### Forecasting with an ARDL ECM

Posted:

**Mon Apr 08, 2019 9:34 pm**Hi Eviews Support,

This should be a simple one. Is there an in-built way to forecast in ECM form?

For example, I can estimate an ARDL in log-levels. Then I can transform it into ECM form using the 'Error correction form', e.g.:

equation Example_1.ardl log(Cons) log(income)

equation Example_1.ecreg

The latter specification is useful to me, because rather than using the spurious relationship in levels, it is explicitly estimating the correction coefficient in log-differences. It is large and statistically significant; so far so good.

But when I click 'Forecast', I get the same forecast series regardless of whether I selected the log differences ECM or the 'spurious ARDL'. I suspect that the forecast function is using the base specification in log-levels and doesn't recognize that I want it in error correction form.

Could you cast some light on this? Is there an in-built way of making ECM forecasts? Or do I need to write an iterative loop for computing dynamic forecasts of the error correction term?

Thank you

This should be a simple one. Is there an in-built way to forecast in ECM form?

For example, I can estimate an ARDL in log-levels. Then I can transform it into ECM form using the 'Error correction form', e.g.:

equation Example_1.ardl log(Cons) log(income)

equation Example_1.ecreg

The latter specification is useful to me, because rather than using the spurious relationship in levels, it is explicitly estimating the correction coefficient in log-differences. It is large and statistically significant; so far so good.

But when I click 'Forecast', I get the same forecast series regardless of whether I selected the log differences ECM or the 'spurious ARDL'. I suspect that the forecast function is using the base specification in log-levels and doesn't recognize that I want it in error correction form.

Could you cast some light on this? Is there an in-built way of making ECM forecasts? Or do I need to write an iterative loop for computing dynamic forecasts of the error correction term?

Thank you