Forecasting papers 2008-10-01

In this issue we have: Forecasting Unemployment Rate Using a Neural Network with Fuzzy Inference System ; Should quarterly government finance statistics be used for fiscal surveillance in Europe? ; The effect of project schedule adherence and rework on the duration forecast accuracy of earned value metrics ; Real-time Prediction with UK Monetary Aggregates in the Presence of Model Uncertainty ; and more.

  • Forecasting Unemployment Rate Using a Neural Network with Fuzzy Inference System
    Date: 2008-09-24
    By: George Atsalakis (Data Analysis and Forecasting Laboratory, Technical University of Crete, GREECE)
    Camelia Ioana Ucenic (University of Crete – Technical University Cluj Napoca)
    Christos Skiadas (Data Analysis and Forecasting Laboratory, Technical University of Crete, GREECE)
    Greece is a low-productivity economy with an ineffective welfare state, relying almost exclusively on low wages and social transfers. Failure to come to terms with this reality hampers both the appropriateness of EU recommendations and the Greek government's capacity to deal with unemployment. Rather than finding a job in a family business or through relationship contacts, young people stay unemployed. Nor can people move back to their village of origin so easily. The underground economy, and the mass of small companies which characterize the Greek economy are booming, on paper. One in three members of the workforce are "self-employed", compared to one in seven in the EU as a whole. (International Viewpoint) An unemployed person in Greece is 2,15 times more likely to suffer poverty than a person in employment. Yet in Greece there are perhaps even more influential factors in determining increased risk of poverty. Th us while unemployment is a crucial factor in the risk of poverty, it is neither the only nor the most significant factor. The paper presents a new technique in the field of unemployment modeling in order to forecast unemployment index. Techniques from the Artificial Neural Networks and from fuzzy logic have been combined to generate a neuro-fuzzy model. The input is a time series. Classical statistics measures are calculated in order to asses the model performance. Further the results are compared with an ARMA and an AR model.
    Keywords: forecasting, neural network, unemployment
  • Should quarterly government finance statistics be used for fiscal surveillance in Europe?
    Date: 2008-09
    By: Javier J. Pérez (European Central Bank, Kaiserstrasse 29, D-60311 Frankfurt am Main, Germany.)
    Diego J. Pedregal (Universidad de Castilla-la-Mancha, Real Casa de La Misericordia, Calle Altagracia, 50, 13003 Ciadad Real, Spain.)
    We use a newly available dataset of euro area quarterly national accounts fiscal data and construct multivariate, state-space mixed-frequencies models for the government deficit, revenue and expenditure in order to assess its information content and its potential use for fiscal forecasting and monitoring purposes. The models are estimated with annual and quarterly national accounts fiscal data, but also incorporate monthly information taken from the cash accounts of the governments. The results show the usefulness of our approach for real-time fiscal policy surveillance in Europe, given the current policy framework in which the relevant official figures are expressed in annual terms. JEL Classification: C53, E6, H6.
    Keywords: Fiscal policies, Mixed frequency data, Forecasting, Unobserved Components Models, State Space, Kalman Filter.
  • The effect of project schedule adherence and rework on the duration forecast accuracy of earned value metrics
    Date: 2008-06
    Earned Value Management (EVM) in project management integrates cost, schedule and technical performance and allows the calculation of cost and schedule variances, performance indices and forecasts of project cost and schedule duration. The earned value method provides early indicators of project performance to reveal opportunities and/or highlight the need for eventual corrective actions.<br>The introduction of the earned schedule (ES) method in 2003 has led to an increasing attention on the forecast accuracy of EVM to predict a project's final duration. Previous research has shown that the ES method outperforms the more traditional predictive metrics for project duration forecasting.<br>In this paper we critically review and test a novel ES extension, the p-factor approach, to measure schedule adherence based on the traditional earned value metrics. A large set of fictive project networks has been cons tructed under a controlled design and performance is measured by means of Monte Carlo simulations.
  • Real-time Prediction with UK Monetary Aggregates in the Presence of Model Uncertainty
    Date: 2008-08
    By: Anthony Garratt
    Gary Koop
    Emi Mise
    Shaun Vahey (Reserve Bank of New Zealand)
    A popular account for the demise of the UK's monetary targeting regime in the 1980s blames the fluctuating predictive relationships between broad money and inflation and real output growth. Yet ex post policy analysis based on heavily-revised data suggests no fluctuations in the predictive content of money. In this paper, we investigate the predictive relationships for inflation and output growth using both real-time and heavily-revised data. We consider a large set of recursively estimated Vector Autoregressive (VAR) and Vector Error Correction models (VECM). These models differ in terms of lag length and the number of cointegrating relationships. We use Bayesian model averaging (BMA) to demonstrate that real-time monetary policymakers faced considerable model uncertainty. The in-sample predictive content of money fluctuated during the 1980s as a result of data revisions in the presence of model uncertainty. Thi s feature is only apparent with real-time data as heavily-revised data obscure these fluctuations. Out of sample predictive evaluations rarely suggest that money matters for either inflation or real output. We conclude that both data revisions and model uncertainty contributed to the demise of the UK's monetary targeting regime. Classification-C11, C32, C53, E51, E52
  • ESeC-Rubin Missing Value Interpretation for a Regional Bottom-Up Hierarchical Forecasting
    Date: 2008-09
    By: Antonio Anselmi (SAS Institute)
    Paola Maddalena Chiodini (Department of Statistics, University of Milano – Bicocca)
    Flavio Verrecchia (ESeC)
    in letteratura, per l'imputazione dei dati mancanti nelle serie storiche, si fa riferimento a statistiche applicate all'intera serie analizzata (e.g. media di tutti i termini della serie), ottenendo una costante d'imputazione generalmente adeguata per una specifica serie. Se le serie sono n (n -> infinito) è impossibile trovare un'unica funzione per le n costanti di imputazione dei missing. Obiettivo del lavoro è proporre un nuovo metodo di imputazione dei dati mancanti – ESeC-Rubin – per basi dati gerarchiche finalizzato alla modellistica temporale. In particolare, la ESeC-Rubin consente di ricostruire il dato mancante tenendo conto di una sequenza di metodi di imputazione e della naturale variabilità degli aggregati studiati. La metodologia proposta in questo lavoro trova ispirazione dalla teoria dei campioni dove non di rado si deve trovare la miglior soluzione possibile al problema del missing. In questo contesto la soluzione che si cerca di dare è quella di ricostruire il dato mancante tenendo conto della naturale variabilità del fenomeno allo studio (Rubin 1987, 1996; Hergoz e Rubin, 1983; Rubin e Shenker, 1986). In effetti la letteratura in tal senso fornisce una gamma piuttosto articolata di strategie che possono di volta in volta essere utilizzate in quanto appare immediatamente evidente che la soluzione non può essere unica e generalizzata. Infine, si presenterà una applicazione della ESeC-Rubin su dati socio-economici di fonte Eurostat. L'applicazione prodotta con SAS Forecast Server consente di comparare i modelli (selezionati in automatico) a partire dalla base dati osservata con missing values e differenti tipologie di imputazione.
    Keywords: Missing Value, Index Number, CAGR Imputation, Stochastic Imputation, ESeC-Rubin Imputation, Regional Bottom-Up Hierarchical Forecasting
    JEL: C8 C43 C16 C22
  • Seasonal Mackey-Glass-GARCH process and short-term dynamics
    Date: 2008-09
    By: Catherine Kyrtsou (Department of Economics, University of Macedonia)
    Michel Terraza (Department of Economics, LAMETA)
    The aim of this article is the study of complex structures which are behind the short-term predictability of stock returns series. In this regard, we employ a seasonal version of the Mackey-Glass-GARCH(p,q) model, initially proposed by Kyrtsou and Terraza (2003) and generalized by Kyrtsou (2005, 2006). It has either negligible or significant autocorrelations in the conditional mean, and a rich structure in the conditional variance. To reveal short or long memory components and non-linear structures in the French Stock Exchange (CAC40) returns series, we apply the test of Geweke and Porter-Hudak (1983), the Brock et al. (1996) and Dechert (1995) tests, the correlation-dimension method of Grassberger and Procaccia (1983), the Lyapunov exponents method of Gencay and Dechert (1992), and the Recurrence Quantification Analysis introduced by Webber and Zbilut (1994). As a confirmation procedure of the dynamics generating future movements in CAC40, we forecast the return series using a seasonal Mackey-Glass-GARCH(1,1) model. The interest of the forecasting exercise is found in the inclusion of high-dimensional non-linearities in the mean equation of returns.
    Keywords: Noisy chaos, short-term dynamics, correlation dimension, Lyapunov exponents, recurrence quantifications, forecasting.
    JEL: C49 C51 C52 C53 D84 G12 G14
  • Disagreement and Biases in Inflation Expectations
    Date: 2008-09-19
    By: Carlos Capistrán
    Allan Timmermann (School of Economics and Management, University of Aarhus, Denmark)
    Disagreement in inflation expectations observed from survey data varies systematically over time in a way that reflects the level and variance of current inflation. This paper offers a simple explanation for these facts based on asymmetries in the forecasters' costs of over- and under-predicting inflation. Our model implies (i) biased forecasts; (ii) positive serial correlation in forecast errors; (iii) a cross-sectional dispersion that rises with the level and the variance of the inflation rate; and (iv) predictability of forecast errors at different horizons by means of the spread between the short- and long-term variance of inflation. We find empirically that these patterns are present in inflation forecasts from the Survey of Professional Forecasters. A constant bias component, not explained by asymmetric loss and rational expectations, is required to explain the shift in the sign of the bias observed for a s ubstantial portion of forecasters around 1982.
    Keywords: asymmetric loss, real-time data, survey expectations
    JEL: C53 C82 E31 E37
  • The Resolution of Macroeconomic Uncertainty: Evidence from Survey Forecast
    Date: 2008-09-19
    By: Andrew J. Patton
    Allan Timmermann (School of Economics and Management, University of Aarhus, Denmark)
    We develop an unobserved components approach to study surveys of forecasts containing multiple forecast horizons. Under the assumption that forecasters optimally update their beliefs about past, current and future state variables as new information arrives, we use our model to extract information on the degree of predictability of the state variable and the importance of measurement errors on that variable. Empirical estimates of the model are obtained using survey forecasts of annual GDP growth and inflation in the US with forecast horizons ranging from 1 to 24 months. The model is found to closely match the joint realization of forecast errors at different horizons and is used to demonstrate how uncertainty about macroeconomic variables is resolved.
    Keywords: Fixed-event forecasts, multiple forecast horizons, Kalman filtering, survey data
  • Forecast Combination With Entry and Exit of Experts
    Date: 2008-09-19
    By: Carlos Capistrán
    Allan Timmermann (School of Economics and Management, University of Aarhus, Denmark)
    Combination of forecasts from survey data is complicated by the frequent entry and exit of individual forecasters which renders conventional least squares regression approaches infeasible. We explore the consequences of this issue for existing combina- tion methods and propose new methods for bias-adjusting the equal-weighted forecast or applying combinations on an extended panel constructed by back-filling missing ob- servations using an EM algorithm. Through simulations and an application to a range of macroeconomic variables we show that the entry and exit of forecasters can have a large effect on the real-time performance of conventional combination methods. The bias-adjusted combination method is found to work well in practice.
    Keywords: Real-time Data, Survey of Professional Forecasters, Bias-adjustment, EM Algorithm.
  • Glossary to ARCH (GARCH)
    Date: 2008-09-04
    By: Tim Bollerslev (School of Economics and Management, University of Aarhus, Denmark)
    The literature on modeling and forecasting time-varying volatility is ripe with acronyms and abbreviations used to describe the many different parametric models that have been put forth since the original linear ARCH model introduced in the seminal Nobel Prize winning paper by Engle (1982). The present paper provides an easy-to-use encyclopedic reference guide to this long list of ARCH acronyms. In addition to the acronyms associated with specific parametric models, I have also included descriptions of various abbreviations associated with more general statistical procedures and ideas that figure especially prominently in the ARCH literature.
    Keywords: (G)ARCH, Volatility models
    JEL: C22
  • Taken from the NEP-FOR mailing list edited by Rob Hyndman.