In running a cointegration test through ARDL model, all the other results are great except the coefficient for the adjustment (error correction) which is -1.5, what could be the reason(s) for this over-correction? Should anything be done?
when ecm -1.5 isn't acceptable. it means adjustment rate towards equilibrium 150% which indicates it is not gonna converse to equilibrium ever. you may add or drop few control variable.
Madam, It's not acceptable any case. Because the adjustment should be less that 100%. That means, if we conduct ECM estimation. The result lies between -0.01 to -0.99. Some time overestimated model can give more than 100% ECT. If you got this type of result, Please try to check your model and try to drop some of the least associated variable. Please check correction matrix before going to run model.
Check to ensure the dependent variable in the ECM is the first difference rather than the level of the series?
What are the variables in the system? Are they oscillating at a high frequency; e.g., switching from positive to negative values (or back again) each period? I suppose that this could explain a coefficient below -1.
In the univariate analogue to the VEC system, that is in a Dickey-Fuller equation (\Delta Y_t = b Y_{t-1} + e_t) , a coefficient on the lagged level in the range between -1 and -2 results from a negative parameter a in the AR(1) transformation (Y_t = a Y_{t-1} + e_t) of a stationary process; b = a - 1. I guess that in a system the loading of the EC term in this range could reflect a similar behaviour of the time series.
If the estimated model is stable and converge towards equilibrium, the error-correction term should be in between '0' and '-1' . If it is less than -1, it implies that the system is explosive and model is not stable.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping