Allocation by Seller's Preference: Rent Stabilization and Housing Discrimination in New York City Dirk W. Early Assistant Professor of Economics Department of Economics and Business Southwestern University Georgetown, TX 78627 phone: (512) 863-1592 fax: (512) 863-1535 earlyd@southwestern.edu December 1999 Abstract New York City's rent stabilization program creates an expenditure ceiling on some units of rental housing. Since price is no longer used to allocate stabilized units, landlords must use a non-price mechanism to determine who will receive the scare number of units available at below market rents. One possible mechanism is allocation by seller's preference, possibly in the form of discrimination. In the absence of rent regulations, landlords could also discriminate. However, an expenditure ceiling makes discrimination less costly since an excess demand for the unit exists. With an excess demand for their unit, landlords can bypass tenants they deem less desirable without losing rental income since additional households are willing to rent the unit. This study will examine the allocation of rent stabilized units in New York City using data from the 1996 New York City Housing and Vacancy Survey. Of interest is whether landlords leasing rent regulated apartments discriminate against certain groups. A logit regression is used to estimate the conditional probability of a New York tenant obtaining a rent stabilized unit as a function of several characteristics of the household, including race, gender, and household size. Preliminary findings suggest that in the borough of Manhattan, Hispanics and immigrants are less likely to obtain a stabilized unit. In addition, landlords seem more willing to rent to smaller households with higher incomes. The findings for Manhattan support the hypothesis that controlled units are allocated by seller's preference under New York's current rent stabilization program. In the other four boroughs of New York City, no evidence exists to suggest that seller's preference plays an important role in the allocation of stabilized housing.
I. Introduction Rent regulations are justified on the grounds that they offer protection to those groups least able to cope with rapidly rising rents. In Santa Monica, CA, for example, low income households, the poor, minorities, young families and senior citizens are all specifically mentioned as groups needing the protection rent controls offer (Levine, et al., 1990). However, since regulations prevent rents from rising to clear the market for stabilized units, an excess demand for regulated units of rental housing exists. Therefore, landlords must use non-price mechanisms to allocate the short supply of controlled units. Possible allocation schemes include excessive deposits, key money, requiring the renting of parking spaces or storage as a prerequisite for obtaining the unit, and allocation by seller s preference. The last of these is the focus of this study. This article uses data from New York City to estimate the role of household characteristics in determining the likelihood of obtaining a rent stabilized unit of housing. The selection process employed by landlords may reduce the likelihood that the groups who were originally intended to be helped actually receive the benefits of lower rents. Specific attention is given to minorities and low income households, groups that economic theory suggests should fair poorly when landlords use their preferences for different types of households to allocation the scarce amount of controlled housing. If landlords do not discriminate in their lending practices, the distribution of the characteristics of renters in controlled units should mimic the distribution
2 in the uncontrolled sector. However, allocation by landlord preference, based on race for example, would imply that African-Americans are less likely to end up in a controlled unit, holding all other factors equal. Although often described as a price ceiling, rent regulations are actually expenditure ceilings. Rent regulations limit the increase in the rent of a unit, but not the price per unit of rental housing service. Price refers to the cost of a standard unit of rental housing service and a market determined rent is the price per unit times the quantity of housing service offered by the unit. (See Arnott (1995) for more detail on this issue.) Under New York's rent stabilization program, the Rent Guidelines Board sets the allowable rent increase for sitting tenants. For 1996, the allowable increase on a one year lease was 2% (NYS DHCR, 1999a). This increase affected nearly all rent stabilized units with, except in extreme situations, no regard to the change in the level of housing service offered by the unit. 1 If the rent stabilization program is effective in controlling the cost of rental housing, allowable rent increases lie below what would prevail in the free sector. Since occupying a unit of controlled housing allow households to obtain housing at below market rents, these units are more desirable than units of uncontrolled housing. Holding unit characteristics constant, tenants will seek out stabilized units first and fall back on the free sector if unable to find a regulated unit of housing. Landlords with vacant units under the rent stabilization program will observe queues forming for their units. If the characteristics of a tenant do not fit the preferences of the landlords, they can deny that tenant with the realization that a number of potential tenants 1 Landlords are required to provide certain basis amenities, e.g. heat and running water, and if substantial improvements are made to the unit, landlords can petition for an increase in rents over what is set by the Rent Guidelines Board (NYS DHCR, 1999b).
3 remain. Landlords with uncontrolled units also have the option of denying tenants they deem undesirable, but at a cost, since the unit may sit vacant until a suitable tenant applies for the unit. Therefore, rent stabilization allows landlords to discriminate at lower costs than what prevails in the free sector. The next section discusses the value of a tenant to a landlord as a function of household characteristics. This is followed by a description of the data and method of estimation. The results are discussed in section four. The paper concludes with a brief summary of the findings, a discussion of the shortcomings of the paper and possible extensions. II. Desirable Tenants The value of a tenant will depend on the expected cost associated with renting to the household and on the landlord s preference for certain groups. For example, it seems reasonable that the rate of unit depreciation increases with the number of occupants. This makes the value of a tenant inversely related to household size. Landlords may also prefer wealthier tenants if they are perceived to be less likely to miss or be late with rent payments. In addition, landlords may also be bias, and therefore, place a lower value on tenants with certain racial and ethnic compositions, holding other things constant. Several authors have offered evidence in support of the hypothesis of racial discrimination existing in housing markets. Using fair housing audits, Yinger (1986) finds compelling evidence of the existence of discrimination in rental housing markets. Specifically, he finds that black auditors in Boston were shown more than a third fewer rental units than similar white auditors. Ladd (1998) discusses the literature on discrimination in mortgage
4 lending and concludes it is clear that mortgage lenders discriminate against minorities (Ladd, 1998, pp. 59). Kain and Quigley (1972) conclude that African-Americans pay a 5 to 10 percent more than whites for similar housing. If landlords renting stabilized units are also bias, minority households will be less likely to obtain units of regulated housing. The behavior of landlords described above will make it very difficult for the households who were intended to reap the rewards of rent stabilization to benefit from the lower rents in the regulated sector. The next section examines data from New York City to determine whether evidence exists in support of the hypothesis that regulated units are allocated by seller s preference. If landlords are not selective in their allocation of units, the probably of obtaining a controlled units should not be significantly affected by race, ethnicity, household size or other characteristics of the household. III. Data and Method of Estimation The 1996 New York City Housing and Vacancy Survey (Bureau of the Census, 1996) is the primary data set for this study. This survey contains information regarding the characteristics of tenants and the rent control status of the unit. The sample is restricted to recent movers (tenants who occupied their unit after 1989) who occupy rent stabilized or unregulated rental units. Rent controlled units are absent since the only controlled units are those that have been continuously occupied by the same tenant since 1971 (NYS DHCR, 1999c). Observations are also omitted if the household is unaware of the rent control status of the unit. With these restrictions, the data set contains 1827 observations on households in unregulated units and 627 observations on households in rent stabilized housing.
5 Logistic regressions are used to estimate the relationship between household characteristics and the probability of obtaining a controlled unit. If landlords renting stabilized units are no more selective than landlords renting unregulated units, household characteristics should not be significant determinants of the probability of obtaining a regulated unit. The coefficient estimates of the logistic regression will determine which characteristics influence the likelihood of obtaining a rent stabilized unit. Running the logistic regression for the entire city may confuse location preferences of tenants with the selection criteria employed by landlords. For example, Manhattan contains nearly 50 percent of all rent stabilized units, but only 11 percent of New York s African- American population resides on the island of Manhattan. Gabriel and Rosenthal (1989), using data from the Washington, D.C. metropolitan area, find that race is a highly significant determinant of residential location choice and that the effect of race was large relative to other variables. If race and ethnicity are also important in New York City, ignoring location choice would complicate the interpretation of the coefficients on race and ethnicity. To control for location, a separate regression is run for each of the five boroughs in New York City. Therefore, the logistic regressions estimate the importance of household characteristics on obtaining a stabilized unit, conditional on moving into a certain borough of New York City. Table 1 gives the household characteristics and descriptive statistics, by borough, of the variables used in the logit regressions.
6 Table 1. Descriptive statistics Dependent Variable Rent regulation status of the unit, = 1 if the household has obtained a rent stabilized unit, otherwise 0 Bronx Brooklyn Manhattan Queens Staten Island Mean Mean Mean Mean Mean (Std. Dev.) (Std. Dev.) (Std. Dev.) (Std. Dev.) (Std. Dev.) 0.16 (0.368) 0.15 (0.360) 0.46 (0.499) 0.22 (0.417) 0.07 (0.263) Independent Variables Age of the head of the household 38.20 (12.030) 38.72 (13.590) 35.78 (11.660) 38.04 (12.170) 40.93 (16.780) Age of the head of the household squared 1603.61 (1149.00) 1683.19 (1284.31) 1416.31 (1072.71) 1595.26 (1121.15) 1954.75 (1735.31) Monthly income in thousands 2.71 (2.236) 3.08 (2.848) 5.93 (6.291) 3.38 (2.632) 3.77 (3.254) Monthly income in thousands squared 12.33 (41.917) 17.58 (60.398) 74.65 (221.989) 18.31 (52.551) 24.70 (74.599) Race of the head of the household, = 1 if non-hispanic African- American, otherwise 0 0.34 (0.474) 0.33 (0.472) 0.09 (0.282) 0.17 (0.378) 0.17 (0.378) Ethnicity of the head of the household, = 1 if Hispanic, otherwise 0 0.50 (0.501) 0.19 (0.392) 0.14 (0.347) 0.29 (0.456) 0.11 (0.317) Immigrant status, = 1 if born outside of the U.S., otherwise 0 0.57 (0.497) 0.57 (0.496) 0.29 (0.456) 0.64 (0.481) 0.12 (0.331) Gender of the head of the household, = 1 if female, otherwise 0 0.39 (0.488) 0.35 (0.477) 0.39 (0.487) 0.27 (0.444) 0.35 (0.479) Number of persons per room 0.80 (0.407) 0.74 (0.395) 0.72 (0.390) 0.78 (0.400) 0.60 (0.363) Number of observations 277 658 668 734 117
7 One justification for the inclusion of age is due to the concern over unit depreciation. It seems reasonable that landlords would prefer older tenants if they are expected to be more careful with the unit. However, landlords prefer high turnover of stabilized units. Under New York's rent stabilization program, allowable rent increases are higher for new tenants. The rent increase for new tenants is the rent increase for sitting tenants, set by the Rent Guidelines Board, plus a vacancy allowance. Landlords with units that were vacated during the first part of 1996 were allowed an 8.5% vacancy allowance on top of the 2% increase for sitting tenants (NYS DHCR, 1999a). Since younger tenants may be perceived to have shorter stays, the benefits associated with higher turnover may offset some of the preference for older tenants. High income tenants, as mentioned above, are more attractive since problems of nonpayment of rent are less likely to occur. Race and ethnicity are captured by dummy variables for non-hispanic African-Americans and for Hispanics, respectively. Since immigrants are an easily identifiable group that could be subject to discrimination by landlords, a dummy variable is included to indicate whether the tenant was born outside the U.S. Female headed households are identified to determine whether landlords have a preference for the gender of their tenants. Household size is a factor due to the level of depreciation increasing with the number of persons. Simply using household size, however, will be problematic since unregulated and regulated units differ by number of rooms. On average, stabilized units have 3.03 rooms per unit while unregulated units contain 3.66 rooms per unit. To control for differences in unit size, while capturing the effects of household size on the probability of obtaining a controlled unit, the number of persons per room is included as an explanatory variable. The next section describes the results from the logit regressions. Coefficient estimates
are then used to simulate the importance of various household characteristics on the probability of obtaining a stabilized unit of housing. 8 IV. Results Table 2 gives the results of the logistic regression for each of the five boroughs. Except for the borough of Manhattan, the results do not support the hypothesis that landlords allocate stabilized units by seller s preference. The only coefficients found to be significant in the regressions using data from the four non-manhattan boroughs are age and age squared for Staten Island. Data on the distribution of rental units in Manhattan fit well with the notion that landlords renting stabilized units select tenants they find most desirable. The coefficient estimates suggest that ethnicity, being born outside the U.S., income, and persons per room are statistically significantly related to the probability of obtaining a unit of stabilized housing on the island of Manhattan. The gender of the household head does not appear to influence a landlord s decision to accept a household's rental application. It is possible to estimate the magnitude of these effects using the coefficient estimates to simulate the probability of obtaining a stabilized unit. The results of the simulation suggest that Hispanics are 30% less likely to be in a stabilized unit, holding all other factors constant, and immigrants to the U.S. are 15% less likely. Changes in income led to small changes in the estimated probability of obtaining stabilized housing. A one-standard deviation increase in income (more than $75,000 annually) increases the probability of living in a stabilized unit by just over 3%.
9 Table 2. Logistic Regressions Variable Age of the head of the household Age of the head of the household squared Monthly income in thousands Monthly income in thousands squared non-hispanic African- American Household Bronx Brooklyn Manhattan Queens Staten Island Coeff. Coeff. Coeff. Coeff. Coeff. Sig Sig Sig Sig Sig (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) -0.0075 (0.0680) 0.9119-0.0076 0.0353 0.0420-0.2743 (0.0428) 0.8589 (0.0366) 0.3352 (0.0411) 0.3075 (0.1543) 0.0755 6.23E-05 (0.0007) 0.0625 (0.1430) -0.0036 (0.0096) -0.3166 (0.4986) 0.9301 0.0002 (0.0004) 0.6621 0.1885 (0.1619) 0.7060-0.0229 (0.0180) 0.5255 0.3773 (0.2452) 0.6644-0.0004 (0.0004) 0.2444 0.0672 (0.0419) 0.2034-0.0031 (0.0015) 0.1238 0.3264 (0.2881) 0.2626-0.0004 (0.0004) 0.1088 0.0483 (0.0686) 0.0397-0.0010 (0.0032) 0.2572-0.1933 (0.2552) 0.3976 0.0028 (0.0015) 0.0607 0.4813 1.0234 (0.9630) 0.2879 0.7590-0.1203 (0.1207) 0.3189 0.4488 1.0783 (0.9762) 0.2693 Hispanic household -0.1472 (0.4694) Born outside the U.S. 0.0472 (0.3578) Female headed household 0.2586 (0.3544) Persons per room -0.0848 (0.4365) 0.8460 Constant -1.4790 (1.6029) 0.7538-0.2739 (0.3414) 0.8950-0.3596 (0.2401) 0.4657 0.1412 (0.2456) 0.2083 (0.2944) 0.3562-2.1078 (0.9664) 0.4224-0.6578 (0.2760) 0.1342-0.3310 (0.1902) 0.5655 0.1716 (0.1728) 0.4793-0.3666 (0.2248) 0.0292-0.6088 (0.7935) 0.0172-0.1523 (0.2189) 0.0817-0.2814 (0.2006) 0.3206 0.1267 (0.2157) 0.1029 0.2602 (0.2384) 0.4429-2.3789 (0.9122) 0.4868 0.7266 (1.1621) 0.5318 0.1608 0.2218 (1.1461) 0.8466 0.5568-0.9261 (0.9200) 0.3141 0.2750-1.3510 (1.4951) 0.3662 0.0091 2.0958 (3.1502) 0.5059 N Chi-Squared Significance 277 1.161 0.999 658 12.000 0.213 668 37.150 0.000 734 6.534 0.686 117 8.031 0.531
The fact that only landlords in Manhattan display behavior indicative of allocation by seller's preference may be due to Manhattan being the most expensive area of New York City. More important, it is conceivable that the free sector of Manhattan has had the highest rates of rent increases of the five boroughs. Since the Rent Guidelines Board sets uniform rent increases for all stabilized units in New York City, the implicit subsidy 2 of units in Manhattan may be larger than in other boroughs. Examining the average rent of stabilized units compared with units in the free sector, offers some evidence in support of this notion. The difference in average rent between the two sectors in Manhattan is $207 per month. This is more than twice as large as the difference in any other borough. If the characteristics of stabilized units are similar to the characteristics of unregulated units, these differences are good estimates of the implicit subsidy from a unit of stabilized housing. The larger the implicit subsidy, the greater the demand for the unit. Landlords in Manhattan may realize that a long list of potential tenants is competing for their units, allowing them to be more selective than landlords in other parts of New York City. V. Conclusion If rationing by price is no longer an option, economic theory predicts that sellers in a market with excess demand must rely on other methods to allocate their good. This study examines data on the characteristics of tenants in stabilized and unregulated apartments in New York City to determine whether landlords use seller's preference in the allocation of stabilized rental housing. For four of the five boroughs of New York City, no evidence exists to suggest that landlords are discriminating according to race, ethnicity, or other factors. However, in the borough of 2 The implicit subsidy from a stabilized unit is the difference between the rent of a stabilized unit and what that unit would rent for in the free sector.
Manhattan, the distribution of stabilized apartments is consistent with landlords allocating their units to the tenants they find most desirable. The results suggest that rent stabilized apartments in Manhattan are less likely to be rented to households with low incomes, households headed by Hispanics and immigrants, and to households that are large relative to the size of the unit. Several shortcomings of this study should be noted. It is hypothesized that Manhattan landlords are able to discriminate with little to no pecuniary costs due to the high implicit subsidies of their units. Whether stabilized units in Manhattan offer high implicit subsidies, relative to the other boroughs of New York City, is a testable hypothesis and should be investigated further. Differences in the characteristics of stabilized and uncontrolled units also need to be examined. If stabilized units are substantially different from units found in the free sector, stabilized and unregulated units may not be easily substitutable. The ability of landlords to discriminate with little cost would be reduced if stabilized units, even at below market rents, are not deemed desirable to most tenants. Tenant choice also needs to be included in the analysis. A unit of housing is only occupied when both parties agree. Landlords must make the unit available and tenants must accept the offer made by the landlord. Care needs to be taken to ensure that the differences in the probability of obtaining stabilized housing are correctly attributed to the actions of one of the two parties. Finally, the New York City Housing and Vacancy Survey does not contain detailed information regarding household composition. In particular, testing whether single women with children are denied stabilized units was impossible given these data. An improved version of the above analysis will be attempted using a more detailed data set, probably the American Housing Survey. Although far from conclusive, this study raises questions regarding the allocation of the benefits of New York's rent stabilization program. Future research will investigate whether
landlord preferences vary through time and across urban areas.
List of References: Boehm, Thomas P. and Richard A. Hofler. A Frontier Approach to Measuring the Effect of Market Discrimination. Southern Economic Journal, 1987, 54, pp. 301-15. Bureau of the Census. New York City Housing and Vacancy Survey: 1996. [machine-readable data file] conducted by the Bureau of the Census for the New York City Department of Housing Preservation and Development, Washington, D.C., Bureau of the Census producer and distributor], 1996. Downs, Anthony. Policy Directions Concerning Racial Discrimination in U.S. Housing Markets. Housing Policy Debate, 1992, 3(2), pp. 685-745. Farley, Reynolds, Elaine L. Fielding, and Maria Krysan. The Residential Preferences of Blacks and Whites: A Four-Metropolis Analysis. Housing Policy Debate, 1997, 8(4), pp. 763-800. Farley, Reynolds. Racial Differences in the Search for Houising: Do Whites and Blacks Use the Same Techniques to Find Housing? Housing Policy Debate, 1996, 7(2), pp. 367-85. Gabriel, Stuart A. and Stuart S. Rosenthal. Household Location and Race: Estimates of a Multinomial Logit Model. The Review of Economics and Statistics, May 1989, 71(2), pp. 240-249. Kain, John F. and John M. Quigley. Housing Market Discrimination, Homeownership, and Savings Behavior. American Economic Review, June 1972, 62(3), pp. 263-77. Ladd, Helen F. Evidence on Discrimination in Mortgage Lending. Journal of Economic Perspectives, 1998, 12(2), pp. 41-62. Levine, Ned, J. Eugene Grisgby III and Allan Heskin. Who Benefits from Rent Control? Effects on Tenants in Santa Monica, California. APA Journal, 56, pp. 140-152. Loikkanen, Heikki A. On Availability Discrimination Under Rent Control. Scandinavian Journal of Economics, 1985, 87(3), pp. 500-20. Momeni, Jamshid. Race, Ethnicity, and Minority Housing in the United States. New York: Greenwood Press, 1986. Myers, Samuel L. and Tsze Chan. Racial Discrimination in Housing Markets Accouting for Credit Risk. Social Science Quarterly, 1995, pp. 543-61.
NYS Division of Housing and Community Renewal (NYS DHCR). Fact Sheet #26 - Guide to Rent Increases for Rent Stabilized Apartments in New York City. World Wide Web page <http://www.dhcr.state.ny.us/ora/pubs/html/orafac26.htm> (last modified October 20, 1999a). NYS Division of Housing and Community Renewal (NYS DHCR). Fact Sheet #11 - Rent Increases for Major Capital Improvements (MCI). World Wide Web page <http://www.dhcr.state.ny.us/ora/pubs/html/orafac11.htm> (last modified October 20, 1999b). NYS Division of Housing and Community Renewal (NYS DHCR). Fact Sheet #1 - Rent Control and Rent Stabilization. World Wide Web page <http://www.dhcr.state.ny.us/ora/pubs/html/orafac1.htm> (last modified October 20, 1999c). Ondrich, J., A. Stricker, and J. Yinger. Do Real Estate Brokers Choose to Discriminate? Evidence from the 1989 Housing Discrimination Study. Southern Economic Journal, 1998, 64(4), pp. 880-901. Rothenberg, Jerome, George Galster, Richard Butler, and John Pitkin. The Maze of Urban Housing Markets. Chicago: University of Chicago Press, 1991. Yinger, John. Measuring Racial Discrimination with Fair Housing Audits: Caught in the Act. American Economic Review, 1986, 76(5), pp. 881-93. Yinger, John. Closed Doors, Opportunities Lost: The Continuing Costs of Housing Discrimination. New York: Russell Sage Foundation, 1995.