Andras Fulop

Self-exciting jumps, learning, and asset pricing implications

The paper proposes a new class of continuous-time asset pricing models where negative jumps play a crucial role. Whenever there is a negative jump in asset returns, it is simultaneously passed on to diffusion variance and the jump intensity, generating self-exciting co-jumps of prices and volatility and jump clustering. To properly deal with parameter uncertainty and in-sample over-fitting, a Bayesian learning approach combined with an efficient particle filter is employed. It not only allows for comparison of both nested and non-nested models, but also generates all quantities necessary for sequential model analysis. Empirical investigation using S&P 500 index returns shows that volatility jumps at the same time as negative jumps in asset returns mainly through jumps in diffusion volatility. We find substantial evidence for jump clustering, in particular, after the recent financial crisis in 2008, even though parameters driving dynamics of the jump intensity remain difficult to identify.The paper proposes a self-exciting asset pricing model that takes into account co-jumps between prices and volatility and self-exciting jump clustering. We employ a Bayesian learning approach to implement real time sequential analysis. We find evidence of self-exciting jump clustering since the 1987 market crash, and its importance becomes more obvious at the onset of the 2008 global financial crisis. It is found that learning affects the tail behaviors of the return distributions and has important implications for risk management, volatility forecasting and option pricing.

Efficient Learning via Simulation: A Marginalized Resample-Move Approach

In state-space models, parameter learning is practically dicult and is still an open issue. This paper proposes an ecient simulation-based parameter learning method. First, the approach breaks up the interdependence of the hidden states and the static parameters by marginalizing out the states using a particle lter. Second, it applies a Bayesian resample-move approach to this marginalized system. The methodology is generic and needs little design eort. Dierent from batch estimation methods, it provides posterior quantities necessary for full sequential inference and recursive model monitoring. The algorithm is implemented both on simulated data in a linear Gaussian model for illustration and comparison and on real data in a L evy jump stochastic volatility model and a structural credit risk model.

Density-Tempered Marginalized Sequential Monte Carlo Samplers

We propose a density-tempered marginalized sequential Monte Carlo (SMC) sampler, a new class of samplers for full Bayesian inference of general state-space models. The dynamic states are approximately marginalized out using a particle filter, and the parameters are sampled via a sequential Monte Carlo sampler over a density-tempered bridge between the prior and the posterior. Our approach delivers exact draws from the joint posterior of the parameters and the latent states for any given number of state particles and is thus easily parallelizable in implementation. We also build into the proposed method a device that can automatically select a suitable number of state particles. Since the method incorporates sample information in a smooth fashion, it delivers good performance in the presence of outliers. We check the performance of the density-tempered SMC algorithm using simulated data based on a linear Gaussian state-space model with and without misspecification. We also apply it on real stock prices using a GARCH-type model with microstructure noise.

A stable estimator of the information matrix under EM for dependent data

This article develops a new and stable estimator for information matrix when the EM algorithm is used in maximum likelihood estimation. This estimator is constructed using the smoothed individual complete-data scores that are readily available from running the EM algorithm. The method works for dependent data sets and when the expectation step is an irregular function of the conditioning parameters. In comparison to the approach of Louis (J. R. Stat. Soc., Ser. B 44:226–233, 1982), this new estimator is more stable and easier to implement. Both real and simulated data are used to demonstrate the use of this new estimator.

Multiperiod Corporate Default Prediction with the Partially-Conditioned Forward Intensity

The forward-intensity model of Duan, {et al} (2012) is a parsimonious and practical way for predicting corporate defaults over multiple horizons. However, it has a noticeable shortcoming because default correlations through intensities are conspicuously absent when the prediction horizon is more than one data period. We propose a new forward-intensity approach that builds in correlations among intensities of individual obligors by conditioning all forward intensities on the future values of some common variables, such as the observed interest rate and/or a latent frailty factor. The new model is implemented on a large sample of US industrial and financial firms spanning the period 1991-2011 on the monthly frequency. Our findings suggest that the new model is able to correct the structural biases at longer prediction horizons reported in Duan et al (2012). Not surprisingly, default correlations are also found to be important in describing the joint default behavior.

Transparency Regime Initiatives and Liquidity in the CDS Market

This paper investigates liquidity changes in the corporate CDS market around two events that increased market transparency in the midst of the financial crisis: the regular dissemination of post-trade data by DTCC starting November 2008, and the implementation of the Small Bang in June 2009. We build an econometric model based on intra-daily bid and ask quotes to measure liquidity and volatility in thinly traded CDS. We find that, after DTCCs release, the market-wide deterioration in CDS liquidity becomes less important for banks and major dealers, consistent with information revelation on counterparty risk. The Small Bang also improved liquidity, particularly for more illiquid CDS.

Bayesian Analysis of Bubbles in Asset Prices

We develop a new model where the dynamic structure of the asset price, after the fundamental value is removed, is subject to two different regimes. One regime reflects the normal period where the asset price divided by the dividend is assumed to follow a mean-reverting process around a stochastic long run mean. The second regime reflects the bubble period with explosive behavior. Stochastic switches between two regimes and non-constant probabilities of exit from the bubble regime are both allowed. A Bayesian learning approach is employed to jointly estimate the latent states and the model parameters in real time. An important feature of our Bayesian method is that we are able to deal with parameter uncertainty and at the same time, to learn about the states and the parameters sequentially, allowing for real time model analysis. This feature is particularly useful for market surveillance. Analysis using simulated data reveals that our method has good power properties for detecting bubbles. Empirical analysis using price-dividend ratios of S&P500 highlights the advantages of our method.

Real-Time Bayesian Learning and Bond Return Predictability

The paper examines statistical and economic evidence of out-of-sample bond return predictability for a real-time Bayesian investor who learns about parameters, hidden states, and predictive models over time. We find some statistical evidence using information contained in forward rates. However, such statistical predictability can hardly generate any economic value for investors. Furthermore, we find that strong statistical and economic evidence of bond return predictability from fully-revised macroeconomic data vanishes when real-time macroeconomic information is used. We also show that highly levered investments in bonds can improve short-run bond return predictability.

Data-Cloning SMC 2 for Applications to Latent Variable Models

A data-cloning SMC2 maximum likelihood estimation algorithm is proposed as a general-purpose optimization routine for models with latent variables. Our algorithm first marginalizes out latent variables by applying one layer of SMC at a fixed parameter value, and then estimates the model parameters by another layer of SMC utilizing density-tempering. Data-cloning is employed to effectively reduce Monte Carlo errors inherent in the SMC2 algorithm, and also to address multi-modality present in typical latent variable models. This new method has wide applicability and can be massively parallelized to take advantage of typical computers today.

Maximum Likelihood Estimation of Latent Variable Models by SMC with Marginalization and Data Cloning

A data-cloning SMC² method is proposed as a general purpose optimization routine for estimating latent variable models by maximum likelihood. The latent variables are first marginalized out by SMC at any fixed parameter value, and the model parameters are then estimated by density tempered SMC. The data-cloning step is employed to efficiently reduce Monte Carlo errors inherent in the SMC² algorithm and also to effectively address multi-modality present in typical objective functions. This new method has wide applicability and can be massively parallelized to take advantage of typical computers today.