Decomposing and Reconstructing Time-Series Momentum with Discrete Wavelet Transform

Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

School of Business | Master's thesis

Authors

Date

2016

Department

Rahoituksen laitos

Major/Subject

Mcode

Degree programme

Finance

Language

en

Pages

31

Series

Abstract

OBJECTIVES OF THE STUDY: The momentum phenomenon is one of the most studied phenomena in finance, covered by research within various geographical markets, time frames, and asset types. The methodologies used for portfolio creation in these research studies are very diverse. This thesis contributes to the research gap by examining time-series momentum using wavelet decomposition in momentum portfolio creation. Wavelet methodologies fit research within time series in finance and economics particularly well, as decomposition into wavelet coefficients allows analyzing processes that occur naturally at different time scales. Traders make decisions at various time intervals ranging from minutes to years, and wavelet decomposition allows examining these intervals closer. DATA AND METHODOLOGY: My sample consists of monthly S&P 500 returns over the period of January 1965 to December 2015. This study replicates time series momentum portfolios of Moskowitz et al. (2012) and supplements the initial portfolio creation strategies with wavelet decomposition with the aim of finding main constituents to momentum and denoising the data used as a basis to form the portfolios. FINDINGS OF THE STUDY: The analysis shows strong evidence that wavelet decomposition improves portfolio creation and provides almost seven times higher returns than the literature standard time-series momentum strategy over the in-sample period. However, the higher returns occur through a higher portfolio turnover rate. The findings suggest that the decomposed trading signal is more sensitive to changes in momentum in the S&P 500 Index, and due to the higher number of transactions required compared to the classic strategy, a proper real-world implementation requires filtering further the statistical significance of the trading signal in order to avoid unnecessary transactions.

Description

Thesis advisor

Jokivuolle, Esa
Suominen, Matti

Keywords

wavelet decomposition, momentum effect, signal processing, time-series

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-201702142342

Collections

[gradu] Kauppakorkeakoulu / BIZ