Output gap and potential output are important concepts in the economic literature. However, most estimates of these quantities are subject to heavy revisions as new data becomes available. For example, the HP-filter suffers from very serious end-of-sample problems, and extending the GPD series forward only partially mitigates the issue. I'm wondering what the current state of the art is in generating more stable output gap and potential output estimates. In particular, I'd be interested in techniques that overweigh estimate stability, at the expense of precision.