![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
Abstract
This article takes a careful look at the recent state of sprawl among America’s 178 largest metropolitan areas through the lens of four sets of questions: (a) Measured at the metropolitan level, is sprawl really declining? Is it declining everywhere, or just in selected metropolitan areas? (b) If sprawl is indeed declining, are more compact growth forms on the rise? (c) If sprawl is indeed declining, is it the result of antisprawl land use and development policies? (d) Which metropolitan-level land market, demographic, and economic factors are most associated with changes in sprawl? It concludes that sprawl is indeed declining when measured by average population densities, but that the decline has been much less widespread if measured in terms of population growth in core-area neighborhoods, changing density gradient intercept and slope estimates, and increased employment clustering. In terms of policy, it finds no evidence that local regulatory regimes or growth management programs have had any effect on sprawl, but finds that the consistent administration of local regulatory programs in ways that incentivize infill development and send consistent signals to developers does contribute to reduced sprawl.
Keywords:
Acknowledgements
The author thanks the editor of Housing Policy Debate and three anonymous reviewers for their focused and thoughtful. comments.
Notes
1. Several researchers, including Malpezzi (Citation1999) and Torrens and Alberti (Citation2000) had previously developed robust metropolitan-scale measures of sprawl, but did not apply them to multiple metropolitan areas.
2. Including tabular and GIS data from the Census Bureau, and tabular data from the NRI, the American Housing Survey (AHS), Claritas Corporation, the Census Transportation Planning Package (CPCC).
3. This includes all U.S. metro areas with a 2000 population of 250,000 or more.
4. The original 1992 NLCD was published at the 100-meter scale, but subsequent versions were published at the 30-meter scale. The 1992 NLCD has since been backdated to the 30-meter scale, but because it uses a slightly different land-cover classification system than later versions do, it is not strictly compatible with those versions.
5. Lacking zoning, Houston is widely regarded as the archetypal laissez-faire U.S. metropolitan area when it comes to urban land-use regulation and sprawl. Portland, by contrast, has both a limited regional government and an urban growth boundary, and is commonly regarded as the most planned of U.S. metropolitan areas.
6. Census tracts with fewer than 100 residents were not included in the land area calculation.
7. The Central Business District, or CBD was identified as the downtown area zip code district (or districts) with the highest level of nongovernmental employment.
8. The estimated intercept and slope values were statistically significant for 1990, 2000, and 2010 for just 110 metro areas.
9. The calculated Moran’s I measures were statistically significant for 135 of the 178 largest metro areas.
10. Because density gradients are estimated using the logarithm of density, the intercept and slope coefficients are reported in terms of logarithms, and cannot be easily interpreted as actual values.
11. The population densities corresponding to these natural logarithm values are 4,770 persons per square mile in 1990, 4,870 persons per square mile in 2000, and 4,630 persons per square mile in 2010.
12. This analysis was limited to those metropolitan areas with a single central city; in which the density gradient intercept and slope estimates were statistically significant; and in which the regression R2 value exceeded 0.20. Among the large metro areas not meeting these criteria and thus not included in this analysis were Dallas-Ft. Worth, San Francisco-Oakland, and Los Angeles-Orange County.
13. As noted previously, Moran’s I varies nonlinearly between −1 and + 1, with positive values indicating greater clustering. Accordingly, an increase of 0.04 in the value of Moran’s I indicates a moderate increase in clustering.
14. Also known as nominal or fixed-effect variables, dummy variables take on just two values: 1 if the observation has the characteristic or attribute in question, and 0 if it does not.
15. The ten subindex areas included in the WRI are: local political pressure, state political involvement, state court involvement, local zoning approval, local assembly, supply restrictions, density restrictions, open space, exactions, and approval delays.
16. Individual index values were also computed for the 47 largest metropolitan areas, and these range from a high of 1.76 for Providence, Rhode Island (based on 16 survey responses), to a low of −.79 for Kansas City (based on 29 survey responses). In most cases, the metro area index values are within +/−.05 of their corresponding state value.
17. Multifamily dwelling units accounted for 62% of housing unit growth during the 1960s, 38% during the 1970s, 25% during the 1980s, 20% during the 1990s, less than 10% between 2000 and 2005, and nearly 50% between 2005 and 2010. (Source: www.census.gov/hhes/www/housing/census/historic/units.html)
18. Location quotient values are usually calculated from employment counts. Values less than 1 indicate that an industry is underrepresented in the local economy relative to the state or nation. Location quotient values greater than 1 indicate that an industry is overrepresented.
19. Among the metro areas included in this analysis, the correlation coefficient between 2006 wage rates and 2007 gross metropolitan product per employee was 0.7.
20. The Bridgeport metropolitan area includes Greenwich and Stamford, both of which are home to major hedge funds and financial firms.
21. Local governments do set property tax rates (and, in some cases, sales tax rates) but rarely in a manner to support particular development forms or densities.
22. Isolating the causal direction of events typically requires the use of more sophisticated econometric techniques such as two-stage least squares (2SLS). In the current case, however, we lack the presence of actual events and sufficient time points to use such techniques.
23. Standardized coefficients are obtained by dividing each variable’s slope coefficient by the standard deviation of the variable.
24. Initial density alone accounts for 38% of metropolitan density change between 2001 and 2011.
25. For the composite rank to be usefully different from the individual component ranks, the component ranks must be uncorrelated. Of the four component ranks (2010 NLCD density, 2010 core-area population share, 2010 density gradient intercept, and 2010 density gradient slope), only the 2010 NLCD density rank and the 2010 density gradient intercept rank were correlated; their Pearson’s correlation coefficient is 0.69.
26. Because they describe nonlinear relationships, logit model results can be difficult to interpret. In terms of signs, a positive coefficient value indicates that a larger value of the independent variable is associated with an observation being correctly classified (e.g., a sprawl leader or laggard). By the same logic, a negative coefficient value indicates that larger values of the independent variable are associated with the observation being incorrectly classified.