Talk:Error correction model

Dr. Dogru's comment on this article

Dr. Dogru has reviewed this Wikipedia page, and provided us with the following comments to improve its quality:

Vecm should be modeled and depicted.

We hope Wikipedians on this talk page can take advantage of these comments and improve the quality of the article accordingly.

Dr. Dogru has published scholarly research which seems to be relevant to this Wikipedia article:

• Reference : dogru, bulent, 2013. "Seigniorage Revenue and Inflation Tax in Turkish economy," MPRA Paper 45538, University Library of Munich, Germany.

ExpertIdeas (talk) 23:45, 23 August 2015 (UTC)

Dr. Pagan's comment on this article

Dr. Pagan has reviewed this Wikipedia page, and provided us with the following comments to improve its quality:

The entry here associates error correction (EC) models with co-integration. Whilst it is true that co-integration involves error correction, EC models were around long before integrated and co-integrated processes were analysed seriously. A.W. Phillips was using them back in 1954, although it may even go back further. Empirical work often featured a partial adjustment model (PAM) that had the idea of an equilibrium value of some variable y of y*(t) to which y(t) would adjust as dy(t)=a(y*(t)-y(t-1)), where dy(t) is the change in y(t). Then one got dy(t)= a(y*-y(t-1)) . This is a special type of EC model. It can be written as dy(t)=a(dy*(t))+a(y*(t-1)-y(t-1)) whereas EC is more general with the form dy(t)=c(dy*(t))+a(y*(t-1)-y(t-1)). Researchers replaced the unobservable target path y*(t) by x(t)b where x(t) was observable and b was a parameter to be estimated. This was the logic that Sargan and Davidson et al used with their empirical work which featured EC models and it did not come from integrated processes. Later it was found that the EC model naturally came up with co-integration, as the unobserved path could be defined via the fact that if y was cointegrated with x the b would be the co-integrating vector. But the EC model is independent of the nature of the data and is just a statement about how one adjusts back to a target. In fact any distributed lag relation of the form A(L)y(t)=B(L)x(t) ( L being the lag operator Lz(t)=x(t-1)) can be written as an EC model, and these were very common in empirical work of the 19060s and 1970s. So I feel that some mention should be made of this rather than to leave the impression that EC models are only about co-integrating relations ( I have even had people tell me they can’t use an EC framework because they don’t have integrated processes).

I think the discussion of the estimation of VECMs is o.k. I am not sure I would spend much time on Engle-Granger as I don’t think it is used much today. Johansen’s approach does have a weakness in that the data must be described by a finite order VAR and we know that this is not always true, but it is probably the most widely used approach.

Re the references – Phillips is 1985 but in the text is 1986. Are you referring to his Econometrica paper or is this meant to be an extra reading?

We hope Wikipedians on this talk page can take advantage of these comments and improve the quality of the article accordingly.

Dr. Pagan has published scholarly research which seems to be relevant to this Wikipedia article:

• Reference : Pagan, A. & Pesaran, M.H., 2007. "On Econometric Analysis of Structural Systems with Permanent and Transitory Shocks and Exogenous Variables," Cambridge Working Papers in Economics 0704, Faculty of Economics, University of Cambridge.

ExpertIdeasBot (talk) 14:33, 19 May 2016 (UTC)

Dr. Lutkepohl's comment on this article

Dr. Lutkepohl has reviewed this Wikipedia page, and provided us with the following comments to improve its quality:

Under further reading I suggest to add a reference to Johansen, S. (1995), Likelihood-based Inference in Cointegrated Vector Autoregressive Models, Oxford University Press

We hope Wikipedians on this talk page can take advantage of these comments and improve the quality of the article accordingly.

Dr. Lutkepohl has published scholarly research which seems to be relevant to this Wikipedia article:

• Reference : Helmut Herwartz & Helmut Luetkepohl, 2009. "Generalized Least Squares Estimation for Cointegration Parameters Under Conditional Heteroskedasticity," Economics Working Papers ECO2009/42, European University Institute.

ExpertIdeasBot (talk) 09:20, 16 June 2016 (UTC)

Dr. Dreger's comment on this article

Dr. Dreger has reviewed this Wikipedia page, and provided us with the following comments to improve its quality:

First para: The variables have a common stochastic trend

Engle and Granger 2-Step approach: Equation is not valid, as it contains two constants. the gamma, and the alpha times beta0. I would suggest to delete the latter. The cointegration test on alpha does not have a standard distribution: Indeed, but also the Dickey-Fuller distribution cannot be used, as the critical values depend on the number of long run regressors

VECM para: Step 3: Test restrictions on the cointegration vectors and feedback mechanisms

Example: Permanent income hypothesis implies a 1:1 relationship between the logs of consumption and income in the long run

We hope Wikipedians on this talk page can take advantage of these comments and improve the quality of the article accordingly.

We believe Dr. Dreger has expertise on the topic of this article, since he has published relevant scholarly research:

• Reference : Christian Dreger & Jurgen Wolters, 2008. "Money Velocity and Asset Prices in the Euro Area," Discussion Papers of DIW Berlin 813, DIW Berlin, German Institute for Economic Research.

ExpertIdeasBot (talk) 19:54, 1 July 2016 (UTC)

Something I don't understand in the example:

If Ct=0.9Yt and the marginal is the derivative, since this is a linear model, how come the author chose delta Ct = 0.5 delta Yt? If this is a linear model then derivative should be constant. Ryaron (talk) 10:51, 20 April 2020 (UTC)