|
on Forecasting |
| By: | Shuping Shi (Department of Economics, Macquarie University); Jun Yu (Faculty of Business Administration, University of Macau); Chen Zhang (Department of Economics, Sun Yat-sen University) |
| Abstract: | Forecasting realized volatility (RV) is central to financial econometrics, with important implications for risk management, asset allocation, and derivative pricing. Motivated by the ongoing debate on volatility modeling, this paper provides a comprehensive empirical comparison of many alternative models. We evaluate leading continuous time models estimated using state-of-the-art methods from the rough volatility literature, together with both standard long-memory autoregressive fractionally integrated moving average (ARFIMA) models and their rough-volatility extensions, as well as several variants of the heterogeneous autoregressive (HAR) model and their logarithmic counterparts. The models are applied to a large panel of equities and cryptocurrencies, with performance assessed using both statistical and economic criteria. Our results show that for equities, continuous time models consistently outperform discrete time alternatives across all evaluation criteria and forecasting horizons. The fractional Brownian motion model for log RV performs best at short horizons, while the fractional Ornstein Uhlenbeck model for log RV dominates in the long run. For cryptocurrencies, a mild divergence emerges between economic and statistical performance: based on realized utility, the quarticity-augmented heterogeneous autoregressive (HARQ) model for RV leads in the short term and the Brownian semistationary models prevail at longer horizons, whereas the HAR-type models for log RV deliver superior statistical accuracy. |
| Keywords: | Realized volatility, Continuous-time models, Discrete-time models, forecasting, economic utility |
| Date: | 2025–10 |
| URL: | https://d.repec.org/n?u=RePEc:boa:wpaper:202537 |
| By: | Mir, Zulfiqar Ali |
| Abstract: | This paper examines the effectiveness of penalized regression techniques in forecasting exchange rate movements. Using daily data for the U.S. Dollar Index (DXY) in 2016, we compare the performance of Ordinary Least Squares (OLS) with Ridge and Lasso regression models. The predictors include gold and silver returns, the S&P 500 Index, short- and long-term Treasury yields, and the EURUSD exchange rate. Results show that while OLS suffers from instability due to multicollinearity, Ridge regression improves coefficient stability and predictive accuracy. Lasso regression provides the best overall performance, with the highest explanatory power and the lowest prediction error, by selecting only the most relevant variables. These findings underscore the value of penalized regression in financial econometrics and highlight its potential for robust exchange rate forecasting. |
| Keywords: | Penalized Regression, Ridge, Lasso, Exchange Rate Forecasting, Dollar Index, Financial Econometrics, Machine Learning in Finance |
| JEL: | C01 C55 C58 F31 F37 F47 G12 G15 G17 |
| Date: | 2025–09–01 |
| URL: | https://d.repec.org/n?u=RePEc:pra:mprapa:125996 |
| By: | Gabriele Casto |
| Abstract: | We introduce the Historical and Dynamic Volatility Ratios (HVR/DVR) and show that equity and index volatilities are cointegrated at intraday and daily horizons. This allows us to construct a VECM to forecast portfolio volatility by exploiting volatility cointegration. On S&P 500 data, HVR is generally stationary and cointegration with the index is frequent; the VECM implementation yields substantially lower mean absolute percentage error (MAPE) than covariance-based forecasts at short- to medium-term horizons across portfolio sizes. The approach is interpretable and readily implementable, factorizing covariance into market volatility, relative-volatility ratios, and correlations. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.23533 |
| By: | Runyao Yu; Ruochen Wu; Yongsheng Han; Jochen L. Cremer |
| Abstract: | Accurate probabilistic forecasting of intraday electricity prices is critical for market participants to inform trading decisions. Existing studies rely on specific domain features, such as Volume-Weighted Average Price (VWAP) and the last price. However, the rich information in the orderbook remains underexplored. Furthermore, these approaches are often developed within a single country and product type, making it unclear whether the approaches are generalizable. In this paper, we extract 384 features from the orderbook and identify a set of powerful features via feature selection. Based on selected features, we present a comprehensive benchmark using classical statistical models, tree-based ensembles, and deep learning models across two countries (Germany and Austria) and two product types (60-min and 15-min). We further perform a systematic generalization study across countries and product types, from which we reveal an asymmetric generalization phenomenon. |
| Date: | 2025–10 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2510.12685 |
| By: | Yiyao Zhang; Diksha Goel; Hussain Ahmad; Claudia Szabo |
| Abstract: | Financial markets are inherently non-stationary, with shifting volatility regimes that alter asset co-movements and return distributions. Standard portfolio optimization methods, typically built on stationarity or regime-agnostic assumptions, struggle to adapt to such changes. To address these challenges, we propose RegimeFolio, a novel regime-aware and sector-specialized framework that, unlike existing regime-agnostic models such as DeepVol and DRL optimizers, integrates explicit volatility regime segmentation with sector-specific ensemble forecasting and adaptive mean-variance allocation. This modular architecture ensures forecasts and portfolio decisions remain aligned with current market conditions, enhancing robustness and interpretability in dynamic markets. RegimeFolio combines three components: (i) an interpretable VIX-based classifier for market regime detection; (ii) regime and sector-specific ensemble learners (Random Forest, Gradient Boosting) to capture conditional return structures; and (iii) a dynamic mean-variance optimizer with shrinkage-regularized covariance estimates for regime-aware allocation. We evaluate RegimeFolio on 34 large cap U.S. equities from 2020 to 2024. The framework achieves a cumulative return of 137 percent, a Sharpe ratio of 1.17, a 12 percent lower maximum drawdown, and a 15 to 20 percent improvement in forecast accuracy compared to conventional and advanced machine learning benchmarks. These results show that explicitly modeling volatility regimes in predictive learning and portfolio allocation enhances robustness and leads to more dependable decision-making in real markets. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2510.14986 |
| By: | Ruslan Tepelyan |
| Abstract: | OHLC bar data is a widely used format for representing financial asset prices over time due to its balance of simplicity and informativeness. Bloomberg has recently introduced a new bar data product that includes additional timing information-specifically, the timestamps of the open, high, low, and close prices within each bar. In this paper, we investigate the impact of incorporating this timing data into machine learning models for predicting volume-weighted average price (VWAP). Our experiments show that including these features consistently improves predictive performance across multiple ML architectures. We observe gains across several key metrics, including log-likelihood, mean squared error (MSE), $R^2$, conditional variance estimation, and directional accuracy. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.16137 |
| By: | Nikolay Gospodinov; Esfandiar Massoumi |
| Abstract: | Policy makers express their views and decisions via the lens of a particular model or theory. But since any model is a highly stylized representation of the unknowable object of interest, all these models are inherently misspecified, and the resulting ambiguity injects uncertainty in the decision-making process. We argue that entropy-based aggregation is a convenient device to confront this uncertainty and summarize relevant information from a set of candidate models and forecasts. The proposed aggregation tends to robustify the decision-making process to various sources of risks and uncertainty. We find compelling evidence for the advantages of entropy-based aggregation for forecasting inflation. |
| Keywords: | model uncertainty; model aggregation; forecast combination; robust policy |
| JEL: | C52 C53 E37 G11 |
| Date: | 2025–10–09 |
| URL: | https://d.repec.org/n?u=RePEc:fip:fedawp:101967 |
| By: | Andrew T. Foerster; Andreas Hornstein; Pierre-Daniel G. Sarte; Mark W. Watson |
| Abstract: | We explore the evolving significance of different production sectors within the U.S. economy since World War II and provide methods for estimating and forecasting these shifts. Using a compositional accounting approach, we find that the well-documented transition from goods to services is primarily driven by two compositional changes: 1) the rise of Intellectual Property Products (IPP) as an input producer, replacing Durable Goods almost one-for-one in terms of input shares in virtually all sectors; and 2) a shift in consumer spending from Nondurable Goods to Services. A structural model replicating these shifts reveals that the rise of IPP at the expense of Durable Goods is largely explained by increases in the efficiency of IPP inputs used in production: input-biased technical change. Trend variations in sectoral total factor productivity, and their attendant effects on relative prices and income, are the main driver of evolving consumption patterns. Both reduced-form and structural forecasts project these trends to continue over the next two decades, albeit at lower rates, indicating a slower pace of structural change. |
| Keywords: | structural changes; forecasting; technical change |
| JEL: | E17 E23 E27 |
| Date: | 2025–10–14 |
| URL: | https://d.repec.org/n?u=RePEc:fip:fedfwp:101936 |
| By: | Stephane Hess; Sander van Cranenburgh |
| Abstract: | Models allowing for random heterogeneity, such as mixed logit and latent class, are generally observed to obtain superior model fit and yield detailed insights into unobserved preference heterogeneity. Using theoretical arguments and two case studies on revealed and stated choice data, this paper highlights that these advantages do not translate into any benefits in forecasting, whether looking at prediction performance or the recovery of market shares. The only exception arises when using conditional distributions in making predictions for the same individuals included in the estimation sample, which obviously precludes any out-of-sample forecasting. |
| Date: | 2025–10 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2510.09185 |
| By: | Emanuel Kohlscheen |
| Abstract: | This article identifies the factors that drove house prices in 13 advanced countries over the past 35 years. It does so based on Breiman s (2001) random forest model. Shapley values indicate that annual house price growth across countries is explained first and foremost by price momentum, initial valuations (proxied by price to rent ratios) and household credit growth. Partial effects of explanatory variables are also elicited and suggest important non-linearities, for instance as to what concerns the effects of CPI inflation on house price growth. The out-of-sample forecast test reveals that the random forest model delivers 44% lower house price variation RMSEs and 45% lower MAEs when compared to an OLS model that uses the same set of 10 pre-determined explanatory variables. Notably, the same model works well for all countries, as the random forest attributes minimal values to country fixed effects. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.21460 |
| By: | Abderraouf Ben Ahmed Mtiraoui (MOFID-Université de Sousse); Nadia Slimene (SU - Shaqra University, Saudi Arabia); Leila Chemli (Faculté des Sciences Economiques et de Gestion de Sousse, Université de Sousse, Tunisia) |
| Abstract: | The purpose of this paper is to assess the ability of a VAR model, used to predict. The results of the estimates lead to adopting a VAR model. However, the performances of this model are quite close, for certain horizons, to those performed by the forecasting organizations for the time series. We will first do a detailed analysis of Bitcoin prices, including the closing price. Next, we will move on to modeling the Bitcoin series using the VAR model, which will then be used for forecasting. We will move on to modeling the Bitcoin series using the VAR model, which consumers will then use. |
| Keywords: | VAR, Time Series, Bitcoin |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:hal:journl:hal-05253337 |
| By: | Shaw Dalen |
| Abstract: | Prediction markets, such as Polymarket, aggregate dispersed information into tradable probabilities, but they still lack a unifying stochastic kernel comparable to the one options gained from Black-Scholes. As these markets scale with institutional participation, exchange integrations, and higher volumes around elections and macro prints, market makers face belief volatility, jump, and cross-event risks without standardized tools for quoting or hedging. We propose such a foundation: a logit jump-diffusion with risk-neutral drift that treats the traded probability p_t as a Q-martingale and exposes belief volatility, jump intensity, and dependence as quotable risk factors. On top, we build a calibration pipeline that filters microstructure noise, separates diffusion from jumps using expectation-maximization, enforces the risk-neutral drift, and yields a stable belief-volatility surface. We then define a coherent derivative layer (variance, correlation, corridor, and first-passage instruments) analogous to volatility and correlation products in option markets. In controlled experiments on synthetic risk-neutral paths and real event data, the model reduces short-horizon belief-variance forecast error relative to diffusion-only and probability-space baselines, supporting both causal calibration and economic interpretability. Conceptually, the logit jump-diffusion kernel supplies an implied-volatility analogue for prediction markets: a tractable, tradable language for quoting, hedging, and transferring belief risk across venues such as Polymarket. |
| Date: | 2025–10 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2510.15205 |
| By: | Fabien Sanchez (G2Elab-SYREL - G2Elab-SYstèmes et Réseaux ELectriques - G2ELab - Laboratoire de Génie Electrique de Grenoble - CNRS - Centre National de la Recherche Scientifique - UGA - Université Grenoble Alpes - Grenoble INP - Institut polytechnique de Grenoble - Grenoble Institute of Technology - UGA - Université Grenoble Alpes); Ahmed Mohamed (G2Elab-SYREL - G2Elab-SYstèmes et Réseaux ELectriques - G2ELab - Laboratoire de Génie Electrique de Grenoble - CNRS - Centre National de la Recherche Scientifique - UGA - Université Grenoble Alpes - Grenoble INP - Institut polytechnique de Grenoble - Grenoble Institute of Technology - UGA - Université Grenoble Alpes); Rémy Rigo-Mariani (G2Elab-SYREL - G2Elab-SYstèmes et Réseaux ELectriques - G2ELab - Laboratoire de Génie Electrique de Grenoble - CNRS - Centre National de la Recherche Scientifique - UGA - Université Grenoble Alpes - Grenoble INP - Institut polytechnique de Grenoble - Grenoble Institute of Technology - UGA - Université Grenoble Alpes); Vincent Debusschere (G2Elab-SYREL - G2Elab-SYstèmes et Réseaux ELectriques - G2ELab - Laboratoire de Génie Electrique de Grenoble - CNRS - Centre National de la Recherche Scientifique - UGA - Université Grenoble Alpes - Grenoble INP - Institut polytechnique de Grenoble - Grenoble Institute of Technology - UGA - Université Grenoble Alpes) |
| Abstract: | This paper discusses the implementation of supervised learning (SL) as a straightforward data-driven technique to compute the day-ahead bids of grid-connected battery energy systems (BESS) participating in energy markets. The objective is to implicitly account for price uncertainty in the BESS schedule before assessing the economic performance. The case study is a 10 MW BESS battery participating in the day-ahead market. Physic-Informed and more traditional loss functions and a large set of tuning parameters are compared based on the generated daily revenues. Either the power injected by the BESS or its state of charge is controlled, illustrating a compromise to find between the accuracy and the resilience of the results, once confronted with the high volatility of energy prices. The performance of AI-based controllers is assessed in terms of precision with a theoretical optimum obtained with a "perfect forecast". A reference bidding strategy using "backcasting" as a forecast is also considered. Simulation over the year 2021 with an hourly training data set of the energy prices of 2020 shows that SL models do not necessarily perform better than reference results (with a minimal error of 58 %). However, discussions about their tuning and design choices shed light on the complex implementation process of the selected case study. |
| Keywords: | Machine Leaning, Energy Market, Storage |
| Date: | 2025–06–29 |
| URL: | https://d.repec.org/n?u=RePEc:hal:journl:hal-05310545 |
| By: | Patrick C. Higgins |
| Abstract: | Although there have been a range of studies investigating the role and importance of global supply and demand shocks in US inflation developments during and since the pandemic, this study uses a heretofore unused dataset for this purpose: a quarterly panel of professional forecasts from Consensus Economics. We use real-time data with daily vintage snapshots since 2005 from the Federal Reserve Board of Governors FAME database to disentangle forecast errors from revisions and to exploit the monthly frequency and partial availability of CPI inflation and industrial production. Our measures of global demand and supply shocks account for nearly 60 percent, and 20 percent, respectively, of the total variability of the five global factors we identify. The global demand shock accounts for a greater share of unanticipated US economic activity growth and inflation than the global supply shock both prior to the pandemic and during and after 2020. Since 2020, however, global demand and global supply shocks have accounted for similar shares of the nowcast errors for US inflation. |
| Keywords: | global shocks; professional forecasts; inflation |
| JEL: | C32 E31 E37 F47 |
| Date: | 2025–10–06 |
| URL: | https://d.repec.org/n?u=RePEc:fip:fedawp:101965 |