The authors address empirically the implications of structural breaks in the variance-covariance matrix of inflation and import prices for changes in pass-through. They define pass-through within a correlated vector autoregression (VAR) framework as the response of domestic inflation to an impulse in import price inflation. This approach allows them to examine changes in both the amount and the duration of pass-through.

The authors develop a test to establish the presence of structural breaks in the error covariance matrix of a multivariate system of equations. The test extends the breaks-in-variance/covariance test proposed by Anderson (1971) by accounting for regression covariates and an unknown break date. It is exact for fixed regressors.

The results of the test reveal evidence of breaks in the covariance matrix in 1984Q3 and in 1991Q1. Estimating the VAR and examining impulse responses over the relevant subsamples, the authors find that, while the initial impact of pass-through is quantitatively similar across the two samples, the impact dissipates twice as fast in the later subsample. In other words, the duration of pass-through has declined from about a two-year period to a one-year period. The authors also document important changes over time in the estimated correlation between domestic and import inflation, both in terms of magnitude and sign.