Where housing outruns local incomes: a comarca-level analysis

Summary of findings

• The affordability spread is wider than the price spread suggests. Measured in years of average per-resident disposable income for an 80 m² home, the range runs from 3.2 years in Garrigues to 15.1 years in Cerdanya, with Barcelonès (14.4) and Aran (14.1) close behind.
• Local income does not protect against unaffordability. Prices do rise with income (rank correlation 0.50), but the burden itself barely follows it (0.31 raw), and at equal density the association essentially vanishes (0.06). Where prices are high, they have outrun the extra income.
• Tourism pressure is the dominant driver of the purchase burden once urbanization is held constant: raw 0.35, but 0.74 at equal density. This is suppression, the mirror image of the usual story: tourist comarques are rural, and rural is cheap, which masks the effect in the raw data.
• Buying and renting answer to different forces. The gap between a comarca's purchase-burden rank and its rental-burden rank follows tourism at 0.79, the strongest coefficient in this report. Rents instead track density (0.78): the rental burden peaks in the metropolitan belt, the purchase burden in the tourist mountains and coast.

The question and the data

The outcome variables are two burdens. The purchase burden divides the registered transaction price per m² (Idescat, 2023 to 2024, the same figure behind our housing price ranking) by the average disposable income per resident (Idescat, 2022), expressed as years of that income for an 80 m² home. The rental burden divides a year of the comarca's average contract rent (Incasòl deposit registrations, 2024) by the same income. Both are available for 42 of 43 comarques; sources and coverage are on the methodology page.

Note what the denominator is: disposable income per resident, not per household. A household usually pools more than one income, so the "years" are not literal years of saving for any real family; they are a comparable ruler across comarques, not a mortgage forecast. Our live affordability index asks the same question with the current Idealista asking price as numerator; this report uses registered sale prices because they are the measured version of what buyers actually paid.

The candidate drivers come from the rest of the dataset: population density (the inverse of most rural), distance to Barcelona (commute ranking), tourist beds per resident (least touristy), distance to the beach, median age, total population and income itself (highest income).

Method: rank correlations and what counts as evidence

All associations are Spearman rank correlations, for the same reason as in our crime report: comarca metrics are heavily skewed (the Barcelonès has roughly twenty times the median density; Pallars Sobirà has 179 tourist beds per 100 residents against a median of 12), and rank correlation is immune to those distortions.

To hold a third variable constant we use partial rank correlation: every variable is converted to ranks, the two ranks of interest are regressed on the control ranks, and the residuals are correlated. The controls here are population density (urbanization) and, in a second specification, distance to Barcelona (the metropolitan gradient). With n = 42, a rank correlation is statistically significant at p < 0.05 (two-sided) when its absolute value exceeds roughly 0.31.

One warning the crime report did not need: controls can also REVEAL an association, not only dissolve one. When a driver is negatively entangled with a confounder (tourist comarques are rural; rural comarques are cheap), the raw correlation understates the effect. Statisticians call it suppression, and it is the central pattern of this report.

Everything here is observational, ecological (comarca-level aggregates, not individuals) and mixes reference periods (income 2022, prices 2023 to 2024, rents 2024). Correlation, with or without controls, is not causation; the limits section spells out exactly which statements the data supports.

Hypothesis 1: expensive housing is simply where incomes are high

Only half true, and the half that fails is the important one. Prices do track incomes across comarques (0.50), but if prices merely followed local purchasing power, the burden (price divided by income) would be flat. It is nothing like flat: it spans 3.2 to 15.1 years of income, and its own correlation with income is a weak 0.31 raw, essentially zero at equal density (0.06).

Figure 1 shows why. Cerdanya and Garrigues sit at similar income levels, yet the price per m² differs several times over. The dispersion around the trend line, not the trend line itself, is where the story lives: something other than local income is setting prices in the expensive comarques.

Hypothesis 2: the metropolitan pull sets the burden

For renting, yes. The rental burden correlates with density at 0.78 and with distance to Barcelona at −0.49: a year of average rent absorbs 61% of one resident's average disposable income in Barcelonès (average contract rent 1,081 euros a month), 58% in Garraf and 54% in Baix Llobregat, against 28% in Terra Alta and 30% in Ribera d'Ebre.

For buying, the metropolitan gradient is real but much weaker: 0.49 against density and −0.27 against distance to Barcelona. The purchase burden's extremes are not metropolitan at all: Cerdanya and Aran are mountain comarques. Something else dominates the purchase market, which is the next hypothesis.

Hypothesis 3: tourism and second homes decouple prices from local incomes

This is the report's central finding, and the raw data actively hides it. The raw correlation between tourist beds per resident and the purchase burden is a modest 0.35. But tourist comarques are overwhelmingly rural, and rural comarques are cheap, so the two effects cancel in the raw number. Hold density constant and the association jumps to 0.74, the strongest driver in the table by a wide margin.

The effect is robust in every direction we probed: 0.74 when distance to Barcelona joins the controls, 0.76 when income does, and a leave-one-out range of 0.72 to 0.76 across all 42 deletions, so no single comarca carries it.

The mechanism is no mystery: in Cerdanya (56 tourist beds per 100 residents), in Aran (92) and on the Empordà coast, buyers of holiday and second homes bring outside incomes into a small local market, and the registered prices those purchases set are then divided by the LOCAL income in our ratio. Tourist beds are our measurable proxy for that outside demand; pure second-home comarques with few commercial beds are underweighted by it, which biases this coefficient down, not up.

Hypothesis 4: buying and renting are different markets

Strongly supported, and it is the sharpest number in this report. Rank every comarca twice, once by purchase burden and once by rental burden, and take the difference. That gap follows tourism at 0.79: in the tourist mountains, buying is expensive out of all proportion to renting (Terra Alta tops the list, followed by Ripollès and Alta Ribagorça), while in the agricultural Lleida plain (Segrià, Pla d'Urgell) the pattern reverses.

The reading is intuitive: second-home buyers compete in the purchase market but mostly not in the year-round rental market, so tourist pressure inflates sale prices more than contract rents. At equal density, tourism does push rents too (0.47), but with roughly half the force it exerts on purchases (0.74).

For a household moving to a tourist comarca, the practical asymmetry matters: renting there is systematically less punishing, relative to the rest of Catalonia, than buying.

Correlations that dissolve under controls

The table below collects every driver we tested against the purchase burden. Three impressive raw correlations turn out to be geography in disguise. Beach distance (−0.35 raw: closer coast, heavier burden) vanishes to 0.04 under the density control: the coast is expensive where it is urban or touristed, not because of the sand itself. Median age (−0.36: younger comarques cost more) collapses to −0.11, and total population (0.40) to −0.16: both were density wearing different clothes.

What this data cannot tell you

The denominator is income per resident, not per household, and Idescat's disposable-income figure (2022) predates the price data (2023 to 2024) and the rent data (2024). Both choices are forced by what exists at comarca level; both mean the "years of income" are a comparative ruler, not a savings plan. If incomes grew unevenly after 2022, the ratios move.

The numerator mixes housing stock. Registered prices per m² average over whatever sold that year: a rural comarca's sales lean toward large old houses, a metropolitan one's toward flats. None of it is quality-adjusted, so a comarca's burden can shift when its sales mix shifts, without any home changing price.

Second homes are not directly measured. Tourist beds per resident counts commercial accommodation; a comarca of pure second homes with few hotels is invisible to the proxy. This biases the tourism coefficients toward zero, so the true outside-demand effect is, if anything, larger than reported. And the usual caveats hold: 42 aggregate observations, one period, no causal identification.

Conclusions

Ranked by how firmly the data supports them: (1) Housing affordability in Catalonia varies almost fivefold, far more than incomes do, so prices, not wages, set the map. (2) The rental burden follows the metropolitan economy: density and distance to Barcelona order it almost completely. (3) The purchase burden follows outside demand: tourism pressure is its dominant driver once urbanization is controlled, an effect the raw data suppresses rather than exaggerates. (4) The two burdens diverge exactly where tourism concentrates, which makes renting, relatively speaking, the cheaper way to live in a tourist comarca.

For a household choosing where to live, the practical reading: the affordability index tells you where local incomes still buy local homes, and this report tells you why its worst performers are not the big city but Cerdanya, Aran and the second-home coast.

Explore the data behind this report

Every variable used here has a public page: housing prices, the affordability index, household income, tourism pressure, population density and distance to Barcelona. Sources, periods and coverage for every dataset are on the methodology page.