Some thoughts (ie, long thread) on this very interesting and very provocative paper by @Astrid_Marinoni and @john_voorheis, which got a lot of attention yesterday. https://twitter.com/john_voorheis/status/1188809442051219458?s=19 /1

John tweets, “the gains from entrepreneurship go overwhelmingly to the top of the income distribution, and there is little to no evidence that there are spillover effects.” https://twitter.com/john_voorheis/status/1188809462192201729?s=19 /2

The paper’s conclusion makes an even bolder claim: /3

These are pretty strong conclusions! Given these statements, one would expect the paper to go through the various hypothesized social benefits of entrepreneurship in exhaustive empirical detail, ruling them out one by one. /4

That’s not what the paper does, so I’ll offer some friendly push-back on the broad claims. /5

I am not going to critique the paper's methodology, except to note the geographical and temporal limits of the analysis. I'll focus more on how Astrid & John are summarizing the results. /6

First, geographic and temporal limits: they find that the founding of high-potential firms doesn't boost incomes at the bottom, *within the same commuting zones & within a short time of founding.* /7

I would not necessarily expect all the effects of high-potential firms to be evident only within the local area and a short time horizon, especially where low-wage workers are concerned. /8

The literature finds that growth plays out over a number of years (and it's plausible that early growth is more about high-skill workers) /9

There’s a bunch of literature on the growth dynamics, but a concrete way to look at it is this neat 2017 paper by Michael Mandel. https://www.progressivepolicy.org/wp-content/uploads/2017/08/PPI_TechJobGrowth_V3.pdf /10

Ok, on to the broader implications of Astrid and John’s paper. /11

Two critiques: (1) boosting the income of locals in the short run is just one of many social benefits we might expect entrepreneurship to provide. (2) The authors’ definition of "entrepreneurship" is useful, interesting, but also extremely narrow. /12

Re (1): when people talk about the benefits of entrepreneurship, they are often thinking of several things including increased competition to incumbent firms; innovations that can lead to increased product variety; /13

maybe some vintage capital effects; contributions to aggregate productivity that have been well documented; aggregate job creation over the lifecycle (also well documented); and others. /14

The paper’s focus on local areas for a research design has some nice benefits but I wouldn’t expect it to capture most of these, particularly if these high-end entrepreneurs are primarily getting into tradeable industries (eg, tech). /15

The authors are basically focusing on the job creation angle, specifically low-income wages, finding no effect. /16

So either the job creation benefits truly aren’t helping folks low on the income ladder, or it takes awhile, or any low-income folks who are benefiting are outside the firm HQ’s local area. /17

I will also add that they *don't* find evidence that entrepreneurship makes "the poor poorer." They find no effect on the poor. /18

In any case, I don’t see this paper as ruling out the many other hypothesized benefits. That would require a different design (and probably some GE structure). /19

For example, do low-income folks benefit from increased product variety? A channel you can’t measure with this design. /20

Another: Lately I have seen a lot of discussion of the benefits of competitive product and labor markets. If startups contribute to competition, do low-income folks benefit? /21

These are a few of the things not captured in this particular research design. Of course, that is fine--no paper studies everything, and the paper tells a compelling story about the things it is studying. /22

Re (2): defining entrepreneurship. Guzman & Stern did the world a huge favor by constructing their measures of high-end entrepreneurship. Check out their project if you haven’t. But note a couple things. /23

First, the specific G&S measure used in the paper (RECPI) measures the registration of firms with high *forecasted* funding outcomes. It does not measure whether firms actually grow or not. /24

(G&S do have another index, REAI, which does measure whether these high potential firms end up having high outcomes). /25

Second, the kind of startups they are identifying are extremely rare (as G&S often note) and represent a tiny subset of all “entrepreneurship.” /26

They seem to even represent a small subset of all *high-growth* entrepreneurship, if the latter is defined as it has sometimes been in the literature. /27

For example, when we look at the top job creators among young firms by defining “top” as the top 10% of the employment-weighted distribution of young firm employment growth rates, we see DHS growth rates exceeding 60%/year. /28

A firm with a DHS growth rate above 60% is nearly doubling in a year. A few years of that, and you are creating a lot of jobs. Are we to think none of them are for low-wage folks? /29

The authors note that they don't want to just study subsistence entrepreneurship, which is smart, but there's a lot of ground between subsistence and the G&S measure. /30

Young firms with DHS growth >60% are accounting for 10% of young firm employment. That is, this is a much larger set of firms than the set being studied by Astrid and John, yet they have a huge impact on labor markets. /31

So it may be that the paper’s conclusions are telling us a lot about a tiny fraction of startups—the ones that have big private valuations or go public—and not much about many other startups, including high-growth ones. /32

There is a literature that studies other startups and indeed finds them providing a lot of benefits. For example, very recent work by Babina et al. finds that young firms are more likely to hire folks with lower previous earnings. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3425596 /33

And, interestingly, Babina et al. find that once you control for worker & firm characteristics, young firms actually pay better than older firms. /34

Or consider this very recent paper by @jmirandal, Sandusky, and Stinson, which finds that working at a young firm boosts later-in-life employment substantially, especially if it’s a high-growth firm. https://drive.google.com/file/d/1_uG2bAjZ7AIPjR0O_sKnz9NWb4YbtuxZ/view /35

We could go on; there is a lot of literature on this. The point is: We should not react to this very interesting and useful paper by concluding that entrepreneurship doesn't benefit the poor. /36

I would not be surprised if high-growth entrepreneurship overall increases inequality; I have read my Quadrini (2000). But I will be surprised if it provides *no benefits* to the bottom of the distribution. /37 end