The first section describes Ehrlich's murder rate function and tests the model for parameter stability over the sample period. The second section considers Ehrlich's choice of explanatory variables and functional forms, and examines sensitivity of his estimate to these choices. Finally, the authors examine the drawbacks of inferring an execution-murder tradeoff from a single estimated equation. They conclude that the time-series model and data used in the original study permit no inference about the deterrent effect of capital punishment on homicide. The parameters of Ehrlich's model are extremely sensitive to the choices of included explanatory variables and the functional form of the model. When the model is recast in alternative, but plausible, structures, it fails to generate a deterrent effect for executions. Ehrlich replies to this critique, reports additional findings concerning the issue of deterrence based upon independent bodies of data, and relates them to results obtained through the time-series analysis.