nep-for New Economics Papers
on Forecasting
Issue of 2020‒01‒27
nine papers chosen by
Rob J Hyndman
Monash University

  1. Capturing Macroeconomic Tail Risks with Bayesian Vector Autoregressions By Andrea Carriero; Todd E. Clark; Massimiliano Marcellino
  2. From fixed-event to fixed-horizon density forecasts: Obtaining measures of multi-horizon uncertainty from survey density forecasts By Gergely Ganics; Barbara Rossi; Tatevik Sekhposyan
  3. Nowcasting East German GDP growth: A MIDAS approach By Claudio, João C.; Heinisch, Katja; Holtemöller, Oliver
  4. Uncertainty measures from partially rounded probabilistic forecast surveys By Alexander, Glas; Matthias, Hartmann
  5. Forecasting own brand sales: Does incorporating competition help? By Li, W.; Fok, D.; Franses, Ph.H.B.F.
  6. The folly of forecasting: The effects of a disaggregated demand forecasting system on forecast error, forecast positive bias, and inventory levels By Brüggen, Alexander; Grabner, Isabella; Sedatole, Karen
  7. Exports Since the International Financial Crisis By Pedro Miguel Avelino Bação; António Portugal Duarte; Diogo Viveiros
  8. Neural Network Associative Forecasting of Demand for Goods By Osipov, Vasiliy; Zhukova, Nataly; Miloserdov, Dmitriy
  9. Comparing Deep Neural Network and Econometric Approaches to Predicting the Impact of Climate Change on Agricultural Yield By Timothy Neal; Michael Keane

  1. By: Andrea Carriero; Todd E. Clark; Massimiliano Marcellino (European University Institute; Universität Commerciale Luigi Bocconi; National Bureau of Economic Research; Centre for Economic Policy Research (CEPR); Universität degli Studi di Firenze; Bocconi University)
    Abstract: A rapidly growing body of research has examined tail risks in macroeconomic outcomes. Most of this work has focused on the risks of significant declines in GDP, and has relied on quantile regression methods to estimate tail risks. In this paper we examine the ability of Bayesian VARs with stochastic volatility to capture tail risks in macroeconomic forecast distributions and outcomes. We consider both a conventional stochastic volatility specification and a specification featuring a common volatility factor that is a function of past financial conditions. Even though the conditional predictive distributions from the VAR models are symmetric, our estimated models featuring time-varying volatility yield more time variation in downside risk as compared to upside risk—a feature highlighted in other work that has advocated for quantile regression methods or focused on asymmetric conditional distributions. Overall, the BVAR models perform comparably to quantile regression for estimating tail risks, with, in addition, some gains in standard point and density forecasts.
    Keywords: forecasting; downside risk; asymmetries
    JEL: C53 E17 E37 F47
    Date: 2020–01–16
  2. By: Gergely Ganics; Barbara Rossi; Tatevik Sekhposyan
    Abstract: Surveys of professional forecasters produce precise and timely point forecasts for key macroeconomic variables. However, the accompanying density forecasts are not as widely utilized, and there is no consensus about their quality. This is partly because such surveys are often conducted for “fixed events”. For example, in each quarter panelists are asked to forecast output growth and inflation for the current calendar year and the next, implying that the forecast horizon changes with each survey round. The fixed-event nature limits the usefulness of survey density predictions for policymakers and market participants, who often wish to characterize uncertainty a fixed number of periods ahead (“fixed-horizon”). Is it possible to obtain fixed-horizon density forecasts using the available fixed-event ones? We propose a density combination approach that weights fixed-event density forecasts according to a uniformity of the probability integral transform criterion, aiming at obtaining a correctly calibrated fixed-horizon density forecast. Using data from the US Survey of Professional Forecasters, we show that our combination method produces competitive density forecasts relative to widely used alternatives based on historical forecast errors or Bayesian VARs. Thus, our proposed fixed-horizon predictive densities are a new and useful tool for researchers and policy makers.
    Keywords: Survey of professional forecasters, density forecasts, forecast combination, predictive density, probability integral transform, uncertainty, real-time.
    JEL: C13 C32 C53
    Date: 2019–12
  3. By: Claudio, João C.; Heinisch, Katja; Holtemöller, Oliver
    Abstract: Economic forecasts are an important element of rational economic policy both on the federal and on the local or regional level. Solid budgetary plans for government expenditures and revenues rely on efficient macroeconomic projections. However, official data on quarterly regional GDP in Germany are not available, and hence, regional GDP forecasts do not play an important role in public budget planning. We provide a new quarterly time series for East German GDP and develop a forecasting approach for East German GDP that takes data availability in real time and regional economic indicators into account. Overall, we find that mixed-data sampling model forecasts for East German GDP in combination with model averaging outperform regional forecast models that only rely on aggregate national information.
    Keywords: business surveys,East Germany,MIDAS model,nowcasting
    JEL: C22 C52 C53 E37 R11
    Date: 2019
  4. By: Alexander, Glas; Matthias, Hartmann
    Abstract: Although survey-based point predictions have been found to outperform successful forecasting models, corresponding variance forecasts are frequently diagnosed as heavily distorted. Forecasters who report inconspicuously low ex-ante variances often produce squared forecast errors that are much larger on average. In this paper, we document the novel stylized fact that this variance misalignment is related to the rounding behavior of survey participants. Rounding may reflect the fact that some survey participants employ a rather judgmental approach to forecasting as opposed to using a formal model. We use the distinct numerical accuracies of panelists' reported probabilities as a means to propose several alternative and easily implementable corrections that i) can be carried out in real time, i.e., before outcomes are observed, and ii) deliver a significantly improved match between ex-ante and ex-post forecast uncertainty. According to our estimates, uncertainty about inflation, output growth and unemployment in the U.S. and the Euro area is higher after correcting for the rounding effect. The increase in the share of non-rounded responses in recent years also helps to understand the trajectory of survey-based average uncertainty during the years since the financial and sovereign debt crisis.
    Keywords: Survey data, probabilistic forecasting, rounding, macroeconomic uncertainty.
    JEL: C32 C52 C53 C83
    Date: 2020–01
  5. By: Li, W.; Fok, D.; Franses, Ph.H.B.F.
    Abstract: This study aims to investigate how much value is added to traditional sales forecast- ing models in marketing by using modern techniques like factor models, Lasso, elastic net, random forest and boosting methods. A benchmark model uses only the focal brand's own information, while the other models include competitive sales and market- ing activities in various ways. An Average Competitor Model (ACM) summarises all competitive information by averages. Factor-augmented models incorporate all or some competitive information by means of common factors. Lasso and elastic net models shrink the coecient estimates of specic competing brands towards zero by adding a shrinkage penalty to the sum of squared residuals. Random forest averages many tree models obtained from bootstrapped samples. Boosting trees grow many small trees sequentially and then average over all the tree models to deliver forecasts. We use these methods to forecast sales of packaged goods one week ahead and compare their pre- dictive performance. Our empirical results for 169 brands across 31 product categories show that the Lasso and elastic net are the safest methods to employ as they are better than the benchmark for most of the brands. The random forest method has better improvement for some of the brands.
    Keywords: Sales forecasting, high-dimensional data, principal components, factor model, Lasso, Elastic Net, random forest, boosting, data mining
    Date: 2019–10–10
  6. By: Brüggen, Alexander; Grabner, Isabella; Sedatole, Karen
    Abstract: Periodic demand forecasts are the primary planning and coordination mechanism within organizations. Because most demand forecasts incorporate human judgment, they are subject to both unintentional error and intentional opportunistic bias. We examine whether a disaggregation of the forecast into various sources of demand reduces forecast error and bias. Using proprietary data from a manufacturing organization, we find that absolute demand forecast error declines following the implementation of a disaggregated forecast system. We also find a favorable effect of forecast disaggregation on finished goods inventory without a corresponding increase in costly production plan changes. We further document a decline in positive forecast bias, except for products whose production is limited owing to scarce production resources. This implies that disaggregation alone is not sufficient to overcome heightened incentives of self-interested sales managers to positively bias the forecast for the very products that an organization would like to avoid tying up in inventory.
    Keywords: Budgeting, forecasting, forecast disaggregation, forecast error, forecast bias, inventory management, sales and operations planning
    Date: 2020–01–10
  7. By: Pedro Miguel Avelino Bação (Centre for Business and Economics CeBER, GEMF and Faculty of Economics, University of Coimbra); António Portugal Duarte (Centre for Business and Economics CeBER and Faculty of Economics, University of Coimbra); Diogo Viveiros (Bank of Portugal)
    Abstract: The aim of this paper is to analyze the performance of several models for forecasting exports. We collected data on Portugal’s real exports of goods and on the variables suggested by models based on the assumption of perfect competition and of monopolistic competition. We estimated Vector Autoregressive (VAR) models with those variables and compared the performance of the forecasts produced by those models with the forecasts obtained from simpler, univariate models, namely ARIMA models and Holt’s linear trend model. We consider four alternative frameworks in which the forecasts might be produced. These scenarios correspond to “static forecasts”, “recursive forecasts”, “dynamic forecasts” and “dynamic forecasts with known exogenous variables”. We also consider the computation of recursive forecasts including dummies related to the international financial crisis. The best model (according to the root mean squared error of the forecasts in the period since the start of the international financial crisis) depends on the scenario considered for the computation of the forecasts. The theory-based models do not produce forecasts that are clearly better than the forecasts produced by the simpler univariate models. In addition, the impact of the international financial crisis appears to be better represented by a temporary shock than by a permanent shock (with a constant effect). The results cast doubts on the relevance of traditional measures of competitiveness for the evolution of exports. As a consequence of the above, it is not clear that discussions of competitiveness that put the emphasis on costs, namely on wages, or on the exchange rate provide useful guides to policy.
    Keywords: Competitiveness, exchange rate, exports, forecasting, productivity, wages.
    JEL: D24 F11 F14 F17 F31
    Date: 2020–01
  8. By: Osipov, Vasiliy; Zhukova, Nataly; Miloserdov, Dmitriy
    Abstract: This article discusses the applicability of recurrent neural networks with controlled elements to the problem of forecasting market demand for goods on the four month horizon. Two variants of forecasting are considered. In the first variant, time series are used to train the neural network, including the real demand values, as well as pre-order values for 1, 2 and 3 months ahead. In the second variant, there is an iterative forecasting method. It predicts the de-mand for the next month at each step, and the training set is supplemented by the values predicted for the previous months. It is shown that the proposed methods can give a sufficiently high result. At the same time, the second ap-proach demonstrates greater potential.
    Keywords: Recurrent Neural Network; Machine Learning; Data Mining; Demand Forecasting
    JEL: C45 L10
    Date: 2019–09–23
  9. By: Timothy Neal (UNSW School of Economics); Michael Keane (UNSW School of Economics)
    Abstract: Predicting the impact of climate change on crop yield is difficult, in part because the production function mapping weather to yield is high dimensional and nonlinear. We compare three approaches to predicting yields: (i) deep neural networks (DNNs), (ii) traditional panel-data models, and (iii) a new panel-data model that allows for unit and time fixed-effects in both intercepts and slopes in the agricultural production function - made feasible by a new estimator developed by Keane and Neal (2020) called MO-OLS. Using U.S. county-level corn yield data from 1950-2015, we show that both DNNs and MO-OLS models outperform traditional panel data models for predicting yield, both in-sample and in a Monte Carlo cross-validation exercise. However, the MO-OLS model substantially outperforms both DNNs and traditional panel-data models in forecasting yield in a 2006-15 holdout sample. We compare predictions of all these models for climate change impacts on yields from 2016 to 2100.
    Keywords: Climate Change, Crop Yield, Panel Data, Machine Learning, Neural Net
    Date: 2020–01

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.