Panel data unit root test fisher type lag length

[Karl Charles]

Hi, im currently working with a dataset of macroeconomic factors for a sample of countries with data ranging from 1995-2015. Im using panel data for the model and want to conduct a unit root test on each of the variables. Im using fisher type with ADF tests to conduct unit root tests for each panel individually. Im using stata but im confused to how many lags i should be using in my tests. I wanted to know is there a procedure to figure out the optimal amount of lags to use on the variables?

Also how do you decide which options to use in the test like time and drift trends etc?

Would be grateful if anyone could offer some advice. Thanks!

[Emad Shehata]

Specifying lags(aic #) requests that the number of lags of the series be chosen such that the Akaike information criterion (AIC) for the regression is minimized. xtunitroot llc will fit ADF regressions with 1 to # lags and choose the regression for which the AIC is minimized. This process is done for each panel so that different panels may use ADF regressions with different numbers of lags.

Specifying lags(bic #) is just like specifying lags(aic #), except that the Bayesian information criterion (BIC) is used instead of the AIC.

Specifying lags(hqic #) is just like specifying lags(aic #), except that the Hannan–Quinn information criterion is used instead of the AIC.

[Emad Shehata]

See next example

[Emad Shehata]

. use http://www.stata-press.com/data/r13/pennxrate
. xtunitroot llc lnrxrate if g7, lags(aic 10)
Levin-Lin-Chu unit-root test for lnrxrate
Ho: Panels contain unit roots Number of panels = 6
Ha: Panels are stationary Number of periods = 34
AR parameter: Common Asymptotics: N/T -> 0
Panel means: Included
Time trend: Not included
ADF regressions: 1.00 lags average (chosen by AIC)
LR variance: Bartlett kernel, 10.00 lags average (chosen by LLC)
Statistic p-value
Unadjusted t -6.7538
Adjusted t* -4.0277 0.0000

[Jesse Wursten]

Just a quick note - the Levin-Lin-Chu panel unit root test (xtunitroot) is not robust to cross-sectional depence, a very common feature in macro data. You can use the xtcd command to test if your data are cross sectionally dependent (available through ssc install). If they are, then you might want to look into chapter 21 of Baltagi (2008)* and the other tests described there.

* Econometric Analysis of Panel Data, 4th Edition, Badi H Baltagi, 2008

[Caterina Brest Lopez]

Hi! following on this post, I'm facing the same question and followed Emad Shehata syntax but for the Fisher test type and run into an error. More specifically, I coded:

xtunitroot fisher tcn_epr, dfuller lags(aic 10)

where tcn_epr is my variable of interest and is the q-o-q change in the Nominal Exchange rate (end of period rate) and I get the following error:

performing unit-root test on first panel using the syntax
dfuller tcn_epr, lags(aic 10)
returned error code 198
r(198);

Do you know why is the syntax wrong in this case? I have the impression that the problem arises when I write "aic" inside the lags option because it runs fine without it (that is, If I code:

xtunitroot fisher tcn_epr, dfuller lags(10)

).

Any help would be much appreciated!

[Jesse Wursten]

The aic bit is not supported for Fisher. See the subsection Fisher_options in -help xtunitroot-