# Pubs and pub companies: the independent check

*Made on 15 October 2026, before publication, by an analyst who didn't build the piece. Every figure was worked out again
from the raw records with separate code, not by rerunning the piece's own.*

## The short version

The findings on closures and on companies going bust hold. Pub sites aren't closing any faster than they were, on both the ONS count and the councils' food
registers, and established pub companies went bust more often in 2023, 2024 and 2025 than in 2019.

The check found one real fault in the first draft, and it was in the going-bust figures. They counted some companies that
weren't insolvent: solvent liquidations, where the owners close a company that can pay its debts, and a few companies
that had only reduced their share capital or had a receiver appointed over a property. They also counted companies that
turned five during the year in the top of the rate but not the bottom. Corrected, the rise for pubs is bigger than the
draft said, and pubs' rate in 2026 so far is still above 2019 rather than back at it. The figures on the page are the
corrected ones.

The jobs figures in the headline (employees of pubs and bars in Great Britain, from ONS's Business Register and Employment
Survey) were added after this check and aren't covered by it; they're read straight from ONS's published series.

## What was checked, and how

**ONS pub and bar sites.** The UK and national counts for every March from 2010 to 2026, and the split by who runs
them, were fetched again from ONS's Nomis service on 15 October 2026. Every figure matched: 44,680 sites in 2010, 41,015
in 2013, 39,145 in 2019 and 37,350 in 2026, with companies running 26,125 in 2019 and 29,570 in 2026. The council counts
were read from the same dataset, and the count of councils with fewer sites (182 of the 303 with 50 or more) matched.

**The food registers.** Every council's archived register was compared spring to spring again, from the archived files:
which pubs closed for good, which changed hands and which were new sites. The UK figures and all thirteen nation and region
rows matched to the pub (one count differs by one, a pub listed twice in one council's file).

Then the closure claim was tested against the cuts most likely to break it. With the same councils in both periods, 17.7%
closed for good from 2013 to 2019 and 19.2% from 2019 to 2026, which is less per year after 2019 than before. Over two
six-year spans of equal length it was 17.7% to 2019 and 16.7% to 2025. With every council read and nothing left out, 18.5%
and 18.9%. With stricter rules for leaving out councils that recoded their pubs, the two periods stay within a point of each
other. And over three-year spans it was 11.1% from 2013 to 2016, 10.7% to 2019, 8.9% to 2022 and 11.1% to 2025.

None of those overturns it. The last one is the caveat. The three years to 2025 closed pubs at the same rate as the three
years to 2016, and the low figure for 2019 to 2022 probably owes something to councils inspecting less during the pandemic
and taking longer to remove closed pubs. So "no faster" is right, and it doesn't mean closures have eased.

The councils left out as recodings were checked against ONS. Nearly all are registers that started or stopped coding their
pubs as pubs (one went from 4 registered pubs to 455 while ONS's count for it barely moved). Leaving them out doesn't flatter
the result: the few left out because their registers fell would, put back, raise the earlier closure share.

**Pub companies.** Every company registered with the trade code for pubs and bars, and for licensed restaurants, was
counted again from the register download: companies formed, dissolved and alive each year, and the survival of each
year's new companies. Every figure matched.

One limit the check added to the method. Before mid-2016 a new company didn't give its trade code until its first annual
return, so a company that closed before then never got one and can't be found by trade. That makes the counts for 2013 to
2015 too low and their survival too high. The piece now compares formations and survival from 2016 on. The record
formations of 2023 to 2025 hold either way.

**Going bust.** About 29,500 Companies House filing histories, every established pub and licensed restaurant company
dissolved since 2019 or in an insolvency process on 9 October 2026, were read again and sorted with separate rules. That's
where the fault above was found. Corrected, established pub companies went bust at 1.80 per 100 in 2019 and 2.69, 2.90 and
2.78 in 2023, 2024 and 2025, up 49%, 61% and 54%; licensed restaurants went from 2.83 to 3.23, 3.54 and 3.14. Restaurants
still go bust more often, and pubs rose further. In 2026 to 9 October pubs are at 2.31 and restaurants at 2.88, annualised
and provisional.

How companies left the register was re-read too: about a third went through an insolvency, a third were struck off at their
directors' request and a third were struck off by the registrar.

**The policy dates.** The test that runs every month from January 2021 as a pretend event was rebuilt three ways: as
drafted, with the corrected count, and with the corrected count and matching ages. In all three none of the policy dates
stands out as worse for pubs than for restaurants, and the three that sit lowest are the same. One caveat: about one pub
company in ten also carries the restaurant code, so the two trades aren't fully separate, which if anything narrows the gap
between them.

**Deprivation and rural areas.** Every cell of the deprivation table and both rural figures matched. The rural finding
also holds when companies at shared addresses are put back in.

**The British Beer and Pub Association.** Its release of 12 October 2026 was read again on 15 October. It gives the three
figures the piece quotes (295 pubs shut by the end of September, about 360 expected by the end of the year, 336 net
closures in 2025) and doesn't say how they're counted, as the piece says.

## What the check can't rule out

"Before the pandemic" means 2019. Earlier years can't be counted in full from these records, because a company that went
bust and was dissolved before 2019 wasn't read.

A company that went into administration or a voluntary arrangement and came out still trading wasn't read, so a few
rescues are missing from every year.

Going bust is dated by the first insolvency filing at Companies House, usually a few days after the process began.

The food register lags: a council removes a closed pub when it notices. That affects both periods, and the pandemic years
most.

None of the three sources says why anything happened, and the check doesn't either.

## Embargo

When the check was made, on 15 October 2026, the piece's public address returned "not found" on the live site, the
preview was marked not to be indexed, and the piece wasn't in the sitemap, the research index or the feeds.