Unit Labour Costs (ULC) are calculated as the ratio of total labour costs to real output. In the OECD System of Unit Labour Cost Indicators the target variable for the quarterly indicator of total labour costs is compensation of employees compiled according to the System of National Accounts 1993 (SNA93). Where this target variable is not available a suitable proxy is sought, with the following general order of preference: gross wages and salaries; labour cost index multiplied by total hours worked; average earnings multiplied by total employment.

Where the use of proxy variables is required, an indicator of their quality can be determined by evaluating the correlation between annual growth rates in the proxy indicator and that of the SNA93 benchmark data. The following table (as collated on 16 March 2007) presents these correlations and also describes the type of proxy variable used.

The correlations give an indication of the reliability of the total labour cost proxy variable used in the compilation of the ULC for the relevant country and economic activity. A low correlation implies that users should take caution when viewing the resultant quarterly ULC prior to reconciliation with annual benchmark data (which generally occurs with a 12-18 month lag). However the following processes used in the OECD System of Unit Labour cost Indicators also need to taken into account when understanding how these proxy variables are treated:

• Benchmarking: As part of the compilation, all quarterly variables are benchmarked (temporally disaggregated) against annual SNA93 values; and for all countries SNA93 data is available. This benchmarking procedure, using the Fernandez method in the software package Ecotrim, will in simple terms ensure that the proxy quarterly total labour cost value has a correlation of 1 with the annual value for the overlapping period. As part of the Fernandez method also, the quarterly end points where there is no supporting annual data will in essence be smoothed and trend towards the long-term series average.
• Seasonal adjustment: The final quarterly ULC are seasonally adjusted using the TRAMO-SEATS method in the software package Demetra. In addition to the seasonally adjusted series, TRAMO-SEATS produces a trend-cycle series that includes all non-seasonal and non-irregular movements in the underlying time series. This series can be regarded as a smoothed seasonally adjusted series, where the degree of smoothing is dependent on the underlying ARIMA model.

The OECD has performed extensive testing of both the Fernandez and TRAMO-SEATS methodologies. These methodologies together ensure that the quarterly series is in line with the countries authoritative annual SNA93 data and the quarterly volatility is reduced but at the same time effectively extracting the underlying signal from the raw data.

Notes:
For Japan, Korea, and the United States, the proxy is the multiplicative sum of: average earnings; and, number of employees.
For Belgium, Portugal and Romania - data is sourced via Eurostat.

Footnotes:
1. Compilation index for Manufacturing is the multiplicative sum of: Average earnings; average weekly hours of work; and, number of employees.
2 Compilation index for Market Services and the Business Sector is the multiplicative sum of: Labour cost index and Number of employees.
3. Compilation index is the product of: Labour cost index for the activity and number of employees for the activity.
4. Market Services for the US is a residual calculation compiled as: the Business Sector minus Industry minus Construction. The calculation (and the correlation) takes place after the benchmarking procedure. For more metadata see: USA Unit Labour Cost Metadata