Test for a unit root using panel data

[Tom van Beek]

Dear Community,

For my master thesis I want to test wether the error term of my model is stationary or not, in order to verify if the regression is spurious.
I am using unbalanced panel data, so based on what I read there are two possibile tests: xtunitroot lps and xtunitroot fisher.

However, when I try the lps test, I get an error saying ''insufficient observations'' and when I run the fisher test it takes ages for my computer to compute the test statistic.
I believe that stata computes the test for every panel (11000 in my case).

Do you know a way to solve these issues?

[Tom van Beek]

I think I solved it myself. Insufficient observations is probably due to some panels that have too few observations. Furthermore missing observations and gaps in the time series also results in an error. Regarding the Fisher test, I guess you just have to be very patient.

[Tom van Beek]

Dear Community,

I have one final question regarding the xtunitroot test. Below is the output of the xtunitroot ips test on the residuals of my regression model. Since the H0 is clearly rejected, does this provide some evidence that the residuals are stationary and hence there is no spurious regression? Or is the rejection of the H0 simply a result of the many panels I have in my dataset, so that there are always a few residuals stationary?

Any reply would be much appreciated,
Best regards,
Tom

xtunitroot ips resid

Im-Pesaran-Shin unit-root test for resid
----------------------------------------
Ho: All panels contain unit roots Number of panels = 9602
Ha: Some panels are stationary Avg. number of periods = 50.80

AR parameter: Panel-specific Asymptotics: T,N -> Infinity
Panel means: Included sequentially
Time trend: Not included

ADF regressions: No lags included
------------------------------------------------------------------------------
Fixed-N exact critical values
Statistic p-value 1% 5% 10%
------------------------------------------------------------------------------
t-bar -7.0027 (Not available)
t-tilde-bar -4.8112
Z-t-tilde-bar -4.1e+02 0.0000
------------------------------------------------------------------------------

[Phil Bromiley]

The answer to your third post is in the output you posted: You'll increase the chances of a helpful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

Ho: All panels contain unit roots Number of panels = 9602
Ha: Some panels are stationary

A significant test indicates some panels are stationary. What you do with that is up to you. You should read up on unit root tests and see how folks in your area treat such tests.