Data vs. information: Using clustering techniques to enhance stock returns forecasting

This paper explores the use of clustering models of stocks to improve both (a) the prediction of stock prices and (b) the returns of trading algorithms.We cluster stocks using k-means and several alternative distance metrics, using as features quarterly financial ratios, prices and daily returns. Then, for each cluster, we train ARIMA and LSTM forecasting models to predict the daily price of each stock in the cluster. Finally, we employ the clustering-empowered forecasting models to analyze the returns of different trading algorithms.We obtain three key results: (i) LSTM models outperform ARIMA and benchmark models, obtaining positive investment returns in several scenarios; (ii) forecasting is improved by using the additional information provided by the clustering methods, therefore selecting relevant data is an important preprocessing task in the forecasting process; (iii) using information from the whole sample of stocks deteriorates the forecasting ability of LSTM models.These results have been validated using data of 240 companies of the Russell 3000 index spanning 2017 to 2022, training and testing with different subperiods.     

An analytic approximation formula for pricing zero-coupon bonds

This paper presents an analytic approximation formula for pricing zero-coupon bonds, when the dynamics of the short-term interest rate are driven by a one-factor mean-reverting process in which changes in the volatility of the interest rate are a function of the level of the interest rate.     

Pricing of credit default index swap tranches with one-factor heavy-tailed copula models

In this paper, we provide two one-factor heavy-tailed copula models for pricing a collateralized debt obligation and credit default index swap tranches: (1) a one-factor double t distribution with fractional degrees of freedom copula model and (2) a one-factor double mixture distribution of t and Gaussian distribution copula model. A time-varying tail-fatness parameter is introduced in each model, allowing one to change the tail-fatness of the copula function continuously. Fitting our model to comprehensive market data, we find that a model with fixed tail-fatness cannot fit market data well over time. The two models that we propose are capable of fitting market data well over time when using a proper time-varying tail-fatness parameter. Moreover, we find that the time-varying tail-fatness parameters change dramatically over a one-year period.     

A One-Factor Conditionally Linear Commodity Pricing Model under Partial Information

A one-factor asset pricing model with an Ornstein–Uhlenbeck process as its state variable is studied under partial information: the mean-reverting level and the mean-reverting speed parameters are modeled as hidden/unobservable stochastic variables. No-arbitrage pricing formulas for derivative securities written on a liquid asset and exponential utility indifference pricing formulas for derivative securities written on an illiquid asset are presented. Moreover, a conditionally linear filtering result is introduced to compute the pricing/hedging formulas and the Bayesian estimators of the hidden variables.     

An analysis of simultaneous company defaults using a shot noise process

During the subprime mortgage crisis, it became apparent that practical models, such as the one-factor Gaussian copula, had underestimated company default correlations. Complex models that attempt to incorporate default dependency are difficult to implement in practice. In this study, we develop a model for a company asset process, based on which we calculate simultaneous default probabilities using an option-theoretic approach. In our model, a shot noise process serves as the key element for controlling correlations among companies’ assets. The risk factor driving the shot noise process is common to all companies in an industry but the shot noise parameters are assumed company-specific; therefore, every company responds differently to this common risk factor. Our model gives earlier warning of financial distress and predicts higher simultaneous default probabilities than commonly used geometric Brownian motion asset model. It is also computationally simple and can be extended to analyze any finite number of companies.     

Good-Deal Option Price Bounds for a Non-Traded Event with Stochastic Return: A Note

Cochrane and Sa'a-Requejo (2000, Journal of Political Economy) proposed the good-deal price bounds for the European call option on an event that is not a traded asset, but is correlated with a traded asset that can be used as an approximate hedge. One remarkable feature of their model is that the return on an event process explicitly appears in the option price bounds formula, which offered a contrast with the standard option pricing model. We show that the good-deal option price bounds on a non-traded event are obtained as a closed-form formula, when the return on an event is governed by a mean reverting process.     

On market price of risk in Asian capital markets around the Asian flu

This study investigates a contemporaneous relationship between realized market risk premia, and conditional variance and covariance in nine Asian markets and the US. The time period for this study is before, during, and after the Asian financial crisis. A contemporaneous state-dependent capital asset pricing model (CAPM) that allows for negative and positive market prices of variance and covariance risk is investigated. In the light of significant upstate and downstate reward to local and world variance risk for all markets and all periods, we conclude that a market return-generating process is a piecewise function of local and world variance over time. Furthermore, a cross-sectional analysis of upstate and downstate market prices of variance and covariance risk indicates that reward to risk is a mix of reward to local and world variance, depending on the ever-changing correlation with the world market. Our findings are consistent with the one-factor conditional international CAPM.     

Analysts versus time-series forecasts of quarterly earnings: A maintained hypothesis revisited

We re-examine the maintained hypothesis of analysts' quarterly earnings per share (EPS) superiority versus ARIMA time-series forecasts. While our empirical results are consistent with overall analysts' dominance, they suggest a more contextual interpretation of this important relationship. Specifically, we find that for a relatively large number of cases (approximately 40%) ARIMA time-series forecasts of quarterly EPS are equal to or more accurate than consensus analysts' forecasts. Moreover, the percentage of time-series superiority increases: (1) for longer forecast horizons, (2) as firm size decreases, and (3) for high-technology firms. Due to the data demands that ARIMA forecasting requires we also examine using a seasonal random walk (SRW) model that requires only one year of data to create quarterly forecasts. Although the ARIMA time-series model results in a significant reduction in sample size it dominates the SRW model. Our findings support the analyst dominance over time series models but suggest that ARIMA time-series models may provide useful input to researchers seeking quarterly EPS expectation models for certain types of firms.     

The Dividend Pricing Model: New Evidence from the Korean Housing Market

It is generally conceded that dividend pricing models are poor predictors of asset prices. This finding is sometimes attributed to excess volatility or to a dividend process manipulated by firm managers. In this paper, we present rather powerful panel tests of the dividend pricing relation using a unique data set in which dividends are set by market forces independent of managers' preferences. We rely on observations on the market for condominium dwellings in Korea—perhaps the only market in which information on dividends and prices is publicly and continuously available to consumers and investors. We extend the “dividend-price ratio model” to panels of housing returns and rents differentiated by type and location. We find broad support for the dividend pricing model during periods both before and after the Asian Financial Crisis of 1997–1998, suggesting that the market for housing assets in Korea has been remarkably efficient. Previous versions of this paper were presented at the Hong Kong-Singapore International Real Estate Research Symposium, August 2004, Hong Kong and the meeting of the Hong Kong Economic Association, January 2005. We are grateful for the comments of Ashok Bardhan, Yuming Fu, Chinmoy Ghosh, Lok Sang Ho, Charles Ka Yui Leung, Sau Kim Lum and Seow Eng Ong. Son's research was supported by the Konkuk University and Hwang's research was supported by the National University of Singapore.     

Stochastic Conditional Intensity Processes

In this article, we introduce the so-called stochastic conditionalintensity (SCI) model by extending Russell’s (1999) autoregressiveconditional intensity (ACI) model by a latent common dynamicfactor that jointly drives the individual intensity components.We show by simulations that the proposed model allows for awide range of (cross-)autocorrelation structures in multivariatepoint processes. The model is estimated by simulated maximumlikelihood (SML) using the efficient importance sampling (EIS)technique. By modeling price intensities based on NYSE trading,we provide significant evidence for a joint latent factor andshow that its inclusion allows for an improved and more parsimoniousspecification of the multivariate intensity process.     

The behavior of risky asset prices in a symmetric least squares learning environment

This paper examines the dynamics of asset prices in a heterogeneous market. Traders are made up of learners who possess limited information and use limited models for predicting the future. The market also includes noise traders in the sense of Black, along with liquidity traders. Learners revise their prediction equations using least squares learning as defined by Marcet and Sargent. We derive the equilibrium price process and show how convergence is obtained. The price process is shown to have a number of interesting properties that are consistent with propositions outlined by Black. Numerical calculations for several examples illuminate how learning takes place in the model.     

A model of discontinuous interest rate behavior,yield curves,and volatility

This paper develops an equilibrium model in which interest rates follow a discontinuous (generalized) gamma process. The gamma process has finite variation, takes an infinite number of “small” jumps in every interval, and includes the Wiener process as a limiting case. The gamma interest rate model produces yield curves that closely resemble those of diffusion models. But in contrast to diffusion models, the curvature of the yield curve does not directly depend on the true volatility of the interest rate process, but instead depends on a different risk-neutral volatility. The gamma model appears to fit the distribution of interest rates changes and the jump characteristics of interest rate paths. Empirical tests reject a diffusion model of interest rates in favor of the more general gamma model because daily interest rate innovations are highly leptokurtic. The author appreciates comments from George Constantinides, Jon Ingersoll, Herbert Johnson, Ray Rishel, and an anonymous referee, computational assistance from Kerry Back and Saikat Nandi, and support from Atlantic Asset Management. Any errors are the responsibility of the author.     

The Construction and Application of a New Exchange Rate Forecast Model Combining ARIMA with a Chaotic BP Algorithm

This article introduces a new model that combines an ARIMA with a chaotic BP (Backforward Propagation Neural Network) algorithm for exchange rate forecasting purposes, which is based on sample data collected from January 4, 2010, to October 20, 2011. The forecast of the exchange rate trend is then provided for the subsequent twenty-five days. Other models are also constructed, such as the ARIMA, BP, ARIMA, and BP algorithms, in order to evaluate the forecast accuracy. Based on our results, the combination of an ARIMA and a chaotic BP algorithm outperforms all other models in terms of the statistical accuracy of short-term forecasts.     

On the range of options prices

In this paper we consider the valuation of an option with time to expiration and pay-off function which is a convex function (as is a European call option), and constant interest rate , in the case where the underlying model for stock prices is a purely discontinuous process (hence typically the model is incomplete). The main result is that, for “most” such models, the range of the values of the option, using all possible equivalent martingale measures for the valuation, is the interval , this interval being the biggest interval in which the values must lie, whatever model is used.     

小波分析结合ARIMA组合模型的2014~2017年石油贸易量预测研究

选取2002~2013年我国石油进出口贸易量的数据进行建模分析。首先运用小波分析理论将贸易量数据进行分解,识别出数据的主要特征和细节特征,针对不同特征进行识别和平稳性检测和参数估计,建立相应的ARIMA模型,并进行预测加权合成。仿真结果表明,小波分析结合ARIMA组合模型的预测精度远远大于为改进的ARIMA预测模型,从而为科学合理的决策提供更为精确的预测模型。     

On a general class of one-factor models for the term structure of interest rates

We propose a general one-factor model for the term structure of interest rates which based upon a model for the short rate. The dynamics of the short rate is described by an appropriate function of a time-changed Wiener process. The model allows for perfect fitting of given term structure of interest rates and volatilities, as well as for mean reversion. Moreover, every type of distribution of the short rate can be achieved, in particular, the distribution can be concentrated on an interval. The model includes several popular models such as the generalized Vasicek (or Hull-White) model, the Black-Derman-Toy, Black-Karasinski model, and others. There is a unified numerical approach to the general model based on a simple lattice approximation which, in particular, can be chosen as a binomial or -nomial lattice with branching probabilities .     

Valuation of Portfolio Credit Derivatives with Default Intensities Using the Vasicek Model

We present a methodology for valuing portfolio credit derivatives under a reduced form model for which the default intensity processes of risk assets follow the one-factor Vasicek model. A closed-form solution of joint survival time distribution is obtained. The solution is applied to value credit derivatives of a credit default swap index and collateralized debt obligation. The limitation of methods using the Vasicek model is discussed. We propose that the method is valid and efficient for a portfolio with small-scale correlated risk assets, for which the acceptable size is much greater than for the traditional method. Numerical examples and parameter analysis are also presented.     

What is the Natural Scale for a Lévy Process in Modelling Term Structure of Interest Rates?

This paper gives examples of explicit arbitrage-free term structure models with Lévy jumps via the state price density approach. By generalizing quadratic Gaussian models, it is found that the probability density function of a Lévy process is a “natural” scale for the process to be the state variable of a market.      

Asset price prediction using seasonal decomposition

We propose a new forecasting procedure for asset prices using seasonal decomposition methods (SD hereafter), e.g., SABL and X-11. Such SD's are based on moving average methods, and they are thus easy to use and are capable of computing the seasonal pattern that changes over time. A SD typically decomposes a series intoT (trend),S (seasonal component), andR (residual or sometimes referred to as the irregular component). We use an ARIMA model onR to obtain its forecast. TheS component is forecasted by an extrapolation taking into account its changing pattern within the sample period. We propose to set up some scenarios on theT component by examining its possibly nonlinear and nonstationary behavior, and in the paper we suggest one possible way for this. Suppose that the forecasting horizon is relatively short compared toT's several cycles just before the end of the sample. Then we may safely extrapolateT linearly into the forecasting period. LinearizingT in such a case, makes sense. As to the slope of the linear line, we suggest the average rate of change of the most recent upward phase of a cycle to be used if we needed an optimistic scenario. Obviously, that of the downward phase may be used for constructing a pessimistic scenario, and that of one entire cycle is suitable for ‘average’ scenario. Once the forecasted values of the three components are obtained, we may put them back to make predictions on the original series based upon various different scenarios. In addition to proposing a new prediction method, we looked into the following issues, among others, in the paper: (1) on what sort of asset prices would our forecasting method work well? (2) Any significant differences if we used X-11 instead of SABL?