Housing versus home sharing: the causal impact of short-term rental regulation on residential rents

TL;DR

Using staggered municipal principal-residence restrictions across Canadian urban areas, this paper finds that short-term rental (STR) activity causally increases residential rents: regulations that reduce STR activity cause rents to decline. The effects operate mainly through commercial STR supply and rental price stickiness, with weaker evidence for a home-sharing demand channel.

Key findings

• Principal-residence restrictions reduce rents: treated neighbourhoods’ log average rents are 0.096 SD lower than the no-regulation counterfactual (ATT = −0.096, SE 0.021, p < 0.001). The same models show declines in FREH listings (−0.251 SD) and non-FREH listings (−0.207 SD), while STR prices are statistically indistinguishable from zero (−0.034, SE 0.035).
• Effects grow over time and spill over: in the year after regulations, rents were on average CA$24 lower (1.7%); by 2023, rents in regulated neighbourhoods were CA$55 (3.3%) lower, and rents in adjacent unregulated areas were CA$40 (3.0%) lower. The authors estimate this saved Canadian renters CA$192.4 million in monthly rent payments in 2023.
• Spillover and residualisation analyses support metropolitan-scale adjustment: untreated neighbourhoods adjacent to treated areas show negative rent effects after about two years (ATT = −0.060, p < 0.1) and declines in STR listings; after removing region-by-year variation, the rent ATT remains negative and significant (ATT = −0.019, SE 0.003, p < 0.001).
• The RE-ESF decomposition finds strong support for the supply channel (FREH_change, p < 0.01) and weak support for the price-stickiness channel (price_change, p < 0.1); the demand channel (non_FREH_change) has the hypothesised direction but a 90% confidence interval crossing zero. Pre-pandemic, STR variation accounted for an average 2.6% of the annual increase in total rent paid; in 2020 this dropped to −5.5%.

Methods

• Time-varying difference-in-differences using Callaway and Sant’Anna’s (2021) group-time ATT estimator, with staggered municipal STR principal-residence restrictions as treatment. Treatment is assigned if a restriction was in place by 1 January of the year; rents are measured in October and STR variables in September.
• Data cover Canadian urban areas: CMHC Rental Market Survey rents and vacancy rates (2015–2023), AirDNA STR listings and nightly prices (2017–2022), Canadian Census controls, and a hand-compiled STR regulation database. The main dataset includes 798 neighbourhoods in British Columbia, Ontario, New Brunswick and Quebec; 38 cities implemented restrictions, covering 309 neighbourhoods.
• Supplementary random-effects eigenvector spatial filtering (RE-ESF) regression decomposes effects into supply (FREH listings), demand (non-FREH listings) and price-stickiness (nightly STR price) channels. Parallel trends are tested via exposure-length ATTs; spatial spillovers and residualised rent models are used as robustness checks.

Limitations

• The paper states that no single clustering level is unambiguously correct because treatment arises from municipal, sub-municipal and partial-jurisdiction regulations; robustness checks use unclustered and municipality-clustered standard errors.
• The RE-ESF pathway analysis is described as exploratory and vulnerable to endogeneity on its own; the authors rely on the DiD for the overall causal claim. The non-FREH demand channel is not statistically significant, and STR prices are treated as exogenously determined by tourism demand, hotel competition and housing quality—a simplifying assumption the authors defend but note is not strictly true.
• Evident from the text: the analysis uses only Airbnb/AirDNA data, excluding other STR platforms, so STR revenue is underestimated; rent data cover purpose-built rentals only, and the CA$8.8 billion total rent figure assumes all tenants’ average rent equals that of purpose-built rental tenants. The authors also note that long-run effects through new residential investment are not captured in the short-run model.