Taxi service is one of the most important modes for urban transportation. In recent years, many taxi companies have been routinely collecting data to track the movement of each taxi for improving security, coordination, and service performance. This paper is intended to use the GPS vehicle positioning data to assess the route choice behavior of taxi drivers and explore if the routes selected by taxi drivers can be incorporated into a traveler information system. It is often perceived that taxi drivers have the ability to select quality routes assuming that: (1) they tend to be more knowledgeable about alternative routes and time-dependent traffic conditions than general public, including some publicly available route guidance systems due to the nature of their profession; and (2) they are typically more motivated to incorporate their knowledge about traffic conditions into their route choice decisions. An experimental study is conducted to examine the validity of these two assumptions. We have developed a framework that can effectively process the data into information about routes selected by taxi drivers and their associated travel times. The performance of the routes selected by taxi drivers is compared with the performance of those recommended by e-maps. Our results indicate that the routes selected by taxi drivers are generally more efficient than the routes recommended by some major e-maps, suggesting that taxi drivers are more active in selecting routes to avoid congestion.