nep-for New Economics Papers
on Forecasting
Issue of 2005‒11‒12
nine papers chosen by
Rob J Hyndman
Monash University

  1. Forecasting Regional Employment in Germany by Means of Neural Networks and Genetic Algorithms By Roberto Patuelli; Simonetta Longhi; Aura Reggiani; Peter Nijkamp
  2. Multicriteria Analysis of Neural Network Forecasting Models: An Application to German Regional Labour Markets By Roberto Patuelli; Simonetta Longhi; Aura Reggiani; Peter Nijkamp
  3. A Rank-order Analysis of Learning Models for Regional Labor Market Forecasting By Roberto Patuelli; Simonetta Longhi; Aura Reggiani; Peter Nijkamp; Uwe Blien
  4. A Rank Approach to Equity Forecast Construction By S.E. Satchell; S.M.Wright
  5. Explaining exchange rate dynamics - the uncovered equity return parity condition By Elizaveta Krylova; Lorenzo Cappiello; Roberto A. De Santis
  6. Monetary Policy with Model Uncertainty: Distribution Forecast Targeting By Lars Svensson; Noah Williams
  7. Foreign Exchange Market Microstructure By Martin D. D. Evans (Georgetown University)
  8. Global inflation By Matteo Ciccarelli; Benoît Mojon
  9. Understanding Order Flow By Martin D. D. Evans; Richard K. Lyons

  1. By: Roberto Patuelli (Vrije Universiteit); Simonetta Longhi (University of Essex); Aura Reggiani (University of Bologna); Peter Nijkamp (Vrije Universiteit)
    Abstract: The aim of this paper is to develop and apply Neural Network (NN) models in order to forecast regional employment patterns in Germany. NNs are statistical tools based on learning algorithms with a distribution over a large amount of quantitative data. NNs are increasingly deployed in the social sciences as a useful technique for interpolating data when a clear specification of the functional relationship between dependent and independent variables is not available. In addition to traditional NN models, a further set of NN models will be developed in this paper, incorporating Genetic Algorithm (GA) techniques in order to detect the networks’ structure. GAs are computer-aided optimization tools that imitate natural biological evolution in order to find the solution that best fits the given case. Our experiments employ a data set consisting of a panel of 439 districts distributed over the former West and East Germany,. The West and East data sets have different time horizons, as employment information by district is available from 1987 and 1993 for West and East Germany, respectively. Separate West and East models are tested, before carrying out a unified experiment on the full data set for Germany. The above models are then evaluated by means of several statistical indicators, in order to test their ability to provide out- of-sample forecasts. A comparison between traditional and GAenhanced models is ultimately proposed. The results show that the West and East NN models perform with different degrees of precision, because of the different data sets’ time horizons.
    Keywords: forecasting; neural networks; regional labour markets
    JEL: C8
    Date: 2005–11–08
  2. By: Roberto Patuelli (Vrije Universiteit); Simonetta Longhi (University of Essex); Aura Reggiani (University of Bologna); Peter Nijkamp (Vrije Universiteit)
    Abstract: This paper develops a flexible multi-dimensional assessment method for the comparison of different statistical-econometric techniques based on learning mechanisms, with a view to analysing and forecasting regional labour markets. The aim of this paper is twofold. A first major objective is to explore the use of a standard choice tool, namely Multicriteria Analysis (MCA), in order to cope with the intrinsic methodological uncertainty on the choice of a suitable statistical- econometric learning technique for regional labour market analysis. MCA is applied here to support choices on the performance of various models – based on classes of Neural Network (NN) techniques – that serve to generate employment forecasts in West Germany at a regional/district level. A second objective of the paper is to analyse the methodological potential of a blend of approaches (NN-MCA) in order to extend the analysis framework to other economic research domains, where formal models are not available, but where a variety of statistical data is present. The paper offers a basis for a more balanced judgement of the performance of rival statistical tests.
    Keywords: multicriteria analysis; neural networks; regional labour markets
    JEL: C9
    Date: 2005–11–08
  3. By: Roberto Patuelli (Vrije Universiteit); Simonetta Longhi (University of Essex); Aura Reggiani (University of Bologna); Peter Nijkamp (Vrije Universiteit); Uwe Blien (Institut fuer Arbeitsmarkt und Berufsforschung)
    Abstract: Using a panel of 439 German regions we evaluate and compare the performance of various Neural Network (NN) models as forecasting tools for regional employment growth. Because of relevant differences in data availability between the former East and West Germany, NN models are computed separately for the two parts of the country. The comparisons of the models and their ex-post forecasts have been carried out by means of a non-parametric test: viz. the Friedman statistic. The Friedman statistic tests the consistency of model results obtained in terms of their rank order. Since there is no normal distribution assumption, this methodology is an interesting substitute for a standard analysis of variance. Furthermore, the Friedman statistic is indifferent to the scale on which the data are measured. The evaluation of the ex-post forecasts suggests that NN models are generally able to correctly identify the fastest-growing and the slowest-growing regions, and hence predict rather well the correct ranking of regions in terms of their employment growth. The comparison among NN models – on the basis of several criteria – suggests that the choice of the variables used in the model may influence the model’s performance and the reliability of its forecasts.
    Keywords: forecasts, regional employment, learning algorithms, rank order test
    JEL: C23 E27 R12
    Date: 2005–11–08
  4. By: S.E. Satchell; S.M.Wright
    Abstract: The purpose of this paper is to present a rank based approach to cross-sectional linear factor modelling. The emphasis is on approximating factor exposures in a consistent manner in order to facilitate the merging of subjective information (from professional investors) with objective information (from accounting data and/or state of the art quantitative models) in a statistically rigorous way without needing to impose the unrealistic simplifying assumptions typical of more standard time series models. We deal with the problems of identifying country and sector returns by an innovative hierarchical factor structure. This is all discussed from the perspective that investment models are not immutable but rather need to be designed with characteristics that are fit for their purpose; for example, returning aggregate county and sector forecasts that are consistent by construction.
    Keywords: : Linear Factor Models, Ranking, Robustness Exposures, Forecasting.
    JEL: G11
    Date: 2005–11
  5. By: Elizaveta Krylova (European Central Bank, Market Operations, Kaiserstrasse 29, 60311 Frankfurt am Main, Germany); Lorenzo Cappiello (DG-Research, European Central Bank, Kaiserstrasse 29, 60311 Frankfurt am Main, Germany); Roberto A. De Santis (DG Economics, European Central Bank, Kaiserstrasse 29, 60311 Frankfurt am Main, Germany)
    Abstract: By employing Lucas’ (1982) model, this study proposes an arbitrage relationship – the Uncovered Equity Return Parity (URP) condition – to explain the dynamics of exchange rates. When expected equity returns in a country/region are lower than expected equity returns in another country/region, the currency associated with the market offering lower returns is expected to appreciate. First, we test the URP assuming that investors are risk neutral and next we relax this hypothesis. The resulting risk premia are proxied by economic variables, which are related to the business cycle. We employ differentials in corporate earnings’ growth rates, short-term interest rate changes, annual inflation rates, and net equity flows. The URP explains a large fraction of the variability of some European currencies vis-à-vis the US dollar. When confronted with the naïve random walk model, the URP for the EUR/USD performs better in terms of forecasts for a set of alternative statistics.
    Keywords: Foreign exchange markets; asset pricing; random walk; UIP; GMM.
    JEL: F31 G15 C22 C53
    Date: 2005–09
  6. By: Lars Svensson; Noah Williams
    Abstract: We examine optimal and other monetary policies in a linear-quadratic setup with a relatively general form of model uncertainty, so-called Markov jump-linear-quadratic systems extended to include forward-looking variables. The form of model uncertainty our framework encompasses includes: simple i.i.d. model deviations; serially correlated model deviations; estimable regime-switching models; more complex structural uncertainty about very different models, for instance, backward- and forward-looking models; time-varying central-bank judgment about the state of model uncertainty; and so forth. We provide an algorithm for finding the optimal policy as well as solutions for arbitrary policy functions. This allows us to compute and plot consistent distribution forecasts---fan charts---of target variables and instruments. Our methods hence extend certainty equivalence and "mean forecast targeting" to more general certainty non-equivalence and "distribution forecast targeting."
    JEL: E42 E52 E58
    Date: 2005–11
  7. By: Martin D. D. Evans (Georgetown University) (Department of Economics, Georgetown University)
    Abstract: This paper provides an overview of the recent literature on Foreign Exchange Market Microstructure. Its aim is not to survey the literature, but rather to provide an introductory tour to the main theoretical ideas and empirical results. The central theoretical idea is that trading is an integral part of the process through which information relevant to the pricing of foreign currency becomes embedded in spot rates. Micro-based models study this information aggregation process and produce a rich set of empirical predictions that find strong support in the data. In particular, micro-based models can account for a large proportion of the daily variation in spot rates. They also supply a rationale for the apparent disconnect between spot rates and fundamentals. In terms of forecasting, micro-based models provide out-of-sample forecasting power for spot rates that is an order of magnitude above that usually found in exchange-rate models. Classification-JEL Codes: F3, F4, G1
    Keywords: Exchange Rates, Microstructure, Information Aggregation, FX Trading.
  8. By: Matteo Ciccarelli (Corresponding author: European Central Bank, DG Research, Kaiserstrasse 29, 60311 Frankfurt am Main, Germany); Benoît Mojon (Université de la Méditerranée and European Central Bank, DG Research, Kaiserstrasse 29, 60311 Frankfurt am Main, Germany)
    Abstract: This paper shows that inflation in industrialized countries is largely a global phenomenon. First, inflations of (22) OECD countries have a common factor that alone account for nearly 70% of their variance. This large variance share that is associated to Global Inflation is not only due to the trend components of inflation (up from 1960 to 1980 and down thereafter) but also to fluctuations at business cycle frequencies. Second, Global Inflation is, consistently with standard models of inflation, a function of real developments at short horizons and monetary developments at longer horizons. Third, there is a very robust "error correction mechanism" that brings national inflation rates back to Global Inflation. This model consistently beats the previous benchmarks used to forecast inflation 1 to 8 quarters ahead across samples and countries.
    Keywords: Inflation; common factor; international business cycle; OECD countries.
    JEL: E31 E37 F42
    Date: 2005–10
  9. By: Martin D. D. Evans; Richard K. Lyons
    Abstract: This paper develops a model for understanding end-user order flow in the FX market. The model addresses several puzzling findings. First, the estimated price-impact of flow from different end-user segments is, dollar-for-dollar, quite different. Second, order flow from segments traditionally thought to be liquidity-motivated actually has power to forecast exchange rates. Third, about one third of order flow's power to forecast exchange rates one month ahead comes from flow's ability to forecast future flow, whereas the remaining two-thirds applies to price components unrelated to future flow. We show that all of these features arise naturally from end-user heterogeneity, in a setting where order flow provides timely information to market-makers about the state of the macroeconomy.
    JEL: F3 F4 G1
    Date: 2005–11

This nep-for issue is ©2005 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.