nep-for New Economics Papers
on Forecasting
Issue of 2020‒08‒24
nine papers chosen by
Rob J Hyndman
Monash University

  1. Hybrid ARFIMA Wavelet Artificial Neural Network Model for DJIA Index Forecasting By Heni Boubaker; Giorgio Canarella; Rangan Guzpta; Stephen M. Miller
  2. Probabilistic Forecast Reconciliation: Properties, Evaluation and Score Optimisation By Anastasios Panagiotelis; Puwasala Gamakumara; George Athanasopoulos; Rob J Hyndman
  3. Nvidia’s stock returns prediction using machine learning techniques for time series forecasting problem By Marcin Chlebus; Michał Dyczko; Michał Woźniak
  4. Predicting the exchange rate path. The importance of using up-to-date observations in the forecasts By Håvard Hungnes
  5. Forecasting macroeconomic risk in real time: Great and Covid-19 Recessions By De Santis, Roberto A.; Van der Veken, Wouter
  6. "Budget Credibility of Subnational Governments: Analyzing the Fiscal Forecasting Errors of 28 States in India" By Lekha Chakraborty; Pinaki Chakraborty; Ruzel Shrestha
  7. Variable Selection in Macroeconomic Forecasting with Many Predictors By Zhenzhong Wang; Zhengyuan Zhu; Cindy Yu
  8. Oil Price and Exchange Rate Behaviour of the BRICS for Over a Century By Afees A. Salisu; Juncal Cunado; Kazeem Isah; Rangan Gupta
  9. China's Missing Pigs: Correcting China's Hog Inventory Data Using a Machine Learning Approach By Yongtong Shao; Minghao Li; Dermot J. Hayes; Wendong Zhang; Tao Xiong; Wei Xie

  1. By: Heni Boubaker (International University of Rabat); Giorgio Canarella (University of Nevada, Las Vegas); Rangan Guzpta (University of Pretoria); Stephen M. Miller (University of Nevada, Las Vegas)
    Abstract: This paper proposes a hybrid modelling approach for forecasting returns and volatilities of the stock market. The model, called ARFIMA-WLLWNN model, integrates the advantages of the ARFIMA model, the wavelet decomposition technique (namely, the discrete MODWT with Daubechies least asymmetric wavelet filter) and artificial neural network (namely, the LLWNN neural network). The model develops through a two-phase approach. In phase one, a wavelet decomposition improves the forecasting accuracy of the LLWNN neural network, resulting in the Wavelet Local Linear Wavelet Neural Network (WLLWNN) model. The Back Propagation (BP) and Particle Swarm Optimization (PSO) learning algorithms optimize the WLLWNN structure. In phase two, the residuals of an ARFIMA model of the conditional mean become the input to the WLLWNN model. The hybrid ARFIMA-WLLWNN model is evaluated using daily closing prices for the Dow Jones Industrial Average (DJIA) index over 01/01/2010 to 02/11/2020. The experimental results indicate that the PSO-optimized version of the hybrid ARFIMA-WLLWNN outperforms the LLWNN, WLLWNN, ARFIMA-LLWNN, and the ARFIMA-HYAPARCH models and provides more accurate out-of-sample forecasts over validation horizons of one, five and twenty-two days.
    Keywords: Wavelet decomposition, WLLWNN, Neural network, ARFIMA, HYGARCH
    JEL: C45 C58 G17
    Date: 2020–08
  2. By: Anastasios Panagiotelis; Puwasala Gamakumara; George Athanasopoulos; Rob J Hyndman
    Abstract: We develop a framework for prediction of multivariate data that follow some known linear constraints, such as the example where some variables are aggregates of others. This is particularly common when forecasting time series (predicting the future), but also arises in other types of prediction. For point prediction, an increasingly popular technique is reconciliation, whereby predictions are made for all series (so-called `base' predictions) and subsequently adjusted to ensure coherence with the constraints. This paper extends reconciliation from the setting of point prediction to probabilistic prediction. A novel definition of reconciliation is developed and used to construct densities and draw samples from a reconciled probabilistic prediction. In the elliptical case, it is proven that the true predictive distribution can be recovered from reconciliation even when the location and scale matrix of the base prediction are chosen arbitrarily. To find reconciliation weights, an objective function based on scoring rules is optimised. The energy and variogram scores are considered since the log score is improper in the context of comparing unreconciled to reconciled predictions, a result also proved in this paper. To account for the stochastic nature of the energy and variogram scores, optimisation is achieved using stochastic gradient descent. This method is shown to improve base predictions in simulation studies and in an empirical application, particularly when the base prediction models are severely misspecified. When misspecification is not too severe, extending popular reconciliation methods for point prediction can result in a similar performance to score optimisation via stochastic gradient descent. The methods described here are implemented in the ProbReco package for R.
    Keywords: scoring rules, probabilistic forecasting, hierarchical time series, stochastic gradient descent
    Date: 2020
  3. By: Marcin Chlebus (Faculty of Economic Sciences, University of Warsaw); Michał Dyczko (Faculty of Mathematics and Computer Science, Warsaw University of Technology); Michał Woźniak (Faculty of Economic Sciences, University of Warsaw)
    Abstract: The main aim of this paper was to predict daily stock returns of Nvidia Corporation company quoted on Nasdaq Stock Market. The most important problems in this research are: statistical specificity of return ratios i.e. time series might occur to be a white noise and the fact of necessity of applying many atypical machine learning methods to handle time factor influence. The period of study covered 07/2012 - 12/2018. Models used in this paper were: SVR, KNN, XGBoost, LightGBM, LSTM, ARIMA, ARIMAX. Features which, were used in models comes from such classes like: technical analysis, fundamental analysis, Google Trends entries, markets related to Nvidia. It was empirically proved that there is a possibility to construct prediction model of Nvidia daily return ratios which can outperform simple naive model. The best performance was obtained by SVR based on stationary attributes. Generally, it was shown that models based on stationary variables perform better than models based on stationary and non-stationary variables. Ensemble approach designed especially for time series failed to make an improvement in forecast precision. It seems that usage of machine learning models for the problem of time series with various explanatory variable classes brings good results.
    Keywords: nvidia, stock returns, machine learning, technical analysis, fundamental analysis, google trends, stationarity, ensembling
    JEL: C32 C38 C44 C51 C52 C61 C65 G11 G15
    Date: 2020
  4. By: Håvard Hungnes (Statistics Norway)
    Abstract: Central banks, private banks, statistical agencies and international organizations such as the IMF and OECD typically use information about the exchange rate some weeks before the publication date as the basis for their exchange rate forecasts. In this paper, we test if forecasts can be made more accurate by utilizing information about exchange rate movements closer to the publication date. To this end, we apply new tests for equal predictability and encompassing for path forecasts. We find that the date when the exchange rate forecast is based on is crucial and this finding should be taken into account when evaluating exchange rate forecasts. Using forecasts made by Statistics Norway over the period 2001 - 2016 we find that the random walk, when based on the exchange rate three days ahead of the publication date, encompassed the predicted path by Statistics Norway. However, when using the exchange rate two weeks before the publication deadline, which is the information used by Statistics Norway in practice when making their forecasts, the random walk path and the predicted exchange rate path by Statistics Norway have equal predictability.
    Keywords: Macroeconomic forecasts; Econometric models; Forecast performance; Forecast evaluation; Forecast comparison
    JEL: C53 F31
    Date: 2020–06
  5. By: De Santis, Roberto A.; Van der Veken, Wouter
    Abstract: We show that financial variables contribute to the forecast of GDP growth during the Great Recession, providing additional insights on both first and higher moments of the GDP growth distribution. If a recession is due to an unforeseen shock (such as the Covid-19 recession), financial variables serve policymakers in providing timely warnings about the severity of the crisis and the macroeconomic risk involved, because downside risks increase as financial stress and corporate spreads become tighter. We use quantile regression and the skewed t-distribution and evaluate the forecasting properties of models using out-of-sample metrics with real-time vintages. JEL Classification: C53, E23, E27, E32, E44
    Keywords: Covid-19 Recession, downside risks, Great Recession, non-linear models, real-time forecast
    Date: 2020–07
  6. By: Lekha Chakraborty; Pinaki Chakraborty; Ruzel Shrestha
    Abstract: Budget credibility, or the ability of governments to accurately forecast macro-fiscal variables, is crucial for effective public finance management. Fiscal marksmanship analysis captures the extent of errors in the budgetary forecasting. The fiscal rules can determine fiscal marksmanship, as effective fiscal consolidation procedures affect the fiscal behavior of the states in conducting the budgetary forecasts. Against this backdrop, applying Theil's technique, we analyze the fiscal forecasting errors for 28 states (except Telangana) in India for the period 2011-16. There is a heterogeneity in the magnitude of errors across subnational governments in India. The forecast errors in revenue receipts have been greater than revenue expenditure. Within revenue receipts, the errors are more significantly pronounced in the grants component. Within expenditure budgets, the errors in capital spending are found to be greater than revenue spending in all the states. Partitioning the sources of errors, we identified that the errors were more broadly random than due to systematic bias, except for a few crucial macro-fiscal variables where improving the forecasting techniques can provide better estimates.
    Keywords: Forecast Errors; Fiscal Policies; Fiscal Forecasting; Political Economy; Fiscal Marksmanship
    JEL: H6 E62 C53
    Date: 2020–07
  7. By: Zhenzhong Wang; Zhengyuan Zhu; Cindy Yu
    Abstract: In the data-rich environment, using many economic predictors to forecast a few key variables has become a new trend in econometrics. The commonly used approach is factor augment (FA) approach. In this paper, we pursue another direction, variable selection (VS) approach, to handle high-dimensional predictors. VS is an active topic in statistics and computer science. However, it does not receive as much attention as FA in economics. This paper introduces several cutting-edge VS methods to economic forecasting, which includes: (1) classical greedy procedures; (2) l1 regularization; (3) gradient descent with sparsification and (4) meta-heuristic algorithms. Comprehensive simulation studies are conducted to compare their variable selection accuracy and prediction performance under different scenarios. Among the reviewed methods, a meta-heuristic algorithm called sequential Monte Carlo algorithm performs the best. Surprisingly the classical forward selection is comparable to it and better than other more sophisticated algorithms. In addition, we apply these VS methods on economic forecasting and compare with the popular FA approach. It turns out for employment rate and CPI inflation, some VS methods can achieve considerable improvement over FA, and the selected predictors can be well explained by economic theories.
    Date: 2020–07
  8. By: Afees A. Salisu (Centre for Econometric & Allied Research, University of Ibadan, Ibadan, Nigeria); Juncal Cunado (Economics Department, University of Navarra, Spain); Kazeem Isah (Centre for Econometric & Allied Research, University of Ibadan, Ibadan, Nigeria); Rangan Gupta (Department of Economics, University of Pretoria, Pretoria, 0002, South Africa)
    Abstract: We attempt to predict the exchange rate returns of BRICS countries with the global oil price using large historical datasets for over a century extending from September 1859 to April 2020. We formulate a predictive model that accounts for the salient features of the predictor and the predicted series in line with the recent literature. We establish, with the aid of asymmetry, that oil price is a good predictor of exchange rate returns for all the net oil-importers (India, South Africa and China) and one of the two net oil-exporters (Russia). We also demonstrate with compelling in-sample and out-of-sample forecast results that accounting for the role of asymmetry is crucial for the oil-based model to beat the benchmark (historical average) model.
    Keywords: Oil price, Exchange rate return, BRICS, Asymmetry, Predictability, Forecast evaluation
    JEL: C22 C53 F31 Q47
    Date: 2020–07
  9. By: Yongtong Shao; Minghao Li; Dermot J. Hayes (Center for Agricultural and Rural Development (CARD)); Wendong Zhang (Center for Agricultural and Rural Development (CARD)); Tao Xiong; Wei Xie
    Abstract: Small sample size often limits forecasting tasks such as the prediction of production, yield, and consumption of agricultural products. Machine learning offers an appealing alternative to traditional forecasting methods. In particular, Support Vector Regression has superior forecasting performance in small sample applications. In this article, we introduce Support Vector Regression via an application to China's hog market. Since 2014, China's hog inventory data has experienced an abnormal decline that contradicts price and consumption trends. We use Support Vector Regression to predict the true inventory based on the price-inventory relationship before 2014. We show that, in this application with a small sample size, Support Vector Regression out-performs neural networks, random forest, and linear regression. Predicted hog inventory decreased by 3.9% from November 2013 to September 2017, instead of the 25.4% decrease in the reported data.
    Date: 2020–08

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.