Quality-adjusted labour input (QALI) is a method of measuring labour input to economic production which takes account of the composition (or "quality") of the workforce as well as the volume of hours worked. This article presents new QALI estimates for 2011 and 2012 and revisions to previously published QALI estimates. Inclusion of estimates for 2012 represents a significant shortening of the publication timetable compared with earlier QALI publications and reflects development of new estimation methods for certain components of the QALI system. In principle, QALI provides a more complete picture of the input of labour into the production process than traditional measures, and therefore provides a broader perspective in assessing productivity performance. Hours worked (and other quantity metrics such as employment and jobs) have been much stronger since the recession in 2008-09 than in comparable earlier periods, so conventional measures of labour productivity have been uncharacteristically weak, a phenomenon that has come to be referred to as the "productivity puzzle". QALI could offer some explanation of the puzzle if falls in labour quality were to have offset some of the movement in hours. However, we find that labour quality has improved in both 2011 and 2012, therefore exacerbating, rather than alleviating, the productivity puzzle.
The authors thank John Allen of ONS for research support in preparing this article.
In simple terms, QALI is a weighting system which weights different categories of labour input (measured by hours worked) according to their relative contribution to economic production. The categories in the weighting structure are based on identifiable characteristics of the labour force (education, age, gender and industry of employment) and the weights are relative hourly earnings of each category. Under competition, economic theory suggests that relative earnings should reflect the marginal productivity of different categories of labour, and intuitively it is reasonable to expect that, on average, more highly paid workers contribute more to economic production than lower paid workers. Using a suitable weighting system, it is possible to subtract movements in hours (sometimes referred to as "unadjusted hours") from QALI indices to identify the pure quality component of labour input. This is also known as "labour composition" but will be referred to as labour quality throughout this article.
The main use of QALI is as a component in a growth accounting framework, where growth of economic output can be decomposed into contributions from labour (split into volume and quality sub-components), contributions from capital services (which measure the capital input to production in an analogous method to QALI, by weighting different types of capital according to their "user costs"), and a residual component ("multi-factor productivity" or MFP in ONS terminology) which reflects the movement in economic output which cannot be accounted for by either labour or capital inputs. MFP estimates to 2010 were published by ONS in September 2012 (Appleton & Franklin, 2012). ONS will publish MFP estimates to 2012 based on the QALI estimates in this article and revised and updated measures of capital services later in 20131.
Whole economy QALI increased by 1.4% in 2011 and by 3.2% in 2012 (see Figure 1). At the whole economy level, hours worked grew by 0.4% in 2011 and by 2.0% in 2012, so the implied contribution of labour quality, in growth rate terms, was 1.0% in 2011 and 1.2% in 20122. It is important to appreciate that, from an economic perspective, it is the movement in QALI rather than the movement in hours which should offer the better representation of what is happening to labour input to economic production. For example, an increase in quality of 1% with hours unchanged should be seen as exactly equivalent (abstracting from distributional effects) to an increase in hours of 1% with quality unchanged.
The layout of the rest of this article is as follows. The following section provides guidance on interpreting QALI, including methodological developments since the last QALI article was published in 2012 and an outline of planned future development work. The next section provides an overview of QALI at the whole economy (and market sector) level over the 2008-09 recession and its aftermath, exploiting the shorter lag between publication as a result of methods changes noted above. This section also looks at how movements in QALI since the 2009-09 recession compare with movements after earlier recessions. The next section examines QALI conditioned on each of the four identified characteristics (that is, by education, age, gender and industry of employment). Lastly we examine revisions since the 2012 QALI article (Acheson & Franklin, 2012).
Throughout this article the contribution of labour quality is generally reported as the difference (measured in index points) between the QALI series and a corresponding index of hours worked. An increase in the index point contribution implies that quality is adding to the growth of a QALI index and vice versa.
Movements in labour quality reflect changes in the distribution of labour input among categories and differentials in hourly remuneration between those categories. Table 1 provides a snapshot of the latter, summarising relative hourly remuneration for each QALI category in 2008, 2010 and 2012. Within QALI, labour input is categorised by industry (10 sub-categories), education (6 sub-categories), gender (2 sub-categories), and age (3 sub-categories), giving a total of 360 QALI sub-categories (10*6*3*2). Table 1 collapses the underlying 360-cell data separately by each sub-category. Not surprisingly, some QALI categories, like working in financial and insurance activities or having a higher degree, provide more return to labour input than others, on average.
|Whole economy = 100||2008||2010||2012|
|Agriculture, forestry and fishing; mining and quarrying; utilities (ABDE)||80.4||71.4||71.7|
|Wholesale and retail trade; accommodation and food services (GI)||86.1||87.3||87.8|
|Transportation and storage (H)||79.7||78.3||78.9|
|Information and communication (J)||141.2||136.6||147.3|
|Financial and insurance activities (K)||160.2||182.1||182.2|
|Real estate activities; professional and scientific activities; administrative and support activities (LMN)||86.0||86.2||88.4|
|Public administration and defence; education; health and social work (OPQ)||118.3||114.4||110.9|
|Arts and entertainment; other services (RSTU)||85.3||87.8||96.4|
|Masters and doctorates||165.0||153.6||148.3|
|First and other degrees||132.5||129.9||127.3|
|Certificates of education or equivalent||109.5||107.3||105.7|
|A – levels or trade apprenticeships||90.7||90.1||88.7|
|GCSEs or equivalent||79.2||78.6||77.5|
Based on LFS self-reported hourly earnings of employees, scaled up for self-employed earnings and benchmarked to macro estimates. See Acheson & Franklin (2012) and Franklin & Mistry (2012) for more information.
Looking at relative remuneration over 2008 to 2012, the distribution by gender and age is broadly stable. The pattern of relative remuneration increasing between the 16-29 and 30-49 age cohorts is usually interpreted as a return to experience and development of on-the-job skills, while one explanation for the fall in relative earnings between the 30-49 and 50+ age groups is that better paid workers can afford to retire earlier. The gender earnings difference is also usually interpreted as a reflection of earnings being correlated with experience and seniority: females being subject to more intervals out of the labour market, on average, due to child rearing. As noted in earlier QALI articles, there is absolutely no suggestion of an underlying gender difference in labour quality.
There is some evidence that relative remuneration across industry of employment has become more unequal over the last 5 years (compare industries K and F for example). By contrast, average returns to higher levels of education seem to have become less pronounced in recent years.
Although Table 1 shows the importance of accounting for different compositional differences when calculating quality-adjusted labour inputs, it does not necessarily foreshadow the QALI results, which also depend on movements in the shares of hours worked by the different QALI categories. For example, an increase in the share of hours worked by those with higher educational qualifications, or from females to males, will tend to increase labour quality as long as these QALI categories enjoy relatively higher remuneration. And of course, volume and relative price movements are not independent of one another. In particular, movements in the shares of hours worked by QALI categories that reflect supply factors (such as an increase in the share of the workforce with higher levels of education due to an expansion of higher education) are likely to be a factor contributing to changes in relative remuneration.
The basic QALI methodology is set out in previous QALI annual articles (most recently in Acheson & Franklin, 2012) and is not repeated here as there have been no significant changes to this methodology and the basic parameterisation of QALI in terms of categories of labour (as in Table 1) is also unchanged.
However, since the last paper, there have been several detailed methodological changes to develop QALI. The main changes are:
extension to more recent quarters, using the same methodology as used in compiling sectional unit labour costs. The main data source for QALI is the Labour Force Survey, which is available about 45 days after each calendar quarter. Historically, the main constraint on QALI (and also on other components of MFP) was the dependence on Blue Book data used to construct income constraints and available around 18 months after the end of the reference year. The methodology used in compiling sectional unit labour costs uses estimation techniques to extend income constraints beyond the last available Blue Book. Note incidentally that this QALI article is based on Blue Book 2012 income constraints (not Blue Book 2013 which will be published at the end of July). Hours are benchmarked to the latest (2013Q1) Labour Productivity release
use of a revised method of dealing with LFS respondents who do not provide information on their education status. In most time periods this is a tiny proportion of responses, so this change makes very little difference to QALI estimates
estimates are published before seasonal adjustment. In previous editions, QALI outputs were subjected to seasonal adjustment even though (i) most of the focus was on annual results (and in particular, MFP is annual), (ii) LFS hours were constrained to seasonally adjusted hours from the ONS labour productivity system and (iii) seasonal adjustment was carried out using a different procedure from that used elsewhere across ONS. For this release, we have not applied seasonal adjustment. This means that our results exhibit more quarter-on-quarter variability than those published last year (this is evident in the revision section of the article). But on the other hand, hours movements in this article are exactly consistent with hours movements published in the latest ONS Labour Productivity release. ONS intends to review the issue of seasonal adjustment of QALI before the next publication
updates to ensure consistency following changes to LFS survey questions. For instance, new LFS educational classifications, as described in the LFS User Guide, which came into effect from 2011Q1 have made it necessary to redefine what LFS education qualifications falls into each QALI education category
QALI estimates are badged as experimental statistics, which are defined in the Code of Practice of the UK Statistics Authority (UKSA) as "... new official statistics undergoing evaluation. They are published in order to involve users and stakeholders in their development ...". The Code of Practice envisages that experimental statistics should not continue indefinitely, and either be subject to formal assessment as official statistics or be discontinued altogether. ONS has received positive feedback from users on the relevance of QALI estimates and is therefore minded to develop QALI estimates with a view to undertaking a formal assessment by the UKSA. We welcome further feedback from users on this matter.
Formal assessment will involve a fundamental review of QALI data sources, methods and processes and is likely to take some time to complete. At this stage we do not envisage making any changes to the basic parameterisation of the QALI system, which is based on a pragmatic assessment of the level of detail that can be supported by the LFS source data. However, some users have expressed an interest in a more detailed industry breakdown, and we welcome feedback on this issue.
The medium term objective would be to compile and publish QALI estimates on a quarterly basis, with a view to providing quarterly estimates of MFP. However, this will also require extensive development of ONS systems for compiling quarterly estimates of capital services. Integration of capital services into the National Accounts is a requirement of the European System of Accounts 2010, so a process of development will also be needed to meet this requirement. ONS plans to incorporate capital services into the National Accounts in Blue Book 2015, which will be published in 2015.
ONS welcomes feedback from users on these development plans. Feedback can be sent by email to email@example.com or by phone (01633 455981 or 01633 455047).
Figure 1 shows indices of QALI and (unadjusted) hours over the period 2007-2012, with movements in labour quality (that is, the differences between the 2 indices) shown as index points on the right hand axis using the same scale1. The shaded portion shows periods of consecutive negative output growth (2008Q2 - 2009Q3). As noted earlier, there has been a good deal of commentary on the relative strength of hours, and the implications for labour productivity given that the recovery in economic output since the recession has been weaker than experienced in earlier recessions. However, while unadjusted hours had only just recovered to its pre-recession level by end-2012, the QALI index was some 5 percentage points above the pre-recession level. Labour quality has been increasing on a trend basis over the whole post-recession period (although interestingly not in 2007). Since in growth accounting terms an index point increase in labour composition is exactly analogous in terms of its impact on productive potential as a 1% increase in hours worked, it is particularly significant at a time when output growth has been heavily negative (in 2008-09) and then very sluggish. On the other hand, improvements in labour quality over 2011-12 are consistent with other recent research such as Blundell et al (2013), and with expereince after earlier recessions (see below).
Figure 2 shows the equivalent data for the market sector, defined as that sector where output is sold at economically meaningful prices, and broadly (though not exactly) equivalent to the private sector. Figure 2 shows a broadly similar pattern to Figure 1, but the recovery in labour quality is not quite as steep as for the whole economy. This would seem to suggest that there has been an especially pronounced increase in labour quality in the non-market sector.
The EUKLEMS growth accounting dataset (www.euklems.net) allows us to compare movements in labour quality following the 2008-09 recession and previous recessions. Comparisons with previous recessions have shown current UK labour productivity to have fallen much further and for a prolonged time with a slower recovery than previously seen.
Looking at movements in labour quality (figure 3) shows this recession to be more in line with previous recessions2. Five years on from the start of the recession, labour quality is now about one index point greater than at the same point after the 1991 recession, and 2 index points greater than at the same point after the 1974 recession. An exception is after the 1980 recession, when labour quality was little changed.
By contrast, figure 4 illustrates the resilience of hours worked recently, compared to both the 1980 and 1991 recessions, and compared with the 1974 recession if we allow for the double dip nature of that recession, with hours worked only falling on an annual basis from 1975.
QALI estimates back to 1993Q1 (1994Q1 for the market sector), together with a full set of supporting data on income weights and hours worked, are available in a reference table (926.5 Kb Excel sheet) accompanying this article.
The EUKLEMS framework for deriving estimates of movements in labour quality differs somewhat from the QALI framework, but as described in Acheson (2011) the differences are not material at the whole economy level.
In addition to aggregate QALI indices for the whole economy and the market sector, it is possible to construct separate QALI indices for each sub-category of each characteristic (education, age, gender and industry of employment). It should be noted that while for each characteristic the sub-categories are exhaustive, weighted averages of QALI sub-indices will generally differ from the whole economy QALI. This is because when conditioning on a single characteristic, the distribution of labour input across the other 3 characteristics will vary. For example, when conditioning on gender we would not expect the distribution of labour input in terms of education, age and industry of employment to evolve in exactly the same manner for males and females.
Education is a key element of human capital and this directly affects productivity by raising the quality of labour input. Modern growth theory assumes human capital facilitates innovation, and therefore that higher levels of education can move the economy onto a permanently higher growth path. ONS measures human capital as part of its programme of Measuring National Well-being.
As noted in previous QALI articles, there has been a dramatic increase in hours worked (and the share of hours worked, albeit from a low base) by workers with higher education qualifications since 1993. As shown in Figures 9 and 10, there has also been strong growth in hours over the last 5 years for these categories of workers. More generally, as shown by Figures 5 to 10, hours worked have fallen among workers with lower levels of education, been little changed among workers with A-levels and Certificates of Education (the spike in the latter in 2010 reflects temporary changes to the routing of LFS survey questions) and risen for the highest qualifications. This pattern, combined with the clear correlation between education and relative remuneration shown in Table 1, provides an intuitive explanation for the overall trend in labour quality over this time period.
The picture for labour quality within education categories is more nuanced however. At this level, movements in labour quality reflect movements (in hours and relative earnings) across industries, age groups and gender. For some education categories such as GCSEs and Certificates of Education, there is little net change in labour quality over 2007-12, while for workers with A-levels, labour quality has declined, particularly in 2012. There has been a small improvement in labour quality among workers with degrees and higher degrees, but the most pronounced trend improvement in labour quality, at least since the recession, has in fact been amongst workers with no formal educational qualifications. Hours worked have fallen very sharply for this labour cohort, so the quality improvement may reflect differentially large reductions in hours among the less well remunerated sub-categories, for example among younger workers and those in poorly- paid industries such as accommodation and food services. Another possible explanation is that workers in this education cohort may be less likely than other education cohorts to move into self-employment where, on average, hourly remuneration is lower than for employees.
Figure 11 throws more light on the influence of education on QALI by showing cumulative changes in shares of hours and earnings by educational category since 2007. There is clear evidence of a shift in the distribution of labour from less qualified to more qualified educational cohorts over this period. In general, one would expect changes in shares of hours worked and earnings to be correlated, and this is indeed the case. However, over the 5 years since 2007, the share of hours worked of those with no formal qualifications has fallen significantly more than their share of earnings, while for the A-level and equivalent cohort, the reverse is the case.
Figures 12 to 14 show that the different age cohorts identified in the QALI framework have demonstrated quite different behaviour in terms of hours worked over the period 2007-12, although the behaviour of labour quality has been broadly similar. The young (16-29 years) cohort experienced a steep reduction in hours worked as the recession took hold in 2008-09, and there has been only a faltering recovery in hours worked since 2009. Hours worked also fell for the 30-49 age cohort, although in this case, the reduction in hours preceded the recession and was significantly less pronounced from peak to trough. Again, there has been only a muted recovery in hours worked since 2009.
By contrast, hours worked have increased significantly among the older (50+) age cohort, with barely any discernible impact from the recession.
The improvement in labour quality since 2007 has been similar across all 3 age cohorts, at around 5 index points. This suggests that distributional changes within each age cohort in terms of education, gender and industry of employment have been broadly neutral. This would also suggest, a priori, that quality improving trends such as the shift to a better educated workforce have been reflected across all 3 age cohorts.
Another perspective is provided by looking at changes in shares of hours and earnings since 2007 (Figure 15). As expected, this shows that the older age cohort has increased its share of hours worked, with - over the whole period - a roughly corresponding movement in its share of earnings. However, for the middle age cohort, the cumulative decline in its share of hours is significantly smaller than the decline in its share of earnings, implying a differential re-allocation towards lower paid employment by education, gender and industry of employment categories, despite the trend improvement in labour quality for this age cohort. The youngest age cohort has experienced the largest fall in its share of hours worked, although for this age cohort the impact on its share of earnings is less pronounced.
The QALI picture for males (Figure 16) is similar to that of the whole economy, which is not too surprising given that males account for more than 60% of hours worked in the UK. Comparing Figure 16 with QALI for females (Figure 17), there is some evidence that hours worked are a little more pro-cyclical for males than for females.
Turning to labour quality, there has been a significantly stronger improvement in female labour quality than in male labour quality over the period 2007-12, reflecting differentially larger movements in the distribution of female hours worked towards better remunerated QALI categories in terms of education, age and industry of employment. This finding is not consistent with the growth of female part-time employment which some commentators have identified as part of the explanation of the productivity puzzle, since part-time employment is on average less well remunerated on an hourly basis than full-time equivalent jobs. However, part-time employment could still be part of the explanation to the extent that there has been a differential shift towards part-time employment among males, although it remains the case that overall quality has improved for both genders
Another noteworthy feature of these data is that, since 2007, there has been a small shift (around 0.2%) in the share of hours worked from females to males. However, the share of earnings by gender has moved very slightly in the opposite direction.
The following sections look at QALI by industry. In these analyses, movements in labour quality reflect compositional and relative price movements within a single industry, between workers with different educational qualifications, different age classes and the gender balance. QALI estimates by industry of employment will be used to derive industry level MFP estimates later in 2013.
This is a heterogeneous industry category combining highly volatile extractive industries (now dominated by oil and gas) with utilities, and is the smallest in terms of hours worked and earnings. As shown in Table 1, it also has the lowest relative hourly remuneration, dragged down by low earnings in agriculture. The growth in hours worked since 2007 (Figure 18) is wholly untypical of the long-term trend. Since 1993, hours worked fell more or less continuously until 2004.
Labour quality in this industry was little changed between 2007 and 2010. The improvement in labour quality in 2011 and 2012 probably reflects compositional movements within this QALI category, notably the increase in hours worked in the oil and gas component, where average hourly remuneration is relatively high.
Hours worked in manufacturing (Figure 19) seem to show a strong cyclical downturn in 2008 and 2009. Again we need to be aware of the long-term downward trend in manufacturing hours, but nevertheless the peak to trough fall in hours was the second largest of all industry categories in the QALI system, behind construction. There has been a flattening-off in hours worked since 2009, and a discernible increase in 2012.
Labour quality in manufacturing does not look spectacular over the period 2007-12, but is relatively rapid compared with the long-term trend.
As noted above, the peak to trough fall in hours worked in construction was the steepest of all QALI industry categories (Figure 20), and this is the only industry category where hours have continued to decline (on an annual basis) since the recession, including in 2012.
Construction is the industry category which demonstrates the least movement in labour quality, presumably reflecting limited opportunities for substitution towards higher skilled and more highly remunerated categories of labour. But as in a number of other QALI industry categories, the improvement in labour quality in recent years has been better than the long-term trend. Labour quality in construction has improved by around 2 index points since 2008, compared with a net increase of only 3 index points over the previous 25 years.
This is the second-largest QALI industry category (after OPQ) in terms of hours and earnings. The time series of hours worked (Figure 21) is similar to that of the economy as a whole.
The improvement in labour quality for industry GI has been a little less pronounced than for the economy as a whole. However, labour quality has still increased by around one index point per years since 2008.
There was a discernible fall in hours worked in this QALI industry category during the recession, and hours worked continued to fall in 2010 and 2011 before staging a partial recovery in 2012 (Figure 22).
Over the long term, the transportation and storage industry has experienced the second weakest growth in labour quality, behind construction. It is also worth recalling from Table 1 that this industry has the second lowest average remuneration, behind ABDE. In this context, the growth in labour quality since 2007, which has been more or less in line with growth at the whole economy level, represents an improvement on the long-term trend.
In terms of hours worked (Figure 23) information and communication is the most cyclical of any service sector industry, suffering the steepest peak-to-trough fall during the recession, and also the strongest recovery since 2009.
In addition to the strong growth in hours, labour quality has also shown one of the most rapid rates of improvement, by around 2 index points per year since 2008. Accordingly the overall QALI index of labour input has risen by around 15% over the last 3 years.
Hours worked in financial and insurance activities grew by 3.6% in 2011 and a further 1.5% in 2012 (Figure 24), to surpass their pre-recession level.
Perhaps surprisingly, given well-publicised job losses among highly paid investment bankers, quality improvement in this QALI industry category has made the largest contribution to overall labour input of any industry over the period 2007-12. This is consistent with longer term trends: along with industries J and RSTU, quality has grown strongly over the whole period covered by the QALI dataset.
Hours worked in QALI industry category LMN fell over the recession but have recovered strongly thereafter, growing by 1.8% in 2011 and 5.2% in 2012.
As is the case for a number of other industries, quality improvement in this group of industries since 2007 has been better than the long-term trend, and particularly better than in the 1990s when labour quality fell for this QALI category.
Hours worked in this QALI category grew sharply in 2009 but there has been little net movement thereafter. This may reflect downward pressure on public sector employment as part of the Government's deficit reduction programme.
There have been year-on-year improvements in labour quality of around 1 index point in 2011 and 2 index points in 2012. This is closely in line with the pattern at the whole economy level.
This is the only QALI category where hours worked have not increased since the recession, and in particular the only service category to experience a decline in hours worked in 2012. This is despite a boost to section R from the Olympic and Paralympic Games in 2012Q3.
As noted earlier, labour quality in this category has performed strongly over the long term. However, the improvement has been rather muted since 2010, with a small deterioration in labour quality in 2011. Nevertheless, quality movements account for substantially all of the quality-adjusted increase in labour input in this category since the recession.
The main sources of revisions are:
Revisions to benchmarks for hours and earnings
Revisions to LFS source data. For example, the regular review of LFS population weights means that LFS data used for quarters after 2009Q3 differ from those used in the 2012 QALI article
Methodological changes. As noted earlier there has been a change to the method of allocating LFS respondents who do not provide a response on their level of education, and unlike previous QALI articles, the estimates in this article have not been seasonally adjusted.
Figure 28 shows percentage differences in the whole economy QALI index between this year and the non-seasonally adjusted data underlying last year's article. It can be seen that this release has tended to revise the QALI series slightly downwards, driven by downward revisions from growth of hours worked. However since 2003, the impact of revisions to income weights have started to negate this bias and the last few quarters of the previous release have been revised upwards. The largest impact is on QALI for 2009, which was reported to have fallen by 1.5% in last year's QALI article; this has now been revised to a fall of 0.6%.
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
Acheson, J. (2011) "Quality–adjusted labour input: new quarterly estimates for 1993 to 2009 and annual estimates from 1970", Economic & Labour Market Review, April 2011
Acheson, J. and Franklin, M. (2012) "Quality-adjusted labour input: estimates to 2010", ONS. Available at:
Appleton, J. and Franklin, M. (2012) "Multi-factor Productivity - Indicative Estimates, 2010", ONS. Available at:
Blundell, R., Crawford, C. and Jin, W. (2013) "What can wages and employment tell us about the UK's productivity puzzle?", IFS Working Paper. Available at: http://www.ifs.org.uk/wps/wp201311.pdf
Franklin, M. & Mistry, P. (2012) "Productivity Measures, Sectional Unit Labour Costs", ONS. Available at: http://www.ons.gov.uk/ons/dcp171766_289064.pdf