Still searching for panel vecm command (PVEC?)

[Fabiana De Cristofaro]

Dear all,

I have found a lot of old posts asking for a Stata command estimating a VECM for panel data, essentially with no definitive answers.

Is there any development on this side? How are you estimating a panel vecm?

Thanks a lot to anyone, every answers would be very useful and welcome.

[George Ford]

here's one suggestion form: https://www.researchgate.net/post/Wh...ECM-stata-code

"In Stata, you can estimate a Panel Vector Error Correction Model (Panel VECM) using the pvec command from the xtabond2 package. Here is a simplified example of how to specify and run a Panel VECM in Stata:
stataCopy code// Load data use "your_data.dta", clear // Install xtabond2 package if not already installed // ssc install xtabond2 // Specify the Panel VECM model xtabond2 y x1 x2, panel(panel_varname) vce(robust) level(2) // Replace 'y' with your dependent variable name // Replace 'x1', 'x2' with your independent variables // Replace 'panel_varname' with the name of your panel identifier variable // Specify other options as needed, such as robust standard errors (vce(robust)) and lag length (level(2))
Make sure to replace y, x1, x2, and panel_varname with your actual variable names in the dataset. Adjust the lag length (level()) and other options according to your specific research needs and data characteristics."

[George Ford]

maybe xtpmg

[Manh Hoang Ba]

here's one suggestion form: https://www.researchgate.net/post/Wh...ECM-stata-code

"In Stata, you can estimate a Panel Vector Error Correction Model (Panel VECM) using the pvec command from the xtabond2 package. Here is a simplified example of how to specify and run a Panel VECM in Stata:
stataCopy code// Load data use "your_data.dta", clear // Install xtabond2 package if not already installed // ssc install xtabond2 // Specify the Panel VECM model xtabond2 y x1 x2, panel(panel_varname) vce(robust) level(2) // Replace 'y' with your dependent variable name // Replace 'x1', 'x2' with your independent variables // Replace 'panel_varname' with the name of your panel identifier variable // Specify other options as needed, such as robust standard errors (vce(robust)) and lag length (level(2))
Make sure to replace y, x1, x2, and panel_varname with your actual variable names in the dataset. Adjust the lag length (level()) and other options according to your specific research needs and data characteristics."

Dear George Ford,

I think using xtabond2 as suggested there is not appropriate. The level(2) option specifies to use a 2% confidence interval instead of 95% (the default), not to specify the lags for the Panel VEC.
Another suggestion in the discussion is to use xtdcce2 (or xtpmg) which is also not entirely appropriate.
These commands estimate a single equation model of the Panel ARDL-ECM type, which implies that there is only one long-run relationship between the endogenous variables in the system. This is only the case when Rank=1 in the VEC, if Rank=2 (there are two long-run relationships) then the results from xtpmg and xtdcce2 will face the problem of misspecification.

[George Ford]

There's a pvecm in R. Maybe it could be coded to Stata, if it's correct.