Do Large Residential Subdivisions Induce Further Development?

Abstract

Problem: Most previous research on land use change has been conducted at coarse scales with aggregated data, and there are relatively few studies that attempt to establish and examine these linkages at a disaggregate level. In addition, the cumulative effects of large residential developments have been little investigated as potential drivers of land use change.

Purpose: We hypothesize that residential subdivisions of 10 acres or more exert detectable priming effects and influence the rate of subsequent development in the vicinity. We use parcel-level data to test whether unilateral development decisions have spillover effects across both space and time.

Methods: We analyze parcel changes and then conduct a spatially explicit hazard analysis of land use change in Mecklenburg County, NC, to identify a set of factors that predict which vacant land parcels are most likely to experience a subdivision event (be subdivided) during a given one-year time period. We also perform a sensitivity analysis to assess how the specification of the priming effect measure affects the analysis, and to gain insight into the spatial extent of the hypothesized relationship between large residential subdivisions and subsequent development.

Results and Conclusions: We find evidence of a positive association between large residential subdivision events and an increased likelihood of subsequent subdivision activity. The sensitivity analysis suggests that this relationship is robust to changes in how the priming effect measure is specified and declines with distance. This corroborates previous assumptions that major residential subdivisions drive land conversion.

Takeaway for Practice: These results suggest that the project review process should consider impacts at a broader scale. Large residential subdivisions warrant attention not only for their localized impacts on traffic, stormwater, schools, and emergency services, but also for the signals they communicate to the larger development community. We find the effects of converting land to residential use cumulate not only over space, but also over time. Thus, they should be examined within the context of existing land use plans as well as infrastructure and service-delivery capabilities. By understanding the factors that explain parcel subdivision events and considering their capacity to induce further residential development, planners and the communities they serve will be in a stronger position to manage growth.

Research Support: None.

Keywords:

• land development
• hazard analysis
• Charlotte
• growth management

Notes

1. Also known as a mixed logit model, this approach accommodates unobserved heterogeneity in regulatory compliance. Train (2003) provides a detailed discussion of model specification and estimation.

2. Vacant parcels are specified as such in the local cadastral records and do not have structures on site.

3. The subdivision event could be considered complete at various points: when the land is purchased or optioned, when the subdivision application is approved, when the plat is final, at groundbreaking, when construction is complete, or when the residential structure is first occupied. The observations are interval censored in that the exact time of event occurrence is not available. Instead, events are only known to have occurred between one of the six time periods and that immediately following. This is also the motivation for adopting a discrete-time hazard analysis approach for studying the pattern of land subdivision in the study area (DeMaris, 2004, p. 430).

4. This is the minimum size at which parcels are subject to the local subdivision ordinance.

5. This occurs when two or more observations experience an event at the same time. The methods used here measure time in discrete intervals rather than continuously and accommodate multiple events within the same period.

6. Parcels that are aggregated are still assessed at each subsequent time period as candidates for an event.

7. The 12 categories are: residential 2+ acres, residential 0.5 to 2 acres, residential 0.25 to 0.5 acres, residential to 0.25 acres, heavy commercial, heavy industrial, institutional, light commercial, light industrial, water, transportation, and woods.

8. This variable was derived from the cumulative length of all streets within a given block group and the median parcel size, with the assumption that holding street length constant, infrastructure is more readily available in areas with smaller lots.

9. Estimates provided by local planners in each jurisdiction.

10. These values ranges from 0 to 1 and measure of the influence of growth moratoria on the probability of land subdivision (Bento, Towe, & Geoghegan, 2007).

11. An odds ratio is a ratio of the probability of an event for one group to the probability of the same event for another group, and can range in value from negative infinity to positive infinity (Pampel, 2000). Thus an odds ratio of 1 indicates equal probability for both groups.

12. We used a series of interaction terms derived from each of the significant predictors and the time indicator variables to assess the validity of this proportionality of odds assumption (Kleinbaum & Klein, 2005) and found no evidence to warrant retaining these interactions.

13. The number of observed events is used to normalize the pseudo R², thereby providing a more realistic representation of explained randomness. Although not equivalent to the familiar ordinary least squares R² statistic, the pseudo R² provides a general measure of the variation explained by a regression model.