Agenda driven research conclusions: Advocates will find what they want, regardless of what the data says.

 From Gender gap in entrepreneurship by Jorge Guzman and Aleksandra (Olenka) Kacperczyk.  From the Abstract.

Using data on the entire population of businesses registered in the states of California and Massachusetts between 1995 and 2011, we decompose the well-established gender gap in entrepreneurship. We show that female-led ventures are 63 percentage points less likely than male-led ventures to obtain external funding (i.e., venture capital). The most significant portion of the gap (65 percent) stems from gender differences in initial startup orientation, with women being less likely to found ventures that signal growth potential to external investors. However, the residual gap is as much as 35 percent and much of this disparity likely reflects investors’ gendered preferences. Consistent with theories of statistical discrimination, the residual gap diminishes significantly when stronger signals of growth are available to investors for comparable female- and male-led ventures or when focal investors appear to be more sophisticated. Finally, conditional on the reception of external funds (i.e., venture capital), women and men are equally likely to achieve exit outcomes, through IPOs or acquisitions.

Intriguing.  They seem to create all the data for one conclusion and then propose something else entirely.

The most powerful finding is in the last sentence.  For men and women who have both passed all the venture capital tests and receive funding, 

Women and men are equally likely to achieve exit outcomes, through IPOs or acquisitions.

Would have been interesting to know some of the effect size details such as:  

What percentage of all those who receive funding complete an exit? - Do 100% end up exiting via IPOs or acquisitions?  50%  35%.  We know it is a small percentage but what percentage?

Do men and women achieve the same levels of exit valuation? -  You don't just want to exit, you want to exit on the most advantageous terms.  So is there an exit differential in valuation?  

I cannot find answers to either question after a quick read of the study.  It is a complex modeling exercise and initial expectations can often bias the outcomes through definitional issues.

Regardless, as stated in the abstract, we can start with the assumption that venture capitalists (VC) are equally good at choosing who is going to be successful as measured by the proxy of exit by IPO or acquisition, regardless of gender.  Valuation would have been the better and more direct measure of success and it is curious why that was not used.

That is strong evidence that VCs are gender blind when making investment decisions.

But the researchers aren't focused on whether VCs are gender-blind in estimating success.  The researchers are interested in determining why women are being disproportionately denied VC funding.  Right off the bat, they rule out 65% of the potential causes as non-gender related.  Women more frequently seek VC funding for ventures VCs are unwilling to fund, i.e. low or slow growth ventures.  Women applicants for VC funding are disproportionately seeking funding for low growth ventures which VCs don't want to fund.  The researchers acknowledge this to be a difference in choices, in female choices, not discrimination against women per se.  

The researchers are left to explain 35% of the gap in funding.  It is at this point that they begin imposing assumptions which determine the conclusions.  What the heck does this mean?

However, the residual gap is as much as 35 percent and much of this disparity likely reflects investors’ gendered preferences.

It means that the researches simply assume into existence that that the 35% difference in outcomes must be the result of gender bias.  But if actual achievement (IPO/Acquisition) is the same between men and women, that strongly suggests that at least some portion (if not all) of the 35% differential must be other factors which the researchers failed to acknowledge or understand.  

The 65% which is acknowledged as non-gender is described in detail:

Our findings confirm the well-established pattern that female-led ventures are significantly less likely to obtain funding. As much as 65 percent of the total disparity in funding can be attributed to differences in startup growth potential at the time of founding. In this regard, our findings suggest that women are significantly less likely than men to found ventures that exhibit growth orientation and are appealing to external investors. Specifically, women are less likely to found and run startups that have appropriable and differentiated technology (as evidenced by patents), to found companies in sectors associated with venture capital such as biotechnology, IT, or semiconductors, and to register the company in Delaware—a jurisdiction associated with an intent to raise external financing. Women are also more likely to start firms in industries associated with a local business activity, rather than traded.

All that seems pretty robust.  

Now they describe the 35% they don't understand:

The residual gap (i.e., 35 percent of the gap, or 18 percentage points) can be attributed to other factors, including, at least in part, investors’ preferences and bias. We provide evidence to link this remaining difference with investors’ biases or beliefs about gender. Specifically, we find that the gap between female and male entrepreneurs closes when signals of growth orientation become stronger (i.e., for female founders at the top of the quality distribution), or when investors are more sophisticated. Both findings are consistent with theories of statistical discrimination, suggesting that gender can be used as a cue to infer information about a new venture when signals of growth potential are weaker or when evaluators are less capable and less experienced.

Always watch for the weasel words - "The residual gap . . . can be attributed to other factors", "at least in part", and "suggesting that gender can be used."   Yes, it might be gender discrimination but we cannot simply assume that into existence.  

If the candidates selected by VC have equal outcomes (equal IPOs/Acquisitions) then there is something in their 35% decision-making which clearly is fulfilling the objective of picking winners.  Guzman and  Kacperczyk don't know what those other factors are, so they assume it must be cretinous old school discrimination.  The equality of outcomes in terms of successful market exits suggest Guzman and  Kacperczyk don't know what they are talking about.

Unless I am very badly misunderstanding the research, always a possibility, this seems like a prime example of cognitive pollution.  It appears that the researchers really wanted to find discrimination against women.  

What they are able to demonstrate is that VC backed ventures are successful regardless of gender.  In other words, their successful winners are equally likely to be male or female led.  The only differential in value creation comes at the funding stage.  Male ventures more frequently receive funding but for readily identifiable reasons based directly observable differences in choices related to potential for value creation (high growth, expansive markets, technology, patents, etc.).  That explains 65% of the difference in funding decisions.

We don't know what constitutes the reasons for the remaining 35% differential.  The researchers want it to be gender discrimination and claim that they find some indicators that it might be simple prejudice but do not elaborate on that.  The fact that there is no difference between genders in terms of success rates argues pretty powerfully that 1) whatever the missing decision factors are, they are strong, and 2) that they are not likely to be gender discrimination per se.  

We cannot ignore that VC investors are portfolio investors just as in other industries such as consumer product companies, real estate companies and the oil industry.  You don't invest in individual ventures, you invest in a portfolio of ventures: you spread your risks.  

Start-ups tend to be winner-take-all and Pareto distributed.  Only a few will succeed, those that succeed will account for all the value created, and no one can consistently pick who will individually succeed.  Equal successful outcomes seems to indicate that VCs are successfully focusing on relevant, non-gender variables.

The researchers claim

Our results have important policy implications. 

And then list out a typical range of policy interventions which are unproven.  Strikingly, where those type of policies are pursued are in the very states from which they are obtaining their data, California and Massachusetts.  

This is all very similar to the gender wage discrimination claim, still widely circulated but long since empirically overturned.  When you take into account education choices, industry sector choices, work persistence, hours worked, etc.,  the nominal wage discrimination shrinks to vanishing point.  

Guzman and  Kacperczyk have accounted for 65% of the nominal apparent discrimination against women at the funding stage.  In a few more years, if they stick with it, they will reluctantly end up finding, just as with the wage discrimination claim, that there are valid factors which they are failing to understand.  All the discrimination is based on eventually identifiable measures of predicate success.  

Until then, this is cognitive pollution masquerading as research. 

Or so it seems.