Research

Commercializing Contrarian Ideas: Evidence from AI Contests

Job Market Paper

This work builds on the notion that some ideas are not only distinctive, but also contrarian: pursued by few, and met with open skepticism by most. Such skepticism hinders contrarians from attracting resources but allows them to experiment openly without fear of immediate imitation. To investigate this dynamic, I leverage hundreds of Artificial Intelligence contests where researchers either employ popular, state-of-the-art methods or pursue alternative, contrarian approaches. These contests act as sudden “moments of clarity” that reveal which methods perform best. They allow researchers to attract resources for commercialization, but also potentially alert competitors of the opportunity. Through a difference-in-differences design, I find that when contrarian contestants win, they are up to 7× more likely to found startups, and their startups attract up to 3× higher venture capital valuations. This cannot be simply explained by the fact that contrarians are more likely to achieve breakthroughs; much of the advantage arises by winning close races. Instead, while close victories validate contrarian entrepreneurs and allow them to attract resources, mainstream researchers tend to adopt contrarian methods only when they conclusively outperform traditional approaches. This reluctance to build on contrarian ideas allows contrarians to leverage public demonstrations to attract resources without provoking immediate competition.

Disagreement Predicts Startup Success: Evidence from Venture Competitions

Accepted, Strategy Science

This paper examines the relationship between disagreement surrounding a startup proposition and its future success. I find that the more venture competition judges disagree on the quality of a startup, the more likely the startup is to succeed, particularly when its proposition is unique. To explain this correlation, I build on the notion that (i) entrepreneurs pursue opportunities based on their subjective beliefs (ii) common opinion alone cannot be a source of competitive advantage. Therefore, value is disproportionately created and captured by founders with unconventional ideas that spark disagreement, and potential investors should harness disagreement as a predictor of success. I leverage data from 67 venture competitions to show that the empirical implications of this theoretical framework are supported by the data, whereas alternative explanations (e.g., that judges disagree more about risky ventures) are not. Additionally, I provide insights into what evaluators tend to disagree more often (e.g., former entrepreneurs) and which aspects of a startup (e.g., business model) are most polarizing. This work has broad implications for investors and institutions that strive to evaluate the potential of startup ideas.

Download Paper

Measuring Welfare Losses in Markets for Ideas: Evidence from Digital Startups Auctions

It is well known that information frictions can make ideas hard to sell. However, due to a lack of pricing data, virtually all studies have focused on how these frictions affect the likelihood of selling ideas rather than on the welfare losses from missed transactions. First, I present a theoretical framework to highlight the importance of this distinction. If the best ideas are the hardest to sell and the idea distribution is skewed, the proportion of welfare lost to frictions could far exceed the proportion of unsold ideas. Next, I demonstrate the practical significance of this distinction by examining online auctions for digital startups. I use the fact that informational frictions are less severe for startups whose revenues are verified by the marketplace. My findings show that revenue-verified startups are only 6.3% more likely to sell but fetch prices that are 2.7 times higher. Simulations reveal that moving from universal access to complete absence of revenue verification reduces gains from trade by 47%. The disproportionate losses are driven by the concentration of potential gains from trade among a few unsold ideas. These results suggest that focusing solely on the number of ideas sold underestimates the true impact of informational frictions.

Do NDAs Help to Sell Ideas? Evidence from 900 Confidentiality Agreements

Common wisdom suggests that non-disclosure agreements (NDAs) can help sellers safely disclose their ideas to potential buyers. However, buyers could potentially refuse to sign a binding agreement prior to disclosure, making them more likely to walk away from the transaction. Using data from an online marketplace for digital startups, I test the role of NDAs in facilitating idea transactions. I observe which sellers require potential buyers to sign a confidentiality agreement and the information disclosed by sellers after signing the NDA. I find that confidentiality agreements enable sellers to share more information: sellers who require NDAs from potential bidders are 24% more likely to share their profit and loss statements compared to those who do not. However, confidentiality agreements discourage bidder participation: requiring confidentiality agreements is associated with 29% fewer bids, making it 6% less likely that the venture will be sold. Overall, the results cast doubt on the effectiveness of NDAs in fostering idea markets.

Are Female Inventors Geographically Constrained? Gender Differences in Team Collocation

Coauthored with Mercedes Delgado

We present a new, stylized fact: female inventors are more likely to be collocated with their co-inventors compared to men. We interpret this correlation as suggestive evidence that women inventors face higher geographical frictions, potentially due to family obligations. This is a concerning possibility, given that 50% of all team patents in 2020 was produced by multi-located teams, and is particularly relevant for the work-from-anywhere debate. In support of our interpretation, we find that the collocation gender gap exists for both new and experienced inventors. The collocation gap increases with age: women are less collocated than men when they are statistically less likely to have children (under 32), but become more collocated past this threshold. Men, instead, become consistently less collocated as they age. Preliminary evidence suggests that women who worked on multi-location patents early in their careers stop patenting earlier. To conclude, the gap (and the gendered geographical frictions it might represent) is mitigated in top technology clusters where teams are typically more collocated.