Facebook friends of Luxembourg

The Grande Région includes the Douro Valley

d'Lëtzebuerger Land du 25.09.2020

International relations are often seen as the domain of sovereign statecraft (delegations, summits, emissaries, missions, and treaties) and measured in national statistics like trade, investment, or migration. It is less visible where people themselves have made connections. Novel data on Facebook friendships reveal new patterns of the Grand Duchy’s connections around the world.

Researchers at Facebook have calculated a Social Connectedness Index (SCI) to measure disproportionally common connections between places. Disproportionally to the two places’ population, that is, as a million-strong city with 2 000 friendships to, say, Luxembourg is probably equally close to us as another of two million with 4 000 links. Our connections would remain equally intense even if we cut the larger city into two equal parts, with a million Facebook users and 2 000 friends here. The raw measure of this intensity would count the number of friendships divided by the number of Facebook users in both areas. This is multiplied by a billion for presentation.

“Place” could mean different things here, from countries down to postal codes in the United States. If we want to see Luxembourg’s relations, it is probably best to focus on areas similar to our country’s size. We could have pockets of friends in larger countries which would not be visible diluted up to the level of the entire country. We have no choice about how Facebook categorized users into locations (think of commuters, jet-setters, or migrants), but in principle, they started from the declared place of residence (on Facebook), with some updating using the device location a user accessed Facebook from. They have used anonymized data from March 2020.

Facebook is making a version of this data public. My interactive map for Luxembourg based on the research version is available online at https://git.io/JUCiS. It shows areas with disproportionally many friends with Luxembourgers in darker colors alongside the SCI numbers. For instance, Luxembourg has the most likely friends within Luxembourg itself, with an SCI of 2.5 million: a 2.5 percent chance that two Luxembourgers are friends on Facebook.

Several patterns are noteworthy from the map, some more surprising than others. First, geography is king: Luxembourg residents have most friends nearby. Friendships are only an order of magnitude less likely with Facebook users from Merzig-Wadern (SCI: 229 685), Trier-Saarburg (445 429), Eifelkreis Bitburg-Prüm (327 335), or Arlon (458 461) than within our borders. Probably because the somewhat larger départements in France dilute the closer friendships around our border, the connections to Moselle (71 389) and Meurthe-et-Moselle (50 468) seem less intense, even less than those to Virton (163 796), Neufchâteau (73 078), Bastogne (178 490), or Verviers (115 105). Many of these connections are also a sign of cross-border work, presumably.

The second general lesson is the importance of a common tongue, though this is less clear-cut for a multilingual place like ours. It is not obvious whether we should expect better connections to the Francophonie or German speakers. But the difference between Walloon or Flemish parts of Belgium likely comes down to linguistic differences on top of geographic distance from our borders. (Though we also have stronger ties to French-speaking cantons of Switzerland than most speaking Schweizerdeutsch.)

A third lesson is immediate from the friendship map: one corner of the world is much closer to us than mere topography suggests. Luxembourg has exceptionally strong ties to Portugal and Cabo Verde (104 331). The bond is particularly close with the north of Portugal (up to 72 287). Still, even with the country’s less well-connected parts (13 841 or more), our connections are stronger than with neighboring countries (apart from our closest neighborhood). Links seem most likely to form (and last) if people migrate, and Luxembourg has many immigrants from these areas as well as a fairly uniform sample from E.U. countries.

Correspondingly, our friendship network in Europe is remarkably uniform otherwise, maybe somewhat weaker with the English, Spanish, Czech, Norwegian and Polish countryside. Some surprising pockets of our friends live in Isernia (10 846), Bari (9 559), and Calabria (5 277) in Southern Italy, or around Innsbruck (12 512), as well as in Montenegro (17 942), Bosnia (5 022), the Raška District of Serbia (13 423), and Peć in Kosovo (7 400). To a lesser extent, many Italian, Greek, or Romanian regions are also closer to Luxembourg than most French or German areas. 

Among major metropolitan regions, our connections to Brussels (16 542) are twice as strong as to Paris (8 002), which are still double those to London (up to 4 755). Other cities you might have expected to stand out, from Milan to Madrid (1 328), from Berlin (3 345) to Frankfurt (3 691), from Rome (1 940) to Copenhagen (2 148), from Amsterdam (3 341) to Stockholm (1 342), from Munich (3 869) to Vienna (3 972), do not have disproportionately many friends in Luxembourg (relative to population). German cities with closer ties include Cologne (5 164), Heidelberg (5 464), Kaiserslautern (7 478), and Freiburg (7 846).

Zooming out, we can find stongs links with some developing nations like Guinea Bissau (up to 46 087) — actually stronger than to most of Belgium. It is equally puzzling how we have exceptional, Portugal-level ties to the Cuvette-Ouest region of the Republic of Congo (57 923). Luxembourg focuses its aid and diplomatic efforts in other countries, only two with remarkable friendship ties: Cabo Verde and Kosovo.

The rest of the world map looks more uniform. There are no apparent strong ties to Latin America, Russia, or the Indo-Pacific, not even to financial hubs or around our embassies. (There is no SCI for Iran or China where it is illegal.) However, we have pockets of friends in Iceland (2 141), Central Africa, Eritrea (up to 3 763), Afghanistan (up to 1 234), or Western Nepal (1 194).

Our ties to many parts of North America are stronger than to other regions outside Europe, with some surprises. We do not have particularly many friends in the corridors of power in Washington, D.C. (942), nor among the masters of the universe in New York (500-600). However, we seem to have some concentration in and to the south of Boston (up to 2 783). If we have friends in the medical establishment there, we should also branch out to the biotech hub and the universities like my alma mater to the north, in Middlesex County (478). There might be some random noise in connections between sparsely populated parts of the States and our not-overly-grand duchy. Yet, it may be more of substance or even a hopeful sign that we have more friends in progressive and innovative San Francisco (961) and Seattle (597, the home of Microsoft and Amazon) than in other major urban hubs like Los Angeles (280), Chicago (255), Dallas (165), or Houston (150). Interestingly, our ties to Cuba (200-700) are not discernibly weaker than to South Florida (300-400), where most Cuban emigrants found refuge.

Modern, data-driven social science has also started to use data like this. These ties’ strength has been found to correlate with trade and investment, which is both a cause and an effect. It could be more surprising that these connections seem to drive the salience and spread of some news more than others, including how people perceive real estate price trends – based on friends’ far-away places! The father of this data, Johannes Stroebel of New York University, presented this work at our department in Luxembourg two years ago. His most recent work with Harvard’s Raj Chetty (disclaimer: my Ph.D. advisor) and our friends at Opportunity Insights shows that the mind-boggling difference between U.S. areas in terms of social mobility is also related to local friendship networks. You can network yourself out of poverty.

Novel data is always fascinating to explore, as there is so much to learn about our world. The lessons might even change how we think about our place in it. If there is any policy conclusion to draw, about who Luxembourgers are and what they want, where the country could leverage our connections and where we might want to reach out more, all the better.

László Sándor has been a research associate at the Department of Finance at the University of Luxembourg. He holds a Ph.D. in Economics from Harvard University. The Social Connectedness Index is available from the Facebook Data for Good Initiative.

László Sándor
© 2020 d’Lëtzebuerger Land