The crude oil market is a major commodity market characterised by the heterogeneity of individual crude oils in terms of production location and physical attributes. This paper applies recent econometric methods to examine how causal relationships between traditional benchmark prices and a broad set of individual crudes have evolved over time. Although the market has become more interconnected, the roles of specific crudes, including established benchmarks, have shifted markedly. The dominance of West Texas Intermediate (WTI) crudes declined for more than a decade after the US shale boom around 2010, while Middle Eastern crudes gained importance, particularly in Asia. Geopolitical disruptions following Russia s invasion of Ukraine in 2022 then prompted a renewed rise in US crude influence, with WTI Cushing, OK, and WTI Midland regaining leadership. Geographical origin and physical characteristics have also shaped these dynamics. Understanding how benchmarks and causality evolve is relevant for pricing, risk management and hedging.

We propose a multiple-equation regression-based method for modeling and forecasting intraday spot volatility. In this approach, intraday intervals are treated as individual time series, deviating from the common practice of treating the data as one continuous sample. Our empirical study, which spans more than two decades and encompasses six US blue-chip stocks, employs the recent OK volatility estimator developed by Li, Wang, and Zhang (2024) to expose the dynamics of latent intraday spot volatility over time. We demonstrate that the proposed method effectively captures the intricate dynamics of intraday spot volatility and find strong evidence that it outperforms a competing regression approach, and popular tree-based machine learning (LightGBM) and deep learning (LSTM) methods, in terms of predictive accuracy as measured by the MSE and QLIKE. These improvements in predictive accuracy extend to logarithmic extensions and across multiple forecast horizons. Overall, our results indicate that the parameter flexibility inherent in the proposed method is advantageous. This flexibility comes without undue computational burden. (c) 2025 The Author(s). Published by Elsevier B.V. on behalf of International Institute of Forecasters. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Forecasting realized volatility (RV) has been widely studied, with numerous techniques developed to enhance predictive accuracy. Among these techniques, the use of RV decompositions based on intraday asset returns has been applied. However, the use of a frequency-based decomposition, which provides unique insights into the dynamics of RV, remains relatively unexplored. This study develops a novel frequency-based (wavelet decomposition) hierarchical forecast reconciliation approach for RV forecasting. Our proposed approach relies on decomposing RV into low, medium, and high frequency components, forecasting these sub-series individually via the HAR model, and applying forecast reconciliation to produce the final RV prediction. Empirical results demonstrate that our proposed wavelet-based framework achieves superior predictive and economically significant performance relative to models employing return-based decompositions and the benchmark HAR, and logarithmic HAR models. The findings also reveal that the wavelet approach provides greater gains at longer forecast horizons, owing to the strong persistence in the low-frequency component of RV. Moreover, this wavelet approach performs well in periods of high volatility and across all market conditions through time.

This study proposes a wavelet-based approach to forecasting Realized Volatility (RV) and evaluates its economic value within a volatility-timing framework. We apply wavelet decomposition to separate short-, medium-, and long-term components and generate forecasts using Heterogeneous Autoregressive (HAR) models. Forecasts based on the low-frequency component consistently lead to better portfolio outcomes, reducing turnover and enhancing investor utility without increasing risk. These results hold even when portfolio weights are forecast directly after being constructed from RV, or when jump-robust volatility estimates are used. The results highlight the importance of aligning forecast evaluation with practical investment objectives. Forecasts delivering the greatest welfare gains may not minimize conventional statistical loss functions.

We analyze whether and to what extent gasoline price shocks influenced headline inflation, one-year gasoline price inflation expectations, and one-year inflation expectations in the United States using a partially identified Bayesian structural vector autoregression model. Results show that headline inflation and one-year inflation expectations increased instantaneously in response to gasoline price shocks, with the effects on one-year inflation expectations being considerably more persistent. The effects on one-year gasoline price inflation expectations were noticeably different. Although these expectations increased initially, they declined two months after the shock, with the effects dissipating shortly thereafter. Variance decomposition reveals that gasoline price shocks explained 66.18 %, 12.06 %, and 22.42 % of the variation in headline inflation, one-year gasoline price inflation expectations, and one-year inflation expectations, respectively.

This paper proposes a new class of marked point process models to capture the clustering behavior in extreme financial events. The idea of multiple dynamic parameters embedded in the context of score driven models is utilized to estimate a dynamic extreme value approach, labeled as the Orthogonal Score-Driven Peaks Over Threshold model. A Monte-Carlo study is conducted to study different time-varying parameter specifications. The results show that this approach can capture a range of different dynamics for the parameters. In an empirical application, we study the dynamics of the tail distribution over time, and in particular on VaR and ES forecasts, for the constituents of the S&P Banks Index. Finally, we study the behavior of extremely adverse returns in the financial system by means of a decomposition of the tail-fl risk measure, giving a deeper understanding of both the dynamics of the risk of an individual bank, and the systemic linkages associated with the stability of the global financial system.

This paper forecasts station-level retail fuel prices using econometric methods, incorporating spatial interdependencies. Error correction models with cross-sectional dependence outperform autoregressive models with wholesale prices or spatial effects, demonstrating the benefits of spatial interdependencies in terms of improved forecasting performance.

We analyze over 1.1 million news articles from The Wall Street Journal to identify the exogenous factors influencing stock market volatility. Using a two-step topic modeling and term frequency approach, we find that news narratives explain a substantial part of the stock market's time variation in volatility. Moreover, incorporating news-based measures into volatility models yields superior out-of-sample forecasts, reducing forecast errors by over 40% at the monthly horizon relative to benchmarks. These improvements translate into significant economic gains through increased realized utility and Sharpe ratios in volatility-targeted strategies. Our approach outperforms ChatGPT-derived predictors and benchmark economic indicators in volatility forecasts.

This paper proposes two new approaches to improve the estimation of the coefficients of the multivariate HAR (MHAR) model with the primary purpose of improving forecast performance. A robust estimator of the covariance matrix is adopted to replace the realized covariance matrix while estimating the MHAR model. The robustness to outliers of the new estimator makes the OLS estimation scheme for the MHAR model more reliable. In addition, a robust estimation scheme is developed for the MHAR model, which is based on the multivariate least-trimmed squares method. Both approaches provide significant improvements in forecasting performance based on both statistical loss and portfolio outcomes. The forecast performance of the multivariate HARQ model can also be improved with the proposed approaches, as evidenced by robustness checks.

Empirical studies on volatility forecasting have predominantly concentrated on point or interval estimates. However, the study of directional changes in volatility remains a relatively underexplored domain. Prediction of directions in volatility is particularly important for timing strategies and asset allocation. In this paper, we consider the predictability of both short-term and long-term volatility directions using a range of machine learning techniques integrating a parsimonious Heterogeneous Autoregressive (HAR) structure. An empirical analysis was undertaken using the S&P500's realized volatility and the Chicago Board Options Exchange's (CBOE) VIX implied volatility. We show that the machine learning techniques consistently enhance the accuracy of volatility directional forecasts. Among these techniques, the Support Vector Machine (SVM) stands out by consistently achieving significant forecasting accuracy and the most substantial economic gains.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

We examine the information flow between equity and credit default swap (CDS) markets using firm-level returns data before and after the global financial crisis. Before the crisis, the information flow was unidirectional, with equity returns leading CDS returns. While equity returns continue to lead CDS returns after the crisis, we find that the speed of adjustment of the CDS market to equity markets has increased during this period. We also find evidence of a bidirectional flow of information between these markets, with equity returns responding to credit protection returns in the postcrisis period. The quicker response of CDS spreads to equity returns during the postcrisis period primarily occurs among entities with lower credit ratings. In contrast, the response of equity returns to lagged CDS returns during the postcrisis period is observed among firms across different credit rating categories; however, the magnitude of the response is higher among those with lower credit ratings.

This paper demonstrates that existing quantile regression models used for jointly forecasting Value-at-Risk (VaR) and expected shortfall (ES) are sensitive to initial conditions. Given the importance of these measures in financial systems, this sensitivity is a critical issue. A new Bayesian quantile regression approach is proposed for estimating joint VaR and ES models. By treating the initial values as unknown parameters, sensitivity issues can be dealt with. Furthermore, new additive-type models are developed for the ES component that are more robust to initial conditions. A novel approach using the open-faced sandwich (OFS) method is proposed which improves uncertainty quantification in risk forecasts. Simulation and empirical results highlight the improvements in risk forecasts ensuing from the proposed methods.

Forecasts of the covariance matrix of returns is a crucial input into portfolio construction. In recent years multivariate version of the Heterogenous AutoRegressive (HAR) models have been designed to utilise realised measures of the covariance matrix to generate forecasts. This paper shows that combining forecasts from simple HAR-like models provide more coefficients estimates, stable forecasts and lower portfolio turnover. The economic benefits of the combination approach become crucial when transactions costs are taken into account. This combination approach also provides benefits in the context of direct forecasts of the portfolio weights. Economic benefits are observed at both 1-day and 1-week ahead forecast horizons.

Self- and cross-excitation in point processes are commonly captured in the financial econometrics literature using a multivariate exponential memory kernel. In this article, the exponential assumption is relaxed and the resultant non-parametric memory kernel is estimated by a method based on second-order cumulants. The estimator is shown to be consistent and asymptotically normally distributed and performs well under simulation. An empirical application based on 10 international stock indices is presented. Two different indices of contagion between markets are constructed from the point process models in order to examine interconnection over time. A conclusion which emerges from these results is the assumption that a parametric kernel may be too restrictive as the application reveals interesting features, and in some cases substantial differences, between the exponential and non-parametric kernels.

This paper develops a new class of dynamic models for forecasting extreme financial risk. This class of models is driven by the score of the conditional distribution with respect to both the duration between extreme events and the magnitude of these events. It is shown that the models are a feasible method for modeling the time-varying arrival intensity and magnitude of extreme events. It is also demonstrated how exogenous variables such as realized measures of volatility can easily be incorporated. An empirical analysis based on a set of major equity indices shows that both the arrival intensity and the size of extreme events vary greatly during times of market turmoil. The proposed framework performs well relative to competing approaches in forecasting extreme tail risk measures.

As we grow more reliant on renewable generation sources to meet our energy requirements, forecasts of the generation output become more important. This paper proposes a time-series econometric framework for modelling and forecasting intraday rooftop solar generation, one that relies only on historical generation data. Some elements have been adapted from financial time series models to deal with the strongly diurnal, and seasonal nature of the data. Based on an application to regional wide aggregate output, it is shown that this framework captures the important features of the data and produces reliable forecasts of the level and volatility of output. This is achieved using traditional time-series methods, and without relying on large-scale weather information.

This paper considers how information from the implied volatility (IV) term structure can be harnessed to improve stock return volatility forecasting within the state-of-the-art HAR model. Factors are extracted from the IV term structure and included as exogenous variables in the HAR framework. We found that including slope and curvature factors leads to significant forecast improvements over the HAR benchmark at a range of forecast horizons, compared with the standard HAR model and HAR model with VIX as IV information set.

Heuristic judgements and cognitive biases characterise the choices that humans make under uncertainty. If these characteristics are considered to be systematic, as indicated in the literature, then they are central to explaining the behaviour of financial markets. In this paper we present a novel framework to model the effects of cognitive biases on financial markets. Specifically, we employ an artificial neural network (ANN) as a proxy for an investor’s decision-making mechanism. A cognitive bias is introduced into the decision-making process by selectively choosing the data on which the ANN is trained and tested. The example presented in this paper considers the effect of hind-sight bias, that events appear more predictable after they have occurred, on financial markets. We show our methodology is able to replicate the return characteristics of the S&P 500 index from 1978 to 2015. In contrast, previous studies have concluded that observed returns are difficult to reconcile with standard economic models, which assume rational investor behaviour. In addition to the hindsight bias, other cognitive biases and systematic behavioural errors may be considered within the framework we present. We therefore anticipate that the methodology will facilitate significant further research to understand how investor behaviours effect financial markets.

This paper is the first to consider the link between information in the facial expressions of economic actors and economic activity.While much research has focused on text based sentiment, little is understood about the possible information conveyed by facial expressions.A collection of media photographs corresponding to the US economy for the period 1996-2018 is used to construct indices of emotions communicated by facial expressions.The indices are correlated with business cycle conditions, and also contain information about the future state of the business cycle, information that is incremental to that contained in commonly used measures of economic conditions.Social cognition means that emotions transmitted via facial expressions can influence expectations regarding economic prospects, decision making and hence future economic outcomes.The results here open up the possibility of using information from facial expressions to augment existing methods for using text-based sentiment for macroeconomic or financial forecasting.