|
Local shopping (Sh) is the number of service and retail jobs per developed area within the
<br />zone, from the U.S. Census. Zones with centroids less than 1/4 mi apart were combined for
<br />this calculation. Local shopping is strongly correlated with density and transit, and did not add
<br />to the RZ after they were taken into account.
<br />The measure of pedestrian bicycle friendliness (Ped) is the number of census blocks per
<br />hectare (scale of the street grid) , plus an adder based upon the mean year the housing was
<br />built, both from the U.S. Census, with bonuses for traffic calming, good pedestrian conditions,
<br />bike lanes, paths, and bike parking, whether as part of the initial design or added later. Had direct
<br />measurements of the continuity, width and quality of sidewalks, nearness of buildings to the
<br />sidewalk, and traffic safety been available, we would have preferred using them to using the
<br />mean year the housing was built. A fine street grid shortens routes and offers more alternatives,
<br />and the frequent intersections slow traffic.' The measure works because older neighborhoods
<br />tend to have a fine street grid, sidewalks, narrow streets, slower traffic and buildings closer to
<br />the sidewalk.
<br />We also tested socioeconomic variables available at the zonal level: average household size,
<br />average household income and average per capita income. These were derived from the U.S.
<br />Census. However, we were not able to explore any independent impact of neighborhood
<br />parking supply or cost.
<br />Analysis of vehicles available and driving
<br />The correlation of each individual independent variable (locational and socio-economic) with auto ownership and
<br />VMT was tested. in the San Francisco area, for instance, Hh/RA explained 63% of the variance in Veh/Hh,
<br />followed by transit job access at 55%, Hh/TA at 52%, transit service at 49%, income/household at 43%, shopping
<br />at 35% and household size at 28%. In each of the three metropolitan areas, Hh/RA explained the most variance in
<br />Veh/Hh and VMT/Hh, see Figure 1. These correlations show a very strong relationship of residential
<br />density to auto ownership and driving in all three regions studied, even before evaluating the other
<br />variables--income, household size, transit service, pedestrian bicycle friendliness, etc. When tested
<br />against each other; density, transit, local shopping, center proximity and pedestrian/bicycle friendliness prove to be
<br />highly correlated. While this makes it harder to pick apart the separate influences, it means that density to some
<br />extent captures the effects of local shopping, transit and pedestrian and bicycle friendliness.
<br />As much as possible, we based the forms of our fits on simple modeling of the physical situations.
<br />For instance, since doubling density doubles the number of nearby destinations, and doubling
<br />transit service doubles the number of destinations you can easily reach, it is reasonable to expect
<br />that for both each doubling would decrease auto ownership and VMT by a similar percentage--a
<br />log-log or power fit. However, there are limitations on how many cars a household can own, and
<br />how many miles even the most auto-entranced folk can drive; therefore we modified the power fit
<br />form to be bounded as the density goes to zero. Similarly, since more people in the household
<br />increase the number of drivers and people to be driven around, we expected a linear relationship
<br />of household size with autos and VMT. However, we had less reason to anticipate a particular
<br />mathematical form (linear, power, bounded power, root, polynomial, exponential, etc.) for the
<br />relationship between income or pedestrian/bicycle friendliness and auto ownership or VMT. So
<br />we tested various mathematically simple forms that had appropriate behavior over the ranges of
<br />1~~
<br />6
<br />
|