Airline reservation systems involve the use of booking policies to implement a predetermined allocation of seats to different fare classes. Models for optimal allocation of seats typically assume one of two commonly used booking policies, often without recognizing the differences between them. In this paper, we present alternative representations of these booking policies, and demonstrate that even with identical seat allocations the two booking policies may result in different expected revenues. We also show conditions under which one of the policies is better. Our Markov chain models facilitate optimization of seat allocations given either booking policy. Examples are given.