Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# Re: st: Panel Unit Root Test

 From Yuval Arbel To statalist@hsphsun2.harvard.edu Subject Re: st: Panel Unit Root Test Date Thu, 6 Dec 2012 06:07:44 -0800

```There are several interesting applications to the unit-root hypothesis:

According to the quantity theory of money M=kPY where M is the
quantity of money, P is price level and Y is the GDP. The prominent
prediction of the model is neutrality of money: if M increases by x%
and Y is fixed P is expected to increase by x%.

Note that if we convert this into a logarithmic form, it turns out
that the model predicts a unit root as a coefficient of P and Y. In
fact, somebody checked the theory based on the unit-root hypothesis.

Another interesting application is based on a study I have just
completed. The government gave public-housing tenants the option to
purchase their renal units after a major discount, which vary from one
period to another. We had the trail of the discount for each and every
tenant, but we had to check this trail is random walk (otherwise
tenants will anticipate the momentum and wait until the last day to
exercise). Under these circumstances we ran a panel unit-root test,
and the unit root hypothesis was not rejected.

On Thu, Dec 6, 2012 at 5:53 AM, Yuval Arbel <yuval.arbel@gmail.com> wrote:
> Frances,
>
> By a nutshell, unit root is a very big problem in time-series
> analysis. If there is a unit root the series is a random walk and
> explosive. This implies that the estimates are inefficient, and by
> backward induction we can show that the SD estimates do not converge
> anywhere. A simple way to explain it is to divide a series to
> sub-samples. Imagine that the mean and SD of each sub-sample is
> different.
>
> A simple way to address unit-root problems is to use a difference
> series, i.e., Yt-Yt-1
>
>
>
>
>
>
> On Thu, Dec 6, 2012 at 5:29 AM, Francesoc Paldini <f.paldini@gmail.com> wrote:
>> Hello Statalist,
>>
>> I'm working with dynamic panel data (unfortunately with small
>> data sets) and I'm performing simple fixed effect estimations. Since T
>> is pretty large (T=30), I don't care about the Nickel Bias.
>>
>> Actually, I don't get the concept of unit root tests for panel data?
>> Under which circumstances are unit roots problematic?
>>
>> Does the asymptotic distribution theory require the estimator to
>> satisfy the usual (time series) conditions that rule out unit and
>> explosive roots? Do I need lots of cross-sectional units (in my case:
>> N=20)?
>>
>> Best wishes,
>> Frances
>> *
>> *   For searches and help try:
>> *   http://www.stata.com/help.cgi?search
>> *   http://www.stata.com/support/faqs/resources/statalist-faq/
>> *   http://www.ats.ucla.edu/stat/stata/
>
>
>
> --
> Dr. Yuval Arbel
> 4 Shaar Palmer Street,
> Haifa 33031, Israel
> e-mail1: yuval.arbel@carmel.ac.il
> e-mail2: yuval.arbel@gmail.com

--
Dr. Yuval Arbel