LAUSR

Abstract Time-varying depreciation rates are estimated for research and development of the United States aggregate economy and innovation-intensive industries. Mean annual R&D depreciation rates are 31.5% for software, 41% for pharmaceuticals, 42% for semiconductors, and 30.4% for aggregate economy during 1978-2014. R&D depreciation rates vary across industries. R&D investment demand has separate elasticities in time-varying depreciation, interest rate and price growth, allowing for different rates of technology shifts across industries. Software R&D investment has the same magnitude of elasticities to depreciation, interest rate and price, supporting a user cost to apply. For pharmaceuticals and semiconductors, depreciation leads to more R&D investment that implies the effect of scale. For aggregate economy, depreciation reduces R&D investment more than interest rate, indicating obsolescence. Forecasting of R&D investment is improved at both industry and aggregate level. Forecastable time-varying depreciation, interest rate and price growth predict R&D investment based on the estimated demand function. The in-sample forecast comparison for 2015-2019 confirms the superiority to the alternative methods. Out-of-sample forecasts of R&D investment are carried out through 2025, and R&D capital stocks are constructed across industries and aggregate U.S. economy.