Fixed Effects High Prob F and low Rsq

[Silvina Sanchez]

Hello everyone,
After performing xtoverid and interpreting its results, I decided to run a fe robust regression for my panel data, clustered by countries (36 years and 35 countries). Unfortunately, my model has a very low R sq and Prob > F = 0.2397.
What are the implications of this? Does this mean I have to find another model? I must mention I do not have a background in statistics or econometrics.

Below my results. The variables are logged.

Thank you very much!


xtreg log_Defence log_RILE log_GDP log_Debt log_Unemployment log_population, fe robust

Fixed-effects (within) regression Number of obs = 774
Group variable: Id_country Number of groups = 35

R-sq: Obs per group:
within = 0.0546 min = 7
between = 0.0353 avg = 22.1
overall = 0.0088 max = 37

F(5,34) = 1.43
corr(u_i, Xb) = -0.0092 Prob > F = 0.2397

(Std. Err. adjusted for 35 clusters in Id_country)
----------------------------------------------------------------------------------
| Robust
log_Defence | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-----------------+----------------------------------------------------------------
log_RILE | -.0647812 .1569548 -0.41 0.682 -.3837518 .2541894
log_GDP | .0354831 .3646179 0.10 0.923 -.7055097 .7764758
log_Debt | .1246861 .1892613 0.66 0.514 -.2599391 .5093114
log_Unemployment | -.2551505 .2147641 -1.19 0.243 -.6916037 .1813028
log_population | 8.941947 4.018865 2.22 0.033 .7746302 17.10926
_cons | -34.12163 17.495 -1.95 0.059 -69.67575 1.432499
-----------------+----------------------------------------------------------------
sigma_u | 3.1562238
sigma_e | .71949173
rho | .95060142 (fraction of variance due to u_i)

[Carlo Lazzaro]

Silvina:
welcome to this forum.
1) as you invoked the community-contributed module -xtoverid- (as FAQ recommend to mention) your standard errors were, in all likelihood, already clustered;
2) as per the negligible within R-sq, your data seems to have a minimal within panel variation. Is this is the case, -fe- will not take you that far;
3) your T dimension is large enough to recommend taking a look at -xtgls- and -xtregar-;
4) check whether your model is correcly specified.

[Silvina Sanchez]

Thank you, Carlo! Correct, I did cluster my standard errors before invoking xtoverid. I will have a look at -xtgls- and -xtregar and check whether my model is misspecified. I have read several other posts and the discussions in this forum have helped me a lot. Really appreciate it.

Best regards,
Silvina