nep-for New Economics Papers
on Forecasting
Issue of 2020‒03‒02
nine papers chosen by
Rob J Hyndman
Monash University

  1. PCA forecast averaging - predicting day-ahead and intraday electricity prices By Katarzyna Maciejowska; Bartosz Uniejewski; Tomasz Serafin
  2. Common Factor Augmented Forecasting Models for the US Dollar-Korean Won Exchange Rate By Hyeongwoo Kim; Soohyon Kim
  3. Text-based crude oil price forecasting By Yun Bai; Xixi Li; Hao Yu; Suling Jia
  4. Beating the naive: Combining LASSO with naive intraday electricity price forecasts By Grzegorz Marcjasz; Bartosz Uniejewski; Rafal Weron
  5. Density forecast combinations: the real-time dimension By McAdam, Peter; Warne, Anders
  6. Nonparametric forecasting of multivariate probability density functions By Dominique Guegan; Matteo Iacopini
  7. Evaluating the forecasting accuracy of the closed- and open economy New Keynesian DSGE models By Van Nguyen, Phuong
  8. Improving S&P stock prediction with time series stock similarity By Lior Sidi
  9. Diverging roads: Theory-based vs. machine learning-implied stock risk premia By Grammig, Joachim; Hanenberg, Constantin; Schlag, Christian; Sönksen, Jantje

  1. By: Katarzyna Maciejowska; Bartosz Uniejewski; Tomasz Serafin
    Abstract: Recently, the development in combining point forecasts of electricity prices obtained with different length of calibration windows have provided an extremely efficient and simple tool for improving predictive accuracy. However, the proposed methods are strongly depended on expert knowledge and may not be directly transferred from one to another model or market. Hence, we consider a novel extension and propose to use Principal Component Analysis (PCA) to automate the procedure of averaging over a rich pool of predictions. We apply PCA to a panel of over 650 point forecasts obtained for different calibration windows. The robustness of the approach is evaluated with three different forecasting tasks, i.e., forecasting day-ahead prices, forecasting intraday ID3 prices one day in advance and finally very short term forecasting of ID3 prices (i.e., six hours before delivery). The empirical results are compared using the Mean Absolute Error measure and Giacomini and White test for conditional predictive ability (CPA). The results indicate that PCA averaging not only yields significantly more accurate forecasts than individual predictions but also outperforms other forecast averaging schemes.
    Keywords: Electricity price forecasting; Day-ahead market; Intraday market; Forecast averaging; Principal component analysis; Decision-making
    JEL: C22 C32 C51 C53 Q41 Q47
    Date: 2020–02–04
  2. By: Hyeongwoo Kim (Auburn University); Soohyon Kim (Bank of Korea)
    Abstract: We propose factor-augmented out of sample forecasting models for the real exchange rate between Korea and the US. We estimate latent common factors by applying an array of data dimensionality reduction methods to a large panel of monthly frequency time series data. We augment benchmark forecasting models with common factor estimates to formulate out-of-sample forecasts of the real exchange rate. Major findings are as follows. First, our factor models outperform conventional forecasting models when combined with factors from the US macroeconomic predictors. Second, our factor models perform well at longer horizons when American real activity factors are employed, whereas American nominal/financial market factors help improve short-run prediction accuracy. Third, models with global PLS factors from UIP fundamentals overall perform well, while PPP and RIRP factors play a limited role in forecasting.
    Keywords: Won/Dollar Real Exchange Rate, Principal Component Analysis, Partial Least Squares, LASSO, Out-of-Sample Forecast
    JEL: C38 C53 C55 F31 G17
    Date: 2020–02–14
  3. By: Yun Bai; Xixi Li; Hao Yu; Suling Jia
    Abstract: Crude oil price forecasting has attracted substantial attention in the field of forecasting. Recently, the research on text-based crude oil price forecasting has advanced. To improve accuracy, some studies have added as many covariates as possible, such as textual and nontextual factors, to their models, leading to unnecessary human intervention and computational costs. Moreover, some methods are only designed for crude oil forecasting and cannot be well transferred to the forecasting of other similar futures commodities. In contrast, this article proposes a text-based forecasting framework for futures commodities that uses only future news headlines obtained from to forecast crude oil prices. Two marketing indexes, the sentiment index and the topic intensity index, are extracted from these news headlines. Considering that the public's sentiment changes over time, the time factor is innovatively applied to the construction of the sentiment index. Taking the nature of the short news headlines into consideration, a short text topic model called SeaNMF is used to calculate the topic intensity of the futures market more accurately. Two methods, VAR and RFE, are used for lag order judgment and feature selection, respectively, at the model construction stage. The experimental results show that the Ada-text model outperforms the Adaboost.RT baseline model and the other benchmarks.
    Date: 2020–01
  4. By: Grzegorz Marcjasz; Bartosz Uniejewski; Rafal Weron
    Abstract: A recent electricity price forecasting study claims that the German intraday, continuous-time market for hourly products is weak-form efficient, i.e., that the best predictor for the so-called ID3-Price index is the most recent transaction price. Here, we undermine this claim and show that we can beat the naive forecast by combining it with a prediction of a parameter-rich model estimated using the least absolute shrinkage and selection operator (LASSO). We further argue, that that if augmented with timely predictions of fundamental variables for the coming hours, the LASSO-estimated model itself can significantly outperform the naive forecast.
    Keywords: Intraday electricity market; ID3-Price index; Price forecasting; Variable selection; Fundamental variables; LASSO; Averaging forecasts
    JEL: C22 C32 C51 C53 Q41 Q47
    Date: 2020–02–02
  5. By: McAdam, Peter; Warne, Anders
    Abstract: Density forecast combinations are examined in real-time using the log score to compare five methods: fixed weights, static and dynamic prediction pools, as well as Bayesian and dynamic model averaging. Since real-time data involves one vintage per time period and are subject to revisions, the chosen actuals for such comparisons typically differ from the information that can be used to compute model weights. The terms observation lag and information lag are introduced to clarify the different time shifts involved for these computations and we discuss how they influence the combination methods. We also introduce upper and lower bounds for the density forecasts, allowing us to benchmark the combination methods. The empirical study employs three DSGE models and two BVARs, where the former are variants of the Smets and Wouters model and the latter are benchmarks. The models are estimated on real-time euro area data and the forecasts cover 2001–2014, focusing on inflation and output growth. We find that some combinations are superior to the individual models for the joint and the output forecasts, mainly due to over-confident forecasts of the BVARs during the Great Recession. Combinations with limited weight variation over time and with positive weights on all models provide better forecasts than those with greater weight variation. For the inflation forecasts, the DSGE models are better overall than the BVARs and the combination methods. JEL Classification: C11, C32, C52, C53, E37
    Keywords: Bayesian inference, euro area, forecast comparisons, model averaging, prediction pools, predictive likelihood
    Date: 2020–02
  6. By: Dominique Guegan (UP1 - Université Panthéon-Sorbonne, CES - Centre d'économie de la Sorbonne - UP1 - Université Panthéon-Sorbonne - CNRS - Centre National de la Recherche Scientifique, Labex ReFi - UP1 - Université Panthéon-Sorbonne, University of Ca’ Foscari [Venice, Italy]); Matteo Iacopini (UP1 - Université Panthéon-Sorbonne, CES - Centre d'économie de la Sorbonne - UP1 - Université Panthéon-Sorbonne - CNRS - Centre National de la Recherche Scientifique, Labex ReFi - UP1 - Université Panthéon-Sorbonne, University of Ca’ Foscari [Venice, Italy])
    Abstract: The study of dependence between random variables is the core of theoretical and applied statistics. Static and dynamic copula models are useful for describing the dependence structure, which is fully encrypted in the copula probability density function. However, these models are not always able to describe the temporal change of the dependence patterns, which is a key characteristic of financial data. We propose a novel nonparametric framework for modelling a time series of copula probability density functions, which allows to forecast the entire function without the need of post-processing procedures to grant positiveness and unit integral. We exploit a suitable isometry that allows to transfer the analysis in a subset of the space of square integrable functions, where we build on nonparametric functional data analysis techniques to perform the analysis. The framework does not assume the densities to belong to any parametric family and it can be successfully applied also to general multivariate probability density functions with bounded or unbounded support. Finally, a noteworthy field of application pertains the study of time varying networks represented through vine copula models. We apply the proposed methodology for estimating and forecasting the time varying dependence structure between the S&P500 and NASDAQ indices.
    Keywords: nonparametric statistics,functional PCA,multivariate densities,copula,functional time series,forecast,unbounded support
    Date: 2018–03
  7. By: Van Nguyen, Phuong
    Abstract: The primary purpose of this paper is to compare the forecasting performance of a small open economy New Keynesian Dynamic Stochastic General Equilibrium (SOE-NK-DSGE) model with its closed-economy counterpart. Based on the quarterly Australian data, these two competing models are recursively estimated, and point forecasts for seven domestic variables are compared. Since Australia is a small open economy, global economic integration and financial linkage play an essential role in this country. However, the empirical findings indicate that the open economy model yields predictions that are less accurate than those from its closed economy counterpart. Two possible reasons could cause this failure of the SOE-NK-DSGE model: (1) misspecification of the foreign sector, and (2) a higher degree of estimation uncertainty. Thus, this research paper examines further how these two issues are associated with this practical problem. To this end, we perform two additional exercises in a new variant of the SOE-NK-DSGE and Bayesian VAR models. Consequently, the findings from these two exercises reveal that a combination of misspecification of the foreign sector and a higher degree of estimation uncertainty causes the failure of the open economy DSGE model in forecasting. Thus, one uses the SOE-NK-DSGE model for prediction with caution.
    Keywords: Small open economy New Keynesian DSGE model; Bayesian estimation; forecasting accuracy; RMSEs
    JEL: B22 C11 E37 E47
    Date: 2020–02
  8. By: Lior Sidi
    Abstract: Stock market prediction with forecasting algorithms is a popular topic these days where most of the forecasting algorithms train only on data collected on a particular stock. In this paper, we enriched the stock data with related stocks just as a professional trader would have done to improve the stock prediction models. We tested five different similarities functions and found co-integration similarity to have the best improvement on the prediction model. We evaluate the models on seven S&P stocks from various industries over five years period. The prediction model we trained on similar stocks had significantly better results with 0.55 mean accuracy, and 19.782 profit compare to the state of the art model with an accuracy of 0.52 and profit of 6.6.
    Date: 2020–02
  9. By: Grammig, Joachim; Hanenberg, Constantin; Schlag, Christian; Sönksen, Jantje
    Abstract: We assess financial theory-based and machine learning-implied measurements of stock risk premia by comparing the quality of their return forecasts. In the low signal-to-noise environment of a one month horizon, we find that it is preferable to rely on a theory-based approach instead of engaging in the computerintensive hyper-parameter tuning of statistical models. The theory-based approach also delivers a solid performance at the one year horizon, at which only one machine learning methodology (random forest) performs substantially better. We also consider ways to combine the opposing modeling philosophies, and identify the use of random forests to account for the approximation residuals of the theory-based approach as a promising hybrid strategy. It combines the advantages of the two diverging paths in the finance world.
    Keywords: stock risk premia,return forecasts,machine learning,theorybased return prediction
    JEL: C53 C58 G12 G17
    Date: 2020

This nep-for issue is ©2020 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.