Combining Disaggregate Forecasts or Combining Disaggregate Information to Forecast an Aggregate

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in WO R K I N G PA PE R S E R I E S N O 1155 / F E B R UA RY 2010This paper can be downloaded without charge from http://www.ecb.europa.eu or from the Social Science Research Network electronic library at http://ssrn.com/abstract_id=1551223In 2010 all ECB publications feature a motif taken from the €500 banknote. Abstract 4 Non-technical summary 5 COMBINING DISAGGREGATE FORECASTS OR COMBINING

Forecasting aggregates and disaggregates with common features

“…Recently, Hendry & Hubrich (2006, 2010, hereafter HH, proposed a procedure for forecasting an aggregate by using a model for that aggregate that includes as regressors its own lags as well as lags of the components. They use autometrics (see Doornik, 2009), and follow the general-to-specific approach to build the model.…”

The Role of Indicator Selection in Nowcasting Euro Area GDP in Pseudo Real Time

“…In forecasting practice, the availability of a very large number of timely monthly indicators (the Appendix lists the 259 indicators used in this study) leads to a curse of dimensionality which prevents the direct estimation of α and j β parameters, unless just a few (k) of these are pre-selected with experience and "art" by the researcher, as implied in the BM large datasets could improve the performance of aggregate GDP models by exploiting factors that embody disaggregate information (see Hendry and Hubrich, 2011). approach. Although unavoidably arbitrary, this extraction of N k << indicators has proved to be quite effective in forecasting euro area GDP in the short-run with BM, see e.g.…”

“…On the basis of the alternatives listed in Hendry and Hubrich (2011), the present paper will consider three ways to tackle the aggregation issue in forecasting euro area GDP: (I) aggregate models using aggregate information for direct aggregate GDP forecasts; (II) component models using disaggregate information for direct disaggregate forecasts (for e.g. consumption and investment) and their aggregation in indirect GDP forecasts; (III) aggregate models using disaggregate information for direct aggregate GDP forecasts.…”

“…Factors extracted through FM combine disaggregate information about the target components that are likely to be included in the indicator data-set and bring this disaggregate information into the aggregate equation (2) through predictors measured by such factors (see Hendry and Hubrich (2011)). For this reason, it is not useful to make disaggregate FM for GDP components given that, as seen above, factor extraction does not explicitly account for the model's target variables.…”

“…Thirdly, we assess the effectiveness of forecasting GDP from the supply and from the demand side, i.e. from a (possibly pre-screened) factor model approach which combines indicator information versus another in which forecasts of GDP expenditure items are combined, in line with the work of Huang and Lee (2010), and Hendry and Hubrich (2011). This paper is set out as follows.…”