Borrowing Residential Trip Generation Rates

Abstract

Residential trip generation rates are a fundamental component of transportation planning. To investigate discrepancies in these rates, residential trip generation rates for nine suburban neighborhoods were computed using four different methods: (1) ground counts conducted at the neighborhoods, (2) household surveys distributed to the neighborhoods, (3) application of national trip generation rates published by the Institute of Transportation Engineers (ITE), and (4) rates derived from the trip generation component of regional urban travel demand models for the neighborhoods. Agencies generally use one of these rates, and by determining all four for the same set of neighborhoods in a controlled study, one can ascertain the extent to which these rates are likely transferable. Rates based on the first three methods were not significantly different. For developments composed solely of single-family detached homes, the average residential trip generation rate was 10.8 based on the site-specific ground counts, 9.2 based on site-specific household surveys, and 9.6 based on ITE’s trip generation rates. However, rates based on the fourth method were significantly different, with a mean rate of 6.4. The greatest differences occurred when the long-range regional model used person trips that were converted to vehicle trips rather than estimating vehicle trips directly. Although the summary statistics presented in this paper will not surprise transportation planners, they illustrate two caveats for balancing data collection costs and need for site-specific information. First, a subtle change in how some rates are calculated limits their utility elsewhere. Second, even when equivalent methods for determining rates are used for similar neighborhoods, differences will occur because of the large and random variation inherent in trip generation. Borrowing rates may indeed be tolerable, but only if one gives the full range of rates possible from this probabilistic process rather than just the expected mean rate. Both caveats are treatable provided they are explicitly addressed through uncertainty analysis or adjustments for expected bias.