nep-for New Economics Papers
on Forecasting
Issue of 2021‒01‒25
eleven papers chosen by
Rob J Hyndman
Monash University

  1. Interest Rate Uncertainty and the Predictability of Bank Revenues By Cepni, Oguzhan; Demirer, Riza; Gupta, Rangan; Sensoy, Ahmet
  2. The g3+ Model: An Upgrade of the Czech National Bank’s Core Forecasting Framework By Frantisek Brazdik; Tibor Hledik; Zuzana Humplova; Iva Martonosi; Karel Musil; Jakub Rysanek; Tomas Sestorad; Jaromir Tonner; Stanislav Tvrz; Jan Zacek
  3. Trading on short-term path forecasts of intraday electricity prices By Tomasz Serafin; Grzegorz Marcjasz; Rafal Weron
  4. A machine learning approach to volatility forecasting By Kim Christensen; Mathias Siggaard; Bezirgen Veliyev
  5. Forecasting in a changing world: from the great recession to the COVID-19 pandemic By Mariia Artemova; Francisco Blasques; Siem Jan Koopman; Zhaokun Zhang
  6. Dynamic Ordering Learning in Multivariate Forecasting By Bruno P. C. Levy; Hedibert F. Lopes
  7. Nowcasting Indonesia’s GDP Growth Using Machine Learning Algorithms By Tamara, Novian; Dwi Muchisha, Nadya; Andriansyah, Andriansyah; Soleh, Agus M
  8. Optimally Imprecise Memory and Biased Forecasts By Rava Azeredo da Silveira; Yeji Sung; Michael Woodford
  9. Are official forecasts of output growth in the EU still biased? Evidence from stability and convergence programmes and the European Commission’s Spring forecasts By Cronin, David; McQuinn, Kieran
  10. Fiscal policy and growth forecasts in the EU: Are official forecasters still misestimating fiscal multipliers? By Cronin, David; McQuinn, Kieran
  11. Loan Delinquency Projections for COVID-19 By Grey Gordon; John Bailey Jones

  1. By: Cepni, Oguzhan (Department of Economics, Copenhagen Business School); Demirer, Riza (Department of Economics and Finance, Southern Illinois University Edwardsville); Gupta, Rangan (Department of Economics, University of Pretoria); Sensoy, Ahmet (Bilkent University, Faculty of Business Administration)
    Abstract: This paper examines the predictive power of interest rate uncertainty over pre-provision net revenues (PPNR) in a large panel of bank holding companies (BHC). Utilizing a linear dynamic panel model based on Bayes predictor, we show that supplementing forecasting models with interest rate uncertainty improves the forecasting performance with the augmented model yielding lower forecast errors in comparison to a baseline model which includes unemployment rate, federal funds rate, and spread variables. Further separating PPNRs into two components that reflect net interest and non-interest income, we show that the predictive power of interest rate uncertainty is concentrated on the non-interest component of bank revenues. Finally, examining the point predictions under a severely stressed scenario, we show that the model can successfully predict the negative effect on overall bank revenues with a rise in the non-interest component of income during 2009:Q1. Overall, the findings suggest that stress testing exercises that involve bank revenue models can benefit from the inclusion of interest rate uncertainty and the cross-sectional information embedded in the panel of BHCs.
    Keywords: Bank stress tests; Empirical Bayes; Interest rate uncertainty; Out-of-sample forecasts
    JEL: C11 C14 C23 G21
    Date: 2021–01–05
  2. By: Frantisek Brazdik; Tibor Hledik; Zuzana Humplova; Iva Martonosi; Karel Musil; Jakub Rysanek; Tomas Sestorad; Jaromir Tonner; Stanislav Tvrz; Jan Zacek
    Abstract: This paper introduces g3+, the new core forecasting model of the Czech National Bank (CNB), which replaced the previous g3 model in July 2019. We present the features of the new core forecasting model together with our motivation for adopting them. The new structural features and extensions were motivated by our experience with using the g3 model for more than a decade as the core forecasting tool at the CNB. The new g3+ model features a novel structural foreign economy block, oil as a production factor, heterogeneous households, and other adjustments. Also, we present a new simulation approach that allows us to emulate limited information for the simulation of conditional forecasts. Furthermore, the introduction of the g3+ model on average preserves the forecasting performance of the CNBs DSGE modeling framework.
    Keywords: Conditional forecast, DSGE, g3 model, oil, small open economy, two country model
    JEL: C51 C53 E27 E37 F41
    Date: 2020–12
  3. By: Tomasz Serafin; Grzegorz Marcjasz; Rafal Weron
    Abstract: We introduce a profitable trading strategy that can support decision-making in continuous intraday markets for electricity. It utilizes a novel forecasting framework, which generates prediction bands from a pool of path forecasts or approximates them using probabilistic price forecasts. The prediction bands then define a time-dependent price level that, when exceeded, indicates a good trading opportunity. Results for the German intraday market show that, in terms of the energy score, our path forecasts beat a well performing similar-day benchmark by over 25%. Moreover, they provide empirical evidence that the increased computational burden induced by generating realistic price paths is offset by higher trading profits. Still, the proposed approximate method offers a reasonable trade-off - it does not require generating path forecasts and yields only slightly lower profits.
    Keywords: Intraday electricity market; Probabilistic forecast; Path forecast; Prediction bands; Energy score; Trading recommendations
    JEL: C22 C32 C51 C53 Q41 Q47
    Date: 2020–12–30
  4. By: Kim Christensen (Aarhus University and CREATES); Mathias Siggaard (Aarhus University and CREATES); Bezirgen Veliyev (Aarhus University and CREATES)
    Abstract: We show that machine learning (ML) algorithms improve one-day-ahead forecasts of realized variance from 29 Dow Jones Industrial Average index stocks over the sample period 2001 - 2017. We inspect several ML approaches: Regularization, tree-based algorithms, and neural networks. Off-the-shelf ML implementations beat the Heterogeneous AutoRegressive (HAR) model, even when the only predictors employed are the daily, weekly, and monthly lag of realized variance. Moreover, ML algorithms are capable of extracting substantial more information from additional predictors of volatility, including firm-specific characteristics and macroeconomic indicators, relative to an extended HAR model (HAR-X). ML automatically deciphers the often nonlinear relationship among the variables, allowing to identify key associations driving volatility. With accumulated local effect (ALE) plots we show there is a general agreement about the set of the most dominant predictors, but disagreement on their ranking. We investigate the robustness of ML when a large number of irrelevant variables, exhibiting serial correlation and conditional heteroscedasticity, are added to the information set. We document sustained forecasting improvements also in this setting.
    Keywords: Gradient boosting, high-frequency data, machine learning, neural network, random forest, realized variance, regularization, volatility forecasting
    JEL: C10 C50
    Date: 2021–01–18
  5. By: Mariia Artemova (Vrije Universiteit Amsterdam); Francisco Blasques (Vrije Universiteit Amsterdam); Siem Jan Koopman (Vrije Universiteit Amsterdam); Zhaokun Zhang (Shanghai University)
    Abstract: We develop a new targeted maximum likelihood estimation method that provides improved forecasting for misspecified linear autoregressive models. The method weighs data points in the observed sample and is useful in the presence of data generating processes featuring structural breaks, complex nonlinearities, or other time-varying properties which cannot be easily captured by model design. Additionally, the method reduces to classical maximum likelihood when the model is well specified, which results in weights which are set uniformly to one. We show how the optimal weights can be set by means of a cross-validation procedure. In a set of Monte Carlo experiments we reveal that the estimation method can significantly improve the forecasting accuracy of autoregressive models. In an empirical study concerned with forecasting the U.S. Industrial Production, we show that the forecast accuracy during the Great Recession can be significantly improved by giving greater weight to observations associated with past recessions. We further establish that the same empirical finding can be found for the 2008-2009 global financial crisis, for different macroeconomic time series, and for the COVID-19 recession in 2020.
    Keywords: Autoregressive Models, Cross-Validation, Kullback-Leibler Divergence, Stationarity and Ergodicity, Macroeconomic Time Series
    JEL: C10 C22 C32 C51
    Date: 2021–01–14
  6. By: Bruno P. C. Levy; Hedibert F. Lopes
    Abstract: In many fields where the main goal is to produce sequential forecasts for decisionmaking problems, the good understanding of the contemporaneous relations among different series is crucial for the estimation of the covariance matrix. In recent years, the modified Cholesky decomposition appeared as a popular approach to covariance matrix estimation. However, its main drawback relies on the imposition of the series ordering structure. In this work, we propose a highly flexible and fast method to deal with the problem of ordering uncertainty in a dynamic fashion with the use of Dynamic Order Probabilities. We apply the proposed method in a dynamic portfolio allocation problem, where the investor is able to learn the contemporaneous relations among different currencies. We show that our approach generates not just significant statistical improvements, but also huge economic gains for a mean-variance investor relative to the Random Walk benchmark and using fixed orders over time.
    Date: 2021–01
  7. By: Tamara, Novian; Dwi Muchisha, Nadya; Andriansyah, Andriansyah; Soleh, Agus M
    Abstract: GDP is very important to be monitored in real time because of its usefulness for policy making. We built and compared the ML models to forecast real-time Indonesia's GDP growth. We used 18 variables that consist a number of quarterly macroeconomic and financial market statistics. We have evaluated the performance of six popular ML algorithms, such as Random Forest, LASSO, Ridge, Elastic Net, Neural Networks, and Support Vector Machines, in doing real-time forecast on GDP growth from 2013:Q3 to 2019:Q4 period. We used the RMSE, MAD, and Pearson correlation coefficient as measurements of forecast accuracy. The results showed that the performance of all these models outperformed AR (1) benchmark. The individual model that showed the best performance is random forest. To gain more accurate forecast result, we run forecast combination using equal weighting and lasso regression. The best model was obtained from forecast combination using lasso regression with selected ML models, which are Random Forest, Ridge, Support Vector Machine, and Neural Network.
    Keywords: Nowcasting, Indonesian GDP, Machine Learning
    JEL: C55 E30 O40
    Date: 2020–06–26
  8. By: Rava Azeredo da Silveira (Biophysique et Neuroscience Théoriques - LPENS (UMR_8023) - Laboratoire de physique de l'ENS - ENS Paris - ENS Paris - École normale supérieure - Paris - PSL - Université Paris sciences et lettres - SU - Sorbonne Université - CNRS - Centre National de la Recherche Scientifique - UP - Université de Paris, Unibas - University of Basel); Yeji Sung (Columbia University [New York]); Michael Woodford (Columbia University [New York])
    Abstract: We propose a model of optimal decision making subject to a memory constraint. The constraint is a limit on the complexity of memory measured using Shannon's mutual information, as in models of rational inattention; but our theory differs from that of Sims (2003) in not assuming costless memory of past cognitive states. We show that the model implies that both forecasts and actions will exhibit idiosyncratic random variation; that average beliefs will also differ from rational-expectations beliefs, with a bias that fluctuates forever with a variance that does not fall to zero even in the long run; and that more recent news will be given disproportionate weight in forecasts. We solve the model under a variety of assumptions about the degree of persistence of the variable to be forecasted and the horizon over which it must be forecasted, and examine how the nature of forecast biases depends on these parameters. The model provides a simple explanation for a number of features of reported expectations in laboratory and field settings, notably the evidence of over-reaction in elicited forecasts documented by Afrouzi et al. (2020) and Bordalo et al. (2020a).
    Keywords: Over-reaction,Survey expectations,Rational inattention
    Date: 2020
  9. By: Cronin, David; McQuinn, Kieran
    Date: 2020
  10. By: Cronin, David; McQuinn, Kieran
    Date: 2020
  11. By: Grey Gordon; John Bailey Jones
    Abstract: The authors forecast the effects of the COVID-19 pandemic on loan delinquency rates under three scenarios for unemployment and house price movements. In the baseline scenario, their model predicts that loan delinquency rises from 2.3 percent in 2019 to a peak of 3.9 percent in 2025 with a total of $580 billion in write-offs. In 2021, absent policy intervention, the model predicts that delinquency would be 3.1 percent. However, mortgage forbearance, student loan forbearance, and fiscal transfers keep delinquency from increasing in 2021. The greatest reductions in delinquency are achieved through mortgage forbearance and student loan forbearance, with fiscal transfers playing a smaller role. In the authors' adverse (favorable) scenario, loan delinquency peaks at 8.1 percent (2.8 percent) and write-offs total $1.1 trillion ($420 billion).
    Keywords: COVID-19; loan delinquency; Survey of Consumer Finances
    Date: 2020–04–15

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.