Wright meets Markowitz: How standard portfolio theory changes when assets are technologies following experience curves

This paper considers how to optimally allocate investments in a portfolio of competing technologies. We introduce a simple model representing the underlying trade-off - between investing enough effort in any one project to spur rapid progress, and diversifying effort over many projects simultaneously to hedge against failure. We use stochastic experience curves to model the idea that investing more in a technology reduces its unit costs, and we use a mean-variance objective function to understand the effects of risk aversion. In contrast to portfolio theory for standard financial assets, the feedback from the experience curves results in multiple local optima of the objective function, so different optimal portfolios may exist simultaneously. We study the two-technology case and characterize the optimal diversification as a function of relative progress rates, variability, initial cost and experience, risk aversion and total demand. There are critical regions of the parameter space in which the globally optimal portfolio changes sharply from one local minimum to another, even though the underlying parameters change only marginally, so a good understanding of the parameter space is essential. We use the efficient frontier framework to visualize technology portfolios and show that the feedback leads to nonlinear distortions of the feasible set.