Dear All,

I am looking for some help as I am quite new to this. Any help or advice would be greatly appreciated. I thank you in advance for taking the time to read this post.



I am currently working with a panel set of global bilateral FDI data from UNCTAD. It covers 12 years, which I know is not a long time frame, but it is all that is available through UNCTAD.

I am attempting to run a standard gravity equation, with LogFDI as my dependent variable. I have a number of institutional variables that I am testing the effects of on FDI. It has come to my attention that nonstationarity and cointegration could pose a problem with my estimations so it was suggested that I under go unit root tests in the form of Phillips-Perron and cointegration tests on the nonstationary variables with Pedroni.

This is where the problem begins. All my variables work fine, with the exception of my dependant variable of LogFDI. Whenever I attempt to run the unit root test, I receive about a thousand errors such as, "could not compute test for panel 3, could not compute test for panel 5, could not compute test for panel 7, etc."

I am assuming this has to do with the fact that there are large number of missing values in bilateral FDI data, apparently this is just the nature of this type of data. However, when I go to compare a panel it says it cannot compute and on that it can, I see no difference between panels. Sometimes both panels are missing data and it works for one but not the other.

I am running the command:

xtunitroot fisher ln_FDIInflows_Host, pperron lags(1)

and I get the following results:

Fisher-type unit-root test for ln_FDIInflows_Host
Based on Phillips-Perron tests
-------------------------------------------------
Ho: All panels contain unit roots Number of panels = 5167
Ha: At least one panel is stationary Avg. number of periods = 6.23

AR parameter: Panel-specific Asymptotics: T -> Infinity
Panel means: Included
Time trend: Not included
Newey-West lags: 1 lag
------------------------------------------------------------------------------
Statistic p-value
------------------------------------------------------------------------------
Inverse chi-squared(7082) P 1.99e+04 0.0000
Inverse normal Z -38.8801 0.0000
Inverse logit t(14104) L* -73.1777 0.0000
Modified inv. chi-squared Pm 107.6729 0.0000
------------------------------------------------------------------------------
P statistic requires number of panels to be finite.
Other statistics are suitable for finite or infinite number of panels.
------------------------------------------------------------------------------




Which according to the literature I've been reading, is not really what I should expect. Which also leads me to my next problem of setting up the Pedroni test. The assumption is that FDI should be nonstationary and cointegrated with several of my time-varying variables. However, if I try to ignore the Phillips-Perron test for now and attempt to see the results of a Pedroni Test, when ever I try to use LogFDI I receive an error.

I am running the command:

xtcointtest pedroni ln_FDIInflows_Host variable1 variable2 variable3

I receive the following:

_hqrdp_la(): 3200 conformability error
_hqrdp(): - function returned error
_hqrd(): - function returned error
hqrd(): - function returned error
getresd(): - function returned error
<istmt>: - function returned error
r(3200);


Now it says its a conformability error, but I fail to see how as all my data is formatted exactly the same. I am obviously missing something.



Any help would be greatly appreciated.

Thank you for your time,

Wes Josefsson