We ask whether failure to control for research and development (R&D)
activity in an output convergence regression affects the coefficient estimates of initial output. We focus on output convergence to an economy’s own steady-state growth path using time series regression framework and convergence across economies using panel estimation. We use data for the 30 member countries of the Organization for Economic Co-operation and Development (OECD) and US state-level real per capita output and per capita patents. The results indicate that after controlling for R&D activity the coefficient estimates increase in magnitude (in absolute terms) and in significance levels. Furthermore, the results are not sensitive to the dataset used or the estimation procedure.