This paper uses station-by-station panel data from 1977-2001 to estimate the demand for passenger boardings on the Metrorail in Washington, D.C. We compare the estimation results using fixed effects regression, random effects generalized least squares, random effects maximum-likelihood, and generalized estimating equations. The results are invariant to the estimation method, but are sensitive to inclusion of an explanatory variable that measures level of roadway congestion. The estimated price elasticity when roadway congestion is excluded is approximately -0.22, slightly lower than the -0.3 elasticity estimate for public transit commonly found in the literature. However, if we account for roadway congestion, the price elasticity increases significantly to -0.9. This suggests a potential downward bias in previous estimates that do not include a measure of roadway congestion. We also find a statistically significant positive, but inelastic short-run cross-price elasticity of demand between gasoline and the Metrorail of approximately 0.3 without, and 0.2 with, the roadway congestion variable. This suggests that the choice of transportation mode may be sensitive even to short-run fluctuations in motor fuel prices. Lastly, although the results indicate a trend away from Metro use over time, Metro ridership is shown to be positively correlated with roadway congestion and per capita income in the D.C. area.