st: Panel Data, Autocorrelation, IV

[Jonathan Stoeterau <jonathanstoeterau@gmail.com>]

Dear helpful Stata-Listers,

I want to estimate a relatively short and highly unbalanced panel (see
xtdes below). I will use first log differences of the dependent and
independent variables. Moreover I expect first order serial
correlation (in the error term of each country) and
heteroskedasticity. Also I would need to include an Instrumental
Variable for one of my independent variables. Random effects seems to
work according to the Hausman-Test for my basic model.

Which would be the correct option to use? Since the newey-west
procedure and xgls cannot include IV? Moreover xtgls with options “,
corr(psar1) panels(correlated) force” cannot be used since the panel
is not balanced.

As a beginner, I appreciate every comment and hope I didn't make basic
mistakes,

Jonathan


    num:  1, 2, ..., 84                                     n =         84
    year:  1993, 1994, ..., 2007                   T =         15
           Delta(year) = 1; (2007-1993)+1 = 15
           (num*year uniquely identifies each observation)

Distribution of T_i:   min      5%     25%       50%       75%     95%     max
                         2       2       2         4         6       8      13

     Freq.  Percent    Cum. |  Pattern
 ---------------------------+-----------------
        2      2.38    2.38 |  .....1......1..
        2      2.38    4.76 |  ...1.....1.....
        2      2.38    7.14 |  ...1..1..1..1..
        2      2.38    9.52 |  1.1....1.......
        1      1.19   10.71 |  ........1..1...
        1      1.19   11.90 |  ......1...11...
        1      1.19   13.10 |  .....1.......1.
        1      1.19   14.29 |  .....1....1....
        1      1.19   15.48 |  .....1...1.1.1.
       71     84.52  100.00 | (other patterns)
 ---------------------------+-----------------
       84    100.00         |  XXXXXXXXXXXXXXX

*
*   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/