In the presence of uncertainty and irreversibility, dynamic decision problems should not be solved using expected net present value analysis. The right to delay a decision can be valuable. We show that the value of this right equals Arrow and Fisher's [Quart. J. Econom. 88, 312-319 (1974)] quasi-option value. In a discrete model we show how to derive quasi-option value using methods from finance. These methods yield the advantage that they permit avoidance of the common pitfall of improperly matching a riskless discount rate with a risky project. In our arbitrage-free model, use of the riskless rate is appropriate. Two main findings are presented. First, if the stochastic dynamic process underlying the problem is known, the Arrow and Fisher [Quart. J. Econom. 88, 312-319 (1974)] and Henry [Amer. Econom. Rev. 64, 1006-1012 (1974)] result, that improper use of net present value too often leads to early development, is correct. Second, if the process is assessed incorrectly, their result can be incorrect in the sense that net present value methods may lead to the correct outcome while the dynamic rule does not.