← All projects

Exploratory analysis · CASE STUDY

TTC ridership: a recovery that hasn't caught up

Two decades of Toronto transit data: the 2016 peak, the pandemic collapse, and why measuring recovery starts with checking what a month means.

PythonpandasNumPymatplotlib
View code on GitHub

TORONTO / A RECOVERY IN CONTEXT

More rides. Still a long way back.

538.1M2016 · calendar-year peak
413.1M2025 · total rides
78.6%2025 as a share of 2019
Annual TTC rides, 2008–2025
TTC annual ridership and incomplete recoveryRidership peaks in 2016 at 538.1 million rides, falls to 197.8 million in 2021, and reaches 413.1 million in 2025. The dashed reference shows 2019 at 525.5 million.0M200M400M600M2019 reference · 525.5M20082012201620212025Complete calendar years · total rides, not unique riders

Recalculated from the project's committed CSV. Annual totals use no inferred reporting-day adjustment; partial 2026 is excluded.

Read the chart as a table
Annual totals from monthly observations
YearRides
2008466,700,000
2009471,233,000
2010477,357,000
2011500,219,000
2012514,007,000
2013525,194,000
2014534,815,000
2015534,005,000
2016538,079,000
2017533,216,000
2018521,404,000
2019525,470,000
2020225,000,000
2021197,841,000
2022318,733,000
2023396,331,000
2024419,860,000
2025413,114,000

The question

How much of Toronto's transit ridership has returned—and are the comparisons themselves fair?

This project examines 508 observations across four TTC indicators in a City of Toronto open-data extract covering 2003–2026. The monthly series start later, in 2007 and 2008. It follows long-term demand, the pandemic shock, seasonal patterns, and the limits of what system-wide data can tell us.

Start with what the data says directly

The strongest calendar year was 2016, with 538.1 million rides. By 2019, annual ridership was 525.5 million—already 2.3% below that peak. The pandemic interrupted a system that had plateaued.

Ridership fell 57.2% in 2020 compared with 2019. The trough followed in 2021, at 197.8 million rides.

In 2025, the monthly observations sum to 413.1 million rides, or 78.6% of 2019. That leaves an annual shortfall of approximately 112.4 million rides. These are totals of rides, not counts of unique passengers.

The extract's 2026 data ends in September. Comparing those nine reporting periods with January–September 2025 gives 303.4 million versus 306.5 million rides, down 1.0%. Comparing partial 2026 with a full calendar year would create a misleading decline.

The analytical work

The most useful discovery: check the denominator

A five-week reporting period can contain 25% more days than a four-week period. A larger ride total may therefore reflect a longer measurement window rather than stronger daily demand.

The notebook finds two distinct groups in monthly rides divided by average weekday rides, consistent with a 4-4-5 reporting calendar. It constructs an assumed table of 28- and 35-day periods to investigate that possibility.

This calendar is inferred, not confirmed by the metadata checked for the project. Daily rates, adjusted seasonal indices, and estimated non-weekday demand depend on it. The annual-total chart above does not use that adjustment.

The difference can materially change the story. For December 2025 versus December 2019, the raw totals suggest 98.5% recovery. Under the assumed period lengths, the per-reporting-day comparison is 78.8%. This is a conditional result that needs the reporting calendar verified before it can support an operational decision.

Findings that need extra care

Peak versus non-peak: the published peak share lands close to a whole percentage in seven of 23 years, including every year from 2022 to 2025. This may reflect rounding or estimation. Small changes in that split should not be interpreted before confirming the methodology.

Weekday versus non-weekday: the notebook compares a directly published weekday measure with a non-weekday rate estimated under the assumed calendar. Those estimates are useful for investigation, but they are not independently observed weekend counts.

Travel purpose: lower weekday recovery does not prove that commuting caused the shortfall. The extract contains no trip-purpose field and cannot distinguish work from education, shopping, leisure, or appointments.

What this could help a transit team investigate

The system-wide gap is a starting point for further questions: which routes or service periods account for the shortfall, how demand relates to service availability, and whether changes in fare revenue tell a consistent story.

This analysis cannot answer those questions alone. It has no route, line, station, service-frequency, or trip-purpose detail. The next step would be to add those sources and verify the reporting calendar before making service recommendations.

Reproducibility and scope

The project repository includes the CSV, analysis notebook, pinned requirements, and exported charts. The dataset is attributed to the City of Toronto under the Open Government Licence – Toronto.

The website's annual chart was recalculated from the committed monthly CSV. Its snapshot covers January 2008 through September 2026; only complete years through 2025 appear in the chart. Upstream revisions may change historical figures.

My takeaway: a convincing chart starts with knowing what was counted, over what interval, and under which assumptions.

This case study documents a portfolio analysis. Findings describe this dataset and should be interpreted within its limitations. Dataset source ↗
Explore more projects