Should you eat where the locals eat?

Two weeks ago I wrote about cuisine diversity in global cities and one of the findings was that Rome is 81 percent Italian. I admitted that I didn't know why it was so high relative to other cities and there were tons of commenters on LinkedIn with guesses. I'd guessed food tourism. A few people agreed: visitors come to Rome for Italian, so Rome serves Italian. Others pushed back with a theory I liked better, that Romans eat out a lot and want Italian when they do.

I can't measure how much either theory drives Rome's restaurant economy. But there was something I could check: are the restaurants that appeal to visitors different from the ones that appeal to people who live there? This is especially interesting because it's the oldest piece of travel advice there is: eat where the locals eat. Everyone says it, but I've never seen anyone test whether it's actually true or why.

What we measured

When Seemor reads a restaurant's reviews, it tries to work out who the restaurant appeals to: neighborhood regulars or people visiting on a trip. It's a judgement from the text. Nobody is counting heads at the door. It draws on the same reviews that give us the food, value and many other scores, so none of these measures is independent of the others. What we have is about 480,000 restaurants across 76 countries with that reading, which is enough to look for patterns.

I split each city's restaurants into three groups: the places that read most local, the places that read most tourist, and the rest. The rest is most of the city. Pooled across 44 major international cities, 24 percent of restaurants lean local, 13 percent lean tourist, and 63 percent read as mixed. The cut-offs are the same in every city, so when I say Paris has more tourist-leaning restaurants than Riyadh, it does. Everything below is a share of restaurants in a group, never a share of diners.

Same cuisine, different restaurant

Looking at Rome first, since that's what started this. Compared to 80 percent of all of Rome's restaurants in this data (81 in the last piece, a slightly different cut), Italian is 89 percent of Rome's local-leaning restaurants, 77 percent of the mixed group and 79 percent of the tourist-leaning ones. So, the local-leaning restaurants are actually more often Italian than the tourist places by a decent margin. This is pretty strong evidence that Italian isn't just there for the tourists. The places that read most local are even more Italian than the tourist-leaning ones. The food tourism theory is less clear; there are certainly a lot of Italian restaurants that tourists frequent, but it's not obvious that food tourism is a major force driving the high concentration of Italian restaurants.

So in Rome the cuisine doesn't change much between the crowds. What does? The kind of place. Digging into the large Italian buckets, pizza al taglio is 20 percent of the local end and 4 percent of the tourist end, while the trattoria rises from 16 to 29. The most local formats are the cheap, specific, everyday ones. The most tourist-leaning formats are the fine, the modern, and the ones that describe themselves only as "Italian".

That's one city. Here are three more, each with its home cuisine opened up the same way.

I'd assumed tourists would be the ones eating foreign food, the way you end up at a pizza place in Bangkok on the fourth night. The data says mostly no. On cuisine, the tourist-leaning places look like the rest of the city: within a few points of the mixed middle in most of the 44 cities. It's the local end that stands apart, more Italian, more Portuguese, more Turkish than anyone else. The exceptions run the other way, mostly cities where the home cuisine has a thinner international reputation or the city sells itself as cosmopolitan: Mexico City, São Paulo, Bogotá, Madrid, Mumbai and Riyadh, where the visitor's restaurant is the international one. So serving the home cuisine doesn't mark a place as a tourist restaurant. If anything, the local places are the most devoted to it. That said, there are lots of other differences.

The type of place moves the same way everywhere. In Lisbon the tasca is 59 percent of local-leaning Portuguese restaurants and 9 percent of tourist-leaning ones. The marisqueira, the seafood house, goes the other way, 6 to 26. Istanbul's esnaf lokantası, the tradesmen's lunch canteen, is 16 percent local and 2 percent tourist. Mexico City's fonda is 16 and 2.

I expected London and New York to be different, since neither has a dominant cuisine. They aren't. Inside their own cuisine they behave exactly like Rome. The fish and chip shop is 20 percent of London's local-leaning British places and 3 percent of the tourist-leaning ones; Modern British runs 3 to 16. The New York diner is 18 percent local and 7 percent tourist. Local food, on this reading, is the everyday habit of a place at least as much as its culinary origin. If you want to look at any other city, the data page has all 44.

Pricing is very different for locals vs. tourists

The obvious next question is what the tourist version costs.

More.

In the places that lean local, 16 percent are expensive or above. In the middle it's 36 percent. At the tourist end it's 58 percent. I expected the direction. I didn't expect it to hold in all 44 cities without a single exception. Istanbul goes from 21 percent expensive among the local places to 72 percent among the tourist ones. Berlin has the smallest gap of any city and it's still 12 to 32.

An interesting question is whether locals prioritize cheaper places or simply avoid the tourist crowds. I'd guess that varies by city and has something to do with local income levels, but I haven't looked into that with any depth.

What does the tourist premium buy?

Hint: it's not the food.

Seemor uses public review text to score several dimensions on a scale of 1-10. Three dimensions are particularly interesting here:

1. Food quality: How good is the food. Not the overall experience, not how noisy it is, not the service, just the food.

2. Value for money: This is not the price level, this is based on what reviewers say about their perception of value. Expensive places can have very high value for money scores and cheap places can have poor ones.

3. Setting appeal: This is about where a place is and how it's set up.

Across all 44 cities, on food quality, 68 percent of local-leaning restaurants score 8 or higher (which is a high bar!). In the middle it's 51 percent. At the tourist end it's 37 percent. Again, local beats tourist in every one of the 44 cities, and in 40 of them the middle sits between the two.

Value for money follows food: 93 percent of local places score well on value, against 52 percent of tourist ones.

So far the old advice is winning comfortably. The places that read local are cheaper, provide a better perception of value and the food is better. And it's true in Tokyo, Mexico City, New York and everywhere else we looked. If that were the whole story I'd have stopped here.

It isn't the whole story. We also score each place on the appeal of being there: the room, the view, the street outside. This is the one measure that flips, and it flips hard. In the local-leaning group, 41 percent score well on setting. In the tourist-leaning group, 83 percent do. Rome's tourist places are twice as likely as its local ones to score well here.

That reframed the whole thing for me. I'd been reading "tourist restaurant" as a synonym for "worse restaurant", and the data says it's a different product. A place facing a cathedral can sell the view alongside the meal, and the reviews say people are happy to buy it. Three streets back, the food has to do more of the work. That's my guess at the economics, and this data can't prove it. But it fits every chart in this piece.

The one-line version, said carefully: tourist-leaning places tend to cost more and score lower on food, but they tend to sit in better settings.

They aren't even in the same neighborhoods

There was a question I'd been avoiding, because I assumed I knew the answer. Are these two groups different restaurants on the same streets, or different parts of the city? Rome settled it in one chart.

Around the Trevi Fountain, 74 percent of restaurants read tourist-leaning and 2 percent local-leaning. The centro storico is 67 and 2. Trastevere, which every guide will tell you is where to go to eat like a Roman, is 45 and 2. Out in Centocelle, east of the center, it's 2 and 47. Central Rome has plenty of restaurants with mixed appeal, but almost none that read strongly local.

Two things to read into that chart. Local-leaning is a high bar. The reviews have to say clearly that locals are the crowd, which is also why mixed appeal is the biggest band everywhere, Centocelle included. A gray bar out there doesn't mean tourists are turning up in Centocelle. It means most of its restaurants don't read as anyone's in particular. And the local-leaning share never gets much past half, even in the most local neighborhood, while the tourist end runs to three quarters. The tourist center is close to a monoculture. The local neighborhoods are mostly just mixed.

Rome made me wonder whether this was happening everywhere, so I plotted every named neighborhood in 43 of the 44 cities, 1,549 of them. Part of the shape is built in: most restaurants everywhere sit in the mixed band, so no neighborhood can be high on both. But if you scattered each city's restaurants across its neighborhoods at random, you'd still expect about sixty neighborhoods with a fifth of their restaurants in each crowd. There are ten. So the two kinds of restaurant aren't just different. They sit in different parts of the city far more often than chance would put them there.

There's more in that chart than I've got room for here. London and New York have most of their neighborhoods at the local end, Istanbul and Paris at the tourist end, and some cities keep their tourist restaurants in a few streets while others spread them across town. Where you actually find the places locals love, and whether you'd want to eat in them, is a future analysis.

Why some cities break the pattern

Tokyo keeps the direction on every measure but the setting gap nearly closes, 79 to 92, against Rome's 40 to 83. Tokyo's local places are already in good settings, or at least the reviews say so. Tokyo is also where the kind of restaurant matters least on its own. Izakaya and ramen read local, sushi and yakiniku lean tourist, but in an earlier pass through the Tokyo data the same format moved with price: a cheap ramen counter read local and an expensive one read tourist, while kaiseki barely moved. In Tokyo, price seems to matter as much as the kind of place.

Dubai has almost no food gap: 52 percent of local-leaning places score 8 or higher, against 47 percent of tourist-leaning ones, the smallest difference of the 44. Dubai's local end is also only 11 percent Middle Eastern, so "local" there means something quite different from what it means in Rome.

Riyadh barely has a tourist end. Only 3 percent of its restaurants read as tourist-leaning, so the gradient there probably reflects differences among resident crowds rather than visitors against locals. São Paulo and the San Francisco Bay Area are nearly as low.

Paris is the other way round, among the most tourist-heavy of the 44 at 23 percent, and the French share barely moves between the groups. I don't have a good explanation beyond the obvious one, that Paris has a lot of visitors and they eat everywhere.

Two limits on all of this. The crowd reading is inferred from what people write, so a place near a monument with a lot of one-visit reviews will read tourist whoever actually eats there. And in several cities a quarter of the tourist-leaning home-cuisine restaurants describe themselves only by their cuisine, "Italian" and nothing more. Some of that is a gap in our tagging. Some of it, I suspect, is that a nondescript restaurant in the center of Rome tends to be a tourist restaurant, and the lack of description is itself the tell.

So should you eat where the locals eat?

Maybe, but the reason is more specific than I'd assumed. In most cities the tourist is eating the same cuisine as the locals, in a more expensive version of it, in a nicer room with a better view and worse odds on the food. Sometimes that's the right trade. If you've got one night in Rome and you want to sit on a piazza, the data says you'll pay for it and probably eat less well, and you may decide that's fine. If you're there for the food, a local-leaning place is the better bet, and that holds in all 44 cities. Whether it's a few streets back or a bus ride away is the next question.

Of course, what locals love isn't necessarily what you'll love. And I don't love the idea of sending a bunch more tourists into the local favorites either. When I travel I mostly skip the tourist places, unless the location is the point of the evening, and look for mixed-appeal places with great food. That's what I use Seemor for.

Look up your city

I built something new for this one. The Seemor data page has every chart in this piece, live: pick up to five of the 44 cities and compare the places that lean local, mixed and tourist, by cuisine, kind of restaurant, price, food quality, value and setting, and now neighborhood by neighborhood. If your city's local end doesn't match where you'd send a friend, tell me. As with the last two pieces, I'm sure this gets something wrong about somewhere you know well, and I'd love to hear about it. I'm also always interested in ideas for new questions to explore.

See this data for other world cities →