Zit 7 Some argue that GMM regressions suffer from weak instrument problems or parameter
identification problems (e.g. Stock and Wright, 2000; Kleibergen and Mavroeidis, 2009; Newey
and Windmeijer, 2009). Lubik and Schorfheide (2007:1073) argue that system-based estimation
methods, such as the Bayesian approach or the full-information maximum likelihood approach,
are superior in dealing with endogeneity identification issues in estimation since it accounts for the
endogeneity by adjusting for the non-zero conditional expectation of the monetary policy shock.
Bayesian methods, however, may end up covering identification problems if improperly used, in
which identification problems emerge because not all the structural parameters are recoverable
from the semi-structural ones and the objective functions are poorly behaved (Canova, 2009:311).