3. The future of work: New evidence on job stability, under-employment and access to good jobs

The complex interplay of globalisation, technological and demographic changes is generating many new opportunities but also challenges for many workers across the OECD. This chapter provides new evidence on three selected topics that have featured prominently in the debate on the future of work: job stability, under-employment and changes in the share of well-paid jobs. The results point to worsening labour market outcomes for those with less than tertiary education and for the young in several countries. In fact, young workers with less than tertiary education stand out as a group that has experienced a pronounced decline in fortunes across a large number of countries. This raises a two-fold challenge. First, policies must promote better opportunities for school-leavers entering the labour market. Second, policies are needed to improve job prospects for the generation of young people who have faced a very tough labour market environment in the past decade.

    

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.

In Brief
Key findings

The complex interplay of globalisation, technological and demographic changes is generating many new opportunities but also challenges for many workers across the OECD. Identifying who is likely to benefit and who may lose out of these deep changes is essential to inform policies contributing to the development of a more inclusive labour market.

This chapter provides new information on three selected topics related to both the quality and quantity of jobs that have featured prominently in the debate on the future of work, but for which hard evidence has been limited – job stability, under-employment and changes in the share of well-paid jobs. First, it investigates whether jobs have truly become less stable and, if so, whether these changes are linked to an increase in the mobility of workers between jobs or between jobs and non-employment. Second, the chapter examines whether there is a growing risk of under-employment (the extent to which people would like to work more hours than they currently do) rather than technologically driven unemployment. More specifically, it looks at how the risk of under-employment has evolved for different socio-demographic groups, as the growth of the service sector, low-skill occupations and atypical forms of employment have contributed to its overall increase in several countries. Finally, the chapter investigates how the chances of getting a middle-paid job have changed for different groups. Again, a key issue here may be that, rather than being confronted by a jobless future, some groups in the labour market may be facing a future where it will be harder to find a well-paid job.

A key finding of the analysis is that the labour market experiences of many young people and of those with less than tertiary education have worsened over the past decade. In fact, the young with less than tertiary education have been particularly affected by these changes, as the share experiencing under-employment, non-employment and low-pay has increased. While these changes have affected different countries to varying degrees, only two countries (Germany and Poland) have not seen a worsening of any of these indicators for young people with less than tertiary education. The evidence suggests that these patterns are unlikely to be only a hangover of the recent global economic crisis.

There is also a clear gender dimension. While the absolute risks of both under-employment and non-employment remain higher for women, the risk of non-employment for men has increased in most countries (particularly for those with less than tertiary education). Men with less than tertiary education have also experienced proportionally large increases in the risk of under-employment. But women remain more likely to be in low-paid jobs and less likely to be in high-paid ones, despite an improvement in the probability of being in middle-paid jobs.

More specifically, the key findings of the chapter are as follows.

Job stability

  • Since 2006, across the OECD, average job stability (as measured by job tenure, the length of time spent in the current job) has increased in a number of countries. This is, however, a compositional effect due to the increase in the share of older workers who tend to have longer job tenure. Once this change in the composition of the workforce is taken into account, job tenure actually declined in most countries.

  • These changes have not affected all workers equally. The largest declines in tenure have occurred for low-educated workers (i.e. those without an upper secondary qualification).

  • There is some evidence that the decline in job tenure stems from the fact that workers move more frequently between jobs rather than from jobs to non-employment. However, the magnitudes of these changes are small and there are many differences between countries.

Under-employment

  • The incidence of under-employment has increased in many countries over the past decade. While the business cycle explains a large part of these changes, more structural and persistent changes have also played a role, including the growth of the service sector, the growing employment share of low-skilled occupations and the spread of non-standard forms of employment with no guaranteed hours.

  • Again, some workers have been more affected than others. Across the OECD, the largest increases in the risk of under-employment have occurred for young people and those with less than tertiary education. Because under-employed workers suffer many disadvantages, these changes represent a substantial challenge on the road to more inclusive labour markets.

  • Across the OECD, women remain much more likely to be under-employed than men (8% vs 3.2%), but under-employment has grown faster among men in most countries, with larger increases affecting those with less than tertiary education.

Job polarisation and jobs of different pay levels

  • In most countries, the share of middle-paid jobs has increased despite the decline in the employment share of middle-skilled occupations (i.e. job polarisation). However, the decline in the share of middle-skilled occupations has meant that the workers without tertiary education are increasingly moving into low-skilled occupations or out of work entirely.

  • On average across the OECD, there have been small increases in the probability of low-paid employment for the young and for workers with medium education. But there has been a pronounced deterioration in the labour market position of the young with less than tertiary education in many countries. In particular, among those who have left education, there have been widespread increases in the incidence of non-employment and of low pay for those in employment.

  • Low-paid employment has increased also among highly educated young people in some countries. On average across the OECD, they are now more likely to be in low-paid jobs than in high-paid ones.

Introduction

This chapter provides new evidence on selected topics relating to both the quality and quantity of work that have featured prominently in the debate on the future of work. In particular, the chapter looks at recent changes in job stability, under-employment and the availability of jobs at different pay levels.

A common conjecture in the debate on the future of work is that job stability might be decreasing as firms adopt business models that favour ad-hoc, short-term hires of workers over traditional long-term employment relationships. On the one hand, this could lead to increased job insecurity and volatile labour income if the job changes involved are largely involuntary for workers. On the other hand, a decline in average job tenure might be the result of increased job hopping by workers, which could lead to better working conditions. Each scenario requires a different policy approach. Therefore, Section 3.1 documents changes in job stability and offers new evidence on the evolution of job-to-job flows and the risk of involuntary separation for different groups of workers.

The second issue discussed in Section 3.2 is that of under-employment (i.e. the extent to which people are working fewer hours than they would like). As argued in Chapter 2, employment losses resulting from job automation are not likely to be as large as has sometimes been suggested. Nevertheless, there has been a process of de-industrialisation and a growth in service sector employment, where low-skilled and less stable jobs are more common. This structural change may contribute to a permanent shift to higher under-employment even if it does not result in a jobless future. Under-employment has already been advanced as an explanation for weak aggregate wage growth over the last decade (OECD, 2018[1]), but it is also a challenge because under-employed workers are at a considerable disadvantage in the labour market: they receive lower wages and experience worse working conditions than similar workers in full-time or voluntary part-time employment (MacDonald, forthcoming[2]). If under-employment is more common for certain groups than for others, it could further increase disparities in the labour market. Understanding how under-employment affects different groups of workers is therefore crucial to inform policies that seek to make work more inclusive in the future.

The third issue addressed in Section 3.3 concerns the risk that jobs are becoming not only more polarised in terms of skill levels (as documented in Chapter 2) but also in terms of pay. A key concern is whether countries are seeing a decline in the share of the middle-paid jobs, which have typically underpinned the living standards of the middle class (OECD, 2019[3]). To address the lack of evidence on this important policy issue, Section 3.3 investigates whether job polarisation has eroded the share of middle-paid jobs. The analysis yields a picture of the changing relationship between occupational skill levels and pay levels and how this has affected different types of workers.

3.1. Are jobs becoming less stable?

In the debate on what the future of work may look like, a common prediction is that lifetime employment will gradually disappear. Instead, job mobility will increase and people will make more frequent transitions into and out of employment and between jobs, which may involve a different employment status (e.g. employee versus self-employment).

A number of megatrends appear to be pushing towards a decline in job stability, which might constitute a positive or a negative development for workers. Rapid advancements in Information Communication Technology (ICT) boost labour mobility by facilitating job search, by encouraging new business models that rely on outsourcing to a greater extent, and by fostering opportunities for independent work and weakening the traditional employer-employee relationship (see also Chapter 2 and Chapter 4). The forces of globalisation play an equally important role. They increase the risk of job displacement by exposing workers to international competition, and they expand the set of available opportunities by granting workers access to the global labour market. In addition, workers’ preferences vis-à-vis employment, flexibility, and independent work are changing and constitute another fundamental driver of labour mobility (Prising, 2016[4]). The rise of the platform economy and of “gig work” epitomises these transformations, and feeds the anxiety that an increasing number of jobs will be shorter-lived.

Are jobs truly becoming less stable? To answer this question, this section uses data on workers who have left education to analyse trends in job stability and labour market mobility across 30 countries over the past decade. 1 Section 3.1.2 investigates whether the observed changes in tenure reflect increasing flows of workers between jobs, or rather between jobs and non-employment.

3.1.1. Job stability has decreased for all age groups

Overall job stability (as measured by average job tenure) has remained stable or has even risen slightly over the past decade in most OECD countries (Figure 3.1). However, after adjusting for changes in the age, education and gender composition of the workforce, tenure has declined on average. The difference between the unadjusted and the adjusted change in tenure is largely due to population ageing which increases the proportion of older workers (with usually longer average job tenure) relative to younger ones. This has been reinforced by recent reforms, including the termination of early-retirement schemes and increases in the official retirement age in several OECD countries, which have led to higher labour force participation at older ages (OECD, 2014[5]).

After adjusting for changes in the demographic structure, job tenure decreased by 4.9% (or around five months) on average (Figure 3.1).2 This modest change masks considerable variation across OECD countries. Nineteen out of thirty countries saw a decrease in tenure (the Czech Republic experienced almost no change). Tenure decreased by more than 17% in Sweden, Luxembourg and Lithuania, while it increased significantly in countries such as Spain and Latvia. Part of the observed decline in average tenure might be driven by the recovery as workers find new jobs with lower tenure. However, while the limited time period makes it difficult to isolate the effect of the cycle, the average decline in tenure across the OECD is larger when adjusting for the cycle.3 This suggests that the main effect of the cycle over the period considered has been to increase job tenure, consistent with the observation that the crisis disproportionally destroyed jobs with low tenure.

Figure 3.1. Job stability has decreased in the majority of countries after accounting for population ageing
Percentage change in job tenure for workers not in education, unadjusted and adjusted, 2006 to 20171
Figure 3.1. Job stability has decreased in the majority of countries after accounting for population ageing

Note: The OECD average is the unweighted average of the displayed countries. The unadjusted change is the percentage change in average tenure between 2006 and 2017. The adjusted change shows the estimated changes once controlling for the composition of the labour force by age, gender and education. The methodology is similar to the one used by Farber (2010[6]).

1. Data for 2017 refer to 2016 for Australia, Germany, and the United States, and to 2014 for Korea.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, and the United States Current Population Survey (CPS) Tenure Supplement.

 StatLink http://dx.doi.org/10.1787/888933966274

The largest declines in job stability have occurred for workers with low education

The decline in job stability was larger for low-educated workers (i.e. with less than upper-secondary education)4 than for other education groups (Figure 3.2). This was the case for all age groups, with the exception of younger male workers where the decline in tenure was almost identical for all education groups.

Over two-thirds of the OECD countries in the sample saw reductions in tenure for workers with low education (see Annex Figure 3.A.1). The reductions in tenure were in some cases large, exceeding 30% in the Slovak Republic, Lithuania, Hungary, Sweden and Poland. In contrast, Norway and Australia, along with Estonia and Latvia, saw workers without an upper-secondary qualification experience increases in tenure of over 15%.

Figure 3.2. The largest declines in job stability have occurred for low-educated workers
Percentage change in job tenure (years) by gender, age and education, 2006 to 20171
Figure 3.2. The largest declines in job stability have occurred for low-educated workers

Note: Each data point is the unweighted average of the countries included in the analysis, excluding Korea due to data quality. The x-axis is the observed average tenure in 2006 and the y-axis is the percent change in average tenure between 2006 and 2017. Young workers are aged 15 to 29 years, prime-aged workers are aged 30 to 54, and older workers are aged 55 to 69.

1. Data for 2017 for Australia, Germany, and the United States refer to 2016.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Canadian Labour Force Survey, and the United States Current Population Survey (CPS) Tenure Supplement.

 StatLink http://dx.doi.org/10.1787/888933966293

3.1.2. Decreasing job tenure may be the result of increased job mobility

Changes in job stability can be linked to more transitions between jobs or between jobs and non-employment. To gauge the relative importance of these different mechanisms, this section studies changes in different types of flows in the labour market.

To assess whether there has been an increase in moves across jobs, the section uses data on yearly transitions from job to job. An increase in this type of transition is often interpreted in the literature as a positive development since the evidence indicates that switching jobs is typically associated with opportunities for career progression and wage gains (Topel and Ward, 1992[7]; Hahn et al., 2018[8]).5 However, job changes might not be voluntary and might not necessarily lead to better outcomes (for example, for workers who move because their fixed-term contract is ending). Unfortunately, the data available do not allow this distinction to be tested.

To assess whether changes in job tenure are linked to increases transitions into non-employment, the section considers: i) transitions from jobs to non-employment (excluding transitions into education); and ii) changes in the probability of moving into unemployment involuntarily, i.e. as a result of a layoff or the end of a contract of limited duration.

Figure 3.3. Job-to-job flows and transitions out of work differ significantly across the OECD
Percentage point changes in rate of transition out of a job into either another job or non-employment, 2006 to 20171
Figure 3.3. Job-to-job flows and transitions out of work differ significantly across the OECD

Note: The OECD average is the unweighted average of all depicted countries. Values have been adjusted for compositional changes. Norway has been excluded for reasons of data quality.

1. Data for 2017 refer to 2016 for Australia, Germany, and the United States, and to 2014 for Korea.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), and the United States Current Population Survey (CPS) Tenure Supplement.

 StatLink http://dx.doi.org/10.1787/888933966312

The evidence that the decline in tenure is due to higher mobility between jobs is mixed across countries

Over the past decade, job-to-job flows have increased in over one-half of 27 countries after adjusting for demographic changes (Figure 3.3).6,7 At the same time, transitions to non-employment fell over the period in most countries. This suggests that the observed declines in tenure are generally linked to increased mobility between jobs rather than between jobs and non-employment. This was particularly evident in Germany, Slovenia, and the United Kingdom. However, in ten countries, the rate of job-to-job movements fell, with the largest drops recorded in Latvia, Spain, Finland, Iceland, and the United States.

The risk of involuntary entry into unemployment has remained stable on average across countries

Over the same period, the share of workers who became unemployed involuntarily (as a result of a layoff or of the end of a contract of limited duration) did not change on average across the OECD, but there were important cross-country differences. After adjusting for demographics, the likelihood of workers becoming unemployed involuntarily increased significantly in countries like the Netherlands, Italy, Greece, Latvia, and Spain, while it fell most prominently in Sweden, France, Germany and Poland (Figure 3.4).

Five of the 12 countries with an increase in the risk of involuntary separations (Figure 3.4) also experienced a decline in job-to-job flows (Figure 3.3). These are Italy, Australia, Denmark, Spain and Latvia. Hence, changes in mobility in these countries appear to be linked to increased risk and uncertainty for workers.8 Conversely, among the twelve countries with declining risk of involuntary separation, all but two (Iceland and Finland) saw an increase in job-to-job flows. Some of the largest drops in involuntary separations and increases in job-to-job flows were observed in Sweden, Germany, France, Great Britain and Hungary.

Figure 3.4. Involuntary separations show very different trends across countries
Percentage point change in probability of an involuntary separation in the past 12 months leading to unemployment, 2006 to 20171
Figure 3.4. Involuntary separations show very different trends across countries

Note: The OECD average is the unweighted average of all depicted countries. Values have been adjusted for compositional changes. Norway has been excluded for reasons of data quality.

1. Data for 2017 refer to 2016 for Australia and Germany.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey.

 StatLink http://dx.doi.org/10.1787/888933966331

Increases in the risk of involuntary unemployment disproportionately affected men

While the changes in involuntary separations were small on average across the OECD, men generally saw larger increases than women. Among men, the increases were of similar magnitude (always below half a percentage point) across education and age groups with two notable exceptions (Figure 3.5). In fact, older workers with high education and younger male workers with low education saw declines in the rate of involuntary separation of just over half a percentage point.

Among women the picture is mixed, with a negative relationship between a group’s pre-crisis rate of involuntary separation and their associated post-crisis change. Younger women with middle and low education had the highest rates of pre-crisis involuntary separations. They also saw the largest fall post-crisis with rates declining by 0.5 and 1.2 percentage points, respectively. Older and prime-age women with middle- and low-education experienced modest to no declines in the rate of involuntary separation. High-educated workers of all age groups – who had the lowest pre-crisis rates – experienced increases in the rate of involuntary separation similar to their male counterparts.

Figure 3.5. Men saw the largest increase in the risk of involuntary entry into unemployment
Percentage point change in likelihood of involuntary separation, 2006 to 20171
Figure 3.5. Men saw the largest increase in the risk of involuntary entry into unemployment

Note: Each data point is the unweighted average of the countries included in the analysis. Data for Norway is excluded from the average due to data quality issues.

1. Data for 2017 refer to 2016 for Australia and Germany.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey.

 StatLink http://dx.doi.org/10.1787/888933966350

Overall, the cross-country picture is mixed and lower stability has not necessarily been associated with increased mobility between jobs

The hypothesis that job stability is decreasing as the result of increased mobility between jobs finds some support across the OECD. Adjusted for compositional changes, tenure is down on average, while transitions to non-employment are down and job-to-job transitions are up. However, the magnitudes of these changes are small and they are a result of significant heterogeneity across countries rather than a clear trend. A trend does emerge among demographic groups. Workers without an upper-secondary diploma experienced the largest decrease in job stability.

These findings point to the need for careful policy interventions to ensure that workers who are affected by increased job insecurity can count on adequate safety nets, effective (and, where possible, pre-emptive) activation policies, and sufficient opportunities for training and re-skilling. Even when declining tenure is linked to greater mobility between jobs (which could, potentially be a positive trend), policies need to ensure that the fragmentation of careers does not penalises workers in terms of access to social protection or training.

3.2. Not unemployed, but under-employed?

The future of work may not be characterised by greater unemployment (see Chapter 2), but will it be one of greater under-employment? This section shows that under-employment has increased in a number of countries in recent times. While heavily affected by the business cycle, under-employment is also linked to persistent structural changes in the labour market. In particular, the incidence of under-employment is higher in the service sector and within low-skill occupations (MacDonald, forthcoming[2]; Valletta, Bengali and van der List, forthcoming[9]), both of which have been growing in recent decades (OECD, 2017[10]).

Under-employed workers are at a particular disadvantage in the labour market.9 They tend to receive lower hourly wages and experience worse working conditions than similar workers in full-time or voluntary part-time employment (MacDonald, forthcoming[2]). Understanding how under-employment affects different groups of workers is therefore crucial to inform policies to promote inclusive labour markets.

This section provides new evidence on how the risk of under-employment has evolved in recent times for different groups of workers, using data from 33 OECD countries and Colombia from 2006 to 2017. Under-employed workers are defined as workers whose main job is part-time and who report either that they could not find a full-time job or that they would like to work more hours.10 Throughout the section, people still studying are excluded from the analysis to ensure that the results are not driven by an increasing number of students seeking part-time employment.

3.2.1. The incidence of under-employment varies across countries and has increased more in those hit hardest by the crisis

The level of under-employment varies considerably across countries (Figure 3.6). On average in OECD countries in 2017, about a third of all part-time workers were under-employed, amounting to around 5.5% of all employees. Italy, Spain, and Australia had 10% or more of employees in under-employment. At the other end of the spectrum, the figure was less than 2% for Colombia, Japan, Estonia, Turkey, Hungary and the Czech Republic.

The Great Recession certainly played an important role in pushing up under-employment in some countries. Under-employment rose in Ireland, Italy, Greece, and Spain, which had been hit particularly hard by the Great Recession, by an average of 6.2 percentage points, much higher than the 1.1 percentage points OECD average.11 In fact, the increases in these countries drive a large share of the overall increase. Without these four countries, the average increase of the remaining ones is 0.4 percentage points.

Nevertheless, levels of under-employment remained higher in 2017 than in 2006 in a number of countries that had long been on a recovery path or had only marginally been affected by the recession (for example in Australia, France, the Netherlands, the United Kingdom, and the United States). This suggests that some of the increase observed is driven by persistent structural changes over and above the temporary fluctuations of the business cycle.

Figure 3.6. The majority of countries have seen increases in under-employment, but particularly those hit hardest by the crisis
Percentage share of dependent workers in under-employment, 2006 and 2017 (or latest year)1
Figure 3.6. The majority of countries have seen increases in under-employment, but particularly those hit hardest by the crisis

Note: The OECD average is the unweighted average of the countries depicted. Under-employed workers are in part-time employment (working 30 hours or less per week) who report either that they could not find a full-time job or that they would like to work more hours.

1. Data for 2017 refer to 2016 for Australia, Germany, and Japan, 2015 for Chile and Turkey, and 2011 for Israel. Data for 2006 refer to 2007 for Colombia and 2009 for Chile.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), United States Current Population Survey (CPS), Canadian Labour Force Survey, Turkey Labour Force Survey, Japan Household Panel Survey (JHPS/KHPS), Colombian Gran encuesta integrada de hogares (GEIH), Chilean National Socio-Economic Characterization Survey (CASEN), Israel Labour Force Survey, Household, Income and Labour Dynamics in Australia (HILDA) Survey.

 StatLink http://dx.doi.org/10.1787/888933966369

3.2.2. The rise in under-employment also reflects permanent structural changes

Across the OECD, under-employment exhibits a positive trend even after accounting for the economic cycle, as shown by the dashed line in Figure 3.7.12 The convergence between the adjusted and unadjusted trends in recent years suggests that under-employment is unlikely to recede much further, unless labour markets become particularly tight as in the years immediately before the recession.

The main structural change contributing to the positive trend adjusted for the economic cycle is the slow but steady growth of the service sector. Figure 3.7 shows that the trend adjusted for the change in the industrial mix is much flatter than the one adjusted for the cycle only. The increasing difference between the two lines since the recession suggests that the importance of the expanding service sector in driving the increase in under-employment has grown in recent times. These results are in line with those obtained for the United States by Valletta, Bengali and Van Der List (forthcoming[9]).

Figure 3.7. Under-employment exhibits an increasing trend even after adjusting for the cycle
Percentage point change in the under-employment rate across the OECD since 2001, adjusted for cyclical and structural factors
Figure 3.7. Under-employment exhibits an increasing trend even after adjusting for the cycle

Note: Time trend shown with reference the baseline level in 2001. The Figure presents unconditional and conditional time trend from a GLM regression model with a logit link. The unconditional model controls for country differences. The model adjusting for cyclical factors adds unemployment and its square to the model. The industry mix is represented by the share of employment in the “Accommodation and food” and the “Arts and leisure” sectors. The results are robust to using a broader definition of the service sector. Estimates are weighted by the average employment for each country over the period. Values are presented as average marginal effects. The model does not include Turkey or Colombia due to data constraints. Similar results are obtained when countries are not weighted by employment, or when a measure of output gap is used to account for the cycle.

Source: OECD calculations.

 StatLink http://dx.doi.org/10.1787/888933966388

Some parts of the service sector have a much higher incidence of under-employment than manufacturing. For example, in 2017 “Accommodation and food services” had a share of under-employed workers of around 12.2%, against only 1.8% in manufacturing (Figure 3.8). One explanation for such large difference is that firms in these sectors often favour part-time employment as a way of dealing with demand variation over the day (Euwals and Hogerbrugge, 2006[11]). In this context, since the choice of a part-time arrangement is driven by the employer’s preference rather than the worker’s, it is more likely to result in involuntary part-time. This argument suggests that under-employment might be particularly important in labour markets where workers have a weak bargaining position vis-à-vis their employers (Chapter 4 further discusses the issue of monopsony power).

Other factors associated with the growth of under-employment are the increasing employment share of low-skilled occupations and the spread of non-standard forms of employment (MacDonald, forthcoming[2]). The increase in the share of low-skilled occupations (which is linked to the expansion of the service sector) is a well-known trend affecting a large number of countries across the OECD – see OECD (2017[10]) and Section 3.3. Chapter 2 discusses the emergence of very atypical part-time contracts, like on-call or zero-hours contracts, where people are not guaranteed any fixed hours or, indeed, any hours at all.

As the structural changes behind the increase in under-employment are likely to continue into the future, it is likely that under-employment will continue to affect a significant (and possibly increasing) number of workers.

Figure 3.8. Under-employment is more common in service sectors
Percentage share of dependent workers indicating under-employment, by broad industry. Unweighted OECD average, 2006 and 2017 (or latest year)1
Figure 3.8. Under-employment is more common in service sectors

Note: The OECD average is an unweighted average. Under-employed workers are in part-time employment (working 30 hours or less per week) who report either that they could not find a full-time job or that they would like to work more hours. Industries are broadly grouped according to a modified NACE Rev.2 A10 classification structure. The category of “Agriculture” broadly corresponds to NACE Rev.2 Section A; “Trade, Logistics, Communications” broadly corresponds to Sections G, H, and J; “Public Services” broadly corresponds to Sections O, P, and Q; and “Arts and leisure” broadly corresponds to Sections R, S, T, and U.

1. Data for 2017 refer to 2016 for Australia, Germany, and Japan, 2015 for Chile and Turkey, and 2011 for Israel. Data for 2006 refer to 2007 for Colombia and 2009 for Chile.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), United States Current Population Survey (CPS), Canadian Labour Force Survey, Turkey Labour Force Survey, Japan Household Panel Survey(JHPS/KHPS), Colombian Gran encuesta integrada de hogares (GEIH), Chilean National Socio-Economic Characterization Survey (CASEN), Israel Labour Force Survey, Household, Income and Labour Dynamics in Australia (HILDA) Survey.

 StatLink http://dx.doi.org/10.1787/888933966407

3.2.3. Under-employment has increased more for youth and those with less than tertiary education

The prevalence of under-employment has increased more for young workers and those with a low or medium level of education (Figure 3.9). Within all age groups of both genders, those with lower education have seen larger increases in the risk of under-employment. And, within all educational groups for both men and women, younger workers have seen larger increases than older ones with one exception. This is the group of prime-age, low-educated women who have seen an increase in the probability of under-employment larger than their younger colleagues and, indeed, than almost any other group (just over 4 percentage points). Overall, three of the four groups with the largest increases include young (male and female) workers with less than tertiary education. For all of these groups the increase is larger than 3.0 percentage points.

By country, young people have seen an increase in the probability of under-employment in 23 of the 34 countries considered in this analysis (see Annex Figure 3.A.2). The average increase for the young among all countries was 2.4 percentage points, but fifteen countries have seen larger increases, and in three (Greece, Italy and, Spain) the increase exceeded 10 percentage points.

3.2.4. Women remain much more likely to be under-employed than men, despite higher than average increases for low-educated men

The share of workers who are under-employed increased by about 1 percentage point for both women and men, but women had much higher initial levels. In 2017, under-employment as a share of dependent workers was almost 8% for women (of all education levels and ages) and only 3.2% for men. For some of the education and age groups reported in Figure 3.9, women’s relative position has improved only because men’s conditions have deteriorated more (for example among young workers with lower education).

While under-employment remains higher among women, its incidence has grown in particular among male workers with less than tertiary education. Among prime-aged male workers with low education, the incidence of under-employment almost doubled from 2.7% in 2006 to 5.1% in 2017. Meanwhile, for low educated young male workers, it increased by almost 80%, reaching 9.7%. Among male workers with medium education, the young ones saw a two-fold increase (from 3.0% to 6.1%), and those in the prime-working age group an increase of 79% (from 1.3% to 2.4%).

Men have seen an increase in the probability of under-employment in 28 of the 34 countries considered (with decreases in Colombia, Poland, Lithuania, Latvia, and Hungary) see Annex Figure 3.A.3. The fortunes of women varied more across countries, as their risk of under-employment actually declined in 13 of the 34 countries. Furthermore, in five of the six countries in which men saw the largest increases in the probability of under-employment (that is, the Slovak Republic, Italy, Spain, Greece and Ireland) women experienced even larger increases. Denmark was an exception, with men experiencing a larger increase.

Figure 3.9. The young and those with low education have seen the largest increases in under-employment
Change in the percentage share of under-employed dependent workers, by age, gender and level of education, unweighted OECD average for 2006-17 (or latest year)1
Figure 3.9. The young and those with low education have seen the largest increases in under-employment

Note: The OECD average is the unweighted average of the countries depicted. Under-employed workers are in part-time employment (working 30 hours or less per week) who report either that they could not find a full-time job or that they would like to work more hours.

1. Data for 2017 refer to 2016 for Australia, Germany, and Japan, 2015 for Chile and Turkey, and 2011 for Israel. Data for 2006 refer to 2007 for Colombia 2009 for Chile.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), United States Current Population Survey (CPS), Canadian Labour Force Survey, Turkey Labour Force Survey, Japan Household Panel Survey(JHPS/KHPS), Colombian Gran encuesta integrada de hogares (GEIH), Chilean National Socio-Economic Characterization Survey (CASEN), Israel Labour Force Survey, Household, Income and Labour Dynamics in Australia (HILDA) Survey.

 StatLink http://dx.doi.org/10.1787/888933966426

3.3. Job polarisation and access to good jobs

The share of middle-skilled occupations has declined in most OECD countries in recent decades (OECD, 2017[10]). A major policy concern is that this might be causing a decline in the number of middle-paid jobs that have typically underpinned the living standards of the middle class. This worry is often echoed in the press (Yglesias, 2014[12]; Elliot, 2017[13]) as well as in the academic and policy debate on the fortunes of the middle class in recent times (Vaughan-Whitehead, Vazquez-Alvarez and Maître, 2016[14]; Pew Research Center, 2015[15]). However, the share of middle-paid jobs might hold up if low- or high-skilled occupations increasingly pay wages close to the median. While job polarisation has received much attention in the academic and policy debate on the future of work, there is surprisingly little evidence on whether the share of middle-paid jobs has held up or not.

Therefore, this section brings new evidence to bear on this issue, using data from up to 32 countries from 2006 to 2016.13 The choice of the time period is constrained by the need to have reliable data on wages for a large number of countries. Jobs are classified into high, middle or low skill depending on the occupation they belong to. The classification adopted is the same as in OECD (2017[10]) and much of the earlier literature cited therein.14 The distribution of jobs into high, middle, or low pay is based on how their hourly wage compares to the median hourly wage in a given year. Following the OECD standard definition,15 low-paid jobs are those paying less than two thirds of the median wage, while high-paid jobs are those paying more than 1.5 times the median wage.16

3.3.1. The decline in the share of middle-skilled jobs has not led to a decline in the share of middle-paid workers

Labour markets across the OECD continued to polarise in the last decade, with most of the decline in the middle-skilled occupations compensated by growth in high-skilled ones (see Annex Figure 3.A.4). This followed a known trend dating back at least to the 1990s (OECD, 2017[10]), which accelerated during the economic crisis (Green, forthcoming[16]). Between 2006 and 2016 the share of middle-skilled jobs declined in all 31 countries considered, except Luxembourg.17 On average, the decline was just over 5 percentage points and was entirely compensated by an increase in the share of high-skilled occupations.

To investigate how job polarisation has affected the share of middle-paid jobs, Figure 3.10 presents a decomposition of changes in the total share of low-paid, middle-paid and high-paid jobs into two components. The first component is the “job polarisation effect”. This is driven by changes in the relative size of different occupations. For example, one might expect that the decline in the share of middle-skilled jobs – which also tend to be middle-paid jobs – might have pushed the share of middle-paid jobs down. The second component is driven by changes in the propensity of different occupations to pay medium wages. For example, the share of middle-paid jobs might increase as a result of an increases in the proportion of high- or low-skilled jobs paying medium wages. Because this would amount to a “shift” in the distribution of these occupations towards middle-pay, this is referred to as “occupational shift” in this analysis.

The share of middle-paid jobs has increased in most countries despite the decline in the share of middle-skilled occupations

Far from decreasing, the share of middle-paid jobs increased by nearly 2 percentage points on average across the 31 countries considered (Panel B of Figure 3.10).18 In 2016, the average share of middle-paid jobs across all countries was 60%, ranging from 45% in Lithuania to 77% in Denmark. Increases took place in 18 countries, with the largest increases of more than 10 percentage points seen in Hungary, Poland and Greece. Among the 13 countries where the share of middle-paid jobs declined, the average change was just under -2 percentage points. Spain saw the largest decline (-7 percentage points) followed by Estonia (-4 percentage points). A decrease also occurred in Australia (-3 percentage points) but this was accounted for by an increase in the share of high-paid jobs rather than in the share of low-paid jobs.

Overall, job polarisation explains a small part of the change in the share of middle-paid jobs. Instead, changes in the propensity of different occupations to pay medium wages account for most of the changes. This is the case regardless of whether the share of middle-paid jobs is increasing or decreasing. On average across all countries, job polarisation reduced the share of middle-pay occupations by 0.8 percentage points, but changes in the propensity of all occupations to pay medium-level wages added over 2.5 percentage points, resulting in a total net increase of just under 2 percentage points.19

The share of high-paid jobs has not grown as fast as the share of high-skilled occupations

Job polarisation has generally contributed to increasing the share of high-paid jobs across countries, but a decline in the propensity of different occupations to pay high wages has limited the growth of high-paid jobs (Panel C of Figure 3.10). On average, job polarisation contributed 1.3 percentage points to the growth of high-paid occupations, but the second component subtracted 1.8 percentage points, resulting in a net small decline of -0.5 percentage points. In other words, the share of high-paid jobs has generally grown less than would have been expected given the shift of the occupational structures towards high-skilled occupations. This pattern is observed in the vast majority of countries.20 In 2016, the average share of high-paid jobs (relative to each country’s median wage) was 21% across all countries, ranging from 11% in Denmark to 29% in Portugal.

On average across the 31 countries, the share of low-paid jobs has declined by 1.3 percentage points (Panel A of Figure 3.10). About 0.5 percentage points of this is explained by the fact that high-skilled occupations have grown more than the other occupations. The remaining 0.8 percentage points is explained by a decline in the propensity of different occupations to pay low wages. The decline in the share of low-pay employment occurred in 20 of the 31 countries considered.

The polarisation of the occupational structure has not led to a hollowing out of the pay distribution

The main conclusion of this analysis is that job polarisation has not resulted in a decline in the share of middle-paid jobs and an increase in the share of high-paid jobs, as might have been expected. Instead, changes in the propensity of different occupations to pay medium-wages have tended to increase the share of middle-paid jobs and decrease that of high-paid ones.

These results point to a changing association between occupational skill levels and relative pay levels which are likely to affect workers in different ways, generating winners and losers. Identifying who these might be is not trivial. In fact, as the propensity of different occupations to pay wages of different levels changes, so do the characteristics of the workers employed in them. This is in part driven by socio-demographic trends, such as increasing female labour force participation, ageing and increasing educational attainment. But the propensity of different groups to work in different occupations is also changing (see Box 3.1).

Hence, to identify those who have lost ground in the midst of all these changes, the next section looks at how the chances of being in a low-paid job have evolved for men and women of different age groups and education levels.

Figure 3.10. Job polarisation explains a small part of the change in the share of middle-paid jobs
Percentage point changes in the share of jobs by level of pay between 2006 and 2016
Figure 3.10. Job polarisation explains a small part of the change in the share of middle-paid jobs

Note: Low-paid jobs are those paying less than two thirds of the median wage, while high-paid jobs are those paying more than 1.5 times the median wage. The OECD average is the unweighted average of all displayed countries. The time period covered is 2006-16, except for Korea (2006-14), Australia (2006-15). Greece, Portugal and Latvia (2007-16), Italy (2007-15), Switzerland (2008-15). Chile, Canada, Ireland and Luxembourg (2006-15), and Iceland (2006-13).

Source: EU statistics on income and living conditions survey (EU-SILC), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, Chilean National Socio-Economic Characterization Survey (CASEN), and CPS Merged Outgoing Rotation Groups (MORG) for the United States.

 StatLink http://dx.doi.org/10.1787/888933966445

Box 3.1.The employment of workers with less than tertiary education has shifted towards low-skill occupations as labour markets have polarised

Where are the workers going who typically would have been employed in middle-skilled occupations? This is a key question for understanding the consequences of job polarisation for workers. Its answer may also illuminate what policies, if any, should be implemented to counter any ill effects of job polarisation.

Although the destruction of middle-skilled jobs might cause distress for workers who held those jobs, this need not be the case. Workers may be able to quickly reskill or work with their employers to transition to high-skilled occupations. Part of the loss of middle-skilled jobs may also occur through a process of attrition as workers in these jobs retire and fewer workers are hired to replace them. Alternatively, workers losing these jobs may only be able to find work in low-skilled occupations, or faced with diminished pay and opportunities, they may decide to exit employment entirely.

Recent OECD research provides the first comparative evidence on how the decline in the share of middle-skilled occupations has affected workers with less than tertiary education, who account for the majority of workers in middle-skilled occupations. (Green, forthcoming[16]). Previous studies have examined individual countries (Jaimovich, Siu and Cortes, 2017[17]; Bachmann, Cim and Green, 2018[18]; Maczulskij and Kauhanen, 2017[19]; Salvatori, 2018[20]).

From the mid-1990s to the mid-2010s across the OECD21, the share of all working-age men (whether in employment or not) with middle-education who are in a middle-skilled occupation declined by 2.9 percentage points from 40.4% to 37.5%, while the share working in low-skilled employment rose from 11.3% to 15.4%. Middle-educated women experienced an even larger shift from middle-skilled employment to low-skilled employment. Their share in middle-skilled jobs dropped by 6.0 percentage points from 22.6% to 16.6% while their share in low-skilled occupations increased from 18.1% to 28.0%. Part of this increase was driven by an increase in women’s labour force participation. The employment-to-population ratio of middle-educated women increased by 4 percentage points over this period.

The proportion of low-educated men working in low-skilled employment also increased from 12.0% to 14.8% while their share in middle-skilled occupations declined by 7.8 percentage points. Over the same period, their share in non-employment rose from 47.1% to 53.3% (see also Section 3.3.3 as well).

The share of low-educated women in middle-skilled employment fell 5.5 percentage points from 19.2% to 14.7%, while their share in low-skilled employment rose by almost the same magnitude from 17.2% to 22.4%. The proportion of low-educated women in employment did not change over this period.

To summarise, there has been a substantial decline in middle-skilled jobs for workers with less than tertiary education. Employment in these middle-skilled occupations provided many workers with a good standard of living. Their decline has meant that workers without a college degree who previously would have held these jobs are increasingly moving into low-skilled occupations, or out of work entirely.

3.3.2. The probability of low-paid employment has increased for some groups of workers

On average, all workers, irrespective of age, education and gender, have seen a decline in the probability of high pay, with the largest decline affecting workers with high education (Figure 3.11).22 For all groups this has translated in an increase in the probability of middle pay, but for some also the probability of low-paid employment has increased slightly. In particular, while the probability of low-paid employment has declined on average for both men and women,23 the results differ across education and age groups.

Figure 3.11. The probability of being in a high-paid job has declined across demographic groups
Percentage point change in the share of jobs by level of pay and by age, gender and level of education between 2006 and 2016, OECD average
Figure 3.11. The probability of being in a high-paid job has declined across demographic groups

Note: The OECD average is the unweighted average of all countries included in Figure 3.10 and Mexico. Low-paid jobs are those paying less than two thirds of the median wage, while high-paid jobs are those paying more than 1.5 times the median wage. The time period covered is 2006-16, except for Korea (2006-14), Australia (2006-2015). Greece, Portugal and Latvia (2007-16), Italy (2007-15), Switzerland (2008-15). Chile, Canada, Ireland and Luxembourg (2006-15), and Iceland (2006-13).

Source: EU statistics on income and living conditions survey (EU-SILC), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, Chilean National Socio-Economic Characterization Survey (CASEN), ENOE longitudinal survey for Mexico and CPS Merged Outgoing Rotation Groups (MORG) for the United States.

 StatLink http://dx.doi.org/10.1787/888933966464

The incidence of low pay increased more for the young and those with medium education

All education groups saw an increase in the probability of being in low-paid jobs across the OECD between 2006 and 2016.24 The changes are all very small, with the largest one reaching a mere 0.8 percentage points for workers with middle-education.25 The small average changes are the result of very different patterns across countries.26

On average, across countries, low- and medium-educated workers have seen more of a shift towards low pay than those with high education. The low-educated saw an increase in the probability of low pay in 18 of the 32 countries (average increase: 4 percentage points), and the medium-educated in 23 (average increase: 3 percentage points) (see Annex Figure 3.A.5). In just over half of the countries (17), these two groups together experienced a net shift of employment towards low-paid jobs. The impact of these shifts across the pay distribution on wage differentials by level of education is discussed in Box 3.2.

By age, an increase in the probability of low pay on average across countries was only recorded for the youngest age group, albeit a very small one (0.2 percentage points, Figure 3.11).27 The increase occurred in 15 of the 32 countries (average of 6.5 percentage points) and in 13 of these resulted in a net shift of young people to low pay. By contrast, for workers aged 30 to 50 the probability of being in low pay declined on average across the OECD by just over 1 percentage point, increasing in only eight countries (see Annex Figure 3.A.6).

Hence, while these changes have affected different countries to varying degrees, the results point to a small increase on average in the probability of working in low-paid jobs for young workers and those with less than tertiary education. These pattern are not driven by increasing enrolment in education as they are obtained excluding those in education and are confirmed when the sample is restricted to full-time workers. This motivates a closer look at the fortunes of young workers with different levels of education.

The probability of low-paid employment has increased for young people with medium education

Young people with low and medium level education both saw shifts of their employment towards low pay. For brevity, Panel A of Figure 3.12 reports the results for those with medium education. On average, this group saw a larger increase in the probability of low pay than young workers with low education (2.6 percentage points vs 0.3). Also, they are a larger and more stable group that has been at the centre of the policy debate in many countries in recent years.28

The probability that a young person with medium education is in a low-paid job increased by 2.6 percentage points on average across countries. This was accompanied by similar declines in the probability of middle- and high-paid employment (-1.1 percentage points and -1.5 percentage points respectively) – with similar patterns between genders. As a result, the share of young workers with medium education who hold a low-paid job reached 38% in 2016. 29

The probability of low-paid employment increased for young workers with medium education in two-thirds (19) of the countries. The average increase in these countries was 8 percentage points (against a decline of 6.6 percentage points in the remaining countries). In six countries, the increase was larger than 10 percentage points (France, Norway, Spain, Austria, Denmark, and Estonia). Among the countries were the probability declined, three had changes in excess of 10 percentage points (Sweden, Poland, and Hungary).

In many countries, even young people with high education have seen an increase in the probability of low-paid employment

Young people with a high-level of education also saw increases in the probability of low-paid employment in many countries (Panel B of Figure 3.12). On average, the increase was 3.5 percentage points across the OECD, accounting for over half of the 6.5 percentage points decline in the probability of high pay. As a result, on average across the OECD, young highly educated people are now more likely to be in low-paid jobs than in high-paid ones (21% vs 14.5%). This was the case in 18 countries already in 2006 – which were joined by Slovenia, Iceland and Austria in 2016.

The increase in the probability of low pay for young people with a high-level of education occurred in as many as 22 countries. This resulted in a net shift of employment to low pay in 13 countries. In 17 countries, the net shift was towards middle pay. Luxembourg was the only country with a net shift of employment towards high pay for this group. The largest increases in the probability of low pay were in Portugal (16 percentage points), Ireland (17 percentage points) and Spain (20 percentage points), while the largest increases in the probability of middle pay were in the Czech Republic (14 percentage points), Sweden (16 percentage points) and Hungary (26 percentage points).

Figure 3.12. Young workers with middle education have seen a shift towards low-paid employment in many countries
Percentage point change in the share of jobs by level of pay by level of education between 2006 and 2016
Figure 3.12. Young workers with middle education have seen a shift towards low-paid employment in many countries

Note: The OECD average is the unweighted average of all countries shown. Low-paid jobs are those paying less than two thirds of the median wage, while high-paid jobs are those paying more than 1.5 times the median wage. The time period covered is 2006-16, except for Korea (2006-14), Australia (2006-15). Greece, Portugal and Latvia (2007-16), Italy (2007-15), Switzerland (2008-15). Chile, Canada, Ireland and Luxembourg (2006-15), and Iceland (2006-13).

Source: EU statistics on income and living conditions survey (EU-SILC), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, Chilean National Socio-Economic Characterization Survey (CASEN), ENOE longitudinal survey for Mexico and CPS Merged Outgoing Rotation Groups (MORG) for the United States.

 StatLink http://dx.doi.org/10.1787/888933966483

3.3.3. The risk of being out-of-employment after leaving education has increased for some groups

The changes in the types of jobs by skill and pay level documented in the previous section raise the possibility that some groups might be struggling to find jobs. For example, Beaudry et al. (2016[21]) argue that the increasing influx of high-educated workers into lower-skilled occupations has pushed more low-educated workers out of employment. Therefore, this section documents changes in the probability of being out-of-employment for workers aged 20 to 60 who have left education.

Box 3.2. The evolution of the wage premium for tertiary education

The contemporaneous occurrence of job polarisation and educational upgrading means the distribution of different educational groups across occupations and, more broadly, the pay distribution is changing (see Section 3.3.2 and Box 3.1). How are these changes affecting the wage premium for tertiary education relative to those with secondary education?

Between 2006 and 2016, the average wage premium across OECD countries with available data for tertiary education declined by about 3.3 percentage points (Figure 3.13).30 There is, however, some variation across countries. Overall, the premium declined in 21 of the 32 countries considered here, but only in 12 was the change larger than 5 percentage points. Six countries experienced declines of at least 10 percentage points (Greece, Poland, Chile, Slovenia, Hungary and Portugal). Five countries had positive or negative changes of less than 1 percentage points (France, Luxembourg, Canada and Latvia). In three countries the premium increased by more than 5 percentage points (Belgium, the United Kingdom and Estonia).

Job polarisation and other compositional changes account for about 40% of the decline in the average premium. When holding occupation, age and gender characteristics constant, the average decline across all countries is 2.1 percentage points. In general, this adjusted estimate is lower than the unadjusted one in the countries with falling premia, especially among those with larger changes. Even among the countries with an overall increase in the wage premium, compositional changes have often tended to push the wage premium down as indicated by the fact that the (positive) adjusted estimate is larger than the unadjusted one.

Differences in wage dynamics across occupations have also played a major role in driving changes in the educational premium in most countries. This is largely because average wage growth between 2006 and 2016 was particularly weak in high-skilled occupations, which tend to employ a high share of workers with high education. Indeed, once differences in occupational wage growth are also accounted for, the average change in the education premium across the 32 countries is positive (1.5 percentage points), remaining negative in only 12 countries (and less than 1 percentage point in two of these, Australia and Italy).

Differences in occupational wage growth have played a particularly important role in the countries with the largest declines in the unadjusted education premium. Among the five countries with the largest declines, controlling for occupation, age and gender, and occupational wage growth reduces the estimated fall in the premium by more than 80% in three (Portugal, Slovenia, and Poland) and by around 60% in the remaining two (Hungary and Chile). The average decline among these countries goes from 18 to around 4 percentage points.

The education wage premium within occupational groups has been stable or increasing on average across countries. In particular, it remained stable within high-skill occupations but increased by about 3 percentage points within low and middle skill occupations. In all three cases, the premium declined in fewer than half of the countries considered.

In summary, while workers with tertiary education retain an earning advantage across the OECD (OECD, 2018[22]), over the past decade the education wage premium fell in a number of countries. This was driven in part by the fact that an increasing number of workers with tertiary education are found in occupations that do not pay high wages. The major driver of the fall has actually been the poor performance of wages of high-skill occupations, which still employ a large share of the highly educated group. The evidence of falling education wage premia within occupations is much weaker. This suggests that, in general, all occupations have been able to absorb the increasing supply of workers with high-education and that low- and middle-skilled occupations in particular might be undergoing a process of upskilling in at least some countries.

Figure 3.13. Changes in the education wage premium across countries
Percentage points change in the wage premium of workers with tertiary education relative to those with secondary education between 2006 and 2016
Figure 3.13. Changes in the education wage premium across countries

Note: The reported values are approximate changes in percentage points obtained from country-specific regressions of log of wages on the relevant covariates The OECD average is the unweighted average of all countries shown. Low-paid jobs are those paying less than two thirds of the median wage, while high-paid jobs are those paying more than 1.5 times the median wage. The time period covered is 2006-16, except for Korea (2006-14), Australia (2006-15). Greece, Portugal and Latvia (2007-16), Italy (2007-15), Switzerland (2008-15). Chile, Canada, Ireland and Luxembourg (2006-15), and Iceland (2006-13).

Source: EU statistics on income and living conditions survey (EU-SILC), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, Chilean National Socio-Economic Characterization Survey (CASEN), ENOE longitudinal survey for Mexico and CPS Merged Outgoing Rotation Groups (MORG) for the United States.

 StatLink http://dx.doi.org/10.1787/888933966502

The risk of non-employment has increased for men but decreased for women

On average, the share of men who have left education and are not employed has risen (2 percentage points), with larger increases among those with lower levels of education. Among women, the increases have been limited to the young (of all education levels) and to those aged 30 to 50 with low education (Figure 3.14). Indeed, on average across all countries the share of all women who have left education but are not working has declined by 3 percentage points.

Figure 3.14. The young with low and medium education suffered the largest increases in the probability of non-employment
Percentage point change in the probability of not being in employment for people who have left education, by age, gender and level of education, unweighted OECD average between 2006 and 2016
Figure 3.14. The young with low and medium education suffered the largest increases in the probability of non-employment

Note: The OECD average is the unweighted average of all countries reported in Figure 3.12. The sample is restricted to people who have left education and are aged 20 to 60. Young are aged 20-30, prime age between 30 to 50 and old from 50 to 60. The time period covered is 2006-2016, except for Korea (2006-2014), Australia (2006-2015). Greece, Portugal and Latvia (2007-2016), Italy (2007-2015), Switzerland (2008-2015). Chile, Canada, Ireland and Luxembourg (2006-2015), and Iceland (2006-2013).

Source: EU statistics on income and living conditions survey (EU-SILC), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, Chilean National Socio-Economic Characterization Survey (CASEN), ENOE longitudinal survey for Mexico and CPS Merged Outgoing Rotation Groups (MORG) for the United States.

 StatLink http://dx.doi.org/10.1787/888933966521

The probability of non-employment increased in 26 of the 32 countries for men and decreased in as many countries for women. In all, except four countries (Poland, Slovenia, Slovakia and Estonia), the position of women relative to men improved. Nevertheless, women remain much more likely to be in non-employment than men (on average 27.7% vs 15.8% in 2016).

Young people with low and medium education suffered the largest increases in the probability of non-employment

For both men and women who have left education, the probability of non-employment rose most for young people, especially for those with low education.31 32 In fact, young men with low education experienced the largest increase of all groups (13 percentage points). Women with high education had the smallest increase among the young (2 percentage points).33 When the changes are computed by age group only (pooling together both genders and all education levels), it is only the young who have experienced an increase in non-employment (4 percentage points) on average across the OECD.34

The probability of a young person being out-of-employment after leaving education increased in 25 of the countries considered (see Annex Figure 3.A.7). The average increase across all countries was 4 percentage points, while the other two age groups saw small declines. The increases were above 10 percentage points in five countries (Iceland, Ireland, Italy, Greece, Spain) and below 1 percentage point in Estonia, Belgium and Korea. Among the seven countries where the probability of a young person in non-employment decreased, the largest changes were seen in Germany (4 percentage points), the Czech Republic (5 percentage points) and Poland (10 percentage points). The United States, the United Kingdom and the Slovak Republic saw small negative changes of less than 1 percentage point.

The other groups to have experienced increases in the probability of non-employment of a magnitude comparable to those of the young were all low-education groups (Figure 3.14).

The fortunes of the young have been declining

The results of this section point to a deterioration of the position of young people in the labour market, which has affected in particular those with less than tertiary education. In fact, the incidence of non-employment has increased among young people who have left education in most countries, with larger increases among those with less than tertiary education. In addition, young workers with medium or low education have also seen widespread increases in the probability of being in low-paid jobs when they do find employment. There are only two countries in which neither of these results hold, i.e. Germany and Poland.

Even young people with high education have seen a worsening of their labour market situation in several countries. Strikingly, following a fall in the probability of being in high-paid jobs, a larger share of them is now found in low-paid than in high-paid jobs on average across the OECD.

While there is some variation in the size of the changes across countries, the main conclusions of this section apply to a large number of countries. These countries were affected to varying degrees by the economic crisis and by 2016 most had been on a recovery path for years. While it is certainly the case that the prolonged and deep recession has played a role in exacerbating the changes in some countries, the results suggest that overall the broad patterns documented in this section cannot be dismissed as the mere temporary product of the recession in all countries. Indeed, most of these results hold on average across the OECD also when adjusting for the economic cycle, even though this evidence can only be seen as tentative since the limited time span covered by the data makes it difficult to isolate the effect of the cycle.35

3.4. Concluding remarks

Understanding how the ongoing transformations of the labour market are affecting different workers is essential to design policies to help make the labour market more inclusive in the future. This chapter has presented new evidence in three key areas of policy relevance, namely job stability, under-employment and the availability of jobs of different pay levels.

While the results differ somewhat between the three topics and across countries, the evidence point to a deterioration of labour market conditions on average across the OECD for two groups, namely the young and those with less than tertiary education. The workers who belong to both of these groups, i.e. the young with less than tertiary education, stand out as those who have seen a deterioration in most of the outcomes considered here and across a large number of countries.

In particular, over the past decade or so, there have been widespread increases in the risk of not being in employment or of being in low-pay employment or under-employment for young people with less than tertiary education. While these changes have affected different countries to varying degrees, only two (Germany and Poland) have not seen a worsening of any of these indicators.36 More highly educated youth have generally fared better than their less-educated peers, but even they have seen increases in the probability of low-pay employment in a number of countries.

These patterns are unlikely to be the short-lived product of the recent global financial crisis. While the OECD countries have been affected by the crisis to a different degree, the majority have been on a recovery path for several years by now and yet the deterioration of labour market conditions for youth is still present in the data. The main conclusions also stand when attempting to adjust for the cycle in spite of the limited time span covered by the data. Overall, therefore, the findings of this chapter point to significant challenges on the road to more inclusive labour markets.

The policy challenges raised by the declining fortunes of young people are twofold. First, there is the need to improve opportunities for the new cohorts entering the labour market. Second, there is the issue of the cohorts of young people who have already suffered worse labour market outcomes than previous generations. A large body of literature has shown that the conditions at the time of labour market entry have a lasting impact on wage and career trajectories over the life cycle.37 This point illustrates that even if part of the changes documented in this chapter are the product of the crisis, they represent a significant policy issue for the future. Helping the cohorts who have faced the tougher environment of the past decade to improve their situation in the future will therefore be a major challenge for inclusive policies.

From a gender perspective, men have seen an increase in the risk of non-employment and under-employment in a number of countries. Nevertheless, the risks of both under-employment and non-employment remain higher for women. Women also remain more likely to be in low-paid jobs and less likely to be in high-paid ones, despite an improvement in the probability of being in middle-paid jobs. While there is a need to better understand the reasons behind the deteriorating labour market outcomes of men, policies to address gender imbalances remain a key priority to build more inclusive labour markets (OECD, 2018[1]).

The scale of the policy challenge at hand requires a multifaceted response. Skill policies can play an important role in improving the labour market experiences of new entrants, as well as supporting career progression for older cohorts. Given the increase in mobility discussed in this Chapter, skills policies will need to adapt to ensure that training programmes also reach those in less stable careers (see Chapter 6). Strengthening social dialogue and better employment regulation can help address imbalances in the employment relationship that may have an adverse impact on more vulnerable workers (see Chapter 4 and 5). Finally, the heightened job insecurity documented in this chapter calls for a review of existing social protection systems to prevent more and more workers from falling through the cracks (see Chapter 7).

References

[18] Bachmann, R., M. Cim and C. Green (2018), “Long-run Patterns of Labour Market Polarisation: Evidence from German Micro Data”, Ruhr Economic Papers, No. 748, RWI.

[21] Beaudry, P., D. Green and B. Sand (2016), “The Great Reversal in the Demand for Skill and Cognitive Tasks”, Journal of Labor Economics, Vol. 34/S1, pp. S199-S247, http://dx.doi.org/10.1086/682347.

[26] Brunner, B. and A. Kuhn (2013), “The impact of labor market entry conditions on initial job assignment and wages”, Journal of Population Economics, Vol. 27/3, pp. 705-738, http://dx.doi.org/10.1007/s00148-013-0494-4.

[25] Burgess, S. et al. (2003), “The class of 1981: the effects of early career unemployment on subsequent unemployment experiences”, Labour Economics, Vol. 10/3, pp. 291-309, http://dx.doi.org/10.1016/S0927-5371(02)00138-0.

[13] Elliot, L. (2017), Robots will not lead to fewer jobs – but the hollowing out of the middle class, The Guardian, https://www.theguardian.com/business/2017/aug/20/robots-are-not-destroying-jobs-but-they-are-hollow-out-the-middle-class (accessed on 11 September 2018).

[11] Euwals, R. and M. Hogerbrugge (2006), “Explaining the Growth of Part-time Employment: Factors of Supply and Demand”, Labour, Vol. 20/3, pp. 533-557, http://dx.doi.org/10.1111/j.1467-9914.2006.00352.x.

[6] Farber, H. (2010), “Job Loss and the Decline in Job Security in the United States”, in Abraham, K., J. Spletzer and M. Harper (eds.), Labor in the New Economy, University of Chicago Press, http://www.nber.org/chapters/c10822 (accessed on 28 July 2017).

[27] Goos, M., A. Manning and A. Salomons (2009), “Job Polarization in Europe”, American Economic Review, Vol. 99/2, pp. 58-63, http://dx.doi.org/10.1257/aer.99.2.58.

[16] Green, A. (forthcoming), “Where are Middle-skill Workers Going?”, OECD Social, Employment and Migration Working Papers, OECD Publishing, Paris.

[8] Hahn, J. et al. (2018), Job Ladders and Growth in Earnings, Hours, and Wages *, https://468ca243-a-0c9971f9-s-sites.googlegroups.com/a/asu.edu/hjanicki/job_ladder_earnings.pdf (accessed on 15 November 2018).

[17] Jaimovich, N., H. Siu and M. Cortes (2017), “Disappearing Routine Jobs: Who, How, and Why?”, Journal of Monetary Economics, Vol. 91, pp. 69-87.

[24] Liu, K., K. Salvanes and E. Sørensen (2016), “Good skills in bad times: Cyclical skill mismatch and the long-term effects of graduating in a recession”, European Economic Review, Vol. 84, pp. 3-17, http://dx.doi.org/10.1016/J.EUROECOREV.2015.08.015.

[2] MacDonald, D. (forthcoming), “Underemployment: Quantity, Quality, and Inclusiveness”, OECD Social, Employment and Migration Working Papers, OECD Publishing, Paris.

[19] Maczulskij, T. and M. Kauhanen (2017), “Where do workers from declining routine jobs go and does migration matter?”, Työpapereita Working Papers, No. 314, Labour Institute for Economic Research.

[3] OECD (2019), Under Pressure: The Squeezed Middle Class, OECD Publishing, Paris, https://dx.doi.org/10.1787/689afed1-en.

[22] OECD (2018), “How does the earnings advantage of tertiary-educated workers evolve across generations?”, Education Indicators in Focus, No. 62, OECD Publishing, Paris, https://dx.doi.org/10.1787/3093362c-en.

[1] OECD (2018), OECD Economic Outlook, Volume 2018 Issue 2: Preliminary version, OECD Publishing, Paris, https://dx.doi.org/10.1787/eco_outlook-v2018-2-en.

[10] OECD (2017), OECD Employment Outlook 2017, OECD Publishing, Paris, http://dx.doi.org/10.1787/empl_outlook-2017-en.

[5] OECD (2014), “Post-crisis pension reforms”, in OECD Pensions Outlook 2014, OECD Publishing, Paris, https://doi.org/10.1787/pens_outlook-2014-5-en.

[15] Pew Research Center (2015), The American middle class is losing ground, https://www.pewsocialtrends.org/2015/12/09/the-american-middle-class-is-losing-ground/.

[4] Prising, J. (2016), Four changes shaping the labour market, The World Economic Forum, https://www.weforum.org/agenda/2016/01/four-changes-shaping-the-labour-market/ (accessed on 8 December 2017).

[20] Salvatori, A. (2018), “The anatomy of job polarisation in the UK”, Journal for Labour Market Research, Vol. 52/1, p. 8, http://dx.doi.org/10.1186/s12651-018-0242-z.

[23] Schwandt, H. and T. von Wachter (2019), “Unlucky Cohorts: Estimating the Long-Term Effects of Entering the Labor Market in a Recession in Large Cross-Sectional Data Sets”, Journal of Labor Economics, Vol. 37/S1, pp. S161-S198, http://dx.doi.org/10.1086/701046.

[7] Topel, R. and M. Ward (1992), “Job Mobility and the Careers of Young Men”, The Quarterly Journal of Economics, Vol. 107/2, pp. 439-479, http://dx.doi.org/10.2307/2118478.

[9] Valletta, R., L. Bengali and C. van der List (forthcoming), “Cyclical and Market Determinants of Involuntary Part-Time Employment”, Journal of Labor Economics.

[14] Vaughan-Whitehead, D., R. Vazquez-Alvarez and N. Maître (2016), “Is the world of work behind middle-class erosion?”, Chapters, pp. 1-61, https://ideas.repec.org/h/elg/eechap/17301_1.html (accessed on 31 January 2018).

[12] Yglesias, M. (2014), Robots won’t destroy jobs, but they may destroy the middle class, Vox, https://www.vox.com/2014/8/23/6057551/autor-job-polarization (accessed on 11 September 2018).

Annex 3.A. Additional results
Annex Figure 3.A.1. Change in adjusted tenure by country and level of education
Percentage change in job tenure (years) by gender, age and education, 2006 to 20171
Annex Figure 3.A.1. Change in adjusted tenure by country and level of education

Note: The OECD average is the unweighted average of the displayed countries. Data are adjusted to control for the composition of the labour force by age, and gender. High education workers have completed a tertiary education. Middle education workers have achieved an upper secondary education and possibly some additional education but less than a bachelor degree. Workers with low education have not completed upper secondary education.

1. Data for Australia, Germany, and the United States are from 2016.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Canadian Labour Force Survey, and the United States Current Population Survey (CPS) Tenure Supplement.

 StatLink http://dx.doi.org/10.1787/888933966540

Annex Figure 3.A.2. Young workers are more likely to be under-employed
Percentage change in the share of dependent workers who indicate they are under-employed, by age, 2006-17 (or latest year)1
Annex Figure 3.A.2. Young workers are more likely to be under-employed

Note: The OECD average is the unweighted average of the countries depicted. Under-employed workers are in part-time employment (working 30 hours or less per week) who report either that they could not find a full-time job or that they would like to work more hours. Young workers are those aged 15 to 29, prime-aged workers are aged 20 to 54, and older workers are aged 55 to 69. The analysis excludes students.

1. Data for Australia, Germany, and Japan are from 2016. Data for Chile and Turkey are from 2015, while Israel data is from 2011. Colombia data for 2006 is from 2007, while Chile uses data from 2009.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), United States Current Population Survey (CPS), Canadian Labour Force Survey, Turkey Labour Force Survey, Japan Household Panel Survey(JHPS/KHPS), Colombia GEIH, Chilean National Socio-Economic Characterization Survey (CASEN), Israel Labour Force Survey, Household, Income and Labour Dynamics in Australia (HILDA) Survey.

 StatLink http://dx.doi.org/10.1787/888933966559

Annex Figure 3.A.3. Change in under-employment by country and gender
Percentage point change in share of dependent workers who indicate they are under-employed, by gender, 2006-17 (or latest year)1
Annex Figure 3.A.3. Change in under-employment by country and gender

Note: The OECD average is the unweighted average of the countries depicted. Under-employed workers are in part-time employment (working 30 hours or less per week) who report either that they could not find a full-time job or that they would like to work more hours. The analysis excludes students.

1. Data for 2017 refer to 2016 for Australia, Germany, and Japan, 2015 for Chile and Turkey are from 2015, and 2011 for Israel. Data for 2006 refer to 2007 for Colombia data and 2009 for Chile.

Source: European labour force survey (EU-LFS), German Socio-Economic Panel (GSOEP), United States Current Population Survey (CPS), Canadian Labour Force Survey, Turkey Labour Force Survey, Japan Household Panel Survey(JHPS/KHPS), Colombian Gran encuesta integrada de hogares (GEIH), Chilean National Socio-Economic Characterization Survey (CASEN), Israel Labour Force Survey, Household, Income and Labour Dynamics in Australia (HILDA) Survey.

 StatLink http://dx.doi.org/10.1787/888933966578

Annex Figure 3.A.4. Jobs by occupation have continued to polarise between 2006 and 2016
Percentage point change in occupational employment shares, 2006-16
Annex Figure 3.A.4. Jobs by occupation have continued to polarise between 2006 and 2016

Note: High-skill occupations are managers, professionals and technicians (ISCO88 codes: 1, 2 and 3). Middle skill occupations are clerks, machine operatives and crafts (codes: 4, 7 and 8). Low-skill occupations are sales and service occupations and elementary occupations (codes: 5 and 9). The OECD average is the unweighted average of all displayed countries. The time period covered is 2006-16, except for: Korea (2006-14), Australia (2006-15). Greece, Portugal and Latvia (2007-2016), Italy (2007-15), Switzerland (2008-15). Chile, Canada, Ireland and Luxembourg (2006-15), Iceland (2006-13)

Source: EU statistics on income and living conditions survey (EU-SILC), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, Chilean National Socio-Economic Characterization Survey (CASEN), and CPS Merged Outgoing Rotation Groups (MORG) for the United States.

 StatLink http://dx.doi.org/10.1787/888933966597

Annex Figure 3.A.5. The risk of low pay has increased for employees with low and medium education in a number of countries
Percentage point change in distribution of jobs by pay level for workers with a low or medium level of education, 2006-16
Annex Figure 3.A.5. The risk of low pay has increased for employees with low and medium education in a number of countries

Note: The OECD average is the unweighted average of all countries shown. Low-paid jobs are those paying less than two thirds of the median wage, while high-paid jobs are those paying more than 1.5 times the median wage. The time period covered is 2006-16, except for: Korea (2006-14), Australia (2006-15). Greece, Portugal and Latvia (2007-16), Italy (2007-15), Switzerland (2008-15). Chile, Canada, Ireland and Luxembourg (2006-15), Iceland (2006-13).

Source: EU statistics on income and living conditions survey (EU-SILC), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, Chilean National Socio-Economic Characterization Survey (CASEN), ENOE longitudinal survey for Mexico and CPS Merged Outgoing Rotation Groups (MORG) for the United States.

 StatLink http://dx.doi.org/10.1787/888933966616

Annex Figure 3.A.6. Young workers have shifted more towards low pay than older ones in some countries
Percentage point change in the distribution of jobs by pay level for younger and prim-age workers, 2006-16
Annex Figure 3.A.6. Young workers have shifted more towards low pay than older ones in some countries

Note: The OECD average is the unweighted average of all countries shown. Low-paid jobs are those paying less than two thirds of the median wage, while high-paid jobs are those paying more than 1.5 times the median wage. Young people are aged 16 to 30 and prime age 31-50. The time period covered is 2006-16, except for: Korea (2006-14), Australia (2006-15). Greece, Portugal and Latvia (2007-16), Italy (2007-15), Switzerland (2008-15). Chile, Canada, Ireland and Luxembourg (2006-15), Iceland (2006-13).

Source: EU statistics on income and living conditions survey (EU-SILC), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, Chilean National Socio-Economic Characterization Survey (CASEN), ENOE longitudinal survey for Mexico and CPS Merged Outgoing Rotation Groups (MORG) for the United States.

 StatLink http://dx.doi.org/10.1787/888933966635

Annex Figure 3.A.7. The risk of being out of employment has risen most for youth who have left education
Percentage point changes in the probability of not being in employment for people who have left education aged 20 to 60 by age, 2006-2016
Annex Figure 3.A.7. The risk of being out of employment has risen most for youth who have left education

Note: The OECD average is the unweighted average of all countries shown. Young people are aged 20 to 30, prime age between 31 to 50 and old from 51 to 60. The time period covered is 2006-2016, except for: Korea (2006-14), Australia (2006-15). Greece, Portugal and Latvia (2007-16), Italy (2007-15), Switzerland (2008-15). Chile, Canada, Ireland and Luxembourg (2006-15), Iceland (2006-13).

Source: EU statistics on income and living conditions survey (EU-SILC), German Socio-Economic Panel (GSOEP), Household, Income and Labour Dynamics in Australia (HILDA) Survey, Korean Labor & Income Panel Study (KLIPS), Canadian Labour Force Survey, Chilean National Socio-Economic Characterization Survey (CASEN), ENOE longitudinal survey for Mexico and CPS Merged Outgoing Rotation Groups (MORG) for the United States.

 StatLink http://dx.doi.org/10.1787/888933966654

Notes

← 1. Choosing an earlier start data does not qualitatively affect the results, and allows consistency with other sections in the Chapter.

← 2. The result is obtained by means of a regression analysis that identifies changes in job tenure between 2006 and 2017, controlling for workers’ age as well as for their level of education and gender.

← 3. Compared with the model that controlled for only demographic shifts, the model controlling for the business cycle (output gap) estimated that tenure had fallen by an additional 4.4 percentage points since 2006. Tenure adjusted for the cycle fell in Canada, Estonia, the Czech Republic, Latvia, and the United Kingdom, whereas tenure adjusted only for demographics had increased in these countries. Controlling for the cycle did not change the direction of adjusted tenure in any countries where adjusted tenure fell when controlling for demographics.

← 4. High education workers have completed a tertiary education. Middle education workers have achieved an upper secondary education and possibly some additional education but less than a bachelor degree. Workers with low education have not completed upper secondary education.

← 5. The available measures of job-to-job transitions capture movements between jobs across consecutive calendar years. Transitions from one year to the next may in fact conceal one (or several) transitions in and out of employment that occur over the course of the reference year. For example, a worker who was employed in the previous year may have experienced a period out of employment before finding the current job, but will be recorded has having experienced a job-to-job transition since job status is recorded yearly. The measure, therefore, should not be interpreted literally as an indicator of direct transitions from a job to another, but rather as a measure of transitions that involve short periods of non-employment between jobs.

← 6. The analysis follows the same technique used for correcting changes in tenure for demographic changes.

← 7. Note that the countries included in the analysis of employment flows does not perfectly match the set of countries examined in the previous tenure analysis due to data limitations. In particular, the flows analysis excludes Canada, Ireland, Norway and Switzerland, and includes Japan.

← 8. There are differences among these countries of course. For example, the changes documented for Denmark here are generally small, and this country is the one with the highest job-to-job transition rate among all those considered in this analysis.

← 9. The term underemployment is sometimes used to refer to workers who are employed in jobs that requires lower qualifications or skills than they possess. As made clear in the text, this section focuses exclusively on underemployment as involuntary part-time employment.

← 10. For a discussion of measurement issues related to underemployment see (MacDonald, forthcoming[2]).

← 11. The unweighted OECD average increased from 4.3% in 2006 to 5.4% in 2017.

← 12. The results reported adjust for the cycle by including a quadratic in unemployment, but the results are qualitatively the same if using a quadratic in output gap. The results reported only control for the share of employment in the service sectors with a high incidence of underemployment, but unsurprisingly the results are very similar when controlling for the overall share of employment in the broader service sector.

← 13. Reliable and comparable data are only available for a handful of countries prior to 2006. At the time of writing, most of the datasets used in the analysis were available up to 2016 at most. The variables considered in this section typically display little year-on-year variation, so it is unlikely that adding one additional year of data would make a substantial difference. In addition, investigations conducted for the analysis in Sections 3.1 and 3.2 (which use different datasets) indicate that the same main patterns are found when using either 2017 or 2016 as end years.

← 14. High skill occupations are managers, professionals and technicians (ISCO88 codes: 1, 2 and 3). Middle skill occupations are clerks, machine operatives and crafts (codes: 4, 7 and 8). Low skill occupations are sales and service occupations and elementary occupations (codes: 5 and 9). The ISCO88 category 6 ("skilled agricultural and fishery workers") are not included in the analysis. This grouping is line with previous practice in the international literature. See OECD (2017[10]) and references therein. Typically, this literature uses average occupational wages as proxy for skills – see for example Goos et al. (2009[27]) for an example covering multiple countries. When using the wage data employed in this Chapter, the average ranking of occupations across countries does indeed reflect the splitting of occupations into low, middle, and high skill groups adopted in this analysis and other contributions.

← 15. See http://www.oecd.org/employment/emp/employmentdatabase-earningsandwages.htm.

← 16. The wage distribution used is that of all employees. However, the main conclusions presented in the analysis also hold when only full-time employees are used to compute median wages.

← 17. Mexico is excluded from this part of the analysis due to inconsistencies in the occupational classification over time.

← 18. As discussed above, middle-paid jobs here are defined as those with wages between 66% and 150% of the median. The finding that the share of middle-paid jobs has not seen a widespread decline across countries holds even when adopting more restrictive definitions of middle-paid jobs. In particular, when eight smaller groups are used (66-80%, 81-90%, and so on until 140-150% of the median), all except one group (jobs with wages between 1.3 times and 1.4 times the median) have seen increases in their share of total employment on average across the 31 countries considered here. Hence, it is clear that, regardless of the precise definition, there has not been a widespread decrease in the share of middle-paid jobs across countries mirroring the ubiquitous decline in the share of middle-skilled jobs that characterises job polarisation.

← 19. The generally small and negative contribution of job polarisation to changes in the share of middle-pay occupations is the result of two partially offsetting forces. The first is the decline in the share of middle-skill occupations that has tended to push the share of middle-paid jobs down. The second is the increase in the share of high-skill jobs that has contributed to the growth of middle-paid jobs because a substantial fraction of these jobs also pays medium-level wages.

← 20. Among the 19 countries where high-paid jobs gained shares, the average growth was less than 1 percentage point. The largest increases occurred in Spain (2.5 percentage points), Luxembourg (2.7 percentage points) and Australia (2.8 percentage points). Only in seven countries was the increase in the share of high-paid jobs in line with (or larger than) the growth expected given the general shift towards high-skill occupations. These countries are Norway, Canada, Germany, the United States, Iceland, Austria and Luxembourg. Among the 12 countries with a declining share of high-paid jobs, the average change was -2.6 percentage points, with the largest values recorded in Italy (-3.4 percentage points), Korea (-7.2 percentage points) and Greece (-8.6 percentage points). All of these countries experienced substantial declines in the propensity of different occupations to pay high wages.

← 21. The countries include all OECD members, which are also members of the European Union, as well as Switzerland, Norway, Iceland and the United States. The analysis uses the following datasets: European labour force survey (EU-LFS), The German Socio-Economic Panel (SOEP) for Germany, and the Current Population Survey (CPS) for the United States.

← 22. The apparent discrepancies between the within-group changes reported in Figure 3.11 and those reported in Figure 3.10 (for the OECD average) are due to the role played by compositional changes. For example, the aggregate decline in the share of workers in high-paid jobs seen for the OECD average in Figure 3.10 is smaller than any of the declines seen for each of the three education groups in Figure 3.11 because the composition of the workforce has shifted towards groups that are more likely to be in high-paid jobs (i.e. older and more educated workers).

← 23. Despite the larger decline in the probability of being in low-pay employment, women remain much more likely to be in low-paid jobs than men (23% vs 16%) and much less likely to be in high-paid jobs (17% vs 25%).

← 24. As referenced in note 21, the apparent discrepancies between the overall (small) decline in the fraction of workers in low pay ( see OECD average in Figure 3.11) and the increase within each of the three education groups (Figure 3.12) is due to the fact the changes in the composition of the workforce have favoured groups that tend to be less concentrated in low pay. A formal decomposition indicates that educational upgrading (i.e. a compositional shift towards workers with higher levels of education) has pushed the share of low-paid workers down by 1.6 percentage points, while the increase propensity of each education group to be in the low-paid employment has added 0.3 percentage points with a net decline of 1.3 percentage points.

← 25. By 2016, the average share employed in low-paid jobs was 37% for the low-education group, 23% for the medium-education group, and 10% for the high-education group.

← 26. One concern is that the changes in the performance of different education groups over time might reflect changes in the unobserved composition of these groups. However, robustness checks show that qualitatively similar results are observed when a single cohort of worker (aged 25 to 45 in 2006 and 35 to 55 in 2016) with a given education level is followed over time. Since the level of education of this cohort is roughly stable over the observation period, this attenuates the concern that the main results might mostly be driven by selection effects.

← 27. The average share of young workers in low pay employment was 35% in 2016, but only 15% and 16% respectively for workers between 30 and 50 and for older workers.

← 28. In the data used in this analysis, the proportion of young people in employment who have low education declined from 22% to 17% between 2006 and 2016 on average across the OECD. The proportion of young people in employment with medium education remained stable at around 47-48%. Such stability makes it implausible that the results reported here stem only or even mostly from the fact that young workers with less than tertiary education are increasingly negatively selected. In fact, as discussed in the main text, the probability of low-paid employment has increased even for young workers with tertiary education in several countries. Moreover, the result in Figure 3.11 show that in the aggregate all education groups have seen some increase in the probability of low-paid employment. Taken together these results point to a generalised change in labour market conditions rather than to a change in selection affecting one particular group. See also note 26 for an additional check that suggests that selection into education levels is unlikely to play a significant role in explaining any of the patterns described here.

← 29. As noted above in the text, only workers whose main economic status is in employment are considered. This attenuates the concern that the results presented might be driven by young people in education taking up part-time jobs. In addition the results on the increase probability of low pay employment for young workers (in general and in particular with low and medium education) hold even when restricting the sample to workers in full-time employment only. This further reduces the risk of the results being heavily affected by young people still in education.

← 30. These estimates are obtained from country-specific regressions of log of wage on the relevant covariates. The reported estimates are therefore in log points and can be interpreted as percentage point changes approximately.

← 31. Note that the figures reported here are not NEET rates, i.e. the proportion of all young people (aged 15-29) who are not in employment, education or training. As described in the main text, the analysis focuses on changes in the incidence of non-employment among those who have left education and the young are defined as workers aged 20 to 30. In addition, due to the data sources used in this analysis, the definition of the employment status also differs from that used in the construction on NEET rates which typically rely on ILO definitions of employment from labour force survey data. Here, for most countries, the definition of employment is instead based on self-reported main economic status.

← 32. A possible observation is that part of the increase in non-employment rates for young people with lower levels of education might be due changes in the composition of this group since, as more people stay in education for longer, those who leave education earlier are increasingly those with poorer labour market prospects. While this analysis cannot rule out this possibility entirely, two facts suggest that the increase in the probability of non-employment for young people who have left education is unlikely to be driven by selection. First, as discussed in note 28, the stability of the share of young people in employment who have medium education suggests that it is unlikely that the composition of this group has changed significantly over this period of time. Second, the deterioration of labour market outcomes described in this section and elsewhere in this chapter is stronger for those with less than tertiary education, but does affect young workers with tertiary education as well (and in fact other age groups are also affected to some extent). While changes in selection may perhaps help explain some of the differences between groups, the fact that the observed changes are not limited to specific subgroups suggests that selection is unlikely to be the main force at play. Finally, it is worth emphasising that even when the deterioration of the labour market conditions for a specific group can be ascribed to changes in its composition driven by selection mechanisms, this remains an important issue that poses a challenge for policies aiming at promoting a more inclusive labour market.

← 33. Young women with high education remain more likely to be in non-employment than their male counterparts (21% vs 16%).

← 34. This figure is effectively a weighted average of the figures for the six young groups reported in Figure 3.14.

← 35. In particular, the average increase in the probability of low pay for the young with medium education across the OECD are robust to controlling either for the output gap or the unemployment rate. The increase in the probability of non-employment for young people who have left education is robust to the inclusion of the unemployment rate, but not to the output gap.

← 36. In particular, Germany and Poland are the only two countries in which the young have not seen an increase in the probability of underemployment, nor in that of non-employment, nor have the young with medium education experienced an increase in the probability of low-pay employment.

← 37. See, among others, Liu et al. (2016[24]), Burgess et al. (2003[25]), Schwandt et al. (2019[23]), Brunner and Kuhn (2013[26]) and references therein.

3. The future of work: New evidence on job stability, under-employment and access to good jobs