Did legalizing multiplexes work?
In May 2023 Toronto made multiplexes, buildings of up to four homes, legal on effectively every residential lot in the city. It was one of the most-watched zoning reforms in North America. Nine years of permit applications hold the answer, with the nuance the headlines skip.
This page comes in two parts: the story in plain language first, then technical notes with the universe definition, the model, and the design's limits. Built on the permit data of study 07.
Applications rose 27% above a path that was already climbing
The honest starting point is that multiplex activity was growing before the reform, about 1.2% a month, year after year. Any fair evaluation has to measure the reform against that moving baseline rather than against a flat one.
How to read the first chart: each vertical bar is one month's permit applications for small multi-unit housing: duplexes, triplexes, fourplexes, and houses converted into several homes, with laneway and garden suites excluded because they were legalized by earlier, separate reforms. The dashed hairline is the model's fit, drawn over the data. Against the rising pre-reform path, the reform's own contribution is a one-time jump of about 27% (95% range +7 to +51), and the climb then continues on the same slope, reaching roughly 150 applications a month by 2025, more than double the old norm.
Multiplex applications jumped 27% above a path that was already rising
monthly permit applications creating units in small multi-unit housing · dashed hairline = the model's fit, drawn over the data (R² = 0.89) – Jan 2017 to May 2026
City of Toronto building permits open data · first applications (revision 00) creating dwelling units · laneway suites excluded
The estimates, with the honesty built in: the second chart shows the three numbers behind the story. The jump is statistically clear. The trend change after the reform is nil, meaning the reform lifted the level and let the old momentum carry on. And the single-family series, checked at the same date, shows no jump at all, which makes a generic city-wide permitting shock in exactly June 2023 an unlikely explanation.
One clean jump, no extra acceleration, and a contrast series that stayed put
Estimated % change vs the pre-reform path – whiskers mark 95% confidence ranges, Newey-West errors – the bottom row is a contrast, never a control: substitution into multiplexes is part of the reform working
City of Toronto building permits · single-series ITS, month-of-year seasonality, 77 pre-reform months
Not just more projects: different projects
If the reform were merely relabeling activity that would have happened anyway, the mix of projects should look the same. It does not.
How to read this chart: before the reform, one small-multi application in eight was for a building of three or more units; after, it is more than one in four. Ground-up new builds tripled their share, from 5% to 17%, where conversions of existing houses had dominated. Projects also got bigger on average, 1.12 units per permit before and 1.59 after, exactly the change a fourplex legalization should produce.
The reform's fingerprint: bigger projects, built from the ground up
before the reform (2017 to May 2023) · after (June 2023 on) – shares of small-multi applications – units per permit rose 1.12 to 1.59
City of Toronto building permits · pre = 77 months, post = 36 months
And in homes rather than applications: the final chart counts the new dwelling units in those applications by year. From a pre-reform norm of roughly 700 to 800 homes a year, the pipeline reached about 2,500 by 2024 and 2025. Against Toronto's tower-dominated supply from study 07, these are small numbers, and against the missing-middle debate, they are the fastest-growing part of the city's housing pipeline.
From about 750 homes a year to about 2,500
New dwelling units in small-multi permit applications, by application year – one rung = 250 homes – before the reform era · 2023 onward
City of Toronto building permits · applications measure intent, not completions · 2026 partial year omitted
The honest fine print
Applications measure builder intent, not finished homes; study 07's permit clock and construction lags sit between these filings and anyone's front door. The design is a single-series before-and-after: nothing else changing for small-multi builders in exactly June 2023 is an assumption, softened but never proven by the flat single-family contrast. That contrast is a contrast and never a control, because a builder switching a house rebuild into a fourplex is the reform working, which would depress single-family numbers for reform-related reasons. Ontario's Bill 23 (late 2022) cut charges for exactly these buildings inside the window, so "the May 2023 reform" is best read as the citywide legalization moment within a broader pro-multiplex policy turn. And a sixplex pilot arrived in nine wards in June 2025, inside the post-period.
Pre-flight catch, rule 4 again: before March 2016 the permit system's structure-type coding makes multiplexes look nearly nonexistent, a classification regime rather than reality. The analysis window starts in 2017 for that reason, and the technical notes show the artifact.
Technical terms, translated
- As-of-right
- Allowed without a rezoning application. Before the reform, a fourplex on most lots meant months at the committee or tribunal from study 05; after, it means a building permit.
- Level shift vs trend change
- A level shift is a one-time step up. A trend change is a new slope. This reform produced the first and left the second alone.
- Contrast vs control
- A control group must be untouched by the policy. Single-family permits are plausibly touched (builders switching over), so they can sharpen the story but never anchor the estimate.
- Application vs permit vs home
- An application is a filing; a permit is its approval; a home exists years later. This study counts filings, the fastest-moving signal of builder behaviour.
Technical notes
The universe, the specification, the estimates, and the limits.
Universe and outcome
From the deduplicated permit file of study 07: first applications (revision
00) with DWELLING_UNITS_CREATED > 0 in small-multi structure
types (2 Unit and 3+ Unit detached, semi-detached, and townhouse; Duplex;
Triplex; Converted House and combination codes). Laneway / Rear Yard Suite is
its own structure type and is excluded. Outcome: monthly application counts,
January 2017 to May 2026 (113 months; the last two available months dropped
for processing lag). Window starts 2017 because structure-type coding shows a
regime change in early 2016 (near-zero small-multi counts before March 2016,
20 to 50 per month immediately after), a classification artifact.
Model
Single-series segmented regression on log monthly counts (Bernal, Cummins and Gasparrini, 2017): intercept, linear pre-trend, level shift and slope change at June 2023, month-of-year indicators; Newey-West standard errors with 6 lags (Newey and West, 1987). Coefficients reported as percentage changes, (e^b − 1) × 100. R² = 0.89. The identical specification is fit to the single-family series as a contrast.
where At = small-multi applications in month t, Postt = 1 from June 2023, τt = months since June 2023 (0 before), and Mmt are month-of-year indicators; Newey–West errors, 6 lags. β2 is the reform's level shift, β3 the slope change.
Estimates
| Term | Estimate | 95% CI | p |
|---|---|---|---|
| Pre-reform trend, per month | +1.2% | +0.9 to +1.4 | < .001 |
| Level shift at June 2023 | +27.1% | +6.9 to +51.2 | .007 |
| Slope change after, per month | −0.1% | −0.7 to +0.6 | .83 |
| Single-family level, same date (contrast) | +8.8% | −13.9 to +37.4 | .48 |
Composition, pre (77 months) vs post (36 months): applications per month
57.6 vs 135.0; new units per month 64.2 vs 215.3; 3+ unit share 12.5% vs
27.8%; ground-up new-build share 5.1% vs 16.5%; mean units per permit 1.12
vs 1.59. Annual units in applications reach roughly 2,500 in 2024 and 2025.
Full estimates in output/results.json.
Sensitivity and diagnostics
The house standard, applied to a result that flatters expectations. Residuals are well behaved (skewness −0.4, excess kurtosis near zero, no residual-on-fitted curvature, p = .28); spread varies with the fitted level (p = .01), which the Newey-West errors already accommodate. The break-date placement barely matters: moving the break anywhere from April to August 2023 keeps the level shift between +20% and +30%, always significant. The break model beats a no-break model on both AIC (−51.7 vs −40.8) and BIC. A Poisson count model with the same regressors gives +33.4% (95% CI +18.7 to +50.0), the same story in a different likelihood. One check is imperfect, and it is reported rather than buried: of two placebo breaks fit inside the pre-period only, June 2021 is a clean null (−6.7%, p = .34) but June 2019 yields +17.9% (p = .068), which says this series has produced level-shift-like wobbles before without any legalization. The reform estimate stands on firmer ground than that placebo: it is larger, more precise, stable across break dates, corroborated by the compositional fingerprint and the flat single-family contrast, but the marginal placebo is a fair reason to hold the causal claim with a slightly looser grip.
Threats to validity
Attribution is to the citywide legalization moment within a broader policy turn: Bill 23's charge reductions (late 2022) and the June 2025 sixplex pilot sit inside the window, and the design cannot isolate the by-law text from the surrounding momentum. The flat single-family contrast argues against a generic June-2023 permitting shock; it cannot rule out multiplex-specific news arriving the same month, which is, of course, mostly the reform itself. Counts are modeled in logs on a series that never approaches zero in-window; the estimates are about filings, and completions will lag by the construction durations measured in study 07.
Reproducibility
Two scripts (run_analysis.py, make_charts.py)
over study 07's committed permit files (run its
get_data.py first). Every number on this page comes from the
committed results.json and monthly CSVs.
References
- City of Toronto. (2026). Building permits: active and cleared permits [data sets]. City of Toronto Open Data Portal. https://open.toronto.ca/dataset/building-permits-cleared-permits/ (retrieved August 2026). Open Government Licence, Toronto.
- City of Toronto. (2023). Expanding Housing Options in Neighbourhoods: multiplex study, Official Plan and citywide Zoning By-law Amendments (adopted by City Council, May 2023). Details in the city's planning records; cited here as the intervention date.
- Bernal, J. L., Cummins, S., & Gasparrini, A. (2017). Interrupted time series regression for the evaluation of public health interventions: a tutorial. International Journal of Epidemiology, 46(1), 348–355.
- Newey, W. K., & West, K. D. (1987). A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica, 55(3), 703–708.