2SLS Model with Autocorrelation and Heteroscedasticity

[Karinna Lai]

Hi everyone,

I have a panel set data and want to test 2 models, both of which are correlated one with another. My data set has 982 observations (71 panel, 14 years) These are the variables in my models :

Model 1 : lnzscore = CAR + FOR + SO + MAS + AS + GGDP + INFL + LVG + LNSIZE

Model 2 : CAR = lnzscore + FOR + SO +MAS + AS + RWATA + ROA + AG

due to the endogeinity in the variables (CAR and lnzscore), i have to run 2sls or GMM test. In this case, i choose 2sls. This is the command i put in :

ivregress 2sls lnzscore Foreign StateOwned MAS AS MA GGDP INFL Leverage lnsize (CAR = RWATA ROA AG)

However, my models are proven to have both heteroscedasticity (tested with estat hettest and xtgls)

i put in command

xtgls lnzscore CAR Foreign StateOwned MAS AS Leverage GGDP INFL lnsize, igls

and get the results

Iteration 1: tolerance = 0


Cross-sectional time-series FGLS regression

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


Estimated covariances = 1 Number of obs = 987
Estimated autocorrelations = 0 Number of groups = 71
Estimated coefficients = 10 Obs per group: min = 12
avg = 13.90141
max = 14
Wald chi2(9) = 65.08

Log likelihood = -1545.672 Prob > chi2 = 0.0000

my model also has autocorrelation, and when i use abar lag test, it shows that i have 8 lags variables (i test it after regressing the model with ivregress 2sls, using abar, lags(10) and p values are not significant until lag 9).

I was wondering how to fix these problems? since i can't found a solution for 2sls model and can't find a step-by step guide (i'm not familiar with stata, so using it was quite difficult for me)

Thanks a lot