In all but the most trivial settings Euler equation estimation in saddlepat
h stable systems is faced with a fundamental identification problem: the Eu
ler equation allows for an unstable root (of its characteristic equation),
while the data used to estimate the Euler equation also obey the transversa
lity condition, which rules out unstable rests. Thus, even if the model is
true, then the data are completely uninformative with respect to the unstab
le root of the Euler equation. But ignorance of the unstable root implies t
hat the parameters are not identified if the relationship between parameter
s and roots is one to one. We illustrate the issue using a linearized Euler
equation and present an application with OECD consumption data. (C) 2001 E
lsevier Science B.V. All rights reserved.