Studies: long questions about the metropolis, with a method of our own

Abstract

Context

Economies do not diversify at random: they enter activities close to the ones they already master. Economic complexity (product space, relatedness, Fitness–Complexity) reads this «adjacent possible», but almost always at country scale. For a metropolitan administration the question is municipal.

Method

From Social Security workers by CNAE-2025 division and municipality (August 2026), specialisation (RCA ≥ 1) is computed for the 295 Catalan municipalities with ≥ 1,000 workers, proximity between 78 sectors by co-specialisation, relatedness density and Fitness–Complexity. A candidate is an explicit rule (no RCA, density in the own top quartile, positive complexity), never a score.

Results

The sector space has 155 links and separates manufacturing, logistics and knowledge families. Barcelona (+3.04) and Sant Cugat (+2.27) lead AMB complexity; Viladecans (−1.68) closes it. Complexity and diversity correlate (Spearman 0.83); the ranking survives the bootstrap (0.975), concrete candidates less so (Jaccard 0.48).

Implications

Candidates are an examination agenda for economic promotion and activity land, not a forecast: they must be checked against land, workforce and firms, and followed month by month to see whether municipalities enter nearby, complex sectors.

Abstract

Context

Innovation districts concentrate research, young firms and skilled jobs in compact, well-connected urban fabric. The literature describes their assets, but metropolitan administrations rarely have an open, place-by-place reading of which assets are already present.

Method

For 197 places in the AMB (65 activity poles, 103 PDU proposals and 29 campuses) the presence of six ingredients is tested with explicit rules: knowledge anchor, startup density, high-value jobs, planned land, rail station and nearby residents. Each answer is yes, no or unknown, with its figure and source; ingredients are counted, never weighted.

Results

57 places meet five or six ingredients, 93 meet three or four and 47 two or fewer. Land (171 places) and nearby residents (173) are almost universal; startup density (49) and high-value jobs (81) are the discriminating ingredients. Only 2 municipalities complete the knowledge–startups–jobs triangle.

Implications

Outside Barcelona, land and transit are already there; young firms and knowledge-intensive jobs are missing. The atlas shows where a single ingredient would change the tier and which indicators to follow every year.

Abstract

Context

The inclusive-growth literature (OECD, Brookings, Lee and Sissons) shows that more output and employment do not guarantee less poverty or better access to housing. The AMB lacked a municipality-by-municipality follow-up built on open data.

Method

Twelve open indicators of growth, prosperity, inclusion and housing are measured as the change over their own window and as a rank among the 36 municipalities. Each dimension is the count of indicators better than the unweighted median; a side scores at half or more. No weights, no mark; a missing value does not count.

Results

12 of the 36 municipalities beat the median on both growth and inclusion; 10 grow without inclusion, 7 include without growing and 7 neither. Income per person grew by a median 34.4% nominal (2015–2023), with 2023 levels from €12,584 to €25,286; employed residents rose 21.6% (2015–2025) and new-lease rent over household income worsened by 1.1 points (2016–2023).

Implications

The monitor shows where growth does not reach the bottom: employment, income and housing policies can be prioritised by the indicators below the median, and the count repeated yearly.

Abstract

Context

Many urban quantities do not grow in proportion to population: income or innovation grow faster (β > 1) and infrastructure more slowly (β < 1). The useful question for a metropolitan administration is how much each municipality’s size would predict.

Method

ln Y = ln Y₀ + β ln N is fitted on the 218 Catalan municipalities above 5,000 inhabitants for nine indicators (the three innovation ones only on the 36 AMB municipalities). Each municipality’s residual (SAMI) is classified with a counted rule per indicator; a 2,000-resample bootstrap, Theil–Sen and the fit without Barcelona serve as checks.

Results

Jobs (β = 1.07) and retail (1.10) are superlinear; total net income (0.97; n = 134), total electricity (0.90) and domestic electricity (0.91) are sublinear. Barcelona sits above its size on 4 of 9 indicators; Badia del Vallès and Santa Coloma de Gramenet below on 6. The classification changes little with the reference set (rank correlation 0.92–1.00).

Implications

Residuals are positions relative to size, not verdicts: they help to prioritise where to look (activity land, energy, R&D capture) and to track whether the gap to the expected value closes.

Abstract

Context

Economic activity estates hold the metropolitan industry and logistics, yet they age, lose land to more profitable uses and are poorly known. The literature on their decline looks for measurable indicators; metropolitan plans need to know where to look first.

Method

Each of the 154 SIPAE estates in the 36 AMB municipalities is read with nine explicit criteria —from occupancy and density to building age, accessibility, transformation pressure and constraints—, each with a public rule giving healthy, watch, at risk or no data. States are counted: no weighted score, and a missing value never counts as good.

Results

Over 6,292 declared hectares, 28 estates have three or more criteria at risk, 114 stay on watch and 12 are healthy. Building age is the criterion that fails most (47 at risk), followed by transformation pressure (33); the median mean age rises from 32 to 40 years between 2015 and 2025. Occupancy is not declared for 42 estates.

Implications

It is a screening to prioritise visits and complete data, not a valuation: the estates with most criteria at risk and the gaps in SIPAE are the first place to act.

Abstract

Context

Royal Decree-law 32/2021 changed the contract modalities from 2022. Metropolitan administrations want to know how hiring has changed in their municipalities, with data that can be updated every month.

Method

SEPE municipal series of registered contracts (2019–2026), by workplace, for the 36 municipalities. Indefinite share of initial contracts, conversions apart; 12 published-month windows; hidden cells as intervals; comparison with calendar 2019.

Results

In the AMB, 10.3% of initial contracts were indefinite in 2019 and 41.9% over the last 12 months (Sep 2025 – Aug 2026). Fewer contracts are registered: 1,260,597 against 1,726,507 in 2019 (−27%). The jump is concentrated between January and April 2022. The share rises more for men than for women.

Implications

The series describes a change in the composition of hiring and does not attribute its cause: fijos discontinuos count as indefinite. It allows each municipality to be followed every month and read beside unemployment, without crossing the two.

Abstract

Context

Metropolitan planning foresees new transformations of industrial land, waterfronts and centralities. The best evidence of what they may bring is the operations already carried out, rarely read with one method.

Method

15 operations from the AMB planning register (9 municipalities, approved 1998–2020). The 2015 and 2023 sections are linked into harmonised units, aggregated with population weights, and the municipality's change is subtracted from the area's, with a 500 m ring. Descriptive, not causal.

Results

Income per person grew more than in the municipality in 10 of the 15 cases (median +3.1 points; from +26.1 in la Marina to −13.4 in Can Sant Joan); population did so in 10. In the rings, income moved within ±3 points of the municipality in 13 cases. 12 cases are measured after their main phase.

Implications

Type, phase and context weigh more than the label. The section-level baseline should be fixed before approval, area and ring tracked yearly, and every analogue read with its caution flags.

Abstract

Context

Plans promise housing and reality delivers it more slowly, yet almost no administration publishes its forecast to look at it again, as the literature on optimism bias and plan conformance recommends.

Method

For the 606 development sectors of the MUC in the AMB, the weighted delivery rate (built = planned − pending) is computed by approval cohort, sector type and urbanisation state; cross-checked against the Cadastre; and a 2026 baseline is registered with a yearly review and a back-test. Unknown is never zero.

Results

In the 362 comparable sectors (80,927 planned homes) the plan data records 10,049 as built (12.4%) and 255 have built none (median 0%). The 198 sectors approved up to 1999 have built 16.2%; those of 2008–2013, 5.3%. The Cadastre/completions ratio since 2007 is 0.93. The baseline expects 37,125 homes in 15 years (2,475 a year) against a 2015–25 pace of 4,343.2.

Implications

In the MUC snapshot (13-02-2024), most comparable open sectors had no home built; a state on a single date cannot tell whether they are stalled or progressing slowly. Planned capacity must be told apart from expectable housing, the expected/actual table published every year with its error, and the baseline never rewritten. The rates are a floor: completed sectors leave the MUC.

Abstract

Context

An already-built metropolis meets new needs by changing the use of what it has: classrooms that empty, premises that never reopen and ageing offices. The literature explains what each reuse decision depends on, not where to start looking across 36 municipalities.

Method

Three lenses with yes / no / unknown rules on open data: public schools with enrolment 2018/2019 → 2024/2025 ≤ −15 % and falling child numbers; Barcelona street runs with ≥ 3 premises empty in both the 2022 and 2024 censuses and low commercial pull; sections with ≥ 2 office parcels meeting 3 of 4 indicators. Criteria are counted, not weighted; an unknown is never a «no».

Results

Of 508 public schools, 94 are candidates (47 with a nursery gap nearby) and 104 remain undetermined. In Barcelona, 2,603 premises were empty in both censuses; of the 246 runs listed, 36 are candidates with 133 persistent empties. Of 734 sections with offices (2,243 parcels), 17 show potential (64 parcels).

Implications

Candidates are places to look at first, not projects: feasibility needs a visit, planning, habitability and an agreement with the owner. They help order a reuse programme and track whether a vacancy persists.

Abstract

Context

When planning allows homes where there was industry or rural land, land value rises and the community is entitled to a share. In Catalonia the protected-housing reserve and the cession of 10 % or 15 % of development rights do so, yet nobody publishes how much uplift the planned transformations of the AMB create or how much of it those obligations collect.

Method

For every unit of the proposals register (582) with homes: static residual value of the residential floor, minus the land value in its existing use, and legal obligations counted at the HPO maximum price. Three scenarios, one-at-a-time sensitivity and a comparison with an uplift-based rule (SoBoN).

Results

Of 348 counted units, 207 have a valued uplift: €4,872m in the central scenario (−2,687 to 15,812). Central legal capture, €5,135m, exceeds the uplift and equals 75,108 affordable homes (59,949–90,276), adding reserve and cession, which may partly coincide (7,026 if the ceded floor carries the plan's mix); in 130 proposals the obligations exceed the uplift. A SoBoN-like rule would give 28,578.

Implications

Obligations set on floor area weigh the same where land gains a lot and where it gains nothing: before raising reserves, residual viability per sector; before quoting figures, new-build prices and industrial land values per estate.

Abstract

Context

Spain has one of the smallest social rental stocks in the OECD and Catalonia set itself in 2007 the goal of 15 % of homes for social policies. Registered demand, protected stock, yearly completions and planning are published separately: no source says how long each municipality would take to house the households that wait.

Method

For each of the 36 municipalities, four quantities per 1,000 households of the 2021 census: registrations in force on the register, HPO homes with protection in force, definitive qualifications per year and protected homes expected from planning within 15 years. Years to cover = registrations ÷ pace; never merged into an index.

Results

In the AMB, 80,384 registered households (62.2 per thousand) against 41,141 HPO homes in force (31.9 per thousand). At 1,372 completions a year (2021–2025 mean) it would take 59–71 years to cover the registrations; in 2025, 36,858 new households registered and 1,791 homes were finished. Pending sectors would add 9,777–19,176 protected homes in 15 years: 12–24 % of demand.

Implications

Neither the recent pace nor planning closes the gap: consolidated urban land, purchases and refurbishment are needed, plus a public series of emergency cases and of public rental stock per municipality. Depending on the assumptions (building pace and share of registrations that is real need), the years to cover range from 11 to 71.

Abstract

Context

Since Kain (1968) and Cervero (1989), the distance between home and work has been read as a sign of metropolitan health: when affordable housing and jobs pull apart, the households with the least slack pay. With rents rising since 2021, the question is whether the AMB is drifting and where.

Method

For the 36 municipalities: 2021→2023 change in self-containment and in the average commute of those working elsewhere (INE matrices, straight-line distance weighted by flows), jobs against resident workers (affiliations) and the rent of new leases (Incasòl). Explicit quadrant rule: rent above or below the median × longer commute or not.

Results

The commuting map barely moves: self-containment from 46.16% to 46.4%, commute from 13.27 to 13.29 km. Volume moves (from 664,941 to 711,114 employed residents working elsewhere) and so does rent (median +14.14%). Six municipalities are «drifting» and three stay unclassified; in 23 jobs grow faster than resident workers.

Implications

Drift is localised: affordable housing and employment land where rent rises and commutes lengthen, with annual monitoring of self-containment, the jobs/workers ratio and rent. Quadrants are associations, not causes.

Abstract

Context

Since 2009 Barcelona has seen housing ownership by legal entities grow while rents rose and tourist accommodation expanded. The literature on financialisation and tourism gentrification describes a chain between them, yet few open sources allow it to be followed neighbourhood by neighbourhood.

Method

Panel of 73 neighbourhoods, 2009–2026: share of premises held by legal entities (Cadastre), new-lease rent (Incasòl), population register and HUT. Cross-correlations of yearly changes at lags −3 to +3 years, a cluster bootstrap over neighbourhoods and an ordered-rule typology. Descriptive only.

Results

The city’s legal-entity share goes from 12.3% to 15.3%; the median neighbourhood rise (2014–2025) is 1.6 points. The correlation between the share’s rise and the rent rise one year later is 0.22 (0.15–0.30), but the asymmetry between directions (0.05; −0.03 to 0.15) includes 0. Population tends to grow after the share (asymmetry 0.33). 14 neighbourhoods combine a corporate rise with many HUT; in 8 the share rises before rent.

Implications

Lagged associations do not prove cause, but they say where ownership, new leases and HUT licences should be watched together. An HUT series by neighbourhood and ownership by owner size would most improve monitoring.

Abstract

Context

Municipal and county limits in the Barcelona metropolitan area do not match the geography of daily life. From the nodal regions of Nystuen and Dacey to the OECD functional urban areas, the question is the same: which municipalities form one commuting market?

Method

The 36 AMB municipalities are the nodes of a directed network weighted by weekday trips (MITMA, October 2025). Directed modularity is maximised with the Louvain heuristic (γ = 1) and the resulting partition is audited with shuffled labels, edge noise, a resolution sweep and the INE residence-by-workplace and study-place matrices (2021–2024).

Results

Five basins emerge (Q = 0.28; 8.1 standard deviations above chance). 59% of trips never leave the municipality and 87% of the trips of the Barcelona basin (2.37 m inhabitants) end inside it. The counties explain the flows worse (Q = 0.25; NMI with the basins 0.63). Under noise the partition holds (mean NMI 0.87) and the 2023 INE census reproduces the structure at fine resolution (NMI 0.66 at γ = 1; 0.82 at γ = 1.25).

Implications

Basins are a reasonable unit for planning transport, services and fares beyond the municipality, always as a robust grouping and never as a perimeter. Their stability across sources and years makes them usable as a yearly indicator.

Abstract

Context

Between 2023 and 2025 the TMB metro gained 9.1% riders and FGC in the AMB 8.8%, but very unevenly across stations. How much is general recovery, how much line service and how much urban change around the station?

Method

Monthly panel of 185 stations (36 months) and station and line × month fixed effects on ln(riders) for the 111 Barcelona stations with known exposure (3,989 observations): newly licensed housing within 800 m, accumulated to 12 months earlier, per 1,000 residents. Station-clustered errors, a placebo and a decomposition of growth into network, line and station.

Results

Per standard deviation of exposure a station grows 1.35% more (95% CI −1.02 to +3.77): positive but not distinguishable from zero, like the placebo (−1.23%). The line explains 22% of the variance of growth on the metro; the local component is uncorrelated with the part attributed to new housing (r = 0.08).

Implications

With granted licences rather than occupied dwellings, a station’s growth is better read through the service of its line than through nearby building works. Licence lists should be open in every municipality, with a service history per station; the associations are not causes.

Abstract

Context

A radial network serves the centre well and leaves tangential trips without a competitive alternative. The literature measures that gap as the ratio of public-transport to car time; the AMB lacked a pair-by-pair reading.

Method

For each pair of the 102 MITMA district zones (October 2025 weekday) the scheduled 07:00–09:00 time is compared with a declared reference car. A counted rule (no competitive PT, or ≥ 2.5× the car with low frequency or ≥ 2 transfers) flags the pair; directions are added into corridors.

Results

Of 8,025,525 daily trips between zones, 3,174,140 are classified and 426,510 (13.4%) fall on pairs with insufficient service. On 107 corridors; the top 10 hold 43.5%. No pair inside Barcelona meets the rule (median 1.58×); in the rest of the AMB it is 2.15× and 72% of exposed trips are there. Other thresholds give 11.2 to 21.7%.

Implications

It is a priority map for tangential lines, frequency and transfers, not a count of lost riders, and an upper bound: 9 municipalities, among them Sant Cugat del Vallès, have no urban bus in the data and hold most of the exposed trips. Local service and modal share must be checked there before anything is sized.

Abstract

Context

Planning in the 36 AMB municipalities enables 70,914 homes still to build in 606 sectors, which will seal soil upstream of a network with 772 inventoried overflow points (187 inside the T100 flood zone). No open sewer catchments exist and rain-tank capacity is not published.

Method

Sectors, overflow points and tanks are assigned to 16 proxy drainage units (nearest main river or stream, 100 m grid) and to the municipality. Five criteria are counted per unit (home, sealed-land and overflow intensity against the AMB; homes in T100; a register showing no tank) as met, not met or unknown, without weights.

Results

49,047 pending homes (69%) fall in proxies meeting three or more criteria; the Llobregat holds 35,560 with 4 of 5. 38 sectors overlap T100: 8,583 known pending homes, and 11 of those sectors have no pending figure (3,408 planned), not counted as zero; 167 sectors have an unknown condition. New sealed land: 1,065 ha (737–1,392). Outside Barcelona, 63,806 homes (90%) have no retention register at all.

Implications

It prioritises where to demand hydraulic studies and drainage conditions before approving sectors, not where an event will occur. Indicators: published tank capacity, a register beyond Barcelona and overflow frequency.

Abstract

Context

The plug-in fleet could multiply by 2 to 7.5 times by 2030, and those who cannot charge at home depend on the street. For a metropolitan administration the question is threefold: how many vehicles will need charging, where the people without a garage live and where the energy will come from, at the scale of the municipality and the census section.

Method

With the DGT fleet (2015–2025) of the 28 municipalities with published fuel type, three arithmetic rules project the plug-in stock to 2030 and the known public power, added up charge point by charge point, is compared with the AFIR yardstick. In 2,151 census sections, three Cadastre criteria give a likely-garage indicator which, crossed with the distance to the nearest public point (400 m), marks the priority sections. Unknown is never zero: 8 municipalities stay outside the classification.

Results

The AMB had 33,198 plug-in cars in 2025 (3.0% of the fleet); in 2030 it would have between 69,974 and 249,719 depending on the rule. The 3,785 public charge points (at 1,228 sites) exceed the yardstick today, at 203%, but with no new points they would cover 40% in 2030 (trend scenario) and only 3 municipalities would reach it; 12 municipalities are under “pressure” and 13 in a “latent gap”. 235 sections (368,541 residents) are priority. Suitable roofs would yield 24 times the 2030 charging energy.

Implications

Public charging should be examined first in the priority sections; the AFIR yardstick gives a scale for concessions even though it does not bind the municipality; and solar is a problem of hourly timing and ownership, not of quantity. No figure is a forecast: they are declared rules that can be followed year after year.

Abstract

Context

In the AMB, forest and scrub meet the city at Collserola, the Garraf, the Ordal and the Marina range. Catalan rules require protection strips around housing estates, but no public source says how many people live right next to forest or how long a fire engine would take to get there.

Method

For each Cadastre parcel within 0–25 m or >25–100 m of forest or scrub of at least 1 ha (ICGC 2024 land cover), dwellings, construction year and modelled driving time from the nearest professional station on the OpenStreetMap network. Residents, 65+ and 0–14 are section figures (INE 2025) shared by dwellings. Each figure is given separately.

Results

Within 0–25 m of forest there are 40,154 dwellings and about 104,131 residents, of whom about 17,763 are 65 or older. Half of these dwellings have the professional station within 10.3 min according to the model, and 51.7 % are beyond 10 min. Corbera de Llobregat reaches a 23.8 min median. 20,301 dwellings in the band are in buildings from before 1980.

Implications

Proximity to forest and distance to the fire station do not peak in the same places: the south-western municipalities, served by volunteer stations, need their own reading of prevention, access and self-protection, and real response-time data should be published.

Abstract

Context

Composite indices summarise well but hide judgements: the choice of weights, the compensation between opportunity and risk and the treatment of what is not known. A metropolitan public decision (36 municipalities, 2,151 census sections, a master plan at initial approval) needs every step to be visible, debatable and repeatable.

Method

Xamfrà fixes eight principles shared by every view and study: counted criteria under Kleene's three-valued logic, unknown ≠ no ≠ zero, whole census sections, a building's colour as context, a provenance and licence manifest per dataset, published sensitivity, local AI with finite plans and a work area independent of the camera. Eleven method families adapt international references to the AMB on one shared pipeline: source → staging → validation → manifest → view or study → print.

Results

The methods paper (v1.0) describes 11 families and, for the AMB, 65 activity poles, 1,626 transformation units (191 with ≥ 4 of 6 opportunity criteria, only 18 with no risk signal) and 25 % of pending homes within a 600 m walk of a well-served station, against 60 % of existing ones. Back-tests of the housing pace miss by 26 to 63 % when the cycle turns.

Implications

Every study in the section declares a named method that describes only what the code does; the principle × study matrix says where each principle is central, partial or not applicable. Where other tools give a score, Xamfrà gives the list of conditions that would justify it and leaves the weighting to the decision-maker.