Fuel-Tax Pass-Through to the Pump

A known-shock event study of Canada's 1 April 2025 carbon-charge removal, with a 2026 excise replication. Pass-through as a ratio, not a coefficient. Open the interactive dashboard →

1. Why this is a clean tax-incidence question

Pass-through — the share of a tax change that reaches the consumer price — is the core object of competition, merger, and tax-incidence economics. It is usually hard to estimate because the "shock" is endogenous or its size is unknown. Canada's consumer carbon price on gasoline is the rare exception: the per-litre amount is published statutorily by the Canada Revenue Agency, and it was removed on a single, pre-announced date — 1 April 2025. That makes pass-through a ratio: the observed retail drop divided by the known  ¢/L charge.

2. The shocks and their statutory size

The federal gasoline fuel charge rose every April from 4.42 ¢/L (2019, $20/t) to  ¢/L (2024, $80/t) before being removed. Because GST/HST is levied on the carbon-inclusive pump price, removing the charge also strips the sales tax on it, so the expected pump drop is charge × (1 + GST/HST) =  ¢/L on average across treated cities. We benchmark pass-through against this GST-adjusted figure, not the bare charge. A second known shock — the federal excise suspension of 14 April 2026 — gives an independent replication.

3. Data and the central confounder

Weekly city retail gasoline prices (NRCan; StatCan monthly for the panel), a crude/wholesale benchmark (FRED WTI / EIA), and CRA statutory rates. The treated units are the federal-backstop provinces; British Columbia ended its own carbon tax the same day, so BC is treated, not a control. The clean no-removal arm is Quebec (cap-and-trade, retained).

The irreducible threat is oil-price confounding: the pump price moves with crude regardless of tax. Every event study below is run with and without the wholesale control; the tax effect must survive it.

4. Event study around 1 April 2025

An interrupted-time-series on treated-city weekly prices estimates the level shift, controlling for contemporaneous and lagged wholesale. The dynamic path (dashboard) is flat before the date and steps down after — pre-trends are flat in magnitude (largest pre-event lead  ¢/L).

The gap between the controlled and naive estimates is the headline methodological point:

5. Corroboration: difference-in-differences vs Quebec

Pooling treated and Quebec cities, the treated×post interaction is  ¢/L ( of the expected drop). Quebec's own margin is stable across the date (shift  ¢/L), supporting its validity as a control.

6. Speed and asymmetry — "rockets and feathers"

An asymmetric error-correction model lets retail prices adjust at different speeds to cost increases vs decreases. The estimated half-life is  weeks upward versus  weeks downward; the formal symmetry test gives .

7. Pass-through ratio

The headline: of the GST-adjusted statutory drop passed through, with a block-bootstrap 95% CI of . Complete pass-through (100%) inside the interval.

8. Replication on the 2026 excise suspension

9. Limitations