nep-for New Economics Papers
on Forecasting
Issue of 2021‒01‒11
seven papers chosen by
Rob J Hyndman
Monash University

  1. Does Big Data Improve Financial Forecasting? The Horizon Effect By Olivier Dessaint; Thierry Foucault; Laurent Frésard
  2. OPEC News and Exchange Rate Forecasting Using Dynamic Bayesian Learning By Xin Sheng; Rangan Gupta; Afees A. Salisu; Elie Bouri
  3. Nonlinear Excess Demand Model for Electricity Price Prediction By Mehmet A. Soytas; Hasan M. Ertugrul; Talat Ulussever
  4. Modelling Realized Covariance Matrices: a Class of Hadamard Exponential Models By Bauwens, Luc; Otranto, Edoardo
  5. Forecasting Mid-price Movement of Bitcoin Futures Using Machine Learning By Akyildirim, Erdinc; Cepni, Oguzhan; Corbet, Shaen; Uddin, Gazi Salah
  6. Uncertainty and Predictability of Real Housing Returns in the United Kingdom: A Regional Analysis By Afees A. Salisu; Rangan Gupta; Ahamuefula E. Ogbonna; Mark E. Wohar
  7. Machine Learning Advances for Time Series Forecasting By Ricardo P. Masini; Marcelo C. Medeiros; Eduardo F. Mendes

  1. By: Olivier Dessaint (INSEAD); Thierry Foucault (HEC Paris - Finance Department); Laurent Frésard (Universita della Svizzera italiana (USI Lugano); Swiss Finance Institute)
    Abstract: We study how data abundance affects the informativeness of financial analysts' forecasts at various horizons. Analysts forecast short-term and long-term earnings and choose how much information to process about each horizon to minimize forecasting error, net of information processing costs. When the cost of obtaining short-term information drops (i.e., more data becomes available), analysts change their information processing strategy in a way that renders their short-term forecasts more informative but that possibly reduces the informativeness of their long-term forecasts. We provide empirical support for this prediction using a large sample of forecasts at various horizons and novel measures of analysts' exposure to abundant data. Data abundance can thus impair the quality of long-term financial forecasts.ty of long-term forecasts.
    Keywords: Big data, Financial analysts' forecasts, Forecasting horizon, Forecasts' informativeness, Social media
    JEL: D84 G14 G17 M41
    Date: 2020–11
  2. By: Xin Sheng (Lord Ashcroft International Business School, Anglia Ruskin University, Chelmsford, CM1 1SQ, United Kingdom); Rangan Gupta (Department of Economics, University of Pretoria, Pretoria, 0002, South Africa); Afees A. Salisu (Centre for Econometric & Allied Research, University of Ibadan, Ibadan, Nigeria); Elie Bouri (Adnan Kassar School of Business, Lebanese American University, Beirut, Lebanon)
    Abstract: We consider whether a newspaper article count index related to the Organization of the Petroleum Exporting Countries (OPEC), which rises in response to important OPEC meetings and events connected with OPEC production levels, contains predictive power for the foreign exchange rates of G10 countries. The applied Bayesian inference methodology synthesizes a wide array of established approaches to modelling exchange rate dynamics, whereby various vector-autoregressive models are considered. Monthly data from 1996:01 to 2020:08 (given an in-sample of 1986:02 to 1995:12), shows that incorporating the OPEC news-related index into the proposed methodology leads to statistical gains in out-of-sample forecasts.
    Keywords: OPEC News, Exchange Rate Forecasting, Bayesian Dynamic Learning
    JEL: C32 C53 Q41
    Date: 2021–01
  3. By: Mehmet A. Soytas (KFUPM Business School); Hasan M. Ertugrul (European University Institute, Department of Economics, Florence, Italy); Talat Ulussever (Energy Exchange Istanbul (EXIST))
    Abstract: Variety of models and estimation techniques have been proposed for electricity price forecasting in the literature. We contribute by introducing a dynamic forecasting model for hourly electricity prices based on nonlinear excess demand specification. Our modelling framework depends on the neoclassical price adjustment equation that necessitates prices adjust toward equilibrium at a rate that is proportional to the excess demand. We approximate the adjustment as a nonlinear (cubic) function of the excess demand which itself is modeled as a latent factor. We show that nonlinear relation of excess demand and price leads to a more accurate description of price evolution toward equilibrium and with this framework, the equilibrium forecast for the price is given by a nonlinear equation of the excess demand that can be modeled as a function of important variables of supply and demand. This generates an advantage to forecasters in employing all the information on supply and demand functions in price pre-diction, rather than simply modelling the price on an ad-hoc manner. We further develop a maximum likelihood estimator with excess demand defined as a normally distributed random variable conditional on observables. We demonstrate our likelihood estimator by using data from Turkish electricity market. Our modelling framework implies time varying volatility for prices which along with the nonlinear mean function, brings two important features of time series modelling dynamics together in parsimonious model.
    Date: 2020–12–20
  4. By: Bauwens, Luc (Université catholique de Louvain, LIDAM/CORE, Belgium); Otranto, Edoardo
    Abstract: Time series of realized covariance matrices can be modelled in the conditional autoregressive Wishart model family via dynamic correlations or via dynamic covariances. Extended parameterizations of these models are proposed, which imply a specific and time-varying impact parameter of the lagged realized covariance (or correlation) on the next conditional covariance (or correlation) of each asset pair. The proposed extensions guarantee the positive definiteness of the conditional covariance or correlation matrix with simple parametric restrictions, while keeping the number of parameters fixed or linear with respect to the number of assets. An empirical study on twenty-nine assets reveals that the extended models have superior forecasting performances than their simpler versions.
    Keywords: realized covariances ; dynamic covariances and correlations ; Hadamard exponential matrix
    JEL: C32 C58
    Date: 2020–11–24
  5. By: Akyildirim, Erdinc (Department of Mathematics, ETH, Zurich, Switzerland and University of Zurich, Department of Banking and Finance, Zurich, Switzerland and Department of Banking and Finance, Burdur Mehmet Akif Ersoy University, Burdur, Turkey); Cepni, Oguzhan (Department of Economics, Copenhagen Business School); Corbet, Shaen (DCU Business School, Dublin City University, Dublin 9, Ireland and School of Accounting, Finance and Economics, University of Waikato, New Zealand); Uddin, Gazi Salah (Department of Management and Engineering, Linköping University, 581 83 Linköping, Sweden)
    Abstract: In the aftermath of the global financial crisis and on-going COVID-19, investors face challenges in understanding price dynamics across assets. In this paper, we explore the applicability of a large scale comparison of machine learning algorithms (MLA) to predict mid-price movement for bitcoin futures prices. We use high-frequency intra-day data to evaluate the relative forecasting performances across various time-frequencies, ranging between 5-minutes and 60-minutes. The empirical analysis is based on six different specifications of MLA methods during periods of pandemic. The empirical results show that MLA outperforms the random walk and ARIMA forecasts in Bitcoin futures markets, which may have important implications in the decision-making process of predictability.
    Keywords: Cryptocurrency; Bitcoin futures; Machine learning; Covid-19; k-Nearest neighbors; Logistic regression; Naive bayes; Random forest; Support vector machine; Extreme gradient; Boosting
    JEL: C60 E50
    Date: 2020–12–22
  6. By: Afees A. Salisu (Centre for Econometric & Allied Research, University of Ibadan, Ibadan, Nigeria); Rangan Gupta (Department of Economics, University of Pretoria, Pretoria, 0002, South Africa); Ahamuefula E. Ogbonna (Centre for Econometric & Allied Research, University of Ibadan; Department of Statistics, University of Ibadan, Ibadan, Nigeria); Mark E. Wohar (College of Business Administration, University of Nebraska at Omaha, 6708 Pine Street, Omaha, NE 68182, USA)
    Abstract: The predictability of uncertainty for real housing returns in the United Kingdom is examined using regional data covering twelve (12) regions namely East Midlands, East of England, London, North East, North West, Northern Ireland, Scotland, South East, South West, Wales, West Midlands, Yorkshire and the Humber. We utilize both housing policy uncertainty (HPU) and economic policy uncertainty (EPU) data while we render analyses for three data samples - full sample and two sub-samples covering the periods before and after the emergence of global financial crisis (GFC). Relying on a predictive model that accounts for the salient characteristics of the data, we find a negative relationship between HPU and real housing returns, on the average, regardless of the region analysed. Also, the model that accounts for HPU outperforms the benchmark model that ignores it while controlling for relevant covariates further improves the forecast performance. Additional analyses involving the EPU measure depict lower predictive contents for house price movements relative to the HPU measure and therefore using sector-specific uncertainty measure is crucial for more precise forecasts of real housing returns.
    Keywords: Real Housing Returns, Economic Policy Uncertainty, United Kingdom, Predictability, Forecast Evaluation
    JEL: C32 C53 R31
    Date: 2021–01
  7. By: Ricardo P. Masini; Marcelo C. Medeiros; Eduardo F. Mendes
    Abstract: In this paper we survey the most recent advances in supervised machine learning and high-dimensional models for time series forecasting. We consider both linear and nonlinear alternatives. Among the linear methods we pay special attention to penalized regressions and ensemble of models. The nonlinear methods considered in the paper include shallow and deep neural networks, in their feed-forward and recurrent versions, and tree-based methods, such as random forests and boosted trees. We also consider ensemble and hybrid models by combining ingredients from different alternatives. Tests for superior predictive ability are briefly reviewed. Finally, we discuss application of machine learning in economics and finance and provide an illustration with high-frequency financial data.
    Date: 2020–12

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.