Instrumental variables estimation is classically employed to avoid simultaneous equations bias in a stable environment. Here we use it to improve upon ordinary least squares estimation of cointegrating regressions between nonstationary and/or long memory stationary variables where the...