My brother is coming to visit London. He wants one thing: beer. Apparently a lot of it. If there is one thing I don't know much about, it's beer and pubs. I'm Gen Z. I'm up at 6:30 for a sunrise run. My idea of a Saturday well spent is an 80-mile ride followed by a flat white. My brother, a Gen Y edge case, considers this a character flaw.
So I did what every little sister would do: instead of finally getting my payback for two decades of him beating me at every sport imaginable (…he was just taller), I opened VS Code. One evening later, I have 49,840 pubs in a dataframe, an NLP classifier on pub names, a pub vulnerability model, and an interactive tool that lets you check whether your own local looks structurally at risk.
His requested pub crawl - my original goal - is in there somewhere. But first: what I found isn’t really a story about beer. It’s a story about what happens to a country when the places where people belong become assets on someone else’s balance sheet.
I’ve lived in four countries in the last few years. I am not, by temperament, a pub person. But some of the most at home I’ve ever felt has been in exactly the kind of places this piece is about.
In Brooklyn (not New York), a tiny town north of Sydney on the Hawkesbury River, the kind of place with one pub and one opinion about outsiders, we entered the quiz as "The Wookies" (yes, that's a Star Wars reference) and somehow, against all odds, narrowly beat the local elderly ladies from Mooney Mooney. When we passed our prize back to them, something shifted. They started talking and suddenly we weren't tourists. We felt at home.
Or in Tuscany, just after I’d moved and didn’t speak a word of Italian, I was cycling behind a local club ride and held the wheel for 40km, eventually pulling on the front for them. After the ride, with some very Italian hand gestures, they invited me to a bar in a tiny village on the piazza. Over a vino rosso we exchanged a few broken sentences about the best routes. I felt at home.
Neither moment was planned. Neither required an app, a membership, or a booking. They happened because a place existed - open, informal, cheap enough that the price of entry was just showing up.
Ray Oldenburg called these spaces third places: not home, not work, but the pubs, cafés, and piazzas where community happens without anyone organising it. Robert Putnam, in Bowling Alone, measured what happens when they vanish: trust collapses, civic life withers, democracy frays. No places, no trust. No trust, no society. But neither asked the question that matters most for Britain in 2026: who is doing the dismantling of these third spaces, and why?
First, let’s put some more numbers to this. I pulled the UK business register snapshots for SIC 56302 (public houses and bars), from 2009 to 2022. Britain went from ~54,000 registered pubs to under 40,000. Over 14,000 gone. A quarter of the total. The trend is relentlessly downward.
(The ONS methodology captures businesses registered for VAT and/or PAYE. Very small pubs below the VAT threshold, the kind of micro-local that matters most, may be undercounted. In other words, the real number is probably worse.)
The timeline is a political economy in miniature: the beer duty escalator (scrapped 2013 after years of above-inflation hikes), the 2017 business rates revaluation, COVID, the 2022 energy crisis. But the point is: the decline predates the shocks. This isn’t one bad year, it’s structural.
Every region is losing. The North East, North West, Yorkshire, and the Midlands fastest - 25-30% of their stock. London least. The correlation between regional pub decline and the 2016 Leave vote is r = 0.67. I won’t claim causation with twelve data points, but the comfortable reading is Putnam's passive erosion of social capital. I’ll soon come to the harder reading of this.
The Brooklyn pub quiz happened because the pub existed. The cycling club aperitivo happened because the village bar existed. What happens in a town where neither does?
So I took all surviving pubs and every pub that has closed since 2016, and asked a simple question: what's structurally different about the ones that died?
I trained a random forest model (an algorithm that builds thousands of decision trees and asks each one what matters most) on several structural characteristics of every pub in the dataset to predict which pubs closed between 2016 and 2024. Then I decomposed the feature importances to see what mattered most. The overwhelming winner was spatial isolation: how far a pub stood from its nearest neighbour and how many other pubs existed within walking distance. The model was telling me something simple. Pubs don't die because of what's inside them. They die because of what's around them.
How exactly did I measure isolation? I calculated the distance to each pub’s nearest neighbour using a haversine BallTree. Translation: straight-line distance over the Earth’s surface and not walking time. (And yes, I used a BallTree instead of a KDTree. The curvature of the Earth matters when your dataset spans from Cornwall to Caithness. I will not be taking questions.) Then I did the same for every pub that has closed since 2016, using archived snapshots of the same dataset.
The pattern is as follows: median nearest-neighbour distance for a surviving pub: ~280 metres. For a pub that closed: ~640 metres. Simply said: pubs that were isolated died while pubs in clusters survived. The red dots on the map are the next to go. They are disproportionately rural, disproportionately northern, disproportionately in the same communities that have already lost the most.
And it’s not just correlation in a static picture. It’s dynamic. Every pub that closes increases the isolation of the remaining pubs, pushing more of them into the danger zone. This is the spatial death spiral.
As promised, the harder reading of the decline of pubs. Brett Christophers (Rentier Capitalism) and Grace Blakeley (Stolen) can offer an answer. The UK economy increasingly extracts value from assets such as land, housing, infrastructure, rather than producing it. The pub is a perfect case study: it sits on land, it has a freehold, it generates modest cash flow. For a private equity firm, that’s not a community asset, it’s a leveraged buyout opportunity. Acquire, load with debt, extract, underinvest, and when the pub becomes unviable, sell the land to developers at a multiple.
We have already seen that isolated pubs die. But pub isolation isn’t a coincidence, someone creates it. When a pubco owns thirty pubs in a region and decides to cut the five least profitable, the remaining twenty-five don’t just lose neighbours. They lose the clustering effect that was keeping them alive. In other words, the just identified spatial death spiral has an author.
The regions that lost most pubs are the regions where ownership was most consolidated under leveraged pubco models. When Ei Group raised rents to service its debt, it was tenants in Sunderland and Burnley who couldn’t pay, not tenants in my dear Islington.
Britain’s largest pub company is called Stonegate: roughly one in eleven falls under this. You know the brands: Slug & Lettuce, Yates’s, Walkabout, all owned by TDR Capital, a private equity firm that also owns Asda. Its corporate structure runs through the Cayman Islands.
Stonegate carries over £3 billion in debt, mostly from its leveraged acquisition of rival Ei Group in 2019, completed months before the pandemic shut every pub in Britain. In 2024: £455 million in interest payments, a £214 million loss, and a citation from the Department for Business for failing to pay minimum wage to 3,650 workers.
It’s not alone. Greene King (~2,700 pubs) was taken private by Li Ka-shing’s Hong Kong property empire for £4.6 billion. Punch Pubs (~1,300 pubs) was bought by Fortress Investment Group, a US private equity firm formerly owned by SoftBank; now majority-owned by Mubadala (Abu Dhabi). Between them, PE-backed and overseas-owned companies control roughly a quarter to a third of every pub in Britain.
So here’s the political economy of pub closures. It is not: people stopped going. It is: pubs became collateral in leveraged buyouts, debt costs were passed down as higher rents and lower investment, and the pubs that couldn’t sustain the extraction closed, while the ones that could were reshaped into higher-margin branded concepts serving a wealthier clientele.
The Bricklayers Arms didn’t die of natural causes. It was killed by a balance sheet in George Town. And if you think London escaped this, look closer…
London looks like the success story. Smallest decline nationally. If you live in Hackney or Peckham, you might think the pub is thriving: new taproom on every corner, natural wine bar where the laundrette used to be.
London didn't lose its pubs. It swapped them. And this is where the spatial story meets the ownership story. In Burnley, private equity closes a pub and nothing replaces it since the land isn't worth enough. In London, the land is worth plenty, so the site reopens: refitted, rebranded, repriced. The count stays stable. But the name changes, and so does who walks through the door. This is why the random forest flagged spatial isolation as the dominant predictor: it's not geography acting alone, it's the financial logic of extraction expressing itself spatially. PE firms don't close pubs at random. They close the ones that sit on cheap land with thin margins and replace the ones that sit on expensive land with higher-margin concepts. The map of pub death is a map of where extraction was profitable and reinvestment wasn't.
I used the pub names to expose it. I classified every surviving London pub as either traditional: such as Red Lion, Kings Arms, Bricklayers Arms, names encoding centuries of heraldry, trades, and coaching roads or as modern, such as The Yard, The Social, Tap Room, names that strip away history for something brandable. Then I mapped them.
Traditional pubs scatter across all of London: inner, outer, rich, poor. Modern pubs cluster in zones 1-2 and the gentrification corridor from Hackney through Peckham. London's count looks stable because craft taprooms in gentrifying postcodes replace closures in outer boroughs. But a £7.50 hazy IPA in Dalston is not the same third place as a £2.50 pint in a working men's club in Barking.
I didn’t want to rely on my own keyword list, so I ran a sanity check using machine learning. (ML people: TF-IDF on character n-grams of every London pub name, K-means clustering projected via PCA. I know, TF-IDF on character n-grams is not state of the art. But when your unsupervised model rediscovers the same class structure that took historians centuries to document, you don't need BERT/ fancy fine-tuned LLM). Simply said, I fed every pub name into an algorithm that breaks words into small character fragments, think of it as teaching a model to recognise the texture of a name rather than its meaning. Then I asked it to sort all names into groups, with no information about history, geography, or what “traditional” even means.
It found the same divide I did. And when I plotted those algorithmically-discovered groups back on the map, they had geography: old-textured names in the suburbs, new-textured names in the centre. The algorithm sees what the eye sees - naming eras are a spatial signature of class.
Teal: the old Britain: heraldic, royal, rural, trades, naval. Pink: the new Britain: bars, kitchens, tap rooms. The countryside is teal and the cities are turning pink. Both are valid places to get a drink. But only one is a third place in Oldenburg’s sense: a space where the full cross-section of a community meets without requiring a particular income, education, or cultural fluency to walk through the door.
So what do you do with this information? I built a companion tool: Is Your Local At Risk?
The model behind it is deliberately simple and that's the point. For every one of Britain's pubs, I calculate two things: how far away is the nearest other pub (nearest-neighbour distance, via haversine BallTree), and how many pubs exist within a walkable radius (pub density). Those two spatial signals get combined into a continuous risk score, then bucketed into green (clustered, resilient), amber (thinning), and red (isolated, vulnerable).
That's it. No pricing data, no opening hours, no operator quality scores. Just structure and space. Because the finding from the closure analysis is stark: the single strongest predictor of whether a pub closed between 2016 and 2024 is how alone it was. Not how good the beer was or who owned it. But whether it had neighbours.
Enter your postcode and see every pub within walking distance, scored green/amber/red by the vulnerability model. The risk scores, not coincidentally, map onto the worst loneliness scores in the ONS wellbeing data.
The tool won't tell you btw whether your local does a decent Sunday roast (I’m working on that model, it just requires some more field research). But it will tell you whether the spatial conditions around it look like the conditions that preceded every closure in the dataset.
The fatalist reading of this all is that: pubs close because economics, because supermarket alcohol, because young people don’t drink as much (pls don’t blame Gen Z for this one, we inherited the wreckage). These are real factors. But they’re the cover story for a series of reversible policy choices. Britain chose to tax a pint in a pub more heavily than wine in a restaurant. Chose permitted development rights that let landlords flip pubs into flats without planning permission. Chose to cut council budgets by 40%, hollowing out the only institution that could have said no.
So here’s my policy brief. I’ll keep it short.
Map it. Build a national Third Place Index. Every neighbourhood scored by density, diversity, and walkability of its social infrastructure. I built a version with free data and a laptop. The ONS has better resources than I do. Presumably.
Protect it. Use the vulnerability model to fast-track the 500 most isolated pubs for Asset of Community Value status (a legal designation that gives communities 6 months to bid when a pub goes up for sale). Require planning permission for any pub conversion, not just where communities have the lawyers and the social capital to fight. The current system protects pubs in Richmond and lets them die in Rochdale.
Fund it. The government spends serious money treating loneliness downstream: social prescribing, befriending services, community connectors. Good programmes. But it’s like funding ambulances while defunding the road barriers. Redirect 10% toward keeping existing third places open. A pub costs less to save than a community centre costs to build, and it comes pre-loaded with something no government programme can manufacture: centuries of accumulated meaning.
Measure it. Add third place access to the Index of Multiple Deprivation (IMD). What gets measured gets managed. Right now a neighbourhood can score perfectly on the IMD while having zero places for its residents to sit together.
Tax it. If TDR Capital wants to own 4,500 British community assets from the Cayman Islands, it can fund the mechanism for communities to buy them back. A levy on offshore-held pub freeholds, ringfenced for community ownership.
Because the counter-model already exists. Community-owned pubs have grown 63% in five years. The model is simple: a community raises shares to buy the freehold. Ownership stays local. Profits get reinvested. No absentee landlord, no debt servicing, no Cayman Islands. The Ivy House in Nunhead became the first pub in Britain listed as an Asset of Community Value, saved from developers by locals who crowdfunded the purchase. It’s still open. The people who own it are the same people who drink in it.
So, the Asset of Community Value framework exists and the co-op model works. What’s missing is money and a government willing to ask where it should come from.
Oh right. My brother.
After all that, here’s what I’ve actually built him: a pub crawl through London that traces the naming eras. I start at a medieval heraldic pub in the City (there are still a few that have served beer since before the Great Fire). Walk east to a Victorian trades pub in Whitechapel - a Bricklayers or a Carpenters. Cross into Hackney for a modern craft spot to see what’s replaced the old guard. South to a coaching inn in Borough. End at a proper South London local that hasn’t changed its name or its prices since Thatcher.
Five pubs. Five centuries. One brother who doesn't know he's about to get a lecture on the financialisation of community infrastructure with his fifth pint. I’ll let you know if he ever forgives me.
Oh, and next month I'm cycling through Australia for six weeks. I want to see whether a country that built its third places differently has built something more durable than what England is losing. Australia's social infrastructure runs on not-for-profit clubs such as RSLs, surf lifesaving clubs, bowls clubs and leagues clubs, all member-owned, community-governed, and structured so that profits stay local. It's essentially the co-op model that Britain is only now discovering, except Australia built thousands of them. The trade-off is that many are funded by poker machines (“pokies”), which is its own political economy story. But the basic architecture of community ownership, democratic governance, no Cayman Islands balance sheet, is what I've been arguing Britain needs. I want to see how it holds up in person. I’ll promise that I’ll bring the dataframe to the pub quiz.
Team name: still The Wookies.
If this was useful, share it - especially with your MP who loves a pint. Hope they love the data too. Also consider buying me a coffee to keep the project alive:)


Energy (top) and emissions (bottom) contributions from different food groups within the three diets modelled by the study. Each column represents a different diet (left to right): lowest-cost, lowest-emission and most common items. The boxes are coloured by food group: animal-sourced foods (red), legumes, nuts and seeds (blue), oils and fats (purple), vegetables (green), fruits (orange) and starchy staples (yellow). Source: Bai et al. (