A general stochastic coefficients model developed by Swamy and Tinsley serves as a reference point for discussion in this second of a series of three articles Other well-known specifications are related to the model. The authors weigh the advantages and disadvantages of stochastic coefficients and suggest procedures to address the identification and estimation problem with weaker and noncontradictory assumptions They argue that the real aim of inference is prediction and that "imprecise" parameter estimates of a coherent model are acceptable if they forecast well.