nep-for New Economics Papers
on Forecasting
Issue of 2012‒10‒13
fifteen papers chosen by
Rob J Hyndman
Monash University

  1. A Dynamic Bivariate Poisson Model for Analysing and Forecasting Match Results in the English Premier League By Siem Jan Koopman; Rutger Lit
  2. Forecasting metal prices: Do forecasters herd? By Pierdzioch, Christian; Rülke, Jan-Christoph; Stadtmann, Georg
  3. A note on forecasting emerging market exchange rates: Evidence of anti-herding By Pierdzioch, Christian; Rülke, Jan-Christoph; Stadtmann, Georg
  4. Combining predictive densities using Bayesian filtering with applications to US economic data By Monica Billio; Roberto Casarin; Francesco Ravazzolo; Herman K. van Dijk
  5. Directional forecasting in financial time series using support vector machines: The USD/Euro exchange rate By Plakandaras, Vasilios; Papadimitriou, Theophilos; Gogas, Periklis
  6. Prediction Markets to Forecast Electricity Demand By Luciano I. de Castro; Peter Cramton
  7. The Economic Value of Realized Volatility: Using High-Frequency Returns for Option Valuation By Peter Christoffersen; Bruno Feunou; Kris Jacobs; Nour Meddahi
  8. Great expectations? Evidence from Colombia’s exchange rate survey By Juan José Echavarría; Mauricio Villamizar
  9. Portfolio risk evaluation: An approach based on dynamic conditional correlations models and wavelet multiresolution analysis By Khalfaoui, R; Boutahar, M
  10. Future Prediction of the Prefectural Economy in Japan: Using a Stochastic Model By Hiroshi Sakamoto
  11. Anticipating Long-Term Stock Market Volatility By Conrad, Christian; Loch, Karin
  12. Forecasting with Unobserved Heterogeneity By Matteo G. Richiardi
  13. Understanding DSGE Filters in Forecasting and Policy Analysis By Andrle, Michal
  14. Forecasting and simulation of the impact of public policies on industrial districts using an agent-based model By Federico Pablo-Marti; Juan Luis Santos; Antonio Gacía-Tabuenca; María Teresa Gallo; Tomás Mancha
  15. Estimating and Forecasting the Yield Curve Using a Markov Switching Dynamic Nelson and Siegel Model By Martin Gonzalez Rozada; Martin sola; Constantino Hevia; Fabio Spagnolo

  1. By: Siem Jan Koopman (VU University Amsterdam); Rutger Lit (VU University Amsterdam)
    Abstract: Attack and defense strengths of football teams vary over time due to changes in the teams of players or their managers. We develop a statistical model for the analysis and forecasting of football match results which are assumed to come from a bivariate Poisson distribution with intensity coefficients that change stochastically over time. This development presents a novelty in the statistical time series analysis of match results from football or other team sports. Our treatment is based on state space and importance sampling methods which are computationally efficient. The out-of-sample performance of our methodology is verified in a betting strategy that is applied to the match outcomes from the 2010/11 and 2011/12 seasons of the English Premier League. We show that our statistical modeling framework can produce a significant positive return over the bookmaker's odds.
    Keywords: Betting; Importance sampling; Kalman filter smoother; Non-Gaussian multivariate time series models; Sport statistics
    JEL: C32 C35
    Date: 2012–09–27
  2. By: Pierdzioch, Christian; Rülke, Jan-Christoph; Stadtmann, Georg
    Abstract: We analyze more than 20,000 forecasts of nine metal prices at four different forecast horizons. We document that forecasts are heterogeneous and report that anti-herding appears to be a source of this heterogeneity. Forecaster anti-herding reflects strategic interactions among forecasters that foster incentives to scatter forecasts around a consensus forecast. --
    Keywords: Metal prices,Forecasting,Forecaster (anti-)herding
    JEL: G17 C33 L61
    Date: 2012
  3. By: Pierdzioch, Christian; Rülke, Jan-Christoph; Stadtmann, Georg
    Abstract: Using survey forecasts of a large number of Asian, European, and South American emerging market exchange rates, we studied empirically whether evidence of herding or antiherding behavior of exchange-rate forecasters can be detected in the cross-section of forecasts. Emerging market exchange-rate forecasts are consistent with herding (anti-herding) if forecasts are biased towards (away from) the consensus forecast. Our empirical findings provide strong evidence of anti-herding of emerging market exchange-rate forecasters. --
    JEL: F31 D84 C33
    Date: 2012
  4. By: Monica Billio (Department of Economics, University Of Venice Cà Foscari); Roberto Casarin (Department of Economics, University Of Venice Cà Foscari); Francesco Ravazzolo (Norges Bank); Herman K. van Dijk (Erasmus University)
    Abstract: Using a Bayesian framework this paper provides a multivariate combination approach to prediction based on a distributional state space representation of predictive densities from alternative models. In the proposed approach the model set can be incomplete. Several multivariate time-varying combination strategies are introduced. In particular, a weight dynamics driven by the past performance of the predictive densities is considered and the use of learning mechanisms. The approach is assessed using statistical and utility-based performance measures for evaluating density forecasts of US macroeconomic time series and of surveys of stock market prices.
    Keywords: Density Forecast Combination, Survey Forecast, Bayesian Filtering, Sequential Monte Carlo.
    JEL: C11 C15 C53 E37
    Date: 2012
  5. By: Plakandaras, Vasilios (Democritus University of Thrace, Department of International Economic Relations and Development); Papadimitriou, Theophilos (Democritus University of Thrace, Department of International Economic Relations and Development); Gogas, Periklis (Democritus University of Thrace, Department of International Economic Relations and Development)
    Abstract: In this paper, we present a novel machine learning based forecasting system of the EU/USD exchange rate directional changes. Specifically, we feed an overcomplete variable set to a Support Vector Machines (SVM) model and refine it through a Sensitivity Analysis process. The dataset spans from 1/1/1999 to 30/11/2011; the data of the last 7 months are reserved for out-of-sample testing. Results show that the proposed scheme outperforms various other machine learning methods treating similar scenarios.
    Keywords: Machine Learning; Support Vector Machines; Exchange Rates; Forecasting
    JEL: C52 C59 F31 G17
    Date: 2012–01–27
  6. By: Luciano I. de Castro; Peter Cramton (Economics Department, University of Maryland)
    Abstract: Forecasting electricity demand for future years is an essential step in resource planning. A common approach is for the system operator to predict future demand from the estimates of individual distribution companies. However, the predictions thus obtained may be of poor quality, since the reporting incentives are unclear. We propose a prediction market as a form of forecasting future demand for electricity. We describe how to implement a simple prediction market for continuous variables, using only contracts based on binary variables. We also discuss specific issues concerning the implementation of such a market.
    Keywords: electricity market design, prediction markets
    JEL: D44
    Date: 2012
  7. By: Peter Christoffersen; Bruno Feunou; Kris Jacobs; Nour Meddahi
    Abstract: Many studies have documented that daily realized volatility estimates based on intraday returns provide volatility forecasts that are superior to forecasts constructed from daily returns only. We investigate whether these forecasting improvements translate into economic value added. To do so we develop a new class of affine discrete-time option valuation models that use daily returns as well as realized volatility. We derive convenient closed-form option valuation formulas and we assess the option valuation properties using S&P500 return and option data. We find that realized volatility reduces the pricing errors of the benchmark model significantly across moneyness, maturity and volatility levels.
    Keywords: Asset pricing; Econometric and statistical methods
    JEL: G13
    Date: 2012
  8. By: Juan José Echavarría; Mauricio Villamizar
    Abstract: In this document we use the Expectations Survey conducted monthly by the Central Bank of Colombia during the period of October 2003 – August 2012. We find that exchange rate revaluations were generally followed by expectations of further revaluation in the short run (1 month), but by expectations of devaluations in the long run (1 year), and that expectations are stabilizing both in the short and long run. The forward rate is generally different from the future spot rate, mainly because forecast errors are on average different from cero. This suggests that exchange rate expectations are not rational. The role of the risk premium is also important, albeit statistically significant only for the 1 year ahead forecasts (not for 1 month). One month expectations are much better predictors than the models of extrapolative, adaptive or regressive expectations or even the forward discount, and all of them outperform a random walk. But results are almost the opposite for 1 year. In this case traders and analysts could actually do much better by following some simple models or by looking at some key variables rather than by following the strategy that they pursue today..
    Keywords: Exchange rate expectations, risk premium, market efficiency, forecasting accuracy, random walk, forward discount, rational expectations hypothesis. Classification JEL: C23, C53, C83, F31, F37.
    Date: 2012–10
  9. By: Khalfaoui, R; Boutahar, M
    Abstract: We analyzed the volatility dynamics of three developed markets (U.K., U.S. and Japan), during the period 2003-2011, by comparing the performance of several multivariate volatility models, namely Constant Conditional Correlation (CCC), Dynamic Conditional Correlation (DCC) and consistent DCC (cDCC) models. To evaluate the performance of models we used four statistical loss functions on the daily Value-at-Risk (VaR) estimates of a diversified portfolio in three stock indices: FTSE 100, S&P 500 and Nikkei 225. We based on one-day ahead conditional variance forecasts. To assess the performance of the abovementioned models and to measure risks over different time-scales, we proposed a wavelet-based approach which decomposes a given time series on different time horizons. Wavelet multiresolution analysis and multivariate conditional volatility models are combined for volatility forecasting to measure the comovement between stock market returns and to estimate daily VaR in the time-frequency space. Empirical results shows that the asymmetric cDCC model of Aielli (2008) is the most preferable according to statistical loss functions under raw data. The results also suggest that wavelet-based models increase predictive performance of financial forecasting in low scales according to number of violations and failure probabilities for VaR models.
    Keywords: Dynamic conditional correlations; Value-at-Risk; wavelet decomposition; Stock prices
    JEL: D53 C53 G11 C52
    Date: 2012–09–24
  10. By: Hiroshi Sakamoto
    Abstract: This study develops an easy forecasting model using prefectural data in Japan. The Markov chain known as a stochastic model corresponds to the vector auto-regressive (VAR) model of the first order. If the transition probability matrix can be appropriately estimated, the forecasting model using the Markov chain can be constructed. Therefore, this study introduces the methodology to estimate the transition probability matrix of the Markov chain using the least-squares optimization. For application, firstly change of the all-prefectures economy by 2020 is analyzed using this model. Secondly, in order to investigate the influence to other prefecture, a specific prefectureÂfs shock is put into a transition probability matrix. Lastly, in order to take out the width of prediction, the Monte Carlo experiment is conducted. Despite this model is very simple, we provide the more sophisticated forecasting information of the prefectural economy in Japan through the complicated extension. JEL classification: C15, C53, C61, O53, R12 Keywords: Prefectural economy, Japan, Stochastic model, Markov chain
    Date: 2012–10
  11. By: Conrad, Christian; Loch, Karin
    Abstract: We investigate the relationship between long-term U.S. stock market risks and the macroeconomic environment using a two component GARCH-MIDAS model. Our results provide strong evidence in favor of counter-cyclical behavior of long-term stock market volatility. Among the various macro variables in our dataset the term spread, housing starts, corporate profits and the unemployment rate have the highest predictive ability for stock market volatility . While the term spread and housing starts are leading variables with respect to stock market volatility, for corporate profits and the unemployment rate expectations data from the Survey of Professional Forecasters regarding the future development are most informative. Our results suggest that macro variables carry information on stock market risk beyond that contained in lagged realized volatilities, in particular when it comes to long-term forecasting.
    Keywords: Volatility Components; MIDAS; Survey Data; Macro Finance Link
    JEL: C53 C58
    Date: 2012–10–05
  12. By: Matteo G. Richiardi
    Abstract: Forecasting based on random intercepts models requires imputation of the individual permanent effects to the simulated individuals. When these individuals enter the simulation with a history of past outcomes this involves sampling from conditional distributions, which might be unfeasible. I present a method for drawing individual permanent effects from a conditional distribution which only requires to invert the corresponding estimated unconditional distribution. While the algorithms currently available in the literature require polynomial time, the proposed method only requires matching two ranks and works therefore in N lnN time.
    Keywords: forecasting, microsimulation, random intercept models, unobserved heterogeneity
    JEL: C15 C53 C63
    Date: 2012
  13. By: Andrle, Michal
    Abstract: The paper introduces methods that allow analysts to (i) decompose the estimates of unobserved quantities into observed data and (ii) impose subjective prior constraints on path estimates of unobserved shocks in structural economic models. For instance, decomposition of output gap to output, inflation, interest rates and other observables contribution is feasible. The intuitive nature and the analytical clarity of procedures suggested are appealing for policy-related and forecasting models. The paper brings some of the power embodied in the theory of linear multivariate filters, namely relatinship between Kalman and Wiener-Kolmogorov filtering, into the area of structural multivariate models, expressed in linear state-space form.
    Keywords: filter; DSGE; state-space; observables decomposition; judgement
    JEL: C10 E50
    Date: 2012–10
  14. By: Federico Pablo-Marti; Juan Luis Santos; Antonio Gacía-Tabuenca; María Teresa Gallo; Tomás Mancha
    Abstract: The research in the topic of industrial districts has been focused on the identification of which industries are forming industrial districts and on the causes behind the development of the clusters. As well as there are historical and efficiency reasons that are behind the current configuration of the industrial districts, up to now it seemed not crucial to clarify how different public policies affect the structure and relationships between the enterprises that are included in the clusters. With the use of an agent-based model we can analyze and forecast how each enterprise will change in stochastic terms. Moreover, it make feasible to predict changes in the size and structure of clusters and possible spillovers. ABMs are based on the assumption in which the economy fluctuates according to the behaviour of agents, which react in a proactive way. This difference makes ABMs an accurate tool for forecasting during crisis taking into account both changes in expectations and in policy instruments. In conventional models interactions are indirect, but agent-based modeling (ABM) allow simulating a plenty of shifts in agents’ behaviour through imitation or in their strategies according to the behaviour of the majority. These capabilities applied to firms permit to modify many not explicit assumptions incorporated into the majority of conventional models with the objective of predicting changes in the size and structure of industrial districts. Moreover, ABM allow making simulations changing parameters included in one or several public policies and obtaining the effects of these policies on clusters, accordingly to their own characteristics. The starting point is the building, trough statistical matching techniques making use of microdata sources, of a general database that replicates the attributes and location of all individuals and companies located in a specific spatial context. Then, behaviours are established for both companies and individuals who are interacting according to their preferences and endowments. In addition to these agents we include a raster of locations, built through downscaling techniques and display the current situation of different policies, in order to measure properly the changes introduced for making simulations. Finally, it would be possible to identify with high accuracy each cluster and its different characteristics. This permits to forecast and simulate the impact of changes in public policies on clusters structure and performance in stochastic terms thus enabling a better assessment of policy outcomes taking into account the robustness of the effect, related to the stochastic nature of the aggregated results. That is, ABM will allow us a better assessment of both policy outcomes and the certainty about the results. JEL: L52, R12, R58 Key words: Agent-based model, policy evaluation, industrial districts
    Date: 2012–10
  15. By: Martin Gonzalez Rozada; Martin sola; Constantino Hevia; Fabio Spagnolo
    Abstract: In this paper we estimate the yield curve of U.S. government bonds using a Markov switching latent variable model. We show how measures such as the level, slope, and curvature of the yield curve are a¤ected by business cycle conditions. We present a switching latent model which not only seem to capture this features in sample but also performs well out of sample.
    Keywords: Yield Curve; Term structure of interest rates, Markov regime switching; Maxi- mum likelihood; Risk premium.
    JEL: C13 C22 E43
    Date: 2012–07

This nep-for issue is ©2012 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.