Addressing COVID-19 Outliers in BVARs with Stochastic Volatility

Abstract

The COVID-19 pandemic has led to enormous data movements that strongly affect parameters and forecasts from standard Bayesian vector autoregressions (BVARs). To address these issues, we propose BVAR models with outlier-augmented stochastic volatility (SV) that combine transitory and persistent changes in volatility. The resulting density forecasts are much less sensitive to outliers in the data than standard BVARs. Predictive Bayes factors indicate that our outlier-augmented SV model provides the best fit for the pandemic period, as well as for earlier subsamples of high volatility. In historical forecasting, outlier-augmented SV schemes fare at least as well as a conventional SV model.

The euro area’s pandemic recession: A DSGE-based interpretation

“… Lenza and Primiceri (2022) estimate a VAR for the US at monthly frequency and explicitly model the change in shock volatility with priors on volatility scaling factors for March-May 2020 and on a decay parameter that determines the speed of convergence of the covariance towards the pre-COVID-19 values. Carriero et al. (2022) conduct a Bayesian VAR analysis and show that allowing for stochastic volatility together with volatility outliers and, possibly, fat-tailed errors produces estimates and forecasts that are less sensitive to COVID-19 realisations.…”

Constructing fan charts from the ragged edge of SPF forecasts

“…19 The additive outliers serve to lower the correlations among forecast updates for different forecast horizons. In contrast, the multiplicative outliers used by Carriero, et al (2021) in vector autoregressions with stochastic volatility lower the persistence of volatility changes.…”

Censored Density Forecasts: Production and Evaluation

“…It is clear that the 2020 pandemic has again raised questions about how forecast errors should be treated. A related literature considers how models used to produce density forecasts should be designed, given the extreme data realizations and consequent forecasting errors observed during the pandemic; for example, see Schorfheide and Song (2020), Carriero et al (2022), Lenza and Primiceri (2022), and Huber et al (forthcoming).…”

“…Our approach thus offers an alternative to model-based methods such as those of Carriero et al (2022), who, when modeling and forecasting macroeconomic data, downweight extreme observations, such as those observed during the COVID-19 pandemic, by allowing for both persistent and temporary heteroscedasticity. Transitory outliers are modeled either by t-distributed error processes or by parameterizing an outlier volatility state as in the so-called stochastic volatility outlier-adjusted (SVO) model of Stock and Watson (2016).…”