Title: Measuring aggregate and sectoral uncertainty
Authors: Luis Uzeda - Bank of Canada (Canada) [presenting]
Efrem Castelnuovo - University of Padova (Italy)
Kerem Tuzcuoglu - Bank of Canada (Canada)
Abstract: An empirical framework is proposed for the estimation of aggregate and sectoral uncertainty that is suitable for data-rich environments. Building upon previous works on multilevel factor and common stochastic volatility models, we jointly decompose the conditional mean and variance of economic time series into an aggregate and a sectoral component. Uncertainty -- aggregate and sectoral -- is captured by common factors driving the conditional variance of economic variables at different levels of aggregation in a large system. We apply this methodology to two large datasets for the U.S. economy. The results indicate that, while aggregate uncertainty dominates during sharp recessions, since the mid-1980s sectoral uncertainty has become more salient than its aggregate counterpart. An early assessment of the ongoing Covid-19 pandemic through the lens of our framework indicates that the current crisis led to an unprecedented spike of aggregate uncertainty, more than double compared with the Great Recession. Also, the durable and non-durable manufacturing sectors witnessed substantial sector-specific uncertainty at the onset of the pandemic.