|
on Forecasting |
| By: | Bouillot, Roland (Maastricht University); Candelon, Bertrand (Université catholique de Louvain, LIDAM/LFIN, Belgium); Kool, Clemens (Maastricht University) |
| Abstract: | Accurate forecasting constitutes a central objective for policymakers. This paper examines the application of advanced machine-learning techniques to predict the 10-year sovereign bond spreads vis-à-vis the German bund, employing a novel high-dimensional dataset covering 10 European countries over the period 2007−2025. An exhaustive comparison of predictive performance, both in-sample and out-of-sample, demonstrates that XGBoost delivers the highest degree of accuracy. Building on these forecasts, we construct fragmentation matrices that capture the extent of asymmetry across Euro area sovereign bond markets. Prior to the COVID-19 crisis, results confirm the well-documented clustering between core and peripheral countries. However, since 2021 this segmentation appears to have weakened, as French and Belgian spreads exhibit a synchronous trajectory. Thesefindingscontribute totheliterature on financialintegrationand fragmentation within the Euro area, offering new insights into the evolving dynamics of sovereign bond markets. |
| Keywords: | Machine learning ; Financial fragmentation risk ; XGBoost ; Sovereign spreads |
| Date: | 2025–11–30 |
| URL: | https://d.repec.org/n?u=RePEc:ajf:louvlf:2025004 |
| By: | Yilin Xiao; Jamie L. Cross |
| Abstract: | We propose a new class of Regularized Random Subspace Regressions (RRSRs) that combine the variance reduction benefits of regularized estimators with the non-linearities of random subspace ensembles. The approach introduces regularization in the selection of predictor subspaces, coefficient estimation within each subspace, or in both, yielding a flexible family of models that nest both RSR and standard penalized regressions as special cases. Using the FRED-MD database as a large predictor space, we show that RRSRs consistently outperform traditional RSR and several widely used econometric and machine learning benchmarks when forecasting four key macroeconomic indicators: inflation, output, unemployment, and the federal funds rate. The most systematic gains arise from the double-regularized specification, underscoring the value of applying shrinkage jointly to subspace selection and coefficient estimation. |
| Keywords: | big data, forecasting, machine learning, model averaging, random subspace, regularization |
| JEL: | C22 C53 C55 E37 |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:een:camaaa:2026-13 |
| By: | Kumar, Labesh |
| Abstract: | This paper compares eight widely used methods for estimating the output gap, ranging from simple deterministic trends to state space models, using both revised and real time U.S. quarterly data from 1980 onward. The resulting measures differ heavily across approaches. Average gap estimates vary by nearly four percentage points, volatility differs by an order of magnitude, and correlations across methods span from strongly positive to negative. Stability across data vintages also varies substantially. Hamilton type filters show relatively strong agreement between real time and final estimates, while simpler trend based methods are considerably less stable. These differences matter for empirical inference. The choice of output gap measure has important implications for Phillips curve estimates and for forecasting performance. Beveridge Nelson decompositions display strong predictive power for inflation when estimated using revised data but perform less well in real time, whereas refined Beveridge Nelson and modified Hamilton filters deliver more consistent results across vintages. Time varying analysis shows that the relationship between economic slack and inflation strengthens during periods of macroeconomic stress, including the early 1990s recession, the global financial crisis, and the post-pandemic period, rather than declining monotonically. For output growth forecasting, HP filter gaps reduce forecast errors using revised data, while unobserved components models perform best in real time. Although Beveridge Nelson based measures are informative for inflation, they tend to worsen growth forecasts. Combining forecasts across gap measures, particularly using Bates Granger weights, yields more reliable performance by offsetting weaknesses of individual methods. Overall, the findings highlight that methodological uncertainty in measuring slack translates directly into policy uncertainty, cautioning against exclusive reliance on any single output gap estimate. |
| Keywords: | Output gap, trend-cycle decomposition, real-time data, Phillips curve, forecast combination |
| JEL: | C52 E32 E37 |
| Date: | 2026–01–08 |
| URL: | https://d.repec.org/n?u=RePEc:pra:mprapa:127829 |
| By: | Mijung Choi (Korea Institute for Industrial Economics and Trade); Sung Keun Park (Korea Institute for Industrial Economics and Trade) |
| Abstract: | The primary purpose of economic forecasting is to support rational decision-making under uncertainty and to provide information for the efficient allocation of limited resources. Mid-term industrial forecasts are essential for formulating national economic strategies in response to structural changes, and model-based approaches must therefore specify variables and functional forms in ways that reflect these dynamics.<p> As demographic shifts, advances in artificial intelligence, and global fragmentation reshape long-run economic fundamentals, forecasting models need to flexibly accommodate evolving relationships among variables while maintaining parsimony to mitigate the growing uncertainty of exogenous inputs.<p> This study proposes time-varying coefficient models and their extension, functional-coefficient panel models, as alternatives that satisfy these requirements. Applying these models to long-term export demand forecasts for South Korea’s materials industries shows that income elasticities have shifted significantly over time and that models that account for structural change achieve superior predictive performance compared to traditional fixed-coefficient or short-term time-series models. |
| Keywords: | exports; export demand; materials industry; industrial forecasting; long-term forecasting; economic forecasting; structural transformation; time-varying parameters; demand modeling; South Korea |
| JEL: | F14 F21 F17 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:ris:kieter:022207 |
| By: | Rajat Masiwal; Colin Aitken; Adam Marchakitus; Mayank Gupta; Katherine Kowal; Hamid A. Pahlavan; Tyler Yang; Y. Qiang Sun; Michael Kremer; Amir Jina; William R. Boos; Pedram Hassanzadeh |
| Abstract: | Artificial intelligence weather prediction (AIWP) models now often outperform traditional physics-based models on common metrics while requiring orders-of-magnitude less computing resources and time. Open-access AIWP models thus hold promise as transformational tools for helping low- and middle-income populations make decisions in the face of high-impact weather shocks. Yet, current approaches to evaluating AIWP models focus mainly on aggregated meteorological metrics without considering local stakeholders' needs in decision-oriented, operational frameworks. Here, we introduce such a framework that connects meteorology, AI, and social sciences. As an example, we apply it to the 150-year-old problem of Indian monsoon forecasting, focusing on benefits to rain-fed agriculture, which is highly susceptible to climate change. AIWP models skillfully predict an agriculturally relevant onset index at regional scales weeks in advance when evaluated out-of-sample using deterministic and probabilistic metrics. This framework informed a government-led effort in 2025 to send 38 million Indian farmers AI-based monsoon onset forecasts, which captured an unusual weeks-long pause in monsoon progression. This decision-oriented benchmarking framework provides a key component of a blueprint for harnessing the power of AIWP models to help large vulnerable populations adapt to weather shocks in the face of climate variability and change. |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2602.03767 |
| By: | Müller, Sebastian; Pugachyov, Nikolay; Weigert, Florian |
| Abstract: | We introduce a simple yet powerful method for enhancing mutual fund performance prediction by combining individual predictors into a composite predictor. This composite approach integrates information from 19 well-established return-based and portfolio holdings-based predictors from the literature. It effectively identifies top decile funds that outperform bottom decile funds by a risk-adjusted 4.56% per annum. Furthermore, it achieves statistically significant outperformance for long-only fund investments against the average active and passive fund. Both return-based predictors (e.g., fund alpha and the t-statistic of alpha) and holdings-based predictors (e.g., skill index and active weight) contribute equally to the composite predictor's success. |
| Keywords: | Mutual funds, performance prediction, composite predictor |
| JEL: | G11 G12 G20 G23 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:cfrwps:336774 |
| By: | Junyu Chen; Tom Boot; Lingwei Kong; Weining Wang |
| Abstract: | Conditional Value-at-Risk (CoVaR) quantifies systemic financial risk by measuring the loss quantile of one asset, conditional on another asset experiencing distress. We develop a Transformer-based methodology that integrates financial news articles directly with market data to improve CoVaR estimates. Unlike approaches that use predefined sentiment scores, our method incorporates raw text embeddings generated by a large language model (LLM). We prove explicit error bounds for our Transformer CoVaR estimator, showing that accurate CoVaR learning is possible even with small datasets. Using U.S. market returns and Reuters news items from 2006--2013, our out-of-sample results show that textual information impacts the CoVaR forecasts. With better predictive performance, we identify a pronounced negative dip during market stress periods across several equity assets when comparing the Transformer-based CoVaR to both the CoVaR without text and the CoVaR using traditional sentiment measures. Our results show that textual data can be used to effectively model systemic risk without requiring prohibitively large data sets. |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2602.12490 |
| By: | Marc Schmitt |
| Abstract: | Risk forecasts drive trading constraints and capital allocation, yet losses are nonstationary and regime-dependent. This paper studies sequential one-sided VaR control via conformal calibration. I propose regime-weighted conformal risk control (RWC), which calibrates a safety buffer from past forecast errors using exponential time decay and regime-similarity weights from regime features. RWC is model-agnostic and wraps any conditional quantile forecaster to target a desired exceedance rate. Finite-sample coverage is established under weighted exchangeability, and approximation bounds are derived under smoothly drifting regimes. On the CRSP U.S.\ equity portfolio, time-weighted conformal calibration is a strong default under drift, while regime weighting can improve regime-conditional stability in some settings with modest conservativeness changes. |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2602.03903 |
| By: | Aijie Shu; Wenbin Wu; Gbenga Ibikunle; Fengxiang He |
| Abstract: | Credit exposure in Decentralized Finance (DeFi) is often implicit and token-mediated, creating a dense web of inter-protocol dependencies. Thus, a shock to one token may result in significant and uncontrolled contagion effects. As the DeFi ecosystem becomes increasingly linked with traditional financial infrastructure through instruments, such as stablecoins, the risk posed by this dynamic demands more powerful quantification tools. We introduce DeXposure-FM, the first time-series, graph foundation model for measuring and forecasting inter-protocol credit exposure on DeFi networks, to the best of our knowledge. Employing a graph-tabular encoder, with pre-trained weight initialization, and multiple task-specific heads, DeXposure-FM is trained on the DeXposure dataset that has 43.7 million data entries, across 4, 300+ protocols on 602 blockchains, covering 24, 300+ unique tokens. The training is operationalized for credit-exposure forecasting, predicting the joint dynamics of (1) protocol-level flows, and (2) the topology and weights of credit-exposure links. The DeXposure-FM is empirically validated on two machine learning benchmarks; it consistently outperforms the state-of-the-art approaches, including a graph foundation model and temporal graph neural networks. DeXposure-FM further produces financial economics tools that support macroprudential monitoring and scenario-based DeFi stress testing, by enabling protocol-level systemic-importance scores, sector-level spillover and concentration measures via a forecast-then-measure pipeline. Empirical verification fully supports our financial economics tools. The model and code have been publicly available. Model: https://huggingface.co/EVIEHub/DeXposure-FM. Code: https://github.com/EVIEHub/DeXposure-FM. |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2602.03981 |
| By: | David Winkelmann; Theresa Elbracht; Jonas Brenker; Arnold Gerzen |
| Abstract: | Grocery retailers frequently apply price discounts to stimulate demand for expiring perishables. However, integrating these discounted sales into future demand forecasts presents a significant challenge. This study investigates the effectiveness of incorporating a fixed share of these sales as \textit{regular} demand into the forecast, as commonly applied in practice. We employ a two-step regression approach on data from a major European grocery retailer, covering over 1, 700 products across 676 stores. We reveal that forecasts underestimate actual demand for most SKUs when discounted sales occur. This residual uplift effect is significantly influenced by the number of sales at reduced prices. Our findings underscore the necessity for more precise approaches to integrate discounted sales into demand forecasts, thereby preventing excess inventory and the associated economic and environmental impacts of spoilage in the grocery sector. |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2602.04464 |
| By: | Layer, Kira; Gutmayer, Stephanie; Sandmeier, Thorben; Ringger, Jonas; Cermak, Jan; Fichtner, Wolf |
| Abstract: | Electricity generation as well as electricity demand are dependent on the weather and climate, and this dependency is expected to further increase in the future. Challenges in energy systems arising from this dependency can be studied using large-scale weather patterns (WPs). These WPs can help reveal the atmospheric drivers of the challenges, but there exist many different classifications of large-scale WPs. Although WPs are widely used in energy-related studies, to our knowledge, no systematic review has yet evaluated the applicability of weather pattern classifications to analyzing extreme events and variability in energy systems. In this study, we aim to fill this gap by reviewing and combining literature dealing with both WP classifications and weather-induced challenges in energy systems. A total of 69 studies are included, which use different classification methods to study weather-induced challenges on energy systems. Overall, most challenges to the energy system arise during blocking weather patterns. Furthermore, we find that stable large-scale WPs allow for better forecasts of wind power generation if combined with other predictors. This review reveals research gaps underscoring the need to consider the whole energy system, including demand and the electricity grid, not only the generation of wind power and photovoltaics. |
| Keywords: | weather patterns, weather regimes, high residual load events, energy drought, weather variability, forecasting, energy system |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:kitiip:336809 |
| By: | Maarten Van Besien (-) |
| Abstract: | Automated valuation models (AVMs) are widely used for large-scale residential rent appraisal, yet standard models do not provide predictive uncertainty measures with guaranteed out-ofsample coverage at prespecified nominal levels, creating risks for institutional decision-making in valuation, risk management, and policy design. Using a transaction-level dataset covering the Flemish rental market in Belgium, we study AVM performance and uncertainty quantification in a large-scale, heterogeneous, and feature-poor setting, where only location, property type, energy performance, number of bedrooms, and rent prices are observed. We show that industry-standard point-prediction accuracy can be achieved by exploiting non-linear spatial structure using coarse geospatial units such as boroughs. For uncertainty quantification, we compare ensemble quantile regression and Inductive Conformal Prediction (ICP). While both improve empirical coverage, ICP is preferred as it guarantees finite-sample marginal coverage without distributional assumptions at substantially lower computational cost. Conditioning ICP calibration on bedroom count (Mondrian ICP) yields the largest efficiency gains, reducing 95% coverage prediction interval width by up to 5.3% relative to absolute residual split conformal prediction. Overall, our results demonstrate that valuation uncertainty can be materially reduced in large-scale, feature-poor housing data with minimal additional modeling complexity. |
| Keywords: | Conformal Prediction, Automated Rent Valuation, Rental Uncertainty Quantification |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:rug:rugwps:26/1136 |
| By: | Cimadomo, Jacopo; Giannone, Domenico; Lenza, Michele; Monti, Francesca; Sokol, Andrej |
| Abstract: | We design a Bayesian Mixed-Frequency vector autoregression (VAR) model for fiscal monitoring, i.e., to nowcast the government deficit-to-GDP ratio in real time and provide a narrative for its dynamics. The model incorporates both monthly cash and quarterly accrual fiscal indicators, together with other high-frequency macroeconomic and financial variables, as well as real GDP and the GDP deflator. Our model produces timely monthly density nowcasts of the annual deficit ratio, while governments and official institutions generally only publish their point predictions bi-annually. Based on a database of real-time vintages of macroeconomic, financial and fiscal variables for Italy, we show that the nowcasts of the annual deficit to GDP ratio of our model are similarly or more accurate than those of the European Commission, depending on the month in which the nowcast is produced. Our scenario analysis compares the dynamics of the deficit ratio associated with a monetary and a typical recession, finding a more muted response in the latter case. JEL Classification: C11, E52, E62, E63, H68 |
| Keywords: | cash data, government deficit, mixed-frequency, monetary-fiscal interactions, monetary policy shock, nowcasting |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:ecb:ecbwps:20263186 |