Tene,
This is a great and very important example. All of the issues you raise are the challenges that have made it difficult to establish conclusively the health effects of transit policies. Conceptually, what would be the *ideal* study? It would randomize different communities to different policies. The unit here is really the transit system, not the individual, so you may want to do the next best thing to a trial as an observational studies with cities that had different policies. But, cities differ in many ways, so it is hard to make a compelling case that health differences between cities are due to transit policy only. The next best thing might be to study what happened when a city changes its policy on transit, e.g., introduces a broader service system or opens a new transit stop. This has been done sometimes. For a nifty example, check out Magda Cerda's study in Colombia:
Magdalena Cerdá, Jeffrey D. Morenoff, Ben B. Hansen, Kimberly J. Tessari Hicks, Luis F. Duque, Alexandra Restrepo, Ana V. Diez-Roux; Reducing Violence by Transforming Neighborhoods: A Natural Experiment in Medellín, Colombia. Am J Epidemiol 2012; 175 (10): 1045-1053. doi: 10.1093/aje/kwr428
I think more opportunities are possible because transit systems do change, e.g., new stations open. One challenge is that such neighborhood changes fairly quickly prompt gentrification, which may lead to better health of people in the neighborhood because healthier people moved in rather than because the same people got healthier. So you really need a longitudinal study that follows the same individuals over time as transit improvements occur.
Also: access to employment is a major consequence of good transportation, so any health outcome potentially influenced by money is also potentially influenced by transit.
Maria