Modeling a micro-structure of agglomeration economies, this article derives a second-best benefit evaluation formula for urban transportation improvements. Without explicitly modeling the sources of agglomeration economies, Venables (JTEP 2007) investigated the same problem. This article examines how his cost-benefit measure should be modified when monopolistic competition with differentiated products provides a micro-foundation of agglomeration economies. Introducing the rural sector and multiple cities explicitly, we show that the result hinges on where the new workers come from. An improvement in urban transportation in one city increases its population but reduces those in other cities. If the population of the rural area (or, equivalently, the total population of the urban areas) is fixed, then the changes in the excess burden cancel out each other and only the direct benefit remains. If migration between the rural area and a city is possible, then a transportation improvement increases the total urban population and there will be positive additional benefits. If the number of cities changes, we have an additional change in the excess burden but the result depends on whether the city size is too large or not. In the former case, the induced effect on the number of cities has a tendency to reduce the social surplus.