Instrumental variables estimation is classically employed to avoid simultaneousequations bias in a stable environment. Here we use it to improve upon ordinaryleast squares estimation of cointegrating regressions between nonstationaryand/or long memory stationary variables where the integration...