Re: st: Repeated time values within panel

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

<>
On Jan 26, 2014, at 2:33 AM, Giovanni wrote:

> I am having a problem setting xt set in stata.I  have a macro panel for 40 counties with 23 variables that include the usual annual macro variable (cpi ,debt , gdp etc), dummies  and some other indicators that in some cases change their value in the same year more than once .
> I do not know the exact date their value changes as those are indicators created by researchers.  Anyway what I did is that for each indicator that takes more than one value in the same year  I copied -pasted for each value the all the observation of the other variables. I provide some of data as an example just to make it more clear only for some of the variables  
> 
> country	 year	 countrycode	 GDPconstant     	 GDPpcurrent	               cpi 	     indicator1	     indicator2	         indicator3	           Dummy1                          
> Australia	1990	1	4.30E+11	3.10E+11	69.2101	-4.5	-4.5	1	0
> Australia	1990	1	4.30E+11	3.10E+11	69.2101	-14.9	-14.9	1	0
> Australia	1991	1	4.20E+11	3.30E+11	71.4406	-14.9	-14.9	1	0
> Australia	1991	1	4.20E+11	3.30E+11	71.4406	-14.9	-14.9	1	0
> Australia	1992	1	4.30E+11	3.30E+11	72.1449	-14.9	-14.9	1	0
> Australia	1993	1	4.40E+11	3.10E+11	73.453	-14.9	-14.9	1	0
> Australia	1993	1	4.40E+11	3.10E+11	73.453	-0.164745	-0.164745	1	0
> Australia	1994	1	4.60E+11	3.20E+11	74.8449	-0.164745	-0.164745	1	0
> Australia	1995	1	4.80E+11	3.70E+11	78.3163	-0.164745	-0.164745	1	0
> Australia	1996	1	5.00E+11	4.00E+11	80.3622	-0.164745	-0.164745	1	0
> Australia	1996	1	5.00E+11	4.00E+11	80.3622	22.5933	22.5933	3	1
> Australia	1997	1	5.20E+11	4.40E+11	80.5635	22.5933	22.5933	3	1
> Australia	1998	1	5.40E+11	4.00E+11	81.251	22.5933	22.5933	3	1
> Australia	1998	1	5.40E+11	4.00E+11	81.251	48.4581	48.4581	2	1
> Australia	1999	1	5.70E+11	3.90E+11	82.4417	48.4581	48.4581	2	1
> Australia	2000	1	5.90E+11	4.20E+11	86.1311	48.4581	48.4581	2	1
> Australia	2001	1	6.00E+11	3.80E+11	89.9044	48.4581	48.4581	2	1
> Australia	2001	1	6.00E+11	3.80E+11	89.9044	33.033	30.7286	2	1
> 
> 
> 
> I am able to xt set and run regressions only when I am not setting the time variable which I really need  in order to conduct afterwards unit root tests. When I try to include the time var I get an error
> 
> xtset countrycode year
> repeated time values within panel
> 
> Two important question
> 1).Have I set up the data correctly?
> 2) How can I xt set including time variable without deleting duplicates?

1) No.
2) With the data you have constructed, you can't.

As you say at the start, some of the variables you are interested in have changed more than once during the year. There is no way to express that with annual data. You either need higher-frequency data (which would probably not be all that useful if you cannot attach dates to the events) or you can just add interacted indicators that take on 1 if there were _any_ changes in the variable that year. They will not be able to capture multiple events. 

When working with IFS (International Financial Statistics) data, it is a very good idea to divide measures that take on values of 10^9, 10^12, etc. and scale them into a more sensible range (e.g., US gdp in trillions or billions, not dollars). That may not matter much in linear regressions, but if you ever use nonlinear estimation techniques, these huge magnitudes may bite you.

Kit

Kit Baum
Professor of Economics and Social Work, Boston College, Chestnut Hill MA, USA
DIW Research Professor, Department of Macroeconomics, DIW Berlin, Berlin, Germany
baum@bc.edu  |  http://ideas.repec.org/e/pba1.html


*
*   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/