Survival Curves of Pandemic-Era Startups: One-Year Share Fell From 80.8% to 76.9% After Relief

New establishments do not fail on a single day. They fail on a clock. The Bureau of Labor Statistics Business Employment Dynamics (BED) age series watches each March birth cohort and asks a blunt question: what share still reports positive employment one year later, two years later, five years later? That one-year survival share is the cleanest early signal we have for whether a vintage of startups is sturdier or more brittle than the ones that came before.

This desk’s question is narrower than “did the pandemic kill small business?” It is: once Paycheck Protection Program (PPP) draws and Employee Retention Credit (ERC) claims stopped propping the payroll tape, how did first-year survival for 2020–21 cohorts compare with pre-pandemic vintages — and what happened to the 2022 cohort that opened into the wind-down? Using BLS The Economics Daily’s division table of 1-year establishment survival through 2022 births, the national desk mean (unweighted across nine Census divisions) prints 80.8% for 2020 births, then 78.5% for 2021 and 76.9% for 2022. The 2017–19 baseline sat near 79.0%. Relief-era births cleared the first year; post-relief births did not.

The interactive dashboard above walks cohort survival curves, the national one-year path with relief markers, division deltas, a 2020×2022 scatter, and a sector ladder. The rest of this post is the narrative behind those panels.

What “one-year survival” actually counts

BED age data are March-to-March. An establishment’s birth year is the year it first reports positive employment. Survival to age one means it still has positive employment in the March one year later. The series is establishment-based, not firm-based: a multi-unit firm can spawn several establishments, and breakouts inherit predecessor birth dates. That is the right object for “how many new locations made it through year one,” and a slightly different object from Census Business Dynamics Statistics (BDS) firm shutdown rates.

The TED table that underpins this post reports 1-year survival rates by Census division for birth years 1994–2022. We take those disclosed cells as given. The national path in the dashboard is an unweighted mean of the nine divisions — a desk estimate, not an employment-weighted BLS national print. Weighting by births would tilt the series toward populous divisions; the unweighted mean treats New England and the Pacific as equal geographic experiments. Both approaches are defensible; we choose the transparent one.

Ages two through five on the curve panel are modeled. We take each cohort’s observed (desk-mean) age-one rate and apply historical BED conditional survival ratios from cohorts with complete follow-up — roughly 0.86 of age-one survivors reach age two, then 0.88, 0.90, and 0.91 for the next steps. That preserves the classic downward bow of entrepreneurship Chart 3 without pretending we re-downloaded every cell of Table 7 under BLS bot restrictions. Modeled ages are scenario geometry, not a second disclosure.

The relief window elevated the 2020 vintage

Eight of nine divisions printed a higher 1-year survival rate for 2020 births than for 2019 births. East North Central jumped from 78.4% to 82.1%. West South Central rose from 79.3% to 81.5%. Mountain climbed from 77.7% to 80.9%. Only the Pacific — already strong at 82.9% in 2019 — stayed roughly flat at 81.6%. The national desk mean moved from 79.0% (2019) to 80.8% (2020), about +1.8 percentage points above the 2017–19 baseline.

That lift is not a miracle of business models. It coincides with the CARES Act window: PPP first-draw liquidity in spring–summer 2020, a second draw into 2021, and ERC claims that kept many thin-margin payrolls solvent through the acute shutdowns. Establishments that would have closed for lack of cash could stay on the QCEW tape long enough to print a March employment report. Survival here is employment continuity, not profitability, founder wellbeing, or product-market fit.

Census BDS firm-age tables tell a related story at the firm layer: young-firm exit rates fell in the acute pandemic years when forbearance and transfers were thickest, then normalized as programs closed. We do not micro-link BDS firms to BED establishments in this post. The two programs answer adjacent questions — firm shutdowns versus establishment employment survival — and both point the same direction for 2020.

Post-relief cohorts slipped below the pre-pandemic baseline

The 2021 birth cohort is transitional. National desk-mean one-year survival fell to 78.5% — already under the 2017–19 baseline — even while the Pacific printed an extraordinary 84.6%, the highest cell in the recent TED window. Strip that Pacific outlier and the other eight divisions average closer to 77.8%. West North Central dropped to 75.7%; South Atlantic to 76.7%. Second-draw PPP and late ERC claims were still in the system for parts of 2021, but the composition of births had shifted toward a larger application boom and a noisier mix of thin entities.

By 2022 births, the relief story is mostly over. The national desk mean lands at 76.9% — −3.9 pp versus the 2020 relief peak and −2.1 pp versus the 2017–19 baseline. Mountain prints the softest cell at 74.4%. West North Central sits at 75.6%. Even the Pacific, which had looked invincible in 2021, falls back to 77.1%. Eight of nine divisions show a negative 2022-minus-2020 delta. The dashboard’s horizontal bar panel and diagonal scatter make that geography visible without a speech.

The table is the whole argument in six rows. Survival did not collapse to Great Recession lows. It gave back the relief premium and then some. That is exactly what you would expect if transfers delayed exits rather than permanently rewriting hazard rates.

Survival curves, not just a single print

A one-year rate is a snapshot. The curve panel stacks ages 0–5 for focus cohorts. Age zero is always 100% by construction. Age one is the TED-based desk mean. Ages two through five bend with historical conditional ratios. On that geometry, the 2020 curve starts higher at age one and stays above the 2018 and 2019 curves through the early ages — a parallel shift more than a shape change. The 2022 curve starts lower and stays lower. If conditional hazards re-steepened after relief (a real risk we cannot fully observe yet for the youngest cohorts), the modeled gap would understate the long-run damage; if hazards mean-reverted, the gap would narrow.

That uncertainty is why the headline stays on one-year survival share. It is disclosed sooner, comparable across vintages, and less dependent on the modeling layer. Five-year survival for 2020 births will not be fully stamped until later BED releases; treat the outer curve as a planning sketch.

Geography: relief was broad, the unwind was broader

Division deltas matter because national averages hide who got the soft landing. East North Central’s +3.7 pp 2020-over-2019 lift was one of the largest relief bumps; its −4.8 pp 2022-versus-2020 drop is among the sharpest givebacks. Mountain’s path — 77.7 → 80.9 → 77.2 → 74.4 — is a miniature of the national story with a steeper final step. Middle Atlantic moved less dramatically but still finished 2022 below 2020.

Pacific is the stubborn exception on the way up and the reminder that regional booms can dominate a single year. Its 84.6% 2021 print is a genuine TED cell, not a desk invention. It may reflect sector mix (tech and professional services establishments), delayed closures, or classification quirks. It does not overturn the national path: by 2022 the Pacific had rejoined the softer pack.

Sector color without overclaiming

The sector ladder is modeled. We anchor the all-private row to the TED desk mean, then spread historical young-establishment exit differentials around it — construction and accommodation & food softer, manufacturing and health care sturdier. Under that sketch, accommodation & food shows the largest relief-to-post-relief drop: a big 2020 lift from dining and lodging support, then a sharp 2022 giveback as transfers faded and labor costs stayed high. Professional services and health care move less. Treat those bars as hypothesis charts, not NAICS microdata. The geographic TED table is the hard evidence; the sector panel is interpretive shading.

Caveats the desk will not sand away

First, establishment ≠ firm ≠ EIN application. A surge in business applications can raise births without raising durable employers. Survival rates condition on establishments that already appear with positive employment; they say nothing about the large pool of non-employing entities.

Second, selection. If relief kept weaker establishments alive into the age-one window, the 2020 survivor pool may be compositionally softer than a pre-pandemic survivor pool at the same age. That would bias multi-year curves down later even if age-one rates looked strong — another reason to keep the modeled ages labeled.

Third, measurement. March employment is a point-in-time concept. Seasonal businesses, multi-establishment reorganizations, and predecessor inheritance in breakouts can move establishments across categories without a “death” in the everyday sense.

Fourth, causal attribution. We align survival prints with PPP/ERC timelines. We do not estimate a treatment effect with a control group of unassisted establishments. Other forces — demand reopening, labor shortages, rent, inflation, and interest rates — also hit 2021–22 births. The coincidence is strong; the identification is not.

Fifth, national construction. Unweighted division means differ from a birth-weighted national rate. If you need the official BLS national Table 7 print for a briefing book, pull it directly; our path is built for transparent geography.

What the desk takes to the briefing

For one-line briefings: the one-year survival share for U.S. private establishments rose to about 80.8% for 2020 births during the relief window, then fell to about 76.9% for 2022 births — below the ~79% pre-pandemic baseline — with eight of nine Census divisions giving back the 2020 lift. That is not a story of total collapse. It is a story of transfers delaying exit, then hazard rates reasserting themselves once the programs closed. Operators and lenders who underwrote 2020–21 vintages on the elevated survival print should re-benchmark to the post-relief path. Policymakers who treat the 2020 spike as a permanent improvement in startup durability are reading a liquidity episode as a productivity episode.

The dashboard’s controls let you flip survival versus exit, filter eras, toggle cohorts on the curve, and isolate a division. Use them. The table above is the spine; the maps of geography and the modeled curves are the ribs.