Dear all,
To study the effect of energy price shifts on energy poverty I am estimating price elasticities of energy demand using a short panel with N > 10.000 and T = 14.
First I used xtabond2 to esitmate long-run elasticities which can consistently estimate dynamic panels where T is small. However, I now realised that my research question does not require the long-run elasticity but the short-run dynamics instead since the research cares about the effect of the level in a predictor at a previous time period on the dependent variable at the current time period.
To estimate short-run estimates I came upon the ARDL model and estimating using xtpmg package (ARDL-PMG estimator), however, as described by Blackburne, this estimator is only consistent for panels in which the number of groups and number of time-series observations are both large.
My question is two-fold.
What is considered to be large T for these estimation methods (how significant is the bias when T = 14)?
Are there any other consistent modelling techniques to estimate short-run dynamics in short panels?
I have read Hsiao et al. 2009 on Bayes estimation of short-run coefficients in dynamic panel data models, but could not find any collaboration on how to implement this estimation method.
All help is welcome!
Kind regards,
Hein Willems
To study the effect of energy price shifts on energy poverty I am estimating price elasticities of energy demand using a short panel with N > 10.000 and T = 14.
First I used xtabond2 to esitmate long-run elasticities which can consistently estimate dynamic panels where T is small. However, I now realised that my research question does not require the long-run elasticity but the short-run dynamics instead since the research cares about the effect of the level in a predictor at a previous time period on the dependent variable at the current time period.
To estimate short-run estimates I came upon the ARDL model and estimating using xtpmg package (ARDL-PMG estimator), however, as described by Blackburne, this estimator is only consistent for panels in which the number of groups and number of time-series observations are both large.
My question is two-fold.
What is considered to be large T for these estimation methods (how significant is the bias when T = 14)?
Are there any other consistent modelling techniques to estimate short-run dynamics in short panels?
I have read Hsiao et al. 2009 on Bayes estimation of short-run coefficients in dynamic panel data models, but could not find any collaboration on how to implement this estimation method.
All help is welcome!
Kind regards,
Hein Willems
