Source
Sarah Shabrina, Elsa Arcaute and Michael Batty (2021), “Airbnb and its potential impact on the London housing market”, Urban Studies, LIT-094, read in full in NOTE-071. Figures came through as captions only, so map values are known only where the text states them.
What was actually shown
Supply. Airbnb listings were 0.75% of London’s stock (about 25,000) in 2015 and 2.38% (about 80,000) in 2019 (Fig. 2). In 2018, 82% of LSOAs (3,982 of 4,835) had at least one active listing.
The proxy. “Possible misuse” is an entire-home listing available more than 180 days a year whose host has two or more listings. The threshold assumes 50% occupancy, taken from Visit Britain hotel bed-space occupancy of 53–55%, so that 180 days offered means more than 90 nights booked, the legal cap under the Deregulation Act 2015, s. 44 (note 1).
Extent, 2018 (Fig. 3). Entire homes available more than 180 days: 0.25% of stock (about 8,200). Possible misuse: 0.16% (about 5,300). In some LSOAs in Camden, Westminster, the City and Hackney possible misuse reaches 3.5–7%; by borough it is at most 1.5% (City of London).
Where it concentrates (LSOA correlations, Figs. 4–5). Purpose-built flats r = +0.28, privately rented +0.33, dwelling-type entropy −0.14, owned −0.31, detached houses −0.25. No significance tests or spatial adjustment are reported, and r² is at most 0.11.
What could have come out otherwise. Possible misuse could have been spread evenly, or tracked owner occupation. It is small citywide and concentrated in flat-dominated, privately rented central areas.
What this does not say
- That 0.16% of homes breached the cap. The proxy counts availability, not bookings. It misses single-listing hosts who exceed 90 nights and includes multi-listing hosts who do not. The occupancy assumption is not tested on Airbnb data.
- That more than 2% of London’s homes are misused. That is the abstract’s wording; the body’s 2% is the share listed at all.
- That Airbnb raises rents, or by how much. The borough regression (n = 33, four predictors, R² 0.62) finds logged 2015 misuse associated with a min–max-normalised 2015–17 asking-rent change (β 0.084, p 0.001). It is a cross-section with no control for prior rent, tourism or trend and no spatial model. Its translation, “a 100% increase … up to 8%” rent, about £90 a year, mixes a log unit with a doubling and a normalised unit with a percentage (NOTE-071’s analysis). The panel and instrumental studies it cites are the better sources for an effect size; the record has not read them.
- That the correlations are causal or robust. They are pairwise, small and untested.
Connections
- The definition follows Schäfer and Braun (2016) on Berlin, whose possible misuse was 0.30% of stock (5,555 units). The record does not hold it.