Hello,

I am pretty new to Stata and econometrics in general so please excuse me if my question seems stupid or is redundant. I really can't manage to find a solution.

I'm working with panel datas (2011 to 2017), on the effect of oil prices on the economic growth with the following pooled OLS regression:
reg growth gvt_spendings fixed_capital_form oil_import_price elec_oil opep beneOpep

The problem is I identified a high correlation between government spendings and the fixed capital formation:

correlate DC_A FBCF
(obs=1,197)

| DC_A FBCF
---------------+------------------------
DC_A | 1.0000
FBCF | 0.8946 1.0000



That could explain the negative effect of government_spendings on growth (which has to be a positive coefficient following GDP=C+I+G):
. reg $ylist $xlist

Source | SS df MS Number of obs = 94
-------------+---------------------------------- F(7, 86) = 6.36
Model | 404.779791 7 57.8256844 Prob > F = 0.0000
Residual | 782.304497 86 9.09656392 R-squared = 0.3410
-------------+---------------------------------- Adj R-squared = 0.2873
Total | 1187.08429 93 12.7643472 Root MSE = 3.0161

------------------------------------------------------------------------------
croissance | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
gvt | -1.90e-11 4.70e-12 -4.04 0.000 -2.83e-11 -9.64e-12
FCF | 2.31e-11 5.20e-12 4.44 0.000 1.27e-11 3.34e-11
Kb_ent | -1.41e-12 3.54e-13 -3.98 0.000 -2.11e-12 -7.05e-13
pOILimport | -.0388874 .0143319 -2.71 0.008 -.0673783 -.0103965
ÉLECoil | -.476328 .1116277 -4.27 0.000 -.6982365 -.2544195
OPEP | -.0234799 .814232 -0.03 0.977 -1.64212 1.59516
béné_OPEP | -.0904356 .14431 -0.63 0.533 -.3773145 .1964432
_cons | 6.269373 1.486281 4.22 0.000 3.314744 9.224002
--------------------------------------------------------------------------




The only way I found to solve my correlation problem could be to drop government_spendings, but p-values skyrocket and my R^2 decreases a lot:

. reg $ylist $xlist

Source | SS df MS Number of obs = 94
-------------+---------------------------------- F(6, 87) = 3.40
Model | 225.735623 6 37.6226038 Prob > F = 0.0046
Residual | 961.348665 87 11.0499847 R-squared = 0.1902
-------------+---------------------------------- Adj R-squared = 0.1343
Total | 1187.08429 93 12.7643472 Root MSE = 3.3242

------------------------------------------------------------------------------
croissance | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
gvt | 4.61e-13 1.87e-12 0.25 0.806 -3.26e-12 4.18e-12
Kb_ent | -9.48e-14 2.13e-13 -0.44 0.658 -5.19e-13 3.29e-13
pOILimport | -.0367706 .0157872 -2.33 0.022 -.0681494 -.0053918
ÉLECoil | -.4131606 .122026 -3.39 0.001 -.6557004 -.1706208
OPEP | .7600824 .8760425 0.87 0.388 -.9811469 2.501312
béné_OPEP | -.0518004 .158762 -0.33 0.745 -.367357 .2637562
_cons | 6.083269 1.637458 3.72 0.000 2.828645 9.337894
-----------------------------------------------------------------------------



Is there a way to solve this problem of correlation? or making those p-values significant?
Keep FCF in the regression would be nice too.



Thanks a lot