Selected Working Papers

Field Experiments

Immigrant Entrepreneurship and Legislator Attention: A Field Experiment

State legislators are less likely to respond and offer help to immigrant entrepreneurs.

Do startup founders respond differently to men and women investors? In two large-scale field experiments targeting over 45,000 US tech startups, we find robust evidence that they do. Startups were 25% more likely to engage with unsolicited outreach from men investors than from equally qualified women. Credentials, meant to signal expertise, further widened this gap. While professional qualifications increased engagement with men investors, they had the opposite effect for women, triggering eschewal. These findings support role congruity theory: when women exhibit markers of authority in men-typed domains like venture capital, they may be perceived as violating gender norms, undermining their legitimacy. We also explore homophily as an alternative explanation—that founders prefer investors of the same gender but find little support for this mechanism. Women-led startups do not favor women investors, and men-led teams show persistent preferences for men, even in the absence of substantive information. These patterns suggest that gender bias operates not only at the investment decision stage but also at the initial gatekeeping point of investor engagement. Our study reveals a hidden asymmetry in entrepreneurial finance: women investors are filtered out early before serious conversations begin. This bias, rooted in perceptions of competence, leadership, and fit, may help explain the persistent gender imbalance in venture capital, not only in who gets funded but in who gets to fund. By shifting the lens to founder behavior, our findings open new directions for understanding and mitigating barriers to gender equity in entrepreneurial ecosystems.

Comparison across Four Randomized Arms (2×2 Between-Subjects Design)

Notes. This figure shows the average reply rates with 95% confidence intervals across the different randomized arms. Panel A shows the difference in reply rates between legislators who received requests from 3rd Generation American entrepreneurs (i.e., nonimmigrant entrepreneurs) and those who received requests from 1st Generation American entrepreneurs (i.e., immigrant entrepreneurs). Panel B presents the reply rates across all four arms.

Cumulative Reply Rates over Time by Randomized Arms

Notes. This figure shows the reply rate measured cumulatively over two weeks for each randomized arm.

Global Capital, Local Minds: Founder Heuristics in International Investor Selection

Exploring the Liability of Foreignness in VC-startup selection.

How does an investor’s foreignness influence founder engagement at the earliest stage of the funding process? We explored this through a field experiment with 29,132 startups in the UK and India, sending unsolicited emails from either domestic or foreign investors. The results were striking. UK founders clearly preferred domestic investors, demonstrating a textbook case of liability of foreignness. Indian founders, however, showed no such bias and sometimes actually favored foreign investors over domestic ones. These contrasting patterns reveal that foreignness isn't universally good or bad for investors. Rather, its impact depends on institutional context. Our findings extend foreignness theory by identifying how founders screen potential investors at the very first point of contact. This cognitive filter shapes cross-border capital flows before formal interactions even begin.

Email View by Recipients

Notes. This figure illustrates the name and subject line of typical emails in the experiment, shown in the recipient’s email inbox on the left and the email content once opened on the right. Each recipient receives only one variation. The opened email is marked blue on the left. WiX’s technology tracks if the recipient viewed the email (i.e., opened it). Before opening, the recipient can see the investment firm, sender’s name, and location (e.g., Helen Anderson at Indigo Oak VC, UK). We randomly selected emails from the cleaned CrunchBase database for each experimental arm. Upon opening the email, the recipient can choose to click the hyperlink ‘Check out our website,’ which opens the website’s main page. WiX records this action as a unique click. Multiple clicks or independent visits from the same IP address do not count as additional clicks.

Email Open and Visit Rates

Notes. This figure displays the email opening rates and website visit rates in descending order for the eight primary experiment arms, categorized by the email recipient’s location (UK or India) and the email sender’s location (man or woman VC investor located in the UK or India).

Strategic Equivalence: How do Entrepreneurs Evaluate Green, Neutral, and Brown Funding?

Founders favor green investors only when credible, otherwise treat them like neutral or brown.

How do entrepreneurs navigate the tension between financial constraints and environmental positioning in their financing decisions? While stakeholder theory suggests that integrating environmental objectives can enhance long-term firm value, persistent brown financing casts doubt on this expectation. Through a large-scale field experiment involving nearly 40,000 technology startups, we examine founders’ considerations of potential financiers’ environmental orientation. We find that founders prefer green financiers over neutral or brown, but only when supported by corroborating signals. Consistent with our strategic equivalence hypothesis, founders tend to treat neutral and fossil-fuel investors similarly. Moreover, startups with greater dependencies on stakeholders demonstrate a pronounced preference for green financing than those operating in less socially sensitive sectors.

Email Labels with Content Preview

Notes. The screenshot demonstrates which part of the email’s content is used as a preview in the label and to what extent. We include all five different types of potential previews. WiX’s technology enables us to determine whether the recipient has opened the email, and hyperlinks embedded in the email’s body can track whether the recipient clicks on the image or the ‘Set up an appointment’ button, resulting in a website visit. We also note that the recipient can see the investment firm name before viewing the body of the email (e.g., CleanGreen or Shale).

Response by Sender Gender Over Startup’s Founding Year

Notes. This table reports the full sample of observations from the second stage of our randomized field experiment (RFE). Panel A shows website visit rates for the full sample, while Panel B breaks them down by investor type: green, neutral, and brown. Each panel reports the total opened emails, the number of recipients who visited the investor’s website, the visit percentage among those who opened the email, and the corresponding odds ratio (%Visited / %Unvisited). Additionally, Panel B includes the green-to-neutral and brown-to-neutral ratios for both percentages and odds ratios. The ratio is adjusted by subtracting one, so a positive value indicates an over-representation of responses from green or brown RFE investors relative to neutral investors. Pearson’s χ2 test is shown in brackets.

Avoiding Female Investors: Experimental Evidence on Gender Bias in Startup Culture

Founders are 25% more likely to answer emails from male investors, and professional credentials help men but hurt women. Serial founder experience and exposure to female leaders (as commanders in the IDF) moderate some of the bias.

Do founders favor male over female investors at first contact? We test this using two large randomized field experiments that sent identical unsolicited emails to over 45,000 U.S. startups and tracked opens and website visits. Founders engaged more with male investors, and professional credentials increased that gap by boosting men and penalizing women. The female penalty appeared across industries, geographies, and stages, rather than being concentrated in all male teams, indicating bias beyond simple homophily. We propose regulatory focus and reputational risk mechanisms and show moderators consistent with each: serial founders exhibit a smaller male advantage, and, in a complementary Israeli survey, male founders who previously served under female commanders are more likely to later receive funding from female investors. Together, the evidence reveals a bottleneck at the first step of investor outreach that restricts women’s access to deals and points to practical interventions that reduce reliance on gendered heuristics.

Experimental Design

Notes. This figure illustrates the experimental design. Because the random assignment was memoryless (also known as pure randomization), we have an unequal number of startups per randomized arm. RCTs 1 and 2 do not have overlapping recruited startups. RCT 1 exhausted all startups with valid email addresses that were incorporated between 2006 and 2019, based on data updated by early 2019. Then, RCT 2 exhausted all startups with valid email addresses that were incorporated between 2017 and 2019 and were originally unavailable in the first data procurement.

Response by Sender Gender Over Startup’s Founding Year

Notes. This figure reports the startup’s email open and website visit percentage for RCT 1 and RCT 2 by pooling experiment arms 1 and 3 (men investors with and without certificates) and arms 2 and 4 (women investors with and without certificates) by the startup’s founding year.

Four-Arm Comparison- Man and Woman With/Without Certificates

Notes. This figure displays the email open rates (black) and website visit percentages (Gray) for startups in RCT 1 and RCT 2, including 95% confidence intervals for each treatment group. It compares the average response rates for experiment arms 1, 2, 3, and 4, highlighting the difference in response rates based on whether startups received emails from men or women investors, both with and without a professional certificate.

Israeli Survey — Exposure to Female Commanders & Gender Pairing

Notes. This table presents multivariate logit estimates where the dependent variable equals one if the startup received funding from a female investor (zero if the investor was male). The key independent variable equals one if the founder had a female commander during military service. Columns (1)–(4) sequentially add controls for startup stage (pre-lettered vs. lettered), founding year, mixed-gender unit assignment, and investor type. Odds ratios are reported with robust standard errors in parentheses. All models restrict the sample to male founders with military service (N=183). ***, **, * denote significance at the 1%, 5%, and 10% levels, respectively.

Theory

Resource Redeployment Theory: Venture Capital Firms as Internal Market Nexus

Elite VCs aren’t at the top because they are great pickers. They got there by orchestrating internal markets that turn failure into value.

The innovation industry is characterized by high failure rates, with few successes justifying the risks. Founders and investors bet on each other, where startups surrender portions of equity and control for capital, advice, reputational signaling, and network access. Yet, for venture capital firms (VCs) operating at the nexus of innovation financing, a persistent puzzle remains: why do certain VCs consistently outperform peers? Existing explanations emphasize superior selection and counseling; however, they fall short when considering elite VCs’ enduring success across cycles, geographies, and industries. We introduce Resource Redeployment Theory (RRT), reframing VC-portfolio firms not as independent bets but interdependent nodes in a constellation of assets, some of which complement, and others substitute one another. RRT argues VCs develop a distinctive capability to identify, extract, and repurpose assets—like talent, code, and customer goodwill—from struggling ventures into high-performing ones. This ability to redeploy intangible assets from one node to another frames VCs as conglomerates that lower transaction costs by creating an internal market, increasing the salvage value of floundering startups. RRT turns VC-startup matching theory on its head, contending that VCs do not win by being great pickers or sage mentors, but by orchestrating internal markets, turning failure into fertilizer.

RRT Analytical Architecture

Notes. This figure illustrates the mechanisms of RRT. Redeployment capability, supported by enabling conditions and strengthened by exposure to failure (Props 2 and 5), enhances VC performance through frequent and effective redeployment (Prop 6), larger and diverse portfolios (Prop 4), and founder concessions on valuation and equity (Prop 3). The effect is strongest in sectors with moderate resource fungibility and strong VC governance rights (Prop 7). Visual conventions: Solid lines represent formal propositions (P1–P7), while dashed lines indicate conceptual relationships discussed in the text but not formalized as propositions. Solid boxes denote core constructs of the theory, whereas dashed boxes represent enabling conditions or boundary conditions that shape how redeployment operates.

Venture Selection Under Uncertainty—Balancing Type I and Type II Errors

Notes. This figure illustrates VC’s decision process when evaluating startups based on noisy signals. The curves represent the probability distributions of perceived quality signals for ultimately successful and unsuccessful ventures. The vertical dashed line marks the VC’s selection threshold (i.e., its self-imposed rejection line). Shaded areas indicate the possible outcomes of the decision rule: Correctly rejecting bad ventures (1-α). Correctly selecting good ventures (1-β). Type I Error—Incorrectly funding an unsuccessful venture (α), and finally, Type II Error—Incorrectly rejecting a successful venture (β). The more selective a VC is, the further to the right its rejection line will be, causing it to minimize Type I errors but at the expense of missing out on promising startups (Type II).

Multiple Choice Survey

Discrimination in Two-sided Matching Market: Experimental and Theoretical Evidence in Entrepreneurial Finance

Founders discriminate against female investors, especially high‑quality ones, creating a “glass ceiling” limiting women’s participation in entrepreneurial finance.

This paper explains the generation process and documents special features of statistical discrimination within two-sided matching markets, focusing specifically on the entrepreneurial financing market. Through an experiment involving real US startup founders, we first identify the presence of statistical gender discrimination against female investors among startup founders. Specifically, female investors’ signals are perceived as less informative compared to those of male investors. This discrimination is predominantly driven by male founders and disproportionately affects high-quality female investors, suggesting the existence of gender homophily and a “glass ceiling” for women in this context. Building upon these experimental findings, we develop a novel search and matching model with endogenous information aggregation and belief formation. This theoretical framework explains how statistical discrimination can arise endogenously within two-sided matching markets, leading to the observed “glass ceiling” distributional effect and perpetuating a persistent low female investor participation rate in equilibrium. Overall, this paper offers novel insights into the nature and distinct characteristics of discrimination within two-sided matching markets.

Ratings-guided matching market

Notes. The search-and-matching framework is illustrated, where blue and red represent the two groups. Darker (lighter) colors represent high (low) quality types, respectively. The “star” represents a good rating. The arrows represent the stochastic transition of investor types and ratings. We introduce an exogenous disruption to matching formation into the framework.

Gender Discrimination Across Profiles (Male Founders vs Female Founders)

Notes. This figure demonstrates how investors’ gender aﬀects the contact interest ratings of male startup founders and female startup founders as the study progresses to the end. Panel A uses evaluations of male startup founders. Panel B uses evaluations of female startup founders. The horizontal line describes the order of each investor profile displayed to the experimental subjects (i.e., the kth displayed investor profile). The vertical line is the coeﬃcient of “Female Investor” of the following regressions: Q4ij = α + β1 Female Investor(ij) + β2 Asian Investor(ij) +ϵ(ij) for all subjects’ evaluation results of the kth displayed investor profiles, with 95% confidence interval. This represents the magnitude of gender discrimination as measured by startup founders’ contact interest ratings (i.e., Q4).

Archival

Code Washing: Evidence from Open-Source Blockchain Startups

Some Blockchain startups fake coding activity to raise funds, but only true code producers deliver lasting returns.

This study examines startups’ management of source code repositories, distinguishing authentic developers (“code-producers”) from those inflating activity to mislead investors (“code-washers”). Using global blockchain startup and GitHub data, we find that code-producers and code-washers achieve greater fundraising success during hot markets than startups without repositories, indicating that investors struggle to evaluate open-source innovation accurately. However, while code-washers experience poorer outcomes post-fundraising, a portfolio of code-producers generates substantial long-term returns. Our study introduces the novel phenomenon of “code-washing,” offering insights into how early ventures navigate (or exploit) information asymmetries during the fundraising phase.

Funds raised by blockchain startups

Notes. This figure illustrates the total amounts raised in fundraising attempts by blockchain startups from 2013 to 2020. Fundraising amounts in all currencies have been converted to U.S. dollars and are expressed in millions.

Code production around fundraising and exchange listing events

Notes. This figure illustrates the evolution of code production, captured as GitHub code commits, around the fundraising start date and the first exchange trading date of blockchain startups in our sample. The time window spans from 270 days before the event to 365 days after the event. The y-axis represents the number of commits made within each time interval. For example, the point labeled ”st” represents the number of commits made in the interval [-30 days, token offering start date]. It shows the code production for code-producers and code-washers around the fundraising event.

Page updated

Google Sites

Report abuse