Empirical estimates of the private value of patent protection are found for four technology areas - computers, textiles, combustion engines and pharmaceuticals - using new patent renewal data of West German patents collected for the period 1953-1988. In Germany, a patentee must pay an annual renewal fee to keep his patent in force. Two dynamic discrete choice models of optimal renewal decisions are developed and used. in conjunction with observed hazard proportions and renewal fee schedules, to estimate the returns to protection. Differences in value across technology, nationality of inventor and time are explored both non-parametrically and parametrically within a deterministic framework. A stochastic formulation of the model, which allows both for learning about the innovation and market and for the possibility of infringements, is estimated using a minimum distance simulation estimator. The evolution of the distribution of returns over the life of a group of patents is calculated for each technology. Results indicate that learning is completed after 6 years, that obsolescence is rapid, and that the distributions of patent value are very skewed. Research and development (R&D) expenditures for each technology area are calculated and patent protection as an implicit subsidy to investment in R&D is discussed. Patent protection is valuable only when there are potential competitors for the use of an innovation. Patent rights must be defended. A game theoretic analysis of litigation explores how these facts influence the decision whether to apply for and keep a patent in force and, in tum, the relationship between the distribution of patent value and that of the underlying innovation. Implications for renewal behaviour are derived from the analysis and the data suggests that the level of potential competition does affect the value of protection. Consideration is given to how these findings bear on the interpretation of empirical estimates of patent value as indicators of innovation.