Several methodologies are used in Main Economic Indicators (MEI) to link time series. The preferred method, and the one most widely used, is the 12-month or 4-quarter link method. This method can be applied when an old and new series (which are being linked) contain data for an overlapping time period of at least one year.
The 12-month (4-quarter) link method calculates a linking factor as the average of the first common year of observations of the new series divided by the equivalent observations of the old series. This method is robust, in particular when revisions have been made with the introduction of the new series. Also, this method ensures as best as possible that consistency is maintained between raw and seasonally adjusted versions of the same series. An exception to this procedure is made in the case of Price Indices, where use of the first common period link method is preferred. This method calculates a linking factor as the ratio of the first observation of the new series to the equivalent observation of the old series in the overlapping period. This method ensures the same monthly growth rate at the link point for the new and old Price Indices and minimises the impact on the long time series if major changes are made to the coverage of the new Price Index.
Use of the first common period link method can cause a spurious relationship between the raw and seasonally adjusted versions of the new series prior to the link point (i.e. after the link has been made), which is the primary reason why the 12-month link method is preferred for most time series where both original and seasonally adjusted series are maintained. However this is not an issue for Price Indices as these are generally not seasonally adjusted. Also, where the series being linked has only an approximate relationship with the current series, it is more appropriate to use the first common period method (e.g. this is often the case for the linking requirements of the OECD system of unit labour cost indicators published in the MEI). Where a full common year between old and new series provided by a country does not exist, the first common period link method will also be used. Also, if there has been no revision to the new series over the first common year, these two linking methods yield the same result. The choice of preferred linking method was decided after detailed methodological and empirical study by the OECD Secretariat.
The implicit link method is used to link series which share no explicit overlapping period, but for which a common reference period can be established by inference or deduction. For example, the method could use the information that for the two series, a common base year value is 100. The two series can then be re-referenced to this common base year and then concatenated (i.e. joined). Note that the information set which necessitates using this link method is rarely found in practice so its use is infrequent.
Some series expressed in levels (such as monetary series measured in currency and series covering western and later unified Germany) are joined by concatenation. This means that values of the new series replace or are added to the old series from the first observation of the new series. As a result of its construction, this method may produce one or more significant ruptures (i.e. series breaks) in the linked series at the link point(s). However, such ruptures are documented in the publications and the database.
Main Economic Indicators investigates each series on a case-by-case basis to determine whether they should be linked or not. If they should, then the most appropriate method is determined by in-depth analysis. Series with little in common methodologically, or those that exhibit significant differences in their evolution are not linked. Furthermore, different methods may be used for different links within the same series, depending on the nature of the old and new data series being linked at the particular point in time. In general it is the policy of the MEI to try and maintain long time series, and therefore our preference is to link series wherever possible and provide appropriate metadata to users.
Linking factors for gross and seasonally adjusted series