Analyzing Some (Miscited) Entrepreneurship Research

My background

While I was growing up my Dad was a postal clerk. Being a mailman is a decent blue collar job, all-in-all, and we were fine financially. But when I was twelve or thirteen, my Dad took a huge risk.

An avid reader, a curious soul, he felt stifled by his work. My Dad filled our house with books and our weekends always included a trip to the local library. He got an Associates degree by night, and then got a grant from (the NJ?) Department of Labor to go get a BA full-time. He quit his job. He spent all of his savings, retirement and otherwise, on graduating with a 4.0 GPA from Rutgers. Passionate about research, he went on to get an MA and PhD. He raised me through my high school years on his stipend alone. Coupon-cutting and free school lunch got us through those years. 

Through this time, my mother was completely out of the picture: I didn't see her between ages 10 and 18. Having said that, she's spent most of her career working in restaurants though: server, cook, etc.  

But it all seems worth it. My dad taught me that taking risks in pursuit of one's dreams was worthwhile, even at a deep economic expense. Since then, the post-great-recession era has not created the best academic market. My Dad has oscillated between short-term academic posts and unemployment. 

Needless to say, neither of my parents comes from any serious money. I didn't fall backwards in to a trust fund. Yet somehow I found a way to be an entrepreneur.

Entrepreneurs come from families with money?

Because of that fact, I was annoyed by the framing and conclusion of this article in Quartz "Entrepreneurs don’t have a special gene for risk—they come from families with money". Although the article is nearly a month old, it's sloppiness bothered me and it's holier-than-thou (counter-cultural?) argument against entrepreneurs is becoming more prevalent. That's fine, lot's of entrepreneurs are entitled, arrogant people — so are many journalists — but I find referring to large communities monolithically so commonplace, and so annoying, that I wrote this short essay.

The article makes a common mistake in journalist reflections of academic research: it turns a statistical fact ("On average, and holding all else equal, entrepreneurs are more likely to have received a gift or inheritance") and turn it into a categorical fact. The absolute divisions make better copy, sure, but reality is messy. I could likely spend my whole life pointing out these types of errors, but this particular instance got under my skin because I'm a fine but by-no-means-atypical counterexample to the "all entrepreneurs come from family money" claim.

The article is so poorly written on so many fronts that maybe I shouldn't be so upset. It convolutes so many different arguments, and makes so many different arguments that sound the same but are not. I do not come from a family with any financial stability. On the other hand, I am white, male, and highly-educated. "Earned" or unearned, I have a huge amount of privilege, pedigree, and connections.[1] I was able to take risks that many people couldn't because of these facts, but it does a disservice to suggest that all the people in a group share the same characteristics. If the author had just written the word "most" or "more than average", etc., then the article would have been well on its way to accuracy. 

The research on entrepreneurship that the article cites is interesting though and it points to some deeper policy points than throwing up your hands and saying you have to come from money to be an entrepreneur. I'm going to write a different article in the future about policy prescriptions, but let me analyze the four research citations given related to entrepreneurship.

Blanchflower and Oswald, "What Makes An Entreprenuer?", 1998.

Linked to from this sentence in the article: "But what often gets lost in these conversations is that the most common shared trait among entrepreneurs is [access to financial] capital—family money, an inheritance, or a pedigree and connections that allow for access to financial stability. "

 Four conclusions from this study:

  1. "consistent with the existence of borrowing constraints on potential entrepreneurs, we find that the probability of self-employment depends markedly upon whether the individual ever received an inheritance or gift"
  2. "when directly questioned in interview surveys, potential entrepreneurs say that raising capital is their principal problem"
  3. "consistent with our theoretical framework's predictions, the self-employed have higher levels of job and life satisfaction than employees"
  4. "childhood personality measurements and psychological test scores are of almost no help in predicting who runs their own business later in life. It is access to start-up capital that matters."

Let's dive in here though. Firstly, the study uses the National Child Development Study (NCDS): "a longitudinal birth cohort study that takes as its subjects all those living in Great Britain who were born between the 3rd and the 9th March, 1958". Before I say anything else about this study, might it be that there are differences between the UK and the US? Those inheritances might have had a larger impact in that society at that time than they might in the US now? That there might be large differences in these facts for between people born in 1958 and 1988?

Putting all that aside, although people who received an inheritance of over GBP5000, the cut-off in their analysis, are twice as likely to be self-employed, most self employed people did not receive a big inheritance. In fact, there are more self employed people who received absolutely no inheritance (1,142) than there are people who received over GBP5000 (692) altogether! [2] If you took a random entrepreneur from the data and asked the question in reverse than Blanchflower and Oswald [1998] does -- how likely are you to have received an inheritance -- the data shows the opposite of the Quartz article's claim.

So, the study cited does not support the sentence that links to it.

Ernst & Young, "Nature or nurture? Decoding the DNA of the entrepreneur"

Linked to from this sentence in the article: "While it seems that entrepreneurs tend to have an admirable penchant for risk, [it’s usually that access to money] which allows them to take risks."

To quote the study, "In the struggle to build momentum and grow their businesses, survey respondents and interviewees agree that founders face three main challenges: funding, people and know-how. And of those three, the biggest obstacle is funding." No doubt that it's true, raising money is difficult. But the citation says nothing about the article's central claim that entrepreneur come from money or have easy access to it. You might well conclude the opposite: so many entrepreneurs note that funding is their biggest obstacle so they must not have access to huge pools of family money or easy cash from connections.

Xu and Ruef, "The myth of the risk-tolerant entrepreneur", 2004.

Linked to from this sentence in the article: "While it seems that entrepreneurs tend to have an admirable penchant for risk, it’s usually that access to money which [allows them to take risks.]"

This is a particular egregious citation. It suggests, on my initial reading, that the linked to article would show that access to money allows entrepreneurs to take risks. The study has nothing to do with that claim! It doesn't relate to the argument one bit. The goal of this article is to "investigate whether entrepreneurs can be assumed to be more risk-tolerant than the general population". Their conclusion: Entrepreneurs are not more risk-tolerant. They found business organization for "non-pecuniary" reasons, like being their own boss. They are in fact more risk-averse because they're trying to peruse profits quickly so they can "lower the risk of business closure" and stay as their own boss. Xu and Ruef [2004] doesn't talk about the backgrounds of entrepreneurs at all, family or otherwise.

Levine and Rubinstein, "Smart and Illicit: Who Becomes an Entrepreneur and Do They Earn More?", 2013.

Paragraph from the article: "University of California, Berkeley economists Ross Levine and Rona Rubenstein analyzed the shared traits of entrepreneurs in a 2013 paper, and found that most were white, male, and highly educated. “If one does not have money in the form of a family with money, the chances of becoming an entrepreneur drop quite a bit,” Levine tells Quartz.

This study is the one that’s closest to supporting the central claim of the Quartz piece, but, again, the categorical nature of the claims is not supported in the empirics — or in Professor Levine's comments. On family background: mothers' education tends to be one year longer (12.6 vs. 11.7 years) for the incorporated self-employed (Levine and Rubinstein's proxy for entrepreneurship[3]), stable two-parent families are true for 83% of entrepreneurs vs. 76% in the general population, and average income for the family is 13k higher, which is a lot (70k vs. 57k). They also do tend to be whiter (83% vs. 70% of the population in the study), more male (72% vs. 52% of the population of the study), more educated by a half year (14.2 years vs. 13.8 in the general population), and slightly more college educated (36% vs. 30% in the general population). The study has some really interesting logit estimates on the probabilities of all of these things, but I'm not going to go in to all that.

I agree that the research here shows that most entrepreneurs are white, male, and highly educated (for some definition of that). But part of the point of all this is to say that statistical significance is not a proxy for actual significance. Saying in an academic paper that the backgrounds of entrepreneurs have more privilege than average, with the numbers plainly available to see, is one matter. Writing a sensational gotcha article that claims that "entrepreneurs ... come from families with money" feels like another.

This isn't even the big take-way from the article though: the big takeaway is that even when you control for whiteness, and richness, and maleness, it still takes something else to be an entrepreneur. We live in a racist, sexist, classist society, I don't think anyone doubts that, but the takeaway from this study — which is almost exactly what the Quartz article is trying to dismiss, is that:

as teenagers, the incorporated tend to have higher learning aptitude and self-esteem scores. But, apparently it takes more to be a successful entrepreneur than having these strong labor market skills: the incorporated self-employed also tend to engage in more illicit activities as youths than other people who succeed as salaried workers. It is a particular mixture of traits that seems to matter for both becoming an entrepreneur and succeeding as an entrepreneur. It is the high-ability person who tends to “break-the-rules” as a youth who is especially likely to become a successful entrepreneur.


There are also some big problems with the datasets looked at, which tend to be longitudinal in nature: they leave out the thriving entrepreneurial spirit in, for example, immigrant communities. To be in the two studies cited above which have serious data, you had to be born in the UK or live in the US at a young age, respectively. That data just does not account for a lot of the entrepreneurship I see. 

Sometime next week, I hope, I'm going to come back to this train of thought and articulate policy ideas around encouraging more entrepreneurship given the observations in these studies. 


[1] "Privilege" can end up an endless enumeration, but let me mention a few others: my vaguely being a Christian is I think not irrelevant, along with my being American (I felt fairly comfortable where I grew up and in all communities I have been part of, well, except Brown to start, but that's a different story); my being cisgender has helped me fit in with men in positions of power; my father valued education which is a privilege, etc. 

[2] This is basic Bayes Theorem reasoning. The conical example is usually given in terms of a medical test. Let's say you have a test that is 99% accurate but a disease that exists in 1% of the population. You use this test on a million people. 10,000 of them actually have the disease of which 9,900 are correctly identified as having the disease and 100 are not. 990,000 people do not have the disease, of which 9,900 are falsely identified as having the disease and 980,100 are correctly identified as not having the disease. So, if I get a negative result from the test, I can be pretty sure I don't have the disease (only 100/980,200 false negatives). But, if I have a positive result, there is only a 50/50 shot I actually do have the disease (9,900/19800)!

[3] This is an imperfect proxy, obviously.