Log-return Time Series Research Articles

Determinism and Non-linear Behaviour of Log-return and Conditional Volatility: Empirical Analysis for 26 Stock Markets

Determinism and non-linear behaviour in log-return and conditional volatility time series of the stock market index is examined for twenty-six countries. For this goal, the principal statistical techniques used in this study are a robust estimator of correlation dimension, a normalized non-linear prediction error, and pseudo-periodic surrogate data method. The proposed approach indicates, first, the stochastic behaviour of all log-return time series. Second, the inability of local linear, ARMA, or state- dependent noise models (such as ARCH, GARCH, and EGARCH) to describe its structure for the frontier, emerging, and developed markets. The same stochastic behaviour of conditional volatility time series, estimated by the stochastic volatility model with moving average innovations, is detected. This finding proves the efficiency of the stochastic volatility model compared with some analysed types of GARCH models for all studied markets. JEL Classification: C12, C52, D53, E44

Forward-looking portfolio selection with multivariate non-Gaussian models

In this study, we suggest a portfolio selection framework based on time series of stock log-returns, option-implied information, and multivariate non-Gaussian processes. We empirically assess a multivariate extension of the normal tempered stable (NTS) model and of the generalized hyperbolic (GH) one by implementing an estimation method that simultaneously calibrates the multivariate time series of log-returns and, for each margin, the univariate observed one-month implied volatility smile. To extract option-implied information, the connection between the historical measure P and the risk-neutral measure Q, needed to price options, is provided by the multivariate Esscher transform. The method is applied to fit a 50-dimensional series of stock returns, to evaluate widely known portfolio risk measures and to perform a forward-looking portfolio selection analysis. The proposed models are able to produce asymmetries, heavy tails, both linear and non-linear dependence and, to calibrate them, there is no need for liquid multivariate derivative quotes.

A Statistical Classification of Cryptocurrencies

The aim of this paper is to derive the main factors that separate cryptocurrencies from the classical assets, by using various classification techniques applied to the daily time series of log-returns. In this sense, a daily time series of asset returns (either cryptocurrencies or classical assets) can be characterized by a multidimensional vector with statistical components like variance, skewness, kurtosis, tail probability, quantiles, conditional tail expectation or GARCH parameters. By using dimension reduction techniques (Factor Analysis) and classification models (Binary Logistic Regression, Discriminant Analysis, Support Vector Machines, K-means clustering, Variance Components Split methods) for a representative sample of cryptocurrencies, stocks, exchange rates and commodities, we are able to classify cryptocurrencies as a new asset class with unique features in the tails of the log-returns distribution. The main result of our paper is the complete separation of the cryptocurrencies from the other type of assets, by using the Maximum Variance Components Split method. In addition, we observe a synchronicity in the evolution of of the cryptocurrencies, compared to the classical assets, mainly due to the tails behaviour of the log-return distribution.

Phenotypic Convergence of Cryptocurrencies

The aim of this paper is to prove the phenotypic convergence of cryptocurrencies, in the sense that individual cryptocurrencies respond to similar selection pressures by developing similar characteristics. In order to retrieve the cryptocurrencies phenotype, we treat cryptocurrencies as financial instruments (genus proximum) and find their specific difference (differentia specifica) by using the daily time series of log-returns. In this sense, a daily time series of asset returns (either cryptocurrencies or classical assets) can be characterized by a multidimensional vector with statistical components like volatility, skewness, kurtosis, tail probability, quantiles, conditional tail expectation or fractal dimension. By using dimension reduction techniques (Factor Analysis) and classification models (Binary Logistic Regression, Discriminant Analysis, Support Vector Machines, K-means clustering, Variance Components Split methods) for a representative sample of cryptocurrencies, stocks, exchange rates and commodities, we are able to classify cryptocurrencies as a new asset class with unique features in the tails of the log-returns distribution. The main result of our paper is the complete separation of the cryptocurrencies from the other type of assets, by using the Maximum Variance Components Split method. More, we observe a divergent evolution of the cryptocurrencies species, compared to the classical assets, mainly due to the tails behavior of the log-returns distribution. The codes used here are available via www.quantlet.de.

Focusing on regions of interest in forecast evaluation

Often, interest in forecast evaluation focuses on certain regions of the whole potential range of the outcome, and forecasts should mainly be ranked according to their performance within these regions. A prime example is risk management, which relies on forecasts of risk measures such as the value-at-risk or the expected shortfall, and hence requires appropriate loss distribution forecasts in the tails. Further examples include weather forecasts with a focus on extreme conditions, or forecasts of environmental variables such as ozone with a focus on concentration levels with adverse health effects. In this paper, we show how weighted scoring rules can be used to this end, and in particular that they allow to rank several potentially misspecified forecasts objectively with the region of interest in mind. This is demonstrated in various simulation scenarios. We introduce desirable properties of weighted scoring rules and present general construction principles based on conditional densities or distributions and on scoring rules for probability forecasts. In our empirical application to log-return time series, all forecasts seem to be slightly misspecified, as is often unavoidable in practice, and no method performs best overall. However, using weighted scoring functions the best method for predicting losses can be identified, which is hence the method of choice for the purpose of risk management.

The impact of estimation methods and data frequency on the results of long memory assessment

The main goal of this paper is to examine the effects of selected methods of estimation (the Geweke and Porter-Hudak, modified Geweke and Porter-Hudak, Whittle, R/S Rescaled Range Statistic, aggregated variance, aggregated absolute value, and Peng’s variance of residuals methods) and data frequency on properties of Hurst exponents for stock returns, volatility, and trading volumes of 43 companies and eight stock market indices. The calculations have been performed for a time series of log-returns, squared log-returns, and log-volume (based on hourly and daily data) by nine methods. Descriptive statistics and distribution laws of Hurst exponents depend on the method of estimation and, to some extent, on data frequency (daily and hourly). While by and large in log-returns no long memory has been detected, some estimation methods confirm the existence of long memory in squared log-returns. All of the applied estimation methods show long memory in log-volume data.

CALIBRATING THE SMILE WITH MULTIVARIATE TIME-CHANGED BROWNIAN MOTION AND THE ESSCHER TRANSFORM

In this study, we investigate two multivariate time-changed Brownian motion option pricing models in which the connection between the historical measure P and the risk-neutral measure Q is given by the Esscher transform. The models incorporate skewness, kurtosis and more complex dependence structures among stocks log-returns than the simple correlation matrix. The two models can be seen as a multivariate extension of the normal inverse Gaussian (NIG) model and the variance gamma (VG) model, respectively. We discuss two possible approaches to estimate historical asset returns and calibrate univariate option implied volatilities. While the first approach considers only time series of log-returns, the second approach makes use of both time series of log-returns and univariate observed volatility surfaces. To calibrate the models, there is no need of liquid multivariate derivative quotes.

Seasonal FIEGARCH processes

Here we develop the theory of seasonal FIEGARCH processes, denoted by SFIEGARCH, establishing conditions for the existence, the invertibility, the stationarity and the ergodicity of these processes. We analyze their asymptotic dependence structure by means of the autocovariance and autocorrelation functions. We also present some properties regarding their spectral representation. All properties are illustrated through graphical examples and an application of SFIEGARCH models to describe the volatility of the S&P500 US stock index log-return time series in the period from December 13, 2004 to October 10, 2009 is provided.

Scale dependence of the directional relationships between coupled time series

Using the cross-correlation of the wavelet transformation, we propose a general method of studying the scale dependence of the direction of coupling for coupled time series. The method is first demonstrated by applying it to coupled van der Pol forced oscillators and coupled nonlinear stochastic equations. We then apply the method to the analysis of the log-return time series of the stock values of the IBM and General Electric (GE) companies. Our analysis indicates that, on average, IBM stocks react earlier to possible common sector price movements than those of GE.

Non-Classical Measures of Investment Risk on the Market of Precious Non-Ferrous Metals Using the Methodology of Stable Distributions

The aim of this article is to present some non-classical risk measures which are commonly used in financial investments, including investments in assets from the market of precious non-ferrous metals. The time series of log-returns of gold, silver, platinum and palladium prices are considered. To properly asses the investment risk the measures based on Value-at-Risk methodology have been used (the VaR estimation approach based on values from the tail of the distribution). Additionally, the measure comparing expected profits to expected losses from the opposite tails distribution has been shown – the Rachev ratio. It was assumed that the log-returns of presented assets belong to the family of stable distributions. The results confirm the validity of the use of stable distributions to asses the risk on the precious non-ferrous metals market.

On a Poisson Subordinated Distribution for Precise Statistical Measurement of Leptokurtic Financial Data

A new Poisson subordinated distribution is proposed to capture major leptokurtic features in log-return time series of financial data. This distribution is intuitive, easy to calculate, and converge quickly. Analytic formula exists for all moments for the entire parameter space. It fits well to the historical daily log-return distributions of currencies, commodities, Treasury yields, VIX, and, most difficult of all, DJIA. It serves as a viable alternative to the Student’s t-distribution and the more sophisticated truncated stable distribution.

A Method to Quantify Risks of Financial Assets: An Empirical Analysis of Japanese Security Prices

This study investigates unconditional distributions of daily log-returns of Japanese security prices from a comprehensive point of view. The purpose of this article is to estimate a risk distribution of stocks in terms of Value-at-Risk (VaR) in order to select low risk securities from many securities. Daily log-return time series of 1,340 Japanese companies listed on the first section of Tokyo Stock Exchange are examined during the last one decade. I develop a method to estimate VaR by both the maximum likelihood estimation procedure under a q-Gaussian assumption and analytical form of its cumulative distribution function. It is confirmed that they are fitted to q-Gaussian distributions (Student t-distributions) with Kolmogorov-Smirnov test. It is found that the complementary cumulative distribution function of VaR has a power-law tail with its characteristic exponent depending on values of the VaR percentile.

Aplikace FIGARCH a EWMA modelů na burzovní indexy PX a BUX

Volatility of the financial time series belongs to the crucial estimated parameters in finance (e.g. in risk management, derivative pricing). It is well known, that volatility varies in time, so that new approaches of volatility modeling have appeared. In this paper two models of the conditional heteroskedasticity - fractionally integrated GARCH (FIGARCH) and EWMA are presented. These models are illustrated on the daily historical returns of stock index PX and index BUX. Standard tests of normality, autocorrelation and conditional heteroskedasticity are applied to these log-return time series and before estimating the models, which confirm a usability of the conditional heteroskedasticity models. Empirical results of the Rescale Range analysis (R/S) indicate a long memory in the volatility process of PX index and the first 40 autocorrelations of the square log-returns show their hyperbolic decay. The volatility models are estimated by quasi-maximum likelihood method with Student´s t-distribution and used to the calculation of the 1-day 95% and 99% Value at Risk values. Finally, the validity of the models is verified by Kupiec´s test, TUFF and Christoffersen´s test. These tests demonstrate, that the FIGARCH model is a suitable alternative to the EWMA model in the Value at Risk calculation.

Temporal evolution into a more efficient stock market

Using the price change and the log return of 10 stock market indices, we examine the temporal evolution of the time scale. The 10 stock markets had similar properties. Their log-return time series had patterns and long-range correlations until the mid-1990s. In the 2000s, however, the long-range correlations for most markets shortened, and the patterns weakened. These phenomena were due to advances in communication infrastructure such as the Internet and internet-based trading systems, which increased the speed of information dissemination. We examined the temporal evolution of the time scale in the markets by comparing the probability density function of log returns for the 2000s with that in the 1990s and by using the minimum entropy density method.

Superstatistical fluctuations in time series: Applications to share-price dynamics and turbulence

We report a general technique to study a given experimental time series with superstatistics. Crucial for the applicability of the superstatistics concept is the existence of a parameter beta that fluctuates on a large time scale as compared to the other time scales of the complex system under consideration. The proposed method extracts the main superstatistical parameters out of a given data set and examines the validity of the superstatistical model assumptions. We test the method thoroughly with surrogate data sets. Then the applicability of the superstatistical approach is illustrated using real experimental data. We study two examples, velocity time series measured in turbulent Taylor-Couette flows and time series of log returns of the closing prices of some stock market indices.