40 JOURNAL FOR ECONOMIC EDUCATORS, 10(1), SUMMER 2010
Public Transportation Ridership Levels
Christopher R. Swimmer and Christopher C. Klein 1 Abstract This article uses linear regression analysis to examine the determinants of public transportation ridership in over 100 U. S. cities in 2007. The primary determinant of ridership appears to be availability of public transportation service. In fact, the relationship is nearly one to one: a 1% increase in availability is associated with a 1% increase in ridership. The relative unimportance of price may be an indicator of the heavy subsidization of fares in most cities, leaving availability as the more effective policy tool to encourage use of public transport. Key Words: identification, public transportation, ridership. JEL Classifications: A22, C81, H42 Introduction What makes one city more apt to use public transportation relative to another? This is an important issue that has been studied by others in various ways. Glaeser et al. (2008), find that the availability of public transportation is a major explanatory factor in urban poverty. Glaeser and Shapiro find evidence that car cities, where a large percentage of people drive themselves to work, grew at the expense of public transportation cities as the percentage of cities’ population taking public transportation declined between 1980 and 2000. Murray et al. (1998), conclude that the performance of a public transport system is determined largely by the proximity of public transport stops to the regional population. Initially, the data were gathered for the top 136 metropolitan statistical areas in the U.S. using the raw number of unlinked trips on public transportation as the measure of ridership. Due to the wide variation in population and ridership across cities, the per-capita unlinked trips were calculated for use as the dependent variable. Missing values reduced the number of observations to 105. The regression analysis utilizes the process of backwards selection by eliminating a single variable per regression based on the highest P-value that is attained. This process will continue until each coefficient’s p-value is less than .10. First, however, Park’s test for heteroscedasticity is performed. After the backwards selection, an F-test is performed to confirm the appropriateness of the resulting equation.
Mr. Swimmer, as an undergraduate Economics major at Middle Tennessee State University, prepared the initial draft of this paper for an Econometrics class taught by Dr.Klein in the Economics and Finance Department at Middle Tennessee State University during the Spring Semester 2009. Christopher C. Klein is Associate Professor, Economics and Finance Department, Middle Tennessee State University, Murfreesboro, TN. firstname.lastname@example.org.
41 JOURNAL FOR ECONOMIC EDUCATORS, 10(1), SUMMER 2010
Data and Variables Eleven independent explanatory variables were chosen as described below. Metropolitan Density: Density in this analysis will be defined as the population divided by the metropolitan area. The reasoning for selecting this is that public transportation is more efficient in areas of higher density. The coefficient should be positive in value. Metropolitan Area: The Metropolitan Statistical Area is generally characterized as the official area of an incorporated city along with its immediate sphere of economic influence. There are two opposing theories for this variable. The first is that a larger area makes public transportation less efficient for those riding. The second is that a larger area simply implies greater economic activity. Average distance: Average distance is the typical number of miles traveled using public transportation. The theory is that a person who expects a longer trip will be less likely to endure the apparent inconvenience of using public transit. The coefficient should be negative. Service availability: Service availability is the maximum...
Please join StudyMode to read the full document