Analyzing social systems, particularly financial markets, using a complex network approach has become one of the most popular fields within econophysics. A similar trend is currently appearing within the econometrics and finance communities, as well. In this study, we present a state-of-the-artmethod for analyzing the structure and risk within stockmarkets, treating them as complex networks using model-free, nonlinear dependency measures based on information theory. This study is the first network analysis of the stockmarket in Shanghai using a nonlinear network methodology. Further, it is often assumed that markets outside the United States and Western Europe are inherently riskier. We find that the Chinese stock market is not structurally risky, contradicting this popular opinion. We use partial mutual information to create filtered networks representing the Shanghai stock exchange, comparing them to networks based on Pearson’s correlation. Consequently, we discuss the structure and characteristics of both the presented methods and the Shanghai stock exchange. This paper provides an insight into the cutting edge methodology designed for analyzing complex financial networks, as well as analyzing the structure of the market in Shanghai and, as such, is of interest to both researchers and financial analysts.
29 August 2015 in Acta Universitatis Lodziensis. Folia Oeconomica
We treat financial markets as complex networks. It is commonplace to create a filtered graph (usually a Minimally Spanning Tree) based on an empirical correlation matrix. In our previous studies we have extended this standard methodology by exchanging Pearson’s correlation coefficient with information—theoretic measures of mutual information and mutual information rate, which allow for the inclusion of non-linear relationships. In this study we investigate the time evolution of financial networks, by applying a running window approach. Since information—theoretic measures are slow to converge, we base our analysis on the Hirschfeld-Gebelein-Rényi Maximum Correlation Coefficient, estimated by the Randomized Dependence Coefficient (RDC). It is defined in terms of canonical correlation analysis of random non-linear copula projections. On this basis we create Minimally Spanning Trees for each window moving along the studied time series, and analyse the time evolution of various network characteristics, and their market significance. We apply this procedure to a dataset describing logarithmic stock returns from Warsaw Stock Exchange for the years between 2006 and 2013, and comment on the findings, their applicability and significance.
In this study we investi gate how bankruptcy aff ects the market behaviour of prices of stocks on Warsaw’s Stock Exchange. As the behaviour of prices can be seen in a myriad of ways, we investi gate a parti cular aspect of this behaviour, namely the predictability of these price formati on processes. We approximate their predictability as the structural complexity of logarithmic returns. This method of analysing predictability of price formati on processes using informati on theory follows closely the mathemati cal defi niti on of predictability, and is equal to the degree to which redundancy is present in the ti me series describing stock returns. We use Shannon’s entropy rate (approximati ng Kolmogorov-Sinai entropy) to measure this redundancy, and esti mate it using the Lempel-Ziv algorithm, computi ng it with a running window approach over the enti re price history of 50 companies listed on the Warsaw market which have gone bankrupt in the last few years. This enables us not only to compare the diff erences between predictability of price formati on processes before and aft er their fi ling for bankruptcy, but also to compare the changes in predictability over ti me, as well as divided into diff erent categories of companies and bankruptcies. There exists a large body of research analysing the effi ciency of the whole market and the predictability of price changes en large, but only a few detailed studies analysing the infl uence of external sti muli on the effi ciency of price formati on processes. This study fi lls this gap in the knowledge of fi nancial markets, and their response to extreme external events.
In our previous studies we have investigated the structural complexity of time series describing stock returns on New York's and Warsaw's stock exchanges, by employing two estimators of Shannon's entropy rate based on Lempel-Ziv and Context Tree Weighting algorithms, which were originally used for data compression. Such structural complexity of the time series describing logarithmic stock returns can be used as a measure of the inherent (model-free) predictability of the underlying price formation processes, testing the Efficient-Market Hypothesis in practice. We have also correlated the estimated predictability with the profitability of standard trading algorithms, and found that these do not use the structure inherent in the stock returns to any significant degree. To find a way to use the structural complexity of the stock returns for the purpose of predictions we propose the Maximum Entropy Production Principle as applied to stock returns, and test it on the two mentioned markets, inquiring into whether it is possible to enhance prediction of stock returns based on the structural complexity of these and the mentioned principle.
Financial markets are complex adaptive systems, and are commonly studied as complex networks. Most of such studies fall short in two respects: they do not account for non-linearity of the studied relationships, and they create one network for the whole studied time series, providing an average picture of a very long, economically non-homogeneous, period. In this study we look at the currency markets by creating networks which can account for non-linearity in the underlying relationships, and are based on short time horizons with the use of running window approach. Since information--theoretic measures are slow to converge, we use Hirschfeld-Gebelein-Renyi Maximum Correlation Coefficient as a measure of the relationships between currencies. We use the Randomized Dependence Coefficient (RDC) as an estimator of the above. It measures the dependence between random samples as the largest canonical correlation between k randomly chosen non-linear projections of their copula transformations. On this basis we create full graphs, and further filter them into minimally spanning trees. We create such networks for each window moving along the studied time series, and analyse the time evolution of various network characteristics, particularly the degree distributions, and their economic significance. We apply this procedure to a dataset describing logarithmic changes in exchange rates in relation to silver for 27 world currencies for the years between 2002 and 2013.
In our previous study we have presented an approach to studying lead--lag effect in financial markets using information and network theories. Methodology presented there, as well as previous studies using Pearson's correlation for the same purpose, approached the concept of lead--lag effect in a naive way. In this paper we further investigate the lead--lag effect in financial markets, this time treating them as causal effects. To incorporate causality in a manner consistent with our previous study, that is including non-linear interdependencies, we base this study on a generalisation of Granger causality in the form of transfer entropy, or equivalently a special case of conditional (partial) mutual information. This way we are able to produce networks of stocks, where directed links represent causal relationships for a specific time lag. We apply this procedure to stocks belonging to the NYSE 100 index for various time lags, to investigate the short-term causality on this market, and to comment on the resulting Bonferroni networks.