By: Drona Gaddam
Drona Gaddam is a rising senior at Phillips Exeter Academy with an interest in macroeconomics. He is the founder and host of The Real Returns Show, an economics radio program that hopes to promote awareness around economic and civic systems. With this series, The Economy We’ll Inherit, Drona hopes to highlight topics relevant to the next generation entering the economy, with a particular focus on innovation and education.
July 17, 2026—In the early 2000s, economist Eric von Hippel argued that “innovation is rapidly becoming democratized,” as users aided by novel computer and communications technology became increasingly able to develop their own products and services. More recently, Mark Cuban has called AI “the great democratizer,” and much of the rhetoric on AI coding tools, low-code platforms, and online learning is similar. A student with a laptop can learn to code on YouTube, ask an AI chatbot to explain calculus, follow scientists on TikTok, or read startup advice online.
At its core, this optimism rests on a theory of exposure. If more young people can encounter technical ideas and recognize innovation as something people like them can participate in, then more of them will innovate. Raj Chetty gave this idea empirical force through his work on “lost Einsteins” in 2018. While Chetty may not have been thinking about Einstein’s own childhood when he coined the phrase, Einstein’s upbringing illustrates the idea almost perfectly.
Who are the Lost Einsteins?
Albert Einstein’s IQ is often estimated at around 160, but his intellect alone did not lead to his success. Einstein’s father and uncle ran an electrotechnical business, and, as a child, Einstein spent time on their factory floor, encountering some of the newest technology of the era. As Einstein grew older, another important influence entered his life: Max Talmey, a medical student who quickly recognized Einstein’s intellectual promise. While studying at Munich Medical School, Talmey often joined the Einstein family for dinner, where he introduced the young Einstein to works such as Bernstein’s Popular Books on Natural Science and Immanuel Kant’s Critique of Pure Reason. These books did more than deepen Einstein’s interest in science; they pushed him toward more abstract ways of thinking. Some historians have even argued that these early conversations helped lay the mental foundation Einstein’s needed for his theory of relativity.
In his 2018 research, Chetty believes that inequalities in exposure to innovation have led to the presence of countless “lost Einsteins” across the United States. According to Chetty, children born into the top 1% of the income distribution are about ten times more likely to become inventors than children in the bottom half, and the disparity is not due to differences in ability. Chetty compared children with high math scores, and low-income students were far less likely to become inventors than their high-income peers. A similar pattern appears across race and gender lines: white children are far more likely to become inventors than black children, and women remain sharply underrepresented among inventors. Race in the United States is closely tied to income, school quality, neighborhood segregation, and access to high-innovation professional networks. On the other hand, girls exist across every income group, and their underrepresentation points more to a specific lack of exposure to female inventors who make invention seem like an attainable path.
The Unfulfilled Promise of Technology as the Great Equalizer
Chetty’s study was published in 2018, which is recent enough to feel contemporary, but old enough that the environment he described has changed meaningfully. In the years since, the basic infrastructure of STEM exposure has expanded, giving new force to von Hippel’s older claim that innovation was becoming democratized. Foundational access to computer science has expanded dramatically over the past decade. In 2018, only 35% of U.S. high schools offered foundational computer science, but more recent Code.org data puts that figure at 57.5% of public high schools. AP Computer Science Principles has grown just as quickly, rising from 72,187 test-takers in 2018 to 175,261 in 2024. Outside the classroom, organizations like Girls Who Code have scaled rapidly as well, growing from 185,000 students served in 2018 to 760,000 by 2024. An important caveat is that these macro-level figures obscure which specific schools and student demographics actually benefited from this growth, leaving it unclear whether these resources reached the most underresourced communities.
Overall, however, beyond data points, the broader digital environment has made technical ideas much easier to access. A student who does not have a local inventor or startup founder in their neighborhood can still seek out coding tutorials, science explainers, or startup guidance online. AI has intensified this shift because it does not merely display information; it can answer questions, explain mistakes, and help with coding or research. More young people can now see innovation, and hence the world looks more democratized than it did when Chetty initially described the “lost Einsteins” problem.
Comprehensive data and federal policy rollouts from 2024 and 2025, however, reveal that this expansion of digital exposure has done little to move the needle on diversity behind innovative output. A 2023 report from the Ewing Marion Kauffman Foundation found that 65% of new employer businesses rely on personal or family savings to cover startup costs. A 2025 report from Social Mobility Ventures illustrated that founders not educated at a private high school were three times less likely to access friends-and-family capital, and 46% of founders said their friends and family simply could not afford to invest. Further, in May 2024, the United States Patent and Trademark Office (USPTO) established the persisting disparity across gender and social lines in its National Strategy for Inclusive Innovation. According to the report, white Americans also remain three times more likely to invent than Black Americans, and women account for only 12.8% of all named inventors on U.S. patents. Perhaps this data is a lagging indicator, and it will take longer for the proliferation of STEM resources and AI to make a meaningful impact. Another way to view this data, however, is through the lens of a production function.
The Economics Behind the Problem
In economics, a production function explains how different inputs combine to produce an output. Economist John List maps out an education production function that treats academic achievement as an output driven by funding, teacher quality, student effort and parental investment. Innovation can be thought of in the same way. The inputs may include raw ability, mentorship, capital, peer networks, geography, etc. The output is invention itself: a patent, startup, or new idea that actually enters the world.
Since Chetty’s study, one input has expanded dramatically: information. More students can now encounter resources online than ever before. But if the other inputs remain unequally distributed, then the output will reflect such inequalities. A student may now be more likely to see innovation, but not necessarily more likely to be mentored into it, funded through it, or connected to the institutions that produce it. This is a problem of information asymmetry and networks. Information asymmetry refers to one when one party in an economic transaction has more accurate information than another. In this case, it is not buyers or sellers withholding information from each other, but an asymmetry between socioeconomic classes’ exposure and access to resources surrounding innovation. Because high-income communities are more likely to be connected to these fields, their children receive more signals about how to enter them. Exposure creates aspiration, and aspiration becomes easier to act on when opportunities are within reach. Resources ranging from well-funded school robotics labs to early internship opportunities provide the scaffolding necessary to turn raw curiosity into realized innovation.
The internet may have lowered the frictions in exposure, but it has not equally distributed the full set of inputs needed to become an innovator. The central issue that Chetty identified still remains.
Chetty’s observation is one of allocative inefficiency. In economics, we often use this term when a mix of goods and services produced does not match consumer demand leading to deadweight loss. In a broader sense, allocative inefficiency is the misallocation of resources. In this case, networks, mentorship, education, and funding are not available to the most capable across the whole population, leading to loss in potential productivity and innovation. If innovative talent is spread evenly across the population, then an equitable and efficient economy should identify and develop that talent wherever it appears. But Chetty’s research suggests that the economy does not allocate innovative opportunities according to ability alone. Resources are allocated through family background, gender, and social networks.
Looking Ahead
This recent evidence is not perfect. Business formation is not necessarily the same thing as innovation, patent counts do not capture every form of innovation, and many new kinds of creativity may never appear in official statistics. A student experimenting with AI may be performing real technical work without starting a company or filing a patent. In fact, over 5.6 million AI-related projects are currently hosted across GitHub and Hugging Face. Nevertheless, the data still indicates that while the front door to innovation may have been widened, the pathways to formal inventive output remain far less democratic. Until the inputs that are needed for innovative output are more equitably available, the distribution of talent will fall short of allocative efficiency. Therefore, the next step is to turn wider exposure into durable pathways for participation. The institutions that shape education and innovation must ensure that opportunity is determined less by family background and more by talent and potential.
Sources:
Chetty’s Lost Einsteins Research: https://opportunityinsights.org/paper/losteinsteins/
Einstein’s Upbringing Background: https://pubmed.ncbi.nlm.nih.gov/9657287/
Eric Von Hippel’s Book: https://direct.mit.edu/books/book/2821/Democratizing-Innovation
Mark Cuban AI Comments: https://www.axios.com/2025/10/02/mark-cuban-ai-great-democratizer
Code.org Data: https://advocacy.code.org/state-of-cs/
CollegeBoard Data: https://apstudents.collegeboard.org/about-ap-scores/score-distributions/ap-computer-science-principles
Girls Who Code Data: https://girlswhocode.com/2024report/
Kauffman Foundation Data: https://www.kauffman.org/wp-content/uploads/2023/06/Access-to-Capital-for-Entrepreneurs-Report-2-June-2023.pdf
Startup Founder Data: https://startupcoalition.substack.com/p/a-missed-opportunity
Education Production Function: https://econ4everyone.uchicago.edu/video/the-economics-of-education/
GitHub and Hugging Face Data: https://hai.stanford.edu/ai-index/2026-ai-index-report/research-and-development