|
on Forecasting |
| By: | Catherine Chen; Chen Gao; Jonathon Hazell; Lihua Lei; Chen Lian |
| Abstract: | Does microeconomic heterogeneity help to forecast aggregate inflation in a non-stationary environment? We develop a scan test for whether one forecast outperforms another, over an interval with unknown starting point and duration. To exploit any occasional forecasting power that the scan test detects, we design an adaptive machine learning pipeline. We encode the distribution of price changes into a high-dimensional vector, which we combine with a gradient boosted trees algorithm. We then combine this micro forecast with other benchmark forecasts, using an adaptive algorithm that makes use of the micro forecast only when it performs well. We apply the pipeline to UK microdata, with four main results. First, the micro forecast outperforms a univariate benchmark, but only in the volatile period after 2020. Second, the scan test detects periods of micro outperformance, so the micro forecast enters the combined forecast. Third, the combined forecast performs comparably to the univariate benchmark before 2020 and better at every horizon after 2020. Fourth, the value of microdata for the combined forecast materializes after 2020. We conclude that microdata are valuable for forecasting aggregate inflation, but only after large shocks. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2607.12345 |
| By: | Nathan Schor; Minchul Shin |
| Abstract: | We introduce ForeComp, an R package for comparing predictive accuracy using Diebold–Mariano type tests of equal predictive ability with standard and fixed-smoothing inference. The package provides a common interface for loss-differential based testing and includes Plot Tradeoff, a visual diagnostic for bandwidth sensitivity and the size–power tradeoff. We illustrate the toolkit with Survey of Professional Forecasters applications and Monte Carlo evidence on finite-sample performance. |
| Keywords: | forecast comparison; Diebold–Mariano test; fixed-b asymptotics; fixed-m asymptotics; long-run variance estimation; R package |
| JEL: | C12 C22 C52 C53 |
| Date: | 2026–08–04 |
| URL: | https://d.repec.org/n?u=RePEc:fip:fedpwp:103596 |
| By: | Likai Chen; Weining Wang |
| Abstract: | Forecasting is a central goal of time-series analysis. This review centers on three major developments in recent AI-based time-series forecasting: transformers, large pretrained models for zero-shot forecasting, and diffusion-based generative forecasters. We connect these methods to the econometric tradition built around the vector autoregression (VAR) through a common object: the conditional distribution of the future given the past. The review is organized around three long-standing challenges: \emph{high dimensionality}, \emph{nonstationarity}, and \emph{nonlinearity}. We argue that modern methods make progress by expanding the classical forecasting template: they allow more flexible dynamics, use larger information sets and training corpora, and represent richer predictive distributions. Yet they often lack the inferential and structural tools that make classical models useful for testing, explanation, and policy analysis. We close by outlining open problems where econometric tools remain important. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2607.14279 |
| By: | Fisher, Jonas; Melosi, Leonardo; Rast, Sebastian |
| Abstract: | Professional forecasters' long-run inflation expectations overreact to news and exhibit persistent, predictable biases in forecast errors. A model incorporating overconfidence in private information and a persistent expectations bias---which generates persistent forecast errors across most forecasters---accounts for these two features of the data, offering a valuable tool for studying long-run inflation expectations. Our analysis highlights substantial, time-varying heterogeneity in forecasters' responses to public information, with sensitivity declining across all forecasters when monetary policy is constrained by the effective lower bound. The model provides a framework to evaluate whether policymakers' communicated inflation paths are consistent with anchored long-run expectations. |
| Keywords: | overreaction; Central bank communication |
| JEL: | E31 D83 E52 E37 |
| Date: | 2025–03 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19991 |
| By: | Adrian, Tobias; Giannone, Domenico; Luciani, Matteo; West, Mike |
| Abstract: | We introduce methodology to bridge scenario analysis and model-based risk forecasting, leveraging their respective strengths in policy settings. Our Bayesian framework addresses the fundamental challenge of reconciling judgmental narrative approaches with statistical forecasting. Analysis evaluates explicit measures of concordance of scenarios with a reference forecasting model, delivers Bayesian predictive synthesis of the scenarios to best match that reference, and addresses scenario set incompleteness. This underlies systematic evaluation and integration of risks from different scenarios, and quantifies relative support for scenarios modulo the defined reference forecasts. The framework offers advances in forecasting in policy institutions that supports clear and rigorous communication of evolving risks. We also discuss broader questions of integrating judgmental information with statistical model-based forecasts in the face of unexpected circumstances. |
| JEL: | C11 C53 E23 E32 |
| Date: | 2025–05 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20219 |
| By: | Martin, Ian |
| Abstract: | I survey work that uses information in derivative and other asset prices to forecast movements in financial markets. |
| Date: | 2025–03 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19998 |
| By: | Hauzenberger, Niko; Marcellino, Massimiliano; Pfarrhofer, Michael; Stelzer, Anna |
| Abstract: | We develop Bayesian machine learning methods for mixed data sampling (MIDAS) regressions. This involves handling frequency mismatches and specifying functional relationships between many predictors and the dependent variable. We use Gaussian processes (GPs) and compress the input space with structured and unstructured MI-DAS variants. This yields several versions of GP-MIDAS with distinct properties and implications, which we evaluate in short-horizon now- and forecasting exercises with both simulated data and data on quarterly US output growth and inflation in the GDP deflator. Our proposed framework leverages macroeconomic Big Data in a computationally efficient way and offers gains in predictive accuracy along several dimensions. |
| JEL: | C11 C22 C53 E31 E37 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19965 |
| By: | Granziera, Eleonora; Larsen, Vegard H.; Meggiorini, Greta; Melosi, Leonardo |
| Abstract: | We examine how speeches by Federal Open Market Committee (FOMC) members, including regional Fed presidents, shape private sector expectations. Speeches that signal rising inflationary pressures prompt both households and professional forecasters to raise their inflation expectations, consistent with Delphic effects. Only professional forecasters respond to Odyssean communications---statements about the Fed's intended policy response---leaving Delphic effects as the dominant channel for households. These household responses are driven by speeches from regional presidents, likely due to greater visibility in regional media coverage. A general equilibrium model, featuring agents who differ in their ability to interpret Odyssean signals, explains this heterogeneity. |
| Keywords: | Central bank communication |
| JEL: | E31 E58 D83 |
| Date: | 2025–03 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20038 |
| By: | Mark W. Watson |
| Abstract: | The persistent surge in U.S. inflation that began in 2021 caught forecasters and policymakers by surprise. The 2021 inflation shocks were viewed as transitory, not persistent, leading to large forecast errors in late 2021 and 2022. This paper asks whether time series models – using only data on current and past inflation, but incorporating stochastic volatility and exhibiting time-varying persistence – performed better. Univariate models, using real-time data, did not. Multivariate models, incorporating sectoral inflation measures, did. |
| JEL: | C32 E37 |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35435 |
| By: | Tae-Hwy Lee (Department of Economics, University of California Riverside); Dingli Wang (University of California, Riverside) |
| Abstract: | Risk managers often work with a fitted forecasting model they cannot replace even when its tail forecasts need adjustment. We propose a median-anchored rule that multiplies the distances from the fitted median to Value-at-Risk (VaR) and Expected Shortfall (ES) by a common positive multiplier. The rule preserves VaR--ES ordering, and minimizing VaR check loss gives a closed-form weighted-quantile estimator. We establish consistency, give an asymptotic distribution under high-level conditions accounting for baseline estimation, and derive a VaR coverage-error bound at the forecast origin. When the same multiplier correctly adjusts both VaR and ES, the adjusted pair has lower conditional zero-homogeneous Fissler--Ziegel (FZ0) risk at that origin. Monte Carlo experiments examine this condition and several forms of misspecification. In rolling S&P~500 forecasts the adjustment lowers FZ0 loss in all four GARCH cells, each pairwise significant, with the largest and most robust gain, 10.4%, for Gaussian GARCH at the 1% tail; the 5% gains do not survive the most conservative family-wise adjustments. A quantile-regression baseline marks the boundary: when the fitted tail is already flexible, one multiplier adds little. Filtered historical simulation quantifies what standardized residuals and conditional scales add when available. |
| Keywords: | Backtesting; Elicitability; Fissler--Ziegel Score; Forecast Evaluation; Model Risk; Quantile Regression |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:ucr:wpaper:202604 |
| By: | Andrea Panozzo, Lorenzo Spadavecchia, Adam Mugume, Elizabeth Kasekende, Samuel Namwanja Musoke, Mariss Nakayaga, Deo Sande, Anita Mpagi, Nzima Ghislain |
| Abstract: | In developing economies, output statistics arrive with long lags and omit a large informal sector, while digital payment systems record a growing share of transactions in near real time. We assess whether these records improve GDP nowcasts in Uganda, exploiting two national systems that observe complementary segments of the economy: Mobile Money, covering retail and informal transactions, and real-time gross settlement (RTGS), covering large-value formal payments. Within a pseudo-real-time design respecting each series’ publication lag, we augment a conventional macroeconomic panel with payment data across linear and machinelearning models. Payment data cut forecast errors by up to 16 percent and rank among the most informative predictors, with up to over three times the weight of a typical macroeconomic indicator. The improvement delivered by payment data is robust to macroeconomic disturbances, such as the COVID-19 contraction. Placebo tests attribute these gains to economic content, not added predictors. Already held by central banks, payment data offer a timely, low-cost input for surveillance where conventional statistics are weakest. |
| Keywords: | Nowcasting, Digital payments, Mobile money, RTGS |
| JEL: | C53 E37 G21 O17 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:baf:cbafwp:cbafwp26282 |
| By: | Liao, Yuan; Ma, Xinjie; Neuhierl, Andreas; Schilling, Linda |
| Abstract: | Machine learning in asset pricing typically predicts expected returns as point estimates, ignoring uncertainty. We develop new methods to construct forecast confidence intervals for expected returns obtained from neural networks. We show that neural network forecasts of expected returns share the same asymptotic distribution as classic nonparametric methods, enabling a closed-form expression for their standard errors. We also propose a computationally feasible bootstrap to obtain the asymptotic distribution. We incorporate these forecast confidence intervals into an uncertainty-averse investment framework. This provides an economic rationale for shrinkage implementations of portfolio selection. Empirically, our methods improve out-of-sample performance. |
| JEL: | G12 C45 C58 |
| Date: | 2025–03 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20080 |
| By: | Taizhen Cheung; SA Kwon |
| Abstract: | Large pretrained time-series models such as TimesFM are attractive for financial forecasting, but raw directional accuracy is a misleading scoreboard in equity markets. An early LoRA adapter in this project appeared to reach roughly 80% directional accuracy; we show this is not evidence of skill. Over a long horizon in a rising market, a trivial "always-up" rule attains comparably high accuracy without using the input at all. To separate genuine skill from this base-rate artifact, we build a reproducible, frozen-data benchmark with expanding walk-forward folds, a stratified held-out-ticker split, honest baselines (zero-shot TimesFM, always-up, random-walk, persistence, AR(1)), and paired significance tests (McNemar, Diebold-Mariano) under Benjamini-Hochberg FDR control. We apply the identical method to two universes -- a tech-heavy NASDAQ-100 and a broad S&P 500 -- reporting excess accuracy over the always-up base rate. Three findings replicate. First, when the historical ~80% condition is recreated, the high number is a base rate of ~0.70 that the fine-tuned model scores below. Second, pooled LoRA shows no directional skill over the base rate at any horizon on either universe (negative at the six-month horizon). Third, per-sector specialization is significantly worse than a single pooled adapter (Diebold-Mariano p |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2607.12248 |
| By: | Yiming Ma; Xinyu Chen |
| Abstract: | Financial observations are continuous, heterogeneous, and noisy, whereas decoder-only next-token models are usually built around discrete symbolic inputs. We introduce Vector-Input Autoregressive Inference for Ordinal-Return Modeling (VAIOM), a decoder-only Transformer for probabilistic next-return modeling on one-hour foreign-exchange bars. VAIOM separates input representation from output likelihood: continuous multivariate financial-event vectors preserve numerical structure at the input, while a categorical distribution over the next volatility-normalized return bucket supports cross-entropy training and likelihood evaluation. The selected 0.9M Hybrid Continuous Input model combines continuous event features with categorical asset metadata, a Mixture-of-Market-States return head, Gap, volatility-regime, and Ordinal auxiliary objectives, and full-sequence supervision. Models and preprocessing are fit using pre-2024 Train data; models are selected on 2024H2 Validation and evaluated without refitting on two 2025 Test periods. Across three independent training seeds, every model outperforms fixed single-bar LightGBM baseline in both Test halves. For the canonical checkpoint, paired gains over LightGBM are 0.029 and 0.043 bits per event. Validation experiments show that continuous input improves over discrete-token input under the same categorical return objective, full-sequence supervision improves over last-position training, and auxiliary representation shaping together with a mixture-structured return head improves return likelihood in controlled comparisons. A supporting capacity study finds that the smallest evaluated complete architecture rung achieves the strongest Validation likelihood on the present corpus. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2607.13929 |
| By: | Cole, Stephen J. (Department of Economics Marquette University); (Department of Economics Marquette University) |
| Abstract: | This paper uses an adaptive learning framework to study FOMC forecasts from the Summary of Economic Projections (SEP) dataset. FOMC expectations are modeled as the sum of two components: (1) an endogenous learning part and (2) a sentiment part capturing waves of optimism and/or pessimism. The results include key policy takeaways. FOMC forecasts are responsive to incoming macroeconomic information, consistent with adaptive learning, while sentiment is persistent, correlated across GDP growth and inflation forecasts, and becomes quantitatively more important during and around recessions. FOMC participants also rely more on their endogenous/learning model to form expectations, but sentiment plays a larger role during and around recessions. Finally, the model-implied sentiment measure is positively and significantly correlated with an external measure of FOMC sentiment and remains robust across alternative forecasting specifications. |
| Keywords: | summary of economic projections, FOMC, constant-gain learning, sentiment shocks, waves of optimism and pessimism, evolving beliefs, monetary policy |
| JEL: | C52 D84 E50 E52 E58 E60 E70 E71 |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:mrq:wpaper:2026-03 |
| By: | Xinxian Chen; Peter Reinhard Hansen; Chen Tong |
| Abstract: | We propose the Split-Session Cluster GARCH model for heavy-tailed multivariate dependence among asset returns decomposed into overnight and intraday components. The model uses convolution-$t$ distributions to allow tail behavior to differ across clusters defined by trading sessions and, within each session, by economic sectors. It also accommodates block-structured conditional correlation matrices, preserving parsimony and scalability in high-dimensional settings. The resulting likelihood remains tractable and yields a score-driven specification for dynamic correlations. We apply the model to U.S. equity returns in six-asset and 100-asset applications. The results reveal pronounced tail heterogeneity between overnight and intraday returns. Model comparisons show that session-specific tail parameters substantially improve fit relative to a common multivariate-$t$ specification, while sector-level tail partitioning delivers additional gains concentrated mainly in the overnight component. In the 100-asset application, asset-level tail heterogeneity delivers the strongest out-of-sample likelihood and global minimum-variance (GMV) portfolio performance. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2607.03669 |
| By: | Asef Y{\i}lk{\i} |
| Abstract: | This paper proposes a novel asset pricing framework that augments large language model (LLM) embeddings of annual report disclosures with supply chain knowledge graph (KG) propagation. Using FinBERT embeddings of 10-K MD&A sections for 255 S&P 500 firms over 2011-2025, two sets of return predictors are constructed: direct LLM embeddings and network-augmented embeddings, where firm-level signals propagate through inter-firm linkages. Fama-MacBeth cross-sectional regressions reveal that the network-augmented factor (net_pc_5) carries significant return predictability with a Newey-West t-statistic of -2.64, even after controlling for momentum, volatility, and firm size. A long-short portfolio sorted on net_pc_5 achieves an annualized Sharpe ratio of 0.86 and a Fama-French five-factor alpha of 7.27% per year (t = 2.30). The predictive power survives out-of-sample tests, placebo experiments, sector-neutralization, and subsample analysis. The findings suggest that inter-firm network structure contains pricing-relevant information beyond firm-level textual disclosures. |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2606.29290 |
| By: | Furceri, Davide; Giannone, Domenico; Kisat, Faizaan; Lam, Raphael; Li, Hongchi |
| Abstract: | This paper proposes a novel framework for analyzing the risks surrounding the public debt outlook, the “Debt-at-Risk.†It employs a quantile panel regression framework to assess how current macro-financial and political conditions impact the entire spectrum of possible future debt outcomes. Many of these factors — including financial conditions and economic variables such as initial debt and GDP growth — predict both the expected level and the uncertainty of future debt, implying pronounced variations in risks, especially in the upper tail of the distribution. By combining the roles of these factors, we find that in a severely adverse scenario — the 95th percentile of the future debt distribution, or debt-at-risk — global public debt could be approximately 20 percentage points higher than currently projected. The magnitudes and sources of debt risks vary over time and across countries, with high initial debt amplifying the effects of economic and financial conditions on debt-at-risk. Furthermore, empirical estimates indicate that debt-at-risk is a key variable for predicting fiscal crises. |
| Keywords: | Debt |
| JEL: | H6 F3 G1 |
| Date: | 2025–05 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:20212 |
| By: | Yurii Sholomytskyi |
| Abstract: | This paper examines the statistical properties of the IMF’s World Economic Outlook (WEO) projections over 1999–2023 for 29 economies. The optimism of WEO growth forecasts is well established; we confirm it and look behind it at two features of how the forecasts are built. First, the growth of the systemic economies (the United States and China) appears to be underutilized in the projections: forecasts embed less of the cross-country growth comovement present in the data, a gap we term forecast fragmentation that did not narrow over the sample. Second, the conditional growth–inflation link present in the historical data is weakly represented in the projections. These patterns suggest that structural models, in which such cross-country and real–nominal linkages can be verified through estimation, could be a useful complement to expert judgment, serving as a baseline check for medium-term anchors. |
| Keywords: | IMF; World Economic Outlook; Forecast error; Optimism bias; Forecast fragmentation |
| Date: | 2026–07–31 |
| URL: | https://d.repec.org/n?u=RePEc:imf:imfwpa:2026/164 |
| By: | Boni Sara (Free University of Bolzano-Bozen); Iseringhausen Martin (European Stability Mechanism); Petrella Ivan (Collegio Carlo Alberto and University of Turin, CEPR); Theodoridis Konstantinos (European Stability Mechanism and Cardiff Business School) |
| Abstract: | We compute a common factor summarising asymmetries in the expected distributions of a large set of survey-based economic data series for the euro area. This expected skewness factor is distinct from lower-moment factors and can help improve forecasts of risks to economic activity and inflation. In addition, within a monthly vector autoregression (VAR), we show that revisions to survey-based expected skewness have macroeconomic and financial implications, even when the average assessment and expected volatility reflected in the surveys remain unchanged. The skewness measure could benefit economic policy institutions by supporting timely quantitative assessments of the balance of risks. |
| Keywords: | Economic sentiment, principal components, quantile regression, skewness |
| JEL: | C22 C38 E66 |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:tur:wpapnw:107 |
| By: | Sankalp Gilda |
| Abstract: | Finance, sensing, and demand streams violate the exchangeability that IID conformal prediction and the IID bootstrap assume, and existing libraries implement either a general resampling engine or conformal calibration without the other. tsbootstrap provides block, residual, sieve, and wild resampling, classical bootstrap confidence intervals, and adaptive conformal calibrators (EnbPI, ACI, NexCP, AgACI) through a single typed API in which a specification object selects each method. In a controlled coverage study the IID bootstrap undercovers sharply under dependence; dependence-aware methods reduce the coverage deficit, the sieve nearest to nominal under short-memory linear dependence. On the shared fixed-statistic path a compiled backend runs several times faster than arch, and a streaming reduce avoids materializing the $O(Bn)$ replicate tensor, limiting peak extra memory to $O(B)$ for the statistic array. The software is MIT licensed (v0.6.1). |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2607.06690 |
| By: | Gonzalez-Casasus, Oriol; Schorfheide, Frank |
| Abstract: | VARs are often estimated with Bayesian techniques to cope with model dimensionality. The posterior means define a class of shrinkage estimators, indexed by hyperparameters that determine the relative weight on maximum likelihood estimates and prior means. In a Bayesian setting, it is natural to choose these hyperparameters by maximizing the marginal data density. However, this is undesirable if the VAR is misspecified. In this paper, we derive asymptotically unbiased estimates of the multi-step forecasting risk and the impulse response estimation risk to determine hyperparameters in settings where the VAR is (potentially) misspecified. The proposed criteria can be used to jointly select the optimal shrinkage hyperparameter, VAR lag length, and to choose among different types of multi-step-ahead predictors; or among IRF estimates based on VARs and local projections. The selection approach is illustrated in a Monte Carlo study and an empirical application. |
| Keywords: | Forecasting; Local projections; Model misspecification; Shrinkage estimation |
| JEL: | C11 C32 C52 C53 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:cpr:ceprdp:19915 |