Three of every four restaurant hires are replacing someone who quit

In June 2026, restaurants, bars and hotels across the United States hired 715,000 people. In the same month, 816,000 left. Of those, 638,000 quit.

That month is the whole first half of the year in miniature. Between January and June the sector hired 4.87 million people and recorded 4.88 million separations. All of that recruiting, trialling and training, and the industry ended June fractionally smaller than it started January.

The reflex explanation is that nobody wants to work in hospitality any more. The federal data says something more useful, and more surprising: people leave restaurant jobs less often than they did before the pandemic. What has changed is what your hiring is buying.

Turnover is lower than it was in 2019

The Job Openings and Labor Turnover Survey reports separations each month as a percentage of employment. Add up twelve months and you have a rough annual turnover rate anyone can reproduce from published figures. For accommodation and food services, that number was 78.7 separations per 100 jobs in 2019. In 2025 it was 65.3.

So the sector still turns over the equivalent of around two thirds of its headcount in a year, which is punishing. But it is a smaller number than before the pandemic, and well down from 83.8 in 2022. The "turnover is over 100% and climbing" line that runs through restaurant trade content has been pointing the wrong way for three years.

But more of your hiring is now pure replacement

Divide a year of quits by a year of hires and you get the share of your recruiting that exists only to backfill someone who walked. In 2015 the sector recorded 66 quits for every 100 hires. In 2019, 72. So far in 2026, 75.

The 2020 collapse in the line is the reopening, when rehiring surged and almost nobody quit. Set that year aside and the picture is a step change after 2021 that has not come back down. Hiring volume fell faster than quitting did, so a larger share of what is left is replacement.

For a single site the arithmetic is blunt. If three of every four people you take on are backfilling a leaver, then three quarters of the job ads, trial shifts, uniform, induction paperwork and the several weeks a new server needs before they stop asking questions buy you no extra capacity at all. They buy you the capacity you had in January.

The applicant drought that defined 2022 has also eased. There were 684,000 open jobs in the sector in June 2026, against a monthly average of 1.41 million in 2022 and 875,000 in 2019. Fewer vacancies chasing the same pool means the binding constraint has moved from finding people to keeping them.

No other sector churns like this

This is why generic retention benchmarks are useless here. A 4.5% monthly quits rate compounds into something no office employer would recognise, and it is the normal operating condition of the industry rather than a symptom of a bad manager. The question worth asking is not whether your churn is high — it is — but how much of it is avoidable.

Work out your own replacement share

The national figure is only useful once you have your own to hold against it. It takes about twenty minutes with a payroll export.

What the evidence actually supports

Almost none of the retention advice written for restaurants has been tested. The strongest experiment in the adjacent world of hourly retail is the Stable Scheduling Study: 28 Gap stores in the San Francisco and Chicago metropolitan areas, randomly assigned to treatment or control, running from November 2015 to August 2016.

Stores that received the scheduling package recorded median sales 7% higher than the control stores, and labor productivity 5% higher — an additional $6.20 of revenue for every hour of labor.

Read that with its limits attached. One apparel retailer, 28 stores, a decade ago, and seven scheduling changes introduced together, so the study cannot tell you which component carried the effect. It is still considerably more evidence than anything else on offer.

Separately, panel data on 1,827 hourly retail and food-service workers found schedule instability to be a strong and robust predictor of turnover among the roughly one third with the least stable schedules. About half of that relationship ran through job satisfaction and about a quarter through conflict between work and family life.

None of this displaces pay. It sits underneath it: the evidence says an unpredictable schedule pushes people out of jobs they would otherwise keep, and predictability is one of the few levers a single site can pull this month without a budget approval.

The numbers this article leaves out

Restaurant content routinely quotes a dollar cost per lost employee. Chasing those figures back rarely reaches a published methodology, so none of them appear here. If you want a cost per quit you can defend, build it from your own records: manager hours spent recruiting and inducting, overtime or agency cover during the gap, the output difference in a new starter’s first weeks, and the waste that comes with training on live service.

The official figures have limits of their own. JOLTS accommodation and food services includes hotels and other lodging alongside restaurants and bars, so it is not a restaurant-only measure. It is national, seasonally adjusted, and June 2026 is preliminary and will be revised. It describes an industry average and nothing about your street.

That kind of check is worth doing before you act on any industry number. We did the same exercise with card processing fees, where the published averages turn out to describe a different merchant than the one reading them.