Startups experiment with technologies and markets to reduce uncertainty. Although experimentation leads to learning, it is costly, raising questions about how to fund it. High-tech startups that operate outside the narrow range of technologies receiving venture capital investments can turn to the U.S. government to fund their experimentation. This study examines how experimentation affects startup performance. Using new data on $42 billion in government R&D contracts awarded to American high-tech startups between 1989 and 2019, I analyze the effect of experimentation on startup employment, sales, and exit through initial public offerings or mergers and acquisitions. I identify the effect through industry-specific exposure to (i) size standards used by federal agencies to determine which firms can compete for government programs reserved for small businesses and (ii) increased government spending before the end of the fiscal year. My results indicate that experimenting with markets, rather than technologies, drives startup performance. However, the effect is heterogeneous: experimenting with markets benefits low-growth startups more than high-growth startups. This finding has significant economic implications for the level and trajectory of experimentation in the U.S. economy.
