COVID-19 and the Youth Labour Market. The Kids are (Probably) Alright

Last week's labour market statistics drew some discussion about rising youth unemployment. It is certainly true that the youth (16-24) unemployment rate is rising faster and further than the unemployment rate for others after Autumn stuttering in the furlough scheme which then unleashed a wave of layoffs focussed on young people (see earlier post ). And yet and yet.... it has been a lot worse before and is unlikely to get very much worse going forward as the beginning of the end of lockdown looks to be approaching and movement - the key to this particular downturn - resumes.

Youth labour market statistics are complicated by the relatively recent rapid growth in the numbers of young people still in education, (since levelled off). The unemployment rate is a measure of the unemployed in the labour force = unemployed/(unemployed+employed). Rising staying on rates will tend to reduce numbers in the labour force and so inflate the unemployment rate for a given number of unemployed.¹ This makes measuring youth unemployment relative to something that will vary less with policy or labour force participation trends a better metric for comparing performance over time. If instead we measure unemployment relative to the entire youth population, we get the u-pop rate. This should be much less susceptible to changes in labour force participation,² and is another useful indicator of youth labour market performance.

In short, it is probably best to look at a range of youth labour market indicators rather than focus on one.

Figure 1 plots both the youth unemployment (left panel) and NEET rates (right panel) in each week throughout 2020 relative to the average of the last 5 years. Anything outside the light grey is a significant departure from recent norms.³ The unemployment and NEET rates are, coincidentally, very similar at around 15% right now. However, even with the latest rise in redundancies, there is not much in either the unemployment or the NEET rate for weeks during October and November and early December (weeks 40 to 49 on the graphs) that we haven't seen in the last 5 years before the COVID downturn.
Figure 1. Youth Unemployment and NEET Rates 2020 relative to 2015-2019 average

Unemployment Rate	NEET Rate

and when we go back further in time, we can see that the youth labour market has been much worse in each of the previous 3 recessions, whether measured by the unemployment rate, the unemployment-to-population rate or the NEET rate - though using the unemployment rate makes the current downturn in the youth labour market look closer to previous downturns. This is not the case with the other two metrics - which suggest now is not as bad as in the past. Rising staying on rates at school, shrinking the labour force and inflating the unemployment rate, also makes the 2010 downturn look the worst of the recent downturns when using the unemployment rate to compare things. Whereas deflating by the population (the u-pop ratio) or using the NEET rate, suggests that the 1980s and 1990 downturns were as bad.
Figure 2. The UK Youth Labour Market Through the Ages

So we have to be careful when trying to assess the state of the youth labour market. The unemployment rate is probably not the best metric to judge things, certainly not by itself. While every unemployment spell is a worry and things aren't brilliant in the youth labour market right now, it is important to recognise that they have been a whole lot worse in the past. And that, the furlough scheme,despite all its faults and blemishes, is probably helping to keep the majority of (young) people in work.

---

1 ^↩ 10 unemployed, 90 employed 50 inactive (education) = unemployment rate =10/(10+90)= 10%, but 10 unemployed, 50 employed 90 inactive = unemployment rate =10/(10+50) = 16.6% .
2 ^↩ Not in EmploymEnt or Training measured as a percentage of the 16-24 year old population.
3 ^↩ More details on how these graphs are worked out can be found here

Long-Term Unemployment and the Covid Downturn

The first signs of rising unemployment under the crisis are beginning to ripple through the labour market. While all unemployment is a worry, we tend to worry even more about the build up of long-term unemployment. We have known at least since the recession of the 1980s that long-term unemployment is harder to escape from and tends to leave lasting scars on both the individuals and the communities in which they live. For governments, this means more resources are needed to try to alleviate the problem.
So, first the good news. We are currently nowhere near the numbers of long-term unemployed that we have seen in the past. Figure 1 tracks the share of unemployed who are long-term, defined as those unemployed for 12 months and more alongside the unemployment rate.¹ You can see that the long-term unemployment share tends to begin rising around 1 year after unemployment starts to rise. This is not too surprising since it takes a year for someone to become long-term unemployed after losing a job. You can also see that long-term unemployment tends to peak around 1 year after the peak of unemployment. This is because the long-term unemployed tend to be at the back of the queue for jobs, so when short-term unemployment begins to drop there are relatively more long-term unemployed still in the queue. Figure 1 shows that, as of 2020 q3, while unemployment had started to rise, the long-term unemployment share was still falling. A renewed inflow of people means relatively more short-term unemployment. So we are only just beginning the process by which long-term unemployment could start to build.
Figure 1. Long-Term Unemployment (12 months +) and Unemployment 1975-2000 So what might we expect to happen in the near future? Figure 2, left panel, looks at (quarterly) inflow rates into unemployment over time.² Despite the large rise in inflows in 2020 compared to 2019, the left panel shows that 2020 is still, as yet, an average year for inflows. Each of the previous three recessions was marked by a larger unemployment inflow that exceeded 2.5% of the labour force. Big inflows generally lead to a rise in long-term unemployment one year later. The relatively good news for 2020 is that inflow rates into unemployment in 2020 are not, yet, nearly as big as the inflows in past recessions. But they are rising rapidly. So it may not augur well if this pattern continues.

Even a relatively moderate inflow could still lead to long-term unemployment later if there are subsequently few job opportunities for the unemployed to access. This is because the build-up of long-term unemployment also depends on the ability to escape unemployment quickly. The right panel of Figure 2 shows the percentage of the newly unemployed in each year who find work over the subsequent 12 months.³ The year 2009 stands out as a year in which the chances of escaping unemployment were much lower than in years when the economy was growing nicely, (like 2005 or 2014). The graph for the 2019 unemployed entry cohort is interesting. In the first few months of unemployment in 2019, the chances of finding work were very high compared to other years. This is consistent with the relatively healthy state of the labour market throughout 2019, offering plenty of job opportunities. However the line starts to taper off in months 9 and 12 as the crisis begins and hiring slows down rapidly (see previous post). So if the economy stalls during an unemployed spell, it becomes harder for all the unemployed to find work and long-term unemployment will then build up.
Figure 2. Unemployment Inflows and the Chances of Finding Work The other thing that matters is the composition of the unemployed. We have known for a long time that certain characteristics, like age, gender, skills, region and ethnicity, are associated with a greater or lesser likelihood of both becoming unemployed and staying unemployed. Young people, for example, are more likely to become unemployed but less likely to become long-term unemployed. But, as Figure 3 shows, it was ever thus. So the composition of the unemployed inflow this time round doesn't look very different.⁴ Which means the likely long term unemployment trend this year depends more on the future level of demand.
Figure 3. Who Becomes Unemployed?

So three things will determine whether long-term unemployment becomes a problem next year. The number of people laid off, the composition of the newly unemployed and the number of job opportunities that arise over the year. At the moment, the latter looks likely to matter more. Which means trying to keep demand bouyant and helping ensure there are enough job opportunities in the subsequent months. It is a lot easier to find work if demand in the wider economy holds up.

---

1 ^↩ Using the official ILO/OECD classification. There is however no official definition of long-term unemployment. The Figures are all compiled using the grossing weights supplied by the Labour Force Survey (LFS).
2 ^↩ They are measured as a percentage of the labour force rather than as a headcount to net out population growth over time.
3 ^↩ The data are derived from the LFS longitudinal panel.
4 ^↩ The numbers are based on a regression of the chances of becoming unemployed. The coefficients are percentage point differences relative to the missing group (eg 25-64 year olds) net of the other characteristics. Figure 3 suggests the newly unemployed are a bit less non-white, a bit more female and graduate perhaps than in the past, but not much.

1 Million Missing Immigrants (And 1.2 Million More UK-Born)

A new Kings College post by Michael O'Connor highlighted the strange case of a rapid fall of around 750,000 in the immigrant population of working age over the last year contained inthe latest ONS labour market statistics . This is very remarkable indeed.

It is all the remarkable still, when one realises that the same survey on which these numbers are based - the Labour Force Survey (LFS) - also shows a fall in the total immigrant population (age 0 upward) of 1 million between 2019 q4 and 2020 q3 AND a rise in the UK-born population of 1.25 million over the same period.

To see how unusual these changes in population are, take a look at the Figures below. Figure 1, left panel, shows the immigrant population (defined as born outside the UK) since 2007.¹ Immigration usually falls a little in recessions. Fewer job opportunities reduce the incentive to come to the UK. Job losses encourage some migrants to leave the UK to seek better fortune elsewhere. During the 2008-2011 downturn there was a little wobble in the immigrant population, but not much to alter the longstanding upward trend in immigration that began in 1995. There were larger falls in the immigrant population after Brexit. But nothing compares to the size of the fall that appears to have gone on during the 9 months of the COVID crisis.

Not only this, the UK-Born population appears to have jumped by around 1.2 million over the same period. This is remarkable indeed. No such period of rapid population growth can be observed in the last 15 years.

Figure 1. UK-Born and Immigrant Populations in the UK 2007-2020 Figure 2 graphs the same data but using the yearly change to get a sense of how unusual these developments are. The yearly growth in the UK-born population was averaging around 350,000 a year prior to 2019 and then...kapow.
The overseas born population was growing at a slighty larger rate and the comes Brexit and then.. kapow.
Figure 2. Yearly Change in the UK-Born and Immigrant Populations in the UK 2007-2020

Figure 3 breaks down the Overseas and UK-born population changes by 10 year age bands.² You can see upticks in the growth of several of the UK-born groups eg. 41-50 after the end of 2019. And downturns in the overseas-born age groups eg 31-40 What is even odder is that if we align these age groups, then the population of UK-born 31-40 year olds ten years ago should equal the population of 41-50 year olds now, give or take some deaths and emigration. A 30 year-old in 2010 is a 40 year old now. Figure 4 plots the actual and predicted population of UK-born 41-50 year olds over time. The lines generally run paralell. The prediction is above the actual population, as expected, but it is paralell - until the start of 2020 when they suddenly diverge.
Figure 4. Predicted and Actual Age Cohort Populations: UK-Born Age 41-50 2017-2020

So something really unusual is going on. It may be that the COVID crisis has simultaneously reduced migrant inflows and increased outflows by an unprecedented amount while at the same time uncovering a new strata of the UK-born population. Or it may be something lurking in the LFS to do with sampling of households over the crisis and the subsequent impact on the population weights.³ Indeed the two issues are related. Stay with this, it gets slightly esoteric for the next two sentences.

The LFS sample numbers are grossed up to the Census projections for the total UK population in each year. If the LFS picks up a decline in immigrants in its sample but the census population projection is unchanged, then the LFS weights will inflate the UK-born population to compensate for the missing immigrants. Got that? The ONS believes that its huge fall in immigrant numbers picked up in the sampling is real. But the Census projection for the total population in 2020 remains unchanged. As such the automatic rebalancing in the weights kicks in. So that explains why the UK-born population grows quite rapidly in 2007/18 in Figure 2 above -a period when it seemed the immigrant population was falling after Brexit. It also explains why the estimated UK-born population takes off into the stratosphere in 2020 in Figure 20. So the population of UK-born individuals is almost certainly off. We will look into how much this affects labour market statistics in future posts. But all this probably demands more attention.

---

1 ^↩The LFS data contain population grossing weights that allow the sample to be grossed up to the population. The Figures are all compiled using the grossing weights supplied by the LFS.
2 ^↩So Age 10 is 0-10 years; 20 is 11-20 years etc
3 ^↩The ONS has acknowledged that a re-weighting process to deal with changing responses to the LFS during the crisis is underway

Who's being made redundant during COVID? Who is staying at work. And is it any different from usual?

Last week's post looked at trends in redundancies over the COVID period relative to previous downturns. It may also be helpful to get a sense of just who is being laid off and whether this is any different from a "normal" recession. Many factors are associated with the risk of layoff, but two features that often get a lot of attention are age and skills. Essentially, UK redundancy law ensures that younger people are easier and cheaper to fire, (because they will have been with a firm for a shorter amount of time, on average). Also people with fewer skills are cheaper to fire, (because they get paid less, on average, and the amount of redundancy pay depends on how much a worker earns prior to being laid off). But all that is, as economists are always saying, holding other things equal and in the UK labour market things are not always that equal.
Look at the left panel in Figure 1 below. The Figure uses Labour Force Survey, (LFS), data to track the percentage of the workforce in each age group who say they have been made redundant in the last 3 months, running from 2005 until August 2020. The Figure peaks at the onset of the last recession and rises again in COVID. Thankfully the peak is no more than 1% of the relevant population. But the thing to take home is that young people, (by young we mean under 35), appear to be only at greater risk of redundancy in big recessions. Outside of recessions, it seems young people may be less likely to be laid off than anyone else.¹ The COVID downturn is notable in that 16 to 24 year olds, in particular, are much more likely to have been laid off. This may have something to do with the sectors in which young people work being more badly hit than usual.²
The right panel does the same but for education levels. Here it is more apparent that redundancies, in whatever period, are less likely to happen for graduates and are instead concentrated among those with Level 4 and Level 3 qualifications. Those with Level 2 and Level 1 qualifications are also less likely to be laid off in "normal" times. Only in a deep recession does the chance of being laid off rise steeply for this group.
Figure 1. UK Layoff Rates by Age and Qualifications 2005-2020 Figure 2 tracks the percentage of the workforce of each of these same groups who leave their jobs voluntarily over the same period. The left panel shows that job quitting is the preserve of the young. This shouldn't be too surprising, since most attempts to find a good job occur when younger. The Figure also shows that quits are also much more seasonal among young people. The peak quitting season is Winter. However, the "quit-age-gap" narrows noticeably in recessions and the COVID recession is no exception.
In contrast, the incidence of job quits do not vary much by education, with the exception of workers with Level 1 or 2 education, quit rates are much the same for all other skill groups, whether falling in a recession or rising in a growth period. Level 1 and 2 groups however are much less likely to quit their jobs whatever the state of the economy. Once again however, COVID appears to be a leveler. All education groups are much less likely to have quit their jobs during COVID.
Figure 2. UK Quit Rates by Age and Qualifications 2005-2020

---

1 ^↩According to the responses contained in the LFS.
2 ^↩Alas, the data on redundancies by industry are not available that would enable us to look at this.

Redundancies and the Labour Market Under COVID. Has Furlough Helped? And Other Related Things You May have Missed

The government's furlough scheme has almost certainly helped prevent a large rise in unemployment over the crisis (so far). This week's announcement of its prolongation under the latest lockdown is almost certainly a good thing for jobs. Despite this, last week's ONS review of the recent patterns of redundancies showed that, by the end of August, layoffs were running at their highest level since the previous recession of 2008-2011.¹ The beginning of the end of the (original) furlough scheme certainly coincides with the recent spike in layoffs, as can be seen from the blue line in the Figure below, which replicates the ONS redundancy data series. Employers were obliged to pay the national insurance contributions of their furloughed employees from August 1st, with further reductions in support and obligation for employers to top up wages in September. It seems that this then was the trigger for the rise. Before this, during the first few months of COVID, there is little sign of an upturn in layoffs, which suggest that the furlough scheme in its original form did its job.
Redundancies are often seen as commensurate with unemployment - but not always. UK labour laws  require most people to have some notice of an impending layoff.² The length of notice rises by one week for every year at the firm, up to a maximum of 12 weeks notice, which does at least allow some time to look for another job while still employed. In normal times, between 40% and 50% of those made redundant are back in work within three months.³ The trouble with recessions is that not many firms hire, particularly not in the current crisis (see earlier SWOB blog ) so it is harder now to find alternative employment for anyone laid off. The Labour Force Survey (LFS) suggests that currently 30% of those laid off in the last three months were in work by August. This percentage is similar to that seen in the 2008-2011 recession. So bad, but not unprecedented.
But no-one is leaving
An important related feature of the labour market is that, alongside redundancies, lots of people move jobs voluntarily all the time. Typically this is because people have found a better job, (better in terms of pay or location or hours or just general stuff), or have quit in the hope of finding something better soon. Firms can often use this "normal" turnover to adjust their workforces without the need for mass layoffs. The red line in the left-hand Figure shows the pattern of these job quits over time. Several things stand out. Quits are typically three to four times as large as layoffs. Quits are also quite seasonal. The blips in the red line correspond to the autumn and winter quarters, when quits are at their highest in the year.⁴
But the key feature in the Figure is that quits have fallen dramatically during the COVID crisis - to their lowest level for 15 years. Job quits always go down in recessions. Fewer firms hire during a crisis, so many fewer people are willing and able to quit their jobs for something different. But this time round, the decline in quits is larger than in the previous crisis.
Why does this matter for unemployment? If we add the numbers who quit into unemployment with the numbers who are laid off and end up in unemployment, then the right-hand panel of the Figure shows that the total entering unemployment is still much, much lower than in the 2008-2011 downturn.⁵ This can help explain why unemployment has not risen so sharply (up to the end of August).
Whether this persists and who is being laid off will be pursued in future blogs

Figure. UK Layoffs and Voluntary Separations 2006-2020 Thing 2

---

1 ^↩ Layoffs in 2008-2011 were in turn much lower than in the 1980s and 1990s recessions because many more people took wage cuts in 2008-2011 and that helped avoid many layoffs
2 ^↩Source: Redundancy pay, however requires at least 2 years at a firm, before rising with length of service at the firm. 3 ^↩Source: my calculations using Labour Force Survey (LFS)
4 ^↩Redundancies are much less seasonal. source: my calculations using LFS
5 ^↩THe ONS definition of redundancies excludes "dismissals" and those whose temporary job has ended. These flows have been included in the total separations graph in the right panel

Two Things About Ethnicity Pay Gaps that May Have Passed You By

Last week the ONS published a study  setting out recent trends in ethnicity pay gaps. Despite lots of caveats in the article, many papers duly reported the end of pay gaps by ethnicity. There are reasons to think, however, that we may not be there yet.
Thing 1
Look at the left hand panel in the Figure below. The Figure essentially reproduces the Figure in the ONS study¹ - which does indeed show that the raw ethnicity pay gap (ie unadjusted for any possible explanatory factors) had fallen to near zero by 2019. However there are many reasons not to look at the raw pay gap. This is largely because ethnic minorities in the UK are still not equally distributed in many dimensions, by geography and education for example. If these factors also determine pay, as region and education do, then we are missing something fundamental when looking at the raw gap. The top line in the left hand panel of the graph duly "nets out" some of these things, (age, gender, qualifications and geography), revealing the pay gap after these factors are taken into account. And it is quite a lot bigger than the raw gap (about 10 percentage points bigger). What it means is that even in 2019, individuals from ethnic minorities were paid worse - about 12% worse - than their white peers with the same age, gender and and region of residence. So not quite there yet.

Figure. Mean Hourly Ethnicity Pay Gap Relative to White: With & Without Controlling for Other Stuff & by Migrant Status Thing 2
Another important feature of the ethnicity pay gap often overlooked in the debate, is that it matters whether the individual was born in the UK or outside the UK. Around half of all non-white ethnic minorities in the UK were born abroad.² The second panel plots the same ethnicity pay gap net of controls as in the left hand panel (middle line) alongside the equivalent for individuals born in the UK and those born outside the UK. The pay gap is much much bigger for non-white ethnicity immigrants than it is for non-white ethnicity individuals born here. Indeed for the latter the pay gap is closer to zero and has been so for some time. This is an issue for any study that looks at average differences between groups. There are, of course, other important dimensions to this debate which we will return to in later blogs. But the message here is that often averages can obscure important features. In this case, just looking at ethnicity as a single group misses out on the interaction of immigration and pay.

---

1 ^↩The Figure uses mean rather than the median hourly pay - but this makes very little difference to the story
2 ^↩Source: Annual Population Survey

UK Labour Market Under COVID Part 3: Layoffs, Hiring & Wages

Recessions, usually signify a rise in layoffs. Even the last recession, which was accompanied by much more wage moderation than previous ones, generated a significant rise in redundancies. This time it is a little different. Furloughing allows firms to postpone difficult decisions. This can be seen in Figure 1 below which tracks layoffs week by week through the first 26 weeks of 2020 relative to the norms of the past 5 years. While there may have been a small upturn in layoffs from around week 16, it is hard to discern a pattern to weekly layoffs that lies outside the norms of the last 5 years. And since the last 5 years were not recessions, it is hard to see much evidence of layoffs so far.

Figure 1. Layoffs in 2020 Week by Week Relative to Norm

Nor is there much evidence of wage cuts in the face of the crisis. We know hours have fallen back and this may affect weekly wages, so to net this hours drop out out we can look at hourly wages. Figure 2 plots the level of hourly wages, adjusted for inflation, week by week. While there is a deal of volatility in these numbers, due to smaller samples, the overall picture is of continued real wage growth over the course of the crisis.

Figure 2. Hourly Wages in 2020 Week by Week Relative to Norm

So people may not have been fired or had their wages cut so far, but there haven't been many people hired. This, in addition to workplace absences, is perhaps the most noticeable feature of the UK labour market during the crisis so far. Figure 3 plots the proportion of employees who have been in new jobs for under one month as a proxy for hiring. Between 2% to 4% of the workforce is usually newly-hired in any week, with fewer hires typically in Spring and Autumn being the main hiring period. In uncertain times, many firms try to make do with their existing workforce rather than expand. They may not even replace staff who leave. Hiring rates in the spring and summer of 2020 were some 1 to 2 percentage points below seasonal norms in the weeks leading up to the lockdown and thereafter. In short, hiring stalled over the crisis. Cumulative hires were some 40.5 percentage points lower than usual by the end of week 26. This means around 4 million fewer hires than might be expected. Many firms are having to make do with existing staff.

Figure 3. Hiring in 2020 Week by Week Relative to Norm