Premium
The data measurement process for UK GNP: stochastic trends, long memory, and unit roots
Author(s) -
Patterson Kerry
Publication year - 2002
Publication title -
journal of forecasting
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.543
H-Index - 59
eISSN - 1099-131X
pISSN - 0277-6693
DOI - 10.1002/for.828
Subject(s) - unit root , vintage , econometrics , process (computing) , computer science , series (stratigraphy) , multivariate statistics , long memory , variable (mathematics) , data series , unit (ring theory) , component (thermodynamics) , economics , mathematics , machine learning , mathematics education , volatility (finance) , paleontology , mathematical analysis , physics , thermodynamics , archaeology , biology , history , operating system
Much published data is subject to a process of revision due, for example, to additional source data, which generates multiple vintages of data on the same generic variable, a process termed the data measurement process or DMP. This article is concerned with several interrelated aspects of the DMP for UK Gross National Product. Relevant questions include the following. Is the DMP well behaved in the sense of providing a single stochastic trend in the vector time series of vintages? Is one of the vintages of data, for example the ‘final’, the sole vintage generating the long‐memory component? Does the multivariate framework proposed here add to the debate on the existence of a unit root in GNP? The likely implicit assumptions of users (that the DMP is well behaved and the final vintage is ‘best’) can be cast in terms of testable hypotheses; and we show that these ‘standard’ assumptions have not always been empirically founded. Copyright © 2002 John Wiley & Sons, Ltd.