Quote:
Originally Posted by dagobert
For your cross-sectional time series analysis on transit ridership to be useful you will want to use more variables than just gas prices, cost of parking downtown, and amount of parking available. Otherwise you'll have omitted variable bias. Preferably you will want to use quarterly data to increase the number of observations since n=30 is the bare minimum for it to be any good and also you should do seasonal adjustments. It might take some time to track down the sources for a lot of pertinent variables but it would make for a fascinating research study. I’m curious of results myself.
|
True.... or one can just review any of the multitude of already-existing studies (and resulting multinomial logistic choice models) regarding mode split; included variables almost invariable include not only things like income, travel time, and car ownership, but also variables like
out-of-pocket cost and
walking/access time, just two of many variables that capture the impact of parking availability on mode choice. All else equal, more parking -> cheaper parking rates + higher availability of parking in proximity to destination. Chicago's core did not empty out in the period 1984-1992, in fact this period encompassed a very substantial real estate boom primarily focused on commercial/office construction....and lots and lots of parking garages in the heart of the loop, in contrast to the previous paradigm of large surface lots on the periphery outside the loop, with no (new) parking allowed inside the loop.
Also, I'm not certain seasonal data would be absolutely necessary to draw conclusions; there is high seasonal cyclic variability of course, but these cycles occur annually, so as long as the annual data measure consistent time periods, they are comparable. To the extent seasonal data would be useful, you could do 12-month rolling averages and more precisely determine the inflection points to correlate to possible contributing events/policies/etc. Unfortunately, the farther back in time you go, the sparser such data, be it transit ridership or employment figures, gets.