Research
Working papers
Time varying kernel densities as dynamic infinite mixture models.
Building on kernel density estimation for time series data, we introduce the family of Dynamic Infinite Mixture Models (DIMMs). DIMMs approximate the time-varying distribution of a time series with that of an infinite mixture of location-scale random variables. Different specifications of a DIMM can capture different properties of the time series of interest, such as different memory properties of the predictive mean and asymmetric effects in the predictive variance. A maximum likelihood estimator is proposed. Its asymptotic properties are studied under a fully misspecified setting and its finite sample behaviour is assessed in a Monte Carlo analysis. An application to US GDP growth shows that DIMMs: i) reliably track the time-varying distribution of interest; ii) perform on par with, if not better than, a fully parametric model when it comes to predicting probability density functions.
First draft to appear soon.
Economic vulnerability is state dependent (with Leopoldo Catania and Alessandra Luati).
This paper shows that different states of the financial system determine a different marginal effect of financial conditions on economic vulnerability. As soon as financial conditions start deteriorating, the economic outlook becomes more pessimistic and uncertain. No increase in macroeconomic uncertainty is expected when financial conditions worsen from an already tighter than usual situation. Additionally, past information on GDP growth turns out to be of paramount importance when it comes to studying and predicting economic vulnerability. Both findings have relevant forecasting and policy-making implications, and persist once other measures of the real economic activity are considered. The analysis relies on a new methodology for the dynamic modelling of multiple quantiles in the presence of an explanatory variable. This new approach exploits all past information on GDP growth and it can accommodate a state dependent marginal effect of financial conditions.
Available on SSRN.Combining dynamic conditional quantile functions with a view towards tail risk management (with Leopoldo Catania and Alessandra Luati).
This paper introduces a new method to model the quantiles of a time series using all past information on a set of explanatory variables and on the time series of interest. The resulting quantiles: i) do not cross over time, ii) have a dynamics which enhances the information set available to extreme quantiles, and iii) incorporate information coming from all regions of the conditional distributions of the explanatory variables. Parameters of the model are estimated through a two-stage quasi maximum likelihood estimator (2SQMLE) whose finite sample properties are investigated through a simulation study. An empirical analysis concerning macro-financial variables reveals a tight connection between financial and macroeconomic tail risk, and shows that the model performs well at making density and tail risk predictions.
First draft to appear soon.
Commonalities in large panels of option prices (with Maria Grith, Paolo Santucci de Magistris and Francesco Violante).
We use a multivariate functional principal component analysis to study commonalities in the implied volatility surfaces of the cross section of US equities. We find that technology-related and financial firms exhibit a higher than average implied volatility. The former accrue this exceedance during the Dot-com bubble, while the latter are more exposed to the Great Financial Crisis. Both findings hold true across different times to maturity and levels of moneyness.
First draft to appear soon.