Quality Adjusted Labour Input (QALI) is a method of measuring changes in the volume of labour input into production which accounts for changes in the composition (or ‘quality’) of the workforce as well as changes in hours worked. This article presents new results for 2013 for the whole economy, the market sector and a range of sub-aggregates by industry, educational qualifications, age and gender, as well as minor revisions to the back series. QALI provides a more complete picture of the input of labour into the production process than traditional metrics, such as hours or jobs. The resilience of hours worked over the downturn period in relation to output has given rise to what has come to be referred to as the ‘productivity puzzle’. If labour quality had fallen over this period, then this may have offered a partial explanation for the puzzle. However, this article reports that labour quality has improved in every year since 2008 (albeit at a slower rate in 2013 than over 2008-12), therefore accentuating rather than alleviating the productivity puzzle. For the first time, this article reports results from stratifying the workforce in terms of occupational groups as an alternative to stratification by educational qualifications as a proxy measure of skills. The broad finding is that using standard classifications of occupations, movements in labour quality are less pronounced than using educational qualifications.
The authors thank Bhavik Patel, Harry Ravi, and John Allen for research support in preparing this article.
QALI differs from traditional measures of labour input as it weights the hours worked by different types of workers by their relative contribution to economic production. Workers are categorised in this weighting structure by identifiable characteristics (age, gender, industry of employment and education level). Changes in the hours worked by these different worker types are then weighted by their shares of total labour income. The rationale for this approach is that under competitive markets, economic theory suggests that the different factors of production (different types of workers, and different types of capital assets) will be paid reflecting their marginal productivity. There are indeed circumstances where this assumption would not be appropriate, such as where there is significant monopsony in employment markets, or where an industry operates as an oligopoly, as noted in O'Mahony and Timmer (2009). But in general, it is reasonable to expect that on average, more highly paid workers contribute more to production than lower paid workers, and that this will be reflective of the distribution of skills in the workforce.
Using a suitable weighting system, it is possible to subtract movements in hours (sometimes referred to as 'unadjusted hours') from movements in QALI indices to identify the pure 'quality' or 'compositional' movement in labour input to production.
The main use of QALI is as an input into a growth accounting framework, in which growth in economic output can be decomposed into contributions from labour (from both the volume of hours worked and a quality component), contributions from capital services (which measures the input of different types of asset in a method analogous to QALI), and multi-factor productivity (MFP), which is captured as a residual component. Multi-factor productivity is also known as ‘total-factor productivity’, ‘disembodied technological change’, and the ‘Solow residual’, and captures changes in output which cannot be attributed to measured changes in the volume of labour or capital inputs. Indicative MFP estimates to 2012 were published by ONS in January 2014 (Field and Franklin, 2014). ONS will publish MFP estimates to 2013 based on the QALI estimates in this article and revised and updated estimates of capital services after the publication of Blue Book 20141.
At the whole economy level, QALI increased by 3.1% in 2012 and by 2.4% in 2013 (see Figure 1 below, which shows cumulative movements since Q1 2008.) Hours worked across the whole economy grew by 2.0% in each of 2012 and 2013, so the implied growth of labour quality was 1.1% in 2012 and 0.5% (after rounding) in 20131. 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.
Changes in labour quality are reflective of the distribution of hours worked among categories of workers, and differentials in the hourly pay of categories of workers. For example, if hours worked by a highly skilled and highly remunerated type of labour (such as brain surgeons) increased, then the volume of labour input as measured by QALI would increase by more than simply the observed proportional increase in hours worked. Conversely, a decrease in hours worked by unskilled workers in elementary occupations who receive lower than average remuneration would result in a fall in QALI by less than the proportional decrease in hours worked.
In brief, QALI weights changes in the hours worked by different ‘categories’ of labour by way of a Törnqvist index (where weights are taken as an average of the current and previous period, and enter geometrically), to construct quality adjusted changes, which can then be compared to movements in unadjusted series of hours worked in order to extract quality movements. The detailed methodology used is outlined in Acheson and Franklin (2012), and is not repeated here, as no material changes to the QALI methodology have occurred since that publication.
Table 1 provides some context, showing relative earnings of the main QALI aggregates. Within QALI, the workforce is cross classified by industry (10 sub-categories), by education (six sub-categories), by age (three sub-categories), and by gender (two sub-categories), giving a total of 360 QALI sub-categories (as given by the product of the four strata).
|Whole economy = 100||2007||2010||2013|
|Agriculture, forestry and fishing; mining and quarrying; utilities (ABDE)||81.9||73.8||85.4|
|Wholesale and retail trade; accommodation and food services (GI)||85.6||86.6||88.7|
|Transportation and storage (H)||78.8||78.3||79.3|
|Information and communication (J)||147.2||145.0||143.7|
|Financial and insurance activities (K)||200.8||189.8||176.3|
|Real estate activities; professional and scientific activities; administrative and support activities (LMN)||88.5||87.5||84.4|
|Public administration and defence; education; health and social work (OPQ)||116.4||115.6||114.0|
|Arts and entertainment; other services (RSTU)||72.5||77.9||84.4|
|Masters and doctorates||159.2||154.9||143.6|
|First and other degrees||134.4||131.0||126.4|
|Certificates of education or equivalent||111.7||107.4||107.3|
|A – levels or trade apprenticeships||91.4||89.5||88.0|
|GCSEs or equivalent||80.2||78.2||77.8|
Looking at relative remuneration over time shows that for the age and gender categories, results remain relatively stable. Generally, age is seen as a proxy for on the job experience, which explains the increase in relative income between 16-29 year olds and 30-49 year olds. The drop-off in relative remuneration for the 50+ group is usually interpreted as better paid workers being able to retire earlier, while lower paid workers will have to work longer. Another interpretation, at least for industries which are physically demanding such as some sub-sections of manufacturing and the other production industries, is that workers may become less physically able with age and consequently have lower marginal productivity.
The gender earnings gap is usually interpreted as a reflection of earnings being correlated with experience and seniority, with females on average (and historically) being subject to more time out of the labour market due to child rearing. There is some support for this view given by the cross classification of age and gender, as on average, income differentials for males and females are narrower for the younger age group and wider for the older age groups.
The income differentials across industries of work and educational groups demonstrate the importance of QALI. For example, relative remuneration in industry K (finance and insurance activities) is more than twice that of industries G and I (retail and wholesale, accommodation and food services) or industry H (transport and storage). Income differentials between workers with differing levels of educational attainment show a very clear picture; that educational levels are strongly correlated with earnings, with workers with no qualifications earning almost 40% less per hour than the UK average. However, in relative terms, the earnings of this group have declined only slightly between 2007 and 2013. By contrast, the earnings premium of higher education groups has narrowed significantly over this period, reflecting increases in the share of workers with higher educational qualifications.
Detailed analysis of employment rates and underemployment for workers of differing educational attainment can be found in a recent ONS article 'Qualifications and Labour Market Participation in England and Wales' (Office for National Statistics, 2014), which shows that proportionally more workers with no qualifications tend to work part time, and that these workers have relatively low participation rates.
Educational attainment and occupations
QALI can be described as a model in which the assumption is made that there are a discrete number of worker ‘types’ which are stratified by statistically identifiable characteristics which reflect differences in labour quality, and that these differences can be measured by relative earnings. A point of contention is whether it is more methodologically sound to stratify the workforce by educational attainment, or by occupation.
Sound arguments can be made for each method, as while educational attainment is a relatively simple variable to collect in labour force surveys compared to occupations, it suffers from a failure to take account of training outside the education system, such as on-the-job training. On the other hand, survey respondents are less likely to misreport their qualifications compared to their job titles and duties. A full discussion of measures of skill from labour force surveys can be found in Lemaître (2002).
In this article we present QALI estimates based upon broad occupational categories, offering comparisons between these and education based QALI. The broad result is that on the standard occupational skill classifications (SOC), QALI fails to capture the magnitude of quality movements over time that can be found when using educational categories. This is interpreted as being largely due to the way in which the broad SOC classifications are determined, and further research will aim to refine the parameterisation of occupational categories to improve on these results.
Figure 1 shows quarterly movements of QALI, unadjusted hours and labour quality for the period 2008-2013. At the end of 2013 QALI was over 8 percentage points higher than before the start of the economic downturn, compared with a cumulative increase of around 3 percentage points in hours worked. Thus growth of labour quality accounts for more than the growth of labour input over this period, albeit that in 2013 labour quality grew by less than 0.5%, which was the slowest rate of growth at the whole economy level since 2004.
There has been considerable commentary on the relative resilience of hours compared to output over the period in Figure 1, and the implications this has on productivity. Broadly, these results imply that for the whole economy, growth of labour productivity over this period would be considerably weaker when using a quality adjusted (or a true 'volume of labour services') measure of labour input, than by using unadjusted hours worked. For example, the Bank of England has recently suggested that whole economy output per hour is around 16% below the level implied by the pre-downturn trend (Barnett et al, 2014). This shortfall would increase to around 21% using QALI.
Figure 2 shows the labour quality component of Figure 1 alongside the corresponding component of a counterfactual QALI weighting structure where occupational classifications are used to stratify the workforce instead of the educational qualifications used in the main QALI system. Occupational groups are listed as a footnote to Figure 2. The broad result as noted earlier is that using broad occupational categories for QALI does not capture the magnitude of labour quality movements that can be found by using education. Under the counterfactual occupational stratification, labour quality grows by about 2 percentage points between 2008 and 2010 (compared with about 3 percentage points when using education strata). Thereafter, labour quality on the occupational stratification cycles around with no clear trend, while quality using the educational strata continues to grow significantly.
These results are broadly consistent with those found in Bell et al (2005), an early version of QALI which found that adjusting labour quality for occupation captured less than a third of the contribution of education over the period 1993-2002.
It is important to note that Figure 2 does not demonstrate that stratification by educational qualification necessarily yields more accurate estimates of QALI than stratifying by occupation. All that can be deduced is that the movements in labour input are larger in one framework than the other. It turns out that this is not because there is less variation in earnings by occupation than by educational qualification. In fact the ratio of earnings of the most skilled to the least skilled occupations is roughly comparable with the ratio of earnings by qualification shown in Table 1. But there is less movement in labour input across occupational boundaries. For example, whereas the share of hours worked by the top two educational categories increased by around 8 percentage points between 2007 and 2013, the share of hours worked by the top two occupational categories increased by only around 1 percentage point.
These results do not necessarily invalidate the use of occupations in QALI, as it is possible that a more selective use of occupational classes may find similar quality movements to the educationally based version of QALI. This might involve using the more specific ‘minor group’ classifications from LFS to create custom occupational skill groupings. Part of the advantage in doing this is that it would result in categories which better reflect earnings differentials. For example, highly skilled technical occupations are on average, more highly remunerated than teaching, but in the broad SOC classifications, highly skilled and remunerated technical occupations are grouped with occupations which are relatively poorly remunerated, and teachers with occupations which are highly remunerated. It is likely that the educational attainment levels of these two groups is more closely related to their earnings, as it would be expected that highly skilled technical workers would have a higher certificate or degree equivalent qualification, aligning them with other relatively well remunerated workers.
Figure 3 shows QALI estimates 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 3 shows that movements in labour quality in the Market Sector have been very similar to movements across the whole economy (Figure 1) since 2008.
The following section looks at QALI by industry. Changes in labour quality by industry reflect compositional and relative income movements within a single industry, which themselves are reflective of changes in the educational, age, and gender makeup of workers in that industry. Annualised QALI results by industry are a primary input used to derive MFP estimates.
ABDE: Production industries excl Manufacturing; Agriculture
This industry aggregate is heterogeneous in that it composes of agriculture, for which incomes are very low, mining and quarrying, a highly volatile industry for which incomes are very high, and utilities, for which incomes are high. It is the smallest industry aggregate in the QALI framework when measured by either total income or total hours. Growth in hours in this aggregate since 2008 is atypical of the longer term trend, which has been driven by falling hours in agriculture and mining and quarrying. Growth in hours picked up strongly in 2013 after a fall in late 2012.
Labour quality for this aggregate has, after a period of little change, grown strongly to the most recent quarter. This is likely to reflect compositional movements within the category, in particular a redistribution of hours worked towards more highly remunerated workers in mining and quarrying and the utilities, and away from agriculture.
In general, the long term trend for manufacturing hours worked has been downwards, but 2008 to 2009 shows a particularly strong fall. Despite this, labour quality picked up over this period, leading to QALI falling by less than hours worked. From 2009 to the end of 2013, hours worked have remained relatively unchanged, but strong quality improvements have led to a partial recovery in QALI.
After a seemingly serial decline in hours worked in construction between 2007 and mid-2012, hours increased strongly through 2013, reflecting a marked upturn in construction activity. Typically, little change in labour quality occurs in the construction industry, largely owing to the fact that it is a physical labour intensive activity with little opportunity for substitution towards more skilled categories of labour. Despite this, the most recent results have shown a pickup in labour quality, leading to QALI for the industry being only around three percentage points below its 2008 level at the end of 2013.
GI: Wholesale and Retail; Accommodation and Food Services
This industry aggregate, the second largest in the QALI framework in terms of both hours and earnings, is typically seen as relatively low skilled and low capital intensive. Movements in hours worked follows a broadly similar path to that of the whole economy. Labour quality for this industry aggregate has shown a general upwards trend for the whole downturn period, but it has been less pronounced than at the whole economy level, perhaps reflecting the lesser degree of heterogeneity across workers in these industries.
H: Transportation and Storage Services
The partial recovery in hours for the transportation industry between 2011 and 2012 has halted over the past year. Over the long term, this industry currently shows the smallest contribution from labour quality to labour input which, like the construction industry, is likely to reflect the relatively low scope for up-skilling of the workforce compared to other industries.
J: Information and Communication Services
Over the long term, the contribution of labour quality for the information and communication industry has typically been the strongest of any industry, largely reflecting the increased requirement for more highly skilled (and consequently, relatively higher remunerated) workers. This can be seen by the strong increase in the proportion of hours worked by people with degrees and higher degrees in the industry, and the corresponding fall in proportions of hours worked by those with A-levels or less. However, labour quality fell slightly in 2013 compared with 2012. Over the downturn, hours worked fell to a trough of around 5 percentage points below their pre-downturn level by the start of 2010, but have increased rapidly and now stand around 7 percentage points above their Q1 2008 level.
K: Financial Services
Hours worked in the finance and insurance industry recovered strongly in 2010 and 2011 but have since remained around their pre-downturn level. Labour quality fell slightly in 2013 compared with 2012. This contrasts with the longer term trend (available in QALI_RFT_SUMM (148.5 Kb Excel sheet) ), which shows that labour quality has grown more in this industry more than in any other industry. This should be borne in mind when analysing long-term productivity estimates for the industry, as they would be significantly depressed over the long term when adjusting labour input for quality.
LMN: Business Services
For this industry aggregate, hours worked have continued their steep increase from Q2 2011 and were almost 10 percentage points above their pre-downturn level at the end of 2013. Labour quality has also picked up during the year, leading to very strong growth in QALI. This industry aggregate is heterogeneous, containing highly skilled professional and scientific activities with medium skilled real estate activities and lower skilled administrative and support activities. Most of the growth of labour quality in recent years has been attributable to compositional changes in professional and scientific activities towards relatively higher remunerated categories.
OPQ: Public Services
This industry aggregate is mostly comprised of government services, but also includes the private sector components of health and education. Hours worked increased by around 2% in 2013 compared with 2012. Labour quality increased by around 0.6% in 2013 compared with 2012, although the quarterly profile was fairly flat through the year.
RSTU: Arts and Recreation; Other Services
An unusually sharp upturn in hours worked in Q4 2013 in this industry aggregate has led to their reaching their pre-downturn level. However this quarterly change followed a period of both falling hours and a small decrease in labour quality, despite the expected boost to hours for industry R from the hosting of the Olympic Games in Q3 2012. Labour quality also picked up sharply in Q4 2013, after showing little net change since late 2011.
Educational attainment is a key variable used in identifying differences in human capital. In economic theory, it is seen as either being a signal to employers that a worker able to attain a higher qualification would be more capable (as the ‘psychic costs’ of gaining a qualification would be lower for a more capable worker than for a less capable worker) and consequently more productive, or alternatively it is seen as an implicit process by which workers gain the knowledge and skills to be more productive through gaining qualifications. In either case, it is apparent that higher qualifications are well correlated with relative earnings, and modern growth theory assumes that human capital facilitates innovation, and that higher levels of educational attainment can put an economy onto a permanently higher growth path.
As noted in previous publications of QALI, there has been a substantial increase in the share of hours worked by people with higher qualifications since the start of the QALI series in 1993. Compositional changes within educational groups tends to be rather muted when compared to other categories, and this is usually interpreted as a signal that movements in other characteristics (industry of employment, age group and gender) tend to remain relatively stable within cohorts of workers grouped by educational qualifications.
No qualifications and GCSEs
QALI for both the no qualifications and GCSE categories show positive contributions from labour quality over the period 2008-2013. This is also true of the longer term trend: labour quality for the no qualifications category being among the strongest of any QALI sub-group. Hours worked for the no qualifications category fell by around 30 percentage points between 2008 and 2012 but showed a small increase in 2013. This combination of rising labour quality and steeply falling hours suggests a shake out of less well remunerated workers with no qualifications, for example because hours worked have fallen more for younger workers with no qualifications than for older workers. Hours worked for the GCSE category have also fallen since 2008 although not as precipitously as the no qualifications group.
A levels and Certificates of education
Hours worked for those with intermediate level educational qualifications (A-levels and Certificates of Education) were not as heavily impacted by the economic downturn as less well qualified workers and had regained their pre-downturn level in 2013.
There has been a small but significant decrease in labour quality among the A-level cohort, reflecting net movement towards relatively less well remunerated employment in terms of the industry, age and gender distributions of these workers.
First degrees and higher degrees
The degree-educated and higher degree-educated groups have seen the largest increases in hours worked since the start of the downturn, with hours worked for these groups barely falling below their pre downturn peaks, and growing by between 20 and 30 percentage points by the end of 2013. In both cases, labour quality has contributed little to the overall movement of the QALI series over this period. And in both cases, there is some evidence of diminishing labour quality in the most recent years.
Results by Age
Across different age bands, movements in both QALI and hours have been varied. Hours worked for the 16-29 years group fell markedly to 2009, and remained around that level until around mid 2012. A partial recovery has been made, but hours are still below their pre downturn level. Labour quality has made a fairly significant contribution to overall labour input, meaning that QALI exceeded its pre-downturn level during 2013.
While hours worked for the 30-49 years group has remained relatively stable over the period, falling lightly until around the middle of 2011 and returning to their pre-downturn levels at the start of 2013, labour quality has contributed to a sharp increase to QALI over the whole period. In contrast to the two younger age groups, hours worked for the 50+ category seem completely unaffected by the downturn, and labour quality has contributed strongly to the QALI series, amounting to around a five percentage point contribution by the end of 2013.
Results by Gender
As noted in previous QALI articles, the use of gender to stratify the workforce in QALI is not intended to imply ‘quality’ differences in male and female labour, but rather is used to account for the empirical differences in earnings between males and females. As males account for a larger share of hours than females, the picture for both QALI and hours worked for males is very similar to that of the whole economy.
There has been a significantly larger contribution from labour quality for females than for males over the period 2008-2013, which reflects larger movements towards better remunerated categories of labour in terms of education, industry of employment, and age. This is typical of the longer term trend, as the share of earnings for females has, in general, increased since 1993.
For QALI, the main sources of revisions are from revisions to the hours and income benchmarks (which come from the ONS Labour Productivity and the Unit Labour Cost systems respectively), revisions to the LFS source data, and methodological changes.
Figure 25 shows revisions in terms of percentage points for the whole economy QALI series between results published in the 2013 article (Franklin and Mistry, 2013) and the current estimate, and a decomposition between hours and labour quality. Revisions are an order of magnitude smaller than between the 2012 and 2013 QALI articles, and largely reflect the adoption of Blue Book 2013 income benchmarks, for which small revisions occur back to the beginning of the time series. These revisions do not cause any fundamental changes to the interpretation of previously published trends in QALI.
The main reason for the much more subdued revisions this year is that no changes have been made to the seasonal adjustment process of QALI since last year. Previously, seasonally adjusted productivity hours were used to benchmark the hours estimates from the QALI system by industry, and then the resulting QALI series were seasonally adjusted. This method was abandoned for the 2013 article, and the system currently uses seasonally adjusted productivity hours only. The role of seasonal adjustment in QALI remains to be an area for development, as the results by characteristics other than industry display clearly seasonal properties. One possibility is to use non seasonally adjusted hours benchmarks and to seasonally adjust QALI indices 'end of pipe'. ONS intends to look further at the issue of seasonal adjustment as part of a thorough-going review of the QALI system in the future.
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 and Franklin M (2012) 'Quality-Adjusted Labour Input (experimental) - Estimates to 2010', Office for National Statistics. Available at: http://www.ons.gov.uk/ons/rel/icp/quality-adjusted-labour-input--experimental-/index.html
Barnett A, Batten S, Chui A, Franklin J & Sebastià-Barriel M (2014) 'The UK Productivity Puzzle'. Bank of England Quarterly Bulletin, 2014 Q2. Volume 54 No. 2. Available at: http://www.bankofengland.co.uk/publications/Documents/quarterlybulletin/2014/qb14q2.pdf
O’Mahony M and Timmer M (2009) 'Output, Input and Productivity Measures at the Industry Level: the EU KLEMS Database', Economic Journal 119(538). Available at: http://onlinelibrary.wiley.com/doi/10.1111/j.1468-0297.2009.02280.x/abstract