This chapter examines the evolution of labour market performance since the onset of the global financial crisis. OECD labour markets are back to pre‑crisis levels in terms of job quantity, with only few notable exceptions, while a more mixed picture emerges as regards job quality and inclusiveness. In spite of this, nominal wage growth remains remarkably lower than it was before the crisis for comparable levels of unemployment, and the shift of the relationship between unemployment and wage growth has continued during the recovery. The chapter investigates the factors accounting for the persistent wage growth slowdown. While low inflation expectations and productivity growth deceleration remain the main drivers of observed patterns, the dynamics of low‑pay jobs and the wages associated to them have also been key factors accounting for the overall decline in wage growth.
OECD Employment Outlook 2018
Chapter 1. Still out of pocket: Recent labour market performance and wage developments
Abstract
The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements in the West Bank under the terms of international law.
Key findings
The recovery from the global financial crisis and the subsequent European debt crisis that affected a number of euro area countries is largely complete. At 2.6 % per year in 2017 and 2.5% projected for 2019, OECD economic growth, while not at a record high, appears stable and even the euro area is experiencing the strongest growth of real gross domestic product (GDP) of the past ten years (OECD, 2018[1]). Employment rates are, on average, above pre‑crisis levels, with the strongest improvements occurring among under‑represented groups. Yet wage growth appears to be lagging behind employment growth, with some signs of acceleration appearing in some countries only towards the end of 2017 or the first quarter of 2018 (OECD, 2018[2]). This soft wage growth suggests that the recovery remains fragile.
This chapter provides an overview of labour market developments since the onset of the global financial crisis with a special emphasis on the possible reasons for unexpectedly low wage growth. The main findings of this chapter are:
OECD labour markets are back to pre‑crisis levels in terms of job quantity, with only a few notable exceptions. Yet, a more mixed picture emerges as regards job quality and inclusiveness, the other two main pillars of the OECD Jobs Strategy together with job quantity. Improvements have occurred over the past decade in many countries as regards the gender gap in labour income, the labour market prospects of disadvantaged groups, and the incidence of job strain – excessive job demands combined with insufficient resources. However, labour market insecurity – the risk of unemployment and its economic cost for workers – is not yet back to pre‑crisis levels and poverty has grown amongst the working‑age population.
Wage growth also remains remarkably lower than it was before the crisis. The OECD average of hourly wage growth rates was between 1.5 and 2 percentage points lower during the Great Recession than it was before for comparable levels of unemployment, and this shift in the relationship between unemployment and wage growth (the so‑called Phillips curve) has continued during the recovery. It is visible even in countries where wage growth seems to be finally picking up a number of years into the recovery, such as the United States.
All in all, in OECD countries, nominal hourly wage growth dropped from 4.8% in the pre‑crisis period to 2.1% in recent years on average. Real wage growth decreased by 1 percentage point over the same period.
The low‑inflation environment and the productivity slowdown have both contributed to the marked deceleration in wage growth. On average, hourly labour productivity growth slowed from 2.3% prior to the crisis to 1.2% in the recent period, while inflation decreased from 2.6% to 0.8%, likely lowering inflation expectations.
The dynamics of low‑pay jobs and the wages they pay have also been key factors accounting for the overall decline in wage growth. In particular, there has been a significant worsening in the average earnings from part‑time jobs relative to that of full‑time jobs, which is associated with the rise of involuntary part‑time employment in a number of countries.
Comparatively poor working conditions among workers regaining employment after an unemployment spell, combined with a large number of transitions from unemployment to employment in some countries, pushed up the number of lower‑paid workers, thereby lowering average wage growth.
Introduction
This chapter provides an overview of labour market developments since the onset of the global financial crisis. After presenting the evolution of the key indicators of labour market performance, developed in the context of the OECD Jobs Strategy in OECD (2017[3]; 2018[4]), special attention is given to wage growth, which appears to be the missing element of the current recovery. Indeed, while unemployment has been on a declining path for a number of years in most OECD countries (OECD, 2016[5]), wage growth remains remarkably lower than it was before the Great Recession for comparable levels of unemployment. This recent downward shift of the wage‑unemployment relationship in a number of countries has raised an increasing interest and concern in the academia and policy fora – see for example (IMF, 2017[6]; Bulligan and Viviano, 2017[7]; OECD, 2016[5]; ECB, 2016[8]; Shambaugh et al., 2017[9]). Beyond the factors typically pointed out in the literature, such as the productivity slowdown and fall of inflation expectations, low‑pay jobs are considered here as an important channel accounting for the disappointing wage growth deceleration.
The remainder of the chapter is divided as follows: Section 1.1 briefly examines the evolution of labour market performance, using a number of standardised indicators; Section 1.2 investigates the statistical factors accounting for the persistent wage growth slowdown; and Section 1.3 presents concluding remarks.
Recent developments in key indicators of labour market performance
Labour market conditions continue to improve. In 2017, the OECD average employment rate was almost 2 percentage points above its pre‑crisis level, (Figure 1.1, Panel A).1 Similarly, unemployment rates continue their slow descent, although in a few countries remain somewhat above their pre‑crisis levels because employment has not increased enough to fully offset rising trends in participation rates (Figure 1.1 Panel B). Yet, in 2016, broad labour underutilisation – adding up inactive and unemployed people as well as involuntary part‑timers – was still, at 28.1%, 1.5 percentage points above 2006 levels (Figure 1.1, Panel C).
The recent performance of OECD countries as regards job quantity has been quite heterogeneous. In 2016, employment rates were more than 8 percentage points above their 2006 levels in Germany, Hungary and Poland. In these countries, these positive employment trends are typically matched by significant reductions in both unemployment and broad labour underutilisation. By contrast, contractions of employment rates as large as 2 percentage points or more occurred in this period in a number of countries hit hard by the Great Recession and the euro debt crisis (Greece, Ireland and Spain) and Denmark. In these countries, negative employment trends are matched by large increases in unemployment and broad underutilisation.
Figure 1.1. Employment performance is back to pre‑crisis levels
Note: Following OECD (2018[4]), broad labour underutilisation is defined in the chart as the sum of inactive, unemployed and involuntary part‑time people.
Source: OECD Employment Database, www.oecd.org/employment/emp/ onlineoecdemploymentdatabase.htm; OECD (2018[4]), Good Jobs for All in a Changing World of Work: The OECD Jobs Strategy, http://www.oecd.org/mcm/documents/C-MIN-2018-7-EN.pdf.
The United States is another country in which the employment rate is still significantly below the pre‑crisis level, despite the longest job recovery in the post‑war period: the unemployment rate is now below the pre‑crisis level but broad labour underutilisation is up by 3.2 percentage points. Despite a relatively stable employment rate, in Italy both the unemployment and labour underutilisation rates were higher in 2016 than in 2006 by 4.6 and 6 percentage points, respectively, due to the opposite effects of increasing labour force participation and soaring involuntary part‑time. Last but not least, the latest available data show a significantly higher labour underutilisation also in Iceland (by 3 percentage points) as well as in Portugal and Slovenia (by 4 percentage points).
The OECD Job Quality framework measures job quality along three dimensions: i) earnings quality, which refers to the extent to which the earnings received by workers in their jobs contribute to their well‑being by taking account of both the average level as well as the way earnings are distributed across the workforce; ii) labour market insecurity, which is measured as the ex‑ante expected monetary loss associated with becoming and staying unemployed as a share of previous earnings; and iii) the quality of working environment, measured as the incidence of job strain that is characterised by a combination of high job demands and few job resources to meet those demands.
Trends in job quality since the mid‑2000s have been contrasted (Figure 1.2). On the one hand, earnings quality has increased, albeit in a limited way, and job strain decreased almost everywhere. On the other hand, labour market insecurity in 2016 was still above 2006 levels in many countries.
Gross hourly earnings expressed in 2010 USD purchasing power parity adjusted by inequality2 have increased modestly in most countries, from 15.59 USD to 16.87 USD between 2006 and 2015. This increase is mainly due to limited growth in real wages (see Section 1.2) and an extremely small reduction in earnings inequality. Earnings quality fell significantly in Greece in this period (with a slump of 1.39 USD), and to a limited extent in Mexico, Turkey and the United States (where adjusted gross hourly earnings decreased by 0.15 to 0.35 USD). Large increases (above 3 USD) occurred in Norway only.
Among the countries for which data are available, the incidence of job strain was 27.5% on average in 2015, against 34.5% in 2005 (Figure 1.2, Panel C). The largest drop, albeit from very high values, occurred in Germany (about 16 percentage points), where job strain incidence is now close to the OECD average. By contrast job strain increased only in Sweden (about 2 percentage points) although, at 25.5% in 2015, the country remains among those with the lowest incidence. It must be kept in mind, however, that these trends may not only be driven by structural improvement but also reflect business‑cycle‑related factors affecting the composition of jobs.3
The increase in labour market insecurity (Figure 1.2, Panel B) is largely driven by the fact that, despite higher employment rates, unemployment in a number of countries in 2016 was not yet at its pre‑crisis levels – see OECD (2018[1]). Reduction in unemployment‑benefit coverage during this period (see Chapter 5), however, played a role in many countries as well. The ex‑ante expected monetary loss associated with becoming and staying unemployed increased by more than one percentage point between 2006 and 2016. The largest increase in labour market insecurity (above 10 percentage points) occurred in Greece and Spain. By contrast, in Germany and the Slovak Republic, labour market insecurity fell by more than 1.5 percentage points.
Figure 1.2. Contrasting trends in job quality
Note: Average earnings adjusted for inequality are obtained as a generalised mean of individual earnings with coefficient ‑3.
Source: OECD (2018[4]), Good Jobs for All in a Changing World of Work: The OECD Jobs Strategy, http://www.oecd.org/mcm/documents/C-MIN-2018-7-EN.pdf.
Figure 1.3. In spite of more inclusive labour markets, poverty remains a concern
Note: Data on low‑income rate refer to 2015 except for Costa Rica and Israel (2016); Denmark, Germany, Hungary, Iceland, Ireland, Italy, Luxembourg, Mexico, New Zealand (2014); Japan (2012), Data on gender labour income gap refer to 2015 except for the United States (2016); Iceland, Ireland, Italy, Luxembourg and Switzerland (2014); Korea (2013). Data on employment gap for disadvantaged groups are a weighted average of the employment gap for mothers with young children, youth (excluding those in education and not in employment), older workers, non‑natives and people with disabilities.
Source: Low‑income rate: Estimates and calculations based on the OECD Income Distribution Database (IDD), http://oe.cd/idd. Gender labour income gap per capita: OECD calculations based on the European Union Statistics on Income and Living Conditions (EU‑SILC) for European countries except Germany, Household, Income and Labour Dynamics in Australia (HILDA) for Australia, German Socio Economic Panel (GSOEP) for Germany, Basic Survey on Wage Structure combined with Labour Force Survey results for Japan, Korean Labor and Income Panel Study (KLIPS) for Korea, and the Current Population Survey (CPS ‑ Annual Social Economic Supplement), for the United States. Employment gap for disadvantaged groups: OECD calculations from the OECD Employment Database, http://www.oecd.org/employment/emp/onlineoecdemploymentdatabase.htm and OECD International Migration Database, http://www.oecd.org/els/mig/oecdmigrationdatabases.htm; for details see footnotes to Figure 1.7 in OECD (2017[3]), OECD Employment Outlook 2017, http://dx.doi.org/10.1787/empl_outlook-2017-en.
Contrasting trends emerge also as regards labour market inclusiveness. On the one hand, poverty has risen significantly since the onset of the crisis: on average, in the OECD, 10.6% of the working‑age population had equivalised household disposable income lower than 50% of the median in 2015 – the so‑called low‑income rate – against 9.6% one decade before (Figure 1.3).4 Low‑income rates have decreased significantly only in Korea as well as Mexico and Chile – albeit from very high levels in the latter two. By contrast, they have increased by more than 2 percentage points in most of the countries that were hit hard by the euro crisis (Greece, Italy, Spain and Slovenia), as well as in a few Eastern European countries (Hungary and the Slovak Republic).
OECD countries, on the other hand, have clearly managed to reduce gender disparities in the labour market. They have also integrated better disadvantaged groups, such as low‑skilled youth, older workers, mothers with young children, immigrants and people with disabilities. Even if women’s annual labour income is still, on average, 39% lower than that of men, this gap fell by 4.5 percentage points between 2006 and 2015.5 Improvements are observable in all OECD countries with available data except Poland, with Luxembourg, Belgium and Ireland showing a reduction even greater than 10 percentage points (see Chapter 6 for a finer analysis of gender labour market disparities and their causes). Similarly, despite the fact that the crisis hit hard on certain groups, the average employment gap of disadvantaged groups6 has decreased in all OECD countries except in Greece and Slovenia, thanks also to a sufficiently long period of restored growth. While the average employment rate of these groups was, on average, 29% lower than that of prime‑age men in 2006, this gap was reduced to 25% ten years later. Remarkable progression was experienced by Chile (10.4%), Poland (9%) and Germany (8.4%).
Wage growth trends since the onset of the crisis
The sharp rise in unemployment brought about by the global financial crisis was followed by a significant slowdown in wage growth in a number of countries. This wage restraint helped limit job losses and set the stage for job growth during the recovery. However, a prolonged period of stagnating wages might significantly reduce worker’s living standards and consumer spending, endangering aggregate demand and growth. Therefore, the decline in unemployment during the recovery should be accompanied by a rebound in wages to allow for it to gain full strength.
The recovery in wage growth lags behind the decline in unemployment
While unemployment has been on a declining path for a number of years in most OECD countries (OECD, 2016[5]), wage growth remains remarkably lower than it was before the recession for comparable levels of unemployment. Underemployment, the productivity slowdown and low inflation expectations are natural candidates to explain this shift of the Phillips curves7 (IMF, 2017[6]; ECB, 2016[8]; Hong et al., 2018[10]). Some additional country specific explanations have been put forward, such as reduced profitability due to the fall in the terms of trade or the high real exchange rate in Australia (Bishop and Cassidy, 2017[11]; Connolly, 2016[12]; Jacobs and Rush, 2015[13]).
The wage‑Phillips curves presented in Figure 1.4 show how nominal hourly wages and unemployment co‑varied, both during the previous cycle (in grey) and during the post‑crisis period (in blue). A rising unemployment gap – defined as the percentage‑point change in unemployment since the start of the global financial crisis – increases competition among workers for jobs and allows employers to lower their wage offers.8 Provided that inflation expectations, productivity growth and the composition of the workforce do not change significantly, that wage adjustments are not made only on the extensive margin (that is for new hires only), and that labour market slack is well proxied by unemployment, the relationship between the change in unemployment since the start of the crisis and wage growth should follow a stable pattern, at least in the short run: wage growth should decline as unemployment rises and then increase back to its previous levels as the unemployment gap shrinks.
OECD‑wide, there has been a clear shift of the Phillips curve following the crisis (top‑left panel of Figure 1.4). During the recession, the average hourly wage growth was between 1.5 and 2 percentage points lower than it was before the recession for comparable levels of unemployment. There is also a gap between the pre‑recovery and post‑recovery curves, showing that this shift has even deepened during the recovery. On average, hourly wage growth in OECD countries was still 0.4 percentage points lower in the last quarter of 2017 than it was in late 2008, while unemployment was at a similar level.
Even in Ireland, the United Kingdom and the United States, where no downward shift of the Phillips curve was observed in the early recovery phase, a softer wage growth with respect to pre‑recovery Phillips curves was observed in 2017. In Germany, the continuous decline in unemployment since 2010 has been accompanied by successive shifts of the Phillips curve. These observations highlight that even in those countries where wage growth seems to be picking up the recovery might be fragile.
Full‑time wage growth has decreased uniformly across the wage distribution between the previous and the current cycle in a number of countries. Figure 1.5 compares the slowdown in nominal wage growth of full‑time employees at the lower decile, the median and the upper decile of the earnings distribution between the periods 2000‑07 and 2007‑16. Average annual growth of median full‑time wages fell by 1.5 percentage points in the OECD area, and slumped by more than 3 percentage points in Ireland, Greece and Portugal as well as many Eastern European countries. Noteworthy, with the only exception of Mexico, in all the countries where wage growth at the median of the wage distribution decelerated by at least one percentage point per year, the wage growth slowdown was significant also at the top decile. Moreover, with the additional exceptions of Latvia and Slovenia, wage growth fell significantly also at the bottom. Yet, the lower deceleration of the bottom decile in a number of countries is by and large a statistical artefact due to composition effects in the context of rising unemployment, particularly strong amongst the low skilled, and should not be taken as evidence that inequality in labour income has decreased since the onset of the crisis. In fact, market income inequality has rather increased over recent years – see OECD (2018[1]).
Figure 1.4. The recovery in wage growth lags behind the decline in unemployment
Note: For ease of interpretation series have been trended using a Hodrick‑Prescott filter.
a. Nominal wage growth: year‑on‑year percentage change in nominal hourly wage (defined as total wages divided by hours worked by employees); unemployment gap: percentage‑points change in the unemployment rate since the start of the crisis in Q4 2007.
b. Unweighted average of 29 OECD countries (excluding Chile, Iceland, Korea, Mexico, New Zealand and Turkey).
Source: OECD calculations based on quarterly national accounts, and the OECD Short‑Term Labour Market Statistics Database, http://dx.doi.org/10.1787/data-00727-en.
Figure 1.5. The slowdown in wage growth was widely spread
Note: Estimates based on gross earnings of full‑time wage and salary workers. However, this definition may vary from one country to another. Further information on the national data sources and earnings concepts used in the calculations can be found at http://dx.doi.org/10.1787/data-00302-en. Results for Estonia, France, Latvia, Lithuania, Luxembourg, Portugal, Spain and Slovenia are based on the European Structure of Earnings Survey (SES).
a. 2000‑07 refers to 2000‑06 for Chile, Italy and Switzerland; 2001‑06 for Poland; 2001‑07 for the Czech Republic and Israel; 2002‑06 for Estonia, France, Latvia, Lithuania, Luxembourg, Portugal, Spain and Slovenia; 2002‑07 for Denmark and the Slovak Republic; 2004‑07 for Austria and Greece; and 2005‑07 for Mexico. 2007‑16 refers to 2006‑14 for Estonia, France, Latvia, Lithuania, Luxembourg, Poland, Portugal, Spain, Slovenia and Switzerland; 2006‑15 for Chile; 2006‑16 for Italy; 2007‑13 for Sweden; and 2007‑15 for Austria, Belgium, Denmark, Finland, Ireland, Israel, Japan and Norway.
b. Unweighted average of the 32 OECD countries shown (not including Iceland, the Netherlands and Turkey).
Source: OECD calculations based on the OECD Earnings Distribution Database, http://dx.doi.org/10.1787/data-00302-en.
The low‑inflation environment and the productivity slowdown have both driven wage growth down
In line with the shift of the Phillips curves, Figure 1.6 (Panel A) shows that the average nominal growth of hourly wages in the OECD experienced a significant decline, from 4.8% prior to the crisis to 2.1% in recent years. However, this decline only partially affected the living standards of workers, due to lower inflation, which decreased from 2.6% to 0.8% (Panel C). As a result, real wage growth decreased by 1 percentage point over this period, from 2.2% to 1.2% (Panel B).
Figure 1.6. Low inflation and the productivity slowdown have both driven wage growth down since the crisis
Note: Countries are ordered by ascending order of the average annualised growth rate in nominal hourly wages in Q1 2000‑Q4 2007.
a. Q4 2012‑Q4 2016 for Switzerland.
b. Total wages divided by total hours worked of employees (and deflated using the private consumption price index in Panel B).
c. OECD is the unweighted average of the 29 OECD countries shown (not including Chile, Iceland, Korea, Mexico, New Zealand and Turkey).
d. Hourly labour productivity refers to real gross domestic product (GDP) divided by total hours worked.
Source: OECD calculations based on quarterly national accounts.
Most of the OECD countries experienced a significant slowdown in wage growth at the depth of the crisis. Real wages even declined in some countries, mostly in the euro area, and especially in countries that were hit hard by the sovereign debt crisis, such as Portugal, Spain, Italy and Greece. The dramatic wage reductions in the Baltic States can be related to the high wage growth that occurred in these countries prior to the crisis and by soaring unemployment at the crisis trough. Outside the euro area, real wages declined in Israel and the United Kingdom, while real wage growth considerably slowed down in the United States, to reach 0.3%, on average, between the fourth quarter of 2007 and the first one of 2009.
While real wages rebounded in most countries after the crisis trough, at about 1.2%, on average, real wage growth remained surprisingly stagnant in the OECD area after the end of the recession despite the progressive reabsorption of labour market slack. In most countries, wage growth did not change much after 2010. Between 2009‑12 and 2012‑17, real wage growth decelerated in Australia, Norway, Switzerland, the Netherlands, the United States and France; and accelerated by less than 0.5 percentage points in Japan, Belgium, Sweden, Greece, Finland, Canada, Austria. More impressive, real wages decreased during the recovery not only in Greece, but also in the Netherlands and Australia.
The low‑inflation environment and the productivity slowdown have both contributed to this deceleration of wage growth (IMF, 2017[6]; Shambaugh et al., 2017[9]). Inflation deceleration has lowered inflation expectations, thereby driving down growth of negotiated wages (see also Chapter 3). Similarly, hourly labour productivity growth has only partially recovered from the negative levels reached during the first phase of the crisis9: it went down from 2.3%, on average, prior to the crisis to 1.2% in the recent period (Figure 1.6, Panel D). In a context of stagnating workers’ bargaining power and strong capital‑labour substitution (see Chapters 2 and 3), this inevitably put a limit to the possibility of raising wages.
Low‑pay jobs have played a role in sluggish wage growth
The recent literature suggests an additional explanation for the recent shift to the left of the wage‑Phillips curve: labour market slack would be greater than what measured by headline unemployment because of greater labour underutilisation due to higher inactivity (Blanchflower and Posen, 2014[14]) and more involuntary part‑time employment (IMF, 2017[6]; Smith, 2014[15]), in particular due to those working part‑time for economic reasons (Altig and Higgins, 2014[16]). For example, Figure 1.7 shows an upsurge of involuntary part‑time in many countries following the recession. Aggregate regressions seem to confirm an impact of the share of involuntary part‑timers on the Phillips curve (IMF, 2017[6]), and the contribution of this effect is particularly large in countries where the unemployment rate is still above pre‑crisis averages. More generally, this additional slack would be related to the stylised fact that, in the aftermath of the recent, long crisis, many jobseekers have been forced to accept jobs that they consider to be worse in terms of working conditions with respect to their expectations and the job they had before the crisis. These workers are still intensively searching for better jobs, thereby raising the number of applications per vacancy for these jobs and exerting downward pressure on wages.
Figure 1.7. The incidence of involuntary part‑time employment increased following the crisis until the early recovery, but then started to decline
Note: Data refer to the share in total employment in Panel A or total part‑time employment in Panel B for Canada, Czech Republic, Israel, Japan, Norway and the United States. Part‑time employment is based on national definitions.
OECD is the unweighted average of the 29 OECD countries shown at each period (excluding Chile, Iceland, Korea, Mexico, Switzerland and the United Kingdom).
Source: OECD Employment Database. http://www.oecd.org/employment/emp/employmentdatabase-employment.htm.
This line of reasoning relies on conjecturing an aggregate phenomenon, by which workers who could find only low‑pay jobs (e.g. those in involuntary part‑time jobs), by searching for higher‑pay jobs (e.g. full‑time jobs), would drive pay on these jobs down due to labour supply in excess of labour demand, resulting in lower average wages. It is possible, however, that part of the slowdown in average wages is only driven by the dynamics of low‑pay jobs and their specific wages. This might occur either because the wages associated with these jobs grow more slowly than those of other jobs (this effect is called heterogeneity effect hereafter), or simply because they are lower and the incidence of the corresponding jobs increases (hereafter called standard composition effect). As opposed to the aggregate story, this scenario involves no effect on remaining jobs. For example, if part‑timers are less paid than full‑timers, then the shift in employment composition towards part‑time jobs would result in lower aggregate wage growth, even with no effect on within‑group wage growth (standard composition effect). Similarly, a decline in the average growth of part‑time wages would result in a slowdown in average wage growth, even in the absence of an effect on the average growth of full‑time wages (heterogeneity effect). The latter effect may even result from an increase in the number of lower‑pay jobs among part‑time jobs (e.g. an increase in the share of involuntary part‑time in total part‑time). The sum of the composition and heterogeneity effects (hereafter called broad composition effect) can be obtained as the difference between the growth rates of the average hourly wage of all workers and of workers in the relatively higher‑pay group of jobs – e.g. full‑timers (OECD, 2018[1]).
Figure 1.8 shows significant broad composition effects of part‑time employment on wage growth in a number of the euro area countries, i.e. the Netherlands, Greece, Germany, Belgium, Italy, Spain and Portugal in the period 2006‑14.10 For example, in Germany, the growth of average hourly real wages for all employees would have been 0.67 percentage point per year greater had it been the same as that of full‑timers. The differential growth of full‑time and part‑time wages (heterogeneity effect) generally played a bigger role in these countries than the standard composition effect – see OECD (2018[1]). This highlights a significant worsening of the earnings of part‑time jobs relative to that of full‑time jobs. Spain and Italy were exceptions, however: most of the significant broad composition effect observed in these two countries was simply driven by the increasing share of part‑time employment and the lower average pay of part‑timers. Such standard composition effects were also at play in Iceland, Norway and Ireland, although the broad effects were mitigated by very weak (or even positive) heterogeneity effects – that is by a relatively dynamic growth of part‑time wages.
The picture is similar when focusing on the early recovery period only (2010‑14): Germany, Greece, Italy, the Netherlands, and Spain still exhibit significant broad composition effects, mainly because of low growth of part‑time wages relative to full‑time wages (with the exception of Italy), which explains part of the stagnation in overall wage growth observed during the early recovery years. Similarly, this type of effects significantly contributed to the wage growth slowdown during the crisis period (2006‑10) in Belgium, Greece, the Netherlands and Portugal.11 By contrast, there is little evidence of significant broad composition effects of part‑time employment during the recovery phase of the previous business cycle (2002‑06)12, thereby highlighting the specific influence of these effects in the post‑crisis sluggish wage growth (OECD, 2018[1]).
In turn, the contribution to the wage growth slowdown of the differential growth between full‑time and part‑time wages (i.e. the heterogeneity effect of part‑time employment) appears to have been significantly associated with the growth of the share of involuntary part‑time in part‑time employment during the early recovery period (Figure 1.9). For example, in Austria, Belgium, Czech Republic, Estonia, Lithuania and Latvia a stable or declining involuntary part‑time employment was coupled with an increase in the wages of their part‑time jobs relative to full‑time wages. By contrast, involuntary part‑time employment grew significantly in many other countries, while part‑time jobs experienced a relative decline in their wages. Rising involuntary part‑time employment might therefore have played a role in the relative decline in part‑time wages that drove overall wage growth down between the onset of the crisis and the early recovery.13
Figure 1.8. Broad composition effects of part‑time employment have continuously driven wage growth down since the crisis
Note: Earnings are deflated using the private consumption price index.
European countries: Data refer to enterprises with at least ten employees in industry, construction and services (except public administration, defence and compulsory social security). OECD is the unweighted average of the 27 OECD countries shown (excluding Canada, Chile, Israel, Japan, Korea, Mexico, New Zealand and Switzerland).
Source: OECD calculations based on Household, Income and Labour Dynamics in Australia (HILDA) for Australia, Labour Force Survey for the United States (CPS ‑ Annual Social Economic Supplement) and Structure of Earnings Survey (SES), Eurostat for other countries.
Figure 1.9. The lower differential growth between full‑time and part‑time wages reflected the expansion of involuntary part‑time employment in the early recovery
Note: The heterogeneity effect reflects the contribution of the differential growth between full‑time and part‑time wages to average wage growth. European countries: Data refer to enterprises with at least ten employees in industry, construction and services (except public administration, defence, compulsory social security). OECD is the unweighted average of the 27 OECD countries shown (excluding Canada, Chile, Iceland, Israel, Japan, Korea, Mexico and New Zealand).
Source: Heterogeneity effect: OECD calculations based on Household, Income and Labour Dynamics in Australia (HILDA) for Australia, Labour Force Survey for the United States (CPS ‑ Annual Social Economic Supplement) and Structure of Earnings Survey (SES), Eurostat for other countries. Growth of involuntary part‑time employment: OECD Employment Database, (www.oecd.org/employment/emp/employmentdatabase-employment.htm).
In a symmetric way, it is possible to conjecture that the role of those heterogeneity effects that are associated with the different wage dynamics of part‑time and full‑time jobs has become less important in the most recent years, due to the recent decline in involuntary part‑time employment in many countries (Figure 1.7).14 One can also expect that this reduction of involuntary part‑time might finally result in wage growth picking up. In fact, there have been recently some timid signs that wage growth has begun to recover (OECD, 2018[2]).15
The effects of the global financial crisis – or the subsequent sovereign debt crisis – were particularly protracted in many countries. As a result a greater number of workers starting a new job spell either just left unemployment or had recent unemployment experiences, likely originated from job loss coupled, in some cases, with subsequent spells of precarious jobs. To the extent that previous research has shown that workers who suffered an unemployment spell are likely to experiment a wage penalty at re‑employment on average – see OECD (2010[17]) and Chapter 4 – or to obtain job offers more frequently in non‑standard forms of employment (Katz and Krueger, 2017[18]), it can be expected that the larger the increase in the share of job‑finders with recent unemployment experience, the lower the growth rate of average wages, even in the absence of a wage growth slowdown for other workers (giving rise to another standard composition effect). In addition, if in the recovery job‑finders tend to accept more frequently lower paid jobs than what they used to do before the crisis, it is likely that the growth of hourly wages of those with recent unemployment experience would be lower, which could again result in lower aggregate wage growth even in the absence of an effect on the growth of average wages of other workers (heterogeneity effect).16
In a number of countries, the overall growth rate of real monthly wages between 2007 and 2014 has been significantly smaller than that of average wages of those who did not experience unemployment spells within the year (Figure 1.10).17 In other words, in these countries wage growth would have been higher in the absence of heterogeneity and composition effects related to more frequent transitions from unemployment to employment and slower growth of the average wage of new hires following an unemployment spell.18 These effects were particularly large in many countries that were hit hard by either the global financial crisis or the euro debt crisis or both, i.e. Estonia, Greece, Italy, Latvia, Slovenia and Spain. For example, in Spain, the annual growth of average monthly wages would have been 0.45 percentage points per year higher in the absence of this type of effects. Standard composition effects generally played the most important role in these countries – see OECD (2018[1]) – since unemployment in 2014 was much higher than in 2007. This type of effects was also important in the Netherlands. In Estonia and Greece, however, a significant part of the impact of this type of broad composition was driven by the relative decline in the wage of those who had a recent unemployment experience, highlighting the relative worsening of the working conditions accepted by job seekers after an unemployment spell compared to those of other workers. Finland and Iceland were also characterised by significant standard composition effects, but their negative impact on average wage growth was mitigated by a relative improvement of the working conditions of those recently unemployed.
Other types of composition effects than those presented here might also have played a role. Additional analysis was therefore carried out to investigate the impact of the changes in the composition of the workforce in terms of age, type of contract or educational attainment. The results of this analysis, however, suggest that all these additional dimensions played a minor role in the wage growth slowdown, on average,19 suggesting that they are at best important only for specific countries.20
Overall, broad composition effects appear to play a significant role. This is particularly the case in countries where unemployment rates are still significantly above pre‑crisis levels. These are the countries where the additional slack effect was found more important in previous research (IMF, 2017[6]). The evidence presented in this chapter suggests that the additional slack effect should not – or at least not completely – be interpreted as an aggregate effect impacting all wages in the same way. The fact that low‑pay jobs have been characterised, in recent years, by increasing incidence and/or lower wage growth mechanically results in lower average wage growth.
Concluding remarks
Employment rates have reached historically high levels in most OECD countries, and the average unemployment rate is back to pre‑crisis level. Yet, the impact of the global financial crisis is still quite visible when one zooms in job quality and inclusiveness. Moreover, wage growth remains significantly subdued compared with pre‑crisis trends and for comparable levels of unemployment; that is to say, the so‑called Phillips curve has shifted during the recession and subsequent recovery.
Figure 1.10. Broad composition effects of unemployment experience have driven wage growth down since the crisis
Note: Wages are deflated using the private consumption price index.
a. 2007‑13 for the United Kingdom and Ireland, 2008‑15 for Australia and the United States. OECD is the unweighted average of the 25 OECD countries shown (excluding Canada, Chile, Germany, Israel, Japan, Korea, Mexico, New Zealand, Switzerland and Turkey).
Source: OECD calculations based on national accounts combined with the European Union Statistics on Income and Living Conditions (EU‑SILC) for European countries, Household, Income and Labour Dynamics in Australia (HILDA) for Australia, Labour Force Survey for the United States (CPS ‑ Annual Social Economic Supplement).
While declining productivity growth as well as lower inflation expectations remain among the primary explanations for the shift in the Phillips curve, this chapter has singled out low‑pay jobs as an important channel accounting for the wage growth deceleration. In particular, earnings of part‑time workers have worsened relative to those of full‑time workers, largely reflected in the rise of involuntary part‑time employment in a number of countries. Moreover, comparatively poor working conditions among those who have regained employment after a joblessness spell, combined with still high unemployment in some countries, pushed up the number of lower‑paid workers, thereby lowering average wage growth. This pattern is probably linked to the fact that, as a result of the protracted economic crisis, many workers were forced to accept low‑pay jobs.
The overall wage growth deceleration therefore hides significant heterogeneity between workers, with a greater impact on vulnerable individuals who are more prone to experience spells of unemployment and/or precarious jobs. In fact, while wages of top 1% income earners have never been so high (Schwellnus, Kappeler and Pionnier, 2017[19]), the share of households at the bottom of the distribution of disposable income is steadily on the rise.21 Wageless growth exacerbates existing inequalities in the labour market, making the need for a more inclusive approach to labour policy – as recommended in the new OECD Jobs Strategy (OECD, 2018[4]) – even more relevant. In this regard, skills policies have a major role to play to ensure that no one is left behind in the context of rapidly evolving skill needs. Indeed, many workers lack basic information‑processing skills that are in high demand in all OECD labour markets, which prevents them from accessing better paid jobs (OECD, 2017[20]). A greater policy effort is therefore required to ensure that every worker is provided with opportunities to develop, maintain and upgrade his/her skills at all ages, thereby reducing the risk of becoming trapped in low‑quality jobs and joblessness, as well as enhancing the ability to adapt to the rapidly changing demand for skills in existing and new jobs.
References
[16] Altig, D. and P. Higgins (2014), The Wrong Question?, Federal Reserve Bank of Atlanta, Macroblog, http://macroblog.typepad.com/macroblog/2014/06/the-wrong-question.html.
[22] Bassanini, A. and E. Caroli (2015), “Is Work Bad for Health? The Role of Constraint versus Choice”, Annals of Economics and Statistics 119/120, pp. 13-37, http://dx.doi.org/10.15609/annaeconstat2009.119-120.13.
[11] Bishop, J. and N. Cassidy (2017), “Insights into Low Wage Growth in Australia”, RBA Bulletin March, pp. 13-20, https://www.rba.gov.au/publications/bulletin/2017/mar/pdf/bu-0317-2-insights-into-low-wage-growth-in-australia.pdf.
[14] Blanchflower, D. and A. Posen (2014), “Wages and Labor Market Slack: Making the Dual Mandate Operational”, Working Paper Series, No. 14-6, Peterson Institute for International Economics, https://ideas.repec.org/p/iie/wpaper/wp14-6.html.
[7] Bulligan, G. and E. Viviano (2017), “Has the wage Phillips curve changed in the euro area?”, IZA Journal of Labor Policy, http://dx.doi.org/10.1186/s40173-017-0087-z.
[12] Connolly, G. (2016), “The Effects of Excess Labour Supply and Excess Labour Demand on Australian Wages”, Paper presented to the 45th Australian Conference of Economists, Flinders University of South Australia, Adelaide, http://esacentral.org.au/images/ConnollyG.pdf.
[24] Daly, M., B. Hobijn and B. Pyle (2016), “What's Up with Wage Growth?”, FRBSF Economic Letter, No. 2016-07, Federal Reserve Bank of San Francisco, San Francisco, CA, https://www.frbsf.org/economic-research/files/el2016-07.pdf (accessed on 30 March 2018).
[8] ECB (2016), “Recent wage trends in the euro area”, ECB Economic Bulletin 3, pp. 21-23, https://www.ecb.europa.eu/pub/pdf/ecbu/eb201603.en.pdf.
[10] Hong, G. et al. (2018), “More Slack than Meets the Eye? Recent Wage Dynamics in Advanced Economies”, IMF Working Paper, No. 18/50, IMF, Washington, D.C., https://www.imf.org/~/media/Files/Publications/WP/2018/wp1850.ashx.
[6] IMF (2017), “Recent Wage Dynamics in Advanced Economies: Drivers and Implications”, in World Economic Outlook, International Monetary Fund, Washington, D.C., https://www.imf.org/en/Publications/WEO/Issues/2017/09/19/world-economic-outlook-october-2017.
[13] Jacobs, D. and A. Rush (2015), “Why is wage growth so low?”, RBA Bulletin, pp. 9-18, https://www.rba.gov.au/publications/bulletin/2015/jun/pdf/bu-0615.pdf#page=11.
[18] Katz, L. and A. Krueger (2017), “The Role of Unemployment in the Rise in Alternative Work Arrangements”, American Economic Review, Papers and Proceedings, Vol. 107/5, pp. 388-392, http://dx.doi.org/10.1257/aer.p20171092.
[23] Nekoei, A. and A. Weber (2017), “Does extending unemployment benefits improve job quality?”, American Economic Review, Vol. 107/2, pp. 527-561, http://dx.doi.org/10.1257/aer.20150528.
[4] OECD (2018), Good Jobs for All in a Changing World of Work: The OECD Jobs Strategy, OECD Publishing, Paris, http://www.oecd.org/mcm/documents/C-MIN-2018-7-EN.pdf.
[2] OECD (2018), OECD Economic Outlook, Volume 2018 Issue 1, OECD Publishing, Paris, http://dx.doi.org/10.1787/eco_outlook-v2018-1-en.
[1] OECD (2018), “Supplementary material for Chapter 1”, in OECD Employment Outlook 2018, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2018-12-en.
[20] OECD (2017), Getting Skills Right: Skills for Jobs Indicators, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264277878-en.
[3] OECD (2017), OECD Employment Outlook 2017, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2017-en.
[5] OECD (2016), OECD Employment Outlook 2016, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2016-en.
[21] OECD (2014), OECD Employment Outlook 2014, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2014-en.
[17] OECD (2010), OECD Employment Outlook 2010: Moving beyond the Jobs Crisis, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2010-en.
[19] Schwellnus, C., A. Kappeler and P. Pionnier (2017), “Decoupling of wages from productivity: Macro-level facts”, OECD Economics Department Working Papers, No. 1373, OECD Publishing, Paris, http://dx.doi.org/10.1787/d4764493-en.
[9] Shambaugh, J. et al. (2017), “Thirteen facts about wage growth”, The Hamilton Project - Economic Facts, September 2017, The Brookings Institution, Washington, D.C., http://www.hamiltonproject.org/assets/files/thirteen_facts_wage_growth.pdf.
[15] Smith, C. (2014), “The Effect of Labor Slack on Wages : Evidence from State-Level Relationships”, FEDS Notes, No. 2014-06-02, Board of Governors of the Federal Reserve System, Washington, D.C., https://www.federalreserve.gov/econresdata/notes/feds-notes/2014/effect-of-labor-slack-on-wages-evidence-from-state-level-relationships-20140602.html.
Supplementary material for Chapter 1
Supplementary material for Chapter 1 is available online only in English at the following DOI: http://dx.doi.org/10.1787/empl_outlook-2018-12-en.
Notes
← 2. Consistent with the OECD Job Quality Framework, average earnings adjusted for inequality are obtained as a generalised mean of individual earnings with coefficient ‑3 – formally this can be written as , where stands for average earnings adjusted for inequality, for income of individual i and for employment headcount; see OECD (2014[21]) for more details.
← 3. For example, during recession years, bad quality jobs are likely to have been destroyed more rapidly, while they might have been more intensively created in the first stage of the recovery. Moreover, work intensity for the same job is likely to vary over the business cycle, with effects on job strain and health – see e.g. Bassanini and Caroli (2015[22]).
← 5. The gender gap in labour income is computed here as the difference between average annual earnings of men and women as a percentage of those of men. Average earnings are computed by considering the whole working age population, independently of whether effectively working or not during the year. A person with no labour income, therefore, contributes to the denominator of average earnings but not to the numerator (see also Chapter 6).
← 6. Defined here as older workers, mothers with young children, youth (excluding those in education and not in employment), immigrants and people with disabilities – see OECD (2017[3]) for more details.
← 7. In the long run, wage growth tends to follow labour productivity growth in the absence of changes in inflation expectations, capital intensity or workers’ bargaining power (see Chapters 2 and 3). Underemployed workers might be still intensively searching for jobs, thereby raising the number of applications per vacancy and exerting downward pressure on wages.
← 8. The unemployment gap is preferred here to the unemployment rate because it allows controlling for cross‑country differences in the structural rate of unemployment.
← 9. Only a few countries did not experience a fall in hourly labour productivity levels just after the crisis: Australia, Canada, Poland, Spain and the United States.
← 10. European Structure of Earnings Survey (SES) data are available only until 2014, which does not allow investigating the role of broad composition effects beyond 2014.
← 11. The same is true for Latvia, although the effects in the longer run (2006‑14) were mitigated by a strong relative increase of part‑time wages relative to full‑time wages in 2010‑14.
← 12. This analysis is undertaken on a restricted set of countries, due to limited data availability.
← 13. Although Figure 1.9 focuses on the 2010‑14 period only, the same pattern can be observed for the 2006‑10 period. However, a few countries are far from the correlation line, making the graph more difficult to read. The graph is therefore not shown here, but is available on request.
← 14. A number of countries, such as Germany and Ireland, are off the correlation line in Figure 1.9, which highlights the role played by the specific institutional contexts in the differential growth of full‑time and part‑time wages.
← 15. This is especially the case in Canada, the Czech Republic, Germany, Hungary, Poland and the United States. For the OECD as a whole, real wages are projected to increase by 1% per year on average in 2018 and 2019 (OECD, 2018[2]). Yet, this is still below pre‑crisis trends for comparable levels of unemployment.
← 16. Job‑seekers may be particularly keen to accept lower wages (and worse working conditions) at re‑employment when they are not entitled to unemployment benefits or when they are approaching maximum potential duration – e.g. Nekoei and Weber (2017[23]) and references cited therein. The negative trend in unemployment benefit coverage in the recovery years (see Chapter 5), by resulting in lower choosiness of jobseekers, could therefore be one factor behind the increase in lower‑paying jobs. To avoid that workers made redundant are exposed to heightened risk of long‑term unemployment, early interventions in the unemployment spells, with appropriate counselling and retraining services, are key. These issues are examined in Chapter 4.
← 17. Overall wage growth rates in Figure 1.8 and Figure 1.10 can hardly be compared, due to differences in the data sampling methodology (SES data refer to firms with more than ten employees only), the definition of wages (hourly earnings versus monthly wages) and the reference period.
← 18. Statistics are constructed from EU SILC, CPS and HILDA microdata. Given that the earnings information available in EU SILC refers to one full calendar year, it is not possible to compute directly the wage growth of those with an experience of unemployment immediately before the job spell. Unemployment experience within the year is therefore used as a proxy. The overall average of monthly wages is trivially equal to the weighted average of monthly wages of those without unemployment spells and of those with some unemployment experience.
← 19. Results are available from the OECD Secretariat upon request. IMF (2017[6]) reaches a similar conclusion as regards industry compositional effects. Further analysis will be carried out in the next editions of the Employment Outlook.
← 20. For example, Daly, Hobijn and Pyle (2016[24]) argue that increased retirement of high‑wage baby‑boomers played a significant role in reducing aggregate wage growth in the United States in recent years.
← 21. See for example Figure 1.3 in Section 1.1 above.