nep-for New Economics Papers
on Forecasting
Issue of 2009‒11‒21
nine papers chosen by
Rob J Hyndman
Monash University

  1. Macroeconomic Forecasting and Structural Change By D'Agostino, Antonello; Gambetti, Luca; Giannone, Domenico; Giannone, Domenico
  2. Forecasting Sales By Franses, Ph.H.B.F.
  3. "Google it!" Forecasting the US unemployment rate with a Google job search index By D'amuri F; Marcucci J
  4. Google Searches as a Means of Improving the Nowcasts of Key Macroeconomic Variables By Konstantin A. Kholodilin; Maximilian Podstawski; Boriss Siliverstovs; Constantin Bürgi
  5. Forecasting long memory time series under a break in persistence By Heinen, Florian; Sibbertsen, Philipp; Kruse, Robinson
  6. Predicting Romanian Financial Distressed Companies By Madalina Andreica
  7. Economic Value of Realized Covariance Forecasts: The European Case By Dimitrios Vortelinos; Dimitrios Thomakos
  8. The Role of Monetary Aggregates in the Policy Analysis of the Swiss National Bank By Gebhard Kirchgässner; Jürgen Wolters
  9. The dynamic analysis and prediction of stock markets through the latent Markov model By De Angelis, L; Paas, L.J.

  1. By: D'Agostino, Antonello (Central Bank and Financial Services Authority of Ireland); Gambetti, Luca (Universitat Autonoma de Barcelona); Giannone, Domenico (ECARES, Université Libre de Bruxelles); Giannone, Domenico
    Abstract: The aim of this paper is to assess whether explicitly modeling structural change increases the accuracy of macroeconomic forecasts. We produce real time out-of-sample forecasts for inflation, the unemployment rate and the interest rate using a Time-Varying Coefficients VAR with Stochastic Volatility (TV-VAR) for the US. The model generates accurate predictions for the three variables. In particular for inflation the TVVAR outperforms, in terms of mean square forecast error, all the competing models: fixed coefficients VARs, Time-Varying ARs and the na¨ýve random walk model. These results are also shown to hold over the most recent period in which it has been hard to forecast inflation.
    JEL: C32 E37 E47
    Date: 2009–10
  2. By: Franses, Ph.H.B.F. (Erasmus Econometric Institute)
    Abstract: This chapter deals with forecasting sales (in units or money), where an explicit distinction is made between sales of durable goods (computers, cars, books) and sales of utilitarian products (SKU level in supermarkets). Invariably, sales forecasting amounts to a combination of statistical modeling and an expert’s touch. Models for durable goods sales are usually based on (variants of) the Bass model, while SKU sales forecasts are typically based on simple extrapolation methods. Forecast evaluation is not standard due to the interaction of model and expert.
    Keywords: sales forecasting;durable goods;SKU-level sales;diffusion;human judgment
    Date: 2009–11–09
  3. By: D'amuri F (Institute for Social and Economic Research); Marcucci J (Bank of Italy)
    Abstract: We suggest the use of an Internet job-search indicator (the Google Index, GI) as the best leading indicator to predict the US unemployment rate. We perform a deep out-of-sample forecasting comparison analyzing many models that adopt both our preferred leading indicator (GI), the more standard initial claims or combinations of both. We find that models augmented with the GI outperform the traditional ones in predicting the monthly unemployment rate, even in most state-level forecasts and in comparison with the Survey of Professional Forecasters.
    Date: 2009–11–18
  4. By: Konstantin A. Kholodilin; Maximilian Podstawski; Boriss Siliverstovs; Constantin Bürgi
    Abstract: The Google Insights data are a collection of recorded Internet searches for a huge number of the keywords, which are available since January 2004. These searches represent a kind of revealed perceptions of Internet users, which are a (possibly not entirely representative) sample of the general public. These data can be used to improve the short-term forecasts or nowcasts of various macroeconomic variables. In this paper, we compare the nowcasts of the growth rates of the real US private consumption based on both the conventional consumer confidence indicators and the Google indicators. The latter are extracted from the Google searches using the principal component analysis. It is shown that the Google indicators are especially successful at predicting private consumption in times of economic trouble, for they are 20% more accurate than the best alternative during the 2008m1-2009m5 forecast period. In addition, Google indicators are available at weekly frequency and not subject to revisions. This makes them an excellent source of information for the macroeconomic forecasting.
    Keywords: Google indicators, forecasting, principal components, US private consumption
    JEL: C22 C53 C82
    Date: 2009
  5. By: Heinen, Florian; Sibbertsen, Philipp; Kruse, Robinson
    Abstract: We consider the problem of forecasting time series with long memory when the memory parameter is subject to a structural break. By means of a large-scale Monte Carlo study we show that ignoring such a change in persistence leads to substantially reduced forecasting precision. The strength of this effect depends on whether the memory parameter is increasing or decreasing over time. A comparison of six forecasting strategies allows us to conclude that pre-testing for a change in persistence is highly recommendable in our setting. In addition we provide an empirical example which underlines the importance of our findings.
    Keywords: Long memory time series, Break in persistence, Structural change, Simulation, Forecasting competition
    JEL: C15 C22 C53
    Date: 2009–11
  6. By: Madalina Andreica
    Abstract: The study consisted in collecting financial information for a group of distressed and non-distressed Romanian listed companies during the period 2006–2008, in order to create early warning signals for financial distressed companies using the following methodologies: the Logistic and the Hazard model, the CHAID decision tree model and the Artificial Neural Network model (ANN). For each company a set of 14 financial ratios, that reflect the company’s profitability, solvency, asset utilization, growth ability and size, were calculated and then used in the study. A Principal Component Analysis was also used to reduce the dimensionality of the data space and to allow seeing that the 2 types of companies do form 2 distinct groups suggesting that the ratios used are useful enough to predict financial distress. The following 4 data sets were separately analyzed: first-year data to predict distress one year ahead, second-year data for a 2 year-ahead prediction, third-year data for a 3 year-ahead prediction, as well as cumulative three-year data to predict distress 1 year ahead by letting the ratios vary in time. For each data set, several prediction models were created using CHAID, the Logit and Hazard models as well as the ANN and the hybrid-ANN. The results are consistent with the theory and also to previous studies and the out-of-sample forecast accuracy of the estimated models of 73%-100% indicates that the proposed early warning models for the Romanian listed companies are quite efficient.
    Keywords: early warning signals, CHAID, ANN
    Date: 2009–11
  7. By: Dimitrios Vortelinos; Dimitrios Thomakos
    Abstract: This paper investigates the economic value of dierent non-parametric realized volatility estimates in Efficient Frontier, Global Minimum Variance,Capital Market Line and Capital Market Line with only positive weights portfolio types. The dataset concerns the CAC40 index, the DAX index and the General Index (GD) of the Athens Stock Exchange. We use the unrestricted realized volatility estimator, the realized optimally sampled volatility estimator and their bias-corrections against the benchmark of the daily squared returns. The value of switching from daily to intraday returns in estimating the covariance matrix is substantial. The type of realized volatility estimator used is also important. This is proven true according to the portfolio statistic measures (mean, standard deviation, Sharp Ratio and Cumulative Return), the basis points that a risk averse investor is willing to pay per year in order to gain from the realized covariance estimates instead of the daily squared returns, the proportion of times that the average portfolio return of the realized covariance forecasts is higher than the benchmark and the proportion of combinations of portfolio parameters for which the above proportion measure is higher than or equals to the 50% of the total combinations of portfolio parameters used.
    Keywords: CAC40, DAX, ASE, portfolios, covariance, realized volatility, bias-correction, optimal sampling, microstructure noise, forecast, evaluation.
    Date: 2009
  8. By: Gebhard Kirchgässner; Jürgen Wolters
    Abstract: Using Swiss data from 1983 to 2008, this paper investigates whether growth rates of the different measures of the quantity of money and or excess money can be used to forecast inflation. After a preliminary data analysis, money demand relations are specified, estimated and tested. Then, employing error correction models, measures of excess money are derived. Using recursive estimates, indicator properties of monetary aggregates for inflation are assessed for the period from 2000 onwards, with time horizons of one, two, and three years. In these calculations, M2 and M3 clearly outperform M1, and excess money is generally a better predictor than the quantity of money. Taking into account also the most (available) recent observations that represent the first three quarters of the economic crisis, the money demand function of M3 remains stable while the one for M2 is strongly influenced by these three observations. While in both cases forecasts for 2010 show inflation rates inside the target zone between zero and two percent, and the same holds for forecasts based on M3 for 2011, forecasts based on M2 provide evidence that the upper limit of this zone might be violated in 2011.
    Keywords: Stability of Money Demand, Monetary Aggregates and Inflation
    JEL: E41 E52
    Date: 2009–11
  9. By: De Angelis, L (Vrije Universiteit Amsterdam, Faculteit der Economische Wetenschappen en Econometrie (Free University Amsterdam, Faculty of Economics Sciences, Business Administration and Economitrics); Paas, L.J.
    Abstract: In this paper we show how the latent Markov model can be used to define different conditions in the stock market, called market- regimes. Changes in regimes can be used to detect financial crises, pinpoint the end of a crisis and predict future developments in the stock market, to some degree. The model is applied to changes in monthly price indexes of the Italian and US stock market in the period from January 2000 to July 2009.
    Keywords: Stock market pattern analysis; Regime-switching; Forecasting; Latent Markov model; Financial crises; Market stability periods
    Date: 2009

This nep-for issue is ©2009 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.