The squeeze on real wages--and what it might take to end it.
Gregg, Paul ; Machin, Stephen ; Fernandez-Salgado, Marina 等
I. Introduction
Compared with earlier economic downturns, the deep recession and
protracted period of economic stagnation that occurred in the aftermath
of the financial crisis of 2008 has seen different behaviour in the UK
labour market. Leaving aside those aged under 25, employment saw only
minor falls--by about 2 per cent--relative to the sizable 7 per cent
fall of GDP that occurred. Employment started to recover alongside
output and stood at 0.5 million higher than pre-recession levels by the
end of 2013, whilst output still remained a little below peak levels.
Again the exception is those aged under 25, for whom the employment
recovery has only just started. This poor output performance coupled
with strong employment means that productivity remains well below peak
levels and lies some 15 per cent below levels seen at a similar stage
after previous recessions. This is what has come to be widely referred
to as the 'productivity puzzle'.
At the same time as this pattern of relatively strong employment
and poor productivity performance, the UK has experienced a significant
fall in real wages. The scale of the real wage falls is historically
unprecedented, certainly in the past fifty years where broadly
comparable records exist (see Taylor et al., 2014). When assessed over
the range of measures and data available, median real weekly wages have
fallen by around 8 per cent since early 2008. This equates to an annual
earnings loss of about 2000 [pounds sterling] in today's prices.
Falls in hourly wages are slightly less marked owing to the increase in
part-time working but the overall impression is similar. The most recent
data suggest little in the way of reversal as the lack of real wage
growth has continued through 2013.
Of course, these patterns are linked. Falling real wages combined
with difficulty accessing loans from banks for investment have combined
to encourage firms to use extra workers rather than new investment in
capital to meet demand. Business investment fell by 14 per cent from
2008 to 2009 and has yet to recover this lost ground. Hence, the
extraordinarily poor period of flat productivity and the (relatively)
good picture for employment that has been seen. In turn poor
productivity performance offers little scope for rises in real wages,
thereby completing the circle.
Academic research around this subject remains limited. In an
earlier paper (Gregg et al., 2014) we undertook an initial analysis of
whether there may have been a structural shift in the evolution of real
wages by studying whether one can garner evidence of a changing role of
unemployment in explaining falling real wages. This, alongside related
contributions by Pessoa and Van Reenen (2014) and Blundell et al.
(2014), who respectively explore productivity movements and cyclical
shifts in the composition of the employed, formed a special session at
the 2013 Royal Economic Society conference.
In this paper, we seek to make three new contributions in terms of
studying what has happened to real wages more recently, compared with
the earlier part of the past twenty five years. We first offer a
thorough, descriptive analysis exploring movements in real wages across
various sources of wage information, concepts of pay and inflation
measures. The aim is to ascertain the robustness of the evidence on
falling real wages and to generate a better understanding as to why the
extent of measured falls vary in different settings. We look at
different data sources, variations across major groups in the workforce
and explore in more detail the evidence for nominal wage stickiness.
The second aim is to rehearse what we know about the reasons why
real pay has been falling. We argue that three factors have been
important drivers of these unprecedented real wage falls. First,
unemployment has been exerting a larger downward pressure on wages than
in previous recessions. Second, median wages have historically tracked
productivity growth in a mutually reinforcing relationship. Productivity
gains then create room for rising real wages and rising real wages
create the incentive for firms to invest in labour saving technologies.
Low wages have in turn created incentives for firms to meet demand by
hiring more workers rather than through investment. Hence the weak
productivity record, though it has been good news for jobs, which has
restricted the room for real wage rises. Third, and predating the
recession, wages of typical British workers have no longer kept up with
productivity gains made in the economy. This stems from a growing
contribution of total compensation going towards supporting pensions,
not just for current but also already retired workers, and because it is
the highest paid (the top 1 or 2 per cent) who have taken a
disproportionate share of the gains from productivity leaving little
room for wage gains by ordinary workers.
The third contribution of the paper is to offer a consideration of
when or if the pattern of observed real wage falls could end. This
entails a discussion of the scope for economic recovery and policy
change to be able to reverse the observed declines. The extent to which
the three reasons that we argue caused falls in real wage growth can be
alleviated is critical to this discussion.
2. Documenting the rise and fall of real wages in the UK
To document what has happened to real wages over time, we analyse
wage data from various sources over the twenty-five year period from
1988 to 2013.1 The start date is determined by the fact that 1988 is the
first year where we have Consumer Price Inflation (CPI) data. The CPI
has become the preferred measure of price inflation. It is used as the
government's inflation target which the Bank of England Monetary
Policy Committee is required to achieve (see ONS, 2012). So in this
paper we principally use the CPI as the measure of consumer prices to
compute the real consumer wage. Where relevant, we also note any
pertinent differences from considering other alternative price series.
What do different wage measures tell us?
Figure 1 shows real wage movements since 1988 at three points of
the wage distribution, namely the 10th, 50th and 90th percentiles. The
figure uses New Earnings Survey/Annual Survey of Hours and Earnings
(NES/ ASHE) data and indexes the three wage growth series to 0 in 1988,
thus showing cumulative growth over time. The ASHE series is derived
from employer pay records and is widely seen as the most accurate wage
data available. It is approximately a 1 per cent sample of all workers.
It is also a panel, covering the same population in every year, and we
exploit this feature later. The data presented is weighted to reflect
the level of employment levels and patterns seen in the Labour Force
Survey. (2)
The left-hand panel of figure 1 expresses the wage growth series in
real terms using the CPI and is thus real consumer wages, whilst the
right-hand panel deflates by the GDP deflator and shows real product
wages (i.e. the real cost of employing workers for firms given the
prices they charge for their output). Figure 2 shows movements over time
in the CPI and the GDP deflator. (3) Looking over the full period shows
that the choice of price deflator makes little difference to the overall
longer run patterns of real wage growth shown in figure 1, although
there are some subtleties that arise in different years. For example,
between 2001 and 2008 the CPI rose more slowly than the GDP deflator and
this reversed from 2008. This means that the real product wage, shown in
the right-hand panel of figure 1, shows slower increases between 2001
and 2008 and smaller falls since then when compared to real consumer
wages.
[FIGURE 1 OMITTED]
However, the broad picture shown by figure 1 is of real weekly
wages of the typical (median) worker rising pretty consistently to
around 2003, except for a brief period of stagnation through the
recession of the early 1990s. This was followed by a period of slower
wage growth between 2003 and 2008, and very sharp declines after this.
(4) By 2003 wages of the typical worker (at the median) reached over 30
per cent above levels seen in the late 1980s. After slowing from then up
to 2008, real wages then fell sharply, falling on this measure by 8 per
cent in just four years, before stabilising in 2013 at levels last seen
in the early 2000s.
The figure also shows that the period from 1988 to 1999 was
associated with faster growth for high earners (the 90th percentile) and
slower growth for lower paid workers at the 10th percentile. (5) Real
wages of the lowest paid did fall through the recession of the early
1990s, although not by as much as in the recent period. From 1999 the
pay of lower wage workers stopped falling further behind median wages,
probably due to the introduction of the UK's national minimum wage
(NMW), even though this directly covered only around 5 per cent of the
workforce (Machin, 2011). The wages of the highest paid continued to
pull away though until 2008. Since then wages have fallen pretty much
equally for all groups.
[FIGURE 2 OMITTED]
Table 1 shows the magnitudes of real wage growth in these three
sub-periods for a range of alternative sources of earnings data and
measures of earnings. From 1988 to 2003, real weekly wages rose at 1.8
per cent per annum for the typical (median) worker according to the ASHE
data used in figure 1. This eased to 1.5 per cent per annum between 2003
and 2008. After that, a sharp decline set in, of -1.6 per cent a year.
One important thing to note, and something we return to later, is that
the 2003 to 2008 period of slow wage growth occurred in a period of good
productivity growth and near full employment. The fall in median real
wages on the CPI measure over the whole 2008-13 period is 8 per cent, or
around 2000 [pounds sterling] for a typical worker in today's
prices in the ASHE data (based on 2008 median annual earnings expressed
in 2013 prices using the CPI, which is almost exactly 25,000 [pounds
sterling]).
Whilst the ASHE database is widely thought of as the mostly
reliable series for earnings, it is not the only data source available
and is somewhat dated compared to other more up to date sources (for
example, April 2013 was only just released at the time of writing). It
also had some revisions to its earnings measures and weighting, most
notably in 2004/5. Table 1 therefore also shows growth for average and
median hourly wage growth since 1988 based on the Labour Force Survey
(LFS) (6) and the ONS Average Weekly Earnings (AWE). The LFS data is
based on a survey of households rather than employer records and thus
may be subject to some reporting biases if people are giving a general
sense of their earnings rather than referring to official records, such
as monthly pay slips. However, it also captures very short-term jobs
which might be missed by the annual ASHE series. The other main ONS
weekly wage data series, Average Weekly Earnings, which is also employer
based, but is reported monthly rather than annually, is thus more up to
date but only average weekly earnings are available.
These alternative series suggest similar wage growth prior to 2008
but slightly less marked falls since 2008 for median and mean weekly
wages than for comparable series in ASHE. Compared to the 8 per cent
ballpark fall in ASHE, the falls in real mean and median weekly wages
since 2008 range from 4 to 7 per cent for these alternative series.
However, an issue here is that in 2013 cuts to higher rate taxation saw
high earners shift some of their earnings into the new tax year, from
April, to attract a lower tax liability. This inflated earnings in the
second quarter of the year and reduced them in the preceding quarter.
Thus there is a noticeable spike in monthly data in April 2013 in the
Average Weekly Earnings series reported by ONS and the second quarter in
the LFS data. The data we report for AWE are the three- monthly average
including March and thus this spike is reduced, but the LFS numbers are
materially affected by the treatment of this period. The annual average
fall in real wages in the LFS data between 2008-13 is larger (by -0.2 to
-0.4 per cent per annum) if an average of quarter 1 and 2 for 2013 is
used. This implies an extra 1 per cent fall in real wages over the
period in the LFS data.
On the nuances that follow from considering different wage/earnings
measures, table 1 also shows that hourly wages on all measures showed
slightly faster growth than weekly wages in the period 1988 to 2003.
This is because average hours of work fell, partly because of increased
part-time working in the workforce but mostly due to full-time workers
cutting back their hours of work as living standards rose. For the ASHE
data we also report numbers on basic earnings, that is excluding bonuses
and overtime earning, and growth at the median is slightly lower,
suggesting declines in incentive pay contributed a little to the fall in
earnings. The ASHE data also report annual earnings for workers who stay
in the same firm through the full year. Here the measured fall in real
wages since 2008 is the largest among the available sources, being down
by 10 to 11 per cent.
Hence, considering all the different sources, a balanced picture is
of falls somewhere in the region of 8 per cent over the five years since
the onset of the financial crisis, or around 2000 [pounds sterling]
lower for the typical worker.
The final column of table 1 shows the latest data available for the
different real wage series. The ASHE data show real wages to be broadly
flat in the year to April 2013. The Labour Force Survey, which is
released quarterly, suggests that real wages continued falling in the
year to quarter 4 of 2013 (we average quarters 3 and 4 to improve
reliability). The ONS Average Weekly Earnings series can be followed up
until the three months to December 2013 and also shows continued real
wage falls. Thus the alternative series to ASHE suggest that real wages
continued to fall in 2013, by something around 1 per cent or 230 [pounds
sterling] per year for median workers). The different series tend to
show similar trends over extended periods but can differ over 1-year
comparisons. Thus the latest data would suggest, on balance, that the
fall in real wages has slowed in 2013 but not yet stopped.
3. The extent of falling real wages across major groups and
individuals
Figure 1 showed that the recent real wage falls have occurred
across the distribution. This is in direct contrast to previous
recessions where, if they occurred at all, they were confined only to
the lowest paid. We therefore turn next to exploring the variation in
the extent of falls across major groups and also to explore movements
for individual workers, exploiting the ASHE panel.
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
Differences across major groups of workers
The picture so far has explored average wage movements among those
in employment in any period. Figures 3 and 4 therefore start to explore
the data to look for variations across demographic groups. The figures
start in 1997 in order to show weighted series and figure 3 starts by
showing median real wage growth for men and women separately. Median
wages for women grew somewhat faster than for men prior to 2009. Indeed
male real wages showed a slight fall in 2008 ahead of women. The fall
from peak levels of real wages is substantially smaller for women at 7.5
per cent since 2009, compared with 10.7 per cent for men since 2008.
Figure 4 shows real wage movements for 18 to 24 year olds, 25-29,
30-34, 35-49, 50-59 and those aged 60+. The older age groups have seen
relatively more modest falls in real wages, in the region of 8 per cent
from peak levels. The falls for the younger age groups amount to real
falls of 11 to 14 per cent. For workers aged 18-25 the fall in real
wages in the recent period has been so extreme that, in real terms,
wages are back to levels not seen since 1998. As these are weekly wages
they will be affected by movements in part-time working, perhaps
connected to increased educational participation. In addition they are
affected by compositional shifts associated with sharp falls in youth
employment in this period. Yet for the slightly older group, those aged
25 to 29, who saw relatively little change in employment and education
participation levels, real wage falls have been around 12 per cent and
have taken wages back to the level last seen fourteen years ago in 1999.
By contrast, for older workers the falls represent a return to wage
levels seen around 2004.
Real wage changes at individual level
We next turn to look at wage movements at the level of the
individual by exploiting the panel aspect of the ASHE data. Figure 5
shows the extent of variation in individual real wages over the period
of the major earnings falls, 2009 to 2012, and for the prior four-year
period, 2005 to 2008.
Our earlier analysis of ASHE showed an 8 per cent fall in median
weekly wages for the workforce as a whole, but when looking at
individual pay growth (as we do in figure 5), we need to study a sample
of people who are in employment in both periods. Thus, young workers who
enter the labour market after 2009 are not included, making those in
employment in both periods older in 2012 than the workforce as a whole.
Likewise workers moving into the workforce by 2012 or leaving after 2009
are not included and these are generally lower paid individuals. It is
normal that as people age we see earnings grow, especially for those
aged under 40 and those retiring at the end of their working lives, who
earn far more than those just entering after leaving education. By
looking at those in employment in both 2009 and 2012 we are thus
focusing on those on the upward part of the wage trajectory and not the
wages of those entering and exiting the workforce in these years. Hence
it comes as no surprise that the real wage falls for those in employment
in both years are smaller than for the workforce as a whole, with the
median wage falling by around 4.5 per cent.
[FIGURE 5 OMITTED]
The figure shows movements between 2005 and 2008 (left-hand side)
and 2009 to 2012 (right-hand side) for comparison. The sharp shift into
negative territory is marked with a clear peak in the more recent period
at around -5 per cent. There is also a very strong spike at -11 per cent
which reflects three years of wage freezes, given CPI inflation. Despite
this positive selection in terms of workers employed both in 2009 and
2012, figure 5 shows that just under 30 per cent of workers experienced
real wage falls of more than 10 per cent, and 15 per cent of workers had
falls in excess of 20 per cent. There is, however, considerable
heterogeneity in the nature of real wage growth. In terms of real
increases, about 40 per cent of workers had real wage gains and 25 per
cent experienced real wage gains in excess of 10 per cent. Between 2005
and 2008, median real wage growth for those in employment in both years
was just under 7 per cent, again showing the effects of positive
selection. So the turnaround in individual wage growth across the two
periods is very large. In the earlier period just 30 per cent
experienced real wage falls and for around 15 per cent these falls were
more than 10 per cent. This picture is essentially the same if we focus
on those who are employed with the same firm over the three-year window.
There has long been discussion of nominal rigidities in pay
adjustment. It is commonly believed that real wages can fall due to high
price inflation, but it is widely thought that nominal wages do not fall
(see, inter alia, Bewley, 1999). Given the recent unprecedented real
wage falls of this magnitude over the short window we have documented
(even for workers staying in the same job), it seems quite plausible
that this stylised fact about the operation of labour markets is being
put to the test in the recent UK experience. Indeed, looking at
individual changes in weekly wages each year from 2009 to 2012, we see
that around 20 per cent of the workforce had nominal wage freezes (i.e.
their weekly earnings were exactly the same one year on, before
factoring in inflation). Furthermore a broadly constant fraction--of
around 20 per cent of workers --had nominal wage falls in excess of 1
per cent (i.e. they earned less one year on than they had started with
before taking inflation into account). Of course this may reflect a
number of changes in pay, such as overtime working, bonuses and shorter
hours of work. If we focus on weekly basic pay, excluding overtime and
bonuses, the fraction seeing nominal wage falls is around 17 per cent in
each year.
[FIGURE 6 OMITTED]
Taking this further, figure 6 shows the distribution of wage growth
for each of the three years for hourly basic pay, thus excluding
overtime and bonuses, for workers employed by the same employer at the
beginning and end of each annual period. This is the tightest measure of
earnings available as it excludes variations in pay due to more volatile
components of earnings (e.g. overtime and bonuses), hours of work or job
changes. Falls in this measure really do represent nominal wage cuts.
Again a sizable group, amounting to somewhat over 20 per cent of the
workforce, experienced nominal wage freezes (defined as wage growth of
between -0.5 and +0.5 per cent). The size of this group actually
increases between 2009-10 and 2011-12, reaching almost 25 per cent in
the year to April 2012. A further 14 per cent see nominal hourly basic
wage falls of more than 0.5 per cent. This is slightly larger in 2009 to
2010 than in the other years, but they are broadly similar. Some 8 to 10
per cent in each year experience nominal wage cuts of 5 per cent or
more.
If we consider workers who are employed by the same employer in all
three years (about 80 per cent of the sample employed in any one of the
three years), we find that 30 per cent have had a nominal wage cut in at
least one of the three years. Note this is for a measure of pay that
strips out changes in hours, overtime and bonus payments etc. A smaller
group experienced nominal wage cuts more than once. Further, some 20 per
cent had a nominal wage cut of at least 5 per cent in one of the three
years. Hence, there is clear evidence that whilst wage freezes are by
far the most common story explaining real wage falls, nearly one third
of workers employed in the same job for three years saw a nominal wage
cut in basic hourly pay in at least one year. Thus it appears that
nominal downward wage rigidities are breaking down in this period of
unprecedented cuts in real wages.
4. Why have real wages fallen?
As we have already stated, the recent falls experienced in real
wages are highly unusual for the UK, even through periods of economic
crisis. Hence the next obvious question is what we can say about why
these falls have occurred. Since as far back as Victorian times, real
wage growth has been broadly in line with productivity growth. As we
make goods and services of higher value in every hour we work, the more
pay levels can rise. Rising wages allow for greater consumption of goods
produced, thus boosting demand. Further, rising wages encourage firms to
boost productivity by investing in labour-saving technology, creating a
virtuous circle.
In addition to wage-productivity links, it has long been
established that the rise and fall of unemployment across the economic
cycle also influences pay movements. Hence to understand the recent real
wage falls we might naturally explore how sensitive wages are to
movements in unemployment, around underlying trend wage growth, and the
relationship between productivity and this underlying level of trend
wage growth.
We consider three (not necessarily mutually exclusive) explanations
of what have been important drivers of the unprecedented real wage
falls. These are:
i) downward pressure on wages from unemployment;
ii) the poor productivity record through the recession and
recovery;
iii) the distribution, across major groups of workers, of the
returns available from productivity gains made in the economy.
Real wages and unemployment
Table 2 builds on earlier research we undertook to explore how
wages have become more sensitive to unemployment in the last decade or
so (see Gregg et al., 2014). It presents 'wage curves' based
on the ASHE data discussed earlier and its regional panel of real median
wages (CPI deflated) and unemployment levels. (7) Panel A estimates the
sensitivity of wages to unemployment around an underlying trend rate of
growth in real wages. The unemployment effect here is thus driven by the
aggregate economic cycle, through the rise and fall in unemployment, and
regional movements around the UK wide average. Panel B introduces year
dummies instead of a trend in wage growth. This nets out the aggregate
cycle and the estimates are derived only from regional variation in
wages and unemployment movements, not the aggregate economic cycle.
We estimated the model over two periods--before 2003 and from 2003
onwards--to explore the extent of underlying wage growth given by the
time trend in Panel A and the sensitivity of wages to movements in
unemployment. The estimation over separate periods allows us to explore
whether both underlying wage trends and the sensitivity to unemployment
have been operating differently over the past decade than previously.
Column 1 of Panel A suggests that prior to 2003 underlying median real
wage growth was 0.8 per cent per annum, and that this fell to 0.6 per
cent per annum in the second time period (see column 2). Thus a large
part of the apparent slowdown disappears when we allow for the
sensitivity of wages to unemployment and for this to change across the
two periods. So, in the later period, unemployment applies a larger
downward pressure on wages (the coefficient on the lagged unemployment
rate going from -0.066 in the first period, to -0.137 in the second, or
a statistically significant drop of -0.071. According to this estimate,
the magnitude is such that a doubling of unemployment from say 4.5 to 9
per cent, slightly more than actually occurred in the downturn, would
lower real wages by 13.7 per cent based on estimates for the recent
period, compared to a little under 7 per cent before 2003. (8)
Such estimates, being based on a short period with just one cycle,
may thus reflect only the correlation in timing between the recent
downturn and the unusual fall in wages, rather than a deeper shift in
the sensitivity of wages to unemployment. Thus the lower panel includes
time dummies for each year rather than trend terms. These will pick up
the economy-wide shift in unemployment and wages in any year and the
unemployment terms now only reflect deviations in unemployment and wages
at the regional level. Thus the estimates are based on regional
variations net of the macroeconomy depending on whether regions with
larger or smaller shifts in unemployment within any year experience
smaller or larger movements in real wages in the following year. As the
results rest only on regional deviations in unemployment and some
regions are quite small, there will be some noise in the data induced by
variation in sampling that occurs in any survey. Hence the magnitudes of
the unemployment effects are smaller, but the key point is that we see
the same increase in the sensitivity of wages to unemployment (the
elasticity changing by a statistically significant -0.036 between the
two time periods). Such a model is asking quite a lot of the data and
that it is robust to being based on regional variations is very
reassuring. Thus we can be confident that through the recent recession
real wages have fallen because of an underlying slowdown in real wage
growth combined with an increased sensitivity to unemployment movements
that have applied greater downward pressure than seen in the past.
Wages and productivity
Next we turn to what drives the underlying trend growth in wages
seen above. The relationship between wages and productivity growth has
been long established. Figure 7 shows the trends in productivity per
hour worked and total compensation per hour. The latter is a broad
measure of total labour costs to firms and includes employer NI and
pension contributions. This figure is an updated version of that
presented in Pessoa and Van Reenen (2013). The figure is expressed in
real terms using the GDP deflator and thus reflects real product wages,
as this is the most appropriate representation for productivity (though,
as figure 2 showed, this difference between producer and consumer prices
over the whole period is small). It shows that total labour costs have
continued to grow in line with productivity. It is thus clear that a
part of the pattern of the slowdown in underlying real wage growth
connects to the poor productivity performance of the UK economy during
the downturn. The poor productivity record reflects the low levels in
business investment seen in this period, which fell by 14 per cent in
the recession and have yet to recover, and the suppression effect of
unemployment on wages noted above. Thus firms are found to be employing
extra workers, as workers priced themselves into work, as their relative
prices made them more attractive than investment, thus preserving
employment and at the same time inducing the stagnation in productivity.
[FIGURE 7 OMITTED]
The distribution of productivity gains
Figure 7 shows no sign of a decoupling of productivity and wages if
wage costs are measured as total compensation. However, figure 8 shows
two additional measures of labour costs based on wages and not on total
compensation. The first is mean wages per hour. This differs from total
compensation in that it does not include employer non-wage labour costs,
such as pension contributions, but only wages received by workers. The
second is median instead of mean wages per hour.
[FIGURE 8 OMITTED]
The point that emerges very clearly is that, over the past decade,
from around 2003, average wages started to grow more slowly than
productivity. The gap between average wages and total compensation per
hour suggests that non-wage labour costs, mostly pensions, took a
growing share of the productivity growth achieved.
Two further points need to be made here. First, these pension costs
are not just those for current workers but all contributions to pension
funds to meet the costs of defined benefit (DB) pension schemes run by
firms. So a portion of the non-wage labour costs are meeting pensions of
already retired workers. Indeed many such DB schemes are now closed to
new workers (Pension Protection Fund, 2013) so current workers will not
see such good deals in the future. Greater longevity and poor stock
market performance, compared with what was expected when such schemes
were set up in the 1950s to 1970s, means that these funds required
higher employer contributions to avoid building up considerable
deficits. The other point to note is that low wage workers often do not
receive the same generosity of pension deals as higher paid workers;
inequalities here are larger than in wages as a whole. Minhat (2008)
suggests that pension contributions made by firms on behalf of senior
executives in the UK represent around 15 per cent of their total
compensation, a figure supported by a recent consultancy report (Lane,
Clarke and Peacock Consulting, 2014). This is far higher than for
typical workers. For instance, the new auto-enrolment system for
uncovered workers--NEST (National Employment Savings Trust)--that is
gradually being introduced in the UK, has employer contributions of just
3 per cent, and even less before the scheme is fully functioning in
2018.
Importantly, figure 8 shows that median wages per hour fell behind
productivity growth far earlier than did the mean, beginning from around
the mid-1990s. Moreover, the gap opened up much faster soon after the
turn of the millennium. The opening of the gap between mean and median
wages is because of rising wage inequality. The faster wage growth of
top earners pulled the average wage up at a faster rate than the median.
Of course, rising wage inequality started before 2000 (see Machin,
2011). Prior to that higher wage growth for high paid workers was
matched by stagnation for the lower paid and so from 1979 to the mid--to
late-1990s, median workers saw pay grow broadly in line with
productivity. Since then the lower paid have matched the middle,
probably due to the National Minimum Wage, but higher pay for top
earners has pulled the average up. Hence it seems that median wages have
become de-coupled from productivity growth, because of rising
inequality, showing that a growing share of the value from productivity
growth has been absorbed by pensions and higher salaries for top
earners.
5. What scope is there for renewed wage growth?
Pulling the different pieces of evidence together, we can start a
discussion of what it might take for broad-based real wage growth to
re-emerge. The evidence of table 2 shows that falling unemployment will
lead to higher wage growth for a period. The evidence presented suggests
that a fall from 8.5 per cent unemployment the peak in this cycle--to
say 5 per cent--the typical level from 1998 to 2008--will result in real
wages rising by about 7 per cent.
The estimated wage curve uses a very simple dynamic structure, (9)
but suggests that wages respond to falling unemployment with a lag of a
year or so. So unemployment peaked right at the end of 2011 (see figure
9) and then edged down until the Summer of 2013, which would suggest
real wages stabilising in 2013. But according to the lag structure, real
wages will not respond to the strong falls in unemployment since the
summer of 2013 till the middle or latter half of 2014. However, the
model suggests that once unemployment has stopped falling for a year or
so then the rise in real wages will return to the underlying growth rate
of just above zero. So the wage recovery will only last till a year or
so after unemployment has stopped falling. Introducing more complex
dynamics suggests that half the effect of falling unemployment is felt
in the second year after the fall occurs, with the rest of the effect
coming at a diminishing rate in the following three years. Thus if
unemployment is on a sustained downward trend, the biggest effects on
wages will be felt in years 3 and 4 after unemployment starts a period
of steady decline, i.e. from the Summer of 2016 if the current downward
trend in unemployment continues. Hence we should expect real wages to
start rising in the second half of 2014 and for growth to get stronger
until a year or two after the fall in unemployment slows significantly.
[FIGURE 9 OMITTED]
The second ingredient for a broad-based wage recovery and the
essential part of a sustained wage recovery is productivity growth. The
weak output performance and strong employment picture mean that
productivity is still 4.5 per cent below that seen at the beginning of
the crash. Such a prolonged period of declining productivity (with the
trough in quarter 4 of 2012) has never been seen before in modern UK
history. Whilst growth has returned, productivity in the past year has
remained relatively flat with a small 0.3 per cent fall in 2013 compared
to 2012.
It is essential that we see a return to the levels of 1.5 to 2 per
cent annual productivity growth seen in the decade before the crash if
real wage growth is to return on a sustained basis. Part of the
extremely poor productivity record reflects low wages, as firms have
weak incentives to invest in labour saving technology when workers are
easy to hire and cheap. So falling unemployment should kickstart
investment; as labour becomes scarcer and real wages stop falling
investment should return and with it productivity. If this does not
occur, there can be little hope of a sustained wage recovery.
Moreover, this is a necessary but not a sufficient condition for a
broad and sustained wage recovery. Figure 8 shows worrying trends in
this regard. If some of the patterns documented there continue, then the
combination of rising total employer pension contributions and growing
wage inequality seen over the last decade before the crash could
continue to extract all the growth in the size of the pie, leaving
little or nothing extra for typical workers. Therefore, policy aiming to
boost wages (outside of the lowest paid where minimum wages have an
effect) needs to focus on boosting productivity, producing sustained
increases in revenues of company pension schemes (e.g. through stock
market returns and real interest rates of company and government bonds)
and addressing the distribution of wage growth not only towards the top
1 per cent of employees (see Bell and van Reenen, 2014, for a good
discussion of this issue). Generous employer pension contributions for
top executives running at around 15 per cent of salary represent a nexus
of the issues of pay inequality and pensions absorbing the bulk of the
gains from productivity.
6. Conclusions
The UK has been experiencing unprecedented falls in real wages and
living standards. In this paper we document the nature of these falls,
discuss reasons why real wages have fallen, and offer a discussion of
what might bring back a return to real wage growth.
In terms of documenting the scale of real wage falls, there are a
number of alternative measures of wages available and indeed measures of
inflation. Taking a broad view of this evidence suggests that since 2008
real wages have fallen by around 8 per cent (with different measures and
sources showing falls in the range of 4 to 11 per cent). This equates to
a fall of around 2000 [pounds sterling] for the typical (median) British
worker. Real wages falls have been widespread and have occurred right
across the wage distribution. Moreover, the broad picture from the
available data suggests that real wages continued to fall through 2013.
The real wages of some groups have been particularly hard hit, most
notably for the young. Those aged 25 to 29 have seen real falls of 12
per cent and those aged 18 to 24 over 14 per cent. Indeed, the fall for
those under 25 is so large it has taken real wages back to levels last
seen in 1998. In part this may reflect greater part-time working whilst
extending educational studies, but even among those aged 25 to 29 (i.e.
those past normal age for finishing education) real weekly wages have
fallen back to levels seen in 1999.
At the individual level there is considerable variation in size of
wage falls. Indeed among those who worked through the recession, in work
in both 2009 and 2012, who are a rather select group, some 60 per cent
experienced falls in real wages. Further, nearly a third saw real wage
falls in excess of 10 per cent and one in six experienced falls in
excess of 20 per cent. Moreover, there is also considerable evidence of
nominal wage cuts. Whilst the bulk of the observed real wage falls stem
from wage freezes combined with erosion by inflation, around one third
of workers who were employed by the same firm between 2009 and 2012
experienced a cut in nominal hourly basic pay (thus excluding overtime
and bonuses) in at least one year of the downturn, and for one in five
workers this cut exceeded 5 per cent of basic hourly pay.
The available evidence suggests these unprecedented real wage falls
are being driven by three factors. First, unemployment is exerting a far
larger downward pressure on wages than in previous recessions. This can
be seen even at the regional level when netting out the economy-wide
rise in unemployment and fall in wages, which strongly suggests it is
not just a coincidence of timing. Second, median wages have historically
tracked productivity growth in a mutually reinforcing relationship.
Productivity gains create room for rising real wages and rising real
wages create the incentive for firms to invest in labour saving
technologies. Low wages have in turn created incentives for firms to
meet demand by hiring more workers, rather than through investment,
hence a weak productivity record, though it has been good news for jobs,
has restricted the room for real wage rises. Third, and this pre-dates
the recession, wages of typical British workers are no longer keeping up
with productivity gains made in the economy. This stems from a growing
contribution of total compensation going towards supporting pensions,
not just for current but also for already retired workers, and that the
highest paid (top 1 or 2 per cent) are taking a disproportionate share
of the gains from productivity, leaving room for few gains by ordinary
workers.
These findings set the scene for a discussion of what conditions
could bring back real wage growth. The recent rapid fall in
unemployment, since the Summer of 2013, should be sufficient to generate
real wage rises in the second half of 2014, as there is a lag of around
a year before wages respond to changing labour market conditions.
Continued falls should lead to a continued wage recovery, but alone such
gains will stop around two to three years after unemployment stops
falling. For a sustained wage recovery the economy also needs to
generate a return to the levels of productivity growth normally seen,
but which have been notably absent over the past six years. As labour
becomes scarce and more expensive we should expect firms to increase
investment, generating productivity improvements. But even this will not
be enough for sustained real wage gains to come about unless the
distribution of the returns from productivity growth can be channelled
back to ordinary workers. This was the historical norm, but it stopped
in the early 2000s--importantly before the downturn--with a
disproportionate share of productivity gains going to support pension
commitments and rapidly rising salaries of very highly paid workers.
Table A1. price indices
Per annum Per annum Per annum
percentage percentage percentage
growth growth growth
1988-2003 2003-8 2008-13
CPI 2.87 2.31 3.08
RPI 3.66 3.50 2.96
GDP deflator 3.07 2.48 2.08
RPIJ 3.12 2.48
CPIH 3.04
Source: ONS published figures.
Notes: The Consumer Price Index (CPI) and the Retail Price Index
(RPI) are the two main measures of consumer prices produced by the
ONS. The RPI and CPI measures differ in the methodology used for
their calculation and also the CPI does not include housing costs
for owner ocupiers. The CPIH and the RPIJ measures aim at
overcoming these two differences between CPI and RPI: the CPIH
includes owner occupiers' housing costs and is available from 2005;
and the RPIJ covers the same basket of goods as the RPI using a
methodology similar to that of the CPI and is available from 1997.
The GDP deflator is the deflator for the Gross Domestic Product.
REFERENCES
Bell, B., Nickell, S. and Quintini, G. (2002), 'Wage
equations, wage curves and all that', Labour Economics, 9, pp.
341-60.
Bell, B. and Van Reenen, J. (2014), 'Bankers' and their
bonuses'. Economic Journal, 124, pp. F1-F21.
Bewley, T. (1999), Why Wages Don't Fall During a Recession,
Harvard University Press.
Blanchflower, D. and Oswald, A. (1994), The Wage Curve, MIT Press.
--(1995), 'An introduction to the wage curve', Journal of
Economic Perspectives, 9, pp. 153-67.
Blundell, R., Crawford, C. and Jin, W. (2014), 'Employment
composition, hourly wages and the productivity puzzle', Economic
Journal, forthcoming.
Gregg, P., Machin, S. and Fernindez-Salgado, M. (2014), 'Real
wages and unemployment in the Big Squeeze', Economic Journal,
forthcoming.
Lane, Clarke and Peacock Consulting (2014), LCP FTSE 100 Executive
Pension Survey 2013.
Machin, S. (201 I), 'Changes in UK wage inequality over the
last forty years', in Gregg, P. and Wadsworth, J. (eds), The Labour
Market in Winter, Oxford University Press.
Machin, S., Murani, S. and Van Reenen, J. (2014), 'Inequality
trends and assortative mating', Centre for Economic Performance,
London School of Economics, mimeo.
Minhat, M. (2008), 'The role of compensation consultants in
executive pay governance', British Accounting Association
conference, April.
Office for National Statistics (2012), Consumer Price Indices
Technical Manual, 2012 Edition,
Pension Protection Fund (2013), The Purple Book: DB Pensions
Universe Risk Profile.
Pessoa. J. and Van Reenen, J. (2013), 'Wage growth and
productivity growth: the myth and reality of 'decoupling",
Centre for Economic Performance Discussion Paper 1246.
--(2014), The UK productivity and jobs mystery: does the answer lie
in labour market flexibility?', Economic Journal, forthcoming.
Taylor, C., Jowett, A. and Hardie, M. (2014), An Examination of
Falling Real Wages, 2010-2013, ONS.
NOTES
(1) Our analysis uses ASHE microdata up to 2012 (the last year
available in the Secure Data Service when we wrote this article) but
where possible also uses the ONS published statistics to update to 2013.
Thus some numbers in the figures and tables end in 2012 and others in
2013.
(2) The weighting only starts in 1997 and shows somewhat faster
earnings growth than is seen in the raw unweighted data. It therefore
probably paints what one might think of as the most optimistic picture
on patterns of real wage growth.
(3) Appendix table AI shows annual price inflation for these two
measures in the three time periods we focus upon, together with the
retail price index (RPI), and two more recently available inflation
indexes (the RPIJ, which covers the same basket of goods as the RPI
using a methodology similar to that of the CPI, and CPIH, which adds
owner occupiers' housing costs to the CPI).
(4) For the longer-term evolution of real wages deflated by the
retail price index (data on which goes back further in time) see the
Figures in Gregg et al. (2014). The overall nature of change (of rising
real wages, followed by a slowdown and then sharp falls is also shown
there. Reduced wage growth in the slowdown period is more marked for the
RPI deflated wages, whilst the 2008-13 falls are similar.
(5) The observed pattern of rising wage inequality in the UK over
the past forty years has been well documented: see Machin (2011) for
more detail.
(6) More precisely, it uses the General Household Survey before
1993 as the LFS did not include wages prior to then (see Machin, Murani
and Van Reenen, 2014, for more details).
(7) See Blanchflower and Oswald (1994, 1995) and Bell et al. (2002)
on wage curves.
(8) Estimating the trend specification for real wages deflated by
the GDP deflator rather than the CPI produced very similar results. The
real wage-unemployment elasticity changes by -0.074 (as compared to the
CPI based change of -0.071 in table 2). Of course, the more general year
dummies specification is identical to that in Panel 2 of table 2 as it
nets out common macroeconomic differences (as the inflation measures are
at the aggregate level) through inclusion of the year dummies.
(9) For more variations, and a number of robustness checks, see
Gregg et al. (2014).
Paul Gregg *, Stephen Machin **, and Marina Fernandez-Salgado *
* Department of Social and Policy Sciences, University of Bath. **
Department of Economics, University College London and Centre for
Economic Performance, London School of Economics. E-mail:
S.Machin@ucl.ac.uk. This work is partly based on data from the New
Earnings Survey/Annual Survey of Hours and Earnings, produced by the
Office for National Statistics (ONS) and supplied by the Secure Data
Service at the UK Data Archive. The data are Crown Copyright and
reproduced with the permission of the controller of HMSO and
Queen's Printer for Scotland. The use of the data in this work does
not imply the endorsement of ONS or the Secure Data Service at the UK
Data Archive in relation to the interpretation or analysis of the data.
This work uses research datasets which may not exactly reproduce
National Statistics aggregates. We would like to thank two anonymous
referees for their helpful comments on an earlier draft.
Table 1. The rise and fall of real wages, 1988-2013
Level Real per
April (or annum
Q2) growth
2013 1988-2003
(CPI)
Average weekly earnings
ONS average weekly earnings (a) 484.00 2.28
ASHE average weekly earnings 511.62 2.26
LFS average weekly earnings (b) 472.04 1.98
ASHE average weekly basic
earnings 485.10 2.57
Median weekly earnings
ASHE median weekly earnings 424.64 1.84
LFS median weekly earnings 392.00 1.78
ASHE median weekly basic
earnings (c) 396.81 2.34
Average hourly earnings
ASHE average hourly earnings 14.90 2.67
LFS average hourly earnings 13.39 1.83
Median hourly earnings
ASHE median hourly earnings 11.70 2.16
LFS median hourly earnings 10.73 1.59
Annual earnings
ASHE average annual earnings(d) 28230.96
ASHE median annual earnings 22790.15
Real per Real per
annum annum
growth growth
2003-8 2008-13
(CPI) (CPI)
Average weekly earnings
ONS average weekly earnings (a) 2.03 -1.34
ASHE average weekly earnings 1.68 -1.70
LFS average weekly earnings (b) 1.50 -1.16
ASHE average weekly basic
earnings 1.53 -1.44
Median weekly earnings
ASHE median weekly earnings 1.53 -1.60
LFS median weekly earnings 0.81 -0.74
ASHE median weekly basic
earnings (c) 1.47 -1.40
Average hourly earnings
ASHE average hourly earnings 1.86 -1.22
LFS average hourly earnings 1.63 -0.66
Median hourly earnings
ASHE median hourly earnings 1.87 -1.20
LFS median hourly earnings 1.75 -0.80
Annual earnings
ASHE average annual earnings(d) 1.96 -2.16
ASHE median annual earnings 1.39 -1.90
Latest
annual
growth
Average weekly earnings
ONS average weekly earnings (a) -0.90 (3 months to Dec)
ASHE average weekly earnings -0.24 (April)
LFS average weekly earnings (b) -1.44 (Q3 & Q4)
ASHE average weekly basic
earnings -0.24 (April)
Median weekly earnings
ASHE median weekly earnings 0.19 (April)
LFS median weekly earnings -2.10 (Q3 & Q4)
ASHE median weekly basic
earnings (c) 0.16 (April)
Average hourly earnings
ASHE average hourly earnings -0.34 (April)
LFS average hourly earnings -1.02 (Q3 & Q4)
Median hourly earnings
ASHE median hourly earnings 0.56 (April)
LFS median hourly earnings -1.82 (Q3 & Q4)
Annual earnings
ASHE average annual earnings(d) -0.82
ASHE median annual earnings -0.53
Notes: (a) The figures on ONS Average Weekly Eearnings (AWE) from
years 1988-99 are ONS backdated estimates using information on the
Average Earnings Index (AEI) as the ONS Average Weekly Earnings series
is only available from January 2000. The estimated series is
considered to be comparable to the published AWE. Yearly changes refer
to the month of April for columns 2 and 3 and to the average of March-
May in column 4. The latest Annual growth refers to average wages for
October-December of 2012 to October-December of 2013. (b) The level
for LFS refers to the 2nd quarter of 2013 (April-June). LFS wages are
weighted averages constructed using waves I and 5 and refer to the 2nd
quarter (April-June) of every year. We use wave 5 from 1993-1996, as
that is the only wave when wages are reported. Prior to 1993 the LFS
did not include wages so we used the General Household Survey from
1988-1992. The latest Annual growth refers to wages for the 3rd and
4th quarter of 2013. (c) ASHE Basic Earnings includes other payments.
It refers to Real per annum growth from 2005-8' due to a change in
definition in 2005. (d) Annual earnings are only available in ASHE
from 1996 and only represent employees in the same job for more than a
year.
Table 2. Regional median real weekly wages and
unemployment (CPI), 1988-2012
1988-2002 2003-12 Change
between
1988-2002
and 2003-12
A. Trend specification
[DELTA]Log (Unemployment -0.013 -0.012 0.001
Rate[t]) (0.010) (0.019) (0.021)
Log (Unemployment -0.066 -0.137 -0.071
Rate [t-I]) (0.009) (0.021) (0.023)
Trend 0.008 0.006 -0.001
(0.002) (0.002) (0.003)
Region dummies Yes Yes Yes
Regional controls Yes Yes Yes
R-squared 0.988 0.965 0.986
Sample size 165 1 10 275
B. Year dummies specification
[DELTA]Log (Unemployment -0.018 -0.011 0.007
Rate[t]) (0.011) (0.013) (0.016)
Log (Unemployment -0.022 -0.058 -0.036
Rate[t-I]) (0.010) (0.012) (0.016)
Region dummies Yes Yes Yes
Year dummies Yes Yes Yes
Regional controls Yes Yes Yes
R-squared 0.994 0.995 0.996
Sample size 165 110 275
Notes: Robust standard errors in parentheses. The time varying
regional controls are from the Labour Force Survey and are the
proportion with a degree, female, young and white in the regional
workforce.
