Young Americans and the Business Formation Surge

This analysis asks a simple question: are Americans in their 20s and early 30s forming businesses at higher rates than in the past? To answer it, we look at two things. First, the stock of self-employed workers in each age group as a share of all employed workers — a snapshot of who is self-employed right now. Second, the flow into self-employment: how many workers transition from wage employment into self-employment each month and year.

We organize the data into age bands (20–34, 35–44, 45–54, 55–64, and 20–64 overall). These are not strict generational boundaries; the goal is to track how self-employment patterns differ across life stages, not to make claims about Millennials or Gen Z as named generations. Where the data show interesting variation within the 20–34 band, we zoom in on the 20–29 group specifically.

Key Findings

1   The Stock Recovery: All Cohorts Above Pre-COVID Baseline

The share of employed workers who are self-employed fell steadily from the early 2000s through 2019 across every age group. The COVID-19 pandemic and its aftermath reversed that trend. As of 2024–2025, every cohort is above its 2016–2019 average and several are approaching or exceeding their long-run 2005–2019 mean.

Age 20–34: A Decade-Long Decline Reversed

The youngest cohort tells the clearest story. Their self-employment share declined from a peak near 5.5% in the early 2000s to a trough around 4.5% in the late 2010s, a 15-year erosion. Post-COVID, that decline has fully reversed. The 2023–2025 annual averages are above the crimson 2016–2019 reference line and tracking near the gray 2005–2019 long-run mean. This is the first sustained period above the pre-COVID baseline since the dot-com era. Within this cohort, it is the youngest workers, those 20–29, who show the sharpest elevation relative to their own historical baseline, suggesting they are the primary driver of the 20–34 signal.

Ages 35–64: Recovery, but From Lower Levels

Older cohorts saw a steeper long decline in self-employment share: 3–5 percentage points from 2004 to 2019 in the 45–54 and 55–64 groups. The post-COVID recovery is real but incomplete. All cohorts are above the 2016–2019 line, but still well below their 2004 peaks. The 55–64 group has recovered the most ground in absolute terms; the 35–44 group the least.

2   Opportunity, Not Necessity: The 20–34 Entry Pathway

Not all self-employment entry is equal. Workers pushed into self-employment by unemployment tend to form less durable, lower-quality businesses than those who leave stable employment by choice. The pathway into self-employment matters for interpreting the post-COVID picture.

To distinguish the two, we compute separate quarterly transition rates from the CPS rotation structure:

• Employment to self-employment: share of employed wage-and-salary workers at month T who are self-employed at month T+1
• Unemployment to self-employment: share of unemployed workers at month T who are self-employed at month T+1

Each panel shows the raw transition rate alongside the shaded 2005–2019 historical range (mean plus or minus one standard deviation). The y-axis is held constant across all four age groups so that absolute levels are directly comparable.

The employment channel is the cleaner signal

The 20–34 age group has a lower base rate of transitioning from employment to self-employment than older cohorts, which is not surprising given that older workers tend to have more capital, professional networks, and industry experience to draw on. What stands out is that the 20–34 rate has been running at or above the top of its historical range for much of the post-COVID period. That kind of persistent elevation is not visible in the same way for the 35–44, 45–54, or 55–64 groups, which look broadly flat relative to their own historical ranges.

The unemployment-to-self-employment channel is harder to read, for two reasons. First, unemployment has remained low since 2022, which means there are fewer unemployed workers in any given CPS rotation cohort to begin with. Fewer people in the denominator means wider swings in the rate even when nothing fundamental has changed. Second, the CPS rotation structure compounds this: only a fraction of respondents are in their matched month in any given quarter, further shrinking the pool. The wide historical bands in Figure 2b reflect both of these constraints.

With that caveat, the 55–64 panel is worth noting. It shows more post-COVID activity above its historical range than the other groups -- a tentative pattern that would be consistent with older workers displaced from jobs entering self-employment as an alternative to continued unemployment. This is the opposite of what we see for 20–34, and it is the kind of contrast that makes the opportunity story for younger workers more credible: the two age groups appear to be responding to different pressures. We hold this observation loosely given the noise, but it is worth watching.

Taken together, the pattern is consistent with opportunity-driven entry among younger workers: employed people with other options choosing self-employment, not unemployed people falling back on it.

3   Self-Employment Is Not Just a Cyclical Response

One natural question: is the post-COVID self-employment recovery simply a function of a tight labor market? The scatter plots below show that for most age groups, the historical relationship between the unemployment rate and the self-employment share is weak. None of the correlation coefficients are particularly strong, and the recent observations tend to cluster above the baseline trend regardless of age group, suggesting that the current self-employment share is higher than unemployment conditions alone would predict.

For the 20–34 cohort, the historical relationship between self-employment share and unemployment is essentially flat (r = −0.09, β = −0.09) — self-employment participation in this age group was never strongly cyclical. The recent crimson points cluster above the OLS baseline, meaning the current self-employment share is higher than the historical unemployment rate alone would predict.

Older cohorts show the expected positive correlation (higher unemployment tends to push up necessity self-employment), but even there, the recent post-COVID points do not fall neatly on the baseline relationship in a way that would make the recovery look purely mechanical.

4   Flow Rates: Entry Is Near Historical Norms

The stock analysis above measures who is self-employed at a point in time. The flow analysis measures who enters self-employment. We use two methods from the CPS rotation structure:

• Month-over-month: tracks person transitions between consecutive survey months. Captures rapid and gig-economy entry; higher volatility.
• Year-over-year: matches the same person at MISH 4 and MISH 8 (roughly 12 months apart). Captures durable business formation; lower volatility but carries a roughly 15-month publication lag.

Year-over-year entry rates are broadly at or near historical baseline for all cohorts. There is no dramatic spike in durable new business formation. The elevated stock is better explained by the compounding effect of several years of flows that are modestly above (or at) their pre-COVID trough, combined with solid persistence rates (the fraction of self-employed workers who remain self-employed 12 months later).

Month-over-month rates for 20–34 show more individual months above their seasonal baseline, particularly in 2024, but the year-over-year series, which filters out short-lived or abandoned attempts, is more muted. Where month-over-month is elevated but year-over-year is not, the signal is workers testing self-employment rather than committing durably. Narrowing to the 20–29 sub-cohort, the year-over-year entry rate sits approximately 7% above its own 2005–2019 baseline mean, a stronger signal than the broader 20–34 group, which sits near its historical average.

Methodology