Overview
Pierre Perron is an econometrician at Boston University (Université de Montréal at the time of this work), specializing in unit root testing, structural change, and asymptotic theory for integrated processes. He is best known for demonstrating that standard unit root tests fail to account for structural breaks (Perron 1989) and for developing modified unit root tests with better size properties in the presence of dependent errors (Perron-Ng 1996). Not to be confused with François Perron, the statistician at Université de Montréal who works on Bayesian nonparametrics.
Key Contributions / Features
- Structural break unit root test (Perron 1989): demonstrated that standard DF tests have zero asymptotic power against trend-stationary alternatives with a single large structural break (Theorem 1); derived Brownian-motion limiting distributions indexed by the break fraction λ=TB/T under the unit root null (Theorem 2); three models — A (crash), B (changing growth), C (both); AO vs. IO test regression forms; reversed 11 of 13 Nelson-Plosser unit root findings at the 1% or 2.5% level by conditioning on the 1929 crash and 1973 oil shock as exogenous events.
- Local asymptotic theory (1994, with S. Nabeya): derived limiting distributions of AR(1) estimators and test statistics in three local-to-unity frameworks (near-white-noise/near-integrated, near-twice-integrated, near-seasonally-integrated).
- M statistics for unit root testing (1996, with S. Ng): proved that M tests (MZα, MSB, MZt) with the AR spectral density estimator are immune to size distortions that plague PP tests in problematic error specifications; provided local asymptotic explanation.
- Cointegration testing (1993, with J.Y. Campbell): showed that Johansen's (1991) cointegration procedure is distorted when a linear trend is present.
Related