It is well established that for some products, increases in cumulative production (i.e. experience) are associated with decreasing unit costs. But does increasing experience imply decreasing costs, or do decreasing costs imply increasing demand? This problem of reverse causality limits the use of experience curves for ex-ante evaluation of policy interventions based on demand stimulus. During World War II, the demand for military products was largely exogenous, providing a useful natural experiment. In addition, since production first rose and then declined, WWII represents a rare setting in which production, cumulative production and an exogenous time trend are uncorrelated enough to lead to precise estimates of their relative effects. We assembled an unprecedented amount of data on U.S. military production during WWII, including several original datasets. Our results indicate that cost decrease can be attributed roughly equally to the growth of experience and to an exogenous time trend. This supports the use of policy-driven demand stimulus to lower the costs of critical technologies. (Joint work with Diana Greenwald and J. Doyne Farmer)