This article outlines a methodology for producing quarterly estimates of unit labour costs below the whole economy level, consistent with other Office for National Statistics (ONS) data on labour inputs and the returns to labour. Subject to user feedback, ONS plans to include some of these series in its quarterly Labour Productivity statistical release, and to replace the current series for manufacturing unit wage costs with a new series for manufacturing unit labour costs. This article also proposes some changes to the methodology for producing the whole economy measures, of which the most significant is a change in the method of seasonal adjustment. The proposed changes have significant effects on the historic profiles for the whole economy and (especially) for manufacturing compared with currently published estimates. New series for the market sector and for the production and services sectors provide additional insight into the performance of the economy. This work also feeds into ONS’s future work plan in the productivity field.
The authors are grateful to John Appleton and Jean Acheson, both formerly of ONS, for preparatory work for this article, and to members of the ONS National Accounts Methods Committee for helpful comments on an earlier version.
Unit labour costs (ULCs) measure the cost of labour input per unit of economic output and are useful indicators of inflationary pressures and competitiveness. Movements in ULCs can be decomposed into movements in costs per unit of labour (which can be approximated by an index of earnings) minus the movement in labour productivity. Thus other things equal, faster growth of earnings will tend to push up ULCs, and increases in labour productivity will tend to reduce ULCs.
Office for National Statistics (ONS) currently publishes estimates of ULCs for the whole economy and estimates of unit wage costs (UWCs, a narrower measure, excluding non-wage labour costs such as employers’ social security and pension contributions) for the whole economy and for the manufacturing sector. These estimates are published quarterly in the Labour Productivity statistical release.
Previous articles by Turvey (2009) and Appleton (2011) introduced a revised methodology for whole economy ULCs and UWCs which better reflect the labour costs attributable to the self-employed. This replaced the previous approach in which the wage and labour costs of the self-employed were assumed to be equal to those of employees, despite evidence from the National Accounts that such an assumption is inconsistent with the income attributed to the self-employed.
However, manufacturing UWCs continue to be based on the same assumption. Moreover, no estimates are currently available for manufacturing ULCs, which are preferable to UWCs in capturing movements in non-wage labour costs. Movements in non-wage labour costs can be significant. For example, there was an exceptionally high level of employers’ pension contributions in the first quarter of 2012.
In this article we extend the methodology set out in Appleton (2011) to manufacturing and other sections of the economy and to the market sector. Subject to feedback from users and further quality assurance work, it is our intention to replace the current series for manufacturing UWCs (DIX4) with a new estimate for manufacturing ULCs and to publish new estimates of ULCs for the production sector, total services and the market sector.
Measures of labour costs (or wage costs) used in the production of ULCs (or UWCs) differ from ONS indices of labour cost per hour (ILCH, see Index of Labour Costs per Hour) by including the labour costs of the self-employed, who comprise around 14 per cent of total labour input according to latest figures from the Labour Force Survey1.
In the National Accounts, compensation of employees (COE) measures the aggregate labour costs of employees, and mixed income measures the income of the self-employed. Conceptually, mixed income can be thought of as a combination of returns to labour and returns to capital. Explicit information on this factor income split is not available.
At the whole economy level we assume that the split in each quarter reflects the nearest analogous observable split. We obtain this by taking the share of COE in a synthetic measure of income measured as the sum of market sector COE and profits (measured in the National Accounts as gross operating surplus – GOS) of incorporated businesses2.
The reasoning behind this approach is that factor income shares for the self-employed are more likely to reflect the share in the rest of the market sector than in the non-market sector, where for example the National Accounts treatment of GOS does not include a return on capital. Equivalently, this methodology assumes that the “production function” of the self-employed resembles that of the (rest of the) market sector.
Denoting the labour share of mixed income as α, then in each time period t
And market sector ULCs can be written as:
Figure 1 shows that growth of ULCs for the market sector broadly tracks growth for the whole economy series. However, there is some evidence since the start of the recession in 2008 that peaks and troughs in market sector ULC growth have been rather more pronounced.
As noted above, ONS currently publishes manufacturing UWCs but these estimates are inconsistent methodologically with the published whole economy measures of UWCs and ULCs. Moreover, the current methodology for manufacturing UWCs is known to be inconsistent with National Accounts estimates of earnings of the self-employed in manufacturing, and UWCs are a less inclusive measure than ULCs.
In order to produce consistent estimates of manufacturing ULCs, and other categories below the whole economy level, we follow the same approach as for the whole economy and market sector, namely to use section level COE data and an apportionment of mixed income to reflect sectional labour shares of income.
ONS currently produces an industry breakdown of COE for Eurostat. However, the manufacturing component of this breakdown is treated as the residual in benchmarking to HMRC administrative data. The resulting series displays significantly different properties from that obtained by using ONS data on manufacturing employment, earnings and non-wage labour costs.
For this article, we have constructed synthetic COE series for manufacturing and other industries which are (a) consistent with section-level employment and earnings data and (b) benchmarked to whole economy COE estimates. We expect the Eurostat series to reflect these methodological improvements, possibly from Blue Book 2013.
To allocate mixed income between industries we use a combination of data from supply-use tables (SUTs) and Labour Force Survey (LFS) data on the distribution and characteristics of self-employment, benchmarked to quarterly National Accounts data on aggregate mixed income. Annual data from the SUTs are profiled across quarters and extrapolated beyond 2010, the last year for which SUT data are currently available, using movements in self-employment from LFS.
Labour shares of mixed income are allowed to vary across sections, using sectional information on COE and GOS and applying sectional equivalents of Equation 1 above. Sectional COEs are the synthetic series described above. Sectional information on GOS is derived in a similar fashion to mixed income, except that the quarterly profiling uses COE profiles rather than LFS. This is because profits attributable to the self-employed represent only a small fraction of overall GOS. As with mixed income, the overall level of GOS is benchmarked to the quarterly National Accounts aggregate.
Annual data for some of the resulting labour shares of mixed income are shown in Figure 2. At the whole economy level, the labour share rose sharply between 1997 and 2001, fell back somewhat in the ‘noughties’ and has increased a little since 2008. The lower profile for the labour share in the production sector is consistent with greater capital intensity in this sector. Greater volatility of the labour share in production is consistent with greater cyclicality of production relative to service sector output.
One further methodological issue is that we are proposing to move to ‘end of pipe’ seasonal adjustment. This entails using non-seasonally adjusted component data (COE, mixed income, GOS, gross value added [GVA] and employment subsidies) and seasonally adjusting the resulting derived ULC statistics, rather than the current approach which uses seasonally adjusted component data to compute ULCs and UWCs.
The end of pipe approach is methodologically preferable because it focuses on the seasonality of the derived statistic in question (ULCs in this case) and avoids any risk that the process of separately seasonally adjusting the numerators and denominators of the ULC expression will not properly reflect the seasonal characteristics of the derived statistic.
Figure 3 illustrates the effect of this change in method of seasonal adjustment by comparing the currently published whole economy ULCs series LNNL with the new series using unadjusted data for COE and other source data and carrying out seasonal adjustment at the end of the process. As noted above, the new series incorporates a change in the source data but this makes almost no difference to the series, either in levels or growth rates.
It follows that substantially all of the differences observed between the two series in Figure 3 arises from the change in seasonal adjustment. The new series shows a more pronounced upturn in growth of ULCs in 2004, and also advances the trough in the series growth rate following the recovery from the 2008-09 recession by a quarter, from 2011Q1 to 2010Q4. According to the new series, ULCs grew by 4.4 per cent in the year to 2012Q2, compared with the published series growth rate of 3.2 per cent3.
We have developed a system to generate section-level ULC estimates for 11 sections or aggregates of sections, as well as for the whole of the production sector (sections B-E) and total services (sections G-U), as shown in Table 1. This aggregation structure is designed to be consistent with other ONS productivity systems including those for quality-adjusted labour indices and multi-factor productivity.
|A||Agriculture, Forestry and Fishing|
|B,D,E||Mining and Quarrying, Utilities|
|G,H,I||Wholesale and retail trade, Transportation and Storage, Accomodation and Food Services|
|J||Information and Communication|
|K||Financial and Insurance Industries|
|M,N||Professional, Scientific and Technical Activites, Administrative and Support Service Activities|
|O,P,Q||Public Administration and Defense, Education, Health and Social Work|
|R,S,T,U||Arts and Entertainment, Other services|
|A-U||Whole Economy (WE)|
All of the results are available in a reference table (53 Kb Excel sheet) to this article, but it is not our intention to publish ULCs for all of these categories on a routine basis. In some cases ONS does not currently publish labour productivity statistics at this breakdown (e.g. construction) and we only publish annual labour productivity for agriculture, forestry and fishing.
Subject to user feedback, we are currently planning to publish new series for manufacturing, production and total services. We welcome feedback from users on this proposal, and also on the question of whether the new series for manufacturing ULCs should replace the existing series for manufacturing UWCs.
A further difference between the two measures is that we make an adjustment for employment subsidies, although the level of such subsidies is currently very small.
This articulation of the market sector in terms of factor incomes entails a very small methodological change to the whole economy ULC and UWC methodology described in Turvey (2009) and Appleton (2011) and used in the production of series LNNL and LNNK. This change makes no discernible difference to either of these series or their growth rates.
Changes arising from ‘end of pipe’ seasonal adjustment will affect the revisions analysis published in the Labour Productivity release, which includes whole economy ULCs and UWCs. We welcome comments from users on this issue.
Figure 4 plots manufacturing ULCs derived using the methodology set out above with the current series for manufacturing UWCs (DIX4). Differences between these two series reflect a combination of (i) more consistent methods for accounting for the labour income of the self-employed in the manufacturing sector, (ii) inclusion of non-wage labour costs in the numerator, and (iii) ‘end of pipe’ seasonal adjustment.
Although the broad picture is similar between the two series, there are material differences in certain time periods. One example is between 2004Q1 and 2005Q1 when the new series shows ULC growth increasing from -4.2 per cent to 2.9 per cent, an increase of 7.1 percentage points, compared with an increase of 3.4 percentage points on the currently published UWC measure. More generally, there is some evidence that cyclical peaks and troughs are more pronounced on the new measure than the existing measure.
Another perspective on the two measures is shown in Figure 5 which plots the quarterly percentage movements in the two series since 2008Q1. Again there is some suggestion that the new measure is more cyclical than the existing measure. And there is a very distinct shift in the timing of the cyclical peak around the trough of the recession, 2008Q4-2009Q1.
ONS does not currently publish ULCs or UWCs for the production sector, so Figure 6 compares the new series with the whole economy series, and Figure 7 compares production with manufacturing. In both cases, the comparison is with the new series described in this article rather than the currently published series.
Unsurprisingly, production ULCs are more volatile than those for the whole economy, and for most of the period, growth of ULCs in the production sector has been lower than in the whole economy. However, since 2008, ULCs in the production sector have generally grown faster than across the whole economy.
Manufacturing accounts for around two-thirds of the production sector in terms of GVA, so it is not surprising that the two series in Figure 7 are clearly correlated. What is perhaps more surprising are the divergences in the series in 2006 and 2011. On both of these occasions, there have been significantly different movements in output and labour costs between manufacturing on the one hand and the rest of the production sector on the other hand1.
As noted in a recent ONS article on the “Productivity Conundrum” this reflects to some extent the particular characteristics of the oil and gas sector. Similar differences are also observed in the published labour productivity statistics for manufacturing and the production sector. For the non manufacturing production sector as a whole (which also includes utilities and waste management) ULCs increased by around 30 percentage points between 2009 and 2011, compared with an increase of just 2 percentage points for the whole economy, and a decrease of more than 3 percentage points in manufacturing.
The service sector accounts for 77 per cent of the economy in terms of GVA, so it is not surprising that ULCs display a similar trend to that of the whole economy. One interesting feature of Figure 8 is that services ULCs were more cyclical than those of the whole economy between 2005 and 2008, but have been significantly less volatile thereafter.
ULCs for seven components of the service sector are published in the Reference Table (53 Kb Excel sheet) component of this release. There have been wide divergences in ULC movements between some of these components. For example, between 2009 and 2011 ULCs increased by some 22 percentage points in section K (Financial and Insurance activities) but have fallen by around 8 percentage points in section L (Real estate activities).
ULCs for the aggregate of sections O. P and Q (which is a rough proxy for the non-market sector) have increased a little less than for the service sector as a whole, consistent with the whole economy and market sector comparison seen in Figure 1.
See the Reference Table (53 Kb Excel sheet) published with this article for more information.
Subject to user feedback, ONS plans to take forward the new estimates described in this article with a view to regular publication in the Labour Productivity release. This will require some additional quality assurance work and other internal processes, so we would not expect to include these new series in the next edition of the release, scheduled for 3 January 2013.
Two further potential spin-offs from the work reported in this article are, firstly, development of a methodology for generating section level income constraints consistent with the Quarterly National Accounts is expected to help in developing a quarterly framework for analysing multi-factor productivity.
This would provide more timely information on the sources of productivity movements. Secondly, this methodology also takes us some way towards development of a set of unit profit indices. Movements in such series can be expected to complement movements in ULCs in highlighting inflationary pressures and movements in profit ratios between different industries.
ONS intends to continue development work and to report progress at a meeting of the Productivity Statistics User Group in early 2013. For further information on this meeting, please email email@example.com.
Statistical contact: Priya Mistry
Tel: +44 (0)1633 455047
We welcome any feedback, please contact us at Productivity@ONS.gsi.gov.uk
Details of the policy governing the release of new data are available by visiting www.statisticsauthority.gov.uk/assessment/code-of-practice/index.html or from the Media Relations Office email: firstname.lastname@example.org
Turvey, A. (2009) ‘Developing a unit labour cost indicator for the UK’, Economic and Labour Market Review, Vol 3, No 6, June 2009, pp 51-56.