somewhere — else.org  /  Systems A falsification
The word
España
vaciada

The emptied interior. Spain’s decade-old vocabulary for rural depopulation.

The pattern
Greek
islands

Where overtourism and population loss actually coincide on the European map.

The Vaciada Paradox

Everyone assumes the tourism map and the depopulation map of Europe are the same map. They are not — and the word for the problem belongs to the wrong country.

In Spain, the word vaciada has carried a specific weight for more than a decade. It names a phenomenon people can feel: the emptying of the rural interior as the young leave for the cities and the coasts, and as the towns they leave behind shrink, age, and slip off the map one shuttered school at a time. It is a powerful word, and powerful words travel. Somewhere along the way, la España vaciada fused with a second European anxiety — overtourism, the sense of places being loved to the point of being hollowed out — until the two became, in the popular imagination, a single story. Tourism fills the coast; the interior empties. One map.

It is an intuitive pairing. You can picture it: the cruise-ship harbours and the abandoned hill villages as two ends of the same imbalance, locals priced out of one place and pulled toward another. The pairing is so natural that it is rarely tested. So I went looking for the overlap — the regions where Europe is simultaneously most saturated by visitors and most drained of residents — expecting the two maps to fall on top of each other.

They don’t.

01 — The number

Across the 277 European regions where the data holds steady from 2018 to 2024 — NUTS-2 units, the mid-level statistical regions Eurostat uses, each carrying a measure of tourism intensity and a measure of population change — the places that are both heavily touristed and losing population come down to a count you can hold in one hand.

Four regions. Not one of them is in Spain.

That is the whole finding in miniature, and the rest of this piece is the work of earning it: showing that the overlap is genuinely small, naming the four regions that make it up, and then explaining why Spain — the country that gave the phenomenon its name — is almost entirely absent from the pattern it is supposed to define.

Test 01 — the assumed correlationn = 277
If the two maps were one, this would slope. It doesn’t.
r = 0.05
Tourism intensity vs. population change
across all 277 NUTS-2 regions.
Statistically indistinguishable from zero.
Scatter of tourism intensity against population change across 277 European regions, showing no correlation
Each point is one European region. A real coupling between tourism and emptying would pull the cloud into a diagonal. Instead it sits as a shapeless mass: knowing how touristed a region is tells you essentially nothing about whether it is gaining or losing people.

A correlation of 0.05 is the statistical equivalent of a shrug. At the scale of the continent, tourism intensity and depopulation are not two readings of one underlying force. They are simply different things happening in different places. The intuitive single map is, as a Europe-wide claim, false.

But a null result is the easy part — it closes a question without opening one. The clean zero hides something more specific underneath, and finding it means letting the regions sort themselves rather than forcing them onto a line.

02 — The four Europes of tourism

Group the regions by how they actually behave — clustering them on tourism intensity and population trajectory together, and letting the structure fall out rather than imposing it — and Europe sorts into four broad kinds of place, with a small fifth tier that refuses to sit quietly in any of them.

Test 02 — the clustersk = 4 + outliers
Where tourism and emptying finally meet
Scatter of all European regions with the four tourism-and-declining outlier regions highlighted in oxblood, far out on the tourism axis and below zero population change

The four base clusters are ordinary combinations — touristed-and-growing, quiet-and-growing, quiet-and-shrinking, moderately-touristed-and-stable. The story is in the tier that breaks off from all of them:

EL42
South Aegean
Greece · Rhodes, Kos, the Cyclades
EL62
Ionian Islands
Greece · Corfu, Zakynthos, Kefalonia
EL43
Crete
Greece · the largest Greek island
HR03
Adriatic Croatia
Croatia · the Dalmatian coast
The four-region tier where intense tourism and population loss share the same footprint. Three are Greek island groups; the fourth is the Croatian Adriatic. The data shows co-occurrence, not cause: these are places where a seasonal visitor monoculture and demographic shrinkage happen to coincide, not places where one is shown to drive the other. The overlap the popular story promises is real — it is just small, and it is Mediterranean-island, not Iberian-interior.

The map where tourism and emptying coincide does exist. It is four islands and a coastline in the eastern Mediterranean — and Spain is nowhere on it.

These four are not a typology — they are too few for that, and their rarity is the point. Out of nearly 280 regions, only this handful sits where intense tourism and genuine population loss share the same ground: small, seasonal economies where a summer flood of visitors and a slow loss of year-round residents occupy the same footprint. They are not a broad European trend. They are an acute, hyper-specialised condition, and there are four of them, clustered in one corner of the continent, speaking Greek and Croatian rather than Spanish.

Before going further, the obvious objection has to be met head-on, because a reader who knows Spain will already be raising it. The vaciada story is village-scale: it is about hamlets of a few dozen residents emptying out across Soria, Teruel, Zamora. This analysis works at NUTS-2 — the macro-regional grid Europe uses to allocate cohesion funds — where a single growing city like Zaragoza or Valladolid mathematically swallows the emptying countryside around it. At that resolution, Spain’s twin crises genuinely never touch; the grid is too coarse to let them. But that cuts the other way for the islands. The reason the South Aegean and Crete and the Croatian Adriatic surface here is that their overlap is large and cohesive enough to survive macro-aggregation — to register even when the map is drawn at its bluntest. Spain’s signal vanishes into the grid; the island signal is strong enough that the grid cannot hide it. The coarseness that erases one is exactly what makes the other remarkable.

03 — The Spanish twist

So where is Spain? This is where the falsification turns from a tidy null into something with an edge — because Spain is not missing from the data for lack of either tourism or depopulation. It has world-leading quantities of both. They simply do not happen in the same regions.

Spain’s tourism saturation is coastal and insular: the Balearics, the Canaries, the Mediterranean costas. Its depopulation is interior and peripheral: Castilla y León on the northern meseta, Asturias on the Atlantic coast, Extremadura along the Portuguese border — the regions where, even averaged across whole NUTS-2 units, the population line still points down. The crowds and the silence sit in different parts of the country, often hundreds of kilometres apart. At the regional scale the two Spanish maps don’t merely fail to correlate; they point in opposite directions — though, as the numbers show, the growth side is loud and the decline side is quiet, and that asymmetry is itself telling.

Test 03 — inside Spain2018–2024 Δpop
Two Spains, pulling opposite directions
Illes Balears — tourism core+7.9%
Comunitat Valenciana — costa+7.3%
Canarias — tourism core+5.0%
Castilla y León — interior−1.0%
Principado de Asturias — northern periphery−1.7%
Extremadura — interior−1.7%
The contrast is real but lopsided. The touristed cores are growing emphatically — five, seven, eight percent over the window. The declining interior regions are slipping far more gently, around a point or two. The single phenomenon the word vaciada implies is, inside Spain itself, two phenomena in two geographies — and at this regional scale the decline side is muted, for reasons the next paragraphs make clear.

That muted decline is not the analysis failing to find the emptying — it is the resolution hiding it. The true vaciada heartland is sub-regional: provinces like Soria, Teruel and Zamora, some of the least densely populated territory in Europe, each emptying hard. But NUTS-2 bundles them into larger units alongside a provincial capital or a growing corridor, and the average comes out flat or only slightly negative. The village-scale catastrophe is real; it simply happens below the grid this map is drawn on. The interior reads as a gentle minus here precisely because the regional lens blurs the sharpest losses into their surroundings — the same blurring that, at the European scale, let only the largest and most cohesive overlaps survive into the four-region tier.

There is a second-order point hiding in the concentration figures, too. When you rank the larger European countries by how unevenly their tourism is distributed across regions — a Gini measure, restricted to countries with enough regions for the comparison to mean anything — Spain comes out near the very top. Its tourism isn’t just coastal; it is concentrated, piled into a handful of regions while the interior sees almost none. The structural reason the two Spanish maps don’t overlap is baked into how unevenly the visitors arrive in the first place.

Spain owns the word. Greece, mostly, carries the pattern.

04 — What the test does and doesn’t show

None of this says the España vaciada literature is wrong. It describes a real, serious, well-documented demographic story about the Spanish interior, and nothing here touches it. What the data falsifies is narrower and more specific: the popular conflation that has flowed out of that literature — the assumption that tourism saturation and depopulation are a single European phenomenon, one map readable from either end. At the scale of the continent, they are not. Where they do overlap, they overlap in four small Mediterranean island and coastal economies, not in Spain.

The remaining limits are worth stating plainly. The clustering is a structure-finding tool, not a law; a third behavioural feature, currently under review, could sharpen the tiers further, and the exact count of the outlier tier depends on where the threshold is drawn — “four” is the result of a defensible cutoff, not a law of nature. The tourism measure is a multi-year mean built to smooth post-pandemic volatility, which trades some recency for stability. And as the Spanish case makes vivid, the whole analysis lives at one resolution: it can tell you where tourism and emptying coincide as regions, not where they collide street by street. The four-region result is robust to the ways I tried to break it, but it is a claim about macro-regions and recent years, not about every place and all time.

What survives all of that is the shape of the thing. The maps are not the same map. The country that named the problem is largely absent from the version of it that involves tourism, and the version that does involve tourism lives somewhere else — four islands and a coast that never got a famous word of their own.

The vaciada paradox is not that Spain is emptying. It is that the word did, and the pattern didn’t.


Sources & method
Eurostat tour_occ_nin2 (regional tourism nights) and demo_r_d2jan (regional population stocks), NUTS-2, 2018–2024. Tourism intensity is a 2022–2024 mean to smooth post-COVID volatility; population change is computed over the boundary-stable region set (277 of 288 regions; 11 dropped for boundary or data-lag reasons, documented in the notebook). Clustering: k-means (k = 4) with a rule-based split on the hyper-tourism cluster, plus a four-region outlier tier.

Full analysis notebook, boundary diagnostic, and methodology: github.com/seanjaya/vaciada-paradox