Statistical inference in regression with heavy-tailed integrated variables

We consider the problem of statistical inference in a bivariate time series regression model when the innovations are heavy-tailed and the ordinary least squares (OLS) estimator is used for parameter estimation. We develop the asymptotic theory for the OLS estimator and the corresponding t-statistics. Limit distributions, that enable us to construct confidence intervals of the estimated parameters, are obtained via Monte Carlo simulations. The approach allows the components of the innovation vector to have different tail behavior.