Herding and zero-intelligence agents in the order book dynamics of an artificial double auction market

Effects of herding on the order book dynamics of a double auction market is studied by an agent-based model. This is done by comparing results from a zero-intelligence model and a model in which herding effect is implemented by aggregation of agents who take market orders into opinion groups. The number of opinion groups in a simulation step is determined from previous volatilities of the market as different agents compare the price change over different time intervals. Besides confirming that when herding is included the tail of the distribution of volatility is enhanced, we found several new results. First, the autocorrelation time of volatility is much shorter than the memory of most of the agents because limit orders have strong influence on the location of best bid and best ask. Second, from the relation between bid-ask imbalance and price return we find that herding reduces the chance for a small imbalance to produce a large price change. Furthermore, herding tends to decrease spread. This is because herding decreases the chance that a market order changes the size of the spread. Finally, we find that the relation between spread and volatility in our models does not agree with empirical data, this indicates a difference between agents with no strategies and agents in real financial markets.     

Long memory features in the high frequency data of the Korean stock market

This paper examines the long memory property in the high frequency data of KOSPI 200 using the FIAPARCH model. The empirical results indicate that the FIAPARCH model can capture asymmetry and long memory in the volatility of intraday KOSPI 200 returns. Interestingly, the presence of long memory is invariant to the temporally aggregated intraday returns, implying that a long memory phenomenon is an inherent characteristic of the data generating process, not a result of structural breaks.     

Herd behaviour experimental testing in laboratory artificial stock market settings. Behavioural foundations of stylised facts of financial returns

Many scholars express concerns that herding behaviour causes excess volatility, destabilises financial markets, and increases the likelihood of systemic risk. We use a special form of the Strongly Typed Genetic Programming (STGP) technique to evolve a stock market divided into two groups—a small subset of artificial agents called ‘Best Agents’ and a main cohort of agents named ‘All Agents’. The ‘Best Agents’ perform best in term of the trailing return of a wealth moving average. We then investigate whether herding behaviour can arise when agents trade Dow Jones, General Electric, and IBM financial instruments in four different artificial stock markets. This paper uses real historical quotes of the three financial instruments to analyse the behavioural foundations of stylised facts such as leptokurtosis, non-IIDness, and volatility clustering. We found evidence of more herding in a group of stocks than in individual stocks, but the magnitude of herding does not contribute to the mispricing of assets in the long run. Our findings suggest that the price formation process caused by the collective behaviour of the entire market exhibit less herding and is more efficient than the segmented market populated by a small subset of agents. Hence, greater genetic diversity leads to greater consistency with fundamental values and market efficiency.     

Characteristics of the volatility in the Korea composite stock price index

We empirically analyze the time series of the Korea Composite Stock Price Index (KOSPI) from March of 1992 to February of 2007 using methods from the hydrodynamic turbulence. To this end, we focus on characteristics of the return and volatility, which are respectively the price change and a measure of the financial market fluctuation over a time interval. With these, we show that the non-Gaussian probability distribution of the return can be modeled by the convolution of the conditional probability distribution of the return given the volatility and the distribution of the volatility per se. From this model, we suggest that the non-Gaussian characteristic of the return results from the fluctuation of the volatility. That is, a large return is partly, if not entirely, due to the market fluctuation in a long time scale influencing the fluctuation in a short time scale via net information flow. We further show that the volatility has a multi-fractal property, which resembles the multifractality of the energy dissipation in the turbulence.     

The endogenous dynamics of financial markets: Interaction and information dissemination

We investigate the process that different interactions between investors will prompt information to propagate along a differentiated path and construct a financial market model. As information spreads, increasingly investors are attracted to participate in trading, then the “herding effect” is magnified gradually, which will induce the topology of market network to change and the price to fluctuate. Especially, under different initial conditions or parameters, the peak and fat-tail property is produced and the obtained statistic values coincide with empirical results: the power-law exponents between the peak value of return probability distribution and the time scales range from 0.579 to 0.747, and the exponents between the accumulation distribution and the return on the tail are close to 3. Besides, the extent of volatility clustering in our produced price series is close to that of S&P 500 and locates between NASDAQ and HSI. All the results obtained here indicate that the continuous variation of the “herding effect” resulting from information propagation among interacting investors may be the origin of stylized facts of price fluctuations.     

Self-organizing Ising model of financial markets

We study a dynamical Ising-like model of agents' opinions (buy or sell) with learning, in which the coupling coefficients are re-assessed continuously in time according to how past external news (time-varying magnetic field) have explained realized market returns. By combining herding, the impact of external news and private information, we find that the stylized facts of financial markets are reproduced only when agents misattribute the success of news to predict return to herding effects, thereby providing positive feedbacks leading to the model functioning close to the Ising critical point.     

Long memory properties in return and volatility: Evidence from the Korean stock market

In this paper, we study the dual long memory property of the Korean stock market. For this purpose, the ARFIMA-FIGARCH model is applied to two daily Korean stock price indices (KOSPI and KOSDAQ). Our empirical results indicate that long memory dynamics in the returns and volatility can be adequately estimated by the joint ARFIMA-FIGARCH model. We also found that the assumption of a skewed Student-t distribution is better for incorporating the tendency of asymmetric leptokurtosis in a return distribution.     

Market memory and fat tail consequences in option pricing on the expOU stochastic volatility model

The expOU stochastic volatility model is capable of reproducing fairly well most important statistical properties of financial markets daily data. Among them, the presence of multiple time scales in the volatility autocorrelation is perhaps the most relevant which makes appear fat tails in the return distributions. This paper wants to go further on with the expOU model we have studied in Ref. [J. Masoliver, J. Perelló, Quant. Finance 6 (2006) 423] by exploring an aspect of practical interest. Having as a benchmark the parameters estimated from the Dow Jones daily data, we want to compute the price for the European option. This is actually done by Monte Carlo, running a large number of simulations. Our main interest is to “see” the effects of a long-range market memory from our expOU model in its subsequent European call option. We pay attention to the effects of the existence of a broad range of time scales in the volatility. We find that a richer set of time scales brings the price of the option higher. This appears in clear contrast to the presence of memory in the price itself which makes the price of the option cheaper.     

Intraday volatility spillovers between spot and futures indices: Evidence from the Korean stock market

This study provides empirical evidence of the relationship between spot and futures markets in Korea. In particular, the study focuses on the volatility spillover relationship between spot and futures markets by using three high-frequency (10 min, 30 min, and 1 h time-scales) intraday data sets of KOSPI 200 spot and futures contracts. The results indicate a strong bi-directional causal relationship between futures and spot markets, suggesting that return volatility in the spot market can influence that in the futures market and vice versa. Thus, the results indicate that new information is reflected in futures and spot markets simultaneously. This bi-directional causal relationship provides market participants with important guidance on understanding the intraday information transmission between the two markets. Thus, on a given trading day, there may be sudden and sharp increases or decreases in return volatility in the Korean stock market as a result of positive feedback and synchronization of spot and futures markets.     

Dynamics of implied volatility surfaces from random matrix theory

We analyze the dynamics of the implied volatility surface of KOSPI 200 futures options from random matrix theory. To extract the informative data, we use random matrix criteria. Implied volatility data have a colossal eigenvalue, and the order of eigenvalues in a noisy regime is distinguishably smaller than a random matrix theory prediction. We discern the marketwide knowledge of the implied volatility surface movement such as the level, skew, and smile effect. These dynamics has the ergodic property and long range autocorrelation. We also study the relationship between the three implied volatility surface dynamics and the underlying asset dynamics, and confirm the existence of leverage effect even in the short time interval.     

Contemporaneous aggregation and long-memory property of returns and volatility in the Korean stock market

The principal objective of this study is to determine whether the long-memory property is real or a spurious result caused by contemporaneous aggregation. In order to assess the presence of long memory in returns and volatility, two different long-memory detection techniques (modified R/S analysis and the GPH test) were applied to the KOSPI 50 index and its 50 constituent individual stock prices. According to the empirical evidence gleaned from the two long-memory tests, we conclude that there exists significant evidence for the long-memory property in volatility in both the market index and in a majority of individual stocks. These findings indicate that the observed evidence of the long-memory feature in volatility of index series is not spurious, and that we can reject the hypothesis that spurious long-memory evidence in the volatility of index series is the consequence of contemporaneous aggregation. However, this conclusion should be considered cautiously, given that a considerable number of the individual stock volatilities in square returns strongly show a short-memory property, as the level of significance in statistical decisions is lowered to the 1% level.     

Leverage bubble

Leverage is strongly related to liquidity in a market and lack of liquidity is considered a cause and/or consequence of the recent financial crisis. A repurchase agreement is a financial instrument where a security is sold simultaneously with an agreement to buy it back at a later date. Repurchase agreement (repo) market size is a very important element in calculating the overall leverage in a financial market. Therefore, studying the behavior of repo market size can help to understand a process that can contribute to the birth of a financial crisis. We hypothesize that herding behavior among large investors led to massive over-leveraging through the use of repos, resulting in a bubble (built up over the previous years) and subsequent crash in this market in early 2008. We use the Johansen–Ledoit–Sornette (JLS) model of rational expectation bubbles and behavioral finance to study the dynamics of the repo market that led to the crash. The JLS model qualifies a bubble by the presence of characteristic patterns in the price dynamics, called log-periodic power law (LPPL) behavior. We show that there was significant LPPL behavior in the market before that crash and that the predicted range of times predicted by the model for the end of the bubble is consistent with the observations.     

What types of investors generate the two-phase phenomenon?

We examine the two-phase phenomenon described by Plerou, Gopikrishnan, and Stanley (2003)  [1] in the KOSPI 200 options market, one of the most liquid options markets in the world. By analysing a unique intraday dataset that contains information about investor type for each trade and quote, we find that the two-phase phenomenon is generated primarily by domestic individual investors, who are generally considered to be uninformed and noisy traders. In contrast, our empirical results indicate that trades by foreign institutions, who are generally considered informed and sophisticated investors, do not exhibit two-phase behaviour.     

A self-adapting herding model: The agent judge-abilities influence the dynamic behaviors

We propose a self-adapting herding model, in which the financial markets consist of agent clusters with different sizes and market desires. The ratio of successful exchange and merger depends on the volatility of the market and the market desires of the agent clusters. The desires are assigned in term of the wealth of the agent clusters when they merge. After an exchange, the beneficial cluster’s desire keeps on the same, the losing one’s desire is altered which is correlative with the agent judge-ability. A parameter R is given to all agents to denote the judge-ability. The numerical calculation shows that the dynamic behaviors of the market are influenced distinctly by R, which includes the exponential magnitudes of the probability distribution of sizes of the agent clusters and the volatility autocorrelation of the returns, the intensity and frequency of the volatility.     

Identifying the structure of group correlation in the Korean financial market

We investigate the structure of the cross-correlation in the Korean stock market. We analyze daily cross-correlations between price fluctuations of 586 different Korean stock entities for the 6-year time period from 2003 to 2008. The main purpose is to investigate the structure of group correlation and its stability by undressing the market-wide effect using the Markowitz multi-factor model and the network-based approach. We find the explicit list of significant firms in the few largest eigenvectors from the undressed correlation matrix. We also observe that each contributor is involved in the same business sectors. The structure of group correlation can not remain constant during each 1-year time period with different starting points, whereas only two largest eigenvectors are stable for 6 years 8-9 eigenvectors remain stable for half-year. The structure of group correlation in the Korean financial market is disturbed during a sufficiently short time period even though the group correlation exists as an ensemble for the 6-year time period in the evolution of the system. We verify the structure of group correlation by applying a network-based approach. In addition, we examine relations between market capitalization and businesses. The Korean stock market shows a different behavior compared to mature markets, implying that the KOSPI is a target for short-positioned investors.     

Option pricing and perfect hedging on correlated stocks

We develop a theory for option pricing with perfect hedging in an inefficient market model where the underlying price variations are autocorrelated over a time τ0. This is accomplished by assuming that the underlying noise in the system is derived by an Ornstein-Uhlenbeck, rather than from a Wiener process. With a modified portfolio consisting in calls, secondary calls and bonds we achieve a riskless strategy which results in a closed and exact expression for the European call price which is always lower than Black-Scholes price. We obtain the same price and a modified delta hedging if we start from an effective one-dimensional market model. We compare these strategies and study the sensitivity of the call price to several parameters where the correlation effects are also observed.     

Measuring capital market efficiency: Global and local correlations structure

We introduce a new measure for capital market efficiency. The measure takes into consideration the correlation structure of the returns (long-term and short-term memory) and local herding behavior (fractal dimension). The efficiency measure is taken as a distance from an ideal efficient market situation. The proposed methodology is applied to a portfolio of 41 stock indices. We find that the Japanese NIKKEI is the most efficient market. From a geographical point of view, the more efficient markets are dominated by the European stock indices and the less efficient markets cover mainly Latin America, Asia and Oceania. The inefficiency is mainly driven by a local herding, i.e. a low fractal dimension.     

A model for the contagion and herding

This work concerns the modeling of contagion and herding effects which can cause significant movements of prices and volatilities. The idea is to adapt some concepts borrowed from the Biological Sciences and that have emerged as useful analogies to model a variety of phenomena in a large variety of fields such as Engineering and Economics. In this work, the allegory of interacting particles is used to describe the contagion and emergence of herding behavior of financial agents leading to the formation of clusters. The main idea is to adapt the schemes originally employed in particle swarm optimization algorithms, together with the concepts of leaders and followers. As an illustration of the applicability of the proposed model, a case study is presented using data from the World Bank.     

Option Pricing in Subdiffusive Bachelier Model

The earliest model of stock prices based on Brownian diffusion is the Bachelier model. In this paper we propose an extension of the Bachelier model, which reflects the subdiffusive nature of the underlying asset dynamics. The subdiffusive property is manifested by the random (infinitely divisible) periods of time, during which the asset price does not change. We introduce a subdiffusive arithmetic Brownian motion as a model of stock prices with such characteristics. The structure of this process agrees with two-stage scenario underlying the anomalous diffusion mechanism, in which trapping random events are superimposed on the Langevin dynamics. We find the corresponding fractional Fokker-Planck equation governing the probability density function of the introduced process. We construct the corresponding martingale measure and show that the model is incomplete. We derive the formulas for European put and call option prices. We describe explicit algorithms and present some Monte-Carlo simulations for the particular cases of α-stable and tempered α-stable distributions of waiting times.     

From market games to real-world markets

This paper uses the development of multi-agent market models to present a unified approach to the joint questions of how financial market movements may be simulated, predicted, and hedged against. We first present the results of agent-based market simulations in which traders equipped with simple buy/sell strategies and limited information compete in speculatory trading. We examine the effect of different market clearing mechanisms and show that implementation of a simple Walrasian auction leads to unstable market dynamics. We then show that a more realistic out-of-equilibrium clearing process leads to dynamics that closely resemble real financial movements, with fat-tailed price increments, clustered volatility and high volume autocorrelation. We then show that replacing the `synthetic' price history used by these simulations with data taken from real financial time-series leads to the remarkable result that the agents can collectively learn to identify moments in the market where profit is attainable. Hence on real financial data, the system as a whole can perform better than random. We then employ the formalism of Bouchaud in conjunction with agent based models to show that in general risk cannot be eliminated from trading with these models. We also show that, in the presence of transaction costs, the risk of option writing is greatly increased. This risk, and the costs, can however be reduced through the use of a delta-hedging strategy with modified, time-dependent volatility structure. Received 30 August 2000