re: st: Zivot-Andrews test

[Christopher Baum <kit.baum@bc.edu>]

<>
     I'm studying two time-series (GDP and energy consumption) and I run the 
Zivot-Andrews test,
which allow for structural breaks in the series.
Does anyone help me in interpret the STATA output? What's the H0 hypothesis?
If I have a t-test < 5% Critical Value, what should I conclude?
Here there is an output for the GDP series:

. zandrews loggdpgkpc, break(trend) graph

Zivot-Andrews unit root test for  loggdpgkpc

Allowing for break in trend

Lag selection via TTest: lags of D.loggdpgkpc included = 0

Minimum t-statistic -1.764 at 1991  (obs 22)

Critical values: 1%: -4.93 5%: -4.42


The test is analogous to a Dickey-Fuller test (dfuller, or the improved version dfgls) in that you are looking to reject the null of a unit root in the process. In the Z-A test, you allow for a breakpoint in the series (which might mistakenly lead you to conclude that the series is nonstationary, whereas it could be stationary with a level or trend shift). To reject the null of I(1) you need a large negative t-stat, larger than the critical values. In the case above you cannot reject. I expect that if you rerun the test on D. loggdpgkpc it will reject.

Kit


Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                              An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
   An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/