nep-for New Economics Papers
on Forecasting
Issue of 2012‒11‒03
fourteen papers chosen by
Rob J Hyndman
Monash University

  1. Was the Recent Downturn in US GDP Predictable? By Mehmet Balcilar; Rangan Gupta; Anandamayee Majumdar; Stephen M. Miller
  2. A Forty Year Assessment of Forecasting the Boat Race By Geert Mesters; Siem Jan Koopman
  3. Moving Average Stochastic Volatility Models with Application to Inflation Forecast By Joshua C C Chan
  4. Can we use seasonally adjusted indicators in dynamic factor models? By Camacho, Maximo; Pérez-Quirós, Gabriel
  5. 1/N versus Mean-Variance: What if we can forecast? By Allen, D.; Lizieri, C.; Satchell, S.
  6. Forecasting the Index of Financial Safety (IFS) of South Africa using neural networks By Matkovskyy, Roman
  7. Reconciling Performance and Interpretability in Customer Churn Prediction using Ensemble Learning based on Generalized Additive Models By K. W. DE BOCK; D. VAN DEN POEL
  8. The Relevant Length of Customer Event History for Churn Prediction: How long is long enough? By M. BALLINGS; D. VAN DEN POEL
  9. Bootstrap prediction mean squared errors of unobserved states based on the Kalman filter with estimated parameters. By Rodríguez, Alejandro; Ruiz, Esther
  10. On the predictive power of implied volatility indexes: A comparative analysis with GARCH forecasted volatility By Bentes, Sonia R; Menezes, Rui
  11. Enhanced Decision Support in Credit Scoring Using Bayesian Binary Quantile Regression By V. L. MIGUÉIS; D. F. BENOIT; D. VAN DEN POEL
  12. The Index of the Financial Safety (IFS) of South Africa and Bayesian Estimates for IFS Vector-Autoregressive Model By Matkovskyy, Roman
  13. A nonparametric Bayesian approach for counterfactual prediction with an application to the Japanese private nursing home market By Sugawara, Shinya
  14. Can we evaluate the predictability of financial markets?. By Ruiz, Esther; Crato, Nuria

  1. By: Mehmet Balcilar (Department of Economics, Eastern Mediterranean University, Famagusta, North Cyprus,via Mersin 10, Turkey); Rangan Gupta (Department of Economics, University of Pretoria); Anandamayee Majumdar (Department of Biostatistics, University of North Texas Health Science Center, School of Public Health, Fort Worth, Texas, 76107, USA); Stephen M. Miller (College of Business, University of Las Vegas, Nevada)
    Abstract: This paper uses small set of variables-- real GDP, the inflation rate, and the short-term interest rate -- and a rich set of models -- athoeretical and theoretical, linear and nonlinear, as well as classical and Bayesian models -- to consider whether we could have predicted the recent downturn of the US real GDP. Comparing the performance by root mean squared errors of the models to the benchmark random-walk model, the two theoretical models, especially the nonlinear model, perform well on the average across all forecast horizons in out-of-sample forecasts, although at specific forecast horizons certain nonlinear athoeretical models perform the best. The nonlinear theoretical model also dominates in our ex ante forecast of the Great Recession, suggesting that developing forward-looking, microfounded, nonlinear, dynamic-stochastic-general-equilibrium models of the economy, may prove crucial in forecasting turning points.
    Keywords: Forecasting, Linear and non-linear models
    JEL: C32 R31
    Date: 2012–10
  2. By: Geert Mesters (VU University Amsterdam); Siem Jan Koopman (VU University Amsterdam)
    Abstract: We study the forecasting of the yearly outcome of the Boat Race between Cambridge and Oxford. We compare the relative performance of different dynamic models for forty years of forecasting. Each model is defined by a binary density conditional on a latent signal that is specified as a dynamic stochastic process with fixed predictors. The out-of-sample predictive ability of the models is compared between each other by using a variety of loss functions and predictive ability tests. We find that the model with its latent signal specified as an autoregressive process cannot be outperformed by the other specifications. This model is able to correctly forecast 30 out of 40 outcomes of the Boat Race.
    Keywords: Binary time series; Predictive ability; Non-Gaussian state space model
    JEL: C32 C35
    Date: 2012–10–23
  3. By: Joshua C C Chan
    Abstract: Moving average and stochastic volatility are two important components for modeling and forecasting macroeconomic and financial time series. The former aims to capture short-run dynamics, whereas the latter allows for volatility clustering and time-varying volatility. We introduce a new class of models that includes both of these useful features. The new models allow the conditional mean process to have a state space form. As such, this general framework includes a wide variety of popular specifications, including the unobserved components and time-varying parameter models. Having a moving average process, however, means that the errors in the measurement equation are no longer serially independent, and estimation becomes more difficult. We develop a posterior simulator that builds upon recent advances in precision-based algorithms for estimating this new class of models. In an empirical application involving U.S. inflation we find that these moving average stochastic volatility models provide better in-sample fitness and out-of-sample forecast performance than the standard variants with only stochastic volatility.
    JEL: C11 C51 C53
    Date: 2012–10
  4. By: Camacho, Maximo; Pérez-Quirós, Gabriel
    Abstract: We examine the short-term performance of two alternative approaches of forecasting from dynamic factor models. The first approach extracts the seasonal component of the individual indicators before estimating the dynamic factor model, while the alternative uses the non seasonally adjusted data in a model that endogenously accounts for seasonal adjustment. Our Monte Carlo analysis reveals that the performance of the former is always comparable to or even better than that of the latter in all the simulated scenarios. Our results have important implications for the factor models literature because they show the that the common practice of using seasonally adjusted data in this type of models is very accurate in terms of forecasting ability. Using five coincident indicators, we illustrate this result for US data.
    Keywords: factor models; seasonal adjustment; short-term forecasting
    JEL: C22 E27 E32
    Date: 2012–10
  5. By: Allen, D.; Lizieri, C.; Satchell, S.
    Abstract: Mean-variance optimisation has been roundly criticised by financial economists and practitioners alike, leading many to advocate a simple 1/N weighting heuristic. We investigate the performance of the Markowitz technique conditional on investor forecasting ability. Using a novel analytical approach, we demonstrate that investors with a modicum of forecasting ability can employ mean-variance to significantly increase their ex ante utility, outperforming the 1/N rule.
    Keywords: Portfolio Choice; Investment Decisions; Financial Forecasting and Simulation
    JEL: G11 G17
    Date: 2012–10–19
  6. By: Matkovskyy, Roman
    Abstract: This paper investigates neural network tools, especially the nonlinear autoregressive model with exogenous input (NARX), to forecast the future conditions of the Index of Financial Safety (IFS) of South Africa. Based on the time series that was used to construct the IFS for South Africa (Matkovskyy, 2012), the NARX model was built to forecast the future values of this index and the results are benchmarked against that of Bayesian Vector-Autoregressive Models. The results show that the NARX model applied to IFS of South Africa and trained by the Levenberg-Marquardt algorithm may ensure a forecast of adequate quality with less computation expanses, compared to BVAR models with different priors.
    Keywords: Index of Financial Safety (IFS); neural networks; nonlinear dynamic network (NDN); nonlinear autoregressive model with exogenous input (NARX); forecast
    JEL: C45 E44 G01
    Date: 2012–08
    Abstract: To build a successful customer churn prediction model, a classification algorithm should be chosen that fulfills two requirements: strong classification performance and a high level of model interpretability. In recent literature, ensemble classifiers have demonstrated superior performance in a multitude of applications and data mining contests. However, due to an increased complexity they result in models that are often difficult to interpret. In this study, GAMensPlus, an ensemble classifier based upon generalized additive models (GAMs), in which both performance and interpretability are reconciled, is presented and evaluated in a context of churn prediction modeling. The recently proposed GAMens, based upon Bagging, the Random Subspace Method and semiparametric GAMs as constituent classifiers, is extended to include two instruments for model interpretability: generalized feature importance scores, and bootstrap confidence bands for smoothing splines. In an experimental comparison on data sets of six real-life churn prediction projects, the competitive performance of the proposed algorithm over a set of well-known benchmark algorithms is demonstrated in terms of four evaluation metrics. Further, the ability of the technique to deliver valuable insight into the drivers of customer churn is illustrated in a case study on data from a European bank. Firstly, it is shown how the generalized feature importance scores allow the analyst to identify the importances of churn predictors in function of the criterion that is used to measure the quality of the model predictions. Secondly, the ability of GAMensPlus to identify nonlinear relationships between predictors and churn probabilities is demonstrated.
    Keywords: Database marketing, customer churn prediction, ensemble classification, generalized additive models (GAMs), GAMens, model interpretability
    Date: 2012–08
    Abstract: The key question of this study is: How long should the length of customer event history be for customer churn prediction? While most studies in predictive churn modeling aim to improve models by data augmentation or algorithm improvement, this study focuses on a another dimension: time window optimization with respect to predictive performance. This paper first presents a formalization of the time window selection strategy, along with a literature review. Next, using logistic regression, classification trees and bagging in combination with classification trees, this study analyzes the improvement in churn-model performance by extending customer event history from 1 to 16 years. The results show that, after the 5th additional year, predictive performance is only marginally increased, meaning that the company in this study can discard 69% of its data with almost no decrease in predictive performance. The practical implication is that analysts can substantially decrease datarelated burdens, such as data storage, preparation and analysis. This is particularly valuable in times of big data where computational complexity is paramount.
    Keywords: Predictive Analytics, Time window, Length of customer event history, predictive customer churn model
    Date: 2012–08
  9. By: Rodríguez, Alejandro; Ruiz, Esther
    Abstract: In the context of linear state space models with known parameters, the Kalman filter (KF) generates best linear unbiased predictions of the underlying states together with their corresponding Prediction Mean Square Errors (PMSE). However, in practice, when the filter is run with the parameters substituted by consistent estimates, the corresponding PMSE do not take into account the parameter uncertainty. Consequently, they underestimate their true counterparts. In this paper, we propose two new bootstrap procedures to obtain PMSE of the unobserved states designed to incorporate this latter uncertainty. We show that the new bootstrap procedures have better finite sample properties than bootstrap alternatives and than procedures based on the asymptotic approximation of the parameter distribution. The proposed procedures are implemented for estimating the PMSE of several key unobservable US macroeconomic variables as the output gap, the Non-accelerating Inflation Rate of Unemployment (NAIRU), the long-run investment rate and the core inflation. We show that taking into account the parameter uncertainty may change their prediction intervals and, consequently, the conclusions about the utility of the NAIRU as a macroeconomic indicator for expansions and recessions.
    Keywords: NAIRU; Output gap; Parameter uncertainty; Prediction intervals; State space models;
    Date: 2012
  10. By: Bentes, Sonia R; Menezes, Rui
    Abstract: This paper examines the behavior of several implied volatility indexes in order to compare them with the volatility forecasts obtained from estimating a GARCH model. Though volatility has always been a prevailing subject of research it has become particularly relevant given the increasingly complexity and uncertainty of stock markets in these days. An important measure to assess the market expectations of the future volatility of the underlying asset is the implied volatility (IV) indexes. Generally, these indexes are calculated based on the prices of out-of-the money put and call options on the underlying asset. Sometimes called the “investor fear gauge”, the IV indexes are a measure of the implied volatility of the underlying index. This study focuses on the implied and GARCH forecasted volatility of some emerging countries and some developed countries. More specifically, it compares the predictive power of the IV indexes with the ones provided by standard volatility models such as the ARCH/GARCH (Autoregressive Conditional Heteroskedasticity Model/ Generalized Autoregressive Conditional Heteroskedasticity Model) type models. Finally, a debate of the results is also provided.
    Keywords: implied volatility; volatility forecasts; GARCH models; volatility indices
    JEL: F37 C32 C01
    Date: 2012–10–24
    Abstract: Fierce competition as well as the recent financial crisis in financial and banking industries made credit scoring gain importance. An accurate estimation of credit risk helps organizations to decide whether or not to grant credit to potential customers. Many classification methods have been suggested to handle this problem in the literature. This paper proposes a model for evaluating credit risk based on binary quantile regression, using Bayesian estimation. This paper points out the distinct advantages of the latter approach: that is (i) the method provides accurate predictions of which customers may default in the future, (ii) the approach provides detailed insight into the effects of the explanatory variables on the probability of default, and (iii) the methodology is ideally suited to build a segmentation scheme of the customers in terms of risk of default and the corresponding uncertainty about the prediction. An often studied dataset from a German bank is used to show the applicability of the method proposed. The results demonstrate that the methodology can be an important tool for credit companies that want to take the credit risk of their customer fully into account.
    Keywords: Credit Scoring, Quantile regression, Classification, Bayesian estimation, Markov Chain Monte Carlo
    Date: 2012–08
  12. By: Matkovskyy, Roman
    Abstract: This paper proposes an approach to explore the strength of the financial system of a country against the possibility of financial perturbations appearing based on the construction of the Index of Financial Safety (IFS) of a country. The Markov Chain Monte Carlo (MCMC) and Gibbs sampler technique is used to estimate a Bayesian Vector Autoregressive Model of the IFS of South Africa for the period 1990Q1-2011Q1 and to forecast its value over the period 2011Q2-2017Q1. It is shown that the IFS could capture the disturbances in the financial system and the BVAR models with the non-informative and Minnesota priors could predict the future dynamics of IFS with sufficient accuracy.
    Keywords: Financial safety; index of financial safety (IFS); Bayesian Vector Autoregressive (BVAR) model; MCMC; Gibbs sampler; South Africa
    JEL: C15 E47 C11 C01 G01
    Date: 2012–04
  13. By: Sugawara, Shinya
    Abstract: This paper proposes a new inferential framework for structural econometric models using a nonparametric Bayesian approach. Although estimation methods based on moment conditions can employ a flexible estimation without distributional assumptions, they have difficulty conducting a prediction analysis. I propose a nonparametric Bayesian methodology for an estimation and prediction analysis. My methodology is applied to an empirical analysis of the Japanese private nursing home market. This market has a sticky economic circumstance, and my prediction simulates an intervention that removes this circumstance. The prediction result implies that the outdated circumstance in this market is harmful for consumers today.
    Keywords: Nonparametric Bayes; Nonlinear simultaneous equation model; Prediction; Industrial organization; Nursing home; Long-term care in Japan
    JEL: J14 L11 C11
    Date: 2012–10–23
  14. By: Ruiz, Esther; Crato, Nuria
    Abstract: In the summer of 2007, the International Institute of Forecasters (IIF) asked us to organize a workshop on ‘‘Predictability of Financial Markets’’, which took place on the 16th and 17th of January, 2009, in the beautiful, historical building of the Institute of Economics and Management (ISEG) in Lisbon. Nine outstanding invited speakers presented papers in the area of time series analysis of financial data, which were then discussed by nine other experts. This special issue provides peerreviewed, corrected and updated versions of seven of these papers, with additional comments by the discussants. Sadly, the paper by the Nobel Laureate Sir Clive Granger could not be finished, as he passed away in May 2009. We will always have the memories of his talk, which has been commented on here by Antonio García-Ferrer. In addition, the paper by Stephen Taylor entitled ‘‘A multi-horizon comparison on density forecasts for the S&P 500 index returns and option prices’’ was unfortunately not made available for this special issue.
    Keywords: Financial markets;
    Date: 2012

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.