Panel Data

[Iryna Kosman]

Hello Everyone!

I'm new to Stata, but I really need help.

In my research, I investigate how pollution affects economic growth in Poland (by region). My dependent variable is GDP and the independent variables are firms' income, total respiratory diseases, number of active population, number of people employed, life expectancy and pollution levels (SO2, NO2 and PM10).

Here my regression:
regress GDP income_firms numb_peop_work life_expect total_respiratdis SO2 NO2 PM10

Source | SS df MS Number of obs = 272
-------------+---------------------------------- F(7, 264) = 6186.95
Model | 3.9920e+13 7 5.7028e+12 Prob > F = 0.0000
Residual | 2.4334e+11 264 921745059 R-squared = 0.9939
-------------+---------------------------------- Adj R-squared = 0.9938
Total | 4.0163e+13 271 1.4820e+11 Root MSE = 30360

-----------------------------------------------------------------------------------
GDP | Coefficient Std. err. t P>|t| [95% conf. interval]
------------------+----------------------------------------------------------------
income_firms | .3473091 .0163807 21.20 0.000 .3150557 .3795626
numb_peop_work | -.0175441 .0038914 -4.51 0.000 -.0252062 -.009882
life_expect | 3417.919 1670.478 2.05 0.042 128.7638 6707.074
total_respiratdis | 49.28193 3.633234 13.56 0.000 42.12813 56.43574
SO2 | 2028.387 892.1707 2.27 0.024 271.7118 3785.063
NO2 | -2196.868 645.3273 -3.40 0.001 -3467.511 -926.2244
PM10 | 876.7038 434.2378 2.02 0.045 21.69363 1731.714
_cons | -263290.2 128898.8 -2.04 0.042 -517090.6 -9489.738
-----------------------------------------------------------------------------------

But my model have multicolinearity


Variable | VIF 1/VIF
-------------+----------------------
total_resp~s | 99.64 0.010036
numb_peop_~k | 55.22 0.018111
income_firms | 25.33 0.039487
PM10 | 2.93 0.341165
NO2 | 2.68 0.373426
SO2 | 2.36 0.423085
life_expect | 1.21 0.823306
-------------+----------------------
Mean VIF | 27.05


What should I better to do with this multicolinearity?

Should I use 2SLS model? Please give me some advice.

Thank you.

[Carlo Lazzaro]

Iryna:
welcome to this forum.
1) you seem to have cross-sectiona data (that is, one wave of data only);
2) you have a sky-rocketing Rsq, which is a sign of overfitting (basically your model needs a better specification to be credible);
3) to have a better understanding of the correlation between your coefficients, type -estat vce,corr- after -regress-;
4) the postestimation routine (-estat hettest-; -linktest-) would be also informative;
5) eventually, have you ruled out reverse causation? If pollution can have an effect on economic growth, the opposite also holds;
6) as an aside, please use CODE delimiters to share what you typed and what Stata gave you back (see the FAQ on this and other topics related to improve posting on this forum). Thanks.