A convenient representation for structural vector autoregressions

Latent variables are used to rewrite a wide class of structural vector autoregressive (SVAR) models. The framework is general enough to include as particular cases all just and over-identified models recently used in applied macroeconomics. The latent variables representation can conveniently be estimated with standard software packages like LISREL, EQS, LINCS and AMOS, for example. The approach is illustrated by using the models of Blanchard and Quah (1989) and Swanson and Granger (1997).