Advanced

Serial Dependence and Portfolio Performance in the Swedish Stock Market

Hummel, Niklas LU (2016) NEKN02 20161
Department of Economics
Abstract
This paper studies the possibility to exploit linear dependence in stock returns of the Swedish OMX 30 index. The main model studied in the paper is a Vector autoregressive (VAR) model. Ten years of data from the OMX 30 index is used, consisting of 27 stocks for the period 2006-2015 which is transformed into daily, weekly and monthly returns. First the significance of the models is tested and it is found that for the daily and weekly data all of the models are significant. For monthly data 77 % of the estimated models are significant. After this the performance of a VAR arbitrage portfolio is evaluated and compared to the performance of a contrarian and an unconditional sample mean arbitrage portfolio. For daily data the VAR model works... (More)
This paper studies the possibility to exploit linear dependence in stock returns of the Swedish OMX 30 index. The main model studied in the paper is a Vector autoregressive (VAR) model. Ten years of data from the OMX 30 index is used, consisting of 27 stocks for the period 2006-2015 which is transformed into daily, weekly and monthly returns. First the significance of the models is tested and it is found that for the daily and weekly data all of the models are significant. For monthly data 77 % of the estimated models are significant. After this the performance of a VAR arbitrage portfolio is evaluated and compared to the performance of a contrarian and an unconditional sample mean arbitrage portfolio. For daily data the VAR model works quite well, yielding a higher Sharpe ratio than the two other arbitrage portfolios. For weekly and monthly data, the performance is not as good. In the weekly case the mean return is negative and in the monthly case the mean return is close to zero. After this a conditional VAR investment portfolio is formed and compared with the benchmark minimum-variance and mean variance portfolios. For daily data the VAR investment portfolio yields a good return, but also exhibits a high variance and very high turnover. For weekly data both the variance and the mean return is very high with a slightly better Sharpe ratio than for daily data. In the case with monthly data the model has a very high negative return, which can best be explained by the fact that the model is not well behaved in this case. The short sale constraints introduced seem inadequate at reducing the turnover of the VAR portfolios in the sense that even though the turnover is reduced so is the Sharpe ratios. Norm constraints was a better alternative to reducing turnover while also improving the Sharpe ratios. However, the tradeoff between mean return and turnover in the presence of transaction cost made it unclear whether the VAR portfolio strategy is a better alternative than using the traditional minimum-variance or mean-variance portfolios (Less)
Please use this url to cite or link to this publication:
author
Hummel, Niklas LU
supervisor
organization
course
NEKN02 20161
year
type
H1 - Master's Degree (One Year)
subject
keywords
Vector autoregression, portfolio choice
language
English
id
8890477
date added to LUP
2016-09-05 14:07:07
date last changed
2016-09-05 14:07:07
@misc{8890477,
  abstract     = {This paper studies the possibility to exploit linear dependence in stock returns of the Swedish OMX 30 index. The main model studied in the paper is a Vector autoregressive (VAR) model. Ten years of data from the OMX 30 index is used, consisting of 27 stocks for the period 2006-2015 which is transformed into daily, weekly and monthly returns. First the significance of the models is tested and it is found that for the daily and weekly data all of the models are significant. For monthly data 77 % of the estimated models are significant. After this the performance of a VAR arbitrage portfolio is evaluated and compared to the performance of a contrarian and an unconditional sample mean arbitrage portfolio. For daily data the VAR model works quite well, yielding a higher Sharpe ratio than the two other arbitrage portfolios. For weekly and monthly data, the performance is not as good. In the weekly case the mean return is negative and in the monthly case the mean return is close to zero. After this a conditional VAR investment portfolio is formed and compared with the benchmark minimum-variance and mean variance portfolios. For daily data the VAR investment portfolio yields a good return, but also exhibits a high variance and very high turnover. For weekly data both the variance and the mean return is very high with a slightly better Sharpe ratio than for daily data. In the case with monthly data the model has a very high negative return, which can best be explained by the fact that the model is not well behaved in this case. The short sale constraints introduced seem inadequate at reducing the turnover of the VAR portfolios in the sense that even though the turnover is reduced so is the Sharpe ratios. Norm constraints was a better alternative to reducing turnover while also improving the Sharpe ratios. However, the tradeoff between mean return and turnover in the presence of transaction cost made it unclear whether the VAR portfolio strategy is a better alternative than using the traditional minimum-variance or mean-variance portfolios},
  author       = {Hummel, Niklas},
  keyword      = {Vector autoregression,portfolio choice},
  language     = {eng},
  note         = {Student Paper},
  title        = {Serial Dependence and Portfolio Performance in the Swedish Stock Market},
  year         = {2016},
}