Innovation has long been seen as central to long-term regional growth. Due to the absence of comprehensive data on the geography of innovation covering long time periods quantifying long-term innovation-development linkages has been challenging. We use newly available patent data from the United States coded to consistent geographies over 150 years to document changing patterns in the geography of innovation. Our analysis reveals three findings. First, the high levels of spatial concentration of innovation today are similar to those in the decades after the Civil War. Second, changes in share of the top 1% locations' innovation drive national spatial concentration trends after 1945. Third, regional innovation leadership displays persistence, but the strength of persistence appears to have fallen over time. We relate our analysis to recent findings in the literature and suggest promising avenues for future inquiry.
Featured on: VoxEU
Journal of Economics and Management Strategy, Vol. 30, No. 2, Summer 2021, pg. 368-397
I provide a primer on six recent large-scale historical patent datasets for use in innovation research. I discuss how each dataset is constructed, the types of patent information included in each, and the quality and completeness of each. Throughout, I emphasize when our knowledge of the history of invention is dependent on the data source used and provide recommendations about which dataset is most likely to be best for different contexts. Overall, these datasets paint a remarkably consistent picture of the history of U.S. invention. When the datasets do disagree, these differences tend to be minor, although I highlight some important exceptions. I further describe several "niche" historical patent datasets that allow researchers to study institutional contexts that cannot be studied using modern patent data. Finally, I discuss features of patent data that are not available for the historical patents but are available for modern patents.
Previous versions of this paper circulated with the titles "Comparing Historical Patent Datasets" and "Historical Patent Data: A Primer and Some Pitfalls."
Explorations in Economic History, Vol. 74, Oct. 2019
We assemble a novel dataset linking inventors listed in the Annual Reports of the Commissioner of Patents to Population Census records spanning 1870 to 1940. We find that inventors are not a random subset of the population. They differ in some unsurprising ways in that they tend to be older, whiter, and more likely male. However, these patterns do change over time. The odds ratio relative to the population as a whole of female inventors increases from a low of 0.07 in 1880 to a high of 0.13 in 1940 and that of non-whites ranges from 0.16 in 1880 to 0.34 in 1940. Both populations remain severely underrepresented throughout the timeframe. We find changes in the occupations of inventors with trends away from farming and towards white collar occupations. We also show the increasing importance of foreign born people in patenting. In 1870, the odds of a foreign born person patenting relative to the population as a whole is nearly 1 and increases to over 1.6 by 1940.
Featured in the USPTO's Study of Underrepresented Classes Chasing Engineering and Science Success report to Congress
In The Economics of Research and Innovation in Agriculture, edited by Petra Moser, University of Chicago Press, 2021, pg. 139-175
To estimate the local effects of establishing land grant colleges, I compare locations that receive a land grant college to "runner-up" locations that were in contention to receive the land grant but did not for as-good-as-random reasons. I find that establishing a land grant college causes an increase in local invention, including in particular agricultural inventions, in college counties relative to the runner-up counties. But land grant college counties see only small and imprecisely estimated improvements in agricultural performance, measured by yield and output, relative to the runner-up counties. I discuss several alternative interpretations of these findings. By comparing the establishment of land grant colleges to non-land grant colleges, I show that land grants appear to cause smaller increases in local invention, population, and agricultural output, but larger increases in agricultural yields and new crop varieties. The effect of land grant colleges on local innovations is largest, even relative to non-land grant colleges, following the passage of legislation that increases funding to agricultural research.
Journal of Economic Literature, Vol. 59, No. 1, March 2021, pg. 294-296
A review of Urquiola's 2020 book, which argues that the decentralized nature of the U.S. higher education system led to the U.S.'s dominance in university research.
Brookings Report, Nov. 2020
The history of Black people’s contributions to the catalog of inventions that marked the Second Industrial Revolution has been largely muted. This period is considered one of the most innovative eras in world history, seeing the birth of major advances in agriculture, transportation, communications, manufacturing, and electricity that fueled rapid economic growth. With the exception of a few notable inventors who are regularly elevated during Black History Month—e.g., George Washington Carver (peanut products) and Madam C. J. Walker (hair products)—the disregard of many of the era’s Black inventors not only whitewashes the historical record, but biases who we perceive to be innovators in the present. Using a new database of inventors, this report demonstrates that Black contributions to the Industrial Revolution were influenced by the disproportionate number of Black Americans who lived in the U.S. South in the late 19th and early 20th century, where their opportunities to acquire and apply skills were severely limited by oppressive institutions. Black Americans living outside the South invented nearly as frequently as white Americans, and at rates that would be considered extremely high by historic or global standards of invention even today.
Journal of Economic History, Vol. 79, No. 2, June 2019, pg. 514-517
A short summary of my dissertation.
Conditionally Accepted, American Economic Journal: Economic Policy
I use data on site selection decisions for a subset of U.S. colleges to identify "runner-up" locations that were strongly considered to become the sites of new colleges but were ultimately not chosen for reasons as good as random assignment. Using the runner-up counties as counterfactuals, I find that establishing a new college causes 45% more patents per year. Controlling for county population explains most of this observed increase. Linking the patent record to a novel dataset of college yearbooks reveals that only 12% of patents in a college's county come from individuals directly affiliated with that college as either alumni or faculty. College counties do not appear to attract especially skilled migrants. I find statistically indistinguishable differences in patenting between locations that establish colleges and locations that establish other types of institutions, as well as between colleges that are more and less focused on technical fields.
with Lauren Russell and Lei Yu
Revise & Resubmit, Review of Economics & Statistics
We investigate how establishing a college affects local educational attainment using historical natural experiments in which "runner-up" locations were strongly considered to become college sites but ultimately not chosen for as-good-as-random reasons. While runner-up counties have since had opportunity to establish their own colleges, winners are still more likely to have a college today. Using this variation, we find that winning counties today have college degree attainment rates 58% higher than runner=up counties and have larger shares of employment in high human capital sectors. These effects are not driven primarily by college employees, migration, or local development.
To understand the importance of informal social interactions for invention, I examine a massive and involuntary disruption of informal social networks from U.S. history: alcohol prohibition. The enactment of state-level prohibition laws differentially treated counties depending on whether those counties were wet or dry prior to prohibition. After the imposition of state-level prohibition, previously wet counties had 8-18% fewer patents per year relative to consistently dry counties. The effect was largest in the first three years after the imposition of prohibition and rebounds thereafter. The effect was smaller for groups that were less likely to frequent saloons, namely women and particular ethnic groups. Next, I use the imposition of prohibition to show that the social network exhibited path dependence in the sense that as individuals rebuilt their networks following prohibition, they connected with new individuals and patented in new technology classes. Thus, while prohibition had only a temporary effect on the rate of invention, it had a lasting effect on the direction of inventive activity. Additionally, I exploit the imposition of prohibition to show that networks increase invention by exposing individuals to others' ideas in addition to simply facilitating collaboration and that informal and formal interactions are complements in the invention production function.
We estimate how the availability of informal gathering places can facilitate knowledge spillovers that lead to innovation. More specifically, we estimate how the spread of coffee shops increases local inventive activity. To account for endogeneity in the number of coffee shops in an area, we use an instrumental variables approach based on the geographic and time patterns of the opening of Starbucks coffee establishments. As Starbucks opened new establishments, these establishments tended to spread outward from the location of the first Starbucks establishment in Seattle. Using this strategy, we find that 10% faster growth in the number of new Starbucks stores in a county increases patent filings in that county by about 4-6%. Average patent quality also increases in counties that see a faster growth in coffee establishments. Collaborative patenting and patenting by firms both increase, as does collaboration across firms, providing suggestive evidence that coffee shops are especially conducive for cross-firm business meetings. We obtain similar results when using instrumental variables based on the geographic and temporal patterns of the opening of other large coffee chains as well, each of which had their first establishments in different parts of the country.
Featured on: New York Times
Forthcoming in the NBER Volume "The Role of Innovation and Entrepreneurship in Economic Growth"
This is an introduction to the forthcoming volume "The Role of Innovation and Entrepreneurship in Economics Growth." The chapters collected in this volume seek to answer the following questions: What is the relationship between innovation/entrepreneurship and economic growth in specific industrial sectors? How has the relationship between innovation/entrepreneurship and economic growth changed over time? How much do policies, programs, and specialized institutions meant to encourage innovation or entrepreneurship ultimately spur economic growth? Does innovation or entrepreneurship affect economic performance and social progress other than through measured productivity and economic growth, and if so, how can these effects be measured? We synthesize the chapters in this volume and present broad conclusions.
During the Industrial Revolution and subsequently, it is widely believed that African Americans contributed disproportionately little to the economic development of the United States, especially in comparison to European Americans and immigrants from Europe. Yet, African Americans lived in entirely different institutional environments than other Americans, particularly in the South under Jim Crow laws. Using a new database that matches inventors to census records, we find that the share of patents invented by Black Americans living in the North matched their share of the U.S. population from 1870 to 1940 and exceeded it in some decades and in some Northern states throughout the period. We find that Black Americans from all regions were responsible for more patents than immigrants from all but two countries (Germany and England). Northern Black Americans patented at rates that exceeded Southern White Americans and were comparable to the highest patenting groups. In total, we estimate that African Americans were the identified inventors of 50,000 patents over the period. Thus, when freed of extreme political oppression, African American demonstrated level of inventiveness that matched the most inventive groups in U.S. history.
By comparing counties that are near and far from land grant colleges using a variety of distance measures, I show that proximity is more important for agricultural productivity and output than for other measures of innovation in other fields. To shed light on how widely innovations from land grant colleges diffuse, I exploit a novel dataset that contains histories of new wheat varieties introduced in the U.S. before 1920. I find that only 10-17% of wheat acreage planted in varieties developed since the establishment of land grant colleges is planted in varieties developed at land grant colleges. Finally, I use data on the publications by researchers affiliated with land grant colleges to present direct evidence that, even more than a century after the land grant colleges were established, land grant research is biased towards crops that were initially most prevalent in land grant college counties, rather than those that were most prevalent in the rest of the state. This suggests that the location of land grant colleges has persistent effects on the direction of agricultural research.
Work in Progress
Drafts available upon request.
Markets for technology are affected by multiple kinds of imperfect information. In particular, asymmetric information (inventors know more about a technology than do potential buyers) and uncertainty (neither party knows whether a new technology will be a success) are both likely to be present, but little is known about the relative importance of these different kinds of information or how they interact. To shed light on this issue, we introduce a novel dataset that contains information about firms' expectations of the future value of their patents as well as information on patent transactions. The data come from firm reports filed with the state of Illinois to comply with a 1919 Blue Sky Law, an early investor protection law, and contain a wealth of information about firms' intangible asset holdings that are not available in modern data. Using this data along with tools from forensic accounting and machine learning, we construct a patent-level measure of imperfect information. We then use this measure to test several predictions of a simple model that incorporates both asymmetric information and uncertainty. We reject the null of zero asymmetric information and provide a framework for ranking technologies by the amount of information asymmetry.
We exploit the fact that historical population censuses were enumerated door-to-door to measure the importance of exposure to neighbors' knowledge and skills for invention and entrepreneurship. We create a measure of intra-city clustering of innovative activity at the microgeographic level by observing how frequently inventors and entrepreneurs have neighbors that are also inventors or entrepreneurs. Such a disaggregated measure allows us to better test theories related to knowledge spillovers, which typically postulate that interactions must occur at close geographic proximity for spillovers to occur. Door-to-door clustering of inventors and entrepreneurs appears to be quite rare and is not correlated with other local features. In household-level regressions, we show that a household's propensity to patent or become an entrepreneur is positively correlated with distance to innovative neighbors. Households are also more likely to invent or become entrepreneurs if they have neighbors with more diverse occupational, educational, or racial backgrounds. Ongoing work exploits cross-census movers to identify the causal effects of neighbors' characteristics on a household's innovativeness.
Go Forth and Serve: Geographic Spillovers from U.S. Colleges
Exploiting a historical college site selection natural experiment, I investigate the extent to which establishing a new college or university affects various outcomes in nearby counties that did not receive the college. Using several alternative measures of distance, I show that counties that are closer to the college see larger increases in patenting and population than do more distant counties. The increase is especially large for less distant counties that were themselves under consideration to become the sites of the new college. I next investigate the establishment of other types of institutions that cause similar increases in local population as colleges. These other types of institutions do not appear to generate the same kinds of positive spillovers to nearby counties. Ongoing work investigates the extent to which the observed effect on surrounding counties is driven by migration decisions of the college's alumni.