An econometric procedure to identify the permanent shocks in vector error c
orrection models is proposed, which allows one to combine long-run and cont
emporaneous restrictions. This procedure is applied to the six-variable mod
el of King, Plosser, Stock and Watson (1991) with a view to providing an al
ternative interpretation to their results based on a different identificati
on scheme. We argue that a real spending shock in the place of the real int
erest rate shock appears to better accommodate their empirical findings.