What does Airbnb's 28-day loophole actually rent for?
Toronto's short-term rental rules stop at 28 nights: require guests to stay at least that long, and the rules no longer apply. Most entire homes on Airbnb now sit past that line. This study asks what they charge.
This page comes in two parts: the story in plain language first, then technical notes with the sample construction, model specification, and full estimates.
One platform, two markets
When a city regulates stays under 28 nights, hosts have a choice: follow the rules, or set a 28-night minimum and step outside them. In the June 2026 snapshot of Toronto's listings, most entire homes had stepped outside.
The data comes from Inside Airbnb, an independent project that regularly collects what is publicly visible on Airbnb's own site. I kept the 9,109 entire homes with a usable price and asked a simple question: do the 28-night listings price like hotels, or like housing?
Raw averages would mislead here, because the two kinds of listings sit in different places and come in different sizes. So the comparison is like-for-like: listings in the same part of the city, with the same number of bedrooms, from the same kind of host. After that matching, the price gap that remains is about the stay length itself.
How to read the first chart: each row is a ring of distance from downtown, from the core to the far edge of the city. The light bar is the typical nightly price of short-stay listings; the dark bar is the typical nightly price of 28-night-minimum listings. The gap barely moves as you travel outward: roughly half price, everywhere.
Loophole listings price 53% below short stays, in every ring of the city
Typical (median) nightly price by distance from downtown – 28-plus-night minimum (outside the by-law) · short stay – like-for-like model estimate: 53% lower
Inside Airbnb, Toronto snapshot June 15 2026 · 9,109 priced entire-home listings
Half of hotel price is still a lot of rent
Pricing at half the nightly rate of a hotel stay sounds cheap, until you multiply by thirty.
How to read the second chart: each dot is $500 of monthly cost. Take the typical 28-night listing's price and stretch it over a month, and a one-bedroom works out to about $3,671. These listings occupy a strange middle ground: month-long minimum stays, furnished, no lease, at asking prices far above what most tenants paying typical long-term rents could manage.
The loophole one-bedroom implies $3,671 a month; a rental condo averages $2,365
Median implied monthly price of entire homes listed with 28-plus-night minimums – one dot = $500 – | = average rent of a rental condo apartment in Toronto (CMHC, October 2025) – 4,047 loophole listings
Inside Airbnb, Toronto snapshot June 15 2026 · listed prices, not completed transactions
How do those figures compare with ordinary rents? Canada's national housing agency, CMHC, surveys landlords every October. Its October 2025 numbers for Toronto put the average rent of a rental condo, the most expensive corner of the ordinary market and the closest cousin to these listings, at $2,365 for a one-bedroom, $2,942 for a two-bedroom, and $3,342 for three or more. The thin markers on the chart place those averages on the same dot scale: every loophole row runs far past its marker, at roughly 1.6 to 2.4 times the condo average. Against regular purpose-built apartments the gap is wider still; the average asking rent on a vacant one-bedroom apartment is $2,020.
This study extends my MA thesis, which found that listings outside the by-law's reach were linked to 2.5 times the rent impact of regulated ones. Two years later, the loophole has grown into the platform's main channel, with most entire homes now using it.
The honest fine print
These are listed prices: what hosts ask, on one June day, in public view. Nobody's signed agreement is in this data, and some listings never find a taker at their asking price. The patterns are strong and consistent across the whole city, but the exact dollar figures should be read as asking prices, never as a rent survey.
Technical terms, translated
- Typical price (median)
- The middle listing when all prices are lined up from lowest to highest. A few extreme luxury listings can drag an average around; the median ignores them.
- Like-for-like comparison
- A model that compares listings matched on location, size, and host type, so the one remaining difference, the minimum stay, explains the gap. It accounts for about two-thirds of all price differences between listings.
- CMHC benchmark rents
- Canada's housing agency surveys landlords every October. "Average rent" mostly reflects sitting tenants; the "asking rent" on a vacant unit is what a new renter would face; condo rentals are surveyed separately and run higher than regular apartment buildings.
- Implied monthly price
- The listing's price stretched over a month of nights, to make a 28-night listing comparable with a rent bill.
Technical notes
Sample construction, specification, and full estimates behind the plain-language story above.
Data and sample
Inside Airbnb's Toronto snapshot of June 15, 2026 (CC BY 4.0): 22,198 listings,
restricted to entire homes. Dropping 2,724 listings without a usable price and
those missing bedrooms, baths, review score, or guest capacity leaves n =
9,109. The loophole indicator is minimum_nights ≥ 28: 58.7% of
entire homes (60.5% of all listings). My thesis measured 66.2% on the same
definition in September 2024, so the share has drifted with the market while
staying well above half.
Design and estimation
A hedonic price model in the tradition of Rosen (1974): OLS on log nightly price with neighbourhood fixed effects, so the loophole contrast is estimated within neighbourhoods rather than across them. Controls: guest capacity, bedrooms, baths, review score, log review count, superhost status, and haversine distance to the downtown core; cluster-robust standard errors by neighbourhood. Effects on price are reported as percentage changes, (e^b − 1) × 100, the standard conversion for indicator variables in semilogarithmic models (Halvorsen and Palmquist, 1980). Implied monthly cost is nightly price × 30, taken as the median within bedroom count among loophole listings (n = 4,047).
where θ is the loophole price signature, reported as (eθ − 1) × 100, and μn(i) are neighbourhood fixed effects; errors clustered by neighbourhood.
| Term | Effect | 95% CI | p |
|---|---|---|---|
| 28-plus-night minimum (the loophole) | −53.0% | −55.0 to −51.0 | < 0.001 |
| Distance to downtown, per km | −3.5% | −5.8 to −1.1 | 0.004 |
| Bedrooms, per bedroom | +15.1% | +11.7 to +18.5 | < 0.001 |
| Superhost | −3.5% | −5.7 to −1.3 | 0.002 |
n = 9,109 entire-home listings. Controls also include guest capacity, baths, review score, and log review count; cluster-robust SEs by neighbourhood; R² = 0.668. Effects converted as (e^b − 1) × 100.
Residual diagnostics
The model is checked graphically against its own residuals, with numeric summaries as adjuncts to the plot. At this sample size the honest framing is the distinction between detectable and problematic: a residual-on-fitted curvature check is statistically unambiguous (p < .001 at n = 9,109), but its magnitude is small. The binned mean residual stays within about 0.1 log points of zero against a typical residual spread of 0.40, with the largest deviation a slight overprediction of the most expensive listings. Against the headline loophole estimate of −53%, a wobble of that size changes nothing that matters, so it is named and kept rather than modeled away; chasing it with extra terms would spend degrees of freedom on the top percentile of the market. Residuals are mildly right-skewed (skewness 0.5) with moderately heavy tails, and spread rises gently with predicted price (mean absolute residual 0.26 to 0.32 across fitted bins); inference already uses cluster-robust standard errors, which do not assume constant spread.
The model against its leftovers: a wobble you can see, small enough to name and keep
Every one of 9,109 listings counted into cells – darker = more listings – dark hairline = mean residual in 22 equal-count bins – residuals in log points, where 0.4 ≈ the typical miss and ±0.1 is the largest binned deviation
Inside Airbnb hedonic model (M2) · n = 9,109 · committed diagnostics output · brightness encodes count
Outlier sensitivity
The published model trims prices at the 1st and 99th percentiles, which is itself an outlier decision, so the sensitivity runs in both directions: the same specification refit with estimators that downweight extremes, and refit with the trim reversed. Comparison metric: the headline loophole effect.
| Estimator | n | Loophole effect |
|---|---|---|
| OLS, 1–99 percentile trim (published) | 9,109 | −53.0% |
| Huber robust, trimmed | 9,109 | −54.3% |
| Median regression, trimmed | 9,109 | −54.5% |
| OLS, no price trim | 9,223 | −54.1% |
The estimate moves by less than two percentage points across all four fits, and the published −53% is the most conservative of them. The trim is not load-bearing, the tails are not driving the result, and the headline stays as written.
Threats to validity
Prices are asking prices from a single snapshot; transactions, occupancy, and discounts for long stays are unobserved, and hosts can change minimum-night settings seasonally. Neighbourhood fixed effects absorb location composition, but within-neighbourhood sorting of loophole listings into cheaper buildings would still shade the estimate. The distance-ring chart in part one is the robustness picture: the gap is stable across the whole city, which is hard to square with purely compositional explanations.
Reproducibility
Three scripts (get_data.py, run_analysis.py,
make_charts.py) on pandas and statsmodels. Every number on this
page comes from the committed results.json and the binned-medians
CSV. Inside Airbnb data used under CC BY 4.0 with attribution.
Rent benchmarks
Benchmark rents come from the CMHC Rental Market Survey, October 2025, City of
Toronto: purpose-built average rents (Table 1.1.2), vacant-unit asking rents
(Table 1.1.9), and rental-condominium averages (Table 4.1.2; the marker series
on the dot chart, chosen because loophole listings are predominantly
condo-type units, which makes it the most conservative comparator). Figures
were extracted programmatically from CMHC's published data tables into
output/cmhc_rents.json; as an internal consistency check, the
condo table's companion RMS column reproduces Table 1.1.2 exactly. Note that
CMHC surveys the primary rental stock in October, while the Airbnb snapshot is
June 2026; the eight-month gap is small relative to the size of the price
differences shown.
References
- Inside Airbnb. (2026). Toronto, Ontario, Canada [data snapshot, June 15, 2026]. https://insideairbnb.com/get-the-data/ (retrieved June 2026). Licensed under CC BY 4.0.
- Canada Mortgage and Housing Corporation. (2025). Rental Market Survey data tables: Toronto, October 2025 [XLSX]. https://assets.cmhc-schl.gc.ca/sites/cmhc/professional/housing-markets-data-and-research/housing-data-tables/rental-market/rental-market-report-data-tables/2025/rmr-toronto-2025-en.xlsx (retrieved July 2026).
- City of Toronto. (2024). By-law 503-2024, short-term rental regulations. City of Toronto.
- Rosen, S. (1974). Hedonic prices and implicit markets: product differentiation in pure competition. Journal of Political Economy, 82(1), 34–55.
- Halvorsen, R., & Palmquist, R. (1980). The interpretation of dummy variables in semilogarithmic equations. American Economic Review, 70(3), 474–475.
- Zhang, J. (2026). Master's thesis on short-term rentals and long-term rents in Toronto, MA Sociology, University of Guelph. Summarized on the research page of this site; full citation on request.