13 Feb Is the Midwest Undercapitalized? A Conditional “Yes.”
A central piece of our investment thesis at Lewis & Clark Ventures is that the Midwest is undercapitalized. But is it?
Our daily conversations with entrepreneurs provide anecdotal evidence that the middle of the country is capital-starved. So too do the many examples of good companies moving to the coasts to raise money.
But it could be argued that entrepreneurs always perceive a shortage of capital. Raising money is onerous, and this simple fact might lead founders to see a shortage of capital. Further, for every story of a Midwestern company moving to the coasts, there could be a similar story of a coastal firm moving to the heartland. Arch Grants alone has lured a handful of companies from the coasts here to St. Louis.
So it is possible that anecdotal evidence has made for shaky inferences. Maybe there is no shortage of venture dollars in the Midwest, just fewer opportunities, and this explains the paucity of VC dollars in the middle of the country.
One way to go beyond the anecdotal evidence is to consider the level of innovation in a state relative to its VC activity. Presumably, states with more innovation should have greater VC activity. States with relatively high levels of innovation but low levels of VC activity are undercapitalized.
Why? Think about it for a second. The whole idea of a state being undercapitalized is that within the state there is latent demand for venture capital – that is, real innovation not being funded. To be clear, this doesn’t just mean that the state has a bunch of companies failing to get VC dollars; this happens in every state. It means the state has genuinely innovative companies not receiving venture funding. Needless to say, this latent demand can’t be observed. So what we’re forced to do is to look at other indications of innovation, like patents.
The figure below is a plot of patents and VC activity. The patents figure is the average number of patents by state from 2001 to 2014. The data is from the US Patent and Trademark Office. VC Deals and VC Dollars are also averages. The data comes from a NVCA/PWC study of VC activity between 2009 and 2014.
So, what does the figure tell us? It provides fairly strong support for the hypothesis that the Midwest is undercapitalized. Among the Midwestern states that received at least five venture investments (shaded in green), only one (Kansas) is above the regression line. Missouri is positioned right where the regression predicts. But the remaining Midwestern states – Iowa, Indiana, Wisconsin, Ohio, Minnesota, Illinois, and Michigan – are all below the regression line, and in some cases well below the regression line.
If capital isn’t flowing to the Midwest, where is it going? Right where we would expect: to the coasts. DC, Virginia, Maryland, Pennsylvania, New York, and Massachusetts are all above where the regression line predicts. (If California were displayed, it would also be above the regression line.) Among the states on the I-95 corridor (from DC to Boston), only New Jersey falls a good deal below the regression line.
The evidence seems to support the claim that the Midwest in undercapitalized, then. So why the conditional “yes” in the title? The reason has to do with what we can safely infer from the figure. The figure tells us that Midwestern states receive less capital than coastal states, given their respective levels of innovation. This evidence is certainly consistent with the conclusion that the Midwest is undercapitalized. But it is also consistent with the coasts being overcapitalized. Perhaps too much money is flowing to Silicon Valley, New York, and Boston – the thought is certainly not a new one.
 There does not seem to be a stronger correlate for VC dollars in a state (and VC deals for that matter) than patents. State GDP and R&D spend also seem to matter, but their effects are minimal compared to Patents. The finding that patents and VC Dollars are correlated is fairly unsurprising, though it raises important questions about causality.
 I’ve taken out the labels for states with near-zero VC activity (Nevada, Arkansas, West Virginia, Wyoming, Delaware, South Dakota, North Dakota, Alaska, New Mexico, Vermont, Hawaii, Louisiana, Nebraska, Montana, Kentucky, Oklahoma, Maine, Mississippi, Alabama). I also removed California from the figure in order to show the variation among the other 49 states. California (on average) had more than 34,000 patents and over 1500 VC Deals for a total value of $15.6 Billion. Hence when I include California in the figure, the rest of the US (with only a couple of exceptions, like Massachusetts) looks like an undifferentiated mass.
 The regression line is for a simple univariate linear model with VC deals on the left side and patents on the left. Patents explain a surprising amount of the variance in VC deals. The adjusted R-squared is over 0.9. I also fit a number of multivariate models, but (as I mentioned above) patents was always the most successful predictor, and the explanatory power of the model never jumped significantly with addition of more variables.