This paper sets up a dynamic general equilibrium model to study how the composition of technical progress affects the asymptotic speed of convergence. The following questions are addressed: Will endogenizing a fraction of the
productivity increases as coming from learning by investing help to generate a low asymptotic speed of convergence in accordance with the empirical evidence? Does it matter whether learning originates in gross or net investment? The answers to both questions turn out to be: yes, a lot. The third question addressed is: Does the speed of convergence significantly depend on the degree to which learning by investing takes the embodied form rather than the disembodied form? The answer turns out to be: no. These results point to a speed of convergence on the small side of 2% per year and possibly tending to a lower level in the future due to the rising importance of investment-specific learning in the wake of the computer revolution
as the empirical evidence suggests.