Time-series studies on the association between daily mortality and air pollution levels have been criticized because they use background air pollution measurement sites to estimate exposure of the whole population, including those living along busy roads. To evaluate whether the exposure of people living along busy roads is estimated with error, we calculated separate effect estimates with separate exposure estimates using background and traffic-influenced measurement stations. We used Poisson regression analysis with generalized additive models to correct for long-term trends, influenza, ambient temperature and relative humidity, and day of the week. Black smoke and nitrogen dioxide were associated with mortality (relative risk of 1.38 and 1.10, respectively, for an increase of 100 microg/m3 on the previous day). Effect estimates were larger in the summer and in the population living along busy roads. Effect estimates were also larger using background stations rather than traffic stations. Overall, we found differences in the association between mortality and air pollution using different methods of exposure assessment; these differences are attributable to exposure misclassification for populations living along busy roads.