures and we find the same results. Why have we found such a small impact of Nehemiah on adjacent property prices? Some have argued that it may be too soon to see the impacts of these projects on commercial real estate prices. However, since these developments are occupied, we believe that the forces influencing neighborhood property values are greatest when the developments are new. In our survey, a major attraction to buyers of these units was that they were new and part of a new development. As they age, these properties may well be less likely to generate new activity.20 It is possible that the Nehemiah complexes have a positive effect on real estate prices in their immediate surroundings. Ellen, Schill, Schwartz, and Susin (2001) find that Nehemiah developments in New York City have a positive impact on house prices in small concentric circles (i.e. ¼ mile radii) around the developments. We focus on the census tract impact of new Nehemiah construction because in the Nehemiah complex census tract areas, there are very few housing transactions taking place. Post-construction of Nehemiah, there were only 23 transactions between 1995 and 1997 in these areas. Measuring the total effect on adjacent census estate markets. 20 To explore this point further, we examined real estate prices in and surrounding the Yorktown community, which is often cited as a successful inner city development. Yorktown was developed in the 1960s and early 1970s (OHCD 1996). Using our sales transaction data for 1986-1997, we examine the impact of Yorktown on the surrounding area 20 years after the development was opened. We find that house prices in Yorktown have been flat over the period. Looking at the census tracts surrounding Yorktown we find no impact of proximity to Yorktown on real estate prices in neighboring areas. This evidence suggests that there are no benefits associated with proximity to Yorktown to the surrounding communities as Yorktown has aged. We do not, however, have comparable data to analyze trends during earlier years. 30 tracts captures the impact of this development on thousands of the households. If this program had a local “macro” impact, this approach would capture this while if the Nehemiah program raises a single block’s value then the census tract as a whole will not experience much appreciation. It is also possible that the New York Nehemiahs have more of an impact because the local economy is booming. The Nehemiah developments may eliminate blight in areas with potentially valuable land. However, Lee, Culhane, and Wachter (1999) find modest positive impacts of FHA housing and Section 8 new construction on nearby areas in the City of Philadelphia. Again, this research examines small concentric circles around the developments. In addition, many of the units developed under these programs are in considerably better neighborhoods than the Nehemiahs examined in this paper. It is possible that the Nehemiah developments are too small to have a significant impact on such devastated neighborhoods. Hedonics based research is a useful tool for establishing the dynamics of real estate prices in the presence of “treatments” such as the construction of the Nehemiah complex. Such an approach cannot answer the question of “why” does a particular urban renewal program work or not work. We use our survey to dig deeper into why the Nehemiah complex has not had a larger impact on its greater community. One possible reason why the Nehemiah has so far had little impact on the surrounding community is that the Nehemiah complex owners have few social interactions with their high poverty neighbors. They may have little impact on the commercial activities in the community if they do not shop there or pursue local amenities such as restaurants. In terms of social interactions, neighborhood children could benefit from interacting with children from the Nehemiah complex. Playing, 31 studying or socializing with children from higher income, more educated families could result in role models that these children would not have access to if the Nehemiah development did not bring these families to the community. This peer group effect would be minimized if Nehemiah residents discouraged their children from playing with children from the community or if the Nehemiah children attend schools outside of the community. While we do not have time diary data on how households allocate their time, we have collected data on the propensity for survey respondent children to attend local public schools. An important fact is that Nehemiah residents are less likely to have school-age children than the average black Settlement Grant survey participant and those Nehemiah residents with children are less likely to have their children attend the local schools. Only 58% of Nehemiah households have school-age children, while 79% of Settlement Grant residents have school-age children. Nearly 60% of Nehemiah households with children send them to the local public schools while 79% of Settlement Grant households send their children to local public schools. This is suggestive evidence that there will not be a large child-to-peer-group effect for local children who live in the Nehemiah vicinity but not in the complex. There are not that many children among the Nehemiah residents and these children are less likely to attend local public schools. This suggests that Nehemiah households may be less concerned about the quality of local schools and, therefore, willing to locate in the Nehemiah developments where they know that local school quality is low. Nehemiah households may define their communities in terms of the other residents of these developments rather than the larger surrounding community. To study this “oasis effect”, in Table 17 we report results from twelve separate probit models. The 32 reported coefficients represent marginal probabilities using stata’s dprobit option. The first six probit models focus on how Nehemiah residents perceive the quality of their community as compared to non-Nehemiah residents’ impressions. The omitted category in these regressions is white non-Nehemiah resident. Nehemiah residents recognize that the schools and the shopping in their area are awful. Relative to a Settlement grant black respondent, the average Nehemiah respondent is 27 percentage points less likely to claim that the local schools are good and is 45 percentage points less likely to claim that local shopping is good. Perhaps more interestingly, relative to the observationally identical black Settlement Grant respondent, the Nehemiah resident is 6.6 percentage points more likely to say that neighbors are helpful and is 12 percentage points less likely to announce that there is undesirable street activity in the community. This is suggestive evidence of the “oasis” effect. Despite the fact that tracts with higher murder rates have statistically significantly higher levels of undesirable street activity (see column 6) and that Nehemiah tracts have high crime, the average Nehemiah resident views the complex not the area as the “community”. The last six columns of Table 17 report crime self-precaution strategies. Note that in these models, we control for the census tract’s murder rate. All else equal, people who live in high crime areas avoid public transit, are less likely to go out alone, and are more likely to own an alarm and not go out at night. Again, we find some evidence of a Nehemiah complex “oasis” effect. Despite the fact that they live in a high crime area, Nehemiah households are less likely to avoid public transport and more willing to go out at night than other black households in equally high crime areas who do not live in the complex. 33 VIII. CONCLUSIONS The City of Philadelphia has made a substantial commitment to promoting homeownership. In this paper, we take a detailed look at the transition from renter to owner for low- and moderate-income households who have participated in two major Philadelphia homeownership programs. Our ability to examine the housing structures and neighborhoods of the same households as both renters and owners allows us to observe the progress these households made by becoming homeowners. For Nehemiah households, their gains in housing structure come at a substantial cost to them through a marked decline in community quality of the surrounding area from their previous neighborhood. Our survey results indicate that Nehemiah households recognize the problems with their location but seem to view the gains in structure as worth the cost. In part, their view of this trade-off may be due to the fact that these households seem to define their community as the Nehemiah development itself, rather than as the larger area around their homes. We find no evidence that these developments have spurred private investment or raised home prices in nearby areas. It is possible that the Nehemiah development’s scale is too small to have an impact on its surrounding community. While the Philadelphia Nehemiahs have not impacted their local community, there is evidence that similar developments have had positive impacts on the surrounding community in New York City. Future research should investigate the conditions in the local economy that might explain these differences. We hypothesize that a local area would experience the largest increase if the local economy is booming and an area with potentially valuable land has not been developed because of some reversible blight such as a Brownfield or abandoned buildings. In addition, future 34 research should explore the extent to which project scale or the initial condition of the neighborhood influence the size of the impact of place based subsidies on surrounding communities. To revitalize high poverty communities, alternative types of place-based subsidies should be considered. The Nehemiah projects have provided deep subsidies to a small set of households. It is possible that other place-based programs, such as land clearance and site improvement (mirroring the Environmental Protection Agency’s Brownfields program), could achieve the goal of urban renewal at a lower cost to the City and provide more widely distributed benefits. Consideration should also be given to the trade-off between using scarce subsidy dollars for homeownership programs versus using those dollars to improve community quality through programs that reduce crime, improve schools, or increase retail activity. Improvements in community quality should increase demand for housing in these communities and have a direct immediate benefit for the families already living there. 35 REFERENCES Black, Sandra. 1999. Do Better Schools Matter? Parental Valuation of Elementary Education. Quarterly Journal of Economics 114(2): 577-599. BUREAU OF THE CENSUS. 1993. 1990 Public Use Microdata Samples Codebook. Chapter 5 and Appendices. Washington, D.C.: The Bureau of the Census. Revised June 1993. Case, Anne and Larry Katz. 1991. The Company You Keep: The Effect of Family and Neighborhood on Disadvantaged Youth. NBER Working Paper 3705. Cutler, David and Ed Glaeser. 1997. Are Ghettos Good or Bad? Quarterly Journal of Economics 112(3): 827-871. DiPasquale, Denise and Ed Glaeser 1999. Incentives and Social Capital: Are Homeowners Better Citizens? Journal of Urban Economics 45: 354-384. DiPasquale, Denise and Matthew Kahn. 1999. Measuring Neighborhood Investments: An Examination of Community Choice. Real Estate Economics, 27(3): 389-424. Ellen Ingrid Gould, Michael Schill, Amy Schwartz and Scott Susin. "Do Homeownership Programs Increase Property Values in Low Income Neighborhoods?" forthcoming in Journal of Housing Research. Green, Richard and Michelle White. 1997. Measuring the Benefits of Homeowning: Effects on Children. Journal of Urban Economics 41(3), 441-461. Gyourko, Joseph; Joseph Tracy, Joseph. 1991. The Structure of Local Public Finance and the Quality of Life. Journal of Political Economy, vol. 99, no. 4, August, pp. 774-806 HUD (DEPARTMENT OF HOUSING AND URBAN DEVELOPMENT). 2000. FY 1998 Income Limits. Washington, D.C.: www.huduser.org. Katz, Lawrence, Jeffrey Kling, and Jeffrey Liebman. Moving to Opportunity In Boston: Early Results of a Randomized Mobility Experiment. NBER Working Paper 7973, 2000. Kromer, John. 1999. Neighborhood Recovery: Reinvestment Policy for the New Hometown. New Brunswick, NJ: Rutgers University Press. Lee, Chang-Moo, Dennis P. Culhane, and Susan M. Wachter. 1999. The Differential Impacts of Federally Assisted Housing Programs on Nearby Property Values A Philadelphia Case Study. Housing Policy Debate Volume 10(1) 75-97. Newburger, Harriet. 1999. Mobility Patterns of Lower Income First-Time Homebuyers in Philadelphia. Cityscape: A Journal of Policy Development and Research 4(3): 201- 220. OFFICE OF HOUSING AND COMMUNITY DEVELOPMENT (OHCD). 1996. Learning from Yorktown. City of Philadelphia Report. 36 OFFICE OF HOUSING AND COMMUNITY DEVELOPMENT (OHCD). 1997a. Settlement Grant Program: Philadelphia Neighborhood Homebuyers. City of Philadelphia Report. OFFICE OF HOUSING AND COMMUNITY DEVELOPMENT (OHCD). 1997b. Year 23 Consolidated Plan (Fiscal Year 1998). City of Philadelphia Report: 9. Rauch, James. 1993. Productivity Gains from Geographic Concentration of Human Capital: Evidence from the Cities. Journal of Urban Economics 34(3): 380-400. RESPONSE CENTER. 1999. The Housing Study Methodology. Correspondence from Dan Margherita and Eleanor Sivick, The Reponse Center, Upper Darby, PA, July 1999. 37 Figure 1 Rates of Return For Housing and Stocks 30% 20% Average Annual Growth Rate 10% 0% -10% 78 80 82 84 86 88 90 92 94 96 98 Year Dow Jones Industrial Average National House Prices Philadelphia Metropolitan House Prices City of Philadelphia House Prices Note: Data are presented as three-year moving averages. Sources: Dow Jones Industrial Average provided by Dow Jones; National and Philadelphia metropolitan area indices from Freddie Mac/Fannie Mae series; City of Philadelphia from authors' calculations from Sales Journal data, City of Philadelphia. Figure 2 House Prices in City of Philadelphia from 1986-1997 (1998$s) $100,000 $80,000 House Price Richest Tracts 2nd Richest Tracts $60,000 2nd Poorest Tracts Poorest Tracts All Tracts $40,000 $20,000 86 87 88 89 90 91 92 93 94 95 96 97 Year Source: Authors' calculations on 146,053 households with data on all variables, from Sales Journal data, City of Philadelphia. Table 1 Nehemiah Projects West Philadelphia (PIA) West Poplar Total for Both Projects Phases I (75 units) & II (101 units) Total Number of Units = 135 Total Number of Units = 176 Total Number of Units = 311 No. of No. of No. of Units Per Unit Totals Units Per Unit Totals Units Per Unit Totals COSTS Bricks and Mortar 135 176 311 Construction 67,053 9,052,161 104,050 18,312,743 87,990 27,364,904 Prof. & Mgt 4,692 633,395 13,823 2,432,800 9,859 3,066,195 Holding Costs 3,096 417,968 1,257 221,240 2,055 639,208 Financing Costs 1,012 136,587 1,932 339,970 1,532 476,557 Bricks and Mortar Subtotal 75,853 10,240,111 121,061 21,306,753 101,437 31,546,864 Land and Site Improvements Public Improvements* 19,259 2,600,000 4,602 810,000 * 10,965 3,410,000 * Land Acquisition 12,624 1,704,250 37,744 6,642,860 26,840 8,347,110 Land and Site Subtotal 31,883 4,304,250 42,346 7,452,860 37,804 11,757,110 TOTAL COSTS 107,736 14,544,361 163,407 28,759,613 139,241 43,303,974 RESOURCES SALES PROCEEDS Unit Type 3 Bedrooms 58 49,500 2,871,000 64 57,000 3,648,000 20,961 6,519,000 39 50,500 1,969,500 11 60,000 660,000 8,455 2,629,500 8 51,000 408,000 44 60,000 2,640,000 9,801 3,048,000 11 52,000 572,000 57 63,000 3,591,000 13,386 4,163,000 4 Bedrooms 16 55,000 880,000 2,830 880,000 3 56,000 168,000 540 168,000 TOTAL SALES PROCEEDS 50,878 6,868,500 59,881 10,539,000 311 55,973 17,407,500 SUBSIDIES Federal Nehemiah Funds 176 15,000 2,640,000 8,489 2,640,000 Additional Public Subidy 56,858 7,675,861 88,526 15,580,613 74,780 23,256,474 TOTAL SUBSIDIES 56,858 7,675,861 103,526 18,220,613 83,268 25,896,474 * Public Improvements totals do not include West Poplar Phase II figures, which are not finalized. Source: Memos from Redevelopment Authority of the City of Philadelphia, February through April 2000. Table 2 Differentials in Community Characteristics Demographic Group Census Tract Attribute All Nehemiah Blacks Whites Tracts % White 0.523 0.041 0.130 0.836 % Black 0.394 0.916 0.818 0.098 % College Graduates 0.147 0.055 0.113 0.176 % Home Owners 0.633 0.235 0.598 0.668 % in Poverty 0.205 0.579 0.278 0.130 Murders Per Thousand People 0.275 0.861 0.452 0.117 Public School Quality 12.047 2.637 5.531 17.265 Average Class Size 19.089 19.805 19.935 18.449 % Commercial Real Estate 0.111 0.289 0.110 0.112 This table is constructed by taking census tract level data and calculating weighted means using each demographic group's tract count as the weights. Its entries represent the average exposure to a community attribute for the average member of the demographic group. Public school quality is defined in the text. Table 3 House Price as a Function of Structure Characteristics and Community Quality Regression Model Based on Philadelphia Sales Transaction Data Dependent Variable: Log of Price Structure Characteristics Log of Total Area 0.227 * (0.017) Number of Stories1,2 2 0.003 (0.027) 3 0.098 * (0.039) 4 or more 0.482 * (0.085) Garage1 0.286 * (0.027) Community Characteristics % Students Scoring Above State Median in Math Tests 1.264 * (0.204) Classroom Size -0.024 * (0.006) Murder Rate per 1000 Persons -0.606 * (0.125) % of Adults with Bachelor's Degree 2.133 * (0.151) % of Total Building Area That Is Commercial 0.285 * (0.156) Miles from City Hall 0.013 (0.009) Constant 7.943 * (0.235) Adjusted R-Squared 0.595 3 Observations 146,053 Notes: Standard errors are in parentheses. *Asterisk is significant at 10% level. 1. Dummy variables. 2. Missing category is one story. 3. Hedonic was run on 146,053 residential properties with at least one arm's length transaction between 1986 and 1997 that have data on all variables in hedonic. In addition, to avoid misleading outliers, transactions with real 1998 dollar prices less than $5,000 or more than $1 million were not included in analysis. Math Scores are percent of 8th graders in census tract scoring above the percent of 8th graders scoring above the state median in the standard math tests; Classroom Size is total enrollment divided by number of teachers; Murder Rate is the average of murders for 1994 and 1995 per 1000 population; Education Level is percent of adults in census tract above 24 years old who have a college degree; Commercial Space is percent of census tract building area that is commercial; Distance from City Hall is distance in miles from center of tract to center of census tract #5, which houses city hall. Source: Authors' calculations of City of Philadelphia Sales Journal and Inventory data, and 1990 Census. Map 1 Community Quality Index by Tract Estimated Value of Community Quality 40,000 to 107,800 (51) 0 to 40,000 (127) -15,000 to 0 (70) -50,000 to -15,000 (96) -323,000 to -50,000 (6) Not Available (17) Location of Nehemiah Projects Table 4 Household Characteristics from Settlement Grant Program Data, 1993-1997 (Status at Time of Grant Application) Settlement Grant Recipients Nehemiah 1 All White Black Hispanic Other All2 Percent of Households 100.0% 19.8% 46.1% 28.1% 4.6% 1.1% Mean Household Size 3.0 2.6 2.9 3.2 4.4 2.9 (1.5) (1.4) (1.4) (1.5) (2.2) (1.5) Percent Female-Headed Household 64.2% 51.1% 72.7% 62.1% 47.4% 75.6% (48.0) (50.0) (44.6) (48.5) (50.0) (43.2) Percent Elderly 1.2% 1.7% 0.8% 1.2% 2.2% 1.2% (10.8) (12.7) (9.1) (11.1) (14.7) (10.8) Percent Handicapped 2.0% 1.5% 1.0% 2.8% 9.3% 0.0% (14.1) (12.3) (10.1) (16.5) (29.1) (0.0) OHCD Median Household Income (98$s)* $19,448 $21,643 $21,405 $15,213 $16,135 $25,379 OHCD Median House Price (98$s)* $42,654 $46,750 $44,992 $36,361 $40,570 $49,129 OHCD Median Downpayment (98$s)* $1,239 $1,523 $1,263 $1,039 $1,309 $1,280 Median House Price/Income 2.3 2.3 2.3 2.5 2.9 1.9 Percent Who Changed Tracts 80.8% 70.5% 85.5% 83.6% 61.8% 90.5% (39.4) (45.6) (35.2) (37.1) (48.7) (29.5) 3 Distance Travelled (miles) 1.88 1.42 2.30 1.63 1.06 2.19 (3.21) (2.99) (3.58) (2.33) (2.83) (3.31) Notes: Based on 8,059 households with address information. Standard deviations are in parentheses. * Household income, house price and downpayment are as reported by OHCD at time of grant, converted to real dollars. These values may vary from values reported by households in authors' survey. 1. 113 households did not identify race. 2. All 86 Nehemiah households had black heads-of-house. 3. Distance from center of previous tract to center of current tract. Source: Settlement Grant Recipient data, the Office of Housing and Community Development of the City of Philadelphia. Table 5 Comparison of Renter (Origin) and Owner (Destination) Neighborhoods by Race Settlement Grant Nehemiah All White Black Hispanic Renter Owner Renter Owner Renter Owner Renter Owner Renter Owner (origin) (destination) (origin) (destination) (origin) (destination) (origin) (destination) (origin) (destination) Mean Distance from Old Census 1.89 1.42 2.30 1.63 2.19 Tract (miles) (2.01) (1.87) (2.24) (1.46) (2.07) Portion of Census Tract Population 38.9% 60.3% 86.1% 91.4% 20.7% 43.1% 35.0% 67.4% 15.1% 3.5% That is White (37.5) (36.8) (20.9) (15.1) (28.0) (37.0) (32.8) (31.8) (22.8) (2.8) Mean Distance from City Center (miles) 4.42 5.10 4.98 5.27 4.49 5.26 4.05 4.88 3.31 2.37 (2.14) (1.97) (2.81) (2.67) (2.08) (1.82) (1.58) (1.49) (1.85) (0.94) Median Census Tract Household Income 25,334 29,143 30,619 31,984 25,972 29,553 20,424 26,684 21,829 10,894 (98 dollars) Median Census Tract House Value 46,074 50,590 62,668 62,706 46,441 49,890 33,582 43,223 49,288 30,624 (98 dollars) Median Census Tract Monthly Rent 509 531 532 538 508 539 491 506 450 196 (98 dollars) Mean Homeownership Rate 62.9% 72.1% 72.3% 76.9% 59.8% 69.9% 60.8% 72.2% 46.0% 26.1% (16.6) (11.7) (13.4) (10.1) (17.8) (13.0) (14.2) (9.8) (24.1) (12.8) White Homeownership Rate 73.5% 77.0% 75.0% 78.5% 69.0% 74.9% 79.8% 79.3% 58.4% 66.0% (21.9) (15.4) (12.8) (9.4) (27.8) (20.2) (13.8) (8.0) (31.9) (41.2) Black Homeownership Rate 49.8% 48.1% 30.5% 32.2% 54.3% 54.3% 50.2% 40.4% 44.7% 26.5% (25.3) (29.2) (31.9) (34.8) (22.3) (25.7) (23.7) (30.6) (25.0) (11.9) Hispanic Homeownership Rate 48.0% 53.7% 50.4% 52.6% 47.6% 54.3% 46.8% 53.1% 35.2% 4.4% (29.6) (27.9) (30.9) (33.0) (36.2) (31.0) (16.0) (18.6) (37.3) (11.2) Mean Percent of Adults in Census 9.0% 8.5% 8.8% 8.1% 11.2% 10.1% 5.3% 5.9% 13.8% 5.0% Tract with B.A.s (9.9) (6.3) (8.4) (6.0) (11.4) (7.0) (7.2) (4.4) (16.4) (1.4) Notes: N=4,517 Settlement Grant and Nehemiah households for whom old and current tracts are known. Standard deviations are in parentheses. Source: Settlement Grant Recipient Data, Office of Housing and Community Development, City of Philadelphia and 1990 Census. Table 6 Community Quality Exposure For City and for Settlement Grant and Nehemiah Households City as Settlement Grant Households1 Nehemiah a Whole All White Black Hispanic Households School Test Scores Percent of 8th grade students scoring 12.0% 10.4% 13.9% 9.3% 8.9% 2.7% above the state median on math tests (10.4) (0.7) (7.6) (6.7) (6.9) (1.0) School Class Size Total student enrollment divided by 18.9 19.8 18.2 20.5 20.0 20.6 number of teachers (2.2) (2.5) (2.1) (2.5) (2.2) (3.6) Crime Number of murders per 1000 population 0.34 0.28 0.11 0.32 0.36 0.89 (0.75) (0.32) (0.20) (0.31) (0.31) (0.55) Education Percent of adults over 24 who have 17.9% 8.4% 8.1% 10.1% 5.8% 5.0% bachelor's degrees (0.2) (0.1) (0.1) (0.1) (0.0) (0.0) Commercial Space Percent of total building area used for 13.7% 9.0% 8.0% 10.1% 8.8% 30.0% commercial space (0.2) (0.1) (0.1) (0.1) (0.1) (0.1) Distance from City Center Miles from center of tract to city hall 5.2 5.1 5.2 5.2 4.8 2.4 (3.0) (1.9) (2.6) (1.8) (1.4) (0.9) Observations2: 350 7,973 1,575 3,762 2,244 86 Notes: 1. 392 households did not identify race. 2. Number of observations for City is 350 census tracts with data on all six measures. For Households columns, number of observations are households of each category in OHCD grant data. Values in Household columns reflect average tract values for tracts with OHCD households of each race, weighted by number of households of that race per tract. Sources: Pennsylvania Department of Education, Philadelphia Police Department Homicide Division, 1990 Census, the Philadelphia Board of Revision of Taxes, and Settlement Grant Recipient data, the Office of Housing and Community Development of the City of Philadelphia. Table 7 Changes Among Community Quality Variables for Settlement Grant Recipients In Becoming Owners All White Black Hispanic Renter Owner Renter Owner Renter Owner Renter Owner Average for All Households Math Scores1 8.1 10.5 13.6 14.2 6.4 9.3 6.5 9.3 Class Size2 19.3 19.8 18.1 18.1 19.9 20.5 19.1 19.8 Murder Rate per 1000 persons 0.42 0.28 0.16 0.11 0.44 0.33 0.58 0.33 3 Education Level 0.09 0.08 0.09 0.08 0.11 0.10 0.05 0.06 Commercial Space4 0.10 0.09 0.10 0.08 0.11 0.10 0.09 0.09 Distance from City Hall (miles)5 4.4 5.0 5.0 5.3 4.4 5.2 4.0 4.9 Notes: 1. Percent of 8th grade students scoring above state median in math tests. 2. Total number of students per teacher. 3. Percent of adults 24 and older who have bachelor's degrees. 4. Percent of total building area used for commercial space. 5. Miles from center of tract to city hall. Based on 4425 households for whom both previous and current census tracts are known; 819 white, 2196 black, 1171 Hispanic, and 239 other. Source: School source, Philadelphia Police Department Homicide Division, 1990 Census, the Philadelphia Boarf of Revision and Taxation, and authors' calculations. Table 8 Determinants of Black Household Locational Choice beta se beta se Tract % Black 0.279 0.046 -0.484 0.143 Tract % black*household income 0.383 0.065 Tract Poverty Rate -3.076 0.146 1.908 0.411 Tract Poverty Rate*household income -2.561 0.201 Tract % College Graduates 0.454 0.238 -5.433 0.801 Tract % College Graduates*household income 2.740 0.330 observations 3280 3280 pseudo R2 0.016 0.02 This table reports estimates of a 204 dimensional conditional logit model. Each black migrant household's choice of census tract is observed. The 204 census tracts represent those tracts in the City of Philadelphia whose median home price in 1990 was less than $60000. The choice of census tract is explained as a function of the tract's poverty rate, % black, and % college graduate. The right column reports results where these attributes are interacted with the migrant household's reported income. Table 9 Survey Data Survey Household Characteristics (Status at Time of Survey, 1998-99) All1 White Black Hispanic Nehemiah Percent of Households 100.0% 23.0% 57.0% 12.0% Household Size 3.4 3.3 3.4 4.0 2.8 Percent Female-Headed Household 43.8% 31.5% 55.3% 31.3% 44.7% Age of Householder 34.4 32.9 35.5 34.0 41.4 Percent Elderly 0.5% 0.0% 0.9% 0.0% 0.0% Percent With Children in House (any relation) 79.3% 73.9% 82.0% 77.1% 57.9% How Many Children Under 18 in House 2.3 1.5 2.3 1.8 1.0 Number of Adults in House 1.7 1.9 1.6 2.1 1.8 Number of Employed Adults in House 1.3 1.4 1.3 1.2 1.4 Number of Adults with B.A. in House 0.2 0.3 0.3 0.1 0.5 Survey Median Household Income (98$s)* $27,500 $32,500 $27,500 $22,500 $37,500 Survey Median House Price (98$s)* $40,612 $48,590 $39,493 $38,360 $52,500 Survey Median Downpayment (98$s)* $2,500 $2,557 $2,500 $2,633 $3,080 Median House Price/Income 2.2 2.1 2.1 2.2 2.0 Years Lived in Philadelphia 26.9 26.1 29.8 19.2 33.1 Percent Who Changed Tracts 82.9% 72.7% 83.3% 97.4% 87.1% Percent Who Stated That They 59.0% 42.9% 63.4% 66.7% 57.3% Changed Neighborhoods Distance Travelled (miles)2 1.9 1.7 2.0 1.9 1.9 Notes: Based on 400 Settlement Grant recipient households. Standard deviations are in parentheses. * Household income, house price and downpayment are as reported by survey respondents at the time of the survey, in 1998 dollars. These values may var from values reported by households to the OHCD at time of grant. 1. 32 households did not identify race. 2. Distance from center of previous tract to center of current tract. Source: Authors' survey. Table 10 Survey Data Structure Characteristics for Nehemiah Households: for Nehemiah Residents Comparison of Renter and Owner Houses Renter Owner Median House Price $53,102 Rooms 6.17 7.60 (2.11) (1.19) Beds 2.37 3.16 (0.96) (0.41) Baths 1.16 2.02 (0.37) (0.38) % Built 1980 or Later 7.5% 100.0% (26.7) (0.0) Air Conditioning 11.3% 100.0% (31.9) (0.0) Garage 14.5% 93.5% (35.5) (24.8) Single Family 67.7% 100.0% (47.1) (0.0) Detached Singe Family 16.7% 3.3% (37.7) (18.0) Attached Single Family 83.3% 96.7% (37.7) (18.0) Soundproof Walls 37.7% 63.8% (48.9) (48.5) Problems with Leaks 58.1% 19.4% (49.7) (39.8) Electrical Problems 29.5% 8.1% (46.0) (27.5) Unsatisfied with House NA 1 1.6% (12.7) New House is Better than Old 85.5% (35.5) Notes: 1. Not available. Standard deviations are in parentheses. Source: Authors' survey of 62 Nehemiah residents for whom previous tract is known. These are self-reported values. Table 11 Survey Data Structure Characteristics for Settlement Grant Recipients: Comparison of Renter and Owner Houses All1 White Black Hispanic Renter Owner Renter Owner Renter Owner Renter Owner Median House Price (98 dollars) $39,493 $48,590 $39,493 $38,927 Rooms 6.26 7.63 6.62 7.58 6.31 7.83 5.62 7.21 (2.33) (1.60) (2.10) (1.56) (2.48) (1.55) (2.10) (1.80) Beds 2.55 3.02 2.52 2.91 2.61 3.10 2.32 2.95 (1.10) (0.65) (1.07) (0.63) (1.12) (0.62) (1.08) (0.80) Baths 1.20 1.34 1.17 1.32 1.21 1.37 1.21 1.29 (0.48) (0.55) (0.41) (0.57) (0.52) (0.56) (0.53) (0.52) % Built 1980 or Later 7.9% 8.1% 5.6% 2.9% 8.8% 11.2% 12.5% 9.7% (27.1) (27.4) (23.1) (16.8) (28.4) (31.6) (33.8) (30.1) Air Conditioning 9.9% 11.2% 14.3% 13.0% 9.1% 10.2% 7.9% 10.5% (30.0) (31.6) (35.2) (33.8) (28.9) (30.4) (27.3) (31.1) Garage 26.2% 42.1% 38.2% 41.6% 22.0% 42.2% 23.7% 42.1% (44.0) (49.4) (48.9) (49.6) (41.6) (49.5) (43.1) (50.0) Single Family 78.3% 100.0% 77.9% 100.0% 78.5% 100.0% 76.3% 100.0% (41.3) 0.0 (41.7) 0.0 (41.2) 0.0 (43.1) 0.0 Detached Single Family 9.1% 7.5% 8.3% 5.3% 8.2% 8.1% 13.8% 10.8% (28.9) (26.4) (27.9) (22.5) (27.6) (27.4) (35.1) (31.5) Attached Single Family 90.9% 92.5% 91.7% 94.7% 91.8% 91.9% 86.2% 89.2% (28.9) (26.4) (27.9) (22.5) (27.6) (27.4) (35.1) (31.5) Soundproof Walls 28.2% 37.1% 29.7% 36.6% 28.3% 34.1% 23.5% 51.4% (45.0) (48.4) (46.0) (48.5) (45.2) (47.5) (43.1) (50.7) Problems with Leaks 33.6% 21.2% 25.0% 24.7% 39.3% 21.1% 23.7% 21.1% (47.3) (40.9) (43.6) (43.4) (49.0) (40.9) (43.1) (41.3) Electrical Problems 18.8% 13.7% 15.8% 16.9% 20.7% 14.5% 15.8% 7.9% (39.1) (34.4) (36.7) (37.7) (40.6) (35.3) (37.0) (27.3) Unsatisfied with House NA 2 3.7% NA 2 3.9% NA 2 3.2% NA 2 7.9% (19.0) (19.5) (17.7) (27.3) 2 2 2 2 New House is Better Than Old NA 76.2% NA 72.0% NA 78.1% NA 78.9% (42.7) (45.2) (41.4) (41.3) Notes: 1. 8% did not identify race of householder. 2. Not available. Standard deviations are in parentheses. Source: Authors' survey of 322 Settlement Grant Recipients for whom previous tract is known. These are self-reported values. Table 12 Survey Data Community Characteristics: Comparison of Current and Previous Neighborhoods for Nehemiah Residents Renter Owner 1990 Census Data1: Average Tract Values for Household Tracts Homeownership Rate 48.5% 33.5% (22.0) (7.8) Vacancy Rate 14.7% 17.6% (6.0) (3.2) Housing Stock Built 1980 or Later 4.0% 3.0% (4.6) (0.9) All Stock That is Commercial 16.3% 35.9% (0.2) (0.1) Adults Who Are College Graduates 14.5% 5.9% (17.2) (1.7) Median Tract House Value (98$s) $33,257 $22,741 Median Tract Household Income (98$s) $24,706 $13,037 Poverty Rate 31.9% 53.4% (15.4) (3.8) Murder Rate per 1000 Persons1 0.50 0.69 (0.38) (0.24) Population White 15.4% 2.3% (25.3) (1.7) Population Black 79.4% 95.8% (31.3) (4.6) Population Hispanic 3.1% 1.4% (9.8) (2.1) 1 Students Scoring Above State Median in Math 5.4% 2.2% (6.5) (0.5) School Class Size1 20.7 22.6 (2.5) (2.2) Distance from City Center2 3.41 2.87 (1.65) (0.60) Survey Data: Average Household Values3 Satisfied with Neighborhood NA 5 64.5% (48.2) New Neighborhood Better Than Old -- 68.6% (47.1) Litter/Abandoned Buildings/etc. a Problem NA 5 83.6% (37.3) Current Neighborhood More Attractive -- 62.9% Than Old Neighborhood (49.0) Take Crime Precautions 35.5% 66.1% (48.2) (47.7) Schools are Good NA 5 25.5% (44.1) Schools in Current Neighborhood Are Better -- 25.7% Than in Old Neighborhood (44.3) Distance from Old Census Tract4 -- 1.92 (1.93) Note: 1. From Census 1990, except murder rates, math scores, and tract distance measures. 2. Distance in miles from center of tract to center of tract #5, where city hall is. 3. These are self-reported values. 4. Distance in miles from center of previous tract to center of current tract. 5. Not available. Sources: Census 1990; Philadelphia Police Department Homicide Division; Pennsylvania Department of Education; and author's survey of 62 Nehemiah residents for whom previous tract is known. Table 13 Survey Data Community Characteristics for Surveyed Grant Recipients: Comparison of Renter and Owner Neighborhoods 1 All White Black Hispanic Renter Owner Renter Owner Renter Owner Renter Owner Census Tract Data2: Average Tract Values For Household Tracts Homeownership Rate 63.1% 71.1% 70.6% 76.9% 60.5% 68.4% 60.1% 71.0% (17.9) (12.6) (15.6) (10.1) (18.2) (13.5) (17.0) (10.0) Vacancy Rate 11.4% 9.3% 8.4% 6.9% 12.7% 10.4% 12.2% 9.5% (5.5) (4.9) (4.8) (3.9) (5.5) (5.1) (4.9) (4.2) Housing Stock Built 1980 or Later 2.7% 1.8% 2.0% 1.8% 2.5% 2.1% 3.7% 0.6% (4.5) (3.4) (3.6) (3.6) (3.5) (3.7) (6.7) (1.6) All Stock That is Commercial 10.2% 9.0% 8.6% 7.2% 11.1% 9.7% 9.5% 10.2% (11.0) (8.8) (7.3) (5.3) (12.6) (10.5) (6.9) (6.3) Adults Who Are College Graduates 10.8% 9.1% 10.7% 8.9% 11.1% 9.5% 10.1% 7.8% (10.9) (6.2) (11.1) (6.9) (10.7) (5.9) (13.0) (6.8) Median Tract House Value (98$s) $45,088 $47,980 $59,548 $67,172 $40,553 $44,168 $27,013 $47,980 Median Tract Household Income (98$s) $28,007 $32,045 $33,132 $33,264 $26,836 $30,555 $22,522 $29,295 Poverty Rate 25.1% 19.1% 16.7% 13.0% 26.4% 21.3% 35.3% 21.7% (14.8) (11.6) (12.7) (9.2) (12.5) (11.6) (19.2) (12.1) Murder Rate per 1000 Persons2 0.37 0.30 0.16 0.11 0.44 0.40 0.48 0.27 (0.35) (0.32) (0.20) (0.17) (0.36) (0.36) (0.41) (0.25) White Population 41.3% 56.3% 82.8% 87.7% 22.1% 38.0% 45.4% 74.2% (38.9) (38.1) (25.6) (19.7) (29.7) (36.2) (34.0) (25.6) Black Population 47.9% 33.4% 9.9% 7.1% 72.0% 52.0% 16.7% 6.0% (40.7) (39.2) (21.8) (17.8) (32.7) (40.3) (18.8) (11.9) Hispanic Population 8.1% 6.3% 4.5% 2.8% 3.7% 5.8% 34.2% 14.4% (17.0) (12.5) (9.7) (7.2) (9.8) (11.6) (27.9) (17.7) Students Scoring Above State Median in Math2 8.9% 10.4% 13.7% 14.3% 6.7% 8.4% 7.9% 10.5% (7.7) (7.5) (9.3) (8.8) (5.6) (6.1) (6.5) (6.9) School Class Size2 19.3 19.6 18.1 18.2 20.0 20.3 18.5 19.3 (2.2) (2.5) (2.0) (2.4) (2.0) (2.2) (2.0) (2.7) Distance from City Center2,3 4.57 5.08 4.89 5.25 4.51 5.06 4.23 5.14 (2.34) (2.06) (2.84) (2.57) (1.97) (1.82) (2.23) (1.60) 4 Survey Data: Average Household Values Satisfied with Neighborhood NA 6 65.1% NA 6 67.5% NA 6 62.9% NA 6 63.2% (47.7) (47.1) (48.4) (48.9) New Neighborhood Better Than Old -- 68.3% -- 70.0% -- 71.4% -- 52.0% (46.7) (46.6) (45.4) (51.0) 6 6 6 6 Litter/Abandoned Buildings/etc. a Problem NA 67.9% NA 64.5% NA 66.8% NA 65.8% (46.8) (48.2) (47.2) (48.1) Current Neighborhood More Attractive -- 58.1% -- 53.3% -- 60.5% -- 52.0% Than Old Neighborhood (49.5) (50.7) (49.1) (51.0) Take Crime Precautions 28.0% 56.2% 16.9% 50.6% 32.8% 58.1% 31.6% 60.5% (44.9) (49.7) (37.7) (50.3) (47.1) (49.5) (47.1) (49.5) Schools are Good NA 6 54.2% NA 6 45.9% NA 6 52.6% NA 6 73.5% (49.9) (50.2) (50.1) (44.8) Schools in Current Neighborhood Are Better -- 51.6% -- 36.7% -- 55.5% -- 48.0% Than in Old Neighborhood (50.1) (49.0) (49.9) (51.0) Distance from Old Census Tract5 -- 1.90 -- 1.69 -- 1.96 -- 1.93 (2.01) (2.41) (1.99) (1.40) Notes: 1. 7% did not identify race of householder. 2. From Census 1990, except murder rates, math scores, and tract distance measures. Includes 147 tracts weighted by the number of Grant Recipient households per tract. Households include only those for whom previous tract is known. 3. Distance in miles from center of tract to center of tract #5, where city hall is. 4. These are self-reported values. 5. Distance in miles from center of previous tract to center of current tract. 6. Not available. Standard deviations are in parentheses. Sources: Census 1990; Philadelphia Police Department Homicide Division; Pennsylvania Department of Education; and authors' survey of 322 Settlement Grant Recipients for whom previous tract is known. Table 14 Survey Data House Price as a Function of Structure Characteristics and Community Quality Regression Model Based on Survey Data on Settlement Grant Recipients Dependent Variable: Log of Price Structure Characteristics from Survey Data Rooms 0.044 * (0.016) Baths 0.047 (0.034) Air Conditioning1 0.157 * (0.071) Garage1 0.143 * (0.046) Soundproof Walls1 0.076 * (0.036) Problems with Leaks1 -0.104 * (0.050) Community Characteristics from Census Data % Students Scoring Above State Median in Math Tests 1.044 * (0.295) Classroom Size -0.002 (0.007) Murder Rate per 1000 Persons -0.092 (0.074) % of Adults with Bachelor's Degree 1.425 * (0.258) % of Total Building Area That Is Commercial 0.223 (0.207) Miles from City Hall 0.019 (0.207) Constant 9.829 * (0.179) Adjusted R-Squared 0.376 Observations 313 Notes: 313 Non-Nehemiah Settlement Grant households in authors' survey with data on all variables. Standard errors are in parentheses. *Asterisks are significant at 10% level. 1. Dummy variables. Sources: Authors' calculations using authors' Survey results and 1990 Census data. Table 15 Survey Results Neighborhood Search All White Black Hispanic Nehemiah Number of houses looked at Range 1-80 1-80 1-60 1-15 1-15 Average 7.1 8.0 7.1 6.0 4.3 (Standard Deviation) (6.8) (9.4) (6.1) (4.5) (4.2) Number of Philadelphia neighborhoods looked at Range 1-8 1-4 1-8 1-4 1-5 Average 1.4 1.2 1.4 1.5 1.4 (Standard Deviation) (0.88) (0.66) (0.98) (0.80) (0.99) Looked in suburbs Average 14.8% 14.1% 14.5% 8.3% 26.3% (Standard Deviation) (35.50) (35.02) (35.26) (27.93) (44.33) Number of suburbs looked at Range 0-4 0-2 0-4 0-1 0-4 Average 0.2 0.2 0.2 0.1 0.5 (Standard Deviation) (0.64) (0.57) (0.62) (0.28) (0.99) Observations 400 92 228 48 76 Source: Authors' survey. Table 16 House Price as a Function of Structure Characteristics and Community Quality Regression Model Based on Philadelphia Sales Transaction Data beta s.e beta s.e Dummy for Nehemiah Tracts -0.401 0.108 -0.390 0.109 Dummy for Nehemiah Tracts Post-Nehemiah 0.120 0.045 0.117 0.046 Dummy for Control Group #1 -0.274 0.107 Dummy for Control Group #1 Post-Nehemiah 0.223 0.064 Dummy for Control Group #2 -0.179 0.075 Dummy for Control Group #2 Post-Nehemiah 0.090 0.051 Log of Total Area 0.228 0.017 0.228 0.017 Number of Stories 2 -0.003 0.027 0.001 0.027 Number of Stories 3 0.092 0.038 0.103 0.038 Number of Stories 4 or more 0.470 0.085 0.484 0.087 Garage 0.287 0.027 0.286 0.026 Community Attributes % Students Scoring Above State Median in Math Tests 1.259 0.202 1.251 0.201 Classroom Size -0.024 0.006 -0.023 0.006 Murder Rate per 1000 persons -0.580 0.123 -0.597 0.125 % of Adults with Bachelor's Degree 2.113 0.152 2.130 0.150 % of Total Buildings That is Commercial 0.326 0.167 0.318 0.159 Miles from City Hall 0.0132 0.009 0.0133 0.009 Constant 7.934 0.235 7.922 0.234 observations 146053 146053 R2 0.596 0.595 Note: This table reports two hedonic home price regressions based on equation (1) in the text. Control Group #1 is defined as the set of census tracts with poverty rates greater than or equal to 50% and these tracts are greater than 1.2 miles away from the closest Nehemiah tract. Control group #2 is defined as the set of census tracts that share a border with at least one Nehemiah tract. The data set covers the years 1986-1997 and year dummies are included in each regression. Post-Nehemiah is a dummy variable that equals one if in that year the Nehemiah housing complex already had been built. Table 17 Survey Data Relationship between Community Quality Objective Measures and Impressions of Community Quality CRIME PRECAUTION Regression #: 1 2 3 4 5 6 7 8 9 10 11 12 Taken Following Crime Precaution: Dependent Variables (Survey Respondents' Good Undesirable Avoid Do not Install Avoid Taken Any Opinion of Their Neigbhorhood): Good Good Social Helpful Good Street Public Go Out Protection Going Out Keep a Crime Schools Schools Connections Neighbors Shopping Activity Transit Alone Devices At Night Weapon Precaution % of 8th grade students scoring above the state 0.047 0.044 median on math tests (0.376) (0.375) Classroom size -0.007 -0.007 (0.010) (0.010) Household inquired about school quality 0.150 * before moving (0.053) Years in current community 0.008 * 0.004 * (0.003) (0.002) Household moved less than 1.25 miles 0.144 (0.091) Percent of neighborhood's total building 0.462 areas that is commercial (0.292) Murder rate (average of 1994 and 1995) 0.244 * 0.064 * 0.083 * 0.111 0.067 -0.022 0.104 (0.054) (0.028) (0.037) (0.083) (0.049) (0.049) (0.088) Nehemiah Resident -0.270 * -0.250 * -0.025 * 0.066 * -0.445 * -0.118 * -0.035 * -0.057 0.143 * -0.077 * 0.046 0.098 * (0.069) (0.070) (0.058) (0.052) (0.091) (0.031) (0.018) (0.039) (0.050) (0.022) (0.052) (0.042) Age of Household Head 0.006 * 0.007 * 0.006 * 0.005 * -0.003 0.004 * 0.001 0.000 -0.007 * -0.003 * -0.002 -0.007 * (0.003) (0.003) (0.003) (0.002) (0.003) (0.002) (0.001) (0.001) (0.003) (0.002) (0.002) (0.002) Race of Household Head Black 0.062 0.064 -0.213 * -0.109 -0.189 * -0.027 0.011 -0.014 0.012 0.071 * -0.034 0.069 (0.071) (0.070) (0.069) (0.066) (0.065) (0.059) (0.028) (0.039) (0.071) (0.033) (0.045) (0.070) Hispanic 0.264 * 0.277 * -0.193 * -0.212 * -0.157 * -0.019 0.062 0.011 -0.075 0.036 -0.048 0.014 (0.081) (0.082) (0.087) (0.093) (0.093) (0.070) (0.065) (0.053) (0.101) (0.064) (0.037) (0.104) Pseudo R-Squared 0.051 0.065 0.081 0.028 0.090 0.044 0.031 0.024 0.024 0.042 0.010 0.022 Observations 384 384 362 428 437 436 432 437 434 436 435 441 Notes: We use probit models in this appendix because the our dependent variable takes on the value 0 or 1. Standard errors are in parentheses. *Asterisk is statistically significant at 10% level. Sources: Pennsylvania Department of Education, Authors' survey, Philadelphia Board of Revision and Taxation.