Research

To increase the number of venture capital-backed startups when successful founders become investors, should government spend on creating non-founder investor slots (non-founder-slot policy) or scaling capital per deal (bigger-deal policy)? In a calibrated search and bargaining model, founders seek investor slots as workers seek jobs; each slot can back one startup and is filled or vacant. Some founders exit successfully by selling or listing startups, pass the accredited-investor wealth test, and open vacancies. Mainly through this recycling loop, each added vacancy multiplies into 2.11 vacancies, counting itself. The model predicts that, at equal spending, the non-founder-slot policy raises founder welfare (certainty equivalent of searching) more than the bigger-deal policy. First, it directly adds vacancies, which multiply and strengthen founders’ bargaining, unlike the bigger-deal policy’s higher average exit value. Second, most startups fail, so risk aversion caps a representative founder’s equity value at 11% of expected payoff. 5% more non-founder slots raise founder welfare by 2.90% (51% through recycling) and participation by 3.84%; equal bigger-deal spending raises them by 0.40% (46% through recycling) and 0.19%. Per dollar spent at these policy sizes, the non-founder-slot policy raises exit value by $1.48, the bigger-deal policy by $0.63. Restoring the wealth test’s real value at its 1982 adoption would cut founder-investors by two-thirds and founder welfare by a third.

The leading evidence on hot venture markets and innovation (Nanda and Rhodes-Kropf, 2013) follows patents for startups that went public and tracks their innovation performance for a three-year window after first funding. We extend the analysis by merging two datasets: SDC Platinum records startups’ venture financing histories, while Orbis tracks their patenting activities. Although the two datasets lack a common identifier, we take the 32,668 US venture capital-backed startups first financed from 1980 to 2019 in SDC Platinum and link 6,467 of them to their patents in Orbis. We also contribute by extending the tracking window for innovation performance of the startups for up to ten years after first funding, which allows us to have a more comprehensive picture of their startup life cycles. With heat measured as the detrended number of startups receiving a first round in the quarter, startups first financed in a quarter one standard deviation hotter receive 0.139 more log total citations over their first ten years (wild bootstrap p = 0.046), about 15% more than startups first financed in an average quarter.

How heavy-tailed founder wealth is depends, in closed form, on how accurately founders are matched with venture funds. Taken as one jump, and with each fortune valued at its current match, the calibrated rise in that accuracy lifts founders’ top wealth shares on the clock of match turnover (renewal of founders’ fund ties): halfway in 2.1 to 7.4 years for match half-lives of 3 to 10 years (illustrative, not measured). Random growth, paced by thirty-year working lives, has half-lives of 40 to 100 years for income moments [Gabaix et al., 2016], computed from asymptotic rates rather than as times to close half a gap: years against decades. Since the shares close half their gap within one match half-life, sorting at this size is the faster margin when matches turn over faster than working lives; the years do not date the rise. Read from venture deals under stated assumptions, without a standard error, sorting strength (the correlation of founders’ and funds’ normal rank scores, zero to one) lies between 0.136 and 0.444 around 1989. A calibration to tail exponents, venture-deal statistics and top-share growth sizes its rise without testing its speed: a one-time rise in talent-growth variance fits about as well if its size is left free, worse if its step is held to the dispersion of household returns on wealth.

[Abstract]