Berry-Esseen bounds for econometric time series

We derive uniform and non--uniform error bounds in the normal approximation under a general dependence assumption. Our method is tailor made for dynamic time series models employed in the econometric literature but it is also applicable for many other dependent processes. Neither stationarity nor any smoothness conditions of the underlying distributions are required. If the introduced weak dependence coefficient decreases with a geometric rate then we obtain, up to a multiplicative logarithmic factor, the same convergence rate as in the centrallimit theorem for independent random variables.