Johansen's test for cointegration is applied to Litterman's original sixvariable Bayesian vector autoregression (BVAR) model to obtain vector error correction mechanism (VECM) and Bayesian error correction (BECM) versions of the model. The Brock, Dechert, and Scheinkman (BDS) test for independence from the non-linear dynamics literature is then applied to the error structures of each estimated equation of the BECM and VECM models, plus two BVAR versions of the model. The results show that none of the models produce independent and identically distributed (ID) errors for all six equations. However, the BDS results suggest the elimination of the Bayesian prior from the BECM model, given that the univariate VECM errors are IID in five equations, compared to only two or three equations under the three Bayesian restricted models. These results combined with previous evidence regarding the superior forecasting performance of BECM over ECM models suggest future experimentation with less restrictive BVAR priors, BECM models corrected for heteroscedasticity, or hybrid specifications based on the nonlinear dynamics literature. KEY WORDS cointegration; error correction; vector autoregression; non-linear dynamics; BDS test; error analysis Many important contributions have been made in time-series analysis since the late 1970s. Two of the more important are Bayesian vector autoregression (BVAR) and error correction models. BVAR models were first introduced in 1979 and have since been shown to be as accurate, if not more accurate, than large structural models and other time-series methodologies (see Litterman, 1986; McNees, 1986). Since that time, cointegration and error correction have received more attention, particularly since Engle and Granger's (1987) cornerstone piece on representing, testing, and estimating error correction models. LeSage (1990a,b) and Shoesmith (1992) apply Engle and Granger's procedures and show that error correction mechanism (ECM) and Bayesian ECM (BECM) models may offer substantial improvements in forecast accuracy over vector autogression (VAR) and BVAR, particularly over longer forecast horizons. More recently, Engle and Granger's (1987) two-step test for cointegration has been shown to be most appropriate for systems of only two or three variables. In the more practical case of several variables, the cointegration test by Johansen (1988) is preferred in that the procedure identifies all possible cointegrating vectors within a multivariate system rather than only the 0277-6693/95/030311-14