nep-for New Economics Papers
on Forecasting
Issue of 2009‒12‒19
fourteen papers chosen by
Rob J Hyndman
Monash University

  1. Putting the New Keynesian DSGE model to the real-time forecasting test. By Marcin Kolasa; Michał Rubaszek; Paweł Skrzypczyński
  2. Forecasting Euro-area recessions using time-varying binary response models for financial. By Bellégo, C.; Ferrara, L.
  3. Forecasting the US Real House Price Index: Structural and Non-Structural Models with and without Fundamentals By Rangan Gupta; Alain Kabundi; Stephen M. Miller
  4. The Role of Financial Variables in Predicting Economic Activity. By Raphael Espinoza; Fabio Fornari; Marco J. Lombardi
  5. Combining VAR and DSGE forecast densities By Ida Wolden Bache; Anne Sofie Jore; James Mitchell; Shaun P. Vahey
  6. Forecasting the Yield Curve for Brazil By Daniel O. Cajueiro; Jose A. Divino; Benjamin M. Tabak
  7. Could We Have Predicted the Recent Downturn in Home Sales of the Four US Census Regions? By Rangan Gupta; Christian K. Tipoy; Sonali Das
  8. Measuring output gap uncertainty By Anthony Garratt; James Mitchell; Shaun P. Vahey
  9. Forecasting Private Consumption: Survey-based Indicators vs. Google Trends By Torsten Schmidt; Simeon Vosen
  10. Directional Prediction of Returns under Asymmetric Loss: Direct and Indirect Approaches By Stanislav Anatolyev; Natalia Kryzhanovskaya
  11. The role of central bank transparency for guiding private sector forecasts By Ehrmann, Michael; Eijffinger, Sylvester C. W.; Fratzscher, Marcel
  12. A Model of West African Millet Prices in Rural Markets By Molly E. Brown; Nathaniel Higgins; Beat Hintermann
  13. Bayesian estimation of an extended local scale stochastic volatility model By Philippe J. Deschamps
  14. Predicting web site audience demographics for web advertising targeting using multi-web site clickstream data By K. W. DE BOCK; D. VAN DEN POEL;

  1. By: Marcin Kolasa (National Bank of Poland, ul. Swietokrzyska 11/21, PL-00-919 Warsaw, Poland.); Michał Rubaszek (National Bank of Poland, ul. Swietokrzyska 11/21, PL-00-919 Warsaw, Poland.); Paweł Skrzypczyński (National Bank of Poland, ul. Swietokrzyska 11/21, PL-00-919 Warsaw, Poland.)
    Abstract: Dynamic stochastic general equilibrium models have recently become standard tools for policy-oriented analyses. Nevertheless, their forecasting properties are still barely explored. We fill this gap by comparing the quality of real-time forecasts from a richly-specified DSGE model to those from the Survey of Professional Forecasters, Bayesian VARs and VARs using priors from a DSGE model. We show that the analyzed DSGE model is relatively successful in forecasting the US economy in the period of 1994-2008. Except for short-term forecasts of inflation and interest rates, it is as good as or clearly outperforms BVARs and DSGE-VARs. Compared to the SPF, the DSGE model generates better output forecasts at longer horizons, but less accurate short-term forecasts for interest rates. Conditional on experts' now casts, however, the forecasting power of the DSGE turns out to be similar or better than that of the SPF for all the variables and horizons. JEL Classification: C11, C32, C53, D58, E17.
    Keywords: Forecasting, DSGE, Bayesian VAR, SPF, Real-time data.
    Date: 2009–11
  2. By: Bellégo, C.; Ferrara, L.
    Abstract: Recent macroeconomic evolutions during the years 2008 and 2009 have pointed out the impact of financial markets on economic activity. In this paper, we propose to evaluate the ability of a set of financial variables to forecast recessions in the euro area by using a non-linear binary response model associated with information combination. Especially, we focus on a time-varying probit model whose parameters evolve according to a Markov chain. For various forecast horizons, we provide a readable and leading signal of recession by combining information according to two combining schemes over the sample 1970-2006. First we average recession probabilities and second we linearly combine variables through a dynamic factor model in order to estimate an innovative factor-augmented probit model. Out-of-sample results over the period 2007-2008 show that financial variables would have been helpful in predicting a recession signal as September 2007, that is around six months before the effective start of the 2008-2009 recession in the euro area.
    Keywords: Macroeconomic forecasting, Business cycles, Turning points, Financial markets, Non-linear time series, Combining forecasts.
    JEL: C53 E32 E44
    Date: 2009
  3. By: Rangan Gupta (Department of Economics, University of Pretoria); Alain Kabundi (Department of Economics and Econometrics, University of Johannesburg); Stephen M. Miller (College of Business, University of Las Vegas, Nevada)
    Abstract: We employ a 10-variable dynamic structural general equilibrium model to forecast the US real house price index as well as its turning point in 2006:Q2. We also examine various Bayesian and classical time-series models in our forecasting exercise to compare to the dynamic stochastic general equilibrium model, estimated using Bayesian methods. In addition to standard vector-autoregressive and Bayesian vector autoregressive models, we also include the information content of either 10 or 120 quarterly series in some models to capture the influence of fundamentals. We consider two approaches for including information from large data sets – extracting common factors (principle components) in a Factor-Augmented Vector Autoregressive or Factor-Augmented Bayesian Vector Autoregressive models or Bayesian shrinkage in a large-scale Bayesian Vector Autoregressive models. We compare the out-ofsample forecast performance of the alternative models, using the average root mean squared error for the forecasts. We find that the small-scale Bayesian-shrinkage model (10 variables) outperforms the other models, including the large-scale Bayesian-shrinkage model (120 variables). Finally, we use each model to forecast the turning point in 2006:Q2, using the estimated model through 2005:Q2. Only the dynamic stochastic general equilibrium model actually forecasts a turning point with any accuracy, suggesting that attention to developing forward-looking microfounded dynamic stochastic general equilibrium models of the housing market, over and above fundamentals, proves crucial in forecasting turning points.
    Keywords: US House prices, Forecasting, DSGE models, Factor Augmented Models, Large-Scale BVAR models
    JEL: C32 R31
    Date: 2009–12
  4. By: Raphael Espinoza (International Monetary Fund, 700 19th Street, N.W., Washington, D.C. 20431, USA.); Fabio Fornari (European Central Bank, Kaiserstrasse 29, 60311 Frankfurt am Main, Germany.); Marco J. Lombardi (European Central Bank, Kaiserstrasse 29, 60311 Frankfurt am Main, Germany.)
    Abstract: Previous research has shown that the US business cycle leads the European cycle by a few quarters, and can therefore help predicting euro area GDP. We investigate whether financial variables provide additional predictive power. We use a VAR model of the US and the euro area GDPs and extend it to take into account common global shocks and information provided by selected combinations of financial variables. In-sample analysis shows that shocks to financial variables influence real activity with a peak around 4 to 6 quarters after the shock. Out-of-sample Root-Mean- Squared Forecast Error (RMFE) shows that adding financial variables yields smaller errors in forecasting US economic activity, especially at a five-quarter horizon, but the gain is overall tiny in economic terms. This link is even less prominent in the euro area, where financial indicators do not improve short and medium term GDP forecasts even when their timely availability, relative to a given GDP release, is exploited. The same conclusion is reached with a dataset of quarterly industrial production indices, although financial variables marginally improve forecasts of monthly industrial production. We argue that the findings that financial variables have no predictive power for future activity in the euro area relate to the unconditional nature of the RMFE metric. When forecasting ability is assessed as if in real time (i.e. conditionally on the information available at the time when forecasts are made), we find that models using financial variables would have been preferred in many episodes, and in particular between 1999 and 2002. Results from the historical decomposition of a VAR model indeed suggest that in that period shocks were predominantly of financial nature. JEL Classification: F30, F42, F47.
    Keywords: VAR, Financial Variables, International Linkages, Conditional Forecast.
    Date: 2009–11
  5. By: Ida Wolden Bache (Norges Bank); Anne Sofie Jore (Norges Bank); James Mitchell (NIESR); Shaun P. Vahey (Melbourne Business School)
    Abstract: A popular macroeconomic forecasting strategy takes combinations across many models to hedge against instabilities of unknown timing; see (among others) Stock and Watson (2004), Clark and McCracken (2010), and Jore et al. (2010). Existing studies of this forecasting strategy exclude Dynamic Stochastic General Equilibrium (DSGE) models, despite the widespread use of these models by monetary policymakers. In this paper, we combine inflation forecast densities utilizing an ensemble system comprising many Vector Autoregressions (VARs), and a policymaking DSGE model. The DSGE receives substantial weight (for short horizons) provided the VAR components exclude structural breaks. In this case, the inflation forecast densities exhibit calibration failure. Allowing for structural breaks in the VARs reduces the weight on the DSGE considerably, and produces well-calibrated forecast densities for inflation.
    Keywords: Ensemble modeling, Forecast densities, Forecast evaluation, VAR models, DSGE models
    JEL: C32 C53 E37
    Date: 2009–11–05
  6. By: Daniel O. Cajueiro; Jose A. Divino; Benjamin M. Tabak
    Date: 2009–11
  7. By: Rangan Gupta (Department of Economics, University of Pretoria); Christian K. Tipoy (Department of Economics, University of Pretoria); Sonali Das (LQM, CSIR, Pretoria)
    Abstract: This paper analyzes the ability of a random walk and, classical and Bayesian versions of autoregressive, vector autoregressive and vector error correction models in forecasting home sales for the four US census regions (Northeast, Middlewest, South, West), using quarterly data over the period of 2001:Q1 to 2004:Q3, based on an in-sample of 1976:Q1 till 2000:Q4. In addition, we also use our models to predict the downturn in the home sales of the four census regions over the period of 2004:Q4 to 2009:Q2, given that the home sales in all the four census regions peaked in 2005:Q3. Based on our analysis, we draw the following conclusions: (i) Barring the South, there always exists a Bayesian model which tends to outperform all other models in forecasting home sales over the out-of-sample horizon; (ii) When we expose our classical and ‘optimal’ Bayesian forecast models to predicting the peaks and declines in home sales, we find that barring the South again, our models did reasonably well in predicting the turning point exactly at 2005:Q3 or with a lead. In general, the fact that different models produce the best forecasting performance for different regions, highlights the fact that economic conditions prevailing at the start of the out-of-sample horizon are not necessarily the same across the regions, and, hence, vindicates our decision to look at regions rather than the economy as a whole. In addition, we also point out that there is no guarantee that the best performing model over the out-of-sample horizon is also well-suited in predicting the downturn in home sales.
    Keywords: Forecast Accuracy, Home Sales, Vector Autoregressive Models
    JEL: C32 R31
    Date: 2009–12
  8. By: Anthony Garratt; James Mitchell; Shaun P. Vahey (Reserve Bank of New Zealand)
    Abstract: We propose a methodology for producing density forecasts for the output gap in real time using a large number of vector autoregessions in inflation and output gap measures. Density combination utilizes a linear mixture of experts framework to produce potentially non-Gaussian ensemble densities for the unobserved output gap. In our application, we show that data revisions alter substantially our probabilistic assessments of the output gap using a variety of output gap measures derived from univariate detrending filters. The resulting ensemble produces well-calibrated forecast densities for US inflation in real time, in contrast to those from simple univariate autoregressions which ignore the contribution of the output gap. Combining evidence from both linear trends and more flexible univariate detrending filters induces strong multi-modality in the predictive densities for the unobserved output gap. The peaks associated with these two detrending methodologies indicate output gaps of opposite sign for some bservations, reflecting the pervasive nature of model uncertainty in our US data.
    JEL: C32 C53 E37
    Date: 2009–12
  9. By: Torsten Schmidt; Simeon Vosen
    Abstract: In this study we introduce a new indicator for private consumption based on search query time series provided by Google Trends. The indicator is based on factors extracted from consumption-related search categories of the Google Trends application Insights for Search. The forecasting performance of the new indicator is assessed relative to the two most common survey-based indicators - the University of Michigan Consumer Sentiment Index and the Conference Board Consumer Confidence Index. The results show that in almost all conducted in-sample and out-of-sample forecasting experiments the Google indicator outperforms the survey-based indicators. This suggests that incorporating information from Google Trends may off er signifi cant benefi ts to forecasters of private consumption.
    Keywords: Google Trends, private consumption, forecasting, Consumer Sentiment Indicator
    JEL: C53 E21 E27
    Date: 2009–11
  10. By: Stanislav Anatolyev (New Economic School); Natalia Kryzhanovskaya (New Economic School)
    Abstract: To predict a return characteristic, one may construct models of different complexity describing the dynamics of different objects. The most complex object is the entire predictive density, while the least complex is the characteristic whose forecast is of interest. This paper investigates, using experiments with real data, the relation between the complexity of the modeled object and the predictive quality of the return characteristic of interest, in the case when this characteristic is a return sign, or, equivalently, the direction-of-change. Importantly, we carry out the comparisons assuming that the underlying loss function is asymmetric, which is more plausible than the quadratic loss still prevailing in the analysis of returns. Our experiments are performed with returns of various frequencies on a stock market index and exchange rate. By and large, modeling the dynamics of returns by autoregressive conditional quantiles tends to produce forecasts of higher quality than modeling the whole predictive density or modeling the return indicators themselves.
    Keywords: Directional prediction, sign prediction, model complexity, prediction quality, asymmetric loss, predictive density, conditional quantiles, binary autoregression
    Date: 2009–11
  11. By: Ehrmann, Michael; Eijffinger, Sylvester C. W.; Fratzscher, Marcel
    Abstract: There is a broad consensus in the literature that costs of information processing and acquisition may generate costly disagreements in expectations among economic agents, and that central banks may play a central role in reducing such dispersion in expectations. This paper analyses empirically whether enhanced central bank transparency lowers dispersion among professional forecasters of key economic variables, using a large set of proxies for central bank transparency in 12 advanced economies. It finds evidence for a significant and sizeable effect of central bank transparency on forecast dispersion, be it by means of announcing a quantified inflation objective, other forms of communication, or by publishing central banks’ inflation and output forecasts. However, there also appear to be limits to central bank transparency, with decreasing marginal returns to enhancing (economic) transparency, and given our findings that disagreement among inflation expectations in the general public is not affected by the various central bank transparency measures analyzed in this paper.
    Keywords: central bank communication; central banking; disagreement; forecasting; inflation targeting; monetary policy; survey expectations; transparency
    JEL: C53 E37 E52
    Date: 2009–12
  12. By: Molly E. Brown (NASA Goddard Space Flight Center, Greenbelt, MD, USA); Nathaniel Higgins (University of Maryland, College Park, MD, USA); Beat Hintermann (Center for Energy Policy and Economics CEPE, Department of Management, Technology and Economics, ETH Zurich, Switzerland)
    Abstract: In this article we specify a model of millet prices in the three West African countries of Burkina Faso, Mali, and Niger. Using data obtained from USAID’s Famine Early Warning Systems Network (FEWS NET) we present a unique regional cereal price forecasting model that takes advantage of the panel nature of our data, and accounts for the flow of millet across markets. Another novel aspect of our analysis is our use of the Normalized Difference Vegetation Index (NDVI) to detect and control for variation in conditions for productivity. The average absolute out-of-sample prediction error for 4-month-ahead millet prices is about 20 %.
    Keywords: Millet, cereal, West Africa, price forecasting, remote sensing, NDVI, regional panel data
    JEL: O13 O18 Q11 Q13 Q17 R32
    Date: 2009–11
  13. By: Philippe J. Deschamps (Department of Quantitative Economics)
    Abstract: A new version of the local scale model of Shephard (1994) is presented. Its features are identically distributed evolution equation disturbances, the incorporation of in-the-mean effects, and the incorporation of variance regressors. A Bayesian posterior simulator and an exact simulation smoother are presented. The model is applied to simulated data and to publicly available exchange rate and asset return data. Simulation smoothing turns out to be essential for the accurate interval estimation of volatilities. Bayes factors show that the new model is competitive with GARCH and Lognormal stochastic volatility formulations. Its forecasting performance is comparable to GARCH.
    Keywords: State space models; Markov chain Monte Carlo; simulation smoothing; generalized error distribution; generalized t distribution
    JEL: C11 C13 C15 C22
    Date: 2009–08–04
  14. By: K. W. DE BOCK; D. VAN DEN POEL;
    Abstract: Several recent studies have explored the virtues of behavioral targeting and personalization for online advertising. In this paper, we add to this literature by proposing a cost-effective methodology for the prediction of demographic web site visitor profiles that can be used for web advertising targeting purposes. The methodology involves the transformation of web site visitors’ clickstream patterns to a set of features and the training of Random Forest classifiers that generate predictions for gender, age, educational level and occupation category. These demographic predictions can support online advertisement targeting (i) as an additional input in personalized advertising or behavioral targeting, in order to restrict ad targeting to demographically defined target groups, or (ii) as an input for aggregated demographic web site visitor profiles that support marketing managers in selecting web sites and achieving an optimal correspondence between target groups and web site audience composition. The proposed methodology is validated using data from a Belgian web metrics company. The results demonstrate that Random Forests demonstrate superior classification performance over a set of benchmark algorithms. Further, the ability of the model set to generate representative demographic web site audience profiles is assessed. The stability of the models over time is demonstrated using out-of-period data.
    Keywords: demography prediction, demographic targeting, web advertising, Random Forests, web user profiling, clickstream analysis
    Date: 2009–11

This nep-for issue is ©2009 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.