Thursday, March 7, 2013

Network analysis of Michelin starred restaurants

The world is full of snobbery, and the food & wine business has more of it than almost anywhere else. Probably the most snobbish part of the restaurant business is the 3-star rating system used in the annual Michelin tourist guides. It seems to me that this is worth trying to explore in some detail.

Michelin stars

Originally (starting in 1926), the Michelin ratings were intended (as noted by Anthony Capella) "to tell haute-bourgeoisie French motorists where to find Parisian-style fine dining"; and that philosophy is still prevalent today. So, if that's not what you're looking for, then Michelin cannot necessarily help you. For example, Paul Levy has observed: "The third Michelin star seems always to have been awarded for surroundings and service as well as food"; and restauranteur Mat Follas (winner of the BBC TV programme MasterChef in 2009) comments: "There's a perceived level of service and over-formality that comes with Michelin, and it's not something that I'd aspire to."

Furthermore, the Michelin guides are frequently accused of a distinct francophile bias. In particular, Anthony Capella notes: "From the type of glassware to the number of amuse-gueules, there is a Michelin way of doing things that often seems to stifle rather than celebrate regional idiosyncrasy." In other words, the inspectors "don't understand" other cooking styles. Nevertheless, as Andy Hayler observes: "Despite its eccentricities, for French food Michelin is generally reliable in its assessments at the top of the restaurant tree".

So, in this analysis I will restrict myself to restaurants that are actually in France (not Monaco!), for which Michelin released the 2013 ratings on February 18. This is appropriate, because in 2010 "Gastronomic Meal of the French" was added to UNESCO's Representative List of the Intangible Cultural Heritage of Humanity, which means that we are supposed to actively safeguard it (along with "Viennese Coffee House Culture").

The stars themselves, according to Michelin, represent (since 1936):
  • One star indicates a very good restaurant in its category, offering cuisine prepared to a consistently high standard. A good place to stop on your journey.
  • Two stars denote excellent cuisine, skillfully and carefully crafted dishes of outstanding quality. Worth a detour.
  • Three stars reward exceptional cuisine where diners eat extremely well, often superbly. Distinctive dishes are precisely executed, using superlative ingredients. Worth a special journey.
The first post-WWII year in which Michelin re-instated their 3-star ratings was 1951, which is thus a convenient starting point for my analysis of haute cuisine dining.

I have compiled a dataset of the number of stars every year since 1951, for every restaurant in France that received 3 stars at least once during that time. Unfortunately, Michelin does not maintain any restaurant lists (you have to search through the individual guides), and so I have compiled the dataset from Marc Vilbois' Art et Gastronomie, supplemented by Andy Hayler's 3 star Restaurant Guide, plus extensive searches through the French edition of Wikipedia and recent Michelin press releases. [The data are apparently also available in: Jean-François Mesplède. 1998. Trois Étoiles au Michelin : une Histoire de la Haute Gastronomie Française et Européenne.]

The analysis

There are 59 restaurants on the list, 26 of which currently still have a 3-star rating; 13 currently have 2 stars, and 6 have 1 star. There are 14 restaurants that currently have no stars, most of which have closed permanently. (It is important to re-emphasize that the 3-star restaurant in Monaco is not included, which reduces the number of restaurants by one, compared to what you will see in the newspapers.)

To start the analysis, we can simply look at the number of three-star restaurants throughout the study period. They come and go, opening and closing, and gaining and losing stars, but there is still an inexorable increase through time.

Next, I have used the manhattan distance and a NeighborNet network to produce the following two network graphs, which summarize the patterns of variation in the stars. Objects that are closely connected in the network are similar to each other based on similar times at which they were at their 3-star peak, and those that are further apart are progressively more different from each other.

As far as the years are concerned, we would expect there to be a gradient from 1951 to 2013, since restaurants come and go continuously. Indeed, this is exactly what the first network shows.

The break from 1962 to 1963 reflects the arrival of "Maison Lameloise" and "L'Oasis la Napoule", the latter of which lost its stars in 1987. The clustering of the years 1972-1999 is associated with the simultaneous arrival on the scene of "La Côte St Jacques", "Au Crocodile", "Hôtel de l'Espérance" and "Le Vivarois", the latter of which disappeared again in 1999. The gap between 1986 and 1987 reflects the arrival of "L'Arpège", "La Vague d'Or" and "l'Hôtel George V".

The break from 2000 to 2001 reflects the move of chef Alain Ducasse from "l'Hôtel du Parc" to "Hôtel Plaza Athénée", taking his three stars with him, combined with the opening of "La Ferme de Mon Père", which immediately acquired three stars. The gap between 2006 and 2007 reflects the closure of both "La Ferme de Mon Père" and "Jamin", the sale of both "Le Buerehiesel" and "Hôtel de l'Espérance", and the resurgence of "Maison Pic".

The pattern among the restaurants themselves is more complex, however, as shown in the second network. There are four groups that might be recognized, based on the timing of their star patterns, and this is clearly reflected in the network (as shown in different colours).

Restaurants with a red asterisk have three stars in 2013.

The small blue group had their Michelin-starred heyday during the first one-third to two-thirds of the time period. The only exception is "Le Café de Paris", which had three stars in the early 1950s but is not placed in this group by the network.

The larger pink group had Michelin stars for all or almost all of the time, although none had three stars for the whole time. The longest was "l'Auberge Du Pont De Collonges" (made famous by chef Paul Bocuse), which had three stars for 49 / 63 years, followed by "l'Auberge de l'Ill" (47 years), "La Maison Troisgros" (46) and "Restaurant de la Tour d'Argent" (44 years, but no longer has three stars).

This group forms two subgroups plus a few outliers. The subgroup from "L'Auberge Du Père Bise" to "Restaurant Lasserre" had their 3-star peak before 1995, whereas the subgroup from "Maison Pic" to "Lucas Carton" had three stars into the 2000s as well. "Hôtel Plaza Athénée" and "Hôtel George V" share the characteristic of having their 3-star period only in the 2000s. "La Petite Auberge" and "Restaurant Charles Barrier" are distinct in having their 3-star period in the 1960s and 1970s, respectively. "La Cote d'Or" is distinct because it had three stars at two very different times, 1951-1963 with chef Alexandre Dumaine, and 1991-2013 with Bernard Loiseau (and then Patrick Bertron).

The large black group had Michelin stars during the final one-third to one-half of the time, except for "Le Café de Paris". They almost all still have three stars in 2013, or they have closed. The subgroup from "Le Petit Nice" to "L'Arpège" had three stars for much longer than did the other members of this group. "Le Café de Paris" had three stars for only a few years (1951-1955) and no stars at all at any other time, just like "l'Hôtel du Parc" (1996-2000) and "La Ferme de Mon Père" (2001-2006).

The green group had Michelin stars during the final two-thirds of the time, but usually peaking before the 2000s. Few still have three stars. "Le Buerehiesel" is distinct in receiving its stars later and peaking later than the others in this group.

In conclusion, this network neatly arranges the restaurants based on when they had stars, and particularly three stars, with the exception of "Le Café de Paris".


As you might imagine, three Michelin stars does not come cheap from the customer's point of view. Even one star increases the meal's price significantly, no matter what it tastes like. Running a 3-star restaurant appears to require about two staff members for every three customers, and someone has to pay these people — so, the bill you get at the end of the meal is guaranteed to take your breath away. Notwithstanding this, a list of any sort provides a challenge to certain types of people.

There appear to be only two people currently advertising on the web that they have eaten in all of the 3-star restaurants in the world that were available to them at the time: Andy Hayler claims to have achieved this on three separate occasions, first in 2004 and then with catch-ups in 2008 and 2010, while the anonymous 3starbackpacker claims to have done it between November 2006 and October 2007. Hayler has reviews of all of his visits (along with a score out of 10), but the Backpacker has only a couple of his reviews. He does, however, use a 13-point scale to rate his visits, as shown in the histogram below. He is particularly critical of the variability among the restaurants, and recommends only 33 of the 73 restaurants as being worth a re-visit (ie. the other 40 do not merit a 3-star rating).

A     **
A/A-  ****
A-/A  *****
A-    *********
A-/B+ ****
B+/A- ********
B+    *******
B+/B  ******
B/B+  ****
B     ********
B/B-  ******
B-/B  *
B-    *********

There is also a story of another person who tried to eat in all of the 3-star restaurants but gave up. Hayler notes that: "Keeping pace with the three-star crowd is not easy ... You can't eat this stuff all the time. You need a break." A New York childrens-wear manufacturer named Leonard S. Bernstein describes how to survive eating in eight 3-star restaurants in eight days, in his classic book The Official Guide to Wine Snobbery. He mentions 17 of the 21 French 3-star restaurants from 1982, two of which no longer exist and only seven of which still have three stars today.

No comments:

Post a Comment