Startups can be counted two ways, and for three decades it did not matter which you chose. New firms as a share of all businesses and the jobs they bring as a share of all employment fell together from the early 1980s through 2012, the decline in business formation that Benjamin Pugsley and Ayşegül Şahin documented and connected to slower employment growth and weaker recoveries.

Line chart of United States startup activity from 1979 to 2023, on two vertical axes. The teal line, read on the left axis, is startups as a percent of all employer firms; the salmon line, read on the right axis, is their percent of total employment. The two rise and fall together on a declining trend from the early 1980s through about 2012, from 12.8 and 3.2 percent to roughly 8 and 2 percent. After about 2015 they diverge: the teal line climbs to 9.1 percent by 2023, its steepest gains around the pandemic, while the salmon line stays near 1.8 percent, close to its lowest reading in the series. A dashed vertical line at 2012 marks the end of the data in the 2019 paper, and grey bands mark recessions.

The two have come apart since about 2015. Startups (teal) have gone from 7.6% of all employer firms in 2010 to 9.1% in 2023, recovering roughly a third of their earlier decline, with the steepest gains around the pandemic. Their share of employment (salmon) has not moved with them: at 1.8% it sits close to its lowest reading in the whole series and essentially where it was a decade ago.

Whatever is generating more new businesses is generating notably smaller ones.