nep-for New Economics Papers
on Forecasting
Issue of 2013‒11‒22
fifteen papers chosen by
Rob J Hyndman
Monash University

  1. Are product spreads useful for forecasting? An empirical evaluation of the Verleger hypothesis By Baumeister, Christiane; Kilian, Lutz
  2. Forecasting the real price of oil in a changing world: A forecast combination approach By Baumeister, Christiane; Kilian, Lutz
  3. “Tourism demand forecasting with different neural networks models” By Oscar Claveria; Enric Monte; Salvador Torra
  4. “Forecasting Business surveys indicators: neural networks vs. time series models” By Oscar Claveria; Salvador Torra
  5. Forecasting and Tracking Real-Time Data Revisions in Inflation Persistence By Tierney, Heather L.R.
  6. Estimates of uncertainty around budget forecasts By John Clark; Caroline Gibbons; Susan Morrissey; Joshua Pooley; Emily Pye; Rhett Wilcox; Luke Willard
  7. Distilling the Macroeconomic News Flow By Alessandro Beber; Michael W. Brandt; Maurizio Luisi
  8. Asymmetric Behaviour of Inflation around the Target in Inflation-Targeting Emerging Markets By Kurmas Akdogan
  9. Multivariate Time Series Model with Hierarchical Structure for Over-dispersed Discrete Outcomes By Nobuhiko Terui; Masataka Ban
  10. A Quest for Leading Indicators of the Turkish Unemployment Rate By H. Burcu Gurcihan; Gonul Sengul; Arzu Yavuz
  11. A Sticky Information Phillips Curve for South Africa By Monique Reid and Gideon Du Rand
  12. Nonparametric estimation of a hedonic price model: A South African case study By M Du Preez, DE Lee and M Sale
  13. Mixed models for predictive modelling in actuarial science. By Antonio, Katrien; Zhang, Yanwei
  14. Why do students leave education early? Theory and evidence on high school dropout rates. By Cabus, S.; De Witte, Kristof
  15. It is harder, not easier, to predict the winner of the Champions League. By Schokkaert, Jeroen; Swinnen, Jo

  1. By: Baumeister, Christiane; Kilian, Lutz
    Abstract: Notwithstanding a resurgence in research on out-of-sample forecasts of the price of oil in recent years, there is one important approach to forecasting the real price of oil which has not been studied systematically to date. This approach is based on the premise that demand for crude oil derives from the demand for refined products such as gasoline or heating oil. Oil industry analysts such as Philip Verleger and financial analysts widely believe that there is predictive power in the product spread, defined as the difference between suitably weighted refined product market prices and the price of crude oil. Our objective is to evaluate this proposition. We derive from first principles a number of alternative forecasting model specifications involving product spreads and compare these models to the no-change forecast of the real price of oil. We show that not all product spread models are useful for out-of-sample forecasting, but some models are, even at horizons between one and two years. The most accurate model is a time-varying parameter model of gasoline and heating oil spot spreads that allows the marginal product market to change over time. We document MSPE reductions as high as 20% and directional accuracy as high as 63% at the two-year horizon, making product spread models a good complement to forecasting models based on economic fundamentals, which work best at short horizons. --
    Keywords: oil price,Futures,WTI,Brent,Acquisition Cost,Refined products,Crack spread,Forecast accuracy,Real-time data
    JEL: Q43 C53 G15
    Date: 2013
  2. By: Baumeister, Christiane; Kilian, Lutz
    Abstract: The U.S. Energy Information Administration (EIA) regularly publishes monthly and quarterly forecasts of the price of crude oil for horizons up to two years, which are widely used by practitioners. Traditionally, such out-of-sample forecasts have been largely judgmental, making them difficult to replicate and justify. An alternative is the use of real-time econometric oil price forecasting models. We investigate the merits of constructing combinations of six such models. Forecast combinations have received little attention in the oil price forecasting literature to date. We demonstrate that over the last 20 years suitably constructed real-time forecast combinations would have been systematically more accurate than the no-change forecast at horizons up to 6 quarters or 18 months. MSPE reduction may be as high as 12% and directional accuracy as high as 72%. The gains in accuracy are robust over time. In contrast, the EIA oil price forecasts not only tend to be less accurate than no-change forecasts, but are much less accurate than our preferred forecast combination. Moreover, including EIA forecasts in the forecast combination systematically lowers the accuracy of the combination forecast. We conclude that suitably constructed forecast combinations should replace traditional judgmental forecasts of the price of oil. --
    Keywords: forecast combination,real-time data,model misspecification,structural change,oil price
    JEL: Q43 C53 E32
    Date: 2013
  3. By: Oscar Claveria (Faculty of Economics, University of Barcelona); Enric Monte (Department of Signal Theory and Communications, Polytechnic University of Catalunya (UPC)); Salvador Torra (Faculty of Economics, University of Barcelona)
    Abstract: This paper aims to compare the performance of different Artificial Neural Networks techniques for tourist demand forecasting. We test the forecasting accuracy of three different types of architectures: a multi-layer perceptron, a radial basis function and an Elman network. We also evaluate the effect of the memory by repeating the experiment assuming different topologies regarding the number of lags introduced. We used tourist arrivals from all the different countries of origin to Catalonia from 2001 to 2012. We find that multi-layer perceptron and radial basis function models outperform Elman networks, being the radial basis function architecture the one providing the best forecasts when no additional lags are incorporated. These results indicate the potential existence of instabilities when using dynamic networks for forecasting purposes. We also find that for higher memories, the forecasting performance obtained for longer horizons improves, suggesting the importance of increasing the dimensionality for long term forecasting.
    Keywords: tourism demand; forecasting; artificial neural networks; multi-layer perceptron; radial basis function; Elman networks; Catalonia. JEL classification: L83; C53; C45; R11
    Date: 2013–11
  4. By: Oscar Claveria (Faculty of Economics, University of Barcelona); Salvador Torra (Faculty of Economics, University of Barcelona)
    Abstract: The objective of this paper is to compare different forecasting methods for the short run forecasting of Business Survey Indicators. We compare the forecasting accuracy of Artificial Neural Networks (ANN) vs. three different time series models: autoregressions (AR), autoregressive integrated moving average (ARIMA) and self-exciting threshold autoregressions (SETAR). We consider all the indicators of the question related to a country’s general situation regarding overall economy, capital expenditures and private consumption (present judgement, compared to same time last year, expected situation by the end of the next six months) of the World Economic Survey (WES) carried out by the Ifo Institute for Economic Research in co-operation with the International Chamber of Commerce. The forecast competition is undertaken for fourteen countries of the European Union. The main results of the forecast competition are offered for raw data for the period ranging from 1989 to 2008, using the last eight quarters for comparing the forecasting accuracy of the different techniques. ANN and ARIMA models outperform SETAR and AR models. Enlarging the observed time series of Business Survey Indicators is of upmost importance in order of assessing the implications of the current situation and its use as input in quantitative forecast models.
    Keywords: Business surveys; Forecasting; Time series models; Nonlinear models; Neural networks.
    Date: 2013–11
  5. By: Tierney, Heather L.R.
    Abstract: The purpose of this paper is to examine the forecasting ability of sixty-two vintages of revised real-time PCE and core PCE using nonparametric methodologies. The combined fields of real-time data and nonparametric forecasting have not been previously explored with rigor, which this paper remedies. The contributions of this paper are on the three fronts of (i.) analysis of real-time data; (ii.) the additional benefits of using nonparametric econometrics to examine real-time data; and (iii.) nonparametric forecasting with real-time data. Regarding the analysis of real-time data revisions, this paper finds that the third quarter releases of real-time data have the largest number of data revisions. Secondly, nonparametric regressions are beneficial in utilizing the information provided by data revisions, which typically are just a few tenths in magnitude but are significant enough to statistically affect regression results. The deviations in window widths can be useful in identifying potential problematic time periods such as a large spike in oil prices. The third and final front of this paper regards nonparametric forecasting and the best performing real-time data release with the three local nonparametric forecasting methods outperforming the parametric benchmark forecasts. Lastly, this paper shows that the best performing quarterly-release of real-time data is dependent on the benchmark revision periods. For vintages 1996:Q1 to 2003:Q3, the second quarter real-time data releases produce the smaller RMSE 58% of the time and for vintages 2003:Q4 to 2011:Q2, the third quarter real-time data releases produce forecasts with smaller RMSE approximately 60% of the time.
    Keywords: Nonparametric Forecasting, Real-Time Data, Monetary Policy, Inflation Persistence
    JEL: C14 C53 E52
    Date: 2013–11–08
  6. By: John Clark (Treasury, Government of Australia); Caroline Gibbons (Treasury, Government of Australia); Susan Morrissey (Treasury, Government of Australia); Joshua Pooley (Treasury, Government of Australia); Emily Pye (Department of Foreign Affairs and Trade, Australian Aid program, Government of Australia); Rhett Wilcox (Treasury, Government of Australia); Luke Willard (Treasury, Government of Australia)
    Abstract: We use past forecast errors to construct confidence intervals around Australian Government Budget forecasts of key economic and fiscal variables. These confidence intervals provide an indication of the extent of uncertainty around the point estimate forecasts presented in the Budget.
    Keywords: Confidence intervals, forecast errors
    JEL: E17 H68
    Date: 2013–11
  7. By: Alessandro Beber; Michael W. Brandt; Maurizio Luisi
    Abstract: We propose a simple cross-sectional technique to extract daily factors from economic news released at different times and frequencies. Our approach can effectively handle the large number of different announcements that are relevant for tracking current economic conditions. We apply the technique to extract real-time measures of inflation, output, employment, and macroeconomic sentiment, as well as corresponding measures of disagreement among economists about these indices. We find that our procedure provides more timely and accurate forecasts of future changes in economic conditions than other real-time forecasting approaches.
    JEL: E0 E17 E27 E32 E37 E44 G0
    Date: 2013–11
  8. By: Kurmas Akdogan
    Abstract: We explore the asymmetric behaviour of inflation around the target level for inflation-targeting emerging markets. The first rationale behind this asymmetry is the asymmetric policy response of the central bank around the target. Central banks could have a stronger bias towards overshooting rather than undershooting the inflation target. Consequently, the policy response would be stronger once the inflation jumps above the target, compared to a negative deviation. Second rationale is the asymmetric inflation persistence. We suggest that recently developed Asymmetric Exponential Smooth Transition Autoregressive (AESTAR) model provides a convenient framework to capture the asymmetric behaviour of inflation driven by these two effects. We further conduct an out-of-sample forecasting exercise and show that the predictive power of AESTAR model for inflation is high, especially at long-horizons.
    Keywords: Inflation, forecasting, nonlinear adjustment
    JEL: C32 E37
    Date: 2013
  9. By: Nobuhiko Terui; Masataka Ban
    Abstract: In this paper, we propose a multivariate time series model for over-dispersed discrete data to explore the market structure based on sales count dynamics. We first discuss the microstructure to show that over-dispersion is inherent in the modeling of market structure based on sales count data. The model is built on the likelihood function induced by decomposing sales count response variables according to products' competitiveness and conditioning on their sum of variables, and it augments them to higher levels by using Poisson-Multinomial relationship in a hierarchical way, represented as a tree structure for the market definition. State space priors are applied to the structured likelihood to develop dynamic generalized linear models for discrete outcomes. For over-dispersion problem, Gamma compound Poisson variables for product sales counts and Dirichlet compound multinomial variables for their shares are connected in a hierarchical fashion. Instead of the density function of compound distributions, we propose a data augmentation approach for more efficient posterior computations in terms of the generated augmented variables particularly for generating forecasts and predictive density. We present the empirical application using weekly product sales time series in a store to compare the proposed models accommodating over-dispersion with alternative no over-dispersed models by several model selection criteria, including in-sample fit, out-of-sample forecasting errors, and information criterion. The empirical results show that the proposed modeling works well for the over-dispersed models based on compound Poisson variables and they provide improved results than models with no consideration of over-dispersion.
    Date: 2013–01
  10. By: H. Burcu Gurcihan; Gonul Sengul; Arzu Yavuz
    Abstract: This paper examines various variables that are likely to be associated with the Turkish non-agricultural unemployment rate in search of indicators to summarize and forecast the state of the labor market. We consider a total of 72 series that reflect aggregate economic activity, labor market conditions, expectations over future economic activity, global economic trends and credit conditions. We use Granger causality tests, correlation analyses and individual out of sample forecast performance of these series to assess their informativeness about the unemployment rate. We find that Business Tendency Survey indicators and some series that measure the global economic conditions satisfy all three criteria of informativeness. Moreover, the composite index constructed from series selected based upon out of sample predictive power improves short-term forecast performance of the autoregressive benchmark model, where we use only lagged values of the unemployment rate.
    Keywords: Leading Indicator,Unemployment Rate,GrangerCausality Test
    JEL: C32 E24
    Date: 2013
  11. By: Monique Reid and Gideon Du Rand
    Abstract: Mankiw and Reis (2002) propose the Sticky Information Phillips Curve as an alternative to the standard New Keynesian Phillips Curve, to address empirical shortcomings in the latter. In this paper, a Sticky Information Phillips curve for South Africa is estimated, which requires data on expectations of current period variables conditional on sequences of earlier period information sets. In the literature the choice of proxies for the inflation expectations and output gap measures are usually not well motivated. In this paper, we test the sensitivity of model fit and parameter estimates to a variety of proxies. We find that parameter estimates for output gap proxies based either on a simple Hodrik-Prescott filter application or on a Kalman filter estimation of an aggregate production function are significant and reasonable, whereas methods employing direct calculation of marginal costs do not yield acceptable results. Estimates of information updating probability range between 0.69 and 0.81. This is somewhat higher than suggested by alternative methods using micro-evidence (0.65 – 0.70 (Reid, 2012)). Lastly, we find that neither parameter estimates nor model diagnostics are sensitive to the choice of expectation proxy, whether it be constructed from surveyed expectations or the ad hoc VAR based forecasting methods.
    Keywords: South Africa, sticky information, Phillips curve
    JEL: E31 E3 E52
    Date: 2013
  12. By: M Du Preez, DE Lee and M Sale
    Abstract: Parametric regression models of hedonic price functions suffer from two main specification issues: the identification of appropriate dependent and independent variables, and the choice of functional form. Although the first issue remains relevant with the use of nonparametric regression models, the second issue becomes irrelevant since these models do not presume functional forms a priori. We estimate a linear parametric model via OLS, which fails a common specification test, before showing that recently developed nonparametric regression methods outperform it significantly. In addition to estimating the models, we compare the out-of-sample prediction performance of the OLS and nonparametric models. Our data reveals that the nonparametric models provide more accurate predictions than the parametric model.
    Keywords: Parametric regression model, hedonic price, South Africa
    Date: 2013
  13. By: Antonio, Katrien; Zhang, Yanwei
    Date: 2013
  14. By: Cabus, S.; De Witte, Kristof
    Abstract: This paper contributes to the growing literature on school dropout by proposing and empirically testing a theoretical framework on the enrollment decision of youngsters in secondary education. The model relates school dropout to time preferences, motivation, opportunity costs, and policy measures, and is empirically tested on longitudinal data of about 5,000 Dutch vocational students. We evaluate the enrollment decision of students for (1) di¤erent intensity levels of dropout prevention policy, and for (2) di¤erent levels of economic development. The results indicate that the model can accurately predict actual enrollment rates over the period 2000-2011. Using the model to forecast the level of school dropout in the Netherlands by the year 2020, it is observed that a very strict dropout prevention policy could yield nearly maximum enrollment rates (i.e., 97%) in schools. However, the annual budget for a similar dropout prevention policy is estimated at e574 million or 0.10% of the Dutch GDP.
    Keywords: Dropout prevention; Economic modeling; Secondary education;
    Date: 2013
  15. By: Schokkaert, Jeroen; Swinnen, Jo
    Abstract: European Cup football has experienced a major change in format with the introduction of the Champions League in 1992 and a major change in admission rules with direct qualification for multiple teams from the highest ranked leagues in 1999. We show that, in line with popular press reports and other studies, qualification in lower rounds has become more predictable in the Champions League. At the same time, however, outcomes at later stages have become less predictable. We provide evidence and an explanation.
    Keywords: Champions League; European Cup; sports tournament; uncertainty of outcome;
    Date: 2013–03

This nep-for issue is ©2013 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.