Forecasting papers 2009-05-26

In this issue we have Exponential smoothing and non-negative data, Using Large Data Sets to Forecast Housing Prices, Evaluating Short-Run Forecasting Properties of the KOF Employment Indicator for Switzerland in Real Time, Estimating and Forecasting Asset Volatility and Its Volatility, and more.

  1. Exponential smoothing and non-negative data
    Date: 2008-07
    By: Muhammad Akram (Department of Econometrics and Business Statistics)
    Rob J Hyndman (Department of Econometrics and Business Statistics,Monash University)
    J. Keith Ord (McDonough School of Business,Georgetown University)
    The most common forecasting methods in business are based on exponential smoothing and the most common time series in business are inherently non-negative. Therefore it is of interest to consider the properties of the potential stochastic models underlying exponential smoothing when applied to non-negative data. We explore exponential smoothing state space models for non-negative data under various assumptions about the innovations, or error, process. We first demonstrate that prediction distributions from some commonly used state space models may have an infinite variance beyond a certain forecasting horizon. For multiplicative error models which do not have this flaw, we show that sample paths will converge almost surely to zero even when the error distribution is non-Gaussian. We propose a new model with similar properties to exponential smoothing, but which does not have these problems, and we develop some distributi! onal properties for our new model. We then explore the implications of our results for inference, and compare the short-term forecasting performance of the various models using data on the weekly sales of over three hundred items of costume jewelry. The main findings of the research are that the Gaussian approximation is adequate for estimation and one-step-ahead forecasting. However, as the forecasting horizon increases, the approximate prediction intervals become increasingly problematic. When the model is to be used for simulation purposes, a suitably specified scheme must be employed.
    Keywords: forecasting; time series; exponential smoothing; positive-valued processes; seasonality; state space models.
    JEL: C1
  2. Using Large Data Sets to Forecast Housing Prices: A Case Study of Twenty US States
    Date: 2009-05
    By: Rangan Gupta (Department of Economics, University of Pretoria)
    Alain Kabundi (Department of Economics and Econometrics, University of Johannesburg)
    Stephen M. Miller (Department of Economics, University of Nevada, Las Vegas)
    We implement several Bayesian and classical models to forecast housing prices in 20 US states. In addition to standard vector-autoregressive (VAR) and Bayesian vector autoregressive (BVAR) models, we also include the information content of 308 additional quarterly series in some models. Several approaches exist for incorporating information from a large number of series. We consider two approaches – extracting common factors (principle components) in a Factor-Augmented Vector Autoregressive (FAVAR) or Factor-Augmented Bayesian Vector Autoregressive (FABVAR) models or Bayesian shrinkage in a large-scale Bayesian Vector Autoregressive (LBVAR) models. In addition, we also introduce spatial or causality priors to augment the forecasting models. Using the period of 1976:Q1 to 1994:Q4 as the in-sample period and 1995:Q1 to 2003:Q4 as the out-of-sample horizon, we compare the forecast performance of the alternative models. Ba! sed on the average root mean squared error (RMSE) for the one-, two-, three-, and four-quarters-ahead forecasts, we find that one of the factor-augmented models generally outperform the large-scale models in the 20 US states examined in this paper.
    Keywords: Housing Prices, Forecasting, Factor Augmented Models, Large-Scale BVAR models
    JEL: C32
  3. Evaluating Short-Run Forecasting Properties of the KOF Employment Indicator for Switzerland in Real Time
    Date: 2009-05
    By: Boriss Siliverstovs (KOF Swiss Economic Institute, ETH Zurich, Switzerland)
    This study investigates the usefulness of the business tendency surveys collected at the KOF institute for short-term forecasting of employment in Switzerland aggregated in the KOF Employment Indicator. We use the real time dataset in order to simulate the actual predictive process using only the information that was available at the time when predictions were made. We evaluate the presence of predictive content of the KOF Employment Indicator both for nowcasts that are published two months before the first official release. We find that inclusion of the KOF Employment Indicator leads to substantial improvement both in in-sample as well as, more importantly, in out-of-sample prediction accuracy. This conclusion holds both for nowcasts and one-quarter ahead forecasts.
    Keywords: Business tendency surveys, Forecasting, Real-time data, Bayesian model averaging, Employment
    JEL: C11
  4. Estimating and Forecasting Asset Volatility and Its Volatility: A Markov-Switching Range Model
    Date: 2009-05
    By: Jan Piplack
    This paper proposes a new model for modeling and forecasting the volatility of asset markets. We suggest to use the log range defined as the natural logarithm of the difference of the maximum and the minimum price observed for an asset within a certain period of time, i.e. one trading week. There is clear evidence for a regime-switching behavior of the volatility of the S&P500 stock market index in the period from 1962 until 2007. A Markov-switching model is found to fit the data significantly better than a linear model, clearly distinguishing periods of high and low volatility. A forecasting exercise leads to promising results by showing that some specifications of the model are able to clearly decrease forecasting errors with respect to the linear model in an absolute and mean square sense.
    Keywords: Volatility, range, Markov-switching, GARCH, forecasting.
    JEL: C53
  5. Modelling Sustainable International Tourism Demand to the Brazilian Amazon
    Date: 2009
    By: Jose Angelo Divino (Department of Economics Catholic University of Brasilia)
    Michael McAleer (Universidad Complutense de Madrid.Department of Quantitative Economics)
    The Amazon rainforest is one of the world's greatest natural wonders and holds great importance and significance for the world's environmental balance. Around 60% of the Amazon rainforest is located in the Brazilian territory. The two biggest states of the Amazon region are Amazonas (the upper Amazon) and Pará (the lower Amazon), which together account for around 73% of the Brazilian Legal Amazon, and are the only states that are serviced by international airports in Brazil's North region. The purpose of this paper is to model and forecast sustainable international tourism demand for the states of Amazonas, Pará, and the aggregate of the two states. By sustainable tourism is meant a distinctive type of tourism that has relatively low environmental and cultural impacts. Economic progress brought about by illegal wood extraction and commercial agriculture has destroyed large areas of the Amazon rainforest. The sust! ainable tourism industry has the potential to contribute to the economic development of the Amazon region without destroying the rainforest. The paper presents unit root tests for monthly and annual data, estimates alternative time series models and conditional volatility models of the shocks to international tourist arrivals, and provides forecasts for 2006 and 2007.
  6. Bayesian nonparametric estimators derived from conditional Gibbs structures
    Date: 2008-06
    By: Antonio Lijoi
    Igor Pruenster
    Stephen G. Walker
    We consider discrete nonparametric priors which induce Gibbs-type exchangeable random partitions and investigate their posterior behavior in detail. In particular, we deduce conditional distributions and the corresponding Bayesian nonparametric estimators, which can be readily exploited for predicting various features of additional samples. The results provide useful tools for genomic applications where prediction of future outcomes is required.
    Keywords: Bayesian nonparametric inference; Exchangeable random partitions; Generalized factorial coeffcients; Generalized gamma process; Poisson-Dirichlet process; Population genetics.
  7. On risk prediction
    Date: 2009-05-11
    By: Lönnbark, Carl (Department of Economics, Umeå University)
    This thesis comprises four papers concerning risk prediction. Paper [I] suggests a nonlinear and multivariate time series model framework that enables the study of simultaneity in returns and in volatilities, as well as asymmetric effects arising from shocks. Using daily data 2000-2006 for the Baltic state stock exchanges and that of Moscow we nd recursive structures with Riga directly depending in returns on Tallinn and Vilnius, and Tallinn on Vilnius. For volatilities both Riga and Vilnius depend on Tallinn. In addition, we find evidence of asymmetric effects of shocks arising in Moscow and in the Baltic states on both returns and volatilities. Paper [II] argues that the estimation error in Value at Risk predictors gives rise to underestimation of portfolio risk. A simple correction is proposed and in an empirical illustration it is found to be economically relevant. Paper [III] studies some approximation approaches to! computing the Value at Risk and the Expected Shortfall for multiple period asset returns. Based on the result of a simulation experiment we conclude that among the approaches studied the one based on assuming a skewed t distribution for the multiple period returns and that based on simulations were the best. We also found that the uncertainty due to the estimation error can be quite accurately estimated employing the delta method. In an empirical illustration we computed five day Value at Risk's for the S&P 500 index. The approaches performed about equally well. Paper [IV] argues that the practise used in the valuation of the portfolio is important for the calculation of the Value at Risk. In particular, when liquidating a large portfolio the seller may not face horizontal demand curves. We propose a partially new approach for incorporating this fact in the Value at Risk and in an empirical illustration we compare it to a competing approach. We find substantial diffe! rences.
    Keywords: Finance; Time series; GARCH; Estimation error; Asymmetry; Supply and demand
    JEL: C22

Taken from the NEP-FOR mailing list edited by Rob Hyndman.