GMM estimates and autocorrelation

[wjgatt]

I'm estimating a single equation model using GMM.

My question is about any potential bias that might result in the coefficients if the residuals display autorcorrelation. For the purpose of inference, if the covariance estimator used employs an HAC (Newey West) heteroscedasticity and autocorrelation covariance correction, can the t-stats and coefficients be used in the normal way?

I'm using a very small sample of some 40 observations, which I guess makes matters only worse. At this point however I'm assuming that the model is correctly specified.

Many thanks