Liquidity risk in capital markets

This thesis focuses on liquidity risk in capital markets. The main aim is to help practitioners to better understand and manage liquidity risk by analyzing the following three topics: modeling correlations in a Liquidity Adjusted VaR (L − VaR) (Chapter Three), impact of regulatory interventions on stock liquidity (Chapter Four) and liquidity commonality and option prices (Chapter Five).
The first topic focuses on an appropriate way to measure expected stock losses by considering liquidity risk (see Chapter Three). The need for a new measure, which also includes stock liquidity, is based on the concern of investors only being able to sell stock at a huge discount or not at all. In reaction, various papers with new methodologies have been published, including the liquidity adjusted Value at Risk (L − VaR) models proposed by Bangia et al. (1998) and Ernst et al. (2012). Based on their approach, we analyze different ways to extend these models and to optimize performance. This is done using advanced conditional volatility models like AR − GARCH and AR − GJR models and by considering correlations between spread and return data. The new model is called correlation and liquidity adjusted VaR (CL – VaR) and shows (based on a five-year observation period) better performance compared to the models by Bangia et al. (1998) and Ernst et al. (2012). The models are calculated and back-tested using unique data called Xetra Liquidity Measure (XLM) provided by Deutsche Börse.
The collapse of Lehman Brothers in 2008 marked the beginning of a financial crisis affecting the entire world of finance. This period is characterized by increasing fear of further defaults by corporations (including banks) or even by countries. In reaction, investors began shifting their assets to more stable and secure investments and this resulted in stock market crashes. Various interventions were made by government institutions to restore stability.
The target of the second topic is to analyze the impact of these interventions on liquidity (measured by volume-weighted bid-ask spreads) and market reaction (measured by returns) at the announcement date (see Chapter Four). In the event, we study abnormal changes of stocks listed on the Dax. The interventions which we consider are published by the Federal Reserve Bank (FED) in the form of a crisis time-line. Here they are further combined to the following categories: bank liability guarantees, liquidity and rescue interventions, unconventional monetary policy and other market intervention. The results show that, for example, the market reacts positively to liquidity and rescue interventions, whereas bank liability guarantees reduce liquidity. In addition, we show that international events have a significant impact on the domestic market in a "spillover effect". By analyzing the spreads of different traded volumes, an asymmetric increase can be detected at the announcement date.
The last topic focuses on the link between equity and option markets (see Chapter Five). There we analyze, on one hand, the link between stock market liquidity and option prices and, on the other hand, the impact of liquidity commonality in equity and option markets. We can show that systematic liquidity (rather than idiosyncratic liquidity) gives a better explanation of changes in “at-the-money” implied volatility. This effect was especially strong during the financial crisis in 2008. Another result is that liquidity risk of higher traded stock volumes is not properly reflected in the option price. This can result in higher hedging costs, as mentioned by Certin et al. (2006). To shed more light into liquidity commonality within the stock market we calculate the LiqCom measure as mentioned by Chordia et al. (2000). The results show a continuously changing liquidity commonality which decreases with increasing traded volume. This is because the market maker focuses for bigger stock positions more on the idiosyncratic liquidity risks while for smaller stock positions the systematic liquidity risk is more important. We confirmed our findings with a robustness check.