|
on Forecasting |
| By: | Timoth\'ee Hornek; Sergio Potenciano Menci; Ivan Pavi\'c |
| Abstract: | The increasing penetration of variable renewable energy and flexible demand technologies, such as electric vehicles and heat pumps, introduces significant uncertainty in power systems, resulting in greater imbalance; defined as the deviation between scheduled and actual supply or demand. Short-term power markets, such as the European continuous intraday market, play a critical role in mitigating these imbalances by enabling traders to adjust forecasts close to real time. Due to the high volatility of the continuous intraday market, traders increasingly rely on electricity price forecasting to guide trading decisions and mitigate price risk. However most electricity price forecasting approaches in the literature simplify the forecasting task. They focus on single benchmark prices, neglecting intra-product price dynamics and price signals from the limit order book. They also underuse high-frequency and cross-product price data. In turn, we propose a novel directional electricity price forecasting method for hourly products in the European continuous intraday market. Our method incorporates short-term features from both hourly and quarter-hourly products and is evaluated using German European Power Exchange data from 2024-2025. The results indicate that features derived from the limit order book are the most influential exogenous variables. In addition, features from neighboring products; especially those with delivery start times that overlap with the trading period of the target product; improve forecast accuracy. Finally, our evaluation of the value captured by our electricity price forecasting suggests that the proposed electricity price forecasting method has the potential to generate profit when applied in trading strategies. |
| Date: | 2025–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.04452 |
| By: | Arif Pathan |
| Abstract: | Short-term sentiment forecasting in financial markets (e.g., stocks, indices) is challenging due to volatility, non-linearity, and noise in OHLC (Open, High, Low, Close) data. This paper introduces a novel CMG (Chaos-Markov-Gaussian) framework that integrates chaos theory, Markov property, and Gaussian processes to improve prediction accuracy. Chaos theory captures nonlinear dynamics; the Markov chain models regime shifts; Gaussian processes add probabilistic reasoning. We enhance the framework with transformer-based deep learning models to capture temporal patterns efficiently. The CMG Framework is designed for fast, resource-efficient, and accurate forecasting of any financial instrument's OHLC time series. Unlike traditional models that require heavy infrastructure and instrument-specific tuning, CMG reduces overhead and generalizes well. We evaluate the framework on market indices, forecasting sentiment for the next trading day's first quarter. A comparative study against statistical, ML, and DL baselines trained on the same dataset with no feature engineering shows CMG consistently outperforms in accuracy and efficiency, making it valuable for analysts and financial institutions. |
| Date: | 2025–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2506.17244 |
| By: | Adam Nelson-Archer; Aleia Sen; Meena Al Hasani; Sofia Davila; Jessica Le; Omar Abbouchi |
| Abstract: | We present a deep learning approach for forecasting short-term employment changes and assessing long-term industry health using labor market data from the U.S. Bureau of Labor Statistics. Our system leverages a Long- and Short-Term Time-series Network (LSTNet) to process multivariate time series data, including employment levels, wages, turnover rates, and job openings. The model outputs both 7-day employment forecasts and an interpretable Industry Employment Health Index (IEHI). Our approach outperforms baseline models across most sectors, particularly in stable industries, and demonstrates strong alignment between IEHI rankings and actual employment volatility. We discuss error patterns, sector-specific performance, and future directions for improving interpretability and generalization. |
| Date: | 2025–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2507.01979 |
| By: | Tanujit Chakraborty; Donia Besher; Madhurima Panja; Shovon Sengupta |
| Abstract: | Accurate forecasting of exchange rates remains a persistent challenge, particularly for emerging economies such as Brazil, Russia, India, and China (BRIC). These series exhibit long memory, nonlinearity, and non-stationarity properties that conventional time series models struggle to capture. Additionally, there exist several key drivers of exchange rate dynamics, including global economic policy uncertainty, US equity market volatility, US monetary policy uncertainty, oil price growth rates, and country-specific short-term interest rate differentials. These empirical complexities underscore the need for a flexible modeling framework that can jointly accommodate long memory, nonlinearity, and the influence of external drivers. To address these challenges, we propose a Neural AutoRegressive Fractionally Integrated Moving Average (NARFIMA) model that combines the long-memory representation of ARFIMA with the nonlinear learning capacity of neural networks, while flexibly incorporating exogenous causal variables. We establish theoretical properties of the model, including asymptotic stationarity of the NARFIMA process using Markov chains and nonlinear time series techniques. We quantify forecast uncertainty using conformal prediction intervals within the NARFIMA framework. Empirical results across six forecast horizons show that NARFIMA consistently outperforms various state-of-the-art statistical and machine learning models in forecasting BRIC exchange rates. These findings provide new insights for policymakers and market participants navigating volatile financial conditions. The \texttt{narfima} \textbf{R} package provides an implementation of our approach. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.06697 |
| By: | Kasymkhan Khubiev; Mikhail Semenov; Irina Podlipnova |
| Abstract: | Deep Learning is evolving fast and integrates into various domains. Finance is a challenging field for deep learning, especially in the case of interpretable artificial intelligence (AI). Although classical approaches perform very well with natural language processing, computer vision, and forecasting, they are not perfect for the financial world, in which specialists use different metrics to evaluate model performance. We first introduce financially grounded loss functions derived from key quantitative finance metrics, including the Sharpe ratio, Profit-and-Loss (PnL), and Maximum Draw down. Additionally, we propose turnover regularization, a method that inherently constrains the turnover of generated positions within predefined limits. Our findings demonstrate that the proposed loss functions, in conjunction with turnover regularization, outperform the traditional mean squared error loss for return prediction tasks when evaluated using algorithmic trading metrics. The study shows that financially grounded metrics enhance predictive performance in trading strategies and portfolio optimization. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.04541 |
| By: | Jayanth Athipatla |
| Abstract: | We introduce a novel rough Bergomi (rBergomi) model featuring a variance-driven exponentially weighted moving average (EWMA) time-dependent Hurst parameter $H_t$, fundamentally distinct from recent machine learning and wavelet-based approaches in the literature. Our framework pioneers a unified rough differential equation (RDE) formulation grounded in rough path theory, where the Hurst parameter dynamically adapts to evolving volatility regimes through a continuous EWMA mechanism tied to instantaneous variance. Unlike discrete model-switching or computationally intensive forecasting methods, our approach provides mathematical tractability while capturing volatility clustering and roughness bursts. We rigorously establish existence and uniqueness of solutions via rough path theory and derive martingale properties. Empirical validation on diverse asset classes including equities, cryptocurrencies, and commodities demonstrates superior performance in capturing dynamics and out-of-sample pricing accuracy. Our results show significant improvements over traditional constant-Hurst models. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.05820 |
| By: | Fabio Milani |
| Abstract: | This paper investigates potential nonlinearities in the gain function, which, under adaptive learning, regulates the updating of agents' beliefs in response to recent forecast errors. I use data on professional survey forecasts to estimate nonparametric functional-coefficient regression models. The estimation results reveal nonlinearities in the relationships between expectations and forecast errors, which are indicative of nonlinear gain functions. Gains increase when forecast errors are historically large, and respond asymmetrically to past overpredictions and underpredictions. The findings suggest incorporating nonlinearities in the modeling of learning gains, instead of relying on the constant-gain assumption. |
| Keywords: | survey forecasts, nonlinear gain, adaptive learning, nonparametric regression, functional coefficient regression model |
| JEL: | C14 E31 E32 E70 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:ces:ceswps:_12124 |
| By: | Tony Chernis; Niko Hauzenberger; Haroon Mumtaz; Michael Pfarrhofer |
| Abstract: | We propose a dynamic factor model (DFM) where the latent factors are linked to observed variables with unknown and potentially nonlinear functions. The key novelty and source of flexibility of our approach is a nonparametric observation equation, specified via Gaussian Process (GP) priors for each series. Factor dynamics are modeled with a standard vector autoregression (VAR), which facilitates computation and interpretation. We discuss a computationally efficient estimation algorithm and consider two empirical applications. First, we forecast key series from the FRED-QD dataset and show that the model yields improvements in predictive accuracy relative to linear benchmarks. Second, we extract driving factors of global inflation dynamics with the GP-DFM, which allows for capturing international asymmetries. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.04928 |
| By: | Patrick J. Laub; Tu Pho; Bernard Wong |
| Abstract: | This paper introduces the Actuarial Neural Additive Model, an inherently interpretable deep learning model for general insurance pricing that offers fully transparent and interpretable results while retaining the strong predictive power of neural networks. This model assigns a dedicated neural network (or subnetwork) to each individual covariate and pairwise interaction term to independently learn its impact on the modeled output while implementing various architectural constraints to allow for essential interpretability (e.g. sparsity) and practical requirements (e.g. smoothness, monotonicity) in insurance applications. The development of our model is grounded in a solid foundation, where we establish a concrete definition of interpretability within the insurance context, complemented by a rigorous mathematical framework. Comparisons in terms of prediction accuracy are made with traditional actuarial and state-of-the-art machine learning methods using both synthetic and real insurance datasets. The results show that the proposed model outperforms other methods in most cases while offering complete transparency in its internal logic, underscoring the strong interpretability and predictive capability. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.08467 |
| By: | Irfan Qureshi (Asian Development Bank); Arief Ramayandi (Asian Development Bank Institute); Ghufran Ahmad (Cardiff University) |
| Abstract: | Standard nowcasting frameworks commonly use weekly or monthly variables to monitor quarterly gross domestic product (GDP). However, this method is not suitable for economies that track GDP annually. We modify the state-space representation of an otherwise standard dynamic factor model to represent annual variables as a linear combination of latent monthly indicators for more frequently released variables. Using data from a lower middle-income country, we derive a monthly activity measure that effectively tracks annual GDP growth. These estimates outperform institutional forecasts and competing approaches to estimate low-frequency data. The model offers broader applications to countries facing data limitations, especially lower-income countries |
| Keywords: | monitoring real activity;Kalman filter;dynamic factor model;annual nowcasting |
| JEL: | C38 C53 E37 O11 O47 |
| Date: | 2025–09–16 |
| URL: | https://d.repec.org/n?u=RePEc:ris:adbewp:021542 |
| By: | Guillaume Coqueret; Martial Laguerre |
| Abstract: | This paper investigates the impact of posterior drift on out-of-sample forecasting accuracy in overparametrized machine learning models. We document the loss in performance when the loadings of the data generating process change between the training and testing samples. This matters crucially in settings in which regime changes are likely to occur, for instance, in financial markets. Applied to equity premium forecasting, our results underline the sensitivity of a market timing strategy to sub-periods and to the bandwidth parameters that control the complexity of the model. For the average investor, we find that focusing on holding periods of 15 years can generate very heterogeneous returns, especially for small bandwidths. Large bandwidths yield much more consistent outcomes, but are far less appealing from a risk-adjusted return standpoint. All in all, our findings tend to recommend cautiousness when resorting to large linear models for stock market predictions. |
| Date: | 2025–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2506.23619 |
| By: | Guo, Hongfei; Marín Díazaraque, Juan Miguel; Veiga, Helena |
| Abstract: | Accurately forecasting volatility is central to risk management, portfolio allocation, and asset pricing. While high-frequency realised measures have been shown to improve predictive accuracy, their value is not uniform across markets or horizons. This paper introduces a class of Bayesian neural network stochastic volatility (NN-SV) models that combine the flexibility of machine learning with the structure of stochastic volatility models. The specifications incorporate realised variance, jump variation, and semivariance from daily and intraday data, and model uncertainty is addressed through a Bayesian stacking ensemble that adaptively aggregates predictive distributions. Using data from the DAX, FTSE 100, and S&P 500 indices, the models are evaluated against classical GARCH and parametric SV benchmarks. The results show that the predictive content of high-frequency measures is horizon- and market-specific. The Bayesian ensemble further enhances robustness by exploiting complementary model strengths. Overall, NN-SV models not only outperform established benchmarks in many settings but also provide new insights into market-specific drivers of volatility dynamics. |
| Keywords: | Ensemble forecasts; GARCH; Neural networks; Realised volatility; Stochastic volatility |
| JEL: | C11 C32 C45 C53 C58 |
| Date: | 2025–09–16 |
| URL: | https://d.repec.org/n?u=RePEc:cte:wsrepe:47944 |
| By: | Ayaan Qayyum |
| Abstract: | This paper will discuss how headline data can be used to predict stock prices. The stock price in question is the SPDR S&P 500 ETF Trust, also known as SPY that tracks the performance of the largest 500 publicly traded corporations in the United States. A key focus is to use news headlines from the Wall Street Journal (WSJ) to predict the movement of stock prices on a daily timescale with OpenAI-based text embedding models used to create vector encodings of each headline with principal component analysis (PCA) to exact the key features. The challenge of this work is to capture the time-dependent and time-independent, nuanced impacts of news on stock prices while handling potential lag effects and market noise. Financial and economic data were collected to improve model performance; such sources include the U.S. Dollar Index (DXY) and Treasury Interest Yields. Over 390 machine-learning inference models were trained. The preliminary results show that headline data embeddings greatly benefit stock price prediction by at least 40% compared to training and optimizing a machine learning system without headline data embeddings. |
| Date: | 2025–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2507.01970 |
| By: | Silvia Goncalves; Michael W. McCracken; Yongxu Yao |
| Abstract: | This paper examines the properties of out-of-sample predictability tests evaluated with real-time data subject to annual benchmark revisions. The presence of both regular and annual revisions can create time heterogeneity in the moments of the real-time forecast evaluation function, which is not compatible with the standard covariance stationarity assumption used to derive the asymptotic theory of these tests. To cover both regular and annual revisions, we replace this standard assumption with a periodic covariance stationarity assumption that allows for periodic patterns of time heterogeneity. Despite the lack of stationarity, we show that the Clark and McCracken (2009) test statistic is robust to the presence of annual benchmark revisions. A similar robustness property is shared by the bootstrap test of Goncalves, McCracken, and Yao (2025). Monte Carlo experiments indicate that both tests provide satisfactory finite sample size and power properties even in modest sample sizes. We conclude with an application to U.S. employment forecasting in the presence of real-time data. |
| Keywords: | real-time data; bootstrap; prediction; forecast evaluation |
| JEL: | C53 C12 C52 |
| Date: | 2025–09–11 |
| URL: | https://d.repec.org/n?u=RePEc:fip:fedlwp:101742 |
| By: | Ruofan Xu; Qingliang Fan |
| Abstract: | We propose a characteristics-augmented quantile factor (QCF) model, where unknown factor loading functions are linked to a large set of observed individual-level (e.g., bond- or stock-specific) covariates via a single-index projection. The single-index specification offers a parsimonious, interpretable, and statistically efficient way to nonparametrically characterize the time-varying loadings, while avoiding the curse of dimensionality in flexible nonparametric models. Using a three-step sieve estimation procedure, the QCF model demonstrates high in-sample and out-of-sample accuracy in simulations. We establish asymptotic properties for estimators of the latent factor, loading functions, and index parameters. In an empirical study, we analyze the dynamic distributional structure of U.S. corporate bond returns from 2003 to 2020. Our method outperforms the benchmark quantile Fama-French five-factor model and quantile latent factor model, particularly in the tails ($\tau=0.05, 0.95$). The model reveals state-dependent risk exposures driven by characteristics such as bond and equity volatility, coupon, and spread. Finally, we provide economic interpretations of the latent factors. |
| Date: | 2025–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2506.19586 |
| By: | Lijie Ding; Egang Lu; Kin Cheung |
| Abstract: | We present a deep learning framework for pricing options based on market-implied volatility surfaces. Using end-of-day S\&P 500 index options quotes from 2018-2023, we construct arbitrage-free volatility surfaces and generate training data for American puts and arithmetic Asian options using QuantLib. To address the high dimensionality of volatility surfaces, we employ a variational autoencoder (VAE) that compresses volatility surfaces across maturities and strikes into a 10-dimensional latent representation. We feed these latent variables, combined with option-specific inputs such as strike and maturity, into a multilayer perceptron to predict option prices. Our model is trained in stages: first to train the VAE for volatility surface compression and reconstruction, then options pricing mapping, and finally fine-tune the entire network end-to-end. The trained pricer achieves high accuracy across American and Asian options, with prediction errors concentrated primarily near long maturities and at-the-money strikes, where absolute bid-ask price differences are known to be large. Our method offers an efficient and scalable approach requiring only a single neural network forward pass and naturally improve with additional data. By bridging volatility surface modeling and option pricing in a unified framework, it provides a fast and flexible alternative to traditional numerical approaches for exotic options. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.05911 |
| By: | Adrian Iulian Cristescu; Matteo Giordano |
| Abstract: | Predicting the probability of default (PD) of prospective loans is a critical objective for financial institutions. In recent years, machine learning (ML) algorithms have achieved remarkable success across a wide variety of prediction tasks; yet, they remain relatively underutilised in credit risk analysis. This paper highlights the opportunities that ML algorithms offer to this field by comparing the performance of five predictive models-Random Forests, Decision Trees, XGBoost, Gradient Boosting and AdaBoost-to the predominantly used logistic regression, over a benchmark dataset from Scheule et al. (Credit Risk Analytics: The R Companion). Our findings underscore the strengths and weaknesses of each method, providing valuable insights into the most effective ML algorithms for PD prediction in the context of loan portfolios. |
| Date: | 2025–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2506.19789 |
| By: | Daniel Graeber; Lorenz Meister; Carsten Schröder; Sabine Zinn |
| Abstract: | Machine learning is increasingly used in social science research, especially for prediction. However, the results are sometimes not as straight-forward to interpret compared to classic regression models. In this paper, we address this trade-off by comparing the predictive performance of random forests and logit regressions to analyze labor market vulnerabilities during the COVID-19 pandemic, and a global surrogate model to enhance our understanding of the complex dynamics. Our study shows that, especially in the presence of non-linearities and feature interactions, random forests outperform regressions both in predictive accuracy and interpretability, yielding policy-relevant insights on vulnerable groups affected by labor market disruptions |
| Keywords: | Machine learning, interpretability, labor market, random forests |
| JEL: | C45 C53 C25 J08 I18 C83 J21 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:diw:diwsop:diw_sp1230 |
| By: | Junjie Guo |
| Abstract: | Portfolio allocation via stock price prediction is inherently difficult due to the notoriously low signal-to-noise ratio of stock time series. This paper proposes a method by integrating wavelet transform convolution and channel attention with LSTM to implement stock price prediction based portfolio allocation. Stock time series data first are processed by wavelet transform convolution to reduce the noise. Processed features are then reconstructed by channel attention. LSTM is utilized to predict the stock price using the final processed features. We construct a portfolio consists of four stocks with trading signals predicted by model. Experiments are conducted by evaluating the return, Sharpe ratio and max drawdown performance. The results indicate that our method achieves robust performance even during period of post-pandemic downward market. |
| Date: | 2025–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2507.01973 |
| By: | Sebastian Fossati (University of Alberta); Xiao Li (University of Alberta) |
| Abstract: | We model the conditional distribution of future exchange rate returns for nine currencies as a function of real-time financial conditions. We show that the lower and upper quantiles of the exchange rate return distribution exhibit significant in-sample co-movement with financial conditions. Similarly, the conditional moments of the out-of-sample forecast display time-varying patterns, with the variance and kurtosis showing the most pronounced changes during and after the 2008-09 financial crisis. Deteriorating financial conditions are associated with an increase in volatility, particularly for commodity currencies. Overall, we conclude that financial conditions capture tail dependencies in exchange rate returns and contain valuable information for out-of-sample prediction. |
| Keywords: | exchange rates; financial conditions; NFCI; density forecasts |
| JEL: | C22 F31 G17 |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:ris:albaec:021546 |
| By: | Heikkinen, Joni; Heimonen, Kari |
| Abstract: | This paper examines online search activity's ability to capture consumers' intentions and enhance short-term forecasting of key economic outcomes. Economic decisions such as consumption and investment are typically preceded by intentions, which, while difficult to observe directly, often manifest as online information-seeking behavior. Using a large, high-frequency dataset of search activity, we nowcast U.S. consumer confidence and private consumption, finding that legal and governmental searches are associated with shifts in consumer confidence, while real estate and news-related searches add value to forecasts of private consumption. We then extend the analysis to GDP nowcasting for selected OECD economies, assessing the predictive performance of search-based indicators across different contexts. Overall, our findings highlight the value of digital attention data as behaviorally grounded signals of consumer intentions, offering a timely complement to traditional economic indicators. |
| Keywords: | nowcasting, online search behavior, consumer confidence, private consumption, GDP growth |
| JEL: | E30 E32 E37 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:bofrdp:325480 |
| By: | Alfred Backhouse; Kang Li; Jakob Foerster; Anisoara Calinescu; Stefan Zohren |
| Abstract: | Simulating limit order books (LOBs) has important applications across forecasting and backtesting for financial market data. However, deep generative models struggle in this context due to the high noise and complexity of the data. Previous work uses autoregressive models, although these experience error accumulation over longer-time sequences. We introduce a novel approach, converting LOB data into a structured image format, and applying diffusion models with inpainting to generate future LOB states. This method leverages spatio-temporal inductive biases in the order book and enables parallel generation of long sequences overcoming issues with error accumulation. We also publicly contribute to LOB-Bench, the industry benchmark for LOB generative models, to allow fair comparison between models using Level-2 and Level-3 order book data (with or without message level data respectively). We show that our model achieves state-of-the-art performance on LOB-Bench, despite using lower fidelity data as input. We also show that our method prioritises coherent global structures over local, high-fidelity details, providing significant improvements over existing methods on certain metrics. Overall, our method lays a strong foundation for future research into generative diffusion approaches to LOB modelling. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.05107 |
| By: | Anthoulla Phella; Vasco J. Gabriel; Luis F. Martins |
| Abstract: | In this paper, using the Bayesian VAR framework suggested by Chan et al. (2025), we produce conditional temperature forecasts up until 2050, by exploiting both equality and inequality constraints on climate drivers like carbon dioxide or methane emissions. Engaging in a counterfactual scenario analysis by imposing a Shared Socioeconomic Pathways (SSPs) scenario of "business as-usual", with no mitigation and high emissions, we observe that conditional and unconditional forecasts would follow a similar path. Instead, if a high mitigation with low emissions scenario were to be followed, the conditional temperature paths would remain below the unconditional trajectory after 2040, i.e. temperatures increases can potentially slow down in a meaningful way, but the lags for changes in emissions to have an effect are quite substantial. The latter should be taken into account greatly when designing response policies to climate change. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.09384 |
| By: | Farkas, Hannah; Linsenmeier, Manuel; Talevi, Marta; Avner, Paolo; Jafino, Bramka Arga; Sidibe, Moussa |
| Abstract: | This study systematically reviews the literature that quantifies the economic benefits of weather observations and forecasts in four weather-dependent economic sectors: agriculture, energy, transport, and disaster-risk management. The review covers 175 peer-reviewed journal articles and 15 policy reports. Findings show that the literature is concentrated in high-income countries and most studies use theoretical models, followed by observational and then experimental research designs. Forecast horizons studied, meteorological variables and services, and monetization techniques vary markedly by sector. Estimated benefits even within specific subsectors span several orders of magnitude and broad uncertainty ranges. An econometric meta-analysis suggests that theoretical studies and studies in richer countries tend to report significantly larger values. Barriers that hinder value realization are identified on both the provider and user sides, with inadequate relevance, weak dissemination, and limited ability to act recurring across sectors. Policy reports rely heavily on back-of-the-envelope or recursive benefit-transfer estimates, rather than on the methods and results of the peer-reviewed literature, revealing a science-to-policy gap. These findings suggest substantial socioeconomic potential of hydrometeorological services around the world, but also knowledge gaps that require more valuation studies focusing on low- and middle-income countries, addres sing provider- and user-side barriers and employing rigorous empirical valuation methods to complement and validate theoretical models. |
| Date: | 2025–09–10 |
| URL: | https://d.repec.org/n?u=RePEc:wbk:wbrwps:11213 |
| By: | Ross Koval; Nicholas Andrews; Xifeng Yan |
| Abstract: | Financial news plays a critical role in the information diffusion process in financial markets and is a known driver of stock prices. However, the information in each news article is not necessarily self-contained, often requiring a broader understanding of the historical news coverage for accurate interpretation. Further, identifying and incorporating the most relevant contextual information presents significant challenges. In this work, we explore the value of historical context in the ability of large language models to understand the market impact of financial news. We find that historical context provides a consistent and significant improvement in performance across methods and time horizons. To this end, we propose an efficient and effective contextualization method that uses a large LM to process the main article, while a small LM encodes the historical context into concise summary embeddings that are then aligned with the large model's representation space. We explore the behavior of the model through multiple qualitative and quantitative interpretability tests and reveal insights into the value of contextualization. Finally, we demonstrate that the value of historical context in model predictions has real-world applications, translating to substantial improvements in simulated investment performance. |
| Date: | 2025–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2509.12519 |