Romania, 2000–2020 · 2,712 localities compared
Emptying of people, not of homes
Of 1,979 localities that had fewer residents in 2020 than in 2000, 1,556 have more dwellings. Satellite images point the same way: in the localities that emptied, the ground covered by buildings grew by almost a third (median). Type the name of your own place and see what happened there.
- 73%of the 2,712 comparable localities had fewer residents in 2020 than in 2000.
- 79%of those (1,556) still have more dwellings, according to the national statistics institute (INS).
- 59%of the new ground under buildings appeared in them. They held 69% of the 2000 population.
existed in 2000added 2000–2015added 2015–2020, estimated
Satellite image, cells of about 90 metres.
Where it happened
Each polygon is a locality: a commune (rural municipality), a town or a city. Colour combines two INS numbers for 2000–2020: residents and dwellings. Click a locality to see it above.
…
The pattern is clearest in Ialomița (35 of 40 localities), Gorj (57 of 70 localities), Călărași (41 of 51 localities). It is much rarer in Ilfov (2 of 38), Brașov (9 of 45), Sibiu (15 of 60).
Dwellings follow population. Built-up ground follows it far less
Each dot is a locality. Horizontal: how its number of residents changed. Vertical: dwellings (INS) or ground covered by buildings (satellite). The thick line is the median of each band.
Dwellings (INS) track population fairly closely.
Dwellings track population closely (correlation coefficient 0.87). Where population fell by over 30%, the median dwelling count falls too (−3%). But in the usual range of decline, 0 to 20%, 82%–91% of localities still have more dwellings than in 2000.
Built-up ground is tied to population much more loosely (0.52). In the 98 localities that lost over 30% of their residents, median growth is +27%. In the 768 that lost 0 to 10%, it is +34%. In the 111 that gained over 30%, it is +73%.
By band of population change
Bands group localities by population change between 2000 and 2020. First bar: share of localities with more dwellings in 2020 than in 2000. Second: median growth of ground covered by buildings.
Where the new ground appeared
In the 2,712 comparable localities, satellite estimates show 544 km² more ground covered by buildings in 2020 than in 2000. Of that increase, 319 km² (59%) lies in localities that lost residents. They held 69% of the 2000 population, so their share of the new ground is smaller than their share of the people, but still a majority.
In localities that emptied, the building footprint per resident rose from 71 to 104 m². The arithmetic is simple (fewer people, buildings that do not vanish), so we do not use it as evidence; we compare how much the ground grew, with the limits below.
Larger towns that lost residents and still grew: Zalău (Sălaj County): residents −3%, dwellings +20%, built-up ground +42%; Pitești (Argeș County): residents −8%, dwellings +8%, built-up ground +33%; Oradea (Bihor County): residents −4%, dwellings +16%, built-up ground +33%; Slobozia (Ialomița County): residents −8%, dwellings +11%, built-up ground +31%.
Who decides where building is allowed
Where building is allowed is set by each locality's general urban plan (PUG, plan urbanistic general), approved by the local council.
- In force
- The Code of territorial planning, urbanism and construction (Law no. 169/2026, Official Gazette no. 661 of 10 August 2026) has applied since 25 August 2026.
- Old plans
- According to legal analyses published after its adoption, a PUG older than seven years must be updated within three years of the Code taking effect, that is by August 2029. Existing plans, including extended ones, stay valid until the update is approved, but no later than that deadline.
- If not updated
- According to the same analysis, an authority that does not update loses access to public funds.
- What you can do
- Ask the town hall when your locality's PUG was adopted and when it will be updated. The council decision approving it is public.
What changes for anyone who builds, article by article, is in the guide to the Urbanism Code.
We found no verifiable national record of PUGs that are up to date, so we give no count of expired plans. Buildings in these data may or may not have been permitted; the data cannot tell the two apart.
How reliable these numbers are
- A satellite does not count houses. JRC GHSL estimates the ground covered by buildings in cells of about 90 metres, from Landsat and Sentinel-2 images. It includes sheds, warehouses and farms; about 7% of the 2020 area is classed by the model as non-residential.
- The 2000 image misses small buildings. Its resolution was 30 metres and low village houses can be missing. Part of the growth in villages therefore comes from better images. That is why we compare groups with each other and do not read values as exact new hectares. INS points the same way.
- Demolitions are invisible. The method reconstructs the past from the newest image and can only remove buildings as it goes back. Built-up ground cannot fall, so we do not claim that “x% of localities grew”; we compare how much they grew.
- 2020 and 2015 are estimates anchored on images from 2018 and 2014. The result is the same over 2000–2015: the median is +23% for localities that lost residents and +32% for those that gained.
- Domicile is not residence. INS publishes the population registered by domicile on 1 January. People who left for abroad can stay on the register, so the numbers do not show how many people actually lived in each locality. Someone who lives in a commune around a city but is registered in the city is counted in the city.
- The dwelling stock is a ledger, not a count. It starts from the census, adds completed dwellings and subtracts reported demolitions; it is rebased at every census, nationally by +3.5% in 2011 and +8.5% in 2021. Rebasing raises the stock in all localities, so we also check the window 2012–2020, with no rebasing: 82% of the 2,148 localities that lost residents have more dwellings. The ledger can record falls (in 2013–2020, 952 localities had at least one year with fewer dwellings), but unreported demolitions only show at the next census.
- 469 localities are left out of the comparison and show grey on the map: 458 communes created or split after 1992 and the units they split from, 10 with sudden jumps in the population series, and one with almost no buildings in the 2000 image.
- Correlation, not explanation. The data do not say who built or for whom, nor whether a dwelling is lived in. We do not estimate upkeep costs: the earlier edition had a per-household figure in lei that could not be supported and was withdrawn.
All localities
Search, sort by any column and click a name to see it at the top. Localities with “–” have a population series that cannot be compared.
| București | 2,154,404 | 2,156,160 | 0% | +13% | +13% |
| Iași | 331,334 | 388,513 | +17% | +23% | +20% |
| Cluj | 319,333 | 327,406 | +3% | +37% | +25% |
| Timiș | 335,549 | 325,583 | −3% | +18% | +16% |
| Constanța | 339,398 | 311,512 | −8% | +18% | +15% |
| Galați | 319,180 | 308,317 | −3% | +9% | +7% |
| Dolj | 314,536 | 300,223 | −5% | +14% | +22% |
| Brașov | 310,536 | 289,778 | −7% | +22% | +24% |
| Prahova | 252,364 | 225,171 | −11% | +4% | +18% |
| Bihor | 230,880 | 221,723 | −4% | +16% | +33% |
| Brăila | 235,372 | 200,342 | −15% | +5% | +6% |
| Bacău | 206,455 | 197,881 | −4% | +13% | +19% |
| Arad | 190,292 | 176,164 | −7% | +12% | +23% |
| Argeș | 188,479 | 173,076 | −8% | +8% | +33% |
| Sibiu | 171,023 | 168,275 | −2% | +26% | +21% |
| Mureș | 163,114 | 147,012 | −10% | +7% | +16% |
| Maramureș | 156,870 | 145,055 | −8% | +12% | +30% |
| Vaslui | 75,885 | 131,830 | – | – | +29% |
| Buzău | 149,885 | 131,231 | −12% | +8% | +13% |
| Suceava | 118,975 | 126,549 | +6% | +16% | +31% |
| Botoșani | 128,408 | 119,832 | −7% | +12% | +21% |
| Satu Mare | 133,820 | 118,932 | −11% | +9% | +24% |
| Vâlcea | 119,099 | 117,569 | −1% | +24% | +27% |
| Neamț | 126,591 | 112,340 | −11% | +9% | +18% |
| Mehedinți | 117,839 | 105,734 | −10% | +8% | +13% |
Method and sources
Localities. The 3,181 administrative units in the INS nomenclature; Bucharest is a single unit, as in TEMPO. Boundaries come from the ANCPI LAU set (geo-spatial.org, 2025).
Residents. INS TEMPO POP107D, population by domicile on 1 January, 2000 and 2020. Dwellings. INS TEMPO LOC101B, dwelling stock on 31 December, 2000 and 2020. Data accessed on 2 October 2026.
Built-up ground. JRC GHS-BUILT-S R2023A, three-arc-second cells, epochs 2000, 2015 and 2020; the total m² per locality, each cell assigned to the polygon containing its centre. The non-residential part: GHS-BUILT-S NRES. Locality areas are computed in the national Stereo 70 projection.
Comparable. 2,712 localities with a population series without breaks over 2000–2020. We exclude units created after 1992 and the units they split from (from the territorial breaks detected in the INS series), any one-year fall above 10% or rise above 40%, any locality whose population grew by over 40% while dwellings grew three times more slowly, and the locality with under 2.5 ha of buildings in 2000.
Calculations. Percentage change is (2020 value / 2000 value − 1) × 100. The median is taken across localities, unweighted. “Emptied” means a 2020 population below the 2000 one. The file of all localities is available as CSV.
Related pages. For one area in detail, see Bucharest and Ilfov seen from orbit. Romania builds. But for whom? compares dwellings with population over 1992–2025 across every locality with data, which is why its percentages differ from those here.
Sources
- INS, TEMPO-Online: POP107D (population by domicile on 1 January, by locality), LOC101B (dwelling stock). statistici.insse.ro, accessed 2 October 2026.
- Pesaresi, M., Politis, P. (2023): GHS-BUILT-S R2023A, built-up surface, 1975–2030. European Commission, JRC. Licence CC BY 4.0. doi:10.2905/9F06F36F-4B11-47EC-ABB0-4F8B7B1D72EA
- ANCPI, administrative-unit boundaries (LAU), via geo-spatial.org, version of 26 March 2025, CC BY-SA 4.0.
- Law no. 169/2026 approving the Code of territorial planning, urbanism and construction, Official Gazette of Romania, Part I, no. 661 of 10 August 2026. text on Lege5 (in Romanian).
- BPV Grigorescu, The new Code of territorial planning, urbanism and construction (on PUGs older than seven years; in Romanian).
Page updated on 2 October 2026.