By Dr. Swati Bhatt, Ph.D., Lecturer of Economics, Princeton University
Wayne Gretzky, often referred to as “The Great One”, is widely considered the greatest ice-hockey player of all time. His career included playing for the Edmonton Oilers (1979-1988), the Los Angeles Kings (1988-1996) and finally the New York Rangers (1996-1999). His skill, vision and leadership on the ice made him an iconic figure in the sport of ice hockey. He is famous for saying, “You miss 100% of the shots you don’t take”, emphasizing the importance of seizing opportunities and being proactive. Gretzky’s focus on creating opportunities for others, rather than merely scoring goals himself, aligns with the mindset required in entrepreneurship. (Sam Walker pointed out that great teams are built around a culture—one that values discipline, shared responsibility and teamwork. Great captains inspire trust and harmony in their teams.)
Introduction: Where are the shots being taken?
Taking shots or risks depends largely on the environment or entrepreneurial culture. The socio-cultural mindset, resource accessibility and prevailing attitudes toward risk, driven by advancements in artificial intelligence technologies (AIT), encourage entrepreneurs to experiment and innovate rapidly. Strong belief in the possibilities offered by AIT can “foster a risk-taking, entrepreneurial culture…. It is a belief system where risk-taking is applauded and even encouraged.” AIT enables startups to achieve tasks traditionally dominated by large corporations, such as analyzing massive datasets and automating complex processes. This levels the playing field, allowing small ventures to compete globally.
It is widely believed that such communities emerge through the clustering of startup companies in specific areas, where neighborhood effects play a crucial role in fostering a risk-taking culture. Witnessing peers take significant risks and succeed reinforces a mindset that such risks are worthwhile, promoting a culture of experimentation and calculated risk-taking. Innovation hubs, such as Silicon Valley, Shenzhen, New York and Bangalore (Bengaluru), exemplify this phenomenon. However, in recent years, startup activity has dispersed geographically, spurred by the adoption of AIT and the shift to remote and hybrid work in the post-pandemic years. The decentralization of innovation has enabled smaller cities, rural areas and international regions to develop startup clusters supported by mentorship and guidance networks.
This article examines the geographical implications for entrepreneurship stemming from the technological and organizational changes of the past decade. Do traditional hubs still dominate in idea generation and implementation? Here, I analyze the state-by-state configuration of startup activity in the United States over the past decade.
Entrepreneurship and productivity
Why does entrepreneurial activity matter? At the individual level, it provides access to the American Dream. But how does it benefit society and the economy?
Innovation lies at the heart of productivity increases as new technologies encourage businesses to adopt new operational styles. For example, restaurants today use QR (quick-response) codes instead of paper menus; manufacturing firms leverage advanced analytics and artificial intelligence to manage inventories; and office workers use AIT to streamline processes such as data entry, scheduling and email management. AI algorithms also analyze user preferences to deliver personalized product recommendations. According to PitchBook, AI startups accounted for 22 percent of first-time venture capital (VC) financing in 2024. (However, this space is becoming crowded as every startup claims to be an AI organization. Perhaps the AI keyword is a teaser to generate interest.)
Startups and new businesses can more easily incorporate new technologies, as legacy frameworks do not constrain them. Consequently, an increase in startups often leads to higher labor productivity and potentially higher wages. Measured as output per labor hour, productivity directly impacts wages and the overall standard of living.
A key question in examining innovation is whether an idea becomes significant only when implemented or if the birth of the idea itself qualifies as innovation. Are startups defined by the establishment of wage-paying firms or by business-formation applications? The influence of startups on labor productivity likely follows employment. Figure 1 below shows annualized nonfarm labor productivity in the US over the years 2010 to 2023. Figure 2 depicts the phenomenal growth in startup activity. Across the US, new business applications increased by 125 percent from 2010 to 2023. Figure 3 represents new business growth of more than 85 percent between 2010 and 2023 in professional and technical services.
The contrasting geography of firm births versus ideas
I, Dr. Swati Bhatt (2023), analyzed US startups from 1994 to 2022 and identified a sharp increase in entrepreneurial activity from 2010 to 2022 compared to the previous decade. States west of the Mississippi River dominated the top 10 performers, including North Dakota, Nebraska, Missouri, Idaho, California, Texas and Washington. Conversely, the 15 weakest performing states were located east of the Mississippi.
My 2023 results are based on wage-paying firms’ births, as measured by the U.S. Bureau of Labor Statistics’s (BLS) survey data. (The BLS’s Handbook of Methods states that births “are establishments with positive third-month employment for the first time in the current quarter and with no links to the previous quarter, or units with positive third-month employment in the current quarter and zero employment in the third month of the previous four quarters.”)
However, recent research suggests that startups are contributing less to employment growth and are smaller in size. (According to Kim et al., “The share of employment among young startups associated with high-growth firms fell from just under 10% in 1985 to below 6% in 2020.”) Assessing startups by idea generation might provide a more comprehensive view than focusing solely on employment metrics. Idea generation can be gauged using the U.S. Census Bureau’s Business Formation Statistics (BFS), which collects data on business applications. Census data often precedes BLS data, as tax registration occurs before employment creation. Thus, Census data highlights idea formation, while BLS data reflects idea implementation.
Using the U.S. Census Bureau’s data, I analyzed the geographical distribution of entrepreneurial activity across the US in recent years. (U.S. Census Bureau data defines startups based on administrative data from the U.S. Department of the Treasury. “The data on business applications are based on applications for an Employer Identification Number [EIN] through filings of IRS Form SS-4.”)
Figure 5 illustrates states ranked by business applications from 2010 to 2023, and Figure 6 shows the corresponding ranking for startups based on employment. (States are ranked by the trend growth rates of applications over the years 2010-2023, normalized by the year 2010.)
There are three key takeaways:
Wyoming leads in growth: Wyoming ranks highest, with a trend growth rate of more than 4 percent per quarter from 2010 to 2023. [Figure 4]
Activity spreading to Southeastern states: Startup activity is rising in Southeastern states—such as South Carolina (1.7 percent per quarter), Georgia (1.43 percent), Mississippi (1.43 percent), North Carolina (1.28 percent) and Tennessee (1.24 percent)—potentially driven by post-pandemic government funding. [Figure 5]
Mismatch between idea and employment hubs: There is a significant disparity between idea-generating states, shown in Figure 5, and employment-generating states, shown in Figure 6. Montana, for example, displays an establishment birth rate of 1.5 percent per quarter. Only South Carolina and Tennessee appear in both rankings, with startup birth rates of around 1 percent per quarter.
AIT-driven startups tend to be capital- and skill-intensive, requiring fewer workers. Additionally, founders’ ideas are often acquired by larger firms, reducing the need for independent growth. For example, Microsoft acquired all of Inflection AI and its entire 70-person workforce in 2024, and in 2022, it acquired Nuance Communications, a leader in speech recognition and conversational AI.
Demographic patterns explain high startup rates
Who were the founders of startups across the US from 2010-2023? Based on Census data, the elasticity of new business formation with respect to all races is positive—the Hispanic population’s share is the highest, followed by blacks, whites and Asians. (All elasticities are positive and significant at the 5 percent level.) This finding partially aligns with Dinlersoz et al. (2023), who examined the pre-pandemic years 2011-2016 across the US using Census data and found positive correlations between the African-American population’s share and new business applications but negative correlations for Hispanics’ and Asian-Americans’ shares.
By contrast, in the top 10 states shown in Figure 5, the white population’s share has the highest elasticity, followed by Hispanics and blacks.
Interestingly, the elasticity with respect to the Asian population’s share is negative.
Recent research
Decker and Haltiwanger (2024), using Census data over the years 1993-2023, found that “while business applications have remained elevated above the pre-pandemic pace, a steady cooling is evident since the fall of 2023”. They attribute this slowdown to an increase in establishment closures “since failure rates are high among young businesses”.
Dinlersoz, Dogan and Zolas (2024) discovered that “AI business applications constitute a very small fraction (0.035%) of the 62+ million business applications … as of late 2023”, but their “classification method is subject to potential selection issues associated with reporting AI-related terms”. (The authors examined U.S. Census Bureau data on applications for business formation in the AI sector over 2004-23. They used “text analysis on an applicant’s reported main business line, business name, and trade name in conjunction to identify business ideas for developing or providing AI technologies and applications”.)
Conclusion
Media attention has focused on large projects, such as the formation of chip-manufacturing facilities or foundries, but these projects could create more than 10,000 jobs in the construction industry and also more jobs in ancillary businesses. For example, Intel is building its foundry in Lorain, Ohio, located near the shores of Lake Erie at the entrance to the Black River. Water is a critical resource and is important for chip manufacturing. The physical infrastructure needed, called a superload, will involve “a bunch of trucks, a lot of permits, all these dudes—from cable companies, power companies, hauling companies—plus a handful of local cops and highway patrol officers,” according to Kelly McEvers.
Excitement about AIT’s possibilities has driven a surge in business applications at the Internal Revenue Service (IRS), located within the U.S. Department of Treasury. These applications are dispersed across the US, particularly in the Southeastern states, rather than concentrated in traditional hubs, such as California or New York. Wyoming leads with more than 85 percent growth in new business applications over the past decade. High-ranking states also include North Carolina, South Carolina, Georgia, Mississippi and Tennessee. No clear racial or ethnic patterns fully explain these results.
This geographic shift may influence future migration patterns, reshaping the population distribution across the US.
Endnote: This article has benefited from data analysis completed by Patrick Chen, Princeton Class of 2027.
Data sources
PitchBook: AI Founder Frenzy.
U.S. Bureau of Labor Statistics (BLS): Business Employment Dynamics (BED).
United States Census Bureau: Business Formation Statistics (BFS) – Methodology.
United States Census Bureau: State Population by Characteristics: 2020-2024, December 2024.