ARDL short run coefficient

[oussama zen]

Dear community,

I am coming back to you because I have a small problem concerning the estimation of the short-term coefficients.

when I have a variable of my model whose lag is zero, I do not receive the short term coefficient.

is it normal?

thank you for advance for you help.

[Sebastian Kripfganz]

When you do not have distributed lags of a variable, then you are estimating only 1 coefficient for that variable. Thus, you cannot separately identify a long-run and a short-run effect. Conditional on the response to long-run relationship, there are no additional short-run effects for this variable. The dependent variable is still responding in the short-run to a change in that independent variable, but only because such a change triggers a deviation from the long-run equilibrium which is then corrected through the equilibrium adjustment mechanism.

[oussama zen]

Thank you dear Sébastien

[Pita Fouta]

Dear oussama zen

I really need your help on rolling Granger causality. I wrote you a private message.

Thanks in advance.

Pita

[Alex Grisdale]

Other papers I have seen still generate short run coefficients when they have 0 lags but stata does not seem to do this. They also seem to generate a coefficient for the error correction term, where as stata does not. This may be a silly question but just curious to why this is the case? As when writing a methodology have zero lags for all 3 of independent variables, so was wondering if there is even a need to write an error correction model specification?

[Alex Grisdale]

Other papers I have seen still generate short run coefficients when they have 0 lags but stata does not seem to do this. They also seem to generate a coefficient for the error correction term, where as stata does not. This may be a silly question but just curious to why this is the case? As when writing a methodology have zero lags for all 3 of independent variables, so was wondering if there is even a need to write an error correction model specification?

Sorry for context, I have used the code "ardl Y X1 X2 X3 X4, lags(1 0 0 0) ec btest"
The btest comes back as reject the null hypothesis, so a long run relationship is present. However, no short run coefficents appear as I have 0 lags but I have seen other paper that produce short run coefficients with zero lags.

[Sebastian Kripfganz]

First of all, the btest option is outdated. Please use the estat ectest postestimation command instead.

ardl Y X1 X2 X3 X4, lags(1 0 0 0) ec
estat ectest

To obtain short-run coefficients with 0 lags, use the ec1 instead of the ec option. Please see the Remarks section in the command's help file and our forthcoming Stata Journal article for a discussion:

• Kripfganz, S., and D. C. Schneider (2023). ardl: Estimating autoregressive distributed lag and equilibrium correction models. Stata Journal, forthcoming.

The ardl command does generate a coefficient for the error correction term; it is called the speed-of-adjustment coefficient and reported in the ADJ section of the output.