Explaining the Support for Homeownership Policy in US Cities: A Political Economy Perspective

Abstract

Local officials in the US often extol the virtues of homeownership. Public proclamations about homeownership are frequent from these officials and echo many of the claims made by federal policy makers. While the federal government has promoted and funded policy to increase homeownership for decades, little is known about the level of financial support devoted to homeownership at the local level or the reasons for any variation in support among cities. This research examines the attitudes of city mayors about local policies and homeownership and investigates explanations for support of homeownership policy. In doing so, the research uses a public choice framework to explore the effects of inter-city competition and other factors on local housing policy making. Using data from surveys of city mayors and professional staff, as well as US Census data, this research finds that high levels of competition are associated with higher levels of support for homeownership. Furthermore, the results suggest that homeownership programs may be considered developmental policy in cities, not redistributive policy like most housing programs. The paper concludes with theoretical and policy implications based on the research results as well as identifying needs for future research on local homeownership policy.

Keywords:

• Homeownership
• tenure
• housing policies
• cities

Acknowledgements

The author wishes to thank William Rohe and Valerie Jenness for helpful comments on an earlier draft of this paper. The author is also thankful to the National Science Foundation (SBER-9630638) and the US Department of Housing and Urban Development for funding the collection of data used in this research. The opinions, interpretations, findings, and conclusions contained in this publication are those of the author and do not necessarily reflect the views of the National Science Foundation or the US government.

Notes

¹ Empirical research indicates that homeownership does have some positive benefits for individuals and the community. However, the evidence is mixed and not all claims of the positive effects of increased homeownership have been tested empirically (see McCarthy et al., 2001; Rohe et al., 2002; and Rossi & Weber, 1996 for findings from research on the benefits of homeownership).

² McCarthy et al. (2001) provide an in-depth discussion of the economic benefits of homeownership to individuals and society. In addition, see US Department of Housing & Urban Development (2002) for a discussion of the economic benefits of homeownership.

³ The definition of affordable housing is vague, but one definition includes programs that serve households up to 120 per cent of an area's median household income (Basolo, 1998). However, many programs are designed to serve households at or below the median income. Regardless of the definition, affordable housing programs generally are not designed to directly serve the more affluent members of the community.

⁴ Tiebout's argument was entirely theoretical and depended on a number of restrictive assumptions, e.g. no barriers to mobility and full knowledge of communities. Despite questions concerning the reasonableness of the assumptions, many scholars generally agree with Tiebout's argument (see Peterson, 1981; Schneider, 1989).

⁵ While Fischel's argument possesses prima facia logic and he presents some convincing evidence from case studies and the literature, a thorough analysis would require knowledge about the preferences of homeowners, evidence that local decision makers were aware of homeowners' preferences, and a clear linkage between homeowner preferences and decisions by local elected officials. It would be interesting to investigate which issues prompt homeowners to organize or not. For example, it is probable that location issues, such as siting of low-income housing developments, raise much more concern among local residents (homeowners and renters) than approval of expenditures for housing programs in general.

⁶ Logistic regressions using response as the dependent variable (1 = responded, 0 = no response) were run using city-wide characteristics, such as per capita income, median housing value and homeownership rate, as explanatory variables (see Basolo, 1998 for full results of these analyses).

⁷ The mean or mode of the relevant variable can be substituted for missing values. Additionally, in some cases, conditional mean imputation may be used to predict the missing values. In general, these approaches are adequate, unless the total sample is relatively small and/or missing values comprise more than a small proportion of the data. In these cases, another imputation technique involving iterative simulation is recommended (Hoyle, 1999). For the multivariate analysis in this paper, the focus was on cases with complete or nearly complete information. Only a few cases on two variables used in the analysis (inter-city competition and fiscal health) had incomplete information and the mean was substituted for the missing values.

⁸ The expenditure figures were collected via a random sample survey. While the survey response rate was relatively high (61.6 per cent), the author of the study notes that cities with higher unemployment rates were less likely to respond to the survey and, therefore, some caution should be exercised in generalizing to the population of all US cities (Basolo, 1999).

⁹ The survey questionnaire to mayors defined affordable housing as housing affordable to different income groups. Specifically, the questionnaire presented the following income categories: “moderate-income is defined as households between 81–120 per cent of the area median income. Low-income is defined as households between 50 per cent and 80 per cent of area median income. Very low-income is defined as households with incomes below 50 per cent of the area median”.

¹⁰ Initially, the dependent variable was divided into quartiles categorized as low (0–34 per cent; n = 47), low-moderate (35–67 per cent; n = 48), moderate-high (68–94 per cent; n = 48), and high (95–100 per cent; n = 48). The categorization of the dependent variable, either as discussed in the body of this paper or by quartiles, might be questioned as arbitrary. Therefore, the regression analysis was run for each dependent variable to compare results. The model using the categorization by quartiles had similar results except for two variables (using p = 0.10 as the criterion): ownership was not statistically significant and percentage non-white was statistically significant. I decided to use the categorization that produced the best fit as measured by the model χ².

¹¹ The recommended analytic approach with an ordered categorical dependent variable is the ordinal logit model (Liao, 1994; Long, 1997). This model requires adherence to the parallel regression (proportional odds) assumption. A test on this study's model showed a violation of the parallel regression assumption. Long (1997) finds that rejection of this assumption is common and recommends the multinomial logit model (MNLM) as an alternative under these circumstances. He notes that using the MNLM, when the data are in fact ordered, results in “…a loss of efficiency, since information is being ignored” (1997, p. 149). However, the efficiency loss is acceptable to overcome the violation of the parallel regression assumption of the ordered logit model.

¹² The results shown here are from the final specification of the model. Initially, housing characteristics including housing physical conditions and the overcrowding rate, as well as two interaction variables, affluence index*ownership rate and percent non-white*ownership rate, were included in the model. These additional variables did not substantially change the difference between the − 2 log likelihoods of the fitted model versus the intercept only model. In addition, the coefficients for these variables were not statistically significant nor did they alter the substantive results of other variables for the contrast of most interest (low-to-high support). Therefore, a decision was made to present the results from the more parsimonious final model.

¹³ See Galster et al. (2004) on Community Development Block Grant expenditures in cities and Tao & Feiock (1999) on the targeting of local economic development activities.

¹⁴ Some housing scholars also favor public incentives to attract middle-income residents back to declining central cities (see Varady & Raffel, 1999). While this strategy might be appropriate in some urban places, the results from the research in this paper indicate that places with higher homeownership rates tend to spend a larger proportion of their funds on homeownership programs. Again, future research needs to examine targeting of housing resources and the distribution of benefits from targeted programs.