Has the UK experienced a lost decade? Over the period 2007-18, both real wages and productivity (measured by output per worker per year) in the UK stagnated, as shown in Figure 1.
Figure 1 UK log real wages and log output per worker per year since 2000 (matched by means)
It is no surprise that real wages have tracked productivity this millennium, as productivity drives real wages. Indeed, the two series have moved closely together since 1860, as can be seen in Figure 2 (Castle and Hendry 2014a).1 The anomalous ‘flat-lining’ evident since 2007 is unprecedented in almost 160 years. Using the model in Castle and Hendry (2014b), there is no evidence that the relationship between real wages and productivity and unemployment shifted after 2007. Indeed, the relationship between the variables is remarkably stable over more than 150 years. However, productivity has manifested many trend shifts, the most recent of which is the negligible growth since 2007.2 Thus, stagnant real wages since 2007 can be explained by a lack of productivity growth (just 0.8% in the decade). But why did productivity suffer such a prolonged slowdown?
One worry might be that changes associated with the rapid reduction of 45% in the UK’s CO2 emissions from 2008 to 2018 impacted GDP growth, and our research on climate econometrics prompted us to search for an explanation.
Figure 2 UK real log wages and log output per worker since 1860 (matched by means)
Figure 3 shows the wage share, which calculates real wages relative to productivity and is the component of national income allocated to wages. The wage share increased considerably from 1995 to 2002 but has since dropped back close to the initial level plotted. This reflects Figure 1 but highlights that real wages did not even keep up with the small increases in productivity. So, workers did not reap even the slight rewards from limited productivity growth since 2009.
Figure 3 Log wage share since 1995
All is not lost
Over the same time period as the stagnation in real wages, unemployment fell to record lows, as seen in Figure 4. There was a steep rise to almost 8% in 2010, but unemployment has since fallen considerably to around 4%. Such behaviour is consistent with past responses of the unemployment rate to events such as recessions. The remarkable constancy of the relationship, despite the sharp fall of 4.5% in GDP in the Great Recession, is shown by the one-step ex post ‘forecasts’, based on the model in Hendry (2001) (also plotted in Figure 4). While the fall in GDP might have been expected to lead to a much sharper rise in unemployment, the considerable cuts to interest rates from approximately 5% to near 0%, with price inflation remaining about 2%, produced negative real rates that acted as a partial offset.
Given the lowest levels of the unemployment rate since the 1970s, the increased bargaining power of labour might have been anticipated to raise real wages relative to productivity. Such an effect is absent, possibly because of a large rise in people willing to work at the going wage. Ostensibly, this is a joyless recovery for workers, albeit with some evidence of ‘catch-up’ at the very end of the period, as the wage share picks up in 2018.
Figure 4 Unemployment rate with model-based one-step ex-post ‘forecasts’ since 2005
Nevertheless, employment has risen by about 4.6 million since 2000, as seen by the dashed blue line in Figure 5. That increase is around 18% of the previous labour force. The present level of employment is the highest ever for the UK and has grown much faster than the population over the last quarter of a century.
Figure 5 Logs of employment and population (matched in 1995)
With so many more workers and real wages relatively constant, total earnings must have increased considerably, and indeed they have risen by almost 35% this millennium and 15% since 2008. In fact, the real GDP per capita of the population has increased by almost 22% since 2000 and 10% since the trough in 2009, as shown by the solid red line in Figure 6 – far from joyless, even if the growth was not evenly shared.
Figure 6 Log real GDP per capita and log real GDP per employee (matched in 1995)
Lost but now found?
The explanation for the dramatic difference from stagnant productivity growth lies in this remarkably different behaviour of aggregate employment and population, seen in Figure 5. We have plotted these time series from 1995, as that was the start of the departure between them. The ratio of employment to population has been mostly rising since the 1980s and is currently at the highest ever peacetime level.
The faster growth of employment relative to the population has led to a divergence between real GDP per capita and real GDP per employee, as in Figure 6. For the latter, total output is spread over a larger number of employees, leading to the stagnant productivity observed in Figure 1. For the former, the steady rate of population growth of about 0.6-0.7% per annum means that total output has expanded faster than the population, so average living standards are more than 20% higher this millennium even over a period that included the Great Recession. The lack of growth in output per worker due to the increase in employment probably reflects a lower marginal product of new employees, which has offset any increases in the productivity of existing workers – the ‘gig economy’ almost certainly played a role.
The substantial rise in real GDP and real GDP per capita also shows that the rapid reduction in the UK’s CO2 emissions from 529 million tonnes (MT) per annum in 2008 to 361 MT in 2018 (a 45% fall) has not greatly slowed GDP growth, and is almost certainly not the cause of the stagnation of real wages. Indeed, the discrepancy between the growth of GDP per capita in our research on climate econometrics (e.g. Hendry 2018) and the stagnation of real wages in our macro-econometrics research prompted this search for an explanation, as offered here.
The puzzle is not fully resolved
The one remaining puzzle is why employment expanded so quickly at a relatively constant real wage. There could have been many forces operating. On the demand side, there was a significant fall in investment following the financial crisis and exacerbated by Brexit uncertainty, which could have led to firms substituting from physical capital to cheap labour. On the supply side, an increase in the number of employees per household could be trying to compensate for low wages for individual incomes, for austerity, or welfare reform. There are doubtless numerous other potential explanations for this trend.
The period since 2007 has seen little improvement in either real wages per employee or productivity, with real wages in 2017 actually being lower than a decade earlier. Nevertheless, total real wages have risen by 12% over that time. The differences between these measures derives from the far larger increase in employment than in population. Not all ‘living standards’ have stagnated.
Authors’ note: The views expressed herein are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Treasury Department or of the US government.
Castle, J L, and D F Hendry (2014a), “The real wage–productivity nexus”, VoxEU.org, 13 January.
Castle, J L, and D F Hendry (2014b), “Semi-automatic non-linear model selection”, in N Haldrup, M Meitz, and P Saikkonen (eds.), Essays in Nonlinear Time Series Econometrics, Oxford University Press.
Castle, J L, J A Doornik, D F Hendry, and F Pretis (2019), “Trend-indicator Saturation”, Working Paper, Nuffield College, Oxford University.
Doornik, J A (2009), “Autometrics”, in J L Castle and N Shephard (eds), The Methodology and Practice of Econometrics, Oxford University Press, 88–121.
Hendry, D F (2001), “Modelling UK inflation, 1875–1991”, Journal of Applied Econometrics 16: 255– 275.
Hendry, D F (2018), “First-in, first-out: Driving the UK's per capita carbon dioxide emissions below 1860 levels”, VoxEU.org, 12 December.
Pretis, F, J-J Reade, and G Sucarrat (2018), “Automated General-to-Specific (GETS) Regression Modeling and Indicator Saturation for Outliers and Structural Breaks”, Journal of Statistical Software 68: 4.
Walker, A, F Pretis, A Powell-Smith, and B Goldacre (2019), “Variation in Responsiveness to Warranted Behaviour Change among NHS Clinicians: a Novel Implementation of Change-Detection Methods in Longitudinal Prescribing Data”, British Medical Journal 367: l5205.
 We use the GDP deflator for calculating both real wages and constant price output, although UK price inflation has been low on most measures since 2000.
 We use Trend Indicator Saturation (TIS) to detect trend shifts, see Castle et al. (2019), and Walker et al. (2019). This method detected 12 trend shifts over the period 1860-2018. TIS is available in Autometrics (Doornik 2009), gets (Pretis et al. 2018) and recently as XLModeler, an Excel Addin: https://xlmodeler.com/.