I propose an econometric model that improves upon existing methods of estimating the natural rate of unemployment (NAIRU) by using information contained in the trend of productivity growth. My approach enhances the recently proposed model of Staiger, Stock and Watson (1997) in several respects. Statistically speaking, the method substantially shrinks the width of the 95% confidence interval, performs better in an out-of-sample inflation forecasting exercise, and is more robust to alternative statistical assumptions. In economic terms, the productivity-augmented model generates a more realistic time profile of the NAIRU, and implies estimates of the Phillips curve slope and the sacrifice ratio that are more in line with conventional wisdom. I also test whether the natural rate is correlated with the level or with the change of the productivity growth trend. I find support for the ?level? hypothesis in both the US and international data.