nep-for New Economics Papers
on Forecasting
Issue of 2005‒12‒09
sixteen papers chosen by
Rob J Hyndman
Monash University

  1. How Useful is Bagging in Forecasting Economic Time Series? A Case Study of US CPI Inflation By Inoue, Atsushi; Kilian, Lutz
  2. Forecast Combinations By Timmermann, Allan G
  3. Forecast Combination and Model Averaging Using Predictive Measures By Eklund, Jana; Karlsson, Sune
  4. Modelling and Forecasting Fiscal Variables for the euro Area By Favero, Carlo A; Marcellino, Massimiliano
  5. The Warsaw Stock Exchange Index WIG: Modelling and Forecasting By Piotr Wdowinski; Aneta Zglinska-Pietrzak
  6. Non-Linearities in the Relation between the Exchange Rate and its Fundamentals By Carlo Altavilla; Paul De Grauwe
  7. Factor model forecasts for New Zealand By Troy Matheson
  8. UIP, Expectations and the Kiwi By Anella Munro;
  9. An Estimated, New Keynesian Policy Model for Australia By Martin Melecky; Daniel Buncic
  10. Short-Run Italian GDP Forecasting and Real-Time Data By Golinelli, Roberto; Parigi, Giuseppe
  11. On the Forecasting Properties of the Alternative Leading Indicators for the German GDP : Recent Evidence By Konstantin A. Kholodilin; Boriss Siliverstovs
  12. Reconciling the Return Predictability Evidenc: In-Sample Forecasts, Out-of-Sample Forecasts, and Parameter Instability By Lettau, Martin; van Nieuwerburgh, Stijn
  13. Survey Expectations By M. Hashem Pesaran; Martin Weale
  14. Term Structure Estimation with Survey Data on Interest Rate Forecasts By Kim, Don H.; Orphanides, Athanasios
  15. An evaluation of labour market forecasts by type of education and occupation for 2002 By Dupuy,Arnaud
  16. Mind your Ps and Qs! Improving ARMA forecasts with RBC priors By Kirdan Lees; Troy Matheson

  1. By: Inoue, Atsushi; Kilian, Lutz
    Abstract: This paper explores the usefulness of bagging methods in forecasting economic time series from linear multiple regression models. We focus on the widely studied question of whether the inclusion of indicators of real economic activity lowers the prediction mean-squared error of forecast models of US consumer price inflation. We study bagging methods for linear regression models with correlated regressors and for factor models. We compare the accuracy of simulated out-of-sample forecasts of inflation based on these bagging methods to that of alternative forecast methods, including factor model forecasts, shrinkage estimator forecasts, combination forecasts and Bayesian model averaging. We find that bagging methods in this application are almost as accurate or more accurate than the best alternatives. Our empirical analysis demonstrates that large reductions in the prediction mean squared error are possible relative to existing methods, a result that is also suggested by the asymptotic analysis of some stylized linear multiple regression examples.
    Keywords: Bayesian model averaging; bootstrap aggregation; factor models; forecast combination; forecast model selection; pre-testing; shrinkage estimation
    JEL: C22 C52 C53
    Date: 2005–10
  2. By: Timmermann, Allan G
    Abstract: Forecast combinations have frequently been found in empirical studies to produce better forecasts on average than methods based on the ex-ante best individual forecasting model. Moreover, simple combinations that ignore correlations between forecast errors often dominate more refined combination schemes aimed at estimating the theoretically optimal combination weights. In this paper we analyse theoretically the factors that determine the advantages from combining forecasts (for example, the degree of correlation between forecast errors and the relative size of the individual models’ forecast error variances). Although the reasons for the success of simple combination schemes are poorly understood, we discuss several possibilities related to model misspecification, instability (non-stationarities) and estimation error in situations where the numbers of models is large relative to the available sample size. We discuss the role of combinations under asymmetric loss and consider combinations of point, interval and probability forecasts.
    Keywords: diversification gains; forecast combinations; model misspecification; pooling and trimming; shrinkage methods
    JEL: C22 C53
    Date: 2005–11
  3. By: Eklund, Jana; Karlsson, Sune
    Abstract: We extend the standard approach to Bayesian forecast combination by forming the weights for the model averaged forecast from the predictive likelihood rather than the standard marginal likelihood. The use of predictive measures of fit offers greater protection against in-sample overfitting and improves forecast performance. For the predictive likelihood we show analytically that the forecast weights have good large and small sample properties. This is confirmed in a simulation study and an application to forecasts of the Swedish inflation rate where forecast combination using the predictive likelihood outperforms standard Bayesian model averaging using the marginal likelihood.
    Keywords: Bayesian model averaging; inflation rate; partial Bayes factor; predictive likelihood; training sample
    JEL: C11 C51 C52 C53
    Date: 2005–10
  4. By: Favero, Carlo A; Marcellino, Massimiliano
    Abstract: In this paper we assess the possibility of producing unbiased forecasts for fiscal variables in the euro area by comparing a set of procedures that rely on different information sets and econometric techniques. In particular, we consider ARMA models, VARs, small scale semi-structural models at the national and euro area level, institutional forecasts (OECD), and pooling. Our small scale models are characterized by the joint modelling of fiscal and monetary policy using simple rules, combined with equations for the evolution of all the relevant fundamentals for the Maastricht Treaty and the Stability and Growth Pact. We rank models on the basis of their forecasting performance using the mean square and mean absolute error criteria at different horizons. Overall, simple time series methods and pooling work well and are able to deliver unbiased forecasts, or slightly upward biased forecast for the debt-GDP dynamics. This result is mostly due to the short sample available, the robustness of simple methods to structural breaks, and to the difficulty of modelling the joint behaviour of several variables in a period of substantial institutional and economic changes. A bootstrap experiment highlights that, even when the data are generated using the estimated small scale multi country model, simple time series models can produce more accurate forecasts, due to their parsimonious specification.
    Keywords: euro area; fiscal forecasting; fiscal rules; forecast comparison
    JEL: C30 C53 E62
    Date: 2005–10
  5. By: Piotr Wdowinski; Aneta Zglinska-Pietrzak
    Abstract: In this paper we have assessed an influence of the NYSE Stock Exchange indexes (DJIA and NASDAQ) and European Stock indexes (DAX and FTSE) on the Warsaw Stock Exchange index WIG within a framework of a GARCH model. By applying a procedure of checking predictive quality of econometric models as proposed by Fair and Shiller (1990), we have found that the NYSE market has relatively more power than European markets in explaining the WSE index WIG.
    Keywords: Warsaw Stock Exchange, stock index, GARCH model, forecasting
    JEL: C20 C50 C60 G10
    Date: 2005
  6. By: Carlo Altavilla; Paul De Grauwe
    Abstract: This paper investigates the relationship between the euro-dollar exchange rate and its underlying fundamentals. First, we develop a simple theoretical model in which chartists and fundamentalists interact. This model predicts the existence of different regimes, and thus non-linearities in the link between the exchange rate and its fundamentals. Second, we account for non-linearity in the exchange rate process by adopting a Markov-switching vector error correction model (MSVECM). Finally, the paper investigates the out-of-sample forecast performance of three competing models of exchange rate determination. The results suggest the presence of nonlinear mean reversion in the nominal exchange rate process. The implications are that different sets of macroeconomic fundamentals act as driving forces of the exchange rates during different time periods. More interestingly, the nonlinear specification significantly improves the forecast accuracy during periods when the deviation between exchange rate and fundamentals is large. Conversely, when the exchange rate is close to its equilibrium value it tends to be better approximated by a naïve random walk.
    Keywords: non-linearity, Markov-switching model, fundamentals
    JEL: C32 F31
    Date: 2005
  7. By: Troy Matheson (Reserve Bank of New Zealand)
    Abstract: This paper focuses on forecasting four key New Zealand macroeconomic variables using a dynamic factor model and a large number of predictors. We compare the (simulated) real-time forecasting performance of the factor model with a variety of other time series models and gauge the sensitivity of our results to alternative variable selection algorithms. We find that the factor model performs particularly well at longer horizons.
    JEL: C32 E47
    Date: 2005–05
  8. By: Anella Munro; (Reserve Bank of New Zealand)
    Abstract: This paper looks at reduced form descriptions of changes in the USD/NZD exchange rate, with emphasis on the interest rate-exchange rate relationship. In the estimated reduced form equations, high domestic short term interest rates relative to foreign interest rates are associated with continued upward pressure on the New Zealand dollar. This effect is most pronounced for the 6-month forward interest differential, and is reinforced by some "inertia" but moderated by deviations from equilibrium as "over or under-valuation" erodes expected returns. Changes in commodity export prices are estimated to have short term effects. Some aspects of the estimated equations are consistent with forward-looking rational expectations, a standard feature of open economy models. Other aspects of the estimated equations suggest random walk exchange rate expectations consistent with Meese and Rogoff (1983). The cross correlation between interest differentials and the exchange rate may be difficult to reconcile with rational expectations. The forecasting performance of a reduced form equation is also assessed.
    JEL: E52 E58 F31 F32
  9. By: Martin Melecky (University of New South Wales, School of Economics); Daniel Buncic (University of New South Wales, School of Economics)
    Abstract: A two-block open economy model is estimated in this paper using Australian and U.S. data. Evaluation of the estimated model is carried out in relation to a simple closed economy alternative. Namely, we inspect the implied transmission mechanisms, and examine the relative out-of-sample forecasting performance of the closed and open economy models.
    Keywords: DSGE Model, Open Economy, Australia, U.S., Bayesian Estimation.
    JEL: F41 E40 E37 C11
    Date: 2005–11–29
  10. By: Golinelli, Roberto; Parigi, Giuseppe
    Abstract: National accounts statistics undergo a process of revisions over time because of the accumulation of information and, less frequently, of deeper changes, as new definitions, new methodologies etc. are implemented. In this paper we try to characterise the revision process of the data of Italian GDP as published by the national statistical office (ISTAT) in the stream of the noise models literature. The analysis shows that this task can be better accomplished by concentrating on the growth rates of the data instead of the levels. Another issue tackled in the paper concerns the informative content of the preliminary releases vis a vis an intermediate vintage supposed to embody all statistical information (or no longer revisable as far as purely statistical changes are concerned) and the latest vintage of the data, supposed to be the definitive one. The analysis of the news models in differences is based on the comparison of the forecasting performance of the preliminary releases with that of a number of one step ahead forecasts computed from alternative models, ranging from very simple univariate to multivariate specifications based on indicators (bridge models). Results show that, for the intermediate vintage, the preliminary version is the better forecast, while the latest vintage, which embodies statistical as well as definitional revisions, may be better characterised by considering both the preliminary version and the bridge models forecasts.
    Keywords: consistent vintages; predictions of 'actual' GDP; preliminary GDP forecasting; real-time data set for Italian GDP
    JEL: C22 C53 C82 E10
    Date: 2005–10
  11. By: Konstantin A. Kholodilin; Boriss Siliverstovs
  12. By: Lettau, Martin; van Nieuwerburgh, Stijn
    Abstract: Evidence of stock return predictability by financial ratios is still controversial as documented by inconsistent results for in-sample and out-of-sample regressions as well as substantial parameter instability. This paper shows that these seemingly incompatible results can be reconciled if the assumption of a fixed steady state mean of the economy is relaxed. We find strong empirical evidence in support of shifts in the steady state and propose simple methods to adjust financial ratios for such shifts. The forecasting relationship of adjusted price ratios and future returns is statistically significant, stable over time and present in out-of-sample tests. We also show that shifts in the steady state are responsible for parameter instability and poor out-of-sample performance of unadjusted price ratios that is found in the data. Our conclusions hold for a variety of financial ratios and are robust to changes in the econometric technique used to estimate shifts in the steady state.
    Keywords: price ratios; dividend price ratio; out-of-sample test; predictibility; Stock returns
    JEL: C12 C22 G1
    Date: 2005–11
  13. By: M. Hashem Pesaran; Martin Weale
    Abstract: This paper focuses on survey expectations and discusses their uses for testing and modeling of expectations. Alternative models of expectations formation are reviewed and the importance of allowing for heterogeneity of expectations is emphasized. A weak form of the rational expectations hypothesis which focuses on average expectations rather than individual expectations is advanced. Other models of expectations formation, such as the adaptive expectations hypothesis, are briefly discussed. Testable implications of rational and extrapolative models of expectations are reviewed and the importance of the loss function for the interpretation of the test results is discussed. The paper then provides an account of the various surveys of expectations, reviews alternative methods of quantifying the qualitative surveys, and discusses the use of aggregate and individual survey responses in the analysis of expectations and for forecasting.
    Keywords: models of expectations formation, survey data, heterogeneity, tests of rational expectations
    JEL: C40 C50 C53 C80
    Date: 2005
  14. By: Kim, Don H.; Orphanides, Athanasios
    Abstract: The estimation of dynamic no-arbitrage term structure models with a flexible specification of the market price of risk is beset by a severe small-sample problem arising from the highly persistent nature of interest rates. We propose using survey forecasts of a short-term interest rate as an additional input to the estimation to overcome the problem. The three-factor pure-Gaussian model thus estimated with the U.S. Treasury term structure for the 1990-2003 period generates a stable estimate of the expected path of the short rate, reproduces the well-known stylized patterns in the expectations hypothesis tests, and captures some of the short-run variations in the survey forecast of the changes in longer-term interest rates.
    Keywords: Dynamic term structure models; expectations hypothesis; interest rate forecasts; survey data; term premia
    JEL: E43 E47 G12
    Date: 2005–11
  15. By: Dupuy,Arnaud (ROA wp)
    Abstract: 1 Introduction1.1 BackgroundThe Research Centre for Education and the Labour Market generates everytwo years medium-term forecast of the labour market prospects of typesof education and occupations. The first forecast were generated in 1989,after a pilot in 1987, under a contract from the Ministry of Education andScience. The project intended in first instance to cover the developmentof an information system of use especially for providing educational andvocational guidance to apprentices and students in secondary and highereducation. Gained experience has shown that the information provided byROA’s forecast was also of primary interest for other labour market agents,namely policy makers and employers.The labour market information provided by ROA’s forecast are used variousinformation products at the national level, for instance by the NationalCareer Guidance Information Centre (LDC) and the Centre for Informationon Higher Education for Consumer and Expert (CHOICE). The first forecastwere used to supplement the labour market module I see!. This wasa computerised information system, established by LDC, bringing togetherinformation from many sources which might be relevant for the choice of a careeror course of study. Vocational guidance by teachers and others involvedin assisting students to make these choices could call up this information viatheir personal computer and obtain, along with other information on studyand vocational choices, an idea of the labour market consequences of thechoices which were available. The LDC brought out another informationsystem, ‘Traject’, which also makes use of labour market information providedby ROA. ROA’s forecast have also been one of the foundations of theLDC’s series of brochures for study and vocational guidance, and both the‘Keuzegids Hoger Onderwijs’ and the ‘Studiekeuze-Informatiedatabase’ publishedby CHOICE. In addition in their own database, the Central for Workand Income (CWI) used the current data and the forecast of the informationsystem to formulate policies on employment in general and vocationalguidance for the unemployed in particular.
    Keywords: education, training and the labour market;
    Date: 2005
  16. By: Kirdan Lees; Troy Matheson (Reserve Bank of New Zealand)
    Abstract: We utilise prior information from a simple RBC model to improve ARMA forecasts of post-war US GDP. We develop three alternative ARMA forecasting processes that use varying degrees of information from the Campbell (1994) flexible labour model. Directly calibrating the model produces poor forecasting performance whereas a model that uses a Bayesian framework to take the model to the data, yields forecasting performance comparable to a purely statistical ARMA process. A final model that uses theory only to restrict the order of the ARMA process (the ps and qs), but that estimates the ARMA parameters using maximum likelihood, yields improved forecasting performance.
    JEL: C11 C22 E37
    Date: 2005–10

This nep-for issue is ©2005 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.