Sources and method
How neighbourhoods are compared, where the data come from, what was tested and what the site does not measure.
1.In short
The site does not grade neighbourhoods. It places each one against the others, theme by theme, and shows what can be reached from home. All the data are public and open.
2.Neighbourhoods
A neighbourhood, here, is an IRIS: INSEE's smallest statistical unit, of about 2,000 residents. Paris and the three inner-suburb départements have 2,752 of them.
Neighbourhoods with fewer than 50 residents (woods, business parks, railway land) are shown but not used in comparisons: their figures would be too fragile.
3.Living conditions
For each theme, neighbourhoods are ranked from least to most affected. A neighbourhood is "in the most affected quarter" if it is among the 25% most exposed.
- Heat
- Heat vulnerability, sealed ground, cool green spaces, cool places within 400 metres.
- Air pollution and noise
- Combined exposure to air pollution and noise, from the Airparif and Bruitparif map.
- Housing
- Share of homes rated F or G, and how strongly electricity consumption rises with the cold.
- Access to care
- GPs and pharmacies reachable on foot or by public transport, relative to the number of residents who can also reach them.
The combined score counts the number of themes, out of 4, in which the neighbourhood is in the most affected quarter. "Highly exposed" means at least 2 of the 3 exposures (heat, air and noise, housing) in the most affected quarter; access to care is a separate axis.
The sentence "Out of 10 neighbourhoods, 8 are less exposed than yours" rounds the ranking to the nearest ten.
Example: Économie 1 neighbourhood, Drancy
- Heat: unfavourable
- Air and noise: unfavourable
- Housing: unfavourable
- Access to care: not unfavourable
3 out of 4
The score is a count, not an average. An average smooths differences away: a very hot but quiet neighbourhood would get the same mark as a moderately hot, moderately noisy one. Counting shows the neighbourhoods where difficulties add up.
Current distribution of the 2,752 neighbourhoods
| Score | Neighbourhoods | Share |
|---|---|---|
| 0 out of 4 | 774 | 28.1% |
| 1 out of 4 | 1,287 | 46.8% |
| 2 out of 4 | 608 | 22.1% |
| 3 out of 4 | 79 | 2.9% |
| 4 out of 4 | 4 | 0.1% |
Sources: Underlaid calculation from the sources above · data as of 1 October 2026
4.Access to care
Being next to a practice is not enough if the GP is already overstretched. The indicator therefore accounts for both distance and demand: each GP and each pharmacy is "shared" among the residents who can reach it on foot and by public transport.
The method follows the DREES's localised potential accessibility indicator (APL), whose weights it uses, applied to a grid of 200 m squares. Facilities without on-site consultations (teleconsultation, home visits, emergency departments) are excluded; the list is published.
Sources: directory of health professionals (RPPS, Agence du numérique en santé), list of pharmacies (FINESS), INSEE Filosofi 2021 gridded population, Île-de-France Mobilités timetables, OpenStreetMap walking network.
5.What is within reach
For each need, the site computes the duration to the nearest place, on foot and by public transport, from the neighbourhood's inhabited squares, on a Tuesday at 10:00. The value shown is the neighbourhood's population-weighted median.
- No constraint
- Walking at 4.5 km/h, on the whole public transport network.
- Slow walking
- Same network, but walking at 3.4 km/h on every walking part: from home, to the stop, in connections and to the destination. It is the usual walking speed measured in people aged 80 and over (Bohannon and Andrews, 2011). This profile can also give an idea of a trip with pain, late in pregnancy or with a young child walking alongside.
- Wheelchair
- Only stops and stations declared step-free are used. The calculation sees neither pavements, slopes, broken lifts nor the help sometimes needed to board a train: real durations are often longer.
These durations are for information: they are not part of the combined score. At night (departure at 1:00), only emergency departments and stations are shown, always on foot and by public transport.
| Place | Source | Good to know |
|---|---|---|
| Crèche (public or private) | INSEE BPE | The base only lists facilities receiving a CAF service grant; it mostly matches crèches on the CAF scale, where fees depend on income, whoever runs them. Micro-crèches whose fees do not depend on income are largely missing. About a two-year lag. |
| Convenience store or supermarket | INSEE BPE | The supermarket category includes non-food shops of the same size, which the base cannot separate; grocery shops under 120 m² are not counted. |
| Public nursery school | INSEE BPE | The nearest school is not necessarily the catchment-area school. |
| Park or garden | L'Institut Paris Region | Green spaces open to the public. Entrances are unknown, so the distance is measured to points every 100 m along the outline. |
| Public toilets, fountains | City of Paris | Paris only: most inner-suburb communes do not publish an open list. Toilets in service; in a wheelchair, the nearest one marked accessible; toilets open 24 hours are shown separately. |
| Emergency department | INSEE BPE | General emergency departments only; paediatric and specialised ones not counted. |
| Town hall, France Travail, post office | INSEE BPE | One town hall per commune (per arrondissement in Paris), annexes not counted; local France Travail agencies; post offices, without postal relays. |
| France Services, CAF, CPAM | ANCT; public administration directory | Fixed sites (mobile buses excluded); offices open to all insured people (sessions at partners' premises excluded). |
| Police station, social centre, library | INSEE BPE | National police open to the public (municipal police excluded); approved social centres; local-authority libraries. |
The places set aside by these rules are listed one by one, with their reason: list of excluded places
6.Residents' resources
Median income, overcrowded homes and the share of second homes describe the means residents have to cope. They are shown separately, for information, and are not part of the combined score: a well-off neighbourhood can be highly exposed to heat. Neighbourhoods are split into three thirds, from lowest to highest.
Since 3 October 2026, the 2022 census replaces the 2021 one. INSEE changed its definition of overcrowding: the median rate goes from 13.5% to 24.6% because of this change of definition, not of a real change. About one neighbourhood in five changes third, always to a neighbouring one. The results already published were checked with these new data (part 7).
Sources: INSEE, Filosofi 2021 (income, latest vintage published at neighbourhood level) and 2022 population census (overcrowding, second homes, population).
7.Hypotheses and results
Before looking at the results, the hypotheses to test were written down and dated, with the rule that decides the verdict. This avoids looking, after the fact, for the story the figures seem to tell.
- Supported
Neighbourhoods whose residents have the fewest means have more difficult access to care.
True in Hauts-de-Seine and Val-de-Marne, including at comparable density. The rule fixed in advance does not allow a conclusion in Seine-Saint-Denis, where almost no neighbourhood is in the metropolis's third with the most means.
Check with the 2022 census: confirmed. The quarter with the most difficult access to care holds 39.3% of the neighbourhoods with the fewest means against 24.0% of those with the most in Hauts-de-Seine, 38.8% against 27.9% in Val-de-Marne; gap over the metropolis 22.9 points (29.4 with the 2021 census).
Hypothesis written on 1 October 2026
- Refuted
Neighbourhoods whose residents have the fewest means lose more access to care in a wheelchair.
No: the loss of access in a wheelchair is concentrated in Paris, where few metro stop points are accessible, whatever the neighbourhood's means. In number of reachable GPs, wheelchair access remains lower in the inner suburbs.
Check with the 2022 census: confirmed. The loss of access in a wheelchair remains concentrated in Paris (52.3% of the neighbourhoods with the fewest means in the most affected quarter, against 66.2% of those with the most); over the metropolis, 13.2% against 35.0%.
Hypothesis written on 1 October 2026
- Refuted, and reversed
Neighbourhoods whose residents have the fewest means are more often far from public services (town hall, France Services, CAF, CPAM, France Travail, post office).
The reverse holds in Hauts-de-Seine, Seine-Saint-Denis and Val-de-Marne, and the result is the same in a wheelchair.
Share of neighbourhoods in the quarter farthest from public services (on foot and by public transport, Tuesday 10:00), by residents' means. Thirds of means computed within each département. Département Fewest means Most means Paris 1.7% 3.8% Hauts-de-Seine 27.2% 48.2% Seine-Saint-Denis 25.4% 45.5% Val-de-Marne 25.1% 46.2% Neighbourhoods with the most means are more often far from public services, including at comparable density in Hauts-de-Seine and Val-de-Marne. The data do not establish why.
This indicator measures travel time on foot and by public transport. It does not take into account the need to use these services, car use or online procedures.
Check with the 2022 census: confirmed. Share in the farthest quarter, fewest means against most: Paris 1.7% against 6.2%, Hauts-de-Seine 26.7% against 48.2%, Seine-Saint-Denis 22.2% against 49.7%, Val-de-Marne 25.1% against 45.3%; same result in a wheelchair.
Hypothesis written on 2 October 2026
8.Limits
- Individual situations. The indicators describe neighbourhoods, not the age, health or circumstances of their residents.
- Private adaptations. Air conditioning, shutters or private solutions do not appear in public data.
- Risks not included. Flooding, soil pollution and industrial risks are not yet taken into account.
- A relative score. Neighbourhoods are compared with each other: adding data can change the results.
- Step-free travel. Step-free times and access are a lower bound: the calculation avoids stairs and stops declared inaccessible, but accounts neither for lifts out of order, nor pavement condition, nor a slower speed, nor whether the destination building is accessible. A stop flagged as accessible may also require assistance booked in advance (as at many RER and train stations).
- Starting point of journeys. In 0.8% of cases, the wheelchair duration shown is 1 to 8 minutes shorter than the duration without constraint. The cause is the point of the street network to which the calculation attaches each start and end: in the full network, it is sometimes a stairway or a path that forces a detour. For the same reason, some durations without constraint are overestimated. This will be fixed in the next calculation, with the same start and end points for every mode.
- Access to care. Each GP counts as one, whatever their working hours: each practitioner's actual activity is not public.
- Public toilets and water points. Only the City of Paris publishes a complete list (toilets and fountains). In the inner suburbs, partial lists exist but cover only a few communes: Grand Paris Seine Ouest (173 fountains and 33 toilets, for 7 communes), Sceaux (21 fountains), Montreuil (22 fountains, 2019 list); Seine-Saint-Denis publishes an OpenStreetMap extract (271 water points) of unknown completeness. These places are therefore shown for Paris neighbourhoods only.
9.Data, code and corrections
The code, the data produced and the detailed method are public. To cite this work or reuse the data, see the About page.
Data sources
The data comes from public bodies and open databases. It is downloaded again at every quarterly update, except access to care, which is recomputed by hand.
| Data | Producer | Edition used |
|---|---|---|
| Heat vulnerability | CSTB (Sat4BDNB project) | Summer 2022 conditions |
| Sealed ground, green spaces | L'Institut Paris Region (land-use survey) | 20212025 edition published, not yet integrated. |
| Facilities (cool spots, services) | INSEE (permanent facilities database) | As of 1 January 2025 |
| Air pollution and noise | Airparif and Bruitparif | 2024 map |
| Homes rated F or G | ADEME (energy performance certificates) | October 2026 extract |
| Electricity consumption | Enedis | 2024 |
| General practitioners, pharmacies | French digital health agency (RPPS); Ministry of Health (FINESS) | 24 September 2026; May 2026 |
| Public transport timetables | Île-de-France Mobilités | 1 October 2026 |
| Streets, footpaths, stairs | OpenStreetMap contributors; IGN (BD TOPO) | 30 September 2026; October 2026 |
| Residents' income | INSEE (Filosofi) | 20212023 edition published at commune level only, with a changed method: not comparable. |
| Population, overcrowded homes, second homes | INSEE (census) | 20222022 edition, reference since 3 October 2026. INSEE changes the definition of overcrowding in it: the median rate goes from 13.5% to 24.6%, with no real change. The differences with 2021 are published as a sensitivity check. |
Contains information from "Horaires prévus sur les lignes de transport en commun d'Île-de-France (GTFS Datahub)", made available by Île-de-France Mobilités under the terms of the "Licence Mobilités".
Corrections made
Indicators that measured something other than what they were meant to measure were removed, including when they widened the observed gaps. Corrections
- Sept. 2026Access to services removed from the score, pending a reliable measureThe data measured distance to services rather than their capacity, and was incomplete in the inner suburbs.
- Oct. 2026Pavement data set asidePavement information exists for 96% of street length in Paris but 26% in Seine-Saint-Denis: no fair comparison was possible.
- Oct. 2026Nearly uninhabited neighbourhoods excluded from the thresholdsNeighbourhoods with fewer than 50 residents (river banks, parks, business areas) skewed the comparison thresholds.
- Oct. 2026Energy ratings of small homes checkedThe labels published by ADEME already include the 2024 correction for homes under 40 m²: no adjustment needed.
- Oct. 2026Access to care: structures without on-site consultations excludedTeleconsultation platforms, SOS Médecins, emergency departments: list published, with the reason for each exclusion.
- Oct. 2026Access to care back in the scoreRebuilt measure: the capacity of GPs and pharmacies shared among the residents who can reach them on foot or by public transport.
- Oct. 20262022 census adoptedPopulation, overcrowded homes and second homes move from the 2021 to the 2022 census; income stays 2021, the latest vintage published at neighbourhood level. The site's key figures are updated. The hypotheses' verdicts, computed on 2021, remain the official results and are checked with 2022. Overcrowding changes definition at INSEE (median rate from 13.5% to 24.6%, with no real change).