Contributions published at Financial Engineering (Markus Leippold)

Given a theoretical pricing model, an implied volatility can be extracted from an option’s market price. Given a set of options with the same maturity and a range of strike prices, it is possible to extract (an approximation to) the entire risk-neutral probability density without having to assume a theoretical pricing model. There are a variety of related methods to do this, but all are subject to certain problems, including the fact that the data never exist to allow full estimation of the tails. Some methods produce improper densities with negative portions. In this article, Ludwig introduces a neural network approach to extract risk-neutral densities from option prices, imposing only a small number of constraints, such as probabilities must be nonnegative and an option’s price must be above intrinsic value. The resulting densities are smooth and sensible, even for days that other approaches find extremely difficult to handle.

Verschiedene wissenschaftliche Arbeiten zeigen, dass eine Covered-Call Strategie auf einen Aktienindex risikoadjustiert besser abschneidet als der Index selbst. Dieses Ergebnis steht im Widerspruch zur Theorie, weshalb in dieser Arbeit mögliche Gründe hierfür zusammengefasst werden. Dazu werden bisherige Studien zu diesem Thema beschrieben und zentrale Ergebnisse vorgestellt. Anschliessend werden gezielt verschiedene Covered-Call Strategien auf den S&P 500 Index untersucht und mit diesem verglichen. Es zeigt sich, dass Covered-Call Portfolios einen risikoadjustierten Mehrwert generieren können, jedoch hängt dies von den verwendeten Call Optionen und dem beobachteten Zeitraum ab.

n this thesis, we examine the challenges of sequentially calibrating options using nonlinear filters. We inves- tigate the accuracy and computational costs of the extended Kalman filter, iterated extended Kalman filter and the particle filter. We compare it to more traditional calibration techniques of (penalized) weighted least squares. Moreover, we apply a heuristic optimization, the Differential Evolution algorithm. A semi-closed expression of European call option prices by the means of a Fourier transforms makes the option pricing fast and suitable for many models. We apply all methods across three stochastic volatility models; the Heston and Double Heston as well as adding jumps in the asset dynamics by using the Bates model. We conclude the thesis with an empirical study on sequential calibration using S&P500 index options for one year time horizon. It is found that the nonlinear filters are able to fit these models to daily European call option data in a considerably fast and accurate manner, even outperforming traditional techniques. They are further able to treat time-varying parameters in a consistent and stable kind with the iterated extended Kalman filter clearly standing out in terms of fit. The implementation of the filters, however, was found to be very involved as the algorithms are very sensitive to noisy input data as the algorithms have no internal mechanism to recover from shocks.

Since the recent financial crisis the financial industries demand for models able to answer ”what-if” questions has considerably increased. One way of answering this type of question is based on scenario analysis. Interestingly, most of the existing scenario generators do not directly incorporate the implied volatility surface as a separate risk factor. This is surprising since implied volatilities are directly ob- servable from option markets and empirical research has accumulated convincing evidence that implied volatility movements are also driven by factors beyond movements in the underlying asset. Accordingly, this thesis addresses the challenge to derive a model that is able to generate scenarios for implied volatil- ity surfaces in conjunction with their underlying assets that are similar to those seen in historical data. Therefore, the entire implied volatility surface is expressed by a small number of (volatility risk) factors. Different approaches including their ability to fit given implied volatility surfaces are discussed. Once this is accomplished, a stochastic model for the volatility risk factors and the underlying asset is specified determining the dynamic behavior of the implied volatility surface. In particular, it is shown how the model can be calibrated to a historical time series of option prices.

We find that stocks exhibiting high dispersion in analysts' earnings forecasts do not only underperform in the U.S. but also in some European countries. However, testing for the dispersion effect in many countries calls for adequate multiple testing controls. Under this paradigm it turns out that none of the naively derived dispersion effects proves to be a sustainable phenomenon. Rationalizing this finding, we document that the dispersion effect's abnormal returns amass in a very narrow time frame and mainly derive from a bet against the technology bubble that would have been rather difficult to implement.

null

null

null

The goal of this thesis is to analyse the currency risk hedging at mutual funds, for which purpose a sample of more than 3’500 so-called (currency) hedged share classes (HSC) are analysed. One of the main research hypotheses was: ”The hedging performance of hedged share classes does not differ significantly from a hedging benchmark”. To approach this question, the actual realized hedged returns of hedged share classes are compared to that of a theoretical perfectly hedged benchmark. This theoretical benchmark is also challenged, with respect to real world implementation difficulties, to identify whether it could actually be achieved. The analysis concludes that the overall average hedged share class results in excess hedging costs, compared to the theoretical benchmark, of almost 40 basispoints per annum. Further analysis shows, that there is great dispersion between the hedged share classes of different asset classes or currency pairs. Especially hedging the foreign exchange (FX) risk of equity funds seems to involve higher cost, than that of any other asset class. Probably the most alarming result is the dispersion among different fund providers. By sorting the HSCs of different fund providers into quartiles, based on their hedging performance, one can find that the worst quartile has a significantly higher hedging cost, than the best quartile. This difference is around 100 basispoints per annum and can only be partially explained by differences in the underlying asset classes among the respective fund providers. Further, the implementation of a theoretically perfectly hedged portfolio was simulated. Its actual implementation was challenged with respect to various known real world hedging problems. The result is, that a perfect hedge is not achievable in reality, and that the analysed implementation challenges are able to explain the major part of the deviation of HSC from the benchmark. The overall conclusion is, that FX hedging is not as trivial as much academical research assumes, and that future academic work, for example around optimal hedging policies, should also have its practicability in mind. Finally proper hedging evaluation and controlling is important and must differentiate between hedging strategies, asset classes and currency pairs.

null

null

null

In Folge der Finanzkrise kam es zu regelrechten Regulationswellen innerhalb der wichtigsten Finanzmärkte. Der Zweck davon war es, die in den Medien als schädlich bezeichneten Leerverkäufe einzuschränken und somit die Krise einzudämmen. Es stellt sich daher die Frage, welche Auswirkungen solche Regulationen auf das Funktionieren der Märkte haben. Es wird in dieser Arbeit überprüft, welchen Einfluss (Teil-)Verbote von Leerverkäufen auf die Effizienz in Aktienmärkten haben. Die Analyse ergibt einen negativen Einfluss, was bedeutet, dass die Verbote dazu führen, dass negative Informationen verzögert in die Preisbildung gelangen. Infolge dessen kommt es zu temporären Überbewertungen von Aktien, was den Markt verzerrt.

Abstract Diese Arbeit untersucht den Effekt des Konjunkturzyklus auf das Beta von 18 SMI Ak- tien. Anhand der SPI-Performance und des BIP-Wachstums werden verschiedene Zu- stände zur Darstellung des Konjunkturzyklus gebildet. Für jede Aktie wird ein Beta pro Zustand berechnet und anschliessend untereinander verglichen. Für die meisten Aktien sind in verschiedenen Zuständen signifikante Unterschiede des Betas zu erkennen. Ein einheitlicher Einfluss auf die Betas verschiedener Aktien ist nicht festzustellen. Daher kann keine allgemeingültige Aussage für eine ganze Branche oder sogar über alle unter- suchten Aktien hinweg gemacht werden. Trotzdem lassen die Untersuchungen den Schluss zu, dass das Beta in verschiedenen Konjunkturzuständen nicht gleich sein muss.

A quality security is defined as a stock that is profitable, growing, safe and well-managed. Across the globe and especially within the US, high-quality stocks do have higher prices on average. Within the Swiss stock market however, this is just the case during crisis. Because of the puzzling phenomenon of asset prices, long/short strategies based on a quality score can earn positive risk-adjusted returns within Switzerland. In general, a score weighted approach outperforms other investment strategies. Opposing previous research, the performance of a QMJ factor, considering the stocks of the Swiss Performance Index, is not significant at a 5% level.

Die vorliegende Arbeit dokumentiert Handelsstrategien, welche Aktien kaufen, die in der Ver-gangenheit gut abgeschnitten haben und die Aktien leerverkaufen, welche in der Vergangenheit schlecht abgeschnitten haben. Die Untersuchungen ergaben für Halteperioden von 3 - 9 Monaten statistisch signifikante, marktadjustierte Überrenditen. Für die Gültigkeit des verwendeten Bewer-tungsmodells konnte zusätzlich gezeigt werden, dass die Erfolge nicht durch zusätzlich eingegan-gene Risiken zustande gekommen sind. Weiter konnten systematische, positive Überrenditen der Momentumstrategien über den gesamten Untersuchungsraum festgehalten werden. Der Korrela-tionsfaktor zwischen den Erfolgen der Momentumstrategien und der Konjunktur, gemessen am realen BIP-Wachstum, bewegten sich nahe Null und deutet auf keinen linearen Zusammenhang hin.

Regelmässig wird die Relevanz von Kapitalmarktmodellen diskutiert. Oft geht es um deren Aussagekraft in Bezug auf die Erklärung und Vorhersage von Investitionsentscheidungen. Anhand einer Analyse von 60 Portfolios einer Schweizer Privatbank wird gezeigt, dass sich die Portfoliozusammensetzung auch bei differenzierter Betrachtung erwarteter Renditen und Verwendung des alternativen Risikomasses Semivarianz nicht abschliessend durch etablierte Kapitalmarktmodelle wie MPT, CAPM oder APT erklären lässt. Trotzdem werden bedeu-tende Unterschiede zwischen EVV- und bankgemanagten Portfolios festgestellt. Anschlies-send wird argumentiert, warum ein Einsatz insbesondere der APT im Rahmen der strategi-schen Portfoliostrukturierung sinnvoll ist und dargelegt, dass durchaus Abwandlungen der Modelle in der Praxis verwendet werden.

Collaterized Debt Obligations (CDOs) have been responsible for massive write-downs in the financial crisis of 2007-2008,- as it seems that many involved companies, professionals and parties underestimated the dangers of the CDOs which was enhanced by the overoptimistic ratings given by credit rating agencies. Using data from the Open Source Model given by Pershing Square Capital Manage- ment, this thesis shall give an empirical approach to the analysis of the CDO market in 2007-2008. The scrutiny shows that some credit ratings had no relationship to their collateral, that the CDO rating market consisted of an inefficient duopoly and that the CDOs performance between the banks varied widely. This thesis shows that the problems in the CDO market are of critical manner and have to be reassessed to prevent a further market crash.