Non-parametric volatility modeling in macroeconomics and finance

The global outbreak of the Covid-19 pandemic led to unprecedented increases in the variance of many key macroeconomic and financial quantities such as output, unemployment and stock prices. Existing volatility models have difficulties in capturing this sort of non-linearities in the conditional variance because they assume smoothly evolving variances. This proposal sets forth a research agenda to develop and apply more flexible volatility models for macroeconomic and financial data. These models will be developed in three steps. First, we focus on univariate non-parametric stochastic volatility models that remain agnostic on the precise law of motion of the variance process. In a
second step, we extend these models to multivariate time series, using insights from the literature on factor and common stochastic volatility modeling techniques. In a final step, our aim is to propose a mixed frequency multivariate stochastic volatility. This model will be able to extract information on how volatilities evolve at higher – daily and weekly – frequencies to infer about the volatility of macroeconomic time series at lower – quarterly and monthly – frequencies. Apart from providing methodological contributions the project also aims to apply these models and techniques to questions of high relevance for central banks and other policy institutions. We aim to use the models and methods developed in the course of the project to questions related to macroe-
conomic forecasting, the analysis of the causes and consequences of macroeconomic volatility, and for developing high frequency measures of macroeconomic stability which can be used as monitoring tools.