This paper studies the asset allocation problem of optimally tracking a target mix of asset categories when there are transaction costs. We consider the trading strategy for an investor who is trying to minimize both fixed and proportional transaction costs while simultaneously minimizing the tracking error with respect to a specified, target asset mix. We use imupulse control theory in a continuous-time, dynamic setting to deal with this problem in a general and analytical way, showing that the optimal trading strategy can be characterized in terms of a quasi-variational inequality. We derive an explicit solution for the two-asset case, and we use this to provide a sensitivity analysis, showing how the optimal strategy depends upon individual input parameters. We also use some theory for one-dimensional diffusion processes to derive analytical expressions for various measures of performance such as the average time between transactions.