LEARNING TO QUIT? A MULTI-YEAR FIELD EXPERIMENT WITH INNOVATION DRIVEN ENTREPRENEURS*

Abstract

We use a randomized experiment with 553 science-and technology-based startups in 12 coworking spaces across the US to evaluate the effects of intensive, short-term entrepreneurial training programs similar to that offered by accelerator programs and executive education programs in US business schools on survival and performance for innovation-driven startups. Treated startups are more likely to shut down their businesses and do so sooner than controls. Conditional on survival, however, treated startups are more likely to raise external funding for their ventures, raise funding faster, and raise more funding than the control group; they also exhibit higher employment. Treated founders are less likely to found a new startup after shutdown. Our findings are consistent with practitioner arguments that early entrepreneurship training interventions can help entrepreneurs with less viable ventures "optimally quit" ("fail fast"). We use machine learning techniques (causal random forest) to provide initial insights on the most impacted subgroups.

Keywords: Entrepreneurship, Fast Failure, Optimal Quitting, Field Experiments, Education JEL Classification: C93, D22, M113, M53, O32