nep-for New Economics Papers
on Forecasting
Issue of 2021‒01‒18
ten papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting Daily Volatility of Stock Price Index Using Daily Returns and Realized Volatility By Takahashi, Makoto; Watanabe, Toshiaki; Omori, Yasuhiro
  2. Regime-dependent commodity price dynamics: A predictive analysis By Crespo-Cuaresma, Jesus; Fortin, Ines; Hlouskova, Jaroslava; Obersteiner, Michael
  3. Day-ahead electricity prices prediction applying hybrid models of LSTM-based deep learning methods and feature selection algorithms under consideration of market coupling By Wei Li; Denis Mike Becker
  4. Smooth Robust Multi-Horizon Forecasts By Andrew B. Martinez; Jennifer L. Castle; David F. Hendry
  5. Forecasting the Olympic medal distribution during a pandemic: a socio-economic machine learning model By Christoph Schlembach; Sascha L. Schmidt; Dominik Schreyer; Linus Wunderlich
  6. The Hard Problem of Prediction for Conflict Prevention By Mueller, H.; Rauh, C.
  7. Improving the Short-term Forecast of World Trade During the Covid-19 Pandemic Using Swift Data on Letters of Credit By Benjamin Carton; Nan Hu; Joannes Mongardini; Kei Moriya; Aneta Radzikowski
  8. Nowcasting in a pandemic using non-parametric mixed frequency VARs By Huber, Florian; Koop, Gary; Onorante, Luca; Pfarrhofer, Michael; Schreiner, Josef
  9. Semi-Structural VAR and Unobserved Components Models to Estimate Finance-Neutral Output Gap By Kátay Gábor; Kerdelhué Lisa; Lequien Matthieu
  10. Using Predictive Analytics for Public Policy: The Case for Lost Work due to the COVID-19 By Cheng, Kent Jason Go

  1. By: Takahashi, Makoto; Watanabe, Toshiaki; Omori, Yasuhiro
    Abstract: This paper compares the volatility predictive abilities of some time-varying volatility models such as thestochastic volatility (SV) and exponential GARCH (EGARCH) models using daily returns, the heterogeneous au-toregressive (HAR) model using daily realized volatility (RV) and the realized SV (RSV) and realized EGARCH(REGARCH) models using the both. The data are the daily return and RV of Dow Jones Industrial Aver-age (DJIA) in US and Nikkei 225 (N225) in Japan. All models are extended to accommodate the well-knownphenomenon in stock markets of a negative correlation between today's return and tomorrow's volatility. Weestimate the HAR model by the ordinary least squares (OLS) and the EGARCH and REGARCH models bythe quasi-maximum likelihood (QML) method. Since it is not straightforward to evaluate the likelihood of theSV and RSV models, we apply a Bayesian estimation via Markov chain Monte Carlo (MCMC) to them. Byconducting predictive ability tests and analyses based on model confidence sets, we confirm that the models us-ing RV outperform the models without RV, that is, the RV provides useful information on forecasting volatility.Moreover, we find that the realized SV model performs best and the HAR model can compete with it. Thecumulative loss analysis suggests that the differences of the predictive abilities among the models are partlycaused by the rise of volatility.
    Keywords: Exponential GARCH (EGARCH) model, Heterogeneous autoregressive (HAR) model, Markov chain Monte Carlo (MCMC), Realized volatility, Stochastic volatility, Volatility forecasting
    JEL: C11 C22 C53 C58 G17
    Date: 2021–01
  2. By: Crespo-Cuaresma, Jesus (Vienna University of Economics and Business, Vienna, International Institute of Applied Systems Analysis (IIASA), Laxenburg, Wittgenstein Center for Demography and Global Human Capital, and Austrian Institute of Economic Research (WIFO), Vienna, Austria); Fortin, Ines (Institute for Advanced Studies, Vienna, Austria); Hlouskova, Jaroslava (Institute for Advanced Studies, Vienna, Austria, International Institute of Applied Systems Analysis (IIASA), Laxenburg, Austria, and University of Economics in Bratislava, Slovakia); Obersteiner, Michael (University of Oxford, Oxford, UK, and International Institute of Applied Systems Analysis (IIASA), Laxenburg, Austria)
    Abstract: We develop an econometric modelling framework to forecast commodity prices taking into account potentially different dynamics and linkages existing at different states of the world and using different performance measures to validate the predictions. We assess the extent to which the quality of the forecasts can be improved by entertaining different regime-dependent threshold models considering different threshold variables. We evaluate prediction quality using both loss minimization and profit maximization measures based on directional accuracy, directional value, the ability to predict adverse movements and returns implied by a trading strategy. Our analysis provides overwhelming evidence that allowing for regime-dependent dynamics leads to improvements in predictive ability for the Goldman Sachs Commodity Index, as well as for its five sub-indices (energy, industrial metals, precious metals, agriculture, livestock). Our results suggest the existence of a trade-off between predictive ability based on loss and profit measures, which implies that the particular aim of the prediction exercise carried out plays a very important role in terms of defining which set of models is the best to use.
    Keywords: Commodity prices, forecasting, threshold models, forecast performance, states of economy
    JEL: Q02 C53 F47
    Date: 2021–01
  3. By: Wei Li; Denis Mike Becker
    Abstract: The availability of accurate day-ahead electricity price forecasts is pivotal for electricity market participants. In the context of trade liberalisation and market harmonisation in the European markets, accurate price forecasting becomes even more difficult to obtain. The increasing power market integration has complicated the forecasting process, where electricity forecasting requires considering features from both the local market and ever-growing coupling markets. In this paper, we apply state-of-the-art deep learning models, combined with feature selection algorithms for electricity price prediction under the consideration of market coupling. We propose three hybrid architectures of long-short term memory (LSTM) deep neural networks and compare the prediction performance, in terms of various feature selections. In our empirical study, we construct a broad set of features from the Nord Pool market and its six coupling countries for forecasting the Nord Pool system price. The results show that feature selection is essential to achieving accurate prediction. Superior feature selection algorithms filter meaningful information, eliminate irrelevant information, and further improve the forecasting accuracy of LSTM-based deep neural networks. The proposed models obtain considerably accurate results.
    Date: 2021–01
  4. By: Andrew B. Martinez (Office of Macroeconomic Analysis, US Department of the Treasury); Jennifer L. Castle (Magdalen College, Climate Econometrics, and Institute for New Economic Thinking at the Oxford Martin School, University of Oxford); David F. Hendry (Nuffield College, Climate Econometrics, and Institute for New Economic Thinking at the Oxford Martin School, University of Oxford)
    Abstract: We investigate whether smooth robust methods for forecasting can help mitigate pronounced and persistent failure across multiple forecast horizons. We demonstrate that naive predictors are interpretable as local estimators of the long-run relationship with the advantage of adapting quickly after a break, but at a cost of additional forecast error variance. Smoothing over naive estimates helps retain these advantages while reducing the costs, especially for longer forecast horizons. We derive the performance of these predictors after a location shift, and confirm the results using simulations. We apply smooth methods to forecasts of U.K. productivity and U.S. 10-year Treasury yields and show that they can dramatically reduce persistent forecast failure exhibited by forecasts from macroeconomic models and professional forecasters.
    Keywords: Location Shifts; Long differencing; Productivity forecasts; Robust forecasts
    JEL: C51 C53
    Date: 2020–12
  5. By: Christoph Schlembach; Sascha L. Schmidt; Dominik Schreyer; Linus Wunderlich
    Abstract: Forecasting the number of Olympic medals for each nation is highly relevant for different stakeholders: Ex ante, sports betting companies can determine the odds while sponsors and media companies can allocate their resources to promising teams. Ex post, sports politicians and managers can benchmark the performance of their teams and evaluate the drivers of success. To significantly increase the Olympic medal forecasting accuracy, we apply machine learning, more specifically a two-staged Random Forest, thus outperforming more traditional na\"ive forecast for three previous Olympics held between 2008 and 2016 for the first time. Regarding the Tokyo 2020 Games in 2021, our model suggests that the United States will lead the Olympic medal table, winning 120 medals, followed by China (87) and Great Britain (74). Intriguingly, we predict that the current COVID-19 pandemic will not significantly alter the medal count as all countries suffer from the pandemic to some extent (data inherent) and limited historical data points on comparable diseases (model inherent).
    Date: 2020–12
  6. By: Mueller, H.; Rauh, C.
    Abstract: There is a growing interest in prevention in several policy areas and this provides a strong motivation for an improved integration of forecasting with machine learning into models of decision making. In this article we propose a framework to tackle conflict prevention. A key problem of conflict forecasting for prevention is that predicting the start of conflict in previously peaceful countries needs to overcome a low baseline risk. To make progress in this hard problem this project combines a newspaper-text corpus of more than 4 million articles with unsupervised and supervised machine learning. The output of the forecast model is then integrated into a simple static framework in which a decision maker decides on the optimal number of interventions to minimize the total cost of conflict and intervention. This exercise highlights the potential cost savings of prevention for which reliable forecasts are a prerequisite.
    Date: 2021–01–06
  7. By: Benjamin Carton; Nan Hu; Joannes Mongardini; Kei Moriya; Aneta Radzikowski
    Abstract: An essential element of the work of the Fund is to monitor and forecast international trade. This paper uses SWIFT messages on letters of credit, together with crude oil prices and new export orders of manufacturing Purchasing Managers’ Index (PMI), to improve the short-term forecast of international trade. A horse race between linear regressions and machine-learning algorithms for the world and 40 large economies shows that forecasts based on linear regressions often outperform those based on machine-learning algorithms, confirming the linear relationship between trade and its financing through letters of credit.
    Keywords: Oil prices;Imports;Exports;Trade finance;Trade balance;SWIFT,trade forecast,machine learning,WP,world trade,trade message,Brent crude oil price,trade advance,letter of credit,linear regression forecast,Merchandise trade,World trade sample
    Date: 2020–11–13
  8. By: Huber, Florian; Koop, Gary; Onorante, Luca; Pfarrhofer, Michael; Schreiner, Josef
    Abstract: This paper develops Bayesian econometric methods for posterior inference in non-parametric mixed frequency VARs using additive regression trees. We argue that regression tree models are ideally suited for macroeconomic nowcasting in the face of extreme observations, for instance those produced by the COVID-19 pandemic of 2020. This is due to their flexibility and ability to model outliers. In an application involving four major euro area countries, we find substantial improvements in nowcasting performance relative to a linear mixed frequency VAR. JEL Classification: C11, C32, C53, E37
    Keywords: Bayesian, macroeconomic forecasting, regression tree models, vector autoregressions
    Date: 2021–01
  9. By: Kátay Gábor; Kerdelhué Lisa; Lequien Matthieu
    Abstract: The paper assesses the impact of adding information on financial cycles on the output gap estimates for eight advanced economies using two unobserved components models: a reduced form extended Hodrick-Prescott filter, and a standard semi-structural unobserved components model. To complement these models, a semi-structural vector autoregression model is proposed in which only supply shocks are identified. The accuracy of the output gap estimates is assessed based on their performance in predicting recessions. The models with financial variables generally produce more accurate output gap estimates at the expense of increased real-time volatility. While the extended Hodrick-Prescott filter is particularly appealing for its real-time stability, it lags behind the two semi-structural models in terms of forecasting performance. The vector autoregression model augmented with financial variables features the best in-sample forecasting performance, and it has similar real-time prediction capabilities to the semi-structural unobserved components model. Overall, financial cycles appear to be relevant in Japan, Spain, the UK, and – to a lesser extent – in the US and in France, while they are relatively muted in Canada, Germany, and Italy.
    Keywords: Unobserved Components model, semi-structural VAR, output gap, financial cycle, sustainable growth, credit, house prices, advanced economies.
    JEL: C32 E32 E44 G01 O11 O1
    Date: 2020
  10. By: Cheng, Kent Jason Go (Syracuse University)
    Abstract: In this brief research article, I demonstrate how predictive analytics or machine learning can be used to predict outcomes that are of interest in public policy. I developed a predictive model that determined who were not able to work during the past four weeks because the COVID-19 pandemic led their employer to close or lose business. I used the Current Population Survey (CPS) collected from May to November 2020 (N=352,278). Predictive models considered were logistic regression and ensemble-based methods (bagging of regression trees, random forests, and boosted regression trees). Predictors included (1) individual-, (2) family-, (3) and community or societal- level factors. To validate the models, I used the random training test splits with equal allocation of samples for the training and testing data. The random forest with the full set of predictors and number of splits set to the square root of the number of predictors yielded the lowest testing error rate. Predictive analytics that seek to forecast the inability to work due to the pandemic can be used for automated means-testing to determine who gets aid like unemployment benefits or food stamps.
    Date: 2021–01–03

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.