Agenda - 10-07-2008- 7b - Laserfiche WebLink

Browse Search

222 the independent variables. We derived the forms of the equations using the San Francisco data and then used these forms to derive the equations for the other two regions. Using the data available in all three geographical regions and the same equation forms, the variables which consistently explain the most variance in vehicles/household (Veh/EIli) are net residential density (I~~/RA), per capita income ($/P), household size (P/I~ and transit accessibility (Tr), see Table 1. For vehicle miles traveled/vehicle (VMT/Veh) the best variables are total residential density (Hh/TA), P/H, pedestrianl6icycle friendliness (Ped) and $/P. These are the equations which when applied in the same form in all three regions give the highest RZ. In both the Veh/Hh and VMT/Veh models, density is raised to a negative power, so doubling density causes a fixed decrease in Veh or VMT. In the Vel~/Hh model, transit service is raised to a negative power, so doubling Tr causes a fixed decrease in Veh. Household size acts as expected, each additional person adds a fixed increase in Veh' or VMT. Improvements in pedestrian and bicycle friendliness reduce VMT/Veh, but with the square root of Ped. The impact of per capita income is a little less straight-forward. It increases with ownership by declining increments as income increases. But VMT/Veh decreases as income increases. The result is that up to an annual income of $25,000 to 30,000, VMT/Hh increases with income as expected, but levels off and falls slightly at higher incomes up to $100,000 per person. The best San Francisco model for VMT/Veh using variables available in all three regions is the same as in Table 1, but with a function for Zonal Transit Density in place of that for Households/Total Acre. Tr gives an RZ of 44.1%, which is slightly better than the 43.8% in the above table. But in the other two regions, Tr in place of H/RA. reduced the RZ to ,40.0% for Chicago and 37.6% for LA, substantially poorer than the I3h/Tot Acre fit. So the recommended equations use Hh/Tot Acre. In all three cities, the R2 for VMT/Veh is much lower than the Rz for Vel~/I-Ili, indicating that neighborhood conditions more strongly impact the decisions on how many vehicles to have available than they do each decision to use the vehicles on hand. However, since Veh/Hh varies much more from zone to zone (1 s.d. equals ±25% for San Francisco) than does VMT/~Ili (±9%), Veh/Hh is more important, and the RZ for the resulting VMT/Hh is almost the same as for Veh/Hh. Data for center proximity was only available for the San Francisco region,, so none of the fits in Table 1 are based on it. The Veh/Ilh fit in Table 3 is already highly significant, and center 7