Did the OECD Composite Leading Indicators see it coming?
Roberto Astolfi, OECD Statistics Directorate
To some Professor Luis Garicano of the London School of Economics is a leading expert in the fields of productivity and industrial organisation, but to many he’s the man Queen Elizabeth asked “Why did no one see it coming?”; “it” being the crisis. In retelling the story, Pr. Garicano pointed out that he welcomed the question as it provided an opportunity to cite many that did see it coming, including Messrs, Krugman and Volcker. Was the OECD among them?
At the OECD, we use a number of techniques to determine what the data are telling us is happening now and what might happen in the future. Dave Turner from the OECD Economics Department describes some of those approaches here. One additional technique used by the OECD Statistics Directorate, is the system of Composite Leading Indicators (CLIs). Simply put, the CLIs combine individual indicators for a given country to anticipate when economic expansion starts entering a downturn, or when growth starts to return. A relevant question in this context therefore is:
How useful were the OECD CLIs just before the crisis?
Perhaps the simplest way to answer the question is by reference to the headline messages announced in each of the monthly CLI Press Releases over the crisis period.
The first indication of potential trouble ahead came in September 2007 (Figure 1) where the headline assessment moved from ‘Mixed outlook’ to “Moderating outlook”. With each successive bulletin, the announcements became more pessimistic. “Weakening outlook” in the months that followed gave way to “Downswing” in January 2008, and even February 2009’s “Lowest level since 70s” was followed by “New low” in March 2009.
From today’s estimates we know that the CLI for the OECD area as a whole reached its pre-crisis high in June 2007, six months before the actual GDP peak that we now locate in December 2007 (vertical solid red line and black dotted line, respectively, in Figure 1).
Using the most recent statistical information, (in other words, including any revisions that may have been made in the interim) Gyomai and Guidetti in 2011 concluded that the “CLI was able to anticipate the downturn in the real economy at least 5 months ahead of its initial materialisation” (detailed results are available in the Statistics Newsletter).
A more stringent approach is to review the performance of the CLI at the time of the crisis using only the statistical information that was then available, as we do in the recently published Statistics Working Paper.
Figure 1: Evolution of CLI Press Release headlines during the Great Recession, OECD area
Note: The vertical lines identify the turning points detected by the CLIs for the OECD area as a whole (peak in June 2007 and trough in February 2009, marked in red) and GDP (marked in dotted black, with a peak in December 2007 and a trough in May 2009).
This approach is more ‘severe’ as formal identification of the turning points can only ever be confirmed some time after they manifest. Nevertheless, even with this more severe examination the latest results confirm the leading properties of the CLI while also indicating that the statistical and methodological revisions that have occurred since the crisis have not shifted CLI turning points to earlier dates, nor have they artificially improved the CLI performance.
Overall then, the OECD CLIs proved to be a robust tool in anticipating the crisis some months before GDP reached its pre-crisis high watermark, and so, perhaps they can be added to the list of illustrious names that can be quoted the next time somebody asks ‘why did no-one see it coming’. Moreover, although, by their very nature and design, CLIs are not able to quantify the magnitude of slowdowns or upturns, and, so, could not quantify the severity of the crisis, the increasing downbeat tone of assessments that followed the first warning in September 2007 provided strong pointers.