nep-for New Economics Papers
on Forecasting
Issue of 2020‒10‒26
fourteen papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting Stock Returns with Large Dimensional Factor Models By Alessandro Giovannelli; Daniele Massacci; Stefano Soccorsi
  2. Nowcasting GDP growth using data reduction methods: Evidence for the French economy By Olivier Darne; Amelie Charles
  3. Optimal probabilistic forecasts: When do they work? By Gael M. Martin; Rub\'en Loaiza-Maya; David T. Frazier; Worapree Maneesoonthorn; Andr\'es Ram\'irez Hassan
  4. Ordinal-response models for irregularly spaced transactions: A forecasting exercise By Dimitrakopoulos, Stefanos; Tsionas, Mike G.; Aknouche, Abdelhakim
  5. Realized Volatility Forecasting Based on Dynamic Quantile Model Averaging By Zongwu Cai; Chaoqun Ma; Xianhua Mi
  6. Encompassing Tests for Value at Risk and Expected Shortfall Multi-Step Forecasts based on Inference on the Boundary By Timo Dimitriadis; Xiaochun Liu; Julie Schnaitmann
  7. Time Series Analyses of Global Oil Prices: Shocks, Effects and Predictability By Ruths Sion, Sebastian
  8. Regional Heterogeneity and U.S. Presidential Elections By Rashad Ahmed; M. Hashem Pesaran
  9. Asset Price Forecasting using Recurrent Neural Networks By Hamed Vaheb
  10. Stock Price Prediction Using Machine Learning and LSTM-Based Deep Learning Models By Sidra Mehtab; Jaydip Sen; Abhishek Dutta
  11. Modeling and Probababilistic Forecasting of Natural Gas Prices By Jonathan Berrisch; Florian Ziel
  12. Forecasting Charge-Off Rates with a Panel Tobit Model: The Role of Uncertainty By Xin Sheng; Rangan Gupta; Qiang Ji
  13. Can Households Predict where the Macroeconomy is Headed? By Kladivko, Kamil; Österholm, Pär
  14. Forecast of Ontario’s housing stock 2020-2046 By Karimova, Amira

  1. By: Alessandro Giovannelli; Daniele Massacci; Stefano Soccorsi
    Abstract: We study equity premium out-of-sample predictability by extracting the information contained in a high number of macroeconomic predictors via large dimensional factor models. We compare the well known factor model with a static representation of the common components with a more general model known as the Generalized Dynamic Factor Model. Using statistical and economic evaluation criteria, we empirically show that the Generalized Dynamic Factor Model helps predicting the equity premium. Exploiting the link between business cycle and return predictability, we find more accurate predictions by combining rolling and recursive forecasts in real-time, with promising results in the aftermath of the Great Financial Crisis.
    Keywords: Stock Returns Forecasting, Factor Model, Large Data Sets, Forecast Evaluation
    JEL: C38 C53 C55 G11 G17
    Date: 2020
  2. By: Olivier Darne (LEMNA - Laboratoire d'économie et de management de Nantes Atlantique - IUML - FR 3473 Institut universitaire Mer et Littoral - UBS - Université de Bretagne Sud - UM - Le Mans Université - UA - Université d'Angers - CNRS - Centre National de la Recherche Scientifique - IFREMER - Institut Français de Recherche pour l'Exploitation de la Mer - UN - Université de Nantes - ECN - École Centrale de Nantes - IEMN-IAE Nantes - Institut d'Économie et de Management de Nantes - Institut d'Administration des Entreprises - Nantes - UN - Université de Nantes); Amelie Charles (Audencia Business School)
    Abstract: In this paper, we propose bridge models to nowcast French gross domestic product (GDP) quarterly growth rate. The bridge models, allowing economic interpretations, are specified by using a machine learning approach via Lasso-based regressions and by an econometric approach based on an automatic general-to-specific procedure. These approaches allow to select explanatory variables among a large data set of soft data. A recursive forecast study is carried out to assess the forecasting performance. It turns out that the bridge models constructed using the both variable-selection approaches outperform benchmark models and give similar performance in the out-of-sample forecasting exercise. Finally, the combined forecasts of these both approaches display interesting forecasting performance.
    Date: 2020–09
  3. By: Gael M. Martin; Rub\'en Loaiza-Maya; David T. Frazier; Worapree Maneesoonthorn; Andr\'es Ram\'irez Hassan
    Abstract: Proper scoring rules are used to assess the out-of-sample accuracy of probabilistic forecasts, with different scoring rules rewarding distinct aspects of forecast performance. Herein, we re-investigate the practice of using proper scoring rules to produce probabilistic forecasts that are `optimal' according to a given score, and assess when their out-of-sample accuracy is superior to alternative forecasts, according to that score. Particular attention is paid to relative predictive performance under misspecification of the predictive model. Using numerical illustrations, we document several novel findings within this paradigm that highlight the important interplay between the true data generating process, the assumed predictive model and the scoring rule. Notably, we show that only when a predictive model is sufficiently compatible with the true process to allow a particular score criterion to reward what it is designed to reward, will this approach to forecasting reap benefits. Subject to this compatibility however, the superiority of the optimal forecast will be greater, the greater is the degree of misspecification. We explore these issues under a range of different scenarios, and using both artificially simulated and empirical data.
    Date: 2020–09
  4. By: Dimitrakopoulos, Stefanos; Tsionas, Mike G.; Aknouche, Abdelhakim
    Abstract: We propose a new model for transaction data that accounts jointly for the time duration between transactions and for the discreteness of the intraday stock price changes. Duration is assumed to follow a stochastic conditional duration model, while price discreteness is captured by an autoregressive moving average ordinal-response model with stochastic volatility and time-varying parameters. The proposed model also allows for endogeneity of the trade durations as well as for leverage and in-mean effects. In a purely Bayesian framework we conduct a forecasting exercise using multiple high-frequency transaction data sets and show that the proposed model produces better point and density forecasts than competing models.
    Keywords: Ordinal-response models, irregularly spaced data, stochastic conditional duration, time varying ARMA-SV model, Bayesian MCMC, model confidence set.
    JEL: C1 C11 C15 C4 C41 C5 C51 C53 C58
    Date: 2020–10–01
  5. By: Zongwu Cai (Department of Economics, The University of Kansas, Lawrence, KS 66045, USA); Chaoqun Ma (School of Business, Hunan University, Changsha, Hunan 410082, China); Xianhua Mi (School of Business, Hunan University, Changsha, Hunan 410082, China)
    Abstract: Heterogeneity, volatility persistence, leverage effect and fat right tails are the most documented stylized features of realized volatility (RV), which introduce substantial difficulties in econometric modeling that requires some rigid distributional assumptions. To accommodate these features without making these assumptions, we study the quantile forecasting of RV by proposing five novel dynamic model averaging strategies designed to combine individual quantile models, termed as dynamic quantile model averaging (DQMA). The empirical results of analyzing high-frequency price data of the S&P 500 index clearly indicate that the stylized facts of RV can be captured by different quantiles, with stronger effects at high-level quantiles. Therefore, DQMA can not only reduce the risk of model uncertainty but also generate more accurate and robust out-of-sample quantile forecasts than those of individual heterogeneous autoregressive quantile models.
    Keywords: Dynamic moving averaging; Model uncertainty; Fat tails; Heterogeneity; Quantile regression; Realized volatility; Time-varying parameters.
    JEL: C12 C13 C14 C23
    Date: 2020–09
  6. By: Timo Dimitriadis; Xiaochun Liu; Julie Schnaitmann
    Abstract: We propose forecast encompassing tests for the Expected Shortfall (ES) jointly with the Value at Risk (VaR) based on flexible link (or combination) functions. Our setup allows testing encompassing for convex forecast combinations and for link functions which preclude crossings of the combined VaR and ES forecasts. As the tests based on these link functions involve parameters which are on the boundary of the parameter space under the null hypothesis, we derive and base our tests on nonstandard asymptotic theory on the boundary. Our simulation study shows that the encompassing tests based on our new link functions outperform tests based on unrestricted linear link functions for one-step and multi-step forecasts. We further illustrate the potential of the proposed tests in a real data analysis for forecasting VaR and ES of the S&P 500 index.
    Date: 2020–09
  7. By: Ruths Sion, Sebastian
    Abstract: This dissertation considers different aspects of crude oil research, primarily based on four independent empirical analyses, interconnected through a common denominator: Time-series analysis methods applied to global oil prices. The first three chapters are of introductory nature. They present the developments on global oil markets since the end of World War II and review the literature on crude oil. More importantly, they show how to estimate global models using vector autoregressive (VAR) and structural vector autoregressive (SVAR) models. The latter allow for the disentanglement and estimation of unexpected oil price shocks required for later analyses. The first analysis reviews the question, originally at the center of economic research on crude oil: How are macroeconomic performance and oil price shocks interrelated? New insights based on longer sample series as well as developments in SVAR models allow to complement the existing literature by estimating global models of oil. Based on a broad set of monthly macroeconomic variables for the United States and Germany, the analysis shows that these two industrialized economies react differently to oil price shocks. The disentanglement of the underlying causes of unexpected oil price movements is crucial. The second empirical analysis concerns the effects of oil embargoes against oil producing countries The same SVAR models are applied in the framework of the sanctions that were imposed on Iran by the international community late 2011 and early 2012. The estimation results show that the direct effects of the Iran sanctions on global oil prices were limited and temporary. By estimating and analyzing the unexpected oil price changes before the implementation of sanctions, we find evidence that sanctions might have important price increasing effects through market expectations long before their official implementation. Departing from the same global model that includes the real price of crude oil as an endogenous variable, the third analysis is concerned with its oil price forecasting properties. We are able to improve the forecasting accuracy by applying regularization methods for variable selection. Originating from the machine learning literature, these methods are now widely used in economic research, especially in cases, where a large number of variables are included in the model. Furthermore, typical lag selection methods, used in the estimation of global models of oil are compared. Finally, the core variable set is augmented by a wide range of possibly relevant regressors as suggested by the literature. The fourth and final analysis concerns another aspect of oil price forecasting when using crude oil futures as forecasts for the spot price of oil. We estimate whether forecasting preferences are asymmetric in a sense that a positive forecast error has a different cost than a negative forecast error of the same magnitude. Using different model specifications and a wide range of instrument sets inspired by the literature on futures, we find robust evidence for asymmetric loss. The market has a preference to underestimate the spot price of crude oil through futures pricing. This indicates the existence of a risk premium on crude oil futures.
    Date: 2020
  8. By: Rashad Ahmed; M. Hashem Pesaran
    Abstract: This paper develops a recursive model of voter turnout and voting outcomes at the U.S. county level to investigate the socioeconomic determinants of recent U.S. presidential elections. It exploits cross-section variations across U.S. counties and investigates the key determinants of the 2016 Presidential Election by allowing for regional heterogeneity and using high-dimensional variable selection algorithms such as Lasso and OCMT. It is shown that the relationship between many socioeconomic variables and voting outcomes are not uniform across U.S. regions. Specifically, allowing for regional heterogeneity explains the unexpected 2016 Republican victory. Moreover, incorporating regional heterogeneity improves electoral predictability among key swing states. Important factors explaining voting outcomes include incumbency effects, voter turnout, local economic performance, unemployment, poverty, educational attainment, house price changes, urban-rural scores, and international competitiveness. Our results also corroborate evidence of ‘short-memory’ among voters: economic fluctuations realized a few months prior to the election are indeed powerful predictors of voting outcomes as compared to their longer-term analogues. The paper also reports forecasts for the 2020 U.S. Presidential Election based on data available at the end of July 2020. The regional models predict a close electoral college outcome. The predictions are split: the Lasso-regional model forecasts a narrow Democratic electoral victory, while the OCMT-regional model forecasts a narrow Republican victory. All models point towards the Democratic candidate winning the popular vote.
    Keywords: voter turnout, popular and electoral college votes, simultaneity and recursive identification, high dimensional forecasting models, Lasso, OCMT
    JEL: C53 C55 D72
    Date: 2020
  9. By: Hamed Vaheb
    Abstract: This thesis serves three primary purposes, first of which is to forecast two stocks, i.e. Goldman Sachs (GS) and General Electric (GE). In order to forecast stock prices, we used a long short-term memory (LSTM) model in which we inputted the prices of two other stocks that lie in rather close correlation with GS. Other models such as ARIMA were used as benchmark. Empirical results manifest the practical challenges when using LSTM for forecasting stocks. One of the main upheavals was a recurring lag which we called "forecasting lag". The second purpose is to develop a more general and objective perspective on the task of time series forecasting so that it could be applied to assist in an arbitrary that of forecasting by ANNs. Thus, attempts are made for distinguishing previous works by certain criteria so as to summarise those including effective information. The summarised information is then unified and expressed through a common terminology that can be applied to different steps of a time series forecasting task. The last but not least purpose of this thesis is to elaborate on a mathematical framework on which ANNs are based. We are going to use the framework introduced in the book "Neural Networks in Mathematical Framework" by Anthony L. Caterini in which the structure of a generic neural network is introduced and the gradient descent algorithm (which incorporates backpropagation) is introduced in terms of their described framework. In the end, we use this framework for a specific architecture, which is recurrent neural networks on which we concentrated and our implementations are based. The book proves its theorems mostly for classification case. Instead, we proved theorems for regression case, which is the case of our problem.
    Date: 2020–10
  10. By: Sidra Mehtab; Jaydip Sen; Abhishek Dutta
    Abstract: Prediction of stock prices has been an important area of research for a long time. While supporters of the efficient market hypothesis believe that it is impossible to predict stock prices accurately, there are formal propositions demonstrating that accurate modeling and designing of appropriate variables may lead to models using which stock prices and stock price movement patterns can be very accurately predicted. In this work, we propose an approach of hybrid modeling for stock price prediction building different machine learning and deep learning-based models. For the purpose of our study, we have used NIFTY 50 index values of the National Stock Exchange (NSE) of India, during the period December 29, 2014 till July 31, 2020. We have built eight regression models using the training data that consisted of NIFTY 50 index records during December 29, 2014 till December 28, 2018. Using these regression models, we predicted the open values of NIFTY 50 for the period December 31, 2018 till July 31, 2020. We, then, augment the predictive power of our forecasting framework by building four deep learning-based regression models using long-and short-term memory (LSTM) networks with a novel approach of walk-forward validation. We exploit the power of LSTM regression models in forecasting the future NIFTY 50 open values using four different models that differ in their architecture and in the structure of their input data. Extensive results are presented on various metrics for the all the regression models. The results clearly indicate that the LSTM-based univariate model that uses one-week prior data as input for predicting the next week open value of the NIFTY 50 time series is the most accurate model.
    Date: 2020–09
  11. By: Jonathan Berrisch; Florian Ziel
    Abstract: In this paper, we examine the problem of modeling and forecasting European Day-Ahead and Month-Ahead natural gas prices. For this, we propose two distinct probabilistic models that can be utilized in risk- and portfolio management. We use daily pricing data ranging from 2011 to 2020. Extensive descriptive data analysis shows that both time series feature heavy tails, conditional heteroscedasticity, and show asymmetric behavior in their differences. We propose state-space time series models under skewed, heavy-tailed distribution to capture all stylized facts in the data. They include the impact of autocorrelation, seasonality, risk premia, temperature, storage levels, the price of European Emission Allowances, and related fuel prices of oil, coal, and electricity. We provide a rigorous model diagnostic and interpret all model components in detail. Additionally, we conduct a probabilistic forecasting study with significance test and compare the predictive performance against literature benchmarks. The proposed Day-Ahead (Month-Ahead) model leads to a $13\%$ ($9$\%) reduction in out of sample CRPS compared to the best performing benchmark model, mainly due to adequate modeling of the volatility and heavy tails.
    Date: 2020–10
  12. 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, South Africa); Qiang Ji (Institutes of Science and Development, Chinese Academy of Sciences, Beijing 100190, China)
    Abstract: Based on a large panel dataset of small commercial banks in the United States, this paper employs a dynamic panel Tobit model to analyze the role of uncertainty in forecasting charge-off rates on loans for credit card (CC) and residential real estate (RRE). When compared to other standard predictors, such as house prices and unemployment rates, we find thatthe effect of uncertainty changes on charge-off rates is more pronounced. Furthermore, it is evident that including heteroskedasticity in the model specification leads to more accurate forecasts.
    Keywords: loan charge-offs, panel data, Tobit model, forecasting
    JEL: C11 C23 C53 G21
    Date: 2020–10
  13. By: Kladivko, Kamil (Örebro University School of Business); Österholm, Pär (Örebro University School of Business)
    Abstract: In this paper, we evaluate households’ directional forecasts of inflation and the unemployment rate in Sweden. The analysis is conducted using monthly forecasts from the National Institute of Economic Research’s Economic Tendency Survey that range from January 1996 until August 2019. Results indicate that households have statistically significant ability to forecast where the unem-ployment is headed, but they fail in predicting the direction of future inflation.
    Keywords: Survey data; Directional forecasts; Inflation; Unemployment
    JEL: E37
    Date: 2020–10–09
  14. By: Karimova, Amira
    Abstract: This study estimates Ontario’s housing stock by Ontario’s Central Metropolitan Areas and by residential dwelling types between 2020-2043. It also evaluates existing housing stock and makes future projections taking into account both the supply-side and demand-side perspectives, as well as the impacts of the Covid-19 pandemic that can lead to various market scenarios.
    Keywords: Ontario, housing stock, infrastructure, real estate, residential
    JEL: R31 R38
    Date: 2020–08–08

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.