Mortgage Stress Tests as Qualification Regimes: A Bayesian DSGE Model of Canadian Macroprudential Policy

Abstract

Between March 2022 and July 2023, the Bank of Canada raised the overnight rate by 475 basis points in its fastest tightening cycle in four decades. Yet, Canada avoided the household- sector distress many feared. This thesis asks through what mechanism Guideline B-20’s minimum qualifying rate reduces household debt vulnerability, and what determines whether the instrument is effective at all. We answer both questions with a Bayesian DSGE model featuring two borrower types and multi-period fixed-rate mortgages, estimated on 18 Canadian quarterly observables spanning 1999Q3–2025Q3. The stress test operates through two channels: a qualification channel that determines which constraint ceiling governs borrowing, and a buffer channel that tightens borrowing within the binding regime. At the baseline calibration, the qualification channel accounts for 90% of the debt reduction. Both channels collapse at a critical income-ceiling parameter: above it, the instrument is inert. Aggregate time series cannot locate the economy relative to this threshold, because the qualification channel’s effect is compositional rather than dynamic; resolving it requires loan-level data inside structural models with richer borrower heterogeneity. Conditional on binding, the stress test reduces aggregate mortgage debt by 2.82% and the debt-service ratio by 0.32 percentage points at a GDP cost of 0.07%. It is welfare-improving, progressively distributed, and more efficient than collateral-based alternatives calibrated to the same debt target. Buffer calibration matters less than whether the income constraint binds.