This study examines whether today's technical change depends on yesterday's technical change. We propose to investigate this feedback effect by using the technical-change component of the Malmquist productivity index. This approach can overcome some problems in alternative patent-citation approaches. We apply the approach by estimating the feedback effect from production data of 25 OECD countries for 1980 through 1997. Our model yields evidence on a positive feedback effect with delays up till eight years. These findings are in line with patent-citation studies and bring us closer to a measure of the social returns to R&D.