VECM - Interpretation of Estimates

[LBS_NK]

Hi All,
I am currently working on finalising part of a thesis, looking into the inter-relationship between flows and 3 asset classes (stocks, bonds and housing indices). I have used the VECM as my series are cointegrated. However i am struggling with understanding the outputs of the VEC Estimates in eviews.
Can anyone shed some light on what the ECT term in the table is and what these output values explain about the short term relationships between the 4 series., i.e what is the correct value to analyse for equilibrium speeds of adjustment, figures that help explain the short term relations....etc
I have also uploaded an excel spreadsheet of the VEC Estimates

VEC Estimates.xlsx

Output of the VEC Estimates from E Views 6

(13.96 KiB) Downloaded 3417 times

Please feel free to comment.
All help much appreciated.
rgds.

[loooooo]

Hi,

Being not familiar with your 4 variables, I can't tell you much about the result, but it looks strange that a differenced term is entered into CE and as a result that the twice-differenced term of that variable appears in the equation. Are all your 4 variables I(1), including d(ASA)? Even if it is, simply once or twice differencing all the variables to match their levels of integration may look ok statistically but maybe not theoretically correct. How would you interpret a differenced variable affecting other variables in level, and/or vice versa? Theoretical justification from the literature, or a firm hypothesis of yours, should precede any statistical testing. I'm not saying what you're doing it wrong, but just saying you should be very careful.

[LBS_NK]

Hi loooooo,
Many thanks for your post. Yes, correct all my variables are I(1) and the variables are indices on bonds, stocks, house prices and flows into a country. I understand where you are coming from, but the justification on asset returns does follow a couple of papers, working with differenced variable and logged series as well. This helps induce the possibility of applying my model. I ran a Granger causality / block wald on the VEC and do realise that the albi is not significant in the interrelationships, but i am still struggling with the speeds of adjustment for each variable on the system and periods to equilibrium.

[donihue]

In principle, the speed of adjustment parameters α(y) (the coefficients on "cointeq1") should be negative and lie between (0, -1). This is because in a VEC model, the point estimate should imply that y(t) converges to the long-run equilibrium relationship – if y is above its long-term value (ecm term >0), y must decline (α(y) <0) and if y is below its long-term value (ecm term <0), y must rise (α(y) <0)

In your case, only one of the α(y) meets this condition (that for D(Albi)), but it is insignificant. This suggests that your model specification needs adjustment and/or that your data is inadequate. The latter is certainly a problem, as you are trying to estimate a total of 72 parameters with only 56 observations.

Regards
Donihue

[gobbble]

I also have a question about the VECM interpretation.

I understand about the speed of adjustment interpretation, which is signifcant in my case. I am just not sure how to interprete in which direction the adjustments goes.

My OLS regression on first differenced variables indicated a negative relation between between exchange rate and price index.

This seemed to be intuiteively wrong to me so I constructed a VECM.

I find a ling-term equilibrium with speed of adjustement of 4 periods, but no significant short term relation.

So what can I conclude from the VECM model how the exchange rate and the Price index behave? I would like to have an indication of a direction. Thanks!

[gobbble]

Can coone help me here with the interpretation?

since there is a long-term equilibrium, does it mean, that in the long-run exchange rates can not influence the price index as they always balance back to the equilibrium?