Open the Michelin Guide to almost any city and the assumption most readers bring with them is that every restaurant on the list has, in some sense, been recognized — anointed, however modestly, by the world’s most self-serious food critics. It has not. Across the roughly 19,400 restaurants currently in the guide, six in ten carry no star and no Bib Gourmand at all. They are simply there: found, described, mapped, and otherwise unrewarded. The Guide, by its own numbers, is mostly a list of restaurants it declined to praise.
This week’s number is two. That’s roughly how many times more likely a restaurant is to actually earn recognition — a star or a Bib Gourmand — if its wine list is good enough that Michelin’s own inspectors bothered to mention it, compared with a restaurant where they didn’t.
Michelin has always insisted this isn’t supposed to be about ambiance. The Guide’s own account of how it inspects a restaurant draws a firm line: recognition is decided on the food, and the food alone. Décor, table setting, service — the entire theater of dining — is explicitly walled off from what earns a place its stars. It is, on its face, an admirable and slightly implausible claim. Ambiance is not nothing. It is most of what a restaurant actually spends its money on.
So we picked six things a restaurant can spend money on that have nothing to do with what’s on the plate — a terrace, a view, a garden, a wine list worth discussing, a counter to eat at, valet parking — and asked a very literal question of Michelin’s own listings, across every restaurant it covers, in every country it covers them in: does having one of these things actually predict recognition? If the Guide is telling the truth about itself, the answer should be no, six times over.
Two of the six actually point the wrong way. Restaurants with a terrace are less likely to carry recognition than those without one, not more — a great view moves the needle by less than two points, in the same losing direction. A garden is a coin flip. Valet parking and counter seating nudge upward, but only barely, and we didn’t chase either one any further, for reasons that will matter in a moment. Which leaves one amenity that isn’t behaving like the other five at all.
Michelin was almost telling the truth.
A restaurant whose wine list earns a mention holds recognition at more than half the rate of one that doesn’t — a twenty-one-point gap, the largest of anything we tested, in either direction. That number, on its own, proves nothing, and the obvious objection is worth saying out loud before we try to answer it: a restaurant with an ambitious wine program is very plausibly just an expensive, ambitious restaurant generally, and expensive restaurants already carry more stars for reasons that have nothing to do with what’s in the cellar. Wine list, in other words, might just be a fancy way of re-measuring price. The obvious next move is to ask whether it’s doing anything price wasn’t already doing.
It survives. Stratified by price tier and tested formally, restaurants with a noted wine list carry roughly double the odds of recognition of otherwise-comparable restaurants at the same price point — not identical across every bracket, but real and positive in three of the four, including the cheapest one, where a good list is rarer and, evidently, more remarkable when a restaurant has one anyway.
The wine list wasn’t just standing in for the price of the meal. Something else was still showing up alongside it.
We sent the analysis to a peer reviewer at that point, reasonably pleased with ourselves. The finding came back with a question we hadn’t asked: what about the country? Wine culture is not evenly distributed. Italy, Switzerland, and Austria are the three countries where an inspector is most likely to note a wine list at all — and all three post recognition rates below the global average. Meanwhile Japan, where a noted wine list is rarer than almost anywhere else in the dataset, posts one of the highest recognition rates in the guide. If wine-loving countries also happen to be stingier with stars for reasons of their own, the whole finding could be an accident of geography wearing a wine list as a disguise.
highest wine-list rate in the dataset
one of the lowest wine-list rates
Adding country to the model didn’t explain the finding away. If anything, it did the opposite: once the model accounts for which country a restaurant sits in, the estimated odds tied to a good wine list go up, not down — at the cheapest price tier, from roughly one-and-a-half times to nearly three times. That’s one estimate from one model, not proof the real-world effect grew larger the moment we looked at it — but it does mean Italy and Japan were pulling hard enough in opposite directions that ignoring geography was more likely muting the pattern than manufacturing it.
One piece of the original story didn’t survive, though, and it’s worth saying so plainly rather than quietly dropping it. We had assumed the wine-list effect grew steadily richer as prices climbed — a tidy, escalating story. Once country entered the model, that escalation mostly disappeared; what’s left is a real effect that doesn’t track cleanly with price at all. The cheapest tier’s result, in particular, leans hard on just twenty-three restaurants, five-sixths of them Italian or French — too thin a base to hang a claim about budget dining on its own.
One aside worth a sentence, not a detour: price itself climbs in predictable steps almost the whole way up the award ladder — and Bib Gourmand, the award created to reward good cooking at a fair price, does in fact price out cheaper on average than restaurants with no award at all. The system is working as advertised, at least on that front.
So, back to the objection we raised at the start: is this just price wearing a disguise? Not entirely, on this evidence — but the far more boring explanation than "inspectors are secretly charmed by a good bottle list" is probably still the right one. A restaurant willing to build and maintain a wine program worth an inspector’s notice is also, on average, the kind of restaurant that takes the rest of its business seriously — sourcing, training, consistency, the unglamorous stuff that food criticism actually claims to reward. A wine list, unlike a terrace, cannot be rented for a season. It requires a cellar, a relationship with growers built over years, and someone on staff who knows what’s in it. It may not be ambiance at all. It may just be one of the few amenities that’s expensive to fake.
A last note on shelf life: the Guide is republished monthly, and this analysis reflects one snapshot of it. The pattern could shift as the Guide itself does — which is exactly why we pinned the exact edition this piece is built on.
Michelin said it wasn’t scoring the room. Mostly, the data agrees — a view doesn’t help, a terrace actively hurts, a garden does nothing at all. But one exception got through, and it got through in every country and at nearly every price we checked it against. If the Guide is grading the food and only the food, something about a good wine list is still leaking onto the scorecard.