Journal of Time Series Analysis最新文献

We introduce a novel class of stochastic volatility models, which can utilize and relate many high-frequency realized volatility (RV) measures to latent volatility. Instrumental variable methods provide a unified framework for estimation and testing. We study parameter inference problems in the proposed framework with nonstationary stochastic volatility and exogenous predictors in the latent volatility process. Identification-robust methods are developed for a joint hypothesis involving the volatility persistence parameter and the autocorrelation parameter of the composite error (or the noise ratio). For inference about the volatility persistence parameter, projection techniques are applied. The proposed tests include Anderson-Rubin-type tests and their point-optimal versions. For distributional theory, we provide finite-sample tests and confidence sets for Gaussian errors, establish exact Monte Carlo test procedures for non-Gaussian errors (possibly heavy-tailed), and show asymptotic validity under weaker assumptions. Simulation results show that the proposed tests outperform the asymptotic test regarding size and exhibit excellent power in empirically realistic settings. The proposed inference methods are applied to IBM's price and option data (2009–2013). We consider 175 different instruments (IVs) spanning 22 classes and analyze their ability to describe the low-frequency volatility. IVs are compared based on the average length of the proposed identification-robust confidence intervals. The superior instrument set mostly comprises 5-min HF realized measures, and these IVs produce confidence sets which show that the volatility process is nearly unit-root. In addition, we find RVs with higher frequency yield wider confidence intervals than RVs with slightly lower frequency, indicating that these confidence intervals adjust to absorb market microstructure noise. Furthermore, when we consider irrelevant or weak IVs (jumps and signed jumps), the proposed tests produce unbounded confidence intervals. We also find that both RV and BV measures produce almost identical confidence intervals across all 14 subclasses, confirming that our methodology is robust in the presence of jumps. Finally, although jumps contain little information regarding the low-frequency volatility, we find evidence that there may be a nonlinear relationship between jumps and low-frequency volatility.

In recognition of authors who have made significant contributions to this Journal, the Journal of Time Series Analysis runs a scheme to honour those authors by naming them as a Journal of Time Series Analysis Distinguished Author. The qualifying criterion for this award is 3.5 points where authors are awarded 1 point for each single-authored article, 1/2 point for each double-authored article, 1/3 point for each triple-authored article, and so on, that they have published in the Journal of Time Series Analysis since its inception. Distinguished Authors are entitled to a one-year free online subscription to the Journal to mark the award. They also receive a certificate commemorating the award.

In addition to the lists of Distinguished Authors announced previously in Volume 41 issue 4 (July 2020), Volume 42 Issue 1 (January 2021), Volume 43 Issue 1 (January 2022), Volume 44 Issue 1 (January 2023), and Volume 45 Issue 1 (January 2024), the Journal of Time Series Analysis is very pleased to welcome Konstantinos Fokianos to the list of Journal of Time Series Analysis Distinguished Authors for 2024, based on his publications in the Journal appearing up to and including Volume 45 Issue 6 (November 2024).

The author declares no conflicts of interest.

QFFE stands for Quantitative Finance and Financial Econometrics conference, an event organized by Sébastien Laurent in Marseille every year since 2018. Each year there are two keynote speakers and two guest speakers, and around 60 selected papers are presented. The program for next year and previous years can be found here. The conference is preceded by a spring school, which offers doctoral students, post-doc, and young academics the opportunity to attend doctoral-level courses.

The QFFE conference is part of the ANR-funded project MLEforRisk (ANR-21-CE26-0007), which stands for Machine Learning and Econometrics for Risk Measurement in Finance. The project seeks to enhance our understanding of the advantages and limitations of integrating econometric methods with machine learning for measuring financial risks. This multidisciplinary initiative bridges the fields of finance and financial econometrics, bringing together a team of junior and senior researchers with expertise in management, economics, applied mathematics, and data science. The project aims to advance both theoretical insights and practical applications, fostering innovation at the intersection of these disciplines.

Since financial data such as stock prices, interest rates, and exchange rates are observed over time, time series analysis is crucial in finance. Finance professionals and academics often rely on fundamental time series models, such as ARMA, as well as essential time series techniques such as spectral analysis. Financial researchers are therefore naturally attracted to any new developments in time series. Econometricians have also developed new time series models and methods to capture the specificities of financial data. Contributions of econometricians include cointegration and error correction models, GARCH and stochastic volatility models, score-driven models, VAR models, Markov switching models, non-causal models, simulation-based inference, state space models, and Kalman filters, realized volatility measures, the Black–Scholes model, and factor models. The field of application of all these time series models and techniques is obviously not limited to finance. The aim of this special issue is to present some recent examples of the interface between time series analysis and finance.

We are very grateful to these authors. We would also like to thank the anonymous reviewers for their valuable review and feedback, which helped to improve the quality of this special issue. Special thanks go to Robert Taylor, Editor-in-Chief of the Journal of Time Series Analysis, for supporting this project, as well as to Priscilla Goldby for her invaluable help.

The puzzling negative relation between liquidity uncertainty and asset returns, originally put forward by Chordia, Subrahmanyam, and Anshuman (2001) and confirmed by the subsequent empirical literature up to date, is neither robust to the aggregation period, nor to the observation frequency used to compute the volatility of trading volume. We demonstrate that their procedure involves an estimation bias due to the persistence and skewness of volumes. When using an alternative approach based on high-frequency data to measure liquidity uncertainty, the relationship turns out to be positive. However, portfolio strategies based on liquidity uncertainty do not appear to be profitable.

Risk parity portfolio optimization, using expected shortfall as the risk measure, is investigated when asset returns are fat-tailed and heteroscedastic with regime switching dynamic correlations. The conditional return distribution is modeled by an elliptical multi-variate generalized hyperbolic distribution, allowing for fast parameter estimation via an expectation-maximization algorithm, and a semi-closed form of the risk contributions. A new method for efficient computation of non-Gaussian risk parity weights sidesteps the need for numerical simulations or Cornish–Fisher-type approximations. Accounting for fat-tailed returns, the risk parity allocation is less sensitive to volatility shocks, thereby generating lower portfolio turnover, in particular during market turmoils such as the global financial crisis or the COVID shock. While risk parity portfolios are rather robust to the misuse of the Gaussian distribution, a sophisticated time series model can improve risk-adjusted returns, strongly reduces drawdowns during periods of market stress and enables to use a holistic risk model for portfolio and risk management.

This article develops statistical inference methods for a class of set-identified models, where the errors are known functions of observations and the parameters satisfy either serial or/and cross-sectional independence conditions. This class of models includes the independent component analysis (ICA), Structural Vector Autoregressive (SVAR), and multi-variate mixed causal–non-causal models. We use the Generalized Covariance (GCov) estimator to compute the residual-based portmanteau statistic for testing the error independence hypothesis. Next, we build the confidence sets for the identified sets of parameters by inverting the test statistic. We also discuss the choice (design) of these statistics. The approach is illustrated by simulations examining the under-identification condition in an ICA model and an application to financial return series.

The mixed causal-non-causal invertible-non-invertible autoregressive moving-average (MARMA) models have the advantage of incorporating roots inside the unit circle, thus adjusting the dynamics of financial returns that depend on future expectations. This article introduces new techniques for estimating, identifying and simulating MARMA models. Although the estimation of the parameters is done using second-order moments, the identification relies on the existence of high-order dynamics, captured in the high-order spectral densities and the correlation of the squared residuals. A comprehensive Monte Carlo study demonstrated the robust performance of our estimation and identification methods. We propose an empirical application to 24 portfolios from emerging markets based on the factors: size, book-to-market, profitability, investment and momentum. All portfolios exhibited forward-looking behavior, showing significant non-causal and non-invertible dynamics. Moreover, we found the residuals to be uncorrelated and independent, with no trace of conditional volatility.

In this article, we analyse the forecasting performance of several parametric extensions of the popular Dynamic Nelson–Siegel (DNS) model for the yield curve. Our focus is on the role of additional and time-varying decay parameters, conditional heteroscedasticity, and macroeconomic variables. We also consider the role of several popular restrictions on the dynamics of the factors. Using a novel dataset of end-of-month continuously compounded Treasury yields on US zero-coupon bonds and frequentist estimation based on the extended Kalman filter, we show that a second decay parameter does not contribute to better forecasts. In concordance with the preferred habitat theory, we also show that the best forecasting model depends on the maturity. For short maturities, the best performance is obtained in a heteroscedastic model with a time-varying decay parameter. However, for long maturities, neither the time-varying decay nor the heteroscedasticity plays any role, and the best forecasts are obtained in the basic DNS model with the shape of the yield curve depending on macroeconomic activity. Finally, we find that assuming non-stationary factors is helpful in forecasting at long horizons.