|
on Forecasting |
| By: | Sergio A. Lago Alves; Waldyr Dutra Areosa; Carlos Viana de Carvalho |
| Abstract: | Professional inflation forecasts contain valuable information but exhibit information frictions. We extract improved forecasts by explicitly modeling these frictions using the US Survey of Professional Forecasters data, and find that forecast rigidity increases systematically with horizon, rising from near zero for backcasts to 0.81 beyond two quarters. In pseudo-real-time tests, our Resetting Nowcasts reduce mean squared errors by 50 percent relative to SPF averages. We derive a novel theoretical criterion showing that improved forecasts dominate when disagreement lies within an optimal interval determined by simple sufficient statistics, easily computable from any survey microdata. The criterion determines in advance the horizons where improved forecasts should dominate, without estimating friction parameters. This generalizes easily to other surveys and variables, providing a tractable method for identifying which forecast horizons offer the greatest potential for improvement. |
| Keywords: | Models and tools, Econometric, statistical and computational methods, Monetary policy, Inflation dynamics and pressures, Real economy and forecasting |
| JEL: | C11 C53 E31 E37 |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:bca:bocawp:26-11 |
| By: | Baumeister, Christiane; Huber, Florian; Lee, Thomas K.; Ravazzolo, Francesco |
| Abstract: | This paper provides a comprehensive analysis of the forecastability of the real price of natural gas in the United States at the monthly frequency considering a universe of models that differ in their complexity and economic content. Our key finding is that considerable reductions in mean-squared prediction error relative to a random walk benchmark can be achieved in real time for forecast horizons of up to two years. A particularly promising model is a six-variable Bayesian vector autoregressive model that includes the fundamental determinants of the supply and demand for natural gas. To capture real-time data constraints of these and other predictor variables, we assemble a rich database of historical vintages from multiple sources. We also compare our model-based forecasts to readily available model-free forecasts provided by experts and futures markets. Given that no single forecasting method dominates all others, we explore the usefulness of pooling forecasts and find that combining forecasts from individual models selected in real time based on their most recent performance delivers the most accurate forecasts. |
| JEL: | C11 C32 C52 Q41 Q47 |
| Date: | 2024–11 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19669 |
| By: | Alice Treesa M; Dr. Arpita Choudhary (Assistant Professor, Madras School of Economics, Chennai, India.) |
| Abstract: | Reliable energy consumption forecasting in the short term is essential for improving building operations efficiency and creating sustainable energy consumption plans. The authors of this study evaluate the forecasting performance of machine learning and deep learning methods which use climate data and time data to predict energy usage at hourly intervals. The study used Linear Regression, Decision Trees, Random Forest, XGBoost and Long Short-Term Memory as comparison methods to assess performance in the same context. The study demonstrated that energy consumption forecasting accuracy depends more on selected features than on the model's complexity. The study found that LSTM model learning capacity remained stable while Random Forest model performance showed superior results in dealing with non-linear features that had temporal attributes. |
| Keywords: | Energy Consumption Prediction, Machine Learning, Ensemble Models, LSTM Model, Feature Engineering, Sustainable Energy Managementsemantics, Neural architectures |
| JEL: | Q47 C53 C45 C38 L94 Q41 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:mad:wpaper:2026-301 |
| By: | Joel P. Flynn (Yale University); Maksim Meinert (Yale University); Karthik A. Sastry (Princeton University) |
| Abstract: | With uncertainty about persistence, we show that forecasts necessarily become more persistent and over-react at long horizons. For these reasons, correctly specified and Bayesian forecasts may under-react at short horizons and over-react at long horizons. These results provide a unified explanation for several asset pricing and forecasting puzzles, including: (i) the excess responsiveness of long-horizon rates to short rates, (ii) the dominance of apparent term premia for long-term rates, (iii) the ex post predictability of bond yields, (iv) the excess volatility of long-horizon forward prices, (v) the excess persistence of long-horizon forecasts, and (vi) the over-reaction of long-horizon forecasts. |
| Date: | 2026–06–26 |
| URL: | https://d.repec.org/n?u=RePEc:cwl:cwldpp:2544 |
| By: | Peter Dixon |
| Abstract: | Macroeconomic forecasts provide a consistent framework for analysing the economy and informing policy decisions. However, their usefulness ultimately depends on their predictive accuracy. This paper assesses NIESR's macroeconomic forecasts, with a particular focus on GDP growth and CPI inflation, produced between 1992 and 2023 across horizons of up to ten years. We document forecast errors, dispersion and formal tests of bias, then evaluate NIESR's forecasts against a hierarchy of benchmarks ‐ including a random walk, autoregressive models and a Bayesian VAR ‐ using Diebold‐Mariano tests. GDP growth forecasts show a significant positive (over‐-prediction) bias that strengthens with horizon; CPI inflation forecasts show no significant bias. NIESR outperforms the benchmark models on inflation at nearly all horizons, but for GDP growth its advantage is concentrated in the short‐to‐medium term and reversed beyond around five years. This coincides closely with the point at which over‐prediction bias becomes significant, suggesting a correctable bias problem rather than a wholesale loss of forecasting content. Strikingly, this long‐horizon weakness is not simply shock‐driven: excluding major crisis years collapses the BVAR's advantage over NIESR forecasts but leaves the AR(p)'s advantage intact. The results suggest that the benefits of structural modelling depend on both the variable being forecast and the forecast horizon. They also indicate that NIESR's longhorizon weakness may reflect systematic optimism rather than a loss of forecasting information, suggesting that horizon‐specific bias correction could provide a promising, empirically testable avenue for improving long‐horizon GDP forecast accuracy. |
| Keywords: | Macroeconomic forecasting, forecasting performance, forecast errors |
| JEL: | C53 E17 |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:nsr:niesrd:582 |
| By: | Abdukakhkhor Abdurakhmonov (Central Bank of Uzbekistan) |
| Abstract: | This paper provides the first systematic assessment of machine learning methods for macroeconomic forecasting in Uzbekistan. Using a comprehensive dataset of more than 170 indicators, we forecast CPI inflation and GDP growth with nine machine learning models and compare them against three traditional benchmarks (ARIMA, VAR, and BVAR). For both targets, the relative performance of machine learning improves as the forecast horizon increases. For inflation, machine learning provides clear and growing gains as the horizon increases, and a simple equal-weighted ensemble of the machine learning models is the most accurate approach overall, achieving the lowest forecast error at nearly every horizon. For GDP growth, by contrast, the traditional benchmarks (ARIMA in particular) remain the most accurate across most horizons, although regularized linear and dimension-reduction machine learning methods are competitive at short horizons. Tree-based models struggle to forecast GDP when growth exceeds the range observed during training because they cannot extrapolate beyond the training data. This limitation is particularly relevant in Uzbekistan's rapidly changing economy, where rapid economic growth in 2024-2025 pushed the level of GDP beyond the range observed in the training sample. We show that forecasting stationary transformations of the target largely removes this weakness. Overall, the findings suggest that machine learning is best used to complement rather than replace the existing forecasting toolkit. It improves the accuracy of medium-term inflation forecasts, whereas traditional models remain more accurate for forecasting GDP. |
| Keywords: | Machine Learning; Macroeconomic Forecasting; Ination; GDP Growth; Model Evaluation and Selection; Uzbekistan |
| JEL: | C22 C45 C53 E31 E37 E52 |
| Date: | 2026–08–03 |
| URL: | https://d.repec.org/n?u=RePEc:gii:giihei:heidwp19-2026 |
| By: | Wagenvoort, Benjamin (Institute for New Economic Thinking at the Oxford Martin School, University of Oxford (INET Oxford)); Lafond, François; Dyer, Joel; Farmer, J. Doyne |
| Abstract: | Many technologies grow along S-curves: diffusion is slow, then rapid, then levels off. Forecasting this growth is vital for renewable energy, AI and other technology transitions, but it has been unclear whether technologies follow a single universal process, and past forecasts have proved unreliable. We assemble a database of 120 mature technologies, from canals to mobile phones, and show their S-curve shapes are remarkably universal. Using Bayesian methods and extensive out-of-sample backtesting, we show that a Bertalanffy-Richards process does a good job of fitting the data and its forecasting outperforms popular alternatives. Its point forecasts are typically accurate to within a factor of two, even from a 5% diffusion origin and decades ahead. This gives a validated method to forecast any technology that follows an S-curve, with known accuracy. Our forecasts for solar PV and wind indicate that by 2050 they will supply approximately 18–290 and 4–17 PWh globally each year (90% prediction intervals). Our median estimate for solar in 2050 is about 85 PWh, similar to all useful energy consumed today. Even the most aggressive IPCC AR6, IEA and NGFS scenarios are too pessimistic about solar, implying that ambitious climate targets will likely be met faster than widely believed. |
| Keywords: | Technology diffusion, Bayesian forecasting, S-curves, Energy transition |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:amz:wpaper:2026-19 |
| By: | Consolo, Agostino; Foroni, Claudia; Lissona, Claudio; Schroeder, Christofer |
| Abstract: | We analyse whether textual information extracted from firms’ earnings calls can improve forecasts of the euro area job vacancy rate. Using transcripts from euro area headquartered firms, we construct a monthly indicator of labour demand based on keywords related to labour market pressures and include it into a mixed frequency Bayesian VAR alongside standard hard and soft indicators. A pseudo–real-time evaluation shows that earnings calls provide timely and valuable signals for tracking vacancy dynamics. Among soft indicators, factors limiting production deliver the largest forecasting gains, while real labour-market indicators such as unemployment add little once qualitative signals are included. Forecast improvements are largely driven by information from the manufacturing sector, whose signals prove substantially more informative than those from services, especially when paired with earnings calls. Taken together, our results highlight the usefulness of high-frequency text-based information for improving short-term labour-demand forecasts in the euro area. JEL Classification: C53, E24, E27 |
| Keywords: | earnings calls, job vacancy rate, mixed-frequency, nowcasting, sectoral heterogeneity |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:ecb:ecbwps:20263269 |
| By: | Stephen Millard |
| Abstract: | In this paper, I examine how scenario analysis can improve the communication of risks in macroeconomic forecasting. Building on the Bernanke review, I evaluate four uses of scenarios: assessing policy interventions, illustrating specific forecast risks, addressing model uncertainty, and decomposing historical forecast errors. I argue that scenarios are most effective when used to convey distinct, narrative‐driven risks around a central forecast, offering clearer insights than fan charts alone in this case. However, I also argue that fan charts are useful for communicating general uncertainty. Overall, I find that narrative‐based scenario analysis is a powerful tool for communicating macroeconomic risks. |
| Keywords: | Scenario analysis, Macroeconomic forecasting, Forecast uncertainty, Risk communication, Narrative-based analysis |
| JEL: | C53 C54 E17 |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:nsr:niesrd:584 |
| By: | Araujo, Douglas; Bokan, Nikola; Comazzi, Fabio; Lenza, Michele |
| Abstract: | Word embeddings are vectors of real numbers associated with words, designed to capture semantic and syntactic similarity between the words in a corpus of text. We estimate the word embeddings of the European Central Bank's introductory statements at monetary policy press conferences by using a simple natural language processing model (Word2Vec), only based on the information and model parameters available as of each press conference. We show that a measure based on such embeddings contributes to improve core inflation forecasts multiple quarters ahead. Other common textual analysis techniques, such as dictionary-based metrics or sentiment metrics do not obtain the same results. The information contained in the embeddings remains valuable for out-of-sample forecasting even after controlling for the central bank inflation forecasts, which are an important input for the introductory statements. |
| Keywords: | Inflation |
| JEL: | E31 E37 E58 |
| Date: | 2024–12 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19784 |
| By: | Weiye Xi; Ciamac C. Moallemi; Mallesh Pai; Shouqiao Want |
| Abstract: | Forward-looking volatility forecasts are central inputs to derivatives pricing, market making, risk management, and volatility-linked trading strategies, with ARCH and GARCH models serving as the canonical workhorses. Such models are natural in standard asset markets, where prices are positive-valued stochastic processes and volatility is typically inferred from return dynamics. Prediction markets have a different structure: prices are bounded probabilities, payoffs are binary, and contracts resolve at known deadlines. We develop and estimate a volatility model tailored to binary prediction markets. The model combines two economic mechanisms: a Wright-Fisher deadline-resolution component, capturing how remaining binary uncertainty is forced to resolve over time, and a Glosten-Milgrom order-flow component, capturing volatility from informed trading as reflected in spreads and volume. Using a large panel of Kalshi contracts, we show that these structural variables carry substantial forecasting power. Plain ARCH/GARCH benchmarks are dominated by structural specifications; combining the structural model with residual GARCH dynamics gives the best overall forecasts. The model also provides an interpretable measurement framework: volatility is highest near fifty-fifty prices, rises near resolution, and varies across categories with the timing and discreteness of information arrival. Economics contracts are closer to smooth deadline-resolution dynamics, while sports contracts exhibit more event-concentrated, jump-like behavior. Across major categories, category-specific fitting does not systematically improve out-of-sample performance, suggesting that the structural specification transfers beyond the pooled headline result. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2607.08199 |
| By: | Mackereth, Ivy; Bora, Siddhartha |
| Abstract: | The World Agricultural Supply and Demand Estimates (WASDE) reports play a central role in shaping expectations for major agricultural commodities, yet their fixed-event structure limits their use in forward-looking policy analysis. This study develops 12-month-ahead fixed-horizon price forecasts for corn, soybeans, and wheat by applying an optimal-weighting approach and a standard ad hoc aggregation benchmark to overlapping current- and next-year WASDE releases. We evaluate the constructed forecasts against realized prices and futures market expectations using Diebold-Mariano tests, Mincer-Zarnowitz efficiency regressions, and forecast encompassing tests. Optimal weighting improves point accuracy for all three commodities and forecast efficiency for corn. Accuracy gains are most pronounced for soybeans, moderate for corn, and modest but consistent for wheat. We further show that the fixed-horizon transformation eliminates statistically significant seasonal patterns in soybean forecast errors that persist under both ad hoc and futures-based specifications, suggesting that a meaningful share of the forecast error seasonality documented in prior literature reflects the fixed event design rather than the underlying information environment. These results demonstrate that reframing WASDE forecasts at a constant horizon improves accuracy, efficiency, and interpretability, with direct applications to farm budgeting, forward pricing, and revenue planning. |
| Keywords: | Demand and Price Analysis, Research Methods/Statistical Methods |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:ags:asea26:404812 |
| By: | Tamkin Nuriyev (Central Bank of the Republic of Azerbaijan); Aygun Garayeva (Central Bank of the Republic of Azerbaijan); Gulzar Tahirova (Central Bank of the Republic of Azerbaijan) |
| Abstract: | Using 800, 000 transaction-level customs records from January 2018 to February 2026, the paper constructs a trade-weighted Imported Food Price Index (IFPI), covering 34 items from the consumer basket with significant import dependence. The index is developed using the Fisher ideal methodology to provide a timely measure of external food price pressures. The results indicate that the IFPI leads official food Consumer Price Index (CPI) by approximately two months, with a maximum correlation of 0.81, highlighting its potential usefulness as an early indicator of domestic food inflation. Building on this, the paper develops a forecasting framework for the IFPI by combining non-parametric Binary Segmentation and Hidden Markov Models with a regularized machine-learning ensemble. The model employs an ensemble approach that combines Histogram-based Gradient Boosting Regression Tree, Random Forest, and Extreme Gradient Boosting, alongside rigorous time-series crossvalidation. The optimized ensemble achieves a 58% out-of-sample R² relative to a random walk benchmark, vastly outperforming traditional linear Autoregressive Distributed Lag (ARDL) (13.60%) and Autoregressive Integrated Moving Average with Exogenous Variables (ARIMAX) (0.18%) baselines. The forecast results are intended to be incorporated into broader inflation forecasting models to improve short-term projections. |
| Keywords: | Import price index; Fisher Ideal index; Food price inflation; Machine learning forecasting; Hidden Markov models |
| JEL: | C43 C53 C55 E31 F14 |
| Date: | 2026–08–03 |
| URL: | https://d.repec.org/n?u=RePEc:gii:giihei:heidwp20-2026 |
| By: | Baley, Isaac; Turen, Javier |
| Abstract: | Professional forecasters adjust their inflation forecasts in a distinctly lumpy pattern, making infrequent but substantial revisions. Strategic concerns play a significant role---forecasters are more likely to adjust, and by larger amounts, when their forecasts deviate from the consensus. Using a fixed-event forecasting framework, we document the impact of lumpiness and consensus pressure on forecast adjustments. Our quantitative model, which integrates Bayesian belief updating with forecast revision costs and strategic concerns, not only replicates the observed lumpiness in survey data but also sheds light on forecasters' apparent overreactions to new information. This structured framework enables us to "cleanse" forecasts, isolating the underlying inflation beliefs that drive these forecasts. |
| Keywords: | Survey of professional forecasters; overreaction; Lumpy adjustment; Inflation expectations; Consensus; Strategic complementarity |
| JEL: | D80 D81 D83 D84 E20 E30 |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19824 |
| By: | Özer, Yeliz; del Barrio Castro, Tomás; Escribano, Álvaro; Sibbertsen, Philipp |
| Abstract: | Deep-time climate records contain deterministic orbital signals and persistent stochastic variation, but how these components jointly affect predictability across climate states remains unclear. We analyze the Cenozoic Global Reference benthic foraminifer oxygen and carbon isotope record spanning 67.1 million years. This very long period is divided into seven climate-state segments. For each segment, we estimate deterministic contemporaneous long-run components combining linear trends, eccentricity, obliquity, climatic precession, and identified harmonic frequencies. The remaining variation is modeled with a bivariate vector autoregressive forecasting framework conditioned on astronomical forcing. The selected deterministic and dynamic structures differ substantially across climate states. Obliquity is the most recurrent orbital predictor, whereas squared obliquity, eccentricity, climatic precession, and harmonic components contribute only in particular segments and differ between the two proxies. Forecast accuracy likewise varies across the record, although observed and predicted values agree closely in several segments. A projection for the next 100, 000 years provides a baseline implied by natural astronomical forcing and continued Icehouse dynamics. Overall, the results show that orbital responsiveness, proxy interactions, and statistical predictability are state dependent. Deep-time climate variability therefore cannot be represented by a single common combination of deterministic forcing and stochastic dynamics across the complete Cenozoic. |
| Keywords: | CENOGRID, Deep-Time Paleoclimate, Forecasting Climate Data. |
| JEL: | C22 C32 C53 |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:han:dpaper:dp-751 |
| By: | Yannick Hoga; Niklas V. Lehmann |
| Abstract: | We study the evaluation of forecasts regarding the timing and occurrence of uncertain future events, such as volcanic eruptions, the start of a war or the beginning of a recession. We show theoretically that a typical approach -- evaluating the forecasts after the event occurred -- incentivizes dishonest predictions if forecasters discount future rewards in favor of more immediate benefits. An empirical application to forecasting tournament data finds strong empirical evidence that forecasters adjust predictions in response to these incentives, implying that existing forecasts of such events are likely systematically overstating the probability of early occurrence. We conclude that rewarding such forecasts in an incentive-compatible way is inherently challenging. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2607.13759 |
| By: | Afees A. Salisu (Centre for Econometrics and Applied Research, Ibadan, Nigeria; Department of Economics, University of Pretoria, Private Bag X20, Hatfield 0028, South Africa); Ahamuefula E. Ogbonna (Centre for Econometrics and Applied Research, Ibadan, Nigeria); Rangan Gupta (Department of Economics, University of Pretoria, Private Bag X20, Hatfield 0028, South Africa); Elie Bouri (School of Business, Lebanese American University, Lebanon) |
| Abstract: | This paper employs the generalized autoregressive conditional heteroscedasticity-mixed data sampling (GARCH-MIDAS) framework to forecast monthly and daily stock return volatility in the United States (US), based on a quarterly news-based Price Conflict Index (PCI) that signals “bad macroeconomic news†. An analysis of historical monthly (1860-2023) and daily (1885-2023) data demonstrates that the GARCH-MIDAS model incorporating PCI outperforms both the benchmark GARCH-MIDAS model with realized volatility (GARCH-MIDAS-RV) and models with macroeconomic variables such as output growth, inflation, unemployment, and interest rates. Furthermore, the inclusion of the PCI in modeling stock return volatility provides higher utility gains compared to models that exclude it. These findings have important implications for both investors and policymakers. |
| Keywords: | Price Conflict, Stock Returns Volatility, Forecasting, GARCH-MIDAS |
| JEL: | C32 C53 E31 G12 |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:pre:wpaper:202620 |
| By: | Guo, Hongfei; Marín Díazaraque, Juan Miguel; Veiga, Helena |
| Abstract: | Adding flexibility to a multivariate volatility model can worsen covariance forecasts; we show when, and how to detect it. We decompose the multivariate QLIKE loss into trace, marginal-scale, and correlation log-determinant components; counterfactual block substitutions attribute gains or losses to either part. In Bayesian dynamic-correlation stochastic volatility, the diagnostic yields a sequence: diffuse dynamic correlations can underperform a constant-correlation baseline; a stabilizing prior repairs the correlation component; once stabilized, lagged realized-volatility inputs are the strongest remaining lever, with neural corrections competitive but not dominant. Under a rolling protocol, the stabilized realized-augmented family rivals realized-covariance benchmarks while retaining full predictive densities. |
| Keywords: | Covariance forecasting; Dynamic correlations; Forecast-object diagnostics; Neural networks; Prior regularization; Realized volatility; Stochastic volatility |
| JEL: | C11 C32 C53 C58 G17 |
| Date: | 2026–07–28 |
| URL: | https://d.repec.org/n?u=RePEc:cte:wsrepe:50561 |
| By: | Chahad, Mohammed; Mogliani, Matteo; Bańbura, Marta; Kulikov, Dmitry; Montes-Galdón, Carlos; Landau, Bettina; Meunier, Baptiste; Odendahl, Florens; Paredes, Joan; Sigwalt, Antoine; Theofilakou, Anastasia; Aristidou, Chryso; Pacella, Claudia; Rodrigues, Paulo; Roth, Markus |
| Abstract: | This paper introduces reduced-form macroeconometric tools, emphasising quantile regression models, to identify key risk drivers for the euro area economy and assess risks around the baseline ECB/Eurosystem staff macroeconomic projections for the euro area inflation and growth. The analysis uses a large number of risk factors, going beyond the usual financial factors, employing a sequential selection approach with robustness checks. To support the analysis a MATLAB toolbox (M@RX) was developed, incorporating several quantile regression-based model classes with a novel parametric tilting methodology and a copula approach for transforming predictive densities across frequencies. This paper contributes to the literature on the treatment of the COVID-era data in quantile regression models. Results indicate that the predictive content of risk factors is horizon, time and objective-dependent. For example, labour market indicators are particularly relevant for assessing upside inflation risks, but to a time-varying extent and with limited predictive power for downside risks. Conversely, uncertainty, money and credit indicators perform better for downside inflation risks. As regards risks to growth, the results confirm the established role of financial conditions, while also highlighting the relevance of monetary indicators, particularly for downside risks. Combined risk factor frameworks – with several different risk indicators – tend to systematically outperform single-factor specifications for density forecasting, due to complementarities across risk indicator groups. An empirical application highlights the policy relevance of these tools, as they provide timely signals and accurately track the direction of realised outcomes. Given the time-varying and state-dependent nature of their predictive performance, a regular performance assessment of the specifications is recommended to maintain reliability. JEL Classification: C22, C53, E27, E37 |
| Keywords: | density forecasts, forecasting, Macro-at-Risk, quantile regression, tail risks |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:ecb:ecbops:2026396 |
| 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: | 2026–01–30 |
| URL: | https://d.repec.org/n?u=RePEc:bri:uobdis:26/840 |
| By: | Stephen Millard |
| Abstract: | In this paper, I discuss some of the challenges associated with communicating uncertainty around forecasts, particularly important given the context of an uncertain world. I first assess how well NIESR's fan charts have captured the distribution of past forecast errors finding that they have some problems with this. I then examine different ways of constructing fan charts, seeing how different measures of uncertainty can be used to adjust the size of the fan in NIESR forecasts. I conclude that fan charts based on historical forecast errors remain a useful way of communicating the degree of uncertainty around a forecast and argue that the width of the fan can be adjusted based on survey measures of uncertainty to provide a realistic illustration of the potential magnitude of forecast errors. I also discuss how, in principle, the fan chart can be skewed to reflect the balance of risks but argue that there is no convincing way of calibrating such an adjustment. |
| Keywords: | Macroeconomic forecasting, Uncertainty, Fan charts |
| JEL: | C53 E17 |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:nsr:niesrd:583 |
| By: | Bartram, Söhnke; Chhaochharia, Vidhi; Kumar, Alok; Mo, Hongwei |
| Abstract: | We examine whether non-local analysts are able to improve the accuracy of their earnings forecasts by observing the forecasts of better-informed local analysts. We find that analysts issue more accurate forecasts for non-local firms in their coverage portfolios by observing the forecasts of local analysts. The improvement in accuracy is larger for non-local analysts at small brokerages and those with more complex portfolios. Further, analysts with superior local information learn more effectively from the locals in their non-local positions. Local analyst forecasts are particularly useful when (i) a firm is far away from the non-local analyst, (ii) the firm’s earnings are harder to predict, (iii) local analysts are more accurate, and (iv) nonlocal analyst forecasts are less accurate. At the aggregate level, learning from the locals improves consensus forecast accuracy and increases favorable career outcomes for non-local analysts. Collectively, these findings suggest that geography-based observational learning reduces informational asymmetry among analysts. |
| Keywords: | Forecast accuracy |
| JEL: | D83 G14 G24 M41 |
| Date: | 2024–11 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19670 |
| By: | Bartram, Söhnke; Grinblatt, Mark; Xu, Yan |
| Abstract: | The relative restrictiveness of a central bank’s supply of money predicts the raw and risk-adjusted returns of its currency—both next month and at least three years into the future. Archived data, known by currency traders at the time, estimates central bank restrictiveness as a scaling of the residual from out-of-sample panel regressions of M1 on macroeconomic variables tied to domestic and international transaction requirements. Carry’s ability to forecast currency returns is subsumed by the central bank restrictiveness signal, which also forecasts inflation. |
| Keywords: | Money supply |
| JEL: | F31 G12 G15 |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19881 |
| By: | Calvo, Richard; Pons, Vincent; Shapiro, Jesse |
| Abstract: | Many influential observers have forecast large partisan shifts in the US electorate based on demographic trends. Such forecasts are appealing because demographic trends are often predictable even over long horizons. We backtest demographic forecasts using data on US elections since 1952. We envision a forecaster who fits a model using data from a given election and uses that model, in tandem with a projection of demographic trends, to predict future elections. Even a forecaster with perfect knowledge of future demographic trends would have performed poorly over this period—worse even than one who simply guesses that each election will have a 50-50 partisan split. Enriching the set of demographics available does not change this conclusion. Slow demographic change, unstable group preferences, and strategic party responses all help to explain why demography has not been destiny in US politics. |
| JEL: | D72 J11 C53 P00 |
| Date: | 2024–10 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19564 |
| By: | Chernov, Mikhail; Elenev, Vadim; Song, Dongho |
| Abstract: | We propose a novel time-series econometric framework to forecast U.S. Presidential election outcomes in real time by combining polling data, economic fundamentals, and political prediction market prices. Our model estimates the joint dynamics of voter preferences across states. Applying our approach to the 2024 Presidential Election, we find a two-factor structure driving the vast majority of the variation in voter preferences. We identify electorally similar state clusters without relying on historical data or demographic models of voter behavior. Our simulations quantify the correlations between state-level election outcomes. Failing to take the correlations into account can bias the forecasted win probability for a given candidate by more than 10 percentage points. We find Pennsylvania to be the most pivotal state in the 2024 election. Our results provide insights for election observers, candidates, and traders. |
| JEL: | C32 C53 D72 P00 |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19836 |