Return to search

The time-variation in style effects in the UK stock market

The thesis extends most previous studies on static version of style effects in the overall period to their time-varying properties in the dynamic macroeconomic conditions and market states in the UK Stock Market. It deals with four research questions on style effects in the UK Stock Market in four empirical chapters, respectively. Firstly, the thesis uses two indicators, long/short return Rlms and style coefficient (Sd(B, to examine whether the time-variation in style effects with short- /medium-/long-term horizon exists in the UK Stock Market. The research finds stronger momentum effect with 6/12 months' formation horizon and 6/12 months' holding horizon from 1956 to 2008, but its volatile pattern over time with the strongest in the 1990s and the weakest in the 1970s. Empirical evidence presents no existence of significant size effect from 1979 to 2008 due to the rotation between (marginally) significant small-cap effect in the 1980s and marginally significant large-c ap effect in the 1990s. It finds significant value effect, strengthening from 5 months' to 24 months' holding horizon, from 1980 to 2008 due to persistent outperformance of shares with low price-to-book ratio over time. Secondly, the thesis applies time-series regression to investigate what factors can significantly explain time-varying style effects in the overall period from 1980 to 2008 and three subperiods. It finds that macroeconomic variables, such as annual change of GDP(GDP(Y)), unexpected annual change of TBILL(UEXTBILL(Y)), annual change of CPI(CPI(Y)), and lagged 8-year's market return (MR(-8)), can offer significant explanatory power for the time-variation in momentum effect (12 12/12), size effect (12), and value effect (24) stronger than corresponding style effects with other holding horizons from 1980 to 2008. The decline in GDP(Y) and the increase in UEXTBILL(Y) imply strong momentum effect and large-cap effect. The decline in CPI(Y) and high MR(-8) means st sensitivity to the same economic forces can explain the findings on correlation among style effects. The thesis shows the variation in significant macroeconomic variables to explain style effects through holding horizons and over time. Thirdly, the thesis discusses whether the dynamic style effects can be indeed predictable by using some lagged macroeconomic variables from 1980 to 2008. It uses statistics PIS, POOS, and PSS to test the performance of recursive in-sample regression model and out-of-sample forecasting model relative to sample mean model. With the model test, the research examines the reliability/or pitfall of insample predictability. Different from momentum effect (12/6, 12/12) and size effect (12), out-of-sample forecasting model for value effect (24) performs poorly relative to the historical mean model even though its in-sample regression model does better. The test shows that the recursive out-of-sample forecasting model offers succes sful signal to capture momentum effect (12/6, 12/12) in higher percentage of all observations, followed by size effect (12) and value effect (24). Finally, the thesis explores whether active trading strategies with filter threshold and signal from out-of-sample forecasting model can capture style rotation from 1980 to 2008. Empirical work shows the significant gains to timing momentum effect. Active trading strategy on any equal-weighted decile momentum portfolio (12/6) and value. Thirdly, the thesis discusses whether the dynamic style effects can be indeed predictable by using some lagged macroeconomic variables from 1980 to 2008. It uses statistics PIS, POOS, and PSS to test the performance of recursive in-sample regression model and out-of-sample forecasting model relative to sample mean model. With the model test, the research examines the reliability/or pitfall of insample predictability. Different from momentum effect (12/6, 12/12) and size effect (12), out-of-sample forecasting model for value effect (24) performs poorly relative to the historical mean model even though its in-sample regression model does better. The test shows that the recursive out-of-sample forecasting model offers successful signal to capture momentum effect (12/6, 12/12) in higher percentage of all observations, followed by size effect (12) and value effect (24). Finally, the thesis explores whether active trading strategies with filter threshold and signal from out-of-sample forecasting model can capture style rotation from 1980 to 2008. Empirical work shows the significant gains to timing momentum effect. Active trading strategy on any equal-weighted decile momentum portfolio (12/6) and value effect (24).

Identiferoai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:576427
Date January 2013
CreatorsLi, Hong
PublisherUniversity of Strathclyde
Source SetsEthos UK
Detected LanguageEnglish
TypeElectronic Thesis or Dissertation
Sourcehttp://oleg.lib.strath.ac.uk:80/R/?func=dbin-jump-full&object_id=18971

Page generated in 0.0085 seconds