The determinants of urban sprawl: theory and estimation

ABSTRACT

We argue that a well-articulated theory, by which we mean a set of structural equations equal in number to the endogenous variables and from which testable hypotheses may be drawn, should be the basis for any effort to estimate the determinants of urban sprawl. Without such a theory, it is not possible to know why a particular determinant ‘works’ to explain a particular definition of urban sprawl, nor is it possible to know whether any particular policy to combat sprawl, however defined, will be successful in achieving that objective without also creating other, possibly adverse, effects. To illustrate our argument, we contrast Burchfield, et al. [2006. Causes of sprawl: A portrait from space. The Quarterly Journal of Economics, 121, 587–633], which is not based on a well-articulated theory, with the urban monocentric model [Brueckner, 1987. The structure of urban equilibrium: A unified treatment of the Muth-Mills model. In E. S. Mills (Ed.), Handbook of regional and urban economics, Vol. II, “urban economics” (pp. 821–845). Amsterdam: Elsevier], which is a well-articulated theory.

KEYWORDS:

• Urban sprawl
• undeveloped land around an average dwelling
• urban spatial size
• monocentric urban model

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Burchfield et al. (2006) use sprawl in a positive, not normative, sense, and we adopt that usage here.

2. ‘To measure the extent of sprawl, for each 30-meter cell of residential development, we calculate the percentage of open space in the immediate square kilometre. We then average across all residential development in each metropolitan area to compute an index of sprawl’ (Burchfield et al., 2006, p. 599).

3. Oueslati et al. (2015, p. 1594) obtain similar results: ‘[T]he fundamental conclusions of the standard monocentric model are valid in the European context for both indices. Although the variables generated by the monocentric model explain a large part of the variation in artificial area [their measure of urban spatial size], their explanatory power for modelling the fragmentation index [their measure of urban fragmentation] is relatively low.

4. As implied by Brueckner (1987, p. 835) and made explicit by Anas and Kim (1993), the distance at which the quotient h/q (and therefore the population density function) pivots (if it does) is farther out than x′.

5. In this subsection, all page and subsection numbers refer to Burchfield et al. (2006), unless otherwise stated.

6. Burchfield et al. (2006) use the term ‘monocentric city model’ even when they are drawing implications from polycentric and city-system models. In all of the following quotes, italics are in the original.

7. As noted earlier, the other two exogenous variables of the model, δ, the radians of a circular urban area available for residential use, and i, the rental rate of the capital input, have not been used in empirical work.

8. Two of the MAs in the Burchfield et al. (2006) sample contain two separate urbanized areas, which makes a total of 277 observations.

9. The coefficients in Table 4 of Burchfield et al. (2006, p. 616) ‘give the impact on the index of a one-standard-deviation increase in the corresponding variable.’ Our coefficients give the impact of a unit change in the corresponding variable, so by ‘consistent’ we mean in terms of sign and statistical significance.