There is little, if any, dispute that R&D is a major source of long-term productivity growth. But there is empirical uncertainty about the magnitude of the productivity gains from R&D. This quantitative uncertainty was again highlighted in a study by two colleagues at the Productivity Commission (Shanks and Zheng 2006).224 They set out to update and extend previous time-series analysis of the effects of R&D on Australia’s productivity performance.225 Previous studies had generated estimates of returns to Australian R&D that seemed implausibly high – a result that is not uncommon in this type of analysis, irrespective of country of investigation (Diewert 2005). With the possibility that limitations on degrees of freedom had been an issue in the previous studies, it was judged that new analysis based on a further 10 years or so of data, plus developments in quantitative tests and techniques, could provide a clearer fi x on the effects of domestic and foreign R&D on Australian productivity performance. As it turned out, the modelling results were fragile – and more so than expected. Estimates of performance effects fell within wide confi dence intervals and were sensitive to seemingly reasonable modifi cations to variable and model specifi cations. Diagnostic tests revealed standard estimating equations to be mis-specifi ed.