Patrice Baubeau () (Agence Nationale de la Recherche—CoDiSyNa, Université Paris Ouest Nanterre La Défense—IDHE, CNRS UMR 8533, 200, Avenue de la République, 92001 Nanterre Cedex, France) Bernard Cazelles (CNRS UMR 7625, Ecole Normale Supérieure, 46 rue d’Ulm, 75230 Paris, France and IRD UR GEODES, 93143 Bondy, France)
Abstract
Although 50 years of scientific work has been invested in building retrospective economic time series, their reliability is still debated, a good example being the two competing nineteenth century French GNP series. Instead of trying to bring up some new details to gauge their respective accuracy, we propose a different route, i.e. testing the intrinsic features of these two series, in absolute terms first, then by benchmarking them to a non-retrospective time series. In order to do that, we rely on new mathematical tools—wavelet spectrum analysis—developed in signal processing. This leads to a new approach, which separates the accuracy of a series between amplitude and time variations, and brings nuanced conclusions as to which of the two series tested is the best: indeed, since a trade-off is almost inescapable between the two criterions of accuracy, the statistical quality of one retrospective time series tends to linger either on one side (amplitude level) or the other (time variations). Our study also shows that variance distribution along the time axis is a good proxy for complex retrospective series accuracy.
Download Info
To download:
If you experience problems downloading a file, check if you have the
proper application to
view it first. Information about this may be contained
in the File-Format links below. In case of further problems read
the IDEAS help
page. Note that these files are not on the IDEAS
site. Please be patient as the files may be large.
As the access to this document is restricted, you may want to look for a different version under "Related research" (further below) or search for a different version of it.
Volume (Year): 3 (2009) Issue (Month): 3 (October) Pages: 275-300 Download reference. The following formats are available: HTML
(with abstract),
plain text
(with abstract),
BibTeX,
RIS (EndNote, RefMan, ProCite),
ReDIF