Capra, Jacopo T.; (2024) Modern statistics applications to systematic trading: low and high frequency perspectives. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Capra_10202690_thesis_revised.pdf Download (14MB) | Preview

Abstract

Macro asset allocation problems can be formulated as a two stage problem process estimating the moments of the return distributions in the first step and solving a portfolio optimization in the second step. As the macro data have a typical low frequency nature the allocation problem resorts either to use low frequency inputs or to rely on price based signals which should reflect the macro data dynamics. In this thesis, I have used a novel approach based on a Bayesian framework to solve macro asset allocation problems which are casted as a multivariate dynamic linear model where the states govern the sensitivity of asset returns to the macro variables. In particular the modelling relies on nowcasted higher frequency macro data enhanced by sentiment based macro news provided by RavenPack. Therefore high frequency macro news are used in a low frequency context for refining the estimation of the macro economic uncertainty. This is a novel attempt in the literature to combine higher frequency nowcasting macro indicators with high frequency macro news data to solve an asset allocation problem. This asset allocation framework is then empirically tested for constructing portfolios of cross asset risk premia strategies. These have become very popular among a growing number of investors, particularly large and sophisticated pension funds, who instead of viewing their portfolios as just a collection of traditional assets have come to view them as a collection of exposures to particular risk factors or market anomalies. Risk premia strategies aim exactly to capture return premiums which are a direct and purer compensation for the systematic exposure to particular risks, or to particular investors’ behaviour anomalies. Moreover this thesis develops and tests a high frequency framework to trade US single stocks leveraging off the information provided by the RavenPack high frequency macro news (which are then used in a more classical/high frequency context). The novelty of this part is also related to the use of reservoir computing in the context of a multi horizon prediction experiment applied to high frequency trading.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item