The Hidden Churn

The economist Joseph Schumpeter coined the phrase “creative destruction” in 1942 to describe capitalism’s essential mechanism: the constant dismantling of old economic structures by new ones. He was theorizing. Eighty years later, we can measure it. The Bureau of Labor Statistics’ Business Dynamics Statistics program — the BDS — has been tracking every private-sector establishment in America since the early 1990s, publishing quarterly data on how many jobs are created and destroyed across the economy. The dataset covers 25 years of continuous measurement, from 2000 through the latest available quarter of 2024, and it tells a story that the monthly jobs report was never designed to reveal.

Consider a concrete example. In the fourth quarter of 2024, the U.S. private sector created 7,802,000 jobs and destroyed 7,496,000. The net — 306,000 — is the kind of number that appears in a headline: “Economy adds 306K jobs.” Accurate, but incomplete. It’s like reporting on an ocean by measuring the tide. The tide tells you whether the water level is up or down; it tells you nothing about the currents, the waves, the undertow, the thermal layers, or the marine life. The 7.8 million created and 7.5 million destroyed are the ocean. The 306,000 net is the tide.

The distinction matters because net job growth can mean very different things depending on the underlying flows. An economy that creates 3 million and destroys 2.5 million has the same net gain as one that creates 8 million and destroys 7.5 million — but the second economy is radically more dynamic, with far more opportunity for workers, far more risk for firms, and far more innovation driving the reallocation. The BDS lets us see which economy we actually have. Spoiler: it’s the high-churn one.

The data source deserves a brief explanation, because it’s one of the least-appreciated statistical programs in the federal government. The BDS is derived from the Quarterly Census of Employment and Wages (QCEW), which captures virtually every private-sector establishment in the United States — roughly 10 million of them — through mandatory unemployment insurance filings. This is not a survey. It’s a census. When the BDS reports that 7.8 million jobs were created in a quarter, that number isn’t extrapolated from a sample; it’s built from actual payroll records filed by actual businesses with state unemployment insurance offices. It is, in practical terms, a complete accounting of the American private economy.

The chart above is the single most important visualization in this series. Two thick lines — gains and losses — march through 25 years roughly in parallel, each measuring between 25 and 35 million jobs per year. The thin green-red ribbon between them is the net change: the number that dominates economic discourse. That ribbon is almost invisible relative to the flows that produce it. In a good year like 2014, the economy gained 29.3 million and lost 26.3 million, netting 2.9 million. In a terrible year like 2009, it gained 25.4 million and lost 30.9 million, netting −5.5 million. The gross flows barely changed — gains fell about 14% from peak — but the net swung from strongly positive to deeply negative. Recessions, in this light, are not events where job creation stops. They are events where destruction slightly outpaces creation.

This is counterintuitive. Most people imagine recessions as periods where the economy grinds to a halt — where businesses stop hiring and start firing en masse. The BDS shows something more subtle and more interesting: even in 2009, the worst year of the Great Recession, the private sector created 25.4 million jobs. That’s only 15% fewer than in the boom year of 2006. The crisis was not that creation collapsed; it was that destruction surged. Losses jumped from about 29 million in normal years to 31 million in 2009 — only a 7% increase — but that modest increase was enough to produce the worst net job loss since the program began. The system is so finely balanced that small shifts in either direction produce large net consequences. It’s an economy running on razor-thin margins.

Where do 7.8 million new jobs come from in a single quarter? The BDS breaks the answer into two channels: expansions and openings. Expansions are existing businesses adding headcount — a restaurant hiring a second line cook, a tech firm adding twenty engineers, a hospital opening a new wing. Openings are new establishments that didn’t exist in the prior quarter — a franchise location opening its doors, a startup making its first hires, a subsidiary launched in a new city. The distinction matters because expansions represent growth within the existing business base, while openings represent entirely new economic activity.

The split is remarkably stable. In 2024, expansions accounted for 79.5% of all job gains and openings accounted for 20.5%. That ratio has barely moved in 25 years — it was 79.3% expansions in the year 2000. The implication is important: four out of five new jobs come from businesses that already exist. The popular narrative that startups and new businesses drive job growth is not exactly wrong, but it dramatically overstates the startup contribution. The vast majority of job creation is boring, incremental, and invisible — existing firms adding a few positions at a time, quarter after quarter, across millions of establishments.

The loss side mirrors the gain side. Contractions — existing businesses shrinking their payrolls — account for roughly 80% of all job losses. Closings — establishments that cease to exist entirely — account for the remaining 20%. Again, remarkably stable over time. The economy creates and destroys jobs overwhelmingly through incremental adjustments at existing businesses, not through dramatic births and deaths. The Blockbuster-to-Netflix story is real and important, but it accounts for about a fifth of the action. The other four-fifths is a dentist’s office adding a hygienist, a trucking company releasing two drivers, a grocery store hiring seasonal staff, a law firm cutting a paralegal.

Within the 20% that does involve births and deaths, there is a smaller subset: establishment births (a subset of openings that represents genuinely new businesses) and establishment deaths (a subset of closings where the business permanently ceases operations). In 2024, establishment births contributed 3.9 million jobs while deaths subtracted 1.7 million — though the death figure is incomplete for recent quarters because it takes time to confirm a business has truly closed rather than merely paused. The birth-death dynamic will be the focus of later episodes, but the key point here is that it’s a sub-plot within the larger story, not the headline.

No event in the 25-year history of the BDS produced anything remotely comparable to the second quarter of 2020. In those three months — April, May, and June — the private sector created 5.7 million jobs and destroyed 20.2 million. The net loss of 14.5 million jobs in a single quarter was more than three times the total annual loss of the Great Recession. If the BDS normally shows two roughly parallel lines, Q2 2020 showed a line that split in two: gains plummeted while losses exploded to levels the data had never recorded.

But the other half of the pandemic story is equally striking. In Q3 2020 — the very next quarter — the economy created 10.9 million jobs and destroyed only 6.8 million, producing a net gain of 4.0 million. That single quarter of gains nearly matched the total annual gain in any pre-pandemic year. The economy didn’t recover gradually over years, as it did after 2009; it snapped back with a velocity that the system had never demonstrated before. The V-shape that economists debated in real time is visible in the BDS data with perfect clarity: a cliff in Q2, a rocket in Q3.

Leisure and hospitality bore the brunt. In Q2 2020, the sector lost a net 5.7 million jobs — more than the total private sector lost in all of 2009. Restaurants shut. Hotels emptied. Theme parks closed. Within that one industry, 6.3 million jobs were destroyed while only 655,000 were created — a loss-to-gain ratio of nearly 10 to 1. Compare this to a normal quarter where gains and losses roughly balance, and the pandemic’s selective cruelty becomes apparent: it didn’t hit the whole economy equally. It hit the parts of the economy that require human proximity — and it hit them with a force that was literally off the chart.

Professional and business services lost 1.8 million net. Retail lost 1.6 million. Education and health services lost 1.7 million — a bitter irony for a health crisis. Manufacturing lost 874,000. Even construction, which would later boom, lost 488,000 in that single quarter. The only sector that barely flinched was utilities, which lost a net 4,000 jobs. Power plants don’t close for a pandemic.

Not all industries churn equally. Some are perpetual motion machines — creating and destroying jobs at furious rates, with the net being a rounding error on the gross. Others are slow, incremental growers or shrinkers where the flows are modest and the net is a large fraction of the total. The table below shows the industry breakdown for 2024, ranked by total gross churn (gains plus losses), with the churn-to-net ratio revealing how much hidden activity underlies each sector’s headline number.

Professional and business services is the churn champion: 5.8 million jobs gained and 5.9 million lost in 2024, for a net of negative 118,000. The total churn was 11.7 million for a result that rounds to zero. This is the staffing-agency effect — temporary workers, consultants, contract employees cycling through placements at extraordinary velocity. A temp who works three months at one firm, moves to another for six months, and takes a month off counts as multiple gains and losses in the BDS. The sector’s churn-to-net ratio of 99:1 is the highest of any major industry. For every net job “lost” in professional services in 2024, 99 others were created and destroyed within the sector.

Leisure and hospitality is the second-highest churner: 5.6 million gained and 5.5 million lost. This is seasonal hiring in its purest form — beach resorts staffing up for summer and releasing workers in fall, ski lodges doing the reverse, restaurants cycling through line cooks at rates that would terrify any other industry. The sector’s churn-to-net ratio of 79:1 reflects an industry where individual job tenure is short but the aggregate flow is enormous.

At the other end of the spectrum, education and health services had the lowest churn-to-net ratio at 9.7:1 — meaning that a relatively large fraction of its gross flows translated into actual net gains. The sector gained 5.0 million and lost 4.0 million, netting 924,000 — the largest net gain of any industry in 2024. This is the healthcare hiring machine, driven by aging demographics and the insatiable demand for nurses, home health aides, and clinic staff. Education and health creates jobs the way a tide comes in: steadily, broadly, and without much churn relative to the gain.

Information stands out as the sector with the most decisive net loss: 567,000 gained and 633,000 lost, netting −66,000 with a relatively low churn ratio of 18:1. This is the tech layoff story of 2023-2024 showing up in the data — Meta, Google, Amazon, and dozens of smaller firms cutting headcount while hiring slowed across the industry. When the information sector sheds jobs, the layoffs are real, permanent, and highly visible in a way that professional-services churn is not.

The existence of this hidden churn has implications that go well beyond statistical curiosity. It reframes three foundational questions about the American economy.

First, it changes what “job creation” means. When a politician claims credit for creating 2 million jobs, the reality is that the economy created roughly 30 million and destroyed 28 million in that year. The 2-million net was not the result of anyone’s policy; it was the residual of an autonomous churning process that is largely beyond any administration’s control. The gains come from millions of individual decisions by millions of business owners — hiring a cashier, adding a shift, opening a second location. The losses come from equally distributed decisions — laying off a driver, automating a process, shutting a branch. The net is an emergent property of the system, not a managed outcome.

Second, it changes what recessions are. The popular image of a recession is an economy that stops creating jobs. The BDS shows that recessions are not about creation stopping — creation merely dips, usually by 10-15%. Recessions are about destruction surging. In 2009, gains fell 16% below their 2006 peak, but losses jumped 6% above the 2006 level. The difference between a boom and a bust is not the difference between creating jobs and not creating jobs; it’s a few percentage points of shift in the balance between two enormous, largely independent flows. The economy in a recession is still furiously creating; it’s just destroying slightly faster.

Third, it changes what economic dynamism means. The rate data — gains and losses as a percentage of total employment — reveals a troubling trend that will be the subject of our next episode. In 2000, the quarterly job gain rate was 7.8% of employment. By 2024, it had fallen to 5.9%. The loss rate declined from 7.4% to 5.7%. Both declined in parallel, meaning the economy is becoming less dynamic — less creation, less destruction, less reallocation. The American economy is still a churn machine, but it is a churn machine that is gradually slowing down. Whether this represents maturation, ossification, or something else entirely is one of the most important economic questions of the next decade.

The chart above shows the churn-to-net ratio for each year since 2000. In a “decisive” year like 2009 or 2020, when the net was large (negative), the ratio drops toward 8-10x — still extraordinary, but lower because the net was meaningful relative to gross. In a “balanced” year like 2003 or 2024, when the net was nearly zero, the ratio soars — hitting 1,033:1 in 2003 and 71:1 in 2024. These are years where the economy churned furiously and arrived back roughly where it started. Sixty million jobs created and destroyed, net result: practically nothing. The machine ran all year long. It just happened to run in circles.

The ratio is not a quality metric — a high ratio doesn’t mean the economy is “bad” any more than a low ratio means it’s “good.” But it does measure something important: how misleading the net number is as a description of underlying reality. In 2024, the ratio of 71:1 means that the net job change captured about 1.4% of the total economic activity that produced it. The other 98.6% was invisible to anyone who only read the headline.