The estimation of production functions for wireless communications markets presents particular challenges including the lumpy nature of spectrum use, and the presence of significant upfront investment. In this paper a systematic study of production functions for wireless markets is undertaken, using data from India. The efficacy of two major functional forms, the Cobb Douglas and the translog functions, are compared by benchmarking the values of spectrum they yield against the values generated by the intuitive but data-intensive cash flow method, and the prices revealed in two auctions. The comparisons allow us to conclude that factoring the lumpy nature of spectrum use by considering ‘effective spectrum’ data rather than raw spectrum data, and relaxing the assumption of constant elasticity of substitution are useful in obtaining accurate estimates of spectrum value.