Methods of tax collection employed by modern governments seem dull when compared to the rich variety observed in history. Whereas most governments today typically use salaried agents to collect taxes, various other types of contractual relationships have been observed in history, including sharing arrangements which divide the tax revenue between the government and collectors at fixed proportions, negotiated payment schemes based on the tax base, and sale of the revenue to a collector in exchange for a lump-sum payment determined at auction. We propose an economic theory of tax collection that can coherently explain the temporal and spatial variation in contractual forms. We begin by offering a simple classification of tax collection schemes observed in history. We then develop a general economic model of tax collection that specifies the cost and benefits of alternative schemes and identifies the conditions under which a government would choose one contractual form over another in maximizing the net revenue. Finally, we use the conclusions of the model to explain some of the well-known patterns of tax collection observed in history and how choices varied over time and space.