nep-for New Economics Papers
on Forecasting
Issue of 2016‒11‒06
four papers chosen by
Rob J Hyndman
Monash University

  1. The Use of Models in Producing OECD Macroeconomic Forecasts By David Turner
  2. Forecasting daily political opinion polls using the fractionally cointegrated VAR model By Morten Ørregaard Nielsen; Sergei S. Shibaev
  3. Lasso-based forecast combinations for forecasting realized variances By Ines Wilms; Jeroen Rombouts; Christophe Croux
  4. Forecasting Chilean Inflation with the Hybrid New Keynesian Phillips Curve: Globalisation, Combination, and Accuracy By Carlos Medel

  1. By: David Turner
    Abstract: This paper firstly describes the role of models in producing OECD global macroeconomic forecasts; secondly, reviews the OECD's forecasting track record; and finally, considers the relationship between forecast performance and models. OECD forecasts are not directly generated from a single global model, but instead rely heavily on expert judgment which is informed by inputs from a range of different models, with forecasts subjected to repeated peer review. For the major OECD economies, current year GDP growth forecasts exhibit a number of desirable properties including that they are unbiased, outperform naïve forecasts and mostly identify turning points. Moreover, there is a trend improvement in current-year forecasting performance which is partly attributed to the increasing use of high frequency ‘now-casting’ indicator models to forecast the current and next quarter’s GDP. Conversely, the track record of one-year-ahead forecasts is much less impressive; such forecasts are biased, often little better than naïve forecasts and are poor at anticipating downturns. Forecasts tend to cluster around those from other international organisations and consensus forecasts; it is particularly striking that differences in one-year-ahead forecasts between forecasters are relatively minor in comparison with the size of average errors made by all of them. This may reflect herding behaviour by forecasters as well as the mean reversion properties of models. These weaknesses in forecasting performance beyond the current year underline the importance of increased efforts to use models to characterise the risk distribution around the baseline forecast, including through the increased use of model-based scenario analysis. Le rôle des modèles dans la production des prévisions macroéconomique de l'OCDE Ce document décrit le rôle des modèles dans la production des prévisions macroéconomiques mondiales de l’OCDE, analyse a posteriori la performance des prévisions passées et examine le lien entre la qualité des prévisions et les modèles utilisés. Les prévisions de l’OCDE ne sont pas élaborées directement à partir d’un modèle mondial unique, mais reposent en grande partie sur des avis d’experts eux-mêmes formés à partir d’éléments provenant de différents modèles. Ces prévisions sont soumises à des spécialistes dans le cadre d’un processus itératif. En ce qui concerne les grandes économies de l’OCDE, les prévisions de croissance du PIB pour l’année en cours présentent un certain nombre de caractéristiques appréciables : elles sont non biaisées, plus exactes que les prévisions « naïves » et, dans la plupart des cas, identifient les points de retournement. En outre, on observe une amélioration tendancielle de la performance des prévisions pour l’année en cours, qui est en partie imputable au recours récent à des modèles d’indicateurs à haute fréquence permettant de prévoir le PIB du trimestre en cours et à venir (now-casting), mais aussi au poids croissant accordé à ces modèles et à l’amélioration de la qualité de leurs résultats. A contrario, l’analyse des prévisions à un an est bien moins convaincante ; ces prévisions sont biaisées, à peine meilleures que les prévisions « naïves » et peu efficaces pour prévoir les retournements de conjoncture. Elles sont généralement proches de celles des autres organisations internationales et du consensus des prévisionnistes, mais il est particulièrement frappant de constater que les disparités existant entre les prévisions à un an des différents prévisionnistes sont moindres comparées à l’ampleur des erreurs moyennes commises par l’ensemble de ces acteurs. Ce constat peut s’expliquer par le comportement moutonnier des prévisionnistes mais également par la tendance au retour à la moyenne qui caractérise les modèles. Ces faiblesses dans les prévisions à plus d’un an montrent qu’il importe d’intensifier les efforts visant à utiliser des modèles pour définir la distribution des risques autour de la prévision de référence, notamment en recourant davantage à l’analyse s’appuyant sur des modèles permettant de construire des scenarios.
    Keywords: forecasting, economic outlook, models, GDP growth
    JEL: E27 E32 F17
    Date: 2016–11–03
  2. By: Morten Ørregaard Nielsen (Queen?s University and CREATES); Sergei S. Shibaev (Queen?s University)
    Abstract: We examine forecasting performance of the recent fractionally cointegrated vector autoregressive (FCVAR) model. We use daily polling data of political support in the United Kingdom for 2010-2015 and compare with popular competing models at several forecast horizons. Our findings show that the four variants of the FCVAR model considered are generally ranked as the top four models in terms of forecast accuracy, and the FCVAR model significantly outperforms both univariate fractional models and the standard cointegrated VAR (CVAR) model at all forecast horizons. The relative forecast improvement is higher at longer forecast horizons, where the root mean squared forecast error of the FCVAR model is up to 15% lower than that of the univariate fractional models and up to 20% lower than that of the CVAR model. In an empirical application to the 2015 UK general election, the estimated common stochastic trend from the model follows the vote share of the UKIP very closely, and we thus interpret it as a measure of Euro-skepticism in public opinion rather than an indicator of the more traditional left-right political spectrum. In terms of prediction of vote shares in the election, forecasts generated by the FCVAR model leading into the election appear to provide a more informative assessment of the current state of public opinion on electoral support than the hung parliament prediction of the opinion poll.
    Keywords: forecasting, fractional cointegration, opinion poll data, vector autoregressive model
    JEL: C32
    Date: 2016–09–22
  3. By: Ines Wilms; Jeroen Rombouts; Christophe Croux
    Abstract: Volatility forecasts are key inputs in financial analysis. While lasso based forecasts have shown to perform well in many applications, their use to obtain volatility forecasts has not yet received much attention in the literature. Lasso estimators produce parsimonious forecast models. Our forecast combination approach hedges against the risk of selecting a wrong degree of model parsimony. Apart from the standard lasso, we consider several lasso extensions that account for the dynamic nature of the forecast model. We apply forecast combined lasso estimators in a comprehensive forecasting exercise using realized variance time series of ten major international stock market indices. We find the lasso extended 'ordered lasso' to give the most accurate realized variance forecasts. Multivariate forecast models, accounting for volatility spillovers between different stock markets, outperform univariate forecast models for longer forecast horizons.
    Keywords: Forecast combination, Hierarchical lasso, Lasso, Ordered Lasso, Realized variance, Volatility forecasting
    Date: 2016–10
  4. By: Carlos Medel
    Abstract: This article analyses the multihorizon predictive power of the Hybrid New Keynesian Phillips Curve (HNKPC) covering the period from 2000.1 to 2014.12, for the Chilean economy. A distinctive feature of this article is the use of a Global Vector Autoregression (GVAR) specification of the HNKPC to enforce an open economy version. Another feature is the use of direct measures of inflation expectations—Consensus Forecasts—differing from a fully-founded rational expectations model. The HNKPC point forecasts are evaluated using the Mean Squared Forecast Error (MSFE) statistic and statistically compared with several benchmarks, including combined forecasts. The results indicate that there is evidence supporting the existence of the HNKPC for the Chilean economy, and robust to alternative specifications. In predictive terms, the results show that in a sample previous to the global financial crisis, the evidence is mixed between atheoretical benchmarks and the HNKPC by itself or participating in a combined prediction. However, when the evaluation sample is extended to include a more volatile inflation period, the results suggest that the HNKPC (and combined with the random walk) delivers the most accurate forecasts at horizons comprised within a year. In the long-run the HNKPC deliver accurate results, but not enough to outperform the candidate statistical models.
    Date: 2016–10

This nep-for issue is ©2016 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.