Understanding the macrofinance dataflow
Lead Research Organisation:
London School of Economics and Political Science
Department Name: Statistics
Abstract
Since the beginning of the twentieth century, the global economic environment has become increasingly interconnected with financial markets. The markets for listed equities and public debt were the main initial drivers of this transition. More recently, the proliferation of derivative products linked to equities and bonds, as well as mortgage-backed securities and credit derivatives, has further increased financial markets' sensitivity to changing macroeconomic conditions and multiplied the speed with which information is transmitted between financial markets and the real economy. Today, all of the traditional macroeconomic variables as well as relatively new indexes and surveys are closely followed by market participants as they are seen to have a significant influence on financial markets. However, finding empirical relationships among financial and macroeconomic variables remains difficult, as these data have complex dynamics
People |
ORCID iD |
| Filippo Pellegrino (Student) |
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| ES/P000622/1 | 01/10/2017 | 30/09/2027 | |||
| 1930500 | Studentship | ES/P000622/1 | 24/09/2017 | 30/06/2021 | Filippo Pellegrino |
| Description | Modelling complex time-varying phenomena is challenging. A common bottleneck with this class of variables is finding a reasonable approach to partition it in a way that does not alter the structure of the data. This is crucial to structure sophisticated techniques for forecasting and performing structural analyses. In the first part of my thesis, I developed a methodology (i.e., the artificial delete-d jackknife) that moves in this direction. I proposed to use it as a way to pre-tune a large class of forecasting models. Furthermore, I proved mathematically its efficiency. Empirical results based on macro-financial data show promising results. This work is available in a paper format and linked in this record. I am in the process of extending this methodology to exploit non-linearities in forecasting macro-financial time series. This is being done by using the artificial delete-d jackknife to bridge models from the classical (statistical) literature with those proposed in machine learning. |
| Exploitation Route | Time series analysis is at the base of a range of fields and studies. The outcome of my thesis could be used to improve the accuracy of forecasting models. A natural way to take forward my research is to use it for building structural analyses based on macroeconomic and financial data. This could help central bankers to understand the effects of the real economy on asset prices (and vice-versa). While the most obvious applications are within social sciences, this research has the potential to impact natural sciences as well. I suppose that the techniques I am proposing could prove to be beneficial for improving the accuracy and reliability of machine learning models. Thus, they might be put in use from researchers in STEM. |
| Sectors | Aerospace, Defence and Marine,Digital/Communication/Information Technologies (including Software),Financial Services, and Management Consultancy,Pharmaceuticals and Medical Biotechnology |
| URL | https://arxiv.org/abs/2002.04697 |