Not logged in.

Quick Search - Contribution

Contribution Details

Type	Master's Thesis
Scope	Discipline-based scholarship
Title	Risk-Based Market Timing Across Asset Classes
Organization Unit	Financial Economics (Thorsten Hens)
Authors	Vincent Maria Zandanell
Supervisors	Alexandre Ziegler
Language	English
Institution	University of Zurich
Faculty	Faculty of Business, Economics and Informatics
Number of Pages	93
Date	2021
Zusammenfassung	Risk measures provide a perspective on the risk associated with the underlying security but cannot measure the true risk of the underlying security. Rather, they measure the risk based on the metric they are based on. Realized volatility as an example is based on price diﬀerences of the underlying asset over time and therefore simply measures the risk of price diﬀerences over time or put otherwise, it measures the changes of the agreed upon price of the underlying security over time. Therefore, if new information enters the market, realized volatility increases if buyers and sellers of the aﬀected security agree upon the diﬀerent new price adjusted for the new information. If the new information is not agreed upon in an instant as eﬃcient markets would suggest, then the new information might increase the realized volatility of the underlying asset even further as there is a diﬀerent agreement on the price for each observation of a given time frequency. This thesis tries to capture the days where new information is lingering in the price ﬁnding process for multiple days or new information constantly enters the market achieving as similar eﬀect. With this in mind, the model will not be able to discriminate on whether the introduced volatility is based on expected risk, unexpected risk, or fear. New information enters the market constantly and is not uncommon and with our risk overlay we do not want to miss the expected return based on the expected daily stream of new information entering the market. We want to capture the rare information entering the market that is capable of producing very unlikely turmoil in the price ﬁnding process for multiple days or even months. In this thesis we analyze whether the chosen risk measures can capture this disturbance and whether they should be avoided. This thesis focuses its market timing eﬀorts on the US market regarding three asset classes, namely, stocks, bonds, and gold. These asset classes will be timed by risk measures based on the realized volatility, the implied volatility, and the variance risk premium from stocks, bonds, and gold. We ﬁt proper distributions to each risk measure in order to determ-ine the risk measurement levels we want to avoid. In the in-sample test it was determined that a 5% unlikely cutoﬀ point is optimal. This allows us not only to be invested around 95% of the time but additionally avoid high risk scenarios. The benchmark is fully passively invested in the same risky asset class at all times and no risk overlay will be applied. The decision by the risk overlay whether to invest the upcoming day will happen daily and will be based on a 10-year rolling window preceding the current day. Additionally, we assume transaction costs of 15 basis points which will not only apply when the risk overlay issues a risk on or risk oﬀ signal but also during the quarterly rebalancing. The rebalancing is necessary as for stocks and bonds we are using multiple equally weighted but free-ﬂoating market indices to simulate a complete asset class. Thus, there will also be rebalancing costs for the passive benchmark. Further turnover will occur at the ﬁrst day of the out-of-sample period when the risky asset has to be bought. We deal with the rebalancing eﬀect by not allowing a risk on signal to re-enter the market with an equally weighted portfolio. Thus, if the risky asset is being bought after a risk oﬀ period, the weights of the indices of the risky asset will be the same as if the asset has never been sold. In the out-of-sample period we avoid any kind of forward-looking bias by assuring that all investment decisions are based on the 10-year rolling window which by the time of the daily investment decision will be historical data. In addition, we made sure to only use knowledge and data that was available at the time of each investment decision and the idea for the risk overlay is justiﬁable with the in-sample analysis and basic economic intuition. We conclude that fear associated with being invested during times of high volatility is reasonable. In this thesis we show how targeting and avoiding these scenarios can be done with the help of simple and readily available risk measure often directly associated with this fear. Achieving higher returns is not guaranteed but a decreased volatility can be achieved with nearly all tested risk measures, within and across asset classes. While the performance was lackluster during the in-sample tests, over the 18 year out-of-sample period, some risk measures stand out. If the desire is to time the stock market, the MOVE index as proxy for bond implied volatility not only nearly doubled the Sharpe ratio from 0.36 to 0.67 but also achieved 4.46 percentage points higher total returns p.a. compared to the benchmark. Timing the bond market can also lead to performance improvements. Here, the stock market variance risk premium as measured by the diﬀerence of the current realized volatility and the current VIX achieved superior Sharpe ratios than the benchmark with higher total returns and lower standard deviation. Close behind is the MOVE index. The performance improvements for bond market timing are however not as noticeable as for the stock market. Timing the gold market did not bring any noticeable improvement that could not otherwise be attributable to luck. None of the risk measures could achieve higher returns and only the gold realized volatility and the bond variance risk premium could achieve slightly higher Sharpe ratios by decreasing the standard deviation.
Export	BibTeX