nep-for New Economics Papers
on Forecasting
Issue of 2021‒04‒26
seven papers chosen by
Rob J Hyndman
Monash University

  1. Predicting Inflation with Neural Networks By Paranhos, Livia
  2. Forecasting Oil and Gold Volatilities with Sentiment Indicators Under Structural Breaks By Jiawen Luo; Riza Demirer; Rangan Gupta; Qiang Ji
  3. “Nowcasting and forecasting GDP growth with machine-learning sentiment indicators” By Oscar Claveria; Enric Monte; Salvador Torra
  4. Neural basis expansion analysis with exogenous variables: Forecasting electricity prices with NBEATSx By Kin G. Olivares; Cristian Challu; Grzegorz Marcjasz; Rafal Weron; Artur Dubrawski
  6. Merits of an Aggregate Futures Price Forecasting Model for the All Wheat U.S. Season-Average Farm Price By Hoffmann, Linwood; Bond, Jennifer K.; Matias, Mariana
  7. Macroeconomic forecasting with statistically validated knowledge graphs By Sonja Tilly; Giacomo Livan

  1. By: Paranhos, Livia (University of Warwick)
    Abstract: This paper applies neural network models to forecast inflation. The use of a particular recurrent neural network, the long-short term memory model, or LSTM, that summarizes macroeconomic information into common components is a major contribution of the paper. Results from an exercise with US data indicate that the estimated neural nets usually present better forecasting performance than standard benchmarks, especially at long horizons. The LSTM in particular is found to outperform the traditional feed-forward network at long horizons, suggesting an advantage of the recurrent model in capturing the long-term trend of inflation. This finding can be rationalized by the so called long memory of the LSTM that incorporates relatively old information in the forecast as long as accuracy is improved, while economizing in the number of estimated parameters. Interestingly, the neural nets containing macroeconomic information capture well the features of inflation during and after the Great Recession, possibly indicating a role for nonlinearities and macro information in this episode. The estimated common components used in the forecast seem able to capture the business cycle dynamics, as well as information on prices.
    Keywords: forecasting ; inflation ; neural networks ; deep learning ; LSTM model
    Date: 2021
  2. By: Jiawen Luo (School of Business Administration, South China University of Technology, Guangzhou, China); Riza Demirer (Department of Economics and Finance, Southern Illinois University Edwardsville, Edwardsville, IL 62026-1102, USA); Rangan Gupta (Department of Economics, University of Pretoria, Private Bag X20, Hatfield, 0028, South Africa); Qiang Ji (Institutes of Science and Development, Chinese Academy of Sciences, Beijing, China)
    Abstract: This paper contributes to the literature on forecasting the realized volatility of oil and gold by (i) utilizing the Infinite Hidden Markov (IHM) switching model within the Heterogeneous Autoregressive (HAR) framework to accommodate structural breaks in the data and (ii) incorporating, for the first time in the literature, various sentiment indicators that proxy for the speculative and hedging tendencies of investors in these markets as predictors in the forecasting models. We show that accounting for structural breaks and incorporating sentiment-related indicators in the forecasting model does not only improve the out-of-sample forecasting performance of volatility models but also has significant economic implications, offering improved risk-adjusted returns for investors, particularly for short-term and mid-term forecasts. We also find evidence of significant cross-market information spilling over across the oil, gold, and stock markets that also contributes to the predictability of short-term market fluctuations due to sentiment-related factors. The results highlight the predictive role of investor sentiment-related factors in improving the forecast accuracy of volatility dynamics in commodities with the potential to also yield economic gains for investors in these markets.
    Keywords: Crude oil, realized volatility forecast, Infinite Hidden Markov model, structural break, speculation
  3. By: Oscar Claveria (AQR-IREA, University of Barcelona); Enric Monte (Polytechnic University of Catalunya); Salvador Torra (Riskcenter-IREA, University of Barcelona)
    Abstract: We apply the two-step machine-learning method proposed by Claveria et al. (2021) to generate country-specific sentiment indicators that provide estimates of year-on-year GDP growth rates. In the first step, by means of genetic programming, business and consumer expectations are evolved to derive sentiment indicators for 19 European economies. In the second step, the sentiment indicators are iteratively re-computed and combined each period to forecast yearly growth rates. To assess the performance of the proposed approach, we have designed two out-of-sample experiments: a nowcasting exercise in which we recursively generate estimates of GDP at the end of each quarter using the latest survey data available, and an iterative forecasting exercise for different forecast horizons We found that forecasts generated with the sentiment indicators outperform those obtained with time series models. These results show the potential of the methodology as a predictive tool.
    Keywords: Forecasting, Economic growth, Business and consumer expectations, Symbolic regression, Evolutionary algorithms, Genetic programming. JEL classification: C51, C55, C63, C83, C93
    Date: 2021–02
  4. By: Kin G. Olivares; Cristian Challu; Grzegorz Marcjasz; Rafal Weron; Artur Dubrawski
    Abstract: We extend the neural basis expansion analysis (NBEATS) to incorporate exogenous factors. The resulting method, called NBEATSx, improves on a well performing deep learning model, extending its capabilities by including exogenous variables and allowing it to integrate multiple sources of useful information. To showcase the utility of the NBEATSx model, we conduct a comprehensive study of its application to electricity price forecasting (EPF) tasks across a broad range of years and markets. We observe state-of-the-art performance, significantly improving the forecast accuracy by nearly 20% over the original NBEATS model, and by up to 5% over other well established statistical and machine learning methods specialized for these tasks. Additionally, the proposed neural network has an interpretable configuration that can structurally decompose time series, visualizing the relative impact of trend and seasonal components and revealing the modeled processes' interactions with exogenous factors.
    Keywords: Deep learning; NBEATS and NBEATSx models; Interpretable neural network; Time series decomposition; Fourier series; Electricity price forecasting
    JEL: C22 C32 C45 C51 C53 Q41 Q47
    Date: 2021–04–19
  5. By: Denis Shibitov (Bank of Russia, Russian Federation); Mariam Mamedli (Bank of Russia, Russian Federation)
    Abstract: We show, how the forecasting performance of models varies, when certain inaccuracies in the pseudo real-time experiment take place. We consider the case of Russian CPI forecasting and estimate several models on not seasonally adjusted data vintages. Particular attention is paid to the availability of the variables at the moment of forecast: we take into account the release timing of the series and the corresponding release delays, in order to reconstruct the forecasting in real-time. In the series of experiments, we quantify how each of these issues affect the out-of-sample error. We illustrate, that the neglect of the release timing generally lowers the errors. The same is true for the use of seasonally adjusted data. The impact of the data vintages depends on the model and forecasting period. The overall effect of all three inaccuracies varies from 8% to 17% depending on the forecasting horizon. This means, that the actual forecasting error can be significantly underestimated, when inaccurate pseudo real-time experiment is run. We underline the need to take these aspects into account, when the real-time forecasting is considered.
    Keywords: inflation, pseudo real-time forecasting, data vintages, machine learning, neural networks.
    JEL: C14 C45 C51 C53
    Date: 2021–04
  6. By: Hoffmann, Linwood; Bond, Jennifer K.; Matias, Mariana
    Abstract: To inform their forecasts, U.S. wheat analysts concerned with production, marketing, and policy issues use the U.S. Department of Agriculture all wheat season-average farm price (SAFP) as reported in World Agricultural Supply and Demand Estimates (WASDE). A futures-based forecasting model linked to hard red winter (HRW) futures prices (Hoffman and Balagtas, 1999) provides important input into the development of the monthly WASDE all wheat SAFP projection. However, in recent years, price relationships among the major classes of wheat have changed, suggesting that additional wheat futures prices should be included in the model. This report presents an alternative, aggregate futures-based forecasting model that utilizes the three available wheat futures contract prices: HRW, soft red winter (SRW), and hard red spring (HRS), which represents the majority of U.S. wheat production. Results show the aggregate futures-based model tends to provide forecasts with a lower mean absolute percent error and a more accurate prediction of positive directional movement than the HRW-only model. Further, the aggregate model more closely tracks the monthly WASDE SAFP projections.
    Keywords: Agribusiness, Agricultural Finance, Crop Production/Industries, Demand and Price Analysis, Financial Economics, Marketing
    Date: 2021–04–20
  7. By: Sonja Tilly; Giacomo Livan
    Abstract: This study leverages narrative from global newspapers to construct theme-based knowledge graphs about world events, demonstrating that features extracted from such graphs improve forecasts of industrial production in three large economies compared to a number of benchmarks. Our analysis relies on a filtering methodology that extracts "backbones" of statistically significant edges from large graph data sets. We find that changes in the eigenvector centrality of nodes in such backbones capture shifts in relative importance between different themes significantly better than graph similarity measures. We supplement our results with an interpretability analysis, showing that the theme categories "disease" and "economic" have the strongest predictive power during the time period that we consider. Our work serves as a blueprint for the construction of parsimonious - yet informative - theme-based knowledge graphs to monitor in real time the evolution of relevant phenomena in socio-economic systems.
    Date: 2021–04

This nep-for issue is ©2021 by Rob J Hyndman. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.