nep-for New Economics Papers
on Forecasting
Issue of 2016‒07‒30
eleven papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting Turkish GDP Growth : Bottom-Up vs Direct? By Mahmut Gunay
  2. Forecasting GDP during and after the Great Recession: A contest between small-scale bridge and large-scale dynamic factor models By Patrice Ollivaud; Pierre-Alain Pionnier; Elena Rusticelli; Cyrille Schwellnus; Seung-Hee Koh
  3. Has the Forecasting Performance of the Federal Reserve’s Greenbooks Changed over Time? By Ozan Eksi; Cuneyt Orman; Bedri Kamil Onur Tas
  4. Forecasting Turkish Real GDP Growth in a Data Rich Environment By Bahar Sen Dogan; Murat Midilic
  5. Forecasting Turkish GDP Growth with Financial Variables and Confidence Indicators By Mahmut Gunay
  6. Automated variable selection and shrinkage for day-ahead electricity price forecasting By Bartosz Uniejewski; Jakub Nowotarski; Rafal Weron
  7. Forecasting Financial Returns with a Structural Macroeconomic Model By Eric Jondeau; Michael Rockinger
  8. Converging on the Medal Stand: Rio 2016 Olympic Forecasts By Marcus Noland
  9. Identification and Critical Time Forecasting of Real Estate Bubbles in the U.S.A and Switzerland By Diego ARDILA; Dorsa SANADGOL; Peter CAUWELS; Didier SORNETTE
  10. Forecasting Future Oil Production in Norway and the UK: A General Improved Methodology By Lucas FIEVET; Zalàn FORRO; Peter CAUWELS; Didier SORNETTE
  11. The predictive power of the implied volatility of interest rates: Evidence from US Dollar, Euro, and Japanese Yen By Takahiro Hattori

  1. By: Mahmut Gunay
    Abstract: [EN] In this note, we compare performance of direct and bottom-up approaches to forecasting Turkish GDP growth. In the bottom-up approach, we forecast each component separately and then aggregate these forecasts to reach GDP growth forecast. In the direct approach, we model and forecast GDP growth itself. Results indicate that bottom-up approach helps reduce forecast errors. Importance of the bottom-up approach becomes more evident when we take into account the storytelling dimension of forecasting. [TR] Bu calismada, milli gelir tahmini icin dogrudan ve dolayli yaklasimlarin performanslari karsilastirilmaktadir. Dolayli yaklasimda milli gelirin alt kalemleri ayri ayri tahmin edilip, bu tahminlerin birlestirilmesiyle milli gelir buyume tahmini olusturulmaktadir. Dogrudan yaklasimda ise milli gelir buyumesinin kendisi modellenmekte ve tahmin edilmektedir. Sonuclar, dolayli yaklasimin tahmin hatalarini azalttigini gostermektedir. Tahminlerin salt rakam sunmaktan ziyade iktisadi bir oyku anlatmak icin de kullanildigi dikkate alindiginda, daha kapsamli analiz yapmaya imkan veren dolayli yaklasimin onemi belirginlesmektedir.
    Date: 2016
  2. By: Patrice Ollivaud; Pierre-Alain Pionnier; Elena Rusticelli; Cyrille Schwellnus; Seung-Hee Koh
    Abstract: This paper compares the short-term forecasting performance of state-of-the-art large-scale dynamic factor models (DFMs) and the small-scale bridge models routinely used at the OECD. Pseudo-real time out-of-sample forecasts for France, Germany, Italy, Japan, United Kingdom and the United States during and after the Great Recession (2008-2014) suggest that large-scale DFMs are not systematically more accurate than small-scale bridge models, especially at short forecast horizons. Moreover, DFM parameters appear to be highly unstable during the Great Recession (2008-2009), making forecast revisions between successive vintages difficult to explain as revisions cannot be fully attributed to news on specific groups of indicators. The implication for OECD forecasting practice is that there would be no gain from switching from the current small-scale bridge models to large-scale DFMs. Prévoir le PIB pendant et après la Grande Récession : Une comparaison des modèles d'étalonnage de petite taille et des modèles à facteurs dynamiques de grande taille Cet article compare les performances en prévision à court terme de modèles à facteurs dynamiques (DFMs) de grande taille standard dans la littérature à celles des modèles d’étalonnage de petite taille couramment utilisés à l’OCDE pour les exercices de prévision. Des prévisions hors échantillon en pseudo temps réel pour la France, l’Allemagne, l’Italie, le Japon le Royaume-Uni et les États-Unis pendant et après la Grande Récession (2008-2014) montrent que les DFMs de grande taille ne sont pas plus performants, en moyenne, que les modèles d’étalonnage de petite taille, notamment aux horizons les plus courts. De plus, les paramètres des DFMs sont très instables pendant la Grande Récession, ce qui rend les révisions des prévisions d’un exercice à l’autre plus difficiles à expliquer et à relier à différents groupes d’indicateurs. En pratique, nous en concluons que l’OCDE n’aurait pas intérêt, pour ses exercices de prévision, à abandonner les modèles d’étalonnage de petite taille pour les DFMs de grande taille.
    Keywords: dynamic factor models, bridge models, big data, nowcasting, prévision en temps réel, modèle d’étalonnage, modèles à facteurs dynamiques, mégadonnées
    JEL: C53 E37
    Date: 2016–07–26
  3. By: Ozan Eksi; Cuneyt Orman; Bedri Kamil Onur Tas
    Abstract: We investigate how the forecasting performance of the Federal Reserve Greenbooks has changed relative to commercial forecasters between 1974 and 2009. To this end, we analyze time-variation in the Greenbook coefficients in forecast encompassing regressions. Assuming that model coefficients change continuously, we estimate unobserved components models using Bayesian inference techniques. To verify that our results do not depend on the specific way change is modeled, we also allow the coefficients to change discretely rather than continuously and test for structural breaks using classical inference techniques. We find that the Greenbook forecasts have been consistently superior to the commercial forecasts at all horizons throughout our sample period. Although the forecasting performance gap has narrowed at more distant horizons after the early-to-mid 1980s, it remains significant.
    Keywords: Greenbook inflation forecasts, SPF inflation forecasts, Evaluating forecasts, Time-variation in coefficients
    JEL: C11 E52 E43
    Date: 2015
  4. By: Bahar Sen Dogan; Murat Midilic
    Abstract: This study generates nowcasts and forecasts for the growth rate of the Gross Domestic Product (GDP) in Turkey using 204 daily financial series with Mixed Data Sampling (MIDAS) framework over the period 2010Q2-2015Q1. Our findings suggest that MIDAS regression models and forecast combinations provide advantage in exploiting information from daily financial data compared to the models using simple aggregation schemes. In addition, incorporating daily financial data into the analysis improves our forecasts substantially. These results indicate that both the information content of the financial data and the flexible data-driven weighting scheme of MIDAS regressions play an essential role in forecasting the future state of the Turkish economy.
    Keywords: Real GDP Growth, Forecasting, MIDAS
    JEL: C22 C53 G10
    Date: 2016
  5. By: Mahmut Gunay
    Abstract: [EN] This note evaluates the forecast performance of the financial variables and confidence indicators for four quarter ahead cumulative growth of Turkish GDP. Our results point out that some indicators can help reduce forecast errors relative to a benchmark, but forecast performance of the variables may change over time. Combining forecasts with equal weight or based on the recent performance does not lead to a significant difference in forecast performance. [TR] Bu calismada Turkiye ekonomisi icin finansal degiskenler ile guven endekslerinin dort ceyrek birikimli GSYIH buyumesi tahmin performanslari degerlendirilmistir. Sonuclar, incelenen degiskenlerin bazilarinin baz bir modele gore tahmin hatalarini dusurdugunu ancak tahmin performansinin zaman icinde degisebildiðini gostermektedir. Tahmin birlestirmesi icin tahminlerin esit agirliklandirilmasi ile son donem performanslarina gore agirliklandirilmasi arasinda onemli bir fark gorulmemistir.
    Date: 2016
  6. By: Bartosz Uniejewski; Jakub Nowotarski; Rafal Weron
    Abstract: In day-ahead electricity price forecasting (EPF) variable selection is a crucial issue. Conducting an extensive empirical study involving state-of-the-art parsimonious expert models as benchmarks, datasets from three major power markets and five classes of automated selection and shrinkage procedures (single-step elimination, stepwise regression, ridge regression, lasso and elastic nets) we show that using the latter two classes can bring significant accuracy gains compared to commonly used EPF models. In particular, one of the elastic nets - a class that has not been considered in EPF before - stands out as the best performing model overall.
    Keywords: Electricity price forecasting; Day-ahead market; Autoregression; Variable selection; Stepwise regression; Ridge regression; Lasso; Elastic net
    JEL: C14 C22 C51 C53 Q47
    Date: 2016–07–05
  7. By: Eric Jondeau (University of Lausanne; Swiss Finance Institute); Michael Rockinger (University of Lausanne - School of Economics and Business Administration (HEC-Lausanne); Centre for Economic Policy Research (CEPR); Swiss Finance Institute)
    Abstract: This paper investigates the ability of a fully structural macro-finance model to forecast long-term financial returns. We estimate a Dynamic Stochastic General Equilibrium (DSGE) model that describes the dynamics of the U.S. economy. The model includes government bond and stock market returns, which allows us to describe bond and stock risk premia. We first show that these risk premia are fundamentally related to other shocks in the economy. Second, the DSGE model reproduces the mean reversion in the term structure of risks for bond and stock returns. It also generates long-term forecasts of financial returns that outperform unrestricted VAR models.
    Keywords: DSGE model, VAR model, Financial returns, Long-term forecast
    JEL: C11 E44 E47
  8. By: Marcus Noland (Peterson Institute for International Economics)
    Abstract: This Policy Brief presents forecasts of medal counts at the 2016 Summer Olympics in Rio de Janeiro, Brazil. Building on previously developed statistical models that analyze the correlations between Olympic success and socioeconomic variables, the models underlying these forecasts go further, adjusting for the distortions in the historical record created by large-scale boycotts (Moscow 1980, Los Angeles 1984) and doping. The latter consideration is critical to forecasting success in Rio insofar, as the forecasts are strongly influenced by performance at the immediately preceding Games. Subject to uncertainties over Russian participation and the possible impact of the Zika virus, the forecasts indicate that the United States is likely to continue to earn the greatest number of medals, but China is closing the medal gap. Brazil should get a boost from hosting the Games, but its home field advantage may not be as great as experienced by prior hosts. Slumping performance in Rio could add to Britain's post-Brexit malaise.
    Date: 2016–07
  9. By: Diego ARDILA (ETH Zurich); Dorsa SANADGOL (ETH Zurich); Peter CAUWELS (ETH Zurich); Didier SORNETTE (ETH Zurich and Swiss Finance Institute)
    Abstract: We present a hybrid model for diagnosis and critical time forecasting of real estate bubbles. The model combines two elements: 1) the Log Periodic Power Law (LPPL) model to describe endogenous price dynamics originated from positive feedback loops between economic agents; and 2) a diffusion index method that creates a parsimonious representation of multiple macroeconomic variables. We examine the behavior of our model on the housing price indices of 380 US metropolitan areas, using 15, 35, and 90 national-level macroeconomic time series and a dynamic forecasting methodology. Empirical results suggests that the model is able to forecast the end of the bubbles and to identify variables highly relevant during the bubble regime. In addition, the same methodology is applied to the national housing price index of Switzerland, diagnosing a bubble in which global imbalances and Switzerland's status as a safe haven seem to be playing a dominant role.
    Keywords: real-estate bubbles, USA and Switzerland, diffusion index, forecasting, log-periodic power law, criticality, positive feedback, sparse partial least squares
    JEL: C12 C22 C52 G01 G17
  10. By: Lucas FIEVET (ETH Zurich); Zalàn FORRO (Independent); Peter CAUWELS (ETH Zurich); Didier SORNETTE (ETH Zurich and Swiss Finance Institute)
    Abstract: We present a new Monte-Carlo methodology to forecast the crude oil production of Norway and the U.K. based on a two-step process, (i) the nonlinear extrapolation of the current/past performances of individual oil fields and (ii) a stochastic model of the frequency of future oil field discoveries. Compared with the standard methodology that tends to underestimate remaining oil reserves, our method gives a better description of future oil production, as validated by our back-tests starting in 2008. Specifically, we predict remaining reserves extractable until 2030 to be 188 ± 10 million barrels for Norway and 98 ± 10 million barrels for the UK, which are respectively 45% and 66% above the predictions using the standard methodology.
    Keywords: Monte-Carlo, oil peak, logistic equation, Poisson process, power law distribution
    JEL: C15 C46 O13 Q40
  11. By: Takahiro Hattori (Faculty of Economics, Keio University)
    Abstract: This is the first paper to analyze the predictability of implied volatility based on swaption for the major currencies US Dollar (USD), Euro (EUR), and Japanese Yen (JPY). Managing interest rate risk is of huge importance for risk management in financial institutions, and swaption is an over-the-counter contract and well-used instrument that enables us to test whether the option contains the information required to predict future realized volatility. Our result shows that implied volatility has greater power to predict future realized volatility compared with the GARCH prediction or HV for the USD and EUR, which is consistent with the equity or futures options markets. However, the GARCH forecast and HV have stronger predictive power for JPY because of the lack of liquidity.
    Keywords: Implied volatility; Predictive power; GARCH; Swaption
    JEL: G13 G14 G12 G13
    Date: 2016–07–11

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.