-estat hettest- and -xttest0-

[Gaby Gabriella]

Hello, I have a question regarding my result in Stata,
I'm currently studying balanced panel data for 5 countries in time periods of 10 years. We are studying the same sample at different points in time.

Here is my -estat hettest- result:

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of LFDIInflow

chi2(1) = 20.42
Prob > chi2 = 0.0000

As far as I know, after this step, we should run a random/fixed effect rather than a P-OLS model.
And after the Hausman test is deployed, we found that random effect is more statistically preferable.

However, I also know that after we run random effect, we need to use -xttest0- to pick between P-OLS or Random effect. Below are the result:

Breusch and Pagan Lagrangian multiplier test for random effects

LFDIInflow[CountryID,t] = Xb + u[CountryID] + e[CountryID,t]

Estimated results:
| Var sd = sqrt(Var)
---------+-----------------------------
LFDIInf~w | .848771 .9212877
e | .5109544 .7148108
u | 0 0

Test: Var(u) = 0
chibar2(01) = 0.00
Prob > chibar2 = 1.0000

Here are my questions:
1. Based on the BP-LM test result that rejects the null hypotheses, Which model is more appropriate? P-OLS or Random?
2. Does the BP-CW (Cook Weisberg) test result determine or indicate a preferable model between P-OLS, FIxed Effect, and Random Effect? Or does it only measure heteroskedasticity?

Thank you in advance

[Carlo Lazzaro]

Gaby:
welcome to this forum.
Some mixing up of different things here, probably due to the fact that both test were named after Breusch and Pagan.
That said:
- estat hettest- checks for heteroskedasticity after -regress- (not -xtreg-):
- -xttest0- check for Group-wise effect after -xtreg,re- (no evidence of that in your example; hence POLS would seem the way to go, but you do not report on the result of -hausman- test, though).
In addition, since you have a T>N panel dataset (T=years; N=countries), you should probably consider something like -xtgls- or -xtregar-.

[Gaby Gabriella]

This is the hausman result:

---- Coefficients ----
| (b) (B) (b-B) sqrt(diag(V_b-V_B))
| fixed random Difference S.E.
-------------+----------------------------------------------------------------
LGDPCap | 3.588705 2.235111 1.353595 .8167364
LCorruption | -.1612722 3.238816 -3.400089 1.091511
LPolStab | .9941053 2.367166 -1.373061 .752492
LPeace | -2.161306 2.367629 -4.528935 2.756454
LDemocracy | 1.693799 -.2942327 1.988032 1.207463
LProperty | 2.31103 4.519923 -2.208893 1.955668
LHumanDev | -10.37336 -36.20384 25.83049 14.89915
LUnemploym~t | -.9729728 -1.087527 .1145543 .6548216
LStartBusi~s | -2.84132 2.100011 -4.94133 1.995951
------------------------------------------------------------------------------
b = consistent under Ho and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under Ho; obtained from xtreg

Test: Ho: difference in coefficients not systematic

chi2(9) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= 8.77
Prob>chi2 = 0.4585
(V_b-V_B is not positive definite)


and here is the xtgls
. xtgls LFDIInflow LGDPCap LCorruption LPolStab LPeace LDemocracy LProperty LHuman
> Dev LUnemployment LStartBusiness

Cross-sectional time-series FGLS regression

Coefficients: generalized least squares
Panels: homoskedastic
Correlation: no autocorrelation

Estimated covariances = 1 Number of obs = 50
Estimated autocorrelations = 0 Number of groups = 5
Estimated coefficients = 10 Time periods = 10
Wald chi2(9) = 40.95
Log likelihood = -51.38493 Prob > chi2 = 0.0000

--------------------------------------------------------------------------------
LFDIInflow | Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------+----------------------------------------------------------------
LGDPCap | 2.235111 1.01949 2.19 0.028 .2369461 4.233275
LCorruption | 3.238816 1.415455 2.29 0.022 .4645749 6.013058
LPolStab | 2.367166 .8931762 2.65 0.008 .6165729 4.117759
LPeace | 2.367629 1.791781 1.32 0.186 -1.144197 5.879456
LDemocracy | -.2942327 .9814343 -0.30 0.764 -2.217809 1.629343
LProperty | 4.519923 2.134752 2.12 0.034 .3358857 8.703961
LHumanDev | -36.20384 10.61973 -3.41 0.001 -57.01814 -15.38954
LUnemployment | -1.087527 .3907388 -2.78 0.005 -1.853361 -.3216932
LStartBusiness | 2.100011 3.439411 0.61 0.541 -4.641112 8.841133
_cons | 104.0629 36.65766 2.84 0.005 32.21523 175.9106
--------------------------------------------------------------------------------

[Carlo Lazzaro]

Gaby:
what is your query then?