|
on Forecasting |
| By: | Duo Zhang; Jiayu Li; Junyi Mo; Elynn Chen |
| Abstract: | Accurate volatility forecasts are vital in modern finance for risk management, portfolio allocation, and strategic decision-making. However, existing methods face key limitations. Fully multivariate models, while comprehensive, are computationally infeasible for realistic portfolios. Factor models, though efficient, primarily use static factor loadings, failing to capture evolving volatility co-movements when they are most critical. To address these limitations, we propose a novel, model-agnostic Factor-Augmented Volatility Forecast framework. Our approach employs a time-varying factor model to extract a compact set of dynamic, cross-sectional factors from realized volatilities with minimal computational cost. These factors are then integrated into both statistical and AI-based forecasting models, enabling a unified system that jointly models asset-specific dynamics and evolving market-wide co-movements. Our framework demonstrates strong performance across two prominent asset classes-large-cap U.S. technology equities and major cryptocurrencies-over both short-term (1-day) and medium-term (7-day) horizons. Using a suite of linear and non-linear AI-driven models, we consistently observe substantial improvements in predictive accuracy and economic value. Notably, a practical pairs-trading strategy built on our forecasts delivers superior risk-adjusted returns and profitability, particularly under adverse market conditions. |
| Date: | 2025–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2508.01880 |
| By: | Ben A. Marconi |
| Abstract: | Financial time series forecasting presents significant challenges due to complex nonlinear relationships, temporal dependencies, variable interdependencies and limited data availability, particularly for tasks involving low-frequency data, newly listed instruments, or emerging market assets. Time Series Foundation Models (TSFMs) offer a promising solution through pretraining on diverse time series corpora followed by task-specific adaptation. This study evaluates two TSFMs (Tiny Time Mixers (TTM) and Chronos) across three financial forecasting tasks: US 10-year Treasury yield changes, EUR/USD volatility, and equity spread prediction. Results demonstrate that TTM exhibits strong transferability. When fine-tuning both the pretrained version of TTM and an untrained model with the same architecture, the pretrained version achieved 25-50% better performance when fine-tuned on limited data and 15-30% improvements even when fine-tuned on lengthier datasets. Notably, TTM's zero-shot performance outperformed naive benchmarks in volatility forecasting and equity spread prediction, with the latter demonstrating that TSFMs can surpass traditional benchmark models without fine-tuning. The pretrained model consistently required 3-10 fewer years of data to achieve comparable performance levels compared to the untrained model, demonstrating significant sample-efficiency gains. However, while TTM outperformed naive baselines, traditional specialised models matched or exceeded its performance in two of three tasks, suggesting TSFMs prioritise breadth over task-specific optimisation. These findings indicate that TSFMs, though still nascent, offer substantial promise for financial forecasting-particularly in noisy, data-constrained tasks-but achieving competitive performance likely requires domain-specific pretraining and architectural refinements tailored to financial time series characteristics. |
| Date: | 2025–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2507.07296 |
| By: | Zihan Lin; Haojie Liu; Randall R. Rojas |
| Abstract: | This study proposes a novel portfolio optimization framework that integrates statistical social network analysis with time series forecasting and risk management. Using daily stock data from the S&P 500 (2020-2024), we construct dependency networks via Vector Autoregression (VAR) and Forecast Error Variance Decomposition (FEVD), transforming influence relationships into a cost-based network. Specifically, FEVD breaks down the VAR's forecast error variance to quantify how much each stock's shocks contribute to another's uncertainty information we invert to form influence-based edge weights in our network. By applying the Minimum Spanning Tree (MST) algorithm, we extract the core inter-stock structure and identify central stocks through degree centrality. A dynamic portfolio is constructed using the top-ranked stocks, with capital allocated based on Value at Risk (VaR). To refine stock selection, we incorporate forecasts from ARIMA and Neural Network Autoregressive (NNAR) models. Trading simulations over a one-year period demonstrate that the MST-based strategies outperform a buy-and-hold benchmark, with the tuned NNAR-enhanced strategy achieving a 63.74% return versus 18.00% for the benchmark. Our results highlight the potential of combining network structures, predictive modeling, and risk metrics to improve adaptive financial decision-making. |
| Date: | 2025–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2507.20039 |
| By: | Sicheng Fu; Fangfang Zhu; Xiangdong Liu |
| Abstract: | This paper investigates the dynamics of risk transmission in cryptocurrency markets and proposes a novel framework for volatility forecasting. The framework uncovers two key empirical facts: the asymmetric amplification of volatility spillovers in both tails, and a structural decoupling between market size and systemic importance. Building on these insights, we develop a state-adaptive volatility forecasting model by extracting time-varying quantile spillover features across different volatility components. These features are embedded into an extended Log-HAR structure, resulting in the SA-Log-HAR model. Empirical results demonstrate that the proposed model outperforms benchmark alternatives in both in-sample fitting and out-of-sample forecasting, particularly in capturing extreme volatility and tail risks with greater robustness and explanatory power. |
| Date: | 2025–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2507.22409 |
| By: | Coppola, Antonio; Clayton, Christopher |
| Abstract: | We examine whether and how granular, real-time predictive models should be integrated into central banks' macroprudential toolkit. First, we develop a tractable framework that formalizes the tradeoff regulators face when choosing between implementing models that forecast systemic risk accurately but have uncertain causal content and models with the opposite profile. We derive the regulator’s optimal policy in a setting in which private portfolios react endogenously to the regulator's model choice and policy rule. We show that even purely predictive models can generate welfare gains for a regulator, and that predictive precision and knowledge of causal impacts of policy interventions are complementary. Second, we introduce a deep learning architecture tailored to financial holdings data—a graph transformer—and we discuss why it is optimally suited to this problem. The model learns vector embedding representations for both assets and investors by explicitly modeling the relational structure of holdings, and it attains state-of-the-art predictive accuracy in out-of-sample forecasting tasks including trade prediction. |
| Date: | 2025–07–25 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:xwsje_v1 |
| By: | Jiawen Luo (School of Business Administration, South China University of Technology, Guangzhou 510640, China); Jingyi Deng (School of Business Administration, South China University of Technology, Guangzhou 510640, China); Juncal Cunado (University of Navarra, School of Economics, Edificio Amigos, E-31080 Pamplona, Spain); Rangan Gupta (Department of Economics, University of Pretoria, Private Bag X20, Hatfield 0028, South Africa) |
| Abstract: | This paper investigates the predictability of supply and demand oil price shocks on U.S. Gross Domestic Product (GDP) using several Mixed Data Sampling (MIDAS) models that link quarterly GDP to monthly oil price shocks for the period 1981-2023. The main findings reveal that oil demand shocks, particularly economic activity and inventory shocks, have higher forecast ability than oil supply shocks, highlighting the importance of disentangling oil price shocks into their underlying components. Additionally, our results suggest that the Time Varying Parameter (TVP)-MIDAS model most effectively captures the dynamic relationship between oil price fluctuations and economic activity, pointing to the heterogeneous impact of oil price shocks over time. Finally, when we extend our analysis to other regions in the world, the results suggest that while oil demand shocks play a significant role in forecasting economic activity in advanced regions, the emerging regions are more vulnerable to oil supply shocks. |
| Keywords: | Oil price shocks, economic activity, Mixed-Data-Sampling (MIDAS), Time-Varying Parameter MIDAS (TVP-MIDAS), Forecast evaluation |
| JEL: | C22 C53 Q41 |
| Date: | 2025–07 |
| URL: | https://d.repec.org/n?u=RePEc:pre:wpaper:202523 |
| By: | Sung Hoon Choi; Donggyu Kim |
| Abstract: | Based on It\^o semimartingale models, several studies have proposed methods for forecasting intraday volatility using high-frequency financial data. These approaches typically rely on restrictive parametric assumptions and are often vulnerable to model misspecification. To address this issue, we introduce a novel nonparametric prediction method for the future intraday instantaneous volatility process during trading hours, which leverages both previous days' data and the current day's observed intraday data. Our approach imposes an interday-by-intraday matrix representation of the instantaneous volatility, which is decomposed into a low-rank conditional expectation component and a noise matrix. To predict the future conditional expected volatility vector, we exploit this low-rank structure and propose the Structural Intraday-volatility Prediction (SIP) procedure. We establish the asymptotic properties of the SIP estimator and demonstrate its effectiveness through an out-of-sample prediction study using real high-frequency trading data. |
| Date: | 2025–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2507.22173 |
| By: | Maretha Roseline Syahnie (Department of Economics, Faculty of Economics & Business, Universitas Gadjah Mada); Muhammad Ryan Sanjaya (Department of Economics, Faculty of Economics & Business, Universitas Gadjah Mada) |
| Abstract: | MSMEs, also known as micro, small, and medium-sized enterprises, are the backbone of the economy in developing countries. Empirical studies indicate that SMEs generally face obstacles, particularly in financing. This study focuses on two main aspects: indexing financial inclusion using principal component analysis (PCA), and analyzing credit and financial inclusion using vector autoregression (VAR) for forecasting. Through a two-stage indexing methodology, the study emphasizes the importance of geographical reach in financial inclusion availability compared to demographic reach, with availability being the most crucial dimension compared to accessibility and usage. VAR models and forecasting were developed for the period from March 2012 to July 2022 in Indonesia, incorporating other variables, such as accessto credit, credit risk, and real GDP. The use of VAR demonstrates consistency, accuracy, and reliability in producing predictions that closely approximate reality, providing a critical basis for policymakers. |
| Keywords: | Micro, small, and medium enterprises (MSMEs) financing, principal component analysis (PCA), financial inclusion index, credit, vector autoregression (VAR), forecasting, Indonesia |
| JEL: | C32 E44 G21 O16 |
| Date: | 2024–07 |
| URL: | https://d.repec.org/n?u=RePEc:gme:wpaper:202407007 |
| By: | Duane, Jackson; Ren, Alicia; Zhang, Wei |
| Abstract: | This paper presents a focused review of recent academic advances in the application of deep learning techniques to algorithmic trading. While traditional machine learning models have long been used in financial forecasting, the last decade has seen a rapid expansion in the use of deep learning architectures due to their ability to model non-linear dependencies, learn hierarchical features, and process high-dimensional sequential data. We categorize and synthesize developments across three primary paradigms: supervised deep learning models for price prediction and signal generation, unsupervised and generative approaches for feature extraction and data augmentation, and reinforcement learning agents for decision-making in trading environments. By analyzing over 30 recent peer-reviewed studies, we highlight how modern models such as attention-based networks, graph neural networks, and deep Q-learning have enhanced the robustness and adaptability of trading algorithms. We also discuss key limitations—including overfitting, data non-stationarity, and lack of interpretability—and summarize efforts to address them. This review serves as a resource for researchers seeking a clear, academically grounded perspective on how deep learning is currently reshaping algorithmic trading systems. |
| Date: | 2025–07–23 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:ctxf9_v1 |
| By: | Afees A. Salisu (Centre for Econometrics & Applied Research, Ibadan, Nigeria; Department of Economics, University of Pretoria, Private Bag X20, Hatfield 0028, South Africa); Ahamuefula E. Ogbonna (Centre for Econometrics & Applied Research, Ibadan, Nigeria); Rangan Gupta (Department of Economics, University of Pretoria, Private Bag X20, Hatfield 0028, South Africa); Yunhan Zhang (Institutes of Science and Development, Chinese Academy of Sciences, Beijing 100190, China; School of Public Policy and Management, University of Chinese Academy of Sciences, Beijing 100049, China) |
| Abstract: | We investigate the predictive content of climate policy uncertainty (CPU) for forecasting the inflation rate of the United States (US) over the monthly period of 1987:05 to 2024:11. We evaluate the performance of our proposed CPU-based predictive model, estimated via the Feasible Quasi Generalized Least Squares (FQGLS) approach, against a historical average benchmark model, with the FQGLS technique adopted to account for heteroscedasticity and autocorrelation in the data. We find statistical evidence in favor of a CPU-based model relative to the benchmark, as well as in case of an extended model involving physical risks of climate change and financial and macroeconomic factors, extracted from a large data set, when CPU is included. The predictive superiority of climate policy-related uncertainties relative to the historical mean continues to be robust under alternative local and global metrics of CPU, as well as in a mixed-frequency set-up, given the availability of high-frequency (weekly) CPU data. Moreover, the importance of local- and global-CPUs is also found to hold in forecasting the inflation rates of 11 other advanced and emerging countries in a statistically significant manner compared to the historical average model. Though across all the 12 economies, own- and global-CPUs perform equally well in forecasting the respective inflation rates. The general importance of uncertainties surrounding policy decisions to tackle climate change in shaping the future path of inflation, understandably, carries implications for the monetary authority. |
| Keywords: | Climate Policy Uncertainty, Inflation, Forecasting |
| JEL: | C22 C53 E31 E37 Q54 |
| Date: | 2025–08 |
| URL: | https://d.repec.org/n?u=RePEc:pre:wpaper:202525 |
| By: | Yusifzada, Tural; Cömert, Hasan; Ahmadov, Vugar |
| Abstract: | This study introduces a novel composite approach to nonlinear inflation dynamics in identifying historical inflation patterns and forecasting future regime shifts. Assuming inflation's responsiveness to its determinants varies across inflation regimes and that inflation shock magnitude shapes the dynamics, we endogenously identify distinct inflation regimes and analyze nonlinear behaviors within such regimes for the BRICS countries (Brazil, Russia, India, China, and South Africa) and Türkiye. In the first stage of our analysis, we employ a Hidden Markov Regime Switching Model combined with Monte Carlo simulations to establish high- and low- inflation thresholds. In the second stage, we utilize an ordered probit model to identify nonlinear probabilistic relationships between inflation regimes and key drivers of inflation such as unit labor costs, exchange rates, and global inflation. Our method achieves over 90% accuracy in predicting inflation regimes based on historical data. It also shows particularly strong out-of-sample performance in the post-pandemic period, outperforming the forecasts of international financial institutions. Even without prior knowledge of exogenous variables, the method anticipates re- gime shifts in five of the six countries analyzed for 2022 and 2023. Our approach offers researchers and central bankers a robust alternative analytical framework for managing high- and low-inflation environments where traditional linear or equilibrium-based models fall short. |
| Keywords: | high inflation, regime switching model, probit model, early warning |
| JEL: | E31 E37 E12 C24 C51 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:bofitp:323946 |
| By: | Kase, Hanno; Rigato, Rodolfo Dinis |
| Abstract: | We introduce an estimated medium scale Heterogeneous-Agent New Keynesian model for forecasting and policy analysis in the Euro Area and discuss the applications of this type of models in central banks, focusing on two main exercises. First, we examine an alternative scenario for monetary policy during the early 2020s inflationary episode, showing that earlier hikes in interest rates would have affected more strongly households at the lower end of the wealth distribution, whose consumption our model suggests was already depressed relative to the rest of the population. To provide intuition for this result, we introduce a new decomposition of the effects of monetary policy on consumption across the wealth distribution. Second, we show that introducing heterogeneous households does not come at the cost of forecasting accuracy by comparing the performance of our model to its exact representative-agent counterpart and demonstrating nearly identical results in predicting key aggregate variables. JEL Classification: D31, E12, E21, E52 |
| Keywords: | forecasting, heterogeneous-agent New Keynesian models, inequality, monetary policy |
| Date: | 2025–08 |
| URL: | https://d.repec.org/n?u=RePEc:ecb:ecbwps:20253086 |
| By: | Schultze, Michelle |
| Abstract: | Kyrgyzstan serves as a key case study for the broader Central Asia–Russia labor pipeline, which supported an estimated 8 million migrants annually in 2020. Prior to the Russo-Ukraine war, remittances from Russia accounted for approximately 30% of Kyrgyzstan’s GDP, driven by over 10% of its population working in Russia. However, understanding wartime migration dynamics is challenging due to suspected political interference in Russian data, restricted foreign access to this data, and the informality that characterizes Central Asian migration patterns. This study incorporates Yandex Wordstat, Google Trends, XGBoost (which outperforms other machine learning methods), and autoregressive models to "nowcast" missing data. The results reveal a push effect linked to war onset in February 2022 and war intensity. However, all three of the analyzed migration datasets suggest a potential delayed labor substitution effect as Central Asian migrants fill vacancies left by conscripted Russian workers, proxied by casualty data from Mediazona and the BBC. The study also examines remittance trends, which seem to increase along with the labor substitution effect after a two-month lag. These results are robust to Russia- and Kyrgyzstan-side socioeconomic controls such as wage levels and population dynamics. This study provides new insight into the largely opaque Central Asia–Russia labor pipeline, a critical element in development policymaking for both regions. It also introduces a novel methodology for nowcasting migration trends, particularly through Yandex Wordstat, which has been largely overlooked in English-language scholarship. |
| Date: | 2025–07–24 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:z2wch_v1 |
| By: | Stefan Sauer; Klaus Wohlrabe |
| Abstract: | The ifo Business Climate Index is one of the most important leading indicators for the German economy. It is based on a monthly survey of approximately 9, 000 firms and reflects responses to two core questions: the assessment of the firms' current business situation and their expectations for the next six months. These questions are deliberately formulated without precise definitions, allowing each respondent to draw on its own relevant factors. This paper investigates which factors firms actually consider, whether they differ across sectors and firm types, and how their importance has changed over time. To this end, we conducted a dedicated meta-survey in 2019 and repeated it as part of the regular ifo Business Survey in 2025. Our results show that internal factors - such as profit situation, demand, and turnover - are the primary drivers of firms' assessments. However, external influences, particularly the economic policy framework and general economic sentiment, have gained importance in recent years. These findings enhance the interpretability of the index and contribute to understanding its strong forecasting performance. The identified factors may also prove valuable for applied business cycle analysis. |
| Keywords: | business climate, economic sentiment, expectation formation, firm survey, ifo Business Survey |
| JEL: | C53 C83 L20 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:ces:ceswps:_12007 |
| By: | Volz, Ute; Wicknig, Florian |
| Abstract: | We propose novel measures to evaluate the risk profile of longer-term inflation expectations, using data on inflation probabilities from the ECB's Survey of Professional Forecasters (SPF). Unlike existing indicators, these measures specifically incorporate the central bank's inflation target. This allows for a more precise assessment of forecasters' perceptions of risks to the central bank's ability to achieve its target. Consequently, these measures provide a valuable additional criterion for assessing the degree of expectation anchoring. In contrast to other metrics, our measures indicate that, between 2014 and 2017 as well as during the Covid-19 crisis, professional forecasters saw the risk that inflation could undershoot the target in the longer term. Moreover, our indicators suggest that, following Russia's invasion of Ukraine, survey participants perceived a risk of inflation overshooting the target four to five years ahead. |
| Keywords: | Inflation, Expectations, Monetary Policy, Survey of Professional Forecasters |
| JEL: | E31 E58 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:bubtps:323949 |