Problem with xtpedroni and xtcointtest testing for cointegration with unbalanced panel data

[matt crocker]

I am struggling to figure out how to test for long run cointegration using xtpedroni or xtcointtest. I am using Stata 17 for windows.

I have an unbalanced panel of data N = 32 with an average T of 80 per panel. The unbalanced data is such that panels may have 20 consecutive observations, a period of missing data and then another period of data.

I am testing for long run cointegration in the context of running a non-linear ARDL model. So far I have confirmed that my variables are either I(0) or I(1) and have used the xtpmg command to estimate the short-run and long-run asymmetric relationships between my variables. I have been following the paper by Shin et al. (2014) who suggest testing for long-run cointegration between variables using a bounds testing procedure.

The paper about the xtpedroni command suggests that it can be used with unbalanced data, but when I run the code below I get an error

xtpedroni avg_cash_rate mortgage_rate, extraobs
matrix has missing values
r(504);

I have also tried the xtcointtest but I also get an error

xtcointtest pedroni avg_cash_rate mortgage_rate
Pedroni test does not allow gaps in data
r(498);

Is it possible to test for cointegration when I have unbalanced panel data?

Thanks