The RATS Software Forum

Hi Tom,
i want to estimate a vertical price transmission model using two prices (wholesale and retail) and test for asymmetry in the the conditional mean using asymmetric VECM and the conditional variance using asymmetric BeKK-GARCH.
the literature review has focused on a single equation asymmetric ECM (i.e. Granger and Lee, 1989; Grasso and Manera 2007) where segmenting the error correction term and the x's (lagged dependent and exogenous variables) into positive and negative values and i want to extend it to be multivariate. is this approach of using asymmetric VECM correct?

thank you for your help

references
Granger, C.W.J., Lee, T.H., 1989. Investigation of production, sales and inventory relationships using multicointegration and non-symmetric error correction models. Journal of Applied Econometrics 4, S145–S159.
Grasso, M., Manera, M., 2007. Asymmetric error correction models for oil-gasolin price relationship. Energy Policy 35, 156-177.

If you're asking if it's a reasonable approach, the fact is that both equations in the VECM have identical explanatory variables, so there is unlikely to be much of a difference due to moving to a multivariate estimation---if there's a tangible difference it would be due to the "GARCH" part, not the MV part. If the literature on this shows that there is decided heteroscedasticity, then you might get results that lead to different conclusions if you allow for that in the estimation. (Of course, the OLS estimates in the ECM form are consistent even in the presense of GARCH effects, so also might not.) Regarding doing an ABEKK, note that the asymmetry in the typical ABEKK model is based upon signs of the residuals, while the asymmetry in the VECM is based upon the signs of the data.