This is how good this post is:
Ah, Spain! The strongest nation in the world; Spaniards have tried to destroy it for centuries, and yet they’ve failed (Bismark dixit). For all its faults (a faults Spain has galore), one thing is incontestable: food is pretty good down there. While in Spain, one does not need to look for a $200-a-head Michelin Star restaurant to have a decent meal. In fact, you can have unforgettable lunches and dinners for $10 (or even less if you are the frugal type). But let’s suppose you are a person of peculiar habits, trained in the school of luxury, accustomed to the ways of the Gods, and you don’t eat in non-Michelin star restaurants… Or may be you are neither refined, nor posh, nor rich and there is nothing god-like about you, but simply want to treat yourself or your significant other to a special meal during your next holiday in sunny (and not so sunny) Spain. Where should you go? Well, do not let me set your agenda; go wherever you fancy! But let me tell you where you will find “more” Michelin Star restaurants.
Michelin and innovation
The Michelin star system might seem a fancy frivolity, an affliction of decadent western nations, too fat and too rich to bother about the plights of the world. I would tend to disagree with that. Rather, I wonder if these stars might tell us something about how innovative a region is overall?
A comparison of Michelin Stars across countries would not really address this question. Places like Israel, Korea and the USA would fair rather poorly -as they have few (if any) Michelin Star restaurants, whereas Belgium and Spain would look untouchable, with their many starred restaurants. The problem, however, is that data from OECD show that Spain spent 22,508 million US dollars in 2020 in research and development (R&D). That amounts to 474.54 dollars per Spaniard. The figures per person for Israel, Korea and the USA are, respectively, 2123.9, 2011.64 and 2006.2. So, it is unlikely that Spain will produce the next big technology anytime soon… but it is the most likely country in the list to earn another Michelin star next year.
What is interesting about Spain is that it has a big pile of these fancy restaurants scattered all over the place, as you can see in this map,
In addition, there are big (but not overwhelming) regional differences in terms of economic development, labour force composition and degrees of innovation. So the question is, does local variation in economic development (and, particularly, innovation levels) correlate with the number of Michelin-star restaurants? Let’s try to figure this out.
The data.
Before we go ahead, a caveat. A best look into the Michelin-innovation question would really require longitudinal data. This is because we could then study the evolution of macro-aggregates which, most likely, will affect the Michelin situation in the regions; the converse effect (Michelin stars driving macro-aggregates) is likely to be much smaller. At present we just have a cross-section which means that, in the grand scale of things, what follows are tentative results among the tentative results.
Throughout, the unit of analysis is the Autonomous Region (which I will often refer to as just ‘region’). There are 17 of these in Spain. They are political units with a huge level of autonomy on areas including taxation, education, health, justice, and even police enforcement. In essence, Spain is a federal state with the degree of ‘federality’ (is that a word or just Edglish?) sitting somewhere between Germany and the USA. I got data from the internet (publicly available data), including the name of each restaurant, the number of stars (one, two or three) and the location of the establishment. To achieve this I wrote a crawler using the old good rvest, alongside SelectorGadget. Once the crawler was deployed, it retrieved data from the 226 restaurants in Spain that have been awarded at least one Michelin Star. In addition to this, I downloaded data from the website of the almighty Instituto Nacional de Estadística1 (INE) including the population in the Autonomous Region, regional GDP per capita in 2018, the proportion of the labour force employed in agriculture, industry, services, building, and the proportion of unemployed people in the region. I also wanted to see the extent to which Michelin stars correlate with how innovative the region is. To this end, the INE provides data of the number of `innovative firms’ in the region (their term, not mine). They split this count depending on whether the a firm innovates with “the product” or the “business model”. The measure raises several questions, but since this is pop science, let’s assume it captures something about how innovative people are in the region.
So, first is first: which region has more Michelin star restaurants?. Let’s have a look at this table:
Looking at the mere number of Michelin Star restaurants, Cataluña is the autonomous region with a largest number of such places (46 restaurants). At the other end we find Extremadura, with just 2. But looking at absolute numbers is a waste of time and leads to misleading conclusions. As with all things in life, we need to look at figures per person. Cataluña has over 7 million inhabitants, and is the second most populous region in Spain. If all regions in Spain were identical in all aspects, this autonomous region would naturally have one of the largest number of restaurants (as one colleague put it once: ‘it’s basic algebra!’… yes, the guy meant ‘basic probability’). When we look at restaurants per inhabitant, things change substantially. Since the number of restaurants is tiny in comparison to the number of people, what we are going to do is to present data per-million people. So, we divide the number of restaurants by the number of inhabitants in the region and then multiply the result by 1,000,000. This yields the fourth and fifth columns in the table.
In per-capita terms, La Rioja, with over 15 restaurants per million people, is the winner of this race. It is followed by the Basque Autonomous Region and Cantabria. Both Rioja and Cantabria are tiny in population terms (300,000 and 500,000 people respectively), while the Basque Autonomous Region has only a moderate size (2 million people). So, if we want to explain what is going on, it would appear that size does not matter. If instead of number of restaurants we look at number of stars, things really do not change much (at the tail of the distribution, Madrid starts to look a bit better).
Quality is probably better measured by the average number of stars per restaurant (not all Michelin restaurants are made the same). This metric changes things a bit. Extremadura is then king (queen?), with 1.5 stars. This result, however, is pretty volatile because there are just 2 Michelin stars in the region, so I’d really take this number with a pinch of salt (pun intended -go on, laugh!). Leaving Extremadura aside, the Basque autonomous region has the a largest number of stars per restaurant, followed by Cantabria (again, a volatile result) and now, Madrid.
Overall, then, we can only conclude up to this point that the Basque Country does live up to its reputation (in terms of quantity and quality of its Michelin Star institutions). Cantabria is an underdog which offers good quality and quantity. La Rioja punches above its weight in supply, whereas Madrid is, probably, the other place were you’ll get a most refined experience.
If size does not matter, then what does?
One would thing that, to explain the distribution of stars, we need to look at money. As the analysis below will show, money matters to explain “numbers” but not to explain “quality” (measured as average number of stars). Overall, anyone familiar with Spain looking at the geographical distribution of restaurants will struggle to find a clear pattern. So, what do Michelin Stars correlate with? Well, difficult to say; here we focus on wealth, innovation, composition of the workforce and unemployment. We only have 17 observations, so regression is pointless. Let’s just check the correlation matrix,
In case you are not familiar, in the above figure, each cell in the grid tells you the strength of the association between any two variables. The larger the circle and the stronger the color, the stronger the association. Blue implies positive correlations (if you increase, I increase) whereas red denotes a negative association (if you increase, I decrease). With this in mind, we can look, first, at the number of restaurants. This seems to be positively correlated with GDP per capita (wealth), proportion of people working in industry and the number of innovative firms per capita. The number of restaurants is negatively correlated with unemployment and the proportion of the workforce employed in agriculture and services. There is a negative association with our quality indicator (average stars per restaurant), but the magnitude of the association seems tenuous. So, the wealthier, more industrialised and more innovative the region, the more Michelin Restaurants (and stars).
When looking at our indicator of quality (stars per restaurant), however, things are more interesting. The only relevant association is with our indicator of regional innovation. The correlation is positive, albeit not too strong. However the finding is greatly compelling (if it turns out to be something more than a statistical artifact). It would come to say that, whereas money brings the restaurants, innovation brings the stars. This conclusion is highly attractive and palatable but our data falls short of telling this story convincingly. More work would be required to look into this more seriously.
Ultimately, however, mission has been accomplished. We successfully built a crawler to get the data we needed in order to see where in Spain you’ll find the ‘largest’ number of Michelin Star restaurants. I confess, I did all this only because I wanted to build a new crawler… got a problem with that? The answer is that Rioja, Cantabria, and the Basque Country lead the league, but if you care about the actual number of stars, you probably want to throw Madrid into the mix. We do not know why certain regions have restaurants of higher quality than those in other regions, but there might be something to do with how innovative the region is overall (and this is partly driven by wealth and the amount of industry). However this latter finding demands a lot more of work before establishing this connection with any degree of confidence.
Hope you’ve enjoyed the read and learned something useful.
Que aproveche!
Almighty because, until 2022 was the only institution in Spain that was truly independent of the political powers -alongside Bank of Spain. Pedro Sanchez’s government did not like the statistics on unemployment and inflation being churned by the Institute, so in 2022 -election year minus one- they deciced take over the place by forcing Prof. Juan Rodríguez-Poo out and replacing him with an apparatchik (another coup to the rather shattered institutional independence of the country)… Politicians and their supporters have become a problem… . ↩︎