My research examines how institutional environments — universities, industrial policy, export controls, AI, and platform design — shape who becomes a founder and what they build. The work is organized below into four themes, each led by a foundational concept I have introduced or helped develop. Recent work centers on cross-border policy interventions, the platform financing of online misinformation, and the methodological infrastructure (alumni entrepreneur surveys, RCTs, AI-enabled methods) that makes large-scale empirical entrepreneurship research possible.
Foundational concept: institutional barriers to entrepreneurship quality — institutions shape not only how many firms get founded but the durability, innovativeness, and growth trajectories of those firms. Subsidies and market catalysts that boost entry volume can quietly hollow out quality.
U.S. export controls were meant to slow China's semiconductor sector. We find they increased VC funding to Chinese semiconductor startups by 6 percentage points and 12% in capital — driven by institutional reclassification, with state-backed investors acting as opportunity coordinators.
The Inflation Reduction Act selectively redirected venture capital toward climate-aligned technologies. Our Policy Credibility Index — scoring statutory specificity, durability, and enforceability — shows that institutional design quality, not just incentive size, determines whether private capital follows industrial policy.
Local government supply-side subsidies across 300 Chinese prefectures do attract more battery-sector entrants — but the marginal entrants are non-corporate, short-lived, generate no patents, and exit at elevated rates even in cities where the underlying battery cluster is thriving. Subsidies without capability screening attract participation without productive commitment.
Trust in government predicts which Stanford alumni become entrepreneurs and which sectors they enter — with implications for South Korea's policy reforms and other contexts where institutional trust varies sharply across demographic groups.
Market catalysts — incubators, science parks, junior stock exchanges — are designed to stimulate venture creation. Analysis of 19,000 firms in Japan shows they produce unintended negative consequences: more competition and information asymmetry, ultimately fewer quality IPOs and less economic growth. The clearest single illustration of the institutional-quality concept above.
Read the paper ↗What China's regional economic transition tells us about which entrepreneurial strategies pay off when institutions shift.
Read the paper ↗How founders in emerging economies use incubators, accelerators, and business associations to acquire public resources they couldn't otherwise reach.
Read the paper ↗Foundational concept: platforms as quasi-regulators — algorithmic monetization and matching decisions act as de facto regulatory infrastructure for the information environment, for the digital divide, and for whose voice shifts firm behavior. Firms inadvertently fund systems they would never knowingly underwrite; founders' access to mentor networks and learning platforms compounds prior advantage.
Programmatic ad-tech systematically directs major-brand advertising dollars to misinformation publishers, even when those brands explicitly want to avoid them. The paper traces the algorithmic pipeline and the consumer-backlash evidence that follows when brand placements become visible.
Evidence from a randomized field experiment on the NovoEd MOOC platform — what kinds of mentor connections, and what kinds of adaptability, actually predict whether early-stage ventures succeed. Among the earliest large-scale platform-RCTs in entrepreneurship research.
Read the paper ↗Activist tactics and how firms get targeted for social change — when secondary stakeholder pressure tactics succeed in shifting firm behavior, and when they don't.
Read the paper ↗A line of work examining how online learning platforms, e-commerce platforms, and gig platforms differ in their downstream effects on rural-to-urban migration patterns, return migration to home provinces, and the entrepreneurial pathways available to founders outside Tier-1 cities — the second-order digital divide in practice.
Foundational contribution: large-scale alumni entrepreneur surveys as a research instrument. Pioneered first with Ed Roberts at MIT (2009), then extended at Stanford (2012) with William F. Miller, the University of Virginia (2014), and Tsinghua University. The surveys created the empirical infrastructure for an entire downstream literature on university entrepreneurship, founder ethnicity, founding-team composition, mentor networks, and the institutional environment around technology entrepreneurship.
Survey of more than 140,000 Stanford alumni. Showed that companies founded by Stanford alumni generate trillions of dollars in revenue and millions of jobs globally — the empirical backbone of the "Stanford economic impact" finding now widely cited in policy and university-impact discourse.
The MIT alumni-impact study — and the methodological forerunner of the Stanford Innovation Survey. Found that companies founded by MIT alumni would, if grouped as an independent nation, generate roughly $2 trillion in annual worldwide sales — the 11th largest economy in the world at the time.
Read the report on Kauffman.org ↗The third in the alumni-impact trilogy (after MIT 2009 and Stanford 2012), applying the same methodology to UVA. Survey of more than 135,000 UVA alumni (22,757 valid responses) found 65,000 alumni-founded companies, 2.3 million employees worldwide, and an estimated $1.6 trillion in annual global revenues.
Read the full report (PDF) ↗Extended the alumni-survey methodology to one of China's leading research universities. Provides the empirical base for comparative work on how university entrepreneurship education differs across institutional contexts.
Survey of 27,783 Stanford alumni asking what CES at the GSB and STVP at the Engineering School actually did to entrepreneurship rates. CES participation had a negative-to-zero effect on rates and STVP had no effect — but CES participants saw lower startup failure and higher long-term firm revenue.
Stanford alumni survey work showing that Asian American alumni are more entrepreneurial than white American alumni, but non-American Asian alumni are substantially less so — with parental entrepreneurship a strong predictor regardless of group, and intergenerational persistence highest among non-American Asians.
Read the paper ↗Survey of 2,067 MIT alumni-founded ventures showing that all-technical founding teams have a 12.8% greater likelihood of favorable exit in cooperative business environments — challenging the conventional wisdom that diverse business skills always matter most.
Read the paper ↗How founding conditions shape whether ventures survive or thrive through environmental change — based on the longitudinal data of the alumni cohort.
Founder interviews recorded as part of the Stanford alumni cohort and the technology entrepreneurship course built around it.
Recorded with former Stanford Provost William Miller in 2012. On founding Sun, how to decide when to pursue a startup opportunity, the discipline of risk, and the story behind writing the first check to Google.
Recorded in 2012. Cheriton on serial founding, the practice of writing the original angel check to Google with Andy Bechtolsheim, and the academic-founder pathway.
Two distinct methodological contributions: large-scale alumni entrepreneur surveys (Theme 3 above), and early adoption of randomized controlled trials, difference-in-differences, and regression discontinuity designs in entrepreneurship research — including the first wave of platform-RCTs in MOOC and accelerator settings, and more recently AI-enabled methods for analyzing regulatory and policy corpora.
Randomized field experiment in a Thai online accelerator (AIS and KBank, 2021–2023, n=166 startups across multiple cohorts). Tests whether structured peer evaluations benefit the advice-giver, not just the receiver. Treatment teams gained +0.217 on independent pitch-quality scores (~⅓ of a SD, p=0.004) and +0.108 on idea usefulness (p=0.042); no significant novelty effect. Mechanisms: rubric-applying memory reinforcement, and metacognitive self-reflection.
Field-experimental study of entrepreneurship training in refugee camps and settlements across Uganda. Foundation-supported program operations and rigorous evaluation, designed in partnership with the practitioners running the program.
Among the earliest large-scale randomized experiments embedded in an online entrepreneurship platform — testing what kinds of mentor connections and what kinds of adaptability actually predict whether early-stage ventures succeed.
Read the paper ↗The export-controls, IRA cleantech, and China battery-subsidies papers (Theme 1) each rely on quasi-experimental designs to identify the causal effect of large-scale policy shocks on venture capital allocation and firm entry composition.
How generative AI is reshaping social-science research design — from automated coding and synthetic experimentation to LLM-based content analysis of regulatory and policy corpora. Reflects the methodological direction of the current research program.
For the full publication list, see Google Scholar. For talks, interviews, and recorded conversations, see Talks & Media. For PhD students and lab members, see Advising.