nep-for New Economics Papers
on Forecasting
Issue of 2019‒10‒21
fourteen papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting Swiss Exports Using Bayesian Forecast Reconciliation By Florian Eckert; Rob J Hyndman; Anastasios Panagiotelis
  2. Futures-based forecasts: How useful are they for oil price volatility forecasting? By Chatziantoniou, Ioannis; Degiannakis, Stavros; Filis, George
  3. Forecasting the Albanian short-term inflation through a Bayesian VAR model By Meri Papavangjeli
  4. Optimal Non-negative Forecast Reconciliation By Shanika L Wickramasuriya; Berwin A Turlach; Rob J Hyndman
  5. Peramalan Penjualan Pupuk Menggunakan Metode Trend Moment By Ulfa, Ulia; Sumijan, Sumijan; Nurcahyo, Gunadi Widi
  6. Large hybrid time-varying parameter VARs By Joshua C.C. Chan
  7. Incentive-driven Inattention By Gaglianone, Wagner Piazza; Giacomini, Raffaela; Issler, João Victor; Skreta, Vasiliki
  8. Nowcasting and forecasting US recessions: Evidence from the Super Learner By Maas, Benedikt
  9. A term structure model under cyclical fluctuations in interest rates By Manuel Moreno; Alfonso Novales; Federico Platania
  10. Sonic Thunder vs Brian the Snail : Fast-sounding racehorse names and prediction accuracy in betting exchange markets By Oliver Merz; Raphael Flepp; Egon Franck
  11. Seasonal Functional Autoregressive Models By Atefeh Zamani; Hossein Haghbin; Maryam Hashemi; Rob J Hyndman
  12. Variational Bayesian Inference in Large Vector Autoregressions with Hierarchical Shrinkage By Deborah Gefang; Gary Koop; Aubrey Poon
  13. Residual Switching Network for Portfolio Optimization By Jifei Wang; Lingjing Wang
  14. Predicting Auction Price of Vehicle License Plate with Deep Residual Learning By Vinci Chow

  1. By: Florian Eckert; Rob J Hyndman; Anastasios Panagiotelis
    Abstract: This paper conducts an extensive forecasting study on 13,118 time series measuring Swiss goods exports, grouped hierarchically by export destination and product category. We apply existing state of the art methods in forecast reconciliation and introduce a novel Bayesian reconciliation framework. This approach allows for explicit estimation of reconciliation biases, leading to several innovations: Prior judgment can be used to assign weights to specific forecasts and the occurrence of negative reconciled forecasts can be ruled out. Overall we find strong evidence that in addition to producing coherent forecasts, reconciliation also leads to improvements in forecast accuracy.
    Keywords: hierarchical forecasting, Bayesian forecast reconciliation, Swiss exports, optimal forecast combination
    JEL: C32 C53 E17
    Date: 2019
  2. By: Chatziantoniou, Ioannis; Degiannakis, Stavros; Filis, George
    Abstract: Oil price volatility forecasts have recently attracted the attention of many studies in the energy finance field. The literature mainly concentrates its attention on the use of daily data, using GARCH-type models. It is only recently that efforts to use more informative intraday data to forecast oil price realized volatility have been made. Despite all these previous efforts, no study has examined the usefulness of futures-based models for oil price realized volatility forecasting, although the use of such models is extensive for oil price predictions. This study fills this void and shows that futures-based forecasts based on intra-day data provide informative forecasts for horizons that span between 1-day and 66-days ahead. More importantly, these results hold true even during turbulent times for the oil market, such as the Global Financial Crisis of 2007-09 and the oil collapse period of 2014-15.
    Keywords: Brent crude oil, realized volatility, forecasting, futures-based forecasts
    JEL: C22 C53 G13 Q47
    Date: 2019
  3. By: Meri Papavangjeli
    Abstract: In the context of the Bank of Albania’s primary objective of achieving and maintaining price stability, generating accurate and reliable forecasts for the future rate of inflation is a necessity for its successful realization. This paper aims to enrich the Bank’s portfolio of short-term inflation forecasting tools through the construction of a Bayesian vector autoregressive (BVAR) model, which unlike standard autoregressive vector (VAR) models, addresses the overparameterization problem, allowing for the inclusion of more endogenous variables, and in this way enabling a more comprehensive explanation of inflation. Several univariate models are estimated to forecast short-term inflation, such as: unconditional mean, random walk, autoregressive integrated moving average (ARIMA) models, and the best performing among them is used as a benchmark to evaluate the forecast performance of the BVAR model. In addition, an unrestricted VAR - the most commonly used tool to obtain projections of the main economic indicators - is constructed as an additional benchmark, based solely on the information that the data series provides. The results show that the BVAR approach, which incorporates more economic information, outperforms the benchmark univariate and the unrestricted VAR models in the different time horizons of the forecast sample, but the differences between models in terms of their forecast performance are not statistically significant.
    Keywords: Bayesian estimation, vector autoregressive, forecasting performance
    JEL: C30 C52 C53 C80
    Date: 2019–10–09
  4. By: Shanika L Wickramasuriya; Berwin A Turlach; Rob J Hyndman
    Abstract: The sum of forecasts of a disaggregated time series are often required to equal the forecast of the aggregate. The least squares solution for finding coherent forecasts uses a reconciliation approach known as MinT, proposed by Wickramasuriya, Athanasopoulos and Hyndman (2019). The MinT approach and its variants do not guarantee that the coherent forecasts are nonnegative, even when all of the original forecasts are non-negative in nature. This has become a serious issue in applications that are inherently non-negative such as with sales data or tourism numbers. While overcoming this difficulty, we consider the analytical solution of MinT as a least squares minimization problem. The non-negativity constraints are then imposed on the minimization problem to ensure that the coherent forecasts are strictly non-negative. Considering the dimension and sparsity of the matrices involved, and the alternative representation of MinT, this constrained quadratic programming problem is solved using three algorithms. They are the block principal pivoting algorithm, projected conjugate gradient algorithm, and scaled gradient projection algorithm. A Monte Carlo simulation is performed to evaluate the computational performances of these algorithms. The results demonstrate that the block principal pivoting algorithm clearly outperforms the rest, and projected conjugate gradient is the second best. The superior performance of the block principal pivoting algorithm can be partially attributed to the alternative representation of the weight matrix in the MinT approach. An empirical investigation is carried out to assess the impact of imposing non-negativity constraints on forecast reconciliation. It is observed that slight gains in forecast accuracy have occurred at the most disaggregated level. At the aggregated level slight losses are also observed. Although the gains or losses are negligible, the procedure plays an important role in decision and policy implementation processes.
    Keywords: aggregation, Australian tourism, coherent forecasts, contemporaneous error correlation, forecast combinations, least squares, non-negative, spatial correlations, reconciliation
    Date: 2019
  5. By: Ulfa, Ulia; Sumijan, Sumijan; Nurcahyo, Gunadi Widi
    Abstract: Aneka Tani Mandiri Trade Unit is a fertilizer sales shop in the city of Padang. From year to year sales of fertilizers in UD. Aneka Tani Mandiri experiences fluctuations where it is difficult to predict sales increases and decreases every month. The problem that most often occurs in this store is often experiencing shortages and excess stock of goods, this is very likely to occur because many of its items are not sold out and many items are needed by consumers but insufficient stock of goods. Another result is that the profits from the store should be more reduced, with this problem the store must be able to predict how many items will be sold and how many items must be provided in the following month, by knowing the number of items to be sold, the deficiency or excess stock of goods can be avoided. So for that the research was conducted using the Trend Moment Method to predict and predict fertilizer stock that will be provided for the following month. So that will increase sales turnover of the store. By building a fertilizer sales forecasting system using the Trend Moment method which is assisted by the PHP and MySQL programming languages can produce ZA fertilizer sales predictions with success rates above 75%
    Keywords: Method of Trend Moment, Fertilizer, Forecasting, Fluctuation, Stock.
    JEL: H00
    Date: 2019–10–14
  6. By: Joshua C.C. Chan
    Abstract: Time-varying parameter VARs with stochastic volatility are routinely used for structural analysis and forecasting in settings involving a few macroeconomic variables. Applying these models to high-dimensional datasets has proved to be challenging due to intensive computations and over-parameterization concerns. We develop an efficient Bayesian sparsification method for a class of models we call hybrid TVP-VARs - VARs with time-varying parameters in some equations but constant coefficients in others. Specifically, for each equation, the new method automatically decides (i) whether the VAR coefficients are constant or time-varying, and (ii) whether the error variance is constant or has a stochastic volatility specification. Using US datasets of various dimensions, we find evidence that the VAR coefficients and error variances in some, but not all, equations are time varying. These large hybrid TVP-VARs also forecast better than standard benchmarks.
    Keywords: large vector autoregression, time-varying parameter, stochastic volatility, trend output growth, macroeconomic forecasting
    JEL: C11 C52 E37 E47
    Date: 2019–10
  7. By: Gaglianone, Wagner Piazza; Giacomini, Raffaela; Issler, João Victor; Skreta, Vasiliki
    Abstract: “Rational inattention” is becoming increasingly prominent in economic modelling, but there is little empirical evidence for its central premise–that the choice of attention results from a cost-benefit optimization. Observational data typically do not allow researchers to infer attention choices from observables. We fill this gap in the literature by exploiting a unique dataset of professional forecasters who update their inflation forecasts at days of their choice. In the data we observe how many forecasters update (extensive margin of updating), the magnitude of the update (intensive margin), and the objective of optimization (forecast accuracy). There are also “shifters” in incentives: A contest that increases the benefit of accurate forecasting, and the release of official data that reduces the cost of information acquisition. These features allow us to link observables to attention and incentive parameters. We structurally estimate a model where the decision to update and the magnitude of the update are endogenous and the latter is the outcome of a rational inattention optimization. The model fits the data and gives realistic predictions. We find that shifts in incentives affect both extensive and intensive margins, but the shift in benefits from the contest has the largest aggregate effect. Counterfactuals reveal that accuracy is maximized if the contest coincides with the release of information, aligning higher benefits with lower costs of attention.
    Date: 2019–10–10
  8. By: Maas, Benedikt
    Abstract: This paper introduces the Super Learner to nowcast and forecast the probability of a US economy recession in the current quarter and future quarters. The Super Learner is an algorithm that selects an optimal weighted average from several machine learning algorithms. In this paper, elastic net, random forests, gradient boosting machines and kernel support vector machines are used as underlying base learners of the Super Learner, which is trained with real-time vintages of the FRED-MD database as input data. The Super Learner’s ability to categorise future time periods into recessions versus expansions is compared with eight different alternatives based on probit models. The relative model performance is evaluated based on receiver operating characteristic (ROC) curves. In summary, the Super Learner predicts a recession very reliably across all forecast horizons, although it is defeated by different individual benchmark models on each horizon.
    Keywords: Machine Learning; Nowcasting; Forecasting; Business cycle analysis
    JEL: C32 C53 E32
    Date: 2019–09
  9. By: Manuel Moreno (Department of Economic Analysis and Finance, University of Castilla-La Mancha, Toledo, Spain.); Alfonso Novales (Instituto Complutense de Análisis Económico (ICAE), and Department of Economic Analysis, Facultad de Ciencias Económicas y Empresariales, Universidad Complutense, 28223 Madrid, Spain.); Federico Platania (Léonard de Vinci Pôle Universitaire, Paris La Défense, France.)
    Abstract: We propose a flexible yet tractable model of the term structure of interest rates (TSIR). Term structure models attempt to explain how interest rates depend on their maturities at a given point in time, characterizing the rela- tionship between short-term and long-term rates. Our model can reproduce and fit a variety of TSIR shapes by capturing cyclical fluctuations of interest rates, different monetary policy reactions as witnessed pre- and post-crisis as well as the effect of the business cycle or exogenous shocks. Our modelling approach also provides a characterization of long-term fluctuations in the mean level of interest rates unveiling the effects of monetary policy in- terventions in interest rates. Furthermore, using daily US data, we compare the empirical ability of our model to both fit and forecast the TSIR under different economic scenarios. We show that our model improves pricing and risk management by fitting and predicting interest rates more accurately and precisely than do existing TSIR models.
    Keywords: Term structure of interest rates; cyclical fluctuations; bond pricing; TSIR fitting performance, interest rates forecast.
    JEL: D53 E43 G13 C58 E32 C31
    Date: 2019–09
  10. By: Oliver Merz (Department of Business Administration, University of Zurich); Raphael Flepp (Department of Business Administration, University of Zurich); Egon Franck (Department of Business Administration, University of Zurich)
    Abstract: This paper examines the influence of objectively irrelevant information on prediction accuracy in horse-racing betting exchange markets. In horse racing, the name of a horse does not depend on the horse’s performance and is thus uninformative. We investigate the impact of fast-sounding horse names on prediction market price accuracy and betting returns. Using over 3 million horse bets, we find evidence that the winning probabilities of bets on horses with fast-sounding names are overstated, which impairs the prediction accuracy of such bets. This finding implies that the prices in betting exchange markets are not efficient, as prices become distorted by incorporating the misleading information from a horse’s fast-sounding name. This bias translates into significantly lower betting returns for horses classified as fast-sounding compared to the returns of all other horses.
    Keywords: Market efficiency, Sports forecasting, Prediction markets, Betting industry, Horse Racing
    JEL: D40 C53 L83
    Date: 2019–10
  11. By: Atefeh Zamani; Hossein Haghbin; Maryam Hashemi; Rob J Hyndman
    Abstract: Functional autoregressive models are popular for functional time series analysis, but the standard formulation fails to address seasonal behaviour in functional time series data. To overcome this shortcoming, we introduce seasonal functional autoregressive time series models. For the model of order one, we derive sufficient stationarity conditions and limiting behavior, and provide estimation and prediction methods. Some properties of the general order P model are also presented. The merits of these models are demonstrated using simulation studies and via an application to real data.
    Keywords: functional time series analysis, seasonal functional autoregressive model, central limit theorem, prediction, estimation
    JEL: C32 C14
    Date: 2019
  12. By: Deborah Gefang; Gary Koop; Aubrey Poon
    Abstract: Many recent papers in macroeconomics have used large Vector Autoregressions (VARs) involving a hundred or more dependent variables. With so many parameters to estimate, Bayesian prior shrinkage is vital in achieving reasonable results. Computational concerns currently limit the range of priors used and render difficult the addition of empirically important features such as stochastic volatility to the large VAR. In this paper, we develop variational Bayes methods for large VARs which overcome the computational hurdle and allow for Bayesian inference in large VARs with a range of hierarchical shrinkage priors and with time-varying volatilities. We demonstrate the computational feasibility and good forecast performance of our methods in an empirical application involving a large quarterly US macroeconomic data set.
    Keywords: Variational inference, Vector Autoregression, Stochastic Volatility, Hierarchical Prior, Forecasting
    JEL: C11 C32 C53
    Date: 2019–05
  13. By: Jifei Wang; Lingjing Wang
    Abstract: This paper studies deep learning methodologies for portfolio optimization in the US equities market. We present a novel residual switching network that can automatically sense changes in market regimes and switch between momentum and reversal predictors accordingly. The residual switching network architecture combines two separate residual networks (ResNets), namely a switching module that learns stock market conditions, and the main module that learns momentum and reversal predictors. We demonstrate that over-fitting noisy financial data can be controlled with stacked residual blocks and further incorporating the attention mechanism can enhance powerful predictive properties. Over the period 2008 to H12017, the residual switching network (Switching-ResNet) strategy verified superior out-of-sample performance with an average annual Sharpe ratio of 2.22, compared with an average annual Sharpe ratio of 0.81 for the ANN-based strategy and 0.69 for the linear model.
    Date: 2019–10
  14. By: Vinci Chow
    Abstract: Due to superstition, license plates with desirable combinations of characters are highly sought after in China, fetching prices that can reach into the millions in government-held auctions. Despite the high stakes involved, there has been essentially no attempt to provide price estimates for license plates. We present an end-to-end neural network model that simultaneously predict the auction price, gives the distribution of prices and produces latent feature vectors. While both types of neural network architectures we consider outperform simpler machine learning methods, convolutional networks outperform recurrent networks for comparable training time or model complexity. The resulting model powers our online price estimator and search engine.
    Date: 2019–10

This nep-for issue is ©2019 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.