The tax on vacant dwellings, extended in 2023

Evidence

The taxe annuelle sur les logements vacants is levied by the State on dwellings left empty for a year or more in the communes a decree lists. It is a tax, not spending; the only figure this site has for its yield, from deputies, is €93m for 2022. The finance law for 2023 raised its rates by more than a third from 2023, in every taxed commune, and extended it from 1 January 2024 beyond the urban areas of more than 50,000 inhabitants, to communes where rents, prices or the share of second homes show housing out of balance: 2,243 metropolitan communes, the zone touristique et tendue. The decree of 25 August 2023 that listed them also redrew the list of tight urban areas of more than 50,000 inhabitants, the zone tendue, whose criterion the law left as it was, and added 292 metropolitan communes to it. The question is whether the newly taxed communes saw fewer dwellings left empty.

• Against the communes taxed since 2013. They stayed taxed, so they shared the 2024 change in how vacancy is counted; but their rates rose in 2023, so against them the estimate is the effect of entering the tax less that of the rate rise. Communes are compared within their region.
• Against communes never taxed. Compared within cells of region, population and vacancy rate in 2019. Never taxed is not untaxed: communes outside the zone may levy a local tax on dwellings vacant for two years, and more of them did in 2024 than in 2022.
• The check. The communes taxed since 2013 against the same never-taxed communes. Nothing started for them in 2024, so any gap after 2023 measures what the two taxed groups share and the others do not.
• Estimators and placebo. The event study of Callaway and Sant'Anna and synthetic difference-in-differences, each with a fake start in 2021 that should find nothing. The outcome is LOVAC's count of private dwellings vacant on 1 January, per 100 private dwellings; 2023 is the base year, 2024 and 2025 the years after.

Against the communes taxed since 2013, the vacancy rate of the 292 communes added to the zone tendue moved by +0.13, estimated per 100 dwellings over 2024 and 2025 by the event study, and by +0.13, estimated by synthetic difference-in-differences. The two groups moved together from 2019 to 2022, and the fake start in 2021 finds +0.10, estimated. No fall here means that entering the tax moved vacancy no more than the 2023 rate rise moved it in the communes already taxed, not that the tax does nothing. Their rate was 9.64, estimated per 100 in 2023, so the fall of 13%, estimated that Segú (2020) measured after the tax was created in 1999 would be −1.25, estimated here, and both intervals exclude it; but that fall sets taxed communes against communes never taxed, while this estimate is entry less the rate rise, so it gives a scale, not a like-for-like test.

Against communes never taxed, the same communes moved by −0.70, estimated by the event study and −1.30, estimated by synthetic difference-in-differences; but the communes taxed since 2013, whose status did not change, moved by −0.76, estimated and −1.46, estimated against the same communes. The fall is shared by the two taxed groups and not by the others: it comes from the change in the files LOVAC counts vacancy from, in the count of 1 January 2024, or from the 2023 rate rise, which these data cannot separate. It is not the effect of entering the tax.

LOVAC counts vacancy from the files that assess the taxes on dwellings. When the taxe d'habitation on main residences ended in 2023, those files came to rest on the occupancy declarations owners have had to make since that year through the tax administration's service « Gérer mes biens immobiliers »; the Cour des comptes reports that more than a million taxpayers were then wrongly charged the taxe d'habitation or the vacancy tax, and LOVAC's producer notes the change in its inputs. Every group's recorded vacancy fell in the count of 1 January 2024 and partly recovered in 2025. Counted in dwellings rather than as a rate, the vacant dwellings of the newly taxed communes of zone tendue even rose by +7.5%, estimated against the communes taxed since 2013, an interval that excludes any fall. A tax on vacancy that raised it is not a credible reading: the count of 2024 fell in proportion more in the communes taxed since 2013, and the rise is read here as a mark of how unevenly the new files moved, not as an effect.

For the 2,243 communes of zone touristique et tendue, those the law itself added, the data allow no verdict. Their recorded vacancy fell in the count of 1 January 2023, before the decree and before its list was known: against the communes taxed since 2013 their gap was +1.12, estimated per 100 dwellings larger in 2022 than in 2023, and the test of equal trends before 2023 rejects them against both comparisons. Their estimate, +0.05, estimated against the communes taxed since 2013, measures that fall as much as the tax.

Vacancy of more than two years tells nothing more. Against never-taxed communes it moved by −0.16, estimated per 100 dwellings, and the communes taxed since 2013 by −0.13, estimated, a comparison whose test of equal trends before 2023 rejects them; against the communes taxed since 2013 it moved by +0.11, estimated, and the test rejects equal trends there too.

The design measures what a commune's listing did, which makes the tax due on every dwelling vacant for a year or more, not what owners paid: how many were taxed in each commune is not published. LOVAC counts the private dwellings the tax files record as vacant, so social housing is outside it. Because the count comes from the tax files, a dwelling billed by mistake whose owner then declares it occupied leaves the count without being occupied: a recorded fall would overstate a real one, and the absence of a fall is the stronger statement. Entering the zone also let a commune raise the taxe d'habitation on second homes, which 1,461 communes did in 2024, so the extension is measured with it. The vacancy rate needs the private dwelling stock, which LOVAC keeps secret in small communes; each figure's panel gives the communes it rests on. Two years after the start, an effect that builds slowly would not show yet. The tax belongs to no spending programme, and the State set no performance indicator on it. The communes on each list are in tlv.csv, the dwelling stock in stock.csv, every estimate by year, comparison and outcome in results.csv and the sources in SOURCES.md. The code is analysis/vacancy.ml, with the estimators of lib/did.ml.