The primary data in this study come from PitchBook, a leading vendor of information on venture capital deals, and the individuals, companies, and investors involved in them. All figures here include venture deals that were completed in each particular calendar year between 2005 and 2017 (inclusive) for companies with headquarters in the United States. Deals completed among the “pre-venture” series (accelerator, incubator, angel, or crowdfunding) are excluded (because they are not included in round sequences in the PitchBook database). More than 95 percent of first financing deals are either Seed or Series A.
Central to the work here, PitchBook tracks information on individual founders for companies in the database, and assigns a gender identifier through a two-step process—first manually via a primary research and secondly through an algorithm that assigns gender based on given names. Where results differ, further research is conducted to resolve discrepancies. Individuals are given gender values of female, male, or other/unknown. As a simple check, we conducted a detailed review of a random sampling of companies identified as having at least one female founder by PitchBook and found their results to be entirely accurate.
Companies in this study are considered “women-founded” if they had at least one verified female founder—as opposed to cases where all founders were female. The former was chosen over the latter for a few reasons. First, the size of founding teams varies widely across industries. In life sciences, for example, the number of founders can be quite large, and imposing an all-founders requirement for gender would skew the results. An all-founders requirement would also classify companies with founder-level contributions by women as “non-women-founded,” which not only feels outside the bounds of what we’re trying to understand here, it is arguably inaccurate. Third, imposing an all-women requirement for companies to be women-founded goes precisely against the entire point of promoting diversity in entrepreneurship. Finally, an all-founders requirement would limit our analyses because the pool of companies would be so small.
While many reports on venture capital deals and on women-founded startup funding focus on the entirety of venture activity (deals or capital invested at all stages), this study focuses primarily on the first round of financing by professional investors (“first financings”) for a few reasons. First, we are primarily interested in understanding the venture-backed companies most closely associated with “starting up” (as opposed to “scaling up”), and we do this by capturing the companies as they enter the venture pipeline (a measure of flow). Second, we wanted to better understand the number of companies that get funded, rather than the amount of capital going into them. Finally, since we want to understand how companies are performing over time in a comparable way, we had to construct annual cohorts and observe key outcomes over a similar time horizon.
To produce annual “first financing” cohorts of companies by gender identity of founding teams, a multistep approach was taken. First, we used the PitchBook platform to tabulate annual lists of companies that completed a first round of venture financing in a calendar year for each of the thirteen years. Next, the lists were sent to PitchBook, which used the back end of its database to flag the companies where at least one female founder could be verified. The list was returned to us with indicators for companies with at least one female founder. We then constructed two corresponding lists for each annual cohort of first financings back on the PitchBook platform—one for companies where at least one female founder could be verified and one where at least one female founder could not be identified. That allowed us to conduct most of the remaining analyses contained in the report (the lone example was geography; see below).
The tabulations and plotting of data across all first financings, by industry, and by geography, were relatively straightforward. The number of first financings in a particular year, naturally, were those completed between January and December. Sector and detailed industry classifications are pre-populated by PitchBook. For geography, PitchBook provided us with lists of first-financing counts by state, city, and zip code. Using files from the U.S. Office of Management and Budget and the Census Bureau, each combination (where the information was available) was mapped back to any one of a metropolitan area, micropolitan area, neither of these, or unknown. This analysis was restricted to metropolitan areas and the United States as a whole.
The analyses for follow-on outcomes were conducted in the following way. Each first-financing cohort (say, 2005 for women-founded companies) was loaded into the PitchBook platform, and search queries were performed based on outcome (e.g., second round of financing, IPO) and the appropriate time lag (three, five, eight, or ten years from first financing). For example, an outcome of acquisition for a company in the 2006 cohort would have had to occur after its first financing in 2006 and before either December 31, 2014 or December 31, 2016 (eight-year and ten-year exits), and so on. Numbers were tabulated for each outcome for the maximum number of cohorts and presented as a share of all first financings for each cohort. All analyses were conducted in the PitchBook platform based on the cohort lists derived as per the above.
Finally, because we took a conservative approach for identifying companies as women-founded and non-women-founded (i.e., those where a female founder could not be confirmed), the latter category may be considered by some as overly expansive, since a number of these companies lacked information on founders entirely. As a check, we replicated our analysis across three groups of founder types: women-founded (at least one verified woman founder), non-women-founded (at least one verified male founder and no women founders), and unknown (where the gender or identity of no founders could be confirmed). The results were strikingly similar.