A Markov model of financial returns

We address the general problem of how to quantify the kinematics of time series with stationary first moments but having non stationary multifractal long-range correlated second moments. We show that a Markov process is sufficient to model important aspects of the multifractality observed in financial time series and propose a kinematic model of price fluctuations. We test the proposed model by analyzing index closing prices of the New York Stock Exchange and the DEM/USD tick-by-tick exchange rates obtained from Reuters EFX. We show that the model captures the characteristic features observed in actual financial time series, including volatility clustering, time scaling and fat tails in the probability density functions, power-law behavior of volatility correlations and, most importantly, the observed nonuniversal multifractal singularity spectrum. Motivated by our finding of strong agreement between the model and the data, we argue that at least two independent stochastic Gaussian variables are required to adequately model price fluctuations.     

Multi-scaling in the Cont–Bouchaud microscopic stock market model

The Cont–Bouchaud percolation model is one of the simplest microsimulation models yet able to account for the main stylized fact of financial markets, e.g. fat tails of the histogram of log-returns. In the present paper we show that for a certain range of the parameters it is possible to generate price time-series that cannot be described in terms of a unique scaling exponent.     

Stochastic volatility of financial markets as the fluctuating rate of trading: An empirical study

We present an empirical study of the subordination hypothesis for a stochastic time series of a stock price. The fluctuating rate of trading is identified with the stochastic variance of the stock price, as in the continuous-time random walk (CTRW) framework. The probability distribution of the stock price changes (log-returns) for a given number of trades N is found to be approximately Gaussian. The probability distribution of N for a given time interval Δt is non-Poissonian and has an exponential tail for large N and a sharp cutoff for small N. Combining these two distributions produces a non-trivial distribution of log-returns for a given time interval Δt, which has exponential tails and a Gaussian central part, in agreement with empirical observations.     

Modelling price dynamics: A hybrid truncated Lévy Flight–GARCH approach

ARCH and GARCH stochastic processes are widely used in finance and are generally accepted as good approximations when modelling the price dynamics with Gaussian conditional probability. It can be seen that certain aspects of the empirical data for asset price changes seems to more closely fit a Truncated Lévy Flight or GARCH model, but each with individual shortfalls. In this paper therefore, we combine the GARCH process with a conditional truncated Lévy distribution in order to build a hybrid model that most notably describes the price change and associated volatility probability density distributions and scaling behaviour over different time horizons.     

“Exact” and Approximate Methods for Bayesian Inference: Stochastic Volatility Case Study

We conduct a case study in which we empirically illustrate the performance of different classes of Bayesian inference methods to estimate stochastic volatility models. In particular, we consider how different particle filtering methods affect the variance of the estimated likelihood. We review and compare particle Markov Chain Monte Carlo (MCMC), RMHMC, fixed-form variational Bayes, and integrated nested Laplace approximation to estimate the posterior distribution of the parameters. Additionally, we conduct the review from the point of view of whether these methods are (1) easily adaptable to different model specifications; (2) adaptable to higher dimensions of the model in a straightforward way; (3) feasible in the multivariate case. We show that when using the stochastic volatility model for methods comparison, various data-generating processes have to be considered to make a fair assessment of the methods. Finally, we present a challenging specification of the multivariate stochastic volatility model, which is rarely used to illustrate the methods but constitutes an important practical application.     

Perturbation expansion for option pricing with stochastic volatility

We fit the volatility fluctuations of the S&P 500 index well by a Chi distribution, and the distribution of log-returns by a corresponding superposition of Gaussian distributions. The Fourier transform of this is, remarkably, of the Tsallis type. An option pricing formula is derived from the same superposition of Black-Scholes expressions. An explicit analytic formula is deduced from a perturbation expansion around a Black-Scholes formula with the mean volatility. The expansion has two parts. The first takes into account the non-Gaussian character of the stock-fluctuations and is organized by powers of the excess kurtosis, the second is contract based, and is organized by the moments of moneyness of the option. With this expansion we show that for the Dow Jones Euro Stoxx 50 option data, a -hedging strategy is close to being optimal.     

An interacting-agent model of financial markets from the viewpoint of nonextensive statistical mechanics

In this paper we present an interacting-agent model of stock markets. We describe a stock market through an Ising-like model in order to formulate the tendency of traders to be influenced by the other traders’ investment attitudes [Kaizoji, Physica A 287 (2000) 493], and formulate the traders’ decision-making regarding investment as the maximum entropy principle for nonextensive entropy [C. Tsallis, J. Stat. Phys. 52 (1988) 479]. We demonstrate that the equilibrium probability distribution function of the traders’ investment attitude is the q-exponential distribution. We also show that the power-law distribution of the volatility of price fluctuations, which is often demonstrated in empirical studies can be explained naturally by our model which originates in the collective crowd behavior of many interacting-agents.     

Anomalous scaling of stock price dynamics within ARCH-models

We show that autoregressive-conditional-heteroskedasticity (ARCH) models can encompass the observed anomalous scaling properties of stock price dynamics remarkably well. We find that with a suitable choice of parameters, simple ARCH models can reproduce the non-standard scaling behavior of the central part of the probability distribution functions of stock prices at different time horizons, as empirically found for the Standard & Poors 500 (S&P 500) index data, but fail to reproduce the shape of the S&P 500 distribution, in particular at the smallest time horizon (1 min). A linear version of ARCH processes, denoted here as LARCH models, still preserving the anomalies observed, permits to fit the 1 min S&P 500 distribution more accurately. Received 12 October 2000 and Received in final form 5 February 2001     

Volatility clustering and scaling for financial time series due to attractor bubbling

A microscopic model of financial markets is considered, consisting of many interacting agents (spins) with global coupling and discrete-time heat bath dynamics, similar to random Ising systems. The interactions between agents change randomly in time. In the thermodynamic limit, the obtained time series of price returns show chaotic bursts resulting from the emergence of attractor bubbling or on-off intermittency, resembling the empirical financial time series with volatility clustering. For a proper choice of the model parameters, the probability distributions of returns exhibit power-law tails with scaling exponents close to the empirical ones.     

Waiting time distributions of the volatility in the Italian MIB30 index: Clustering or Poisson functions?

We investigate the time behaviour of the Italian MIB30 stock index collected every minute during two months in the period from May 17, 2006, up to July 24, 2006. We find short-range correlations in the price returns and, on the contrary, a long persistent time lag and slow decay in the autocorrelation functions of volatility. Besides, we find that the probability density functions (PDFs) of returns show fat tails, which are well fit by the log-normal model of Castaing [B. Castaing, Y. Gagne, E.J. Hopfinger, Physica D 46 (1990) 177], and a convergence toward a normal distribution for large time scales; we also find that the PDFs of volatility, for short time horizons, fit better with a log-normal distribution than with a Gaussian. Most of these features characterize the indexes and stocks of the largest American, European and Asian markets.We also investigate the distribution of stochastic separation between isolated strong events in the volatility signal. This is interesting because this gives us a deeper understanding about the price formation process. By using a test for the occurrence of local Poisson hypothesis, we show that the process we examined strongly departs from a Poisson statistics, the origin of this failure stemming from the presence of temporal clustering and of a certain amount of memory.     

A study about the existence of the leverage effect in stochastic volatility models

The empirical relationship between the return of an asset and the volatility of the asset has been well documented in the financial literature. Named the leverage effect or sometimes risk-premium effect, it is observed in real data that, when the return of the asset decreases, the volatility increases and vice versa.Consequently, it is important to demonstrate that any formulated model for the asset price is capable of generating this effect observed in practice. Furthermore, we need to understand the conditions on the parameters present in the model that guarantee the apparition of the leverage effect.In this paper we analyze two general specifications of stochastic volatility models and their capability of generating the perceived leverage effect. We derive conditions for the apparition of leverage effect in both of these stochastic volatility models. We exemplify using stochastic volatility models used in practice and we explicitly state the conditions for the existence of the leverage effect in these examples.     

Clustering of volatility as a multiscale phenomenon

The dynamics of prices in financial markets has been studied intensively both experimentally (data analysis) and theoretically (models). Nevertheless, a complete stochastic characterization of volatility is still lacking. What is well known is that absolute returns have memory on a long time range, this phenomenon is known as clustering of volatility. In this paper we show that volatility correlations are power-laws with a non-unique scaling exponent. This kind of multiscale phenomenology has some analogies with fully developed turbulence and disordered systems and it is now pointed out for financial series. Starting from historical returns series, we have also derived the volatility distribution, and the results are in agreement with a log-normal shape. In our study, we consider the New York Stock Exchange (NYSE), daily composite index closes (January 1966 to June 1998) and the US Dollar/Deutsche Mark (USD-DM) noon buying rates certified by the Federal Reserve Bank of New York (October 1989 to September 1998). Received 1 February 2000     

The volatility in a multi-share financial market model

Single index financial market models cannot account for the empirically observed complex interactions between shares in a market. We describe a multi-share financial market model and compare characteristics of the volatility, that is the variance of the price fluctuations, with empirical characteristics. In particular we find its probability distribution is similar to a log normal distribution but with a long power-law tail for the large fluctuations, and that the time development shows superdiffusion. Both these results are in good quantitative agreement with observations.     

Empirical scaling laws and the aggregation of non-stationary data

Widely cited evidence for scaling (self-similarity) of the returns of stocks and other securities is inconsistent with virtually all currently-used models for price movements. In particular, state-of-the-art models provide for ubiquitous, irregular, and oftentimes high-frequency fluctuations in volatility (“stochastic volatility”), both intraday and across the days, weeks, and years over which data is aggregated in demonstrations of self-similarity of returns. Stochastic volatility renders these models, which are based on variants and generalizations of random walks, incompatible with self-similarity. We show here that empirical evidence for self-similarity does not actually contradict the analytic lack of self-similarity in these models. The resolution of the mismatch between models and data can be traced to a statistical consequence of aggregating large amounts of non-stationary data.     

Modelling fluctuations of financial time series: from cascade process to stochastic volatility model

In this paper, we provide a simple, “generic” interpretation of multifractal scaling laws and multiplicative cascade process paradigms in terms of volatility correlations. We show that in this context 1/f power spectra, as recently observed in reference [23], naturally emerge. We then propose a simple solvable “stochastic volatility” model for return fluctuations. This model is able to reproduce most of recent empirical findings concerning financial time series: no correlation between price variations, long-range volatility correlations and multifractal statistics. Moreover, its extension to a multivariate context, in order to model portfolio behavior, is very natural. Comparisons to real data and other models proposed elsewhere are provided. Received 22 May 2000     

Fractional Brownian motion, fractional Gaussian noise, and Tsallis permutation entropy

In this work, we analyze two important stochastic processes, the fractional Brownian motion and fractional Gaussian noise, within the framework of the Tsallis permutation entropy. This entropic measure, evaluated after using the Bandt & Pompe method to extract the associated probability distribution, is shown to be a powerful tool to characterize fractal stochastic processes. It allows for a better discrimination of the processes than the Shannon counterpart for appropriate ranges of values of the entropic index. Moreover, we find the optimum value of this entropic index for the stochastic processes under study.     

Role of scaling in the statistical modelling of finance

Modelling the evolution of a financial index as a stochastic process is a problem awaiting a full, satisfactory solution since it was first formulated by Bachelier in 1900. Here it is shown that the scaling with time of the return probability density function sampled from the historical series suggests a successful model. The resulting stochastic process is a heteroskedastic, non-Markovian martingale, which can be used to simulate index evolution on the basis of an autoregressive strategy. Results are fully consistent with volatility clustering and with the multiscaling properties of the return distribution. The idea of basing the process construction on scaling, and the construction itself, are closely inspired by the probabilistic renormalization group approach of statistical mechanics and by a recent formulation of the central limit theorem for sums of strongly correlated random variables.