Εντοπίστηκε ένα σφάλμα στη λειτουργία της ΠΥΞΙΔΑΣ όταν χρησιμοποιείται μέσω του προγράμματος περιήγησης Safari. Μέχρι να αποκατασταθεί το πρόβλημα, προτείνουμε τη χρήση εναλλακτικού browser όπως ο Chrome ή ο Firefox. A bug has been identified in the operation of the PYXIDA platform when accessed via the Safari browser. Until the problem is resolved, we recommend using an alternative browser such as Chrome or Firefox.
Λογότυπο αποθετηρίου

Διδακτορικές διατριβές

Μόνιμο URI για αυτήν τη συλλογήhttps://pyxida.aueb.gr/handle/123456789/34

Περιήγηση

Πλοήγηση Διδακτορικές διατριβές ανά Συγγραφέα "Dimitrakopoulos, Dimitris"

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω
Φιλτράρετε τα αποτελέσματα πληκτρολογώντας τα πρώτα γράμματα
Τώρα δείχνει 1 - 1 από 1
  • Μικρογραφία εικόνας
    Τεκμήριο
    Measuring market risk in financial and freight markets
    Dimitrakopoulos, Dimitris; Athens University of Economics and Business, Department of Accounting and Finance; Kavussanos, Manolis
    The recent global financial crisis has amply highlighted the importance of prudent management of financial risks for the well functioning of firms and the prevention of a systemic crisis. Risk measurement constitutes an mandatory step towards risk management which is linked tightly to other layers of the risk management process of such as risk reporting, limit setting, performance evaluation and risk budgeting. This thesis focuses on market risk by investigating the critical issue of market risk measurement. A main concern for risk taking agents and regulators is the accurate estimation of market risk which is directly linked to the process of the selection of a risk measurement model. The latter turns out to be a matter of empirical investigation depending on several factors complicating decisions on the specification of a preferable risk model, such as the properties the underlying assets for which market risk is estimated, the time period and the investment horizon over which the risk measurement models are estimated and the confidence level of risk estimations, amongst others. Additionally, although financial literature on market risk measurement has progressed substantially during the last decade leading to the development of a vast number of market risk measurement models and validation techniques, it remains inconclusive as to which method is preferable for the estimation of market risk. The aim of this thesis is to investigate the critical issue of market risk quantification by shedding some light on the controversial issue of market risk model selection. For this purpose two alternative risk metrics were utilized: The value at risk (VaR) and the expected tail loss (ETL). Broadly speaking, VaR and ETL are summary statistics that quantify in a probabilistic manner the exposure of an asset or portfolio to market risks. More specifically, VaR measures the loss that shall not be exceeded with a given probability over a fixed investment horizon; see for e.g. Jorion (1997). The ETL is defined as the expectation of the loss beyond the VaR level; see for e.g. Artzner et al. (1997). The markets chosen for the empirical analysis have a prominent role in the world economy and are extremely important for the global economic stability. Furthermore, little or no empirical research exists investigating the issue of market risk measurement for these markets. These are emerging equity markets and ocean freight markets for the liquid and the dry bulk segments of shipping. In order to elucidate different patterns in the behavior of the market risk measurement models and revisit the documented performance of the alternative methods in the case of mainstream financial assets, an additional sample of four developed markets equity indices was included. To this end, the employed sample consists of 18 international financial corporation emerging stock market indices as well as 4 developed equity market indices from the regions of Asia, America and Europe adding up to a diverse and representative sample of emerging and developed equity portfolios. As regards the dry bulk and the liquid bulk freight rate markets, aggregate and specific route indices of the Baltic Exchange are considered. A large number of VaR and ETL forecasting models are compared in an empirical evaluation assessment. These include: traditional models such as the random walk, the riskmetrics, GARCH variants and historical simulation based methods, extreme value methods such as the peaks over threshold and the block maxima method, and hybrid methods such as the EGARCH-VaR-X approach. These models were evaluated in terms of statistical accuracy by employing Christoffersen’s 1998, Lopez (1998) and Harvey et al. (1997) backtesting methods. There are certain idiosyncratic and in some cases unique attributes that pertain to the markets considered rendering the investigation of market risk measurement particularly interesting and challenging. These involve the turbulent nature of emerging and freight markets, the low correlation of emerging markets with the developed markets, the contagion effect in between emerging markets during periods of financial turmoil, the highly cyclical and seasonal behavior of freight rate returns and the longer term investment horizons being part of risk taking agents concerns who activate in freight rate markets. This thesis makes a number of important contributions: first, it demonstrates which market risk measurement methods are appropriate for the aforementioned markets. The importance of these markets for the world economy in conjunction with the fact that little or no empirical research exists investigating the issue of market risk measurement renders the investigation of this thesis particularly interesting; second, it addresses the issues of market risk estimation in association with the unique attributes present in these markets; third, it revisits the issue of market risk model selection for developed equity markets, allowing direct comparisons to be made with the results obtained from the assets investigated in this thesis; fourth, additional issues related to the risk measurement such as the estimation of medium term market risk and model selection during periods of crises or when the data generating process incorporates seasonality and cyclicality are addressed. Fifth, findings may guide decisions regarding risk estimation for other markets which share common characteristics with the aforementioned markets, such as the electricity and the agricultural commodity markets. The findings of this thesis can be summarized in the following: Overall, it was found that the fatter the tail of the empirical profit/loss distribution, the more difficult the estimation of market risk becomes. Specifically, few or no models managed to survive the backtesting for the markets which exhibit the fattest tails as measured by the tail index. This highlights the importance of using extreme value based diagnostic tools (such as the tail index estimates used in this thesis) in the practice of risk management to determine the properties of the tails of the profit/loss distributions and adapt risk modeling accordingly. The most successful selected VaR models were common for both emerging and developed equity markets despite the documented differences between these two asset classes standing out. However, in contrast to many studies on emerging markets selecting extreme value market risk measurement methods, the hereby thesis selects the simple non-parametric historical simulation based method of Boudoukh et al. (1998). In general, it was found that the tradeoff between statistical accuracy and complexity in the estimation of VaR was advantageous in the case of non-parametric models; in other words, these models seem to provide statistically sufficient risk forecasts just like the extreme value or other sophisticated VaR models do with comparatively fewer resources needed for their estimation. However, it is notable that although the same VaR model was selected for both emerging and developed markets, diversities concerning the performance of the same VaR models with respect to the estimation window were demonstrated: Specifically, most VaR models tend to overestimate and underestimate the realized VaR for the emerging and developed markets, respectively. Thus, if long estimation windows are employed for the estimation of the market risk in an emerging markets context, risk reporting for imposing capital reserve requirements may turn out to be costly due to the overestimation of risk. Another subtle point which deserves some attention is that special care should be taken in the case of post crisis periods during which the performance of the VaR models which included the crisis in the estimation sample deteriorated substantially. Assuming that this finding describes a typical pattern in the estimation of VaR during post-crisis periods, then it is very significant for the periods succeeding the current financial crisis as it suggests that most VaR models incorporating the crisis in their estimation sample, are likely to produce inaccurate risk forecasts. In the case of ocean freight rates findings indicate strong segregation between the markets examined as the performance of the alternative risk measurement models differed significantly leading to the selection of different models for each market. This finding contradicts literature for the majority of asset classes where the alternative risk measurement methods seem to perform similarly across the various assets belonging to the same asset class (i.e. equity portfolios of emerging and developed markets). In general, despite the salient unique empirical regularities evident in freight rate markets, it was found that the liquid bulk freight rates can be properly modeled as conventional financial assets for risk measurement purposes. However, different risk measurement models are selected for quantifying market risk in the liquid bulk freight markets: Specifically, the best performing models for quantifying daily liquid bulk freight rate risk exposures were the historical simulation, the GARCH, and the random walk VaR and ETL methods. In contrast to the liquid bulk freight rates, in the case of risk quantification for the dry bulk freight rates most conventional VaR and ETL models fell short in providing sufficiently accurate risk forecasts, demanding different treatment as far as market risk estimation is concerned. To this end, more sophisticated risk measurement models such as the switching ARCH, the filtered peaks over threshold and the combination of an exponential GARCH with student’s-t innovations parameterized by the Huisman et al. (2001) algorithm, outperformed other models in terms of statistical sufficiency and accuracy. Finally, this thesis makes an attempt to resolve the problem of estimating longer term risk with limited data by comparing the scaling laws of the square root and the tail index root of time with an empirically determined scaling law based on the theory of self similarity. The latter is proven to yield the most accurate medium term risk forecasts in contrast to scaling by the square or the tail index root of time which were found to understate the realized VaR systematically. As medium term risk exposures are relevant in many applications in finance (for e.g. when estimating market risk for imposing capital reserve requirements or in the case of estimating risk of insurance company portfolios) the proposed scaling law can be used effectively for extrapolating longer term risk forecasts with high frequency data by increasing the accuracy of risk forecasts.