I use novel data on defense contracts to study the effects of government purchases in the US and develop
new stylized facts about their transmission mechanism. My methodology leverages the construction of a new
quarterly series of US military prime contracts, available from 1947:1. Defense contracts: (i) are exogenous
to output fluctuations; (ii) retain statistical power and robustness across various samples; (iii) accurately
measure the timing of the shocks; and (iv) obviate the need for narrative analysis. My findings indicate that
a positive shock to defense contracts, ordered first in a VAR, bolsters output, inventories, non-durable-plus-service
consumption, hours worked, employment, labor earnings, disposable income, the price-cost markup, the product-wage,
and labor productivity. I argue that the observed gains in labor productivity stem from ``learning-by-doing," a
feature particularly relevant to the production of military items. Further, leveraging a two-sector RBC model, I
demonstrate that the learning-by-doing induced productivity enhancements in the manufacturing sector suffice to
increase aggregate consumption, rationalizing the VAR evidence.
