Latest resesarch from the independent American think tank The Conference Board paints a worrying picture about the growth of UK labour productivity. While global growth in labour productivity has weakened following the financial crisis, its weakness in the UK is singled out in the Board’s 2019 Productivity Brief. It finds that amongst large mature economies the decline in labour productivity growth rates has been greatest in the UK. This has important implications for the country’s longer-term well-being and, specifically, it peoples’ living standards.
The UK saw the growth in real GDP (national output) fall from 1.8 per cent in 2017 to 1.4 per cent in 2018. The Conference Board predicts that this will fall further to 0.8 per cent in 2019. In the context of living standards, the growth in real GDP per capita is particularly important. An increase in the population will, other things being equal, lower living standards because more people will be sharing a given amount of real national income. The growth in real GDP per capita fell from 1.1 per cent in 2017 to 0.7 per cent in 2018 and is predicted to fall to just 0.1 per cent in 2019.
Chart 1 shows the annual rates of growth in real GDP and real GDP per capita from the 1950s. The average growth rates are 2.4 and 1.9 per cent respectively. The other series shown is the annual growth in real GDP per person employed. This is a measure of the growth in labour productivity. Its average annual growth rate is also 1.9 per cent. This illustrates the intrinsic long-run relationship between labour productivity growth and the growth rate of GDP per capita and hence in general living stanadards. (Click here to download a PowerPoint copy of the chart.)
In the short term, rates of growth in output per worker (labour productivity) and GDP per capita (general living standards) can be less similar. For example, when unemployment rates rise labour productivity rates may be little affected despite GDP per capita falling. Nonetheless, the important point here is the close long-run relationship between the growth in labour productivity and GDP per capita. This then raises an important question: what factors contribute to the growth in output and labour productivity?
An approach known as growth accounting helps to identify four key contributors to the growth of total output. The first is the quantity of labour, commonly measured in labour hours. The second is the quality of labour, also known as labour composition. Third is capital services which are physical inputs into production and include machinery, structures and IT capital. Capital services are affected by quantity and quality, but, unlike labour, it is practically more difficult to separate out these dimensions. Fourth, is Total Factor Productivity (TFP).
TFP it is essentially the residual contribution to output growth that cannot be explained by changes in the quantity and quality of the individual inputs. Hence, in principle, it is capturing changes in how effectively the labour and capital inputs are being employed and combined in production. The Conference Board’s Productivity Brief describes the growth in TFP as providing ‘a more accurate picture of the overall efficiency by which capital, labour and skills are combined in the production process’.
Chart 2 shows Conference Board estimates of the percentage point contribution of these four sources of growth since 1990. Over this period, output growth averaged 2 per cent per year. The contribution of capital services and, hence, what is known as capital accumulation is particularly significant at 1.5 percentage points per year. This has been significantly larger than the contribution of labour hours which averaged only 0.3 percentage points per year since 1990. This evidences the importance played by capital deepening for output growth in the UK. (Click here to download a PowerPoint copy of the chart.)
Capital deepening captures the growth in capital services relative to the growth in the labour input. It takes on even greater significance when we think about the growth in labour productivity since, after all, this is the growth in output relative to the quantity of labour. It is significant though that since 2015 the growth of capital services has contributed only 1 percentage point to output growth while the growth of labour hours has contributed an average of 0.7 percentage points. This points to a slowdown in capital deepening and hence in the growth of labour productivity.
Chart 2 also illustrates the importance of TFP growth to overall output growth. It is also important (along with capital deepening and the growth in labour quality) for the growth in labour productivity. Interestingly, we observe significant fluctuations in the growth of TFP. This is thought to reflect fluctuations in the utilisation of inputs. For example, if the utilisation of inputs falls (rises) when output falls (increases) this will be mirrored by a disproportionately large fall (increase) in TFP. In the longer-term, however, changes in TFP capture aspects of technological progress and advancement that enable more effective production methods and techniques to be deployed. In other words, the growth of TFP captures the ability of production to benefit from the advancement in ideas, products, processes and know-how.
A decline in the growth in TFP growth following the financial crisis is found quite widely in mature economies. The annual rate of growth of TFP across mature economies fell from 0.5 per cent year in 2000-2007 to 0.2 per cent in 2010-2017. In the UK this fall was from 0.5 per cent to -0.1 per cent. Hence, the decline in TFP growth of 0.6 percentage points between 2010 and 2017 was double the 0.3 percentage point fall across all mature economies. In 2018 the Conference Board estimate that TFP in the UK fell by 0.1 percent further exacerbating the downward pressure on labour productivity.
As our final chart shows, it is the magnitude to which labour productivity has eased following the financial crisis that sets the UK apart. While across all mature economies the growth of output per labour hour (another measure of labour productivity growth) fell from an average of 2.3 per cent per year in 2000-2007 to 1.2 per cent in 2010-2017, in the UK the fall was from 2.2 per cent to 0.5 per cent per year. (Click here to download a PowerPoint copy of the chart.)
While the productivity problem facing the UK is not new, the latest figures comes as a very timely reminder of the extent of the problem. To some extent the uncertainty around Brexit and the negative impact on capital accumulation has only helped to exacerbate the problem. But, this may mask a more systemic problem facing the UK. Getting to the root of this problem matters. It matters most significantly for our long-term wellbeing and prosperity. The productivity gap with our major industrial competitors is a gap that policymakers need not only to be mindful of but one that needs closing.
- What do you understand by the term labour productivity. How could we measure it?
- Why is it important to look at the growth of output per capita when assessing the benefits of long-term growth?
- Why is labour productivity important for the long-term well-being of a country?
- What do you understand by the method of growth accounting?
- What is the distinction between capital accumulation and capital deepening?
- What might explain why the growth of labour productivity has been lower in the years following the post-financial crisis?
- What do you understand by Total Factor Productivity (TFP)?
- What does the long-term growth of TFP attempt to capture?
- If you were an economic advisor to the government, what types of policy initiatives might you recommend for a government concerned about low rates of growth of labour productivity?
The distinction between nominal and real values is an incredibly important one in economics. We apply the latest GDP numbers from the ONS to show how the inflation-adjusted numbers help to convey the twin characteristics of growth: positive longer-term growth but variable short-term rates of growth. It is real GDP numbers that help us to understand better the macroeconomic environment and, not least, its inherent volatility. To use nominal GDP numbers means painting a less than clear, if not inaccurate, picture of the macroeconomic environment.
The provisional estimate for GDP (the value of output) in the UK in 2018 is £2.115 trillion, up 3.2 per cent from £2.050 trillion in 2017. These are the actual numbers, or what are referred to as nominal values. They make no adjustment for inflation and reflect the prices of output that were prevailing at the time. Hence, the figures are also referred to as GDP at current prices.
The use of nominal GDP data can be something of a problem when we compare historical values. In 1950, for example, as we can see from Chart 1, nominal GDP in 1950 was a mere £12.926 billion. In other words, the nominal figures show that the value of the country’s output was 163.595 times greater in 2018 (or an increase of 162,595 per cent). However, if we want to make a more meaningful comparison of the country’s national income we need to adjust for inflation. (Click here to download a PowerPoint copy of the chart.)
If we measure GDP at constant prices we eliminate the effect of inflation. This allow us to make a more meaningful comparison of national income. Consider first the real GDP numbers for 1950 and 2018. GDP in 1950 at 2016 prices was £373.9 billion. This is higher than the nominal (current-price) value because prices in 2016 were higher than those in 1950. Meanwhile, GDP in 2018 when measured at 2016 prices was £2.034 trillion. This real value is smaller than the corresponding nominal value because prices in 2016 where lower than those in 2018.
Between 1950 and 2018 there was a proportionate increase in real GDP of 5.439 (or a 443.9 per cent increase). Because we have removed the effect of inflation the real growth figure is much lower than the nominal growth figure. Crucially, what we are left with is an indicator of the growth in the volume of output. Whereas nominal growth rates are affected both by changes in volumes and prices, real growth rates reflect only changes in volumes.
Consider now output growth between 2017 and 2018. As we saw earlier, the nominal figures suggest growth of 3.2 per cent. In fact, GDP at constant 2016 prices increased from £2005.4 trillion in 2017 to £2,033.6 trillion in 2018: an increase of 1.4 per cent. This was the lowest rate of growth in national output since 2012 when output also grew by 1.4 per cent. In 2017 national output had increased by 1.8 per cent, the same increase as in 2016.
To put the recent growth in national output into context, Chart 2 shows the annual rate of growth in real GDP each year since 1950. Across the period, the average annual rate of growth in real GDP and, hence, in the volume of national output was 2.5 per cent. In the current decade growth has averaged only 1.9 per cent. This followed falls of 0.3 per cent and 4.2 per cent in 2008 and 2009 respectively as the effects of the financial crisis on the economy were felt. (Click here to download a PowerPoint copy of the chart.)
By plotting the percentage changes in real GDP from year to year, we get a much clearer sense of the inherent instability that we identified at the outset as a characteristic of growth. This is true not only for the UK, but economies more generally. This instability is the key characteristic of the macroeconomic environment. It influences and informs much of what we study in economics.
The variability of growth rates that create the instability of economies again requires an understanding of the distinction between nominal and real GDP. Chart 3 illustrates the growth in GDP both in nominal and real terms. The average annual rate of growth of nominal GDP is 7.8 per cent, considerably higher than the average real growth rate of 2.5 per cent per year. The difference again reflects the effect of rising prices. (Click here to download a PowerPoint copy of the chart.
Chart 3 clearly shows the wrong conclusions that can be drawn if one was to focus on the growth in nominal GDP from year to year. Perhaps the best example is 1975. In this year nominal GDP grew by 24.2 per cent. However, the volume of national output contracted: real GDP fell by 1.5 per cent. The growth in nominal GDP reflects the rapid growth in prices seen in that year. The economy’s average price level (the GDP deflator) rose by 26.1 per cent. Hence, the growth in nominal GDP reflected not an increase in the volume of output – that fell – but instead a large increase in prices.
The importance of the distinction between nominal and real GDP is further demonstrated by the fact that since 1950 nominal GDP has fallen in only one year. In 2009 nominal GDP fell by 2.7 per cent. The 1.6 per cent rise in the economy’s average price level was not enough to offset the fall in the volume of output of just over 4.2 per cent. In other years when the volume of output (real GDP) fell, the effect of rising prices meant that the value of output (nominal GDP) nonetheless rose.
So to conclude, the distinction between nominal and real GDP is crucial when analysing economic growth. To understand the distinction gives you a truly real advantage in making sense of the macroeconomic environment.
- What do you understand by the term ‘macroeconomic environment’? What data could be used to describe the macroeconomic environment?
- When a country experiences positive rates of inflation, which is higher: nominal economic growth or real economic growth?
- Does an increase in nominal GDP mean a country’s production has increased? Explain your answer.
- Does a decrease in nominal GDP mean a country’s production has decreased? Explain your answer.
- Why does a change in the growth of real GDP allow us to focus on what has happened to the volume of production?
- What does the concept of the ‘business cycle’ have to do with real rates of economic growth?
- When would falls in real GDP be classified as a recession?
- Distinguish between the concepts of ‘short-term growth rates’ and ‘longer-term growth’.
- Why might the distinction between nominal and real be important when analysing changes in people’s pay? What would be the significance of an increase in real pay?
How would your life be without the internet? For many of you, this is a question that may be difficult to answer – as the internet has probably been an integral part of your life, probably since a very young age. We use internet infrastructure (broadband, 4G, 5G) to communicate, to shop, to educate ourselves, to keep in touch with each other, to buy and sell goods and services. We use it to seek and find new information, to learn how to cook, to download music, to watch movies. We also use the internet to make fast payments, transfer money between accounts, manage our ISA or our pension fund, set up direct debits and pay our credit-card bills.
I could spend hours writing about all the things that we do over the internet these days, and I would probably never manage to come up with a complete list. Just think about how many hours you spend online every day. Most likely, much of your waking time is spent using internet-based services one way or another (including apps on your phone, streaming on your phone, tablet or your smart TV and similar). If your access to the internet was disrupted, you would certainly feel the difference. What if you just couldn’t afford to have computer or internet access? What effect would that have on your education, your ability to find a job, and your income?
Martin Jenkins, a former homeless man, now entrepreneur, thinks that the magnitude of this effect is rather significant. In fact, he is so convinced about the importance of bringing the internet to poorer households, that he recently founded a company, Neptune, offering low-income households in the Bronx district of New York free access to online education, healthcare and finance portals. His venture was mentioned in a recent (and very interesting) BBC article – a link to which can be found at the end of this blog. But is internet connectivity really that important when it comes to economic and labour market outcomes? And is there a systematic link between economic growth and internet penetration rates?
These are all questions that have been the subject of intensive debate over the last few years, in the context of both developed and developing economies. Indeed, the ‘digital divide’ as it is known (the economic gap between the internet haves and have nots) is not something that concerns only developing countries. According to a recent policy brief published by the New York City Comptroller:
More than one-third (34 percent) of households in the Bronx lack broadband at home, compared to 30 percent in Brooklyn, 26 percent in Queens, 22 percent in Staten Island, and 21 percent in Manhattan.
The report goes on to present data on the percentage of households with internet connection at home by NYC district, and it does not take advanced econometric skills for one to notice that there is a clear link between median district income and broadband access. Wealthier districts (e.g. Manhattan Community District 1 & 2 – Battery Park City, Greenwich Village & Soho PUMA), tend to have a significantly higher share of households with broadband access, than less affluent ones (e.g. NYC-Brooklyn Community District 13 – Brighton Beach & Coney Island PUMA) – 88% of total households compared with 58%.
But, do these large variations in internet connectivity matter? The evidence is mixed. On the one hand, there are several studies that find a clear, strong link between internet penetration and economic growth. Czernich et al (2011), for instance, using data on OECD countries over the period 1996–2007, find that “a 10 percentage point increase in broadband penetration raised annual per capita growth by 0.9–1.5 percentage points”.
Another study by Koutroumpis (2018) examined the effect of rolling out broadband in the UK.
For the UK, the speed increase contributed 1.71% to GDP in total and 0.12% annually. Combining the effect of the adoption and speed changes increased UK GDP by 6.99% cumulatively and 0.49% annually on average”. (pp.10–11)
The evidence is less clear, however, when one tries to estimate the benefits between different types of workers – low and high skilled. In a recent paper, Atasoy (2013) finds that:
gaining access to broadband services in a county is associated with approximately a 1.8 percentage point increase in the employment rate, with larger effects in rural and isolated areas.
But then he adds:
most of the employment gains result from existing firms increasing the scale of their labor demand and from growth in the labor force. These results are consistent with a theoretical model in which broadband technology is complementary to skilled workers, with larger effects among college-educated workers and in industries and occupations that employ more college-educated workers.
Similarly, Forman et al (2009) analyse the effect of business use of advanced internet technology and local variation in US wage growth, over the period 1995–2000. Their findings show that:
Advanced internet technology is associated with larger wage growth in places that were already well off. These are places with highly educated and large urban populations, and concentration of IT-intensive industry. Overall, advanced internet explains over half of the difference in wage growth between these counties and all others.
How important then is internet access as a determinant of growth and economic activity and what role does it have in bridging economic disparities between communities? The answer to this question is most likely ‘very important’ – but less straightforward than one might have assumed.
- Comptroller, New York City, Internet Inequality
- Czernich, N., Falck, O., Kretschmer, T. and Woessmann, L., 2011, Broadband infrastructure and economic growth, The Economic Journal, 121(552), pp.505–32
- Koutroumpis, P., 2018, The economic impact of broadband: evidence from OECD countries, Ofcom
- Atasoy, H., 2013, The effects of broadband internet expansion on labor market outcomes, ILR Review, 66(2), pp.315–45
- Forman, C., Goldfarb, A. and Greenstein, S., 2009, The Internet and Local Wages: Convergence or Divergence? (No. w14750), National Bureau of Economic Research
- Is there a link between economic growth and internet access? Discuss, using examples.
- Explain the arguments for and against government intervention to subsidise internet access of poorer households.
- How important is the internet to you and your day to day life? Take a day offline (yes, really – a whole day). Then come back and write about it.
One of the announcements in the recent UK Budget was the ending of the Private Finance Initiative (PFI), including its revised form, PF2. PFI was introduced by the Conservative government in 1992. Subsequently, it was to become central to the Labour government’s ‘Third-way’ approach of using the private sector to deliver public projects and services.
PFI involves a public–private partnership (PPP). The private sector builds and/or runs public projects, such as new schools, hospitals, roads, bridges, student accommodation, and so on. The public sector, in the form of government departments, NHS foundation trusts, local authorities, etc., then pays the private sector company a rent for the infrastructure or pays the company to provide services. The benefit of PFI is that it allows private-sector capital to be used for new projects and thus reduces the need for government to borrow; the disadvantage is that it commits the public-sector body to payments over the long-term to the company involved.
As the chart shows, PFI became an important means of funding public service provision during the 2000s. In the 10-year period up to financial year 2007/08, more than 50 new projects were being signed each year.
As the number of projects grew and with them the long-term financial commitments of the public sector, so criticisms mounted. These included:
- Quality and cost. It was claimed that PFI projects were resulting in poorer quality of provision and that cost control was often poor, resulting in a higher burden for the taxpayer in the long term.
- Credit availability. PFI projects are typically dependent on the private partner using debt finance to acquire the necessary funds. Therefore, credit conditions affect the ability of PFI to fund the delivery of public services. With the credit crunch of 2008/9, many firms operating PFI projects found it difficult to raise finance.
- The financial health of the private partner. What happens if the private company runs into financial difficulties. In 2005, the engineering company Jarvis only just managed to avoid bankruptcy by securing refinancing on all 14 of its PFI deals.
Recognising these problems, in 2011 the government set up a review of PFI. The result was a revised form of PFI, known as ‘PF2’. PF2 projects involved tighter financial control, with the government acting as a minority co-investor; more robust tendering processes, with bidders required to develop a long-term financing solution, where bank debt does not form the majority of the financing of the project; the removal of cleaning, catering and other ‘soft services’.
Despite the government’s intention that PPPs remain an important plank of its funding of public services, the number of new PFI/PF2 projects has nonetheless declined sharply during the 2010s as the chart shows. Of the 715 PPP projects as of 31 March 2017, 631 had been signed before May 2010. Indeed, in 2016/17 only 1 new project was signed.
The collapse of Carillion
Concerns over PPPs remained despite the reforms under PF2. These were brought dramatically into focus with the collapse of Carillion plc (see the blog, Outsourcing, PFI and the demise of Carillion). Carillion was a British company focused on construction and facilities management (i.e. support services for organisations). It was a significant private-sector partner in PPP projects. By 2014 it had won 60 PPP projects in the UK and Canada, including hospitals, schools, university buildings, prisons, roads and railways.
However, Carillion had increasing burdens of debt, caused, in part, by various major acquisitions, including McAlpine in 2008. Events came to a head when, on 15 January 2018, an application was made to the High Court for a compulsory liquidation of the company.
A subsequent report for the House of Commons Public Administration and Constitutional Affairs Committee in light of the collapse of Carillion found that procurement procedures were fundamentally flawed. It found that contracts were awarded based on cost rather than quality. This meant that some contracts were not sustainable. Between 2016 and the collapse of Carillion the government had been forced to renegotiate more than £120m of contracts so that public services could continue.
The ending of PPPs?
On 18 January 2018, the National Audit Office published an assessment of PFI and PF2. The report stated that there were 716 PFI and PF2 projects at the time, either under construction or in operation, with a total capital value of £59.4 billion. In recent years, however, ‘the government’s use of the PFI and PF2 models had slowed significantly, reducing from, on average, 55 deals each year in the five years to 2007/8 to only one in 2016/17.’
At its conference in September 2018, the Labour shadow chancellor, John McDonnell, said that, if elected, a Labour government would not award any new PFI/PF2 contracts. He claimed that PFI/PF2 contracts were set to cost the taxpayer £200bn over the coming decade. Labour policy would be to review all existing PFI/PF2 contracts and bring the bulk of them fully back into the public sector.
Then in the Budget of 29 October 2018, the Chancellor announced that no further PFI/PF2 projects would be awarded, although existing ones would continue.
I have never signed off a PFI contract as chancellor, and I can confirm today that I never will. I can announce that the government will abolish the use of PFI and PF2 for future projects.
We will honour existing contracts. But the days of the public sector being a pushover, must end. We will establish a centre of excellence to actively manage these contracts in the taxpayers’ interest, starting in the health sector.
But does this mean that there will be no more public-private partnerships, of which PFI is just one example? The answer is no. As the Chancellor stated:
And in financing public infrastructure, I remain committed to the use of public-private partnership where it delivers value for the taxpayer and genuinely transfers risk to the private sector.
But just what form future PPPs will take is unclear. Clearly, the government will want to get value for money, but that depends on the mechanisms used to ensure efficient and high-quality projects. What is more, there is still the danger that the companies involved could end up with unsustainable levels of debt if economic circumstances change and it will still involve a burden on the taxpayer for the future.
- Find out how PF2 differs from PFI and assess the extent to which it overcame the problems identified with PFI.
- The government is not bringing back existing PFI contracts into the public sector, whereas the Labour Party would do so – at least with some of them. Assess the arguments for and against bringing PFI contracts ‘in-house’.
- Find out why Carillion collapsed. To what extent was this due to its taking on PFI contracts?
- What were the main findings of the National Audit Office’s assessment of PFI and PF2?
- The government still supports the use of public-private partnerships (PPPs). What form could these take other than as PFI/PF2 contracts? Would the problems associated with PFI/PF2 also apply to PPPs in general?
I recently found myself talking about my favourite TV shows from my childhood. Smurfs aside, the most popular one for me (and I suppose for many other people from my generation) had to be Knight Rider. It was a story about a crime fighter (David Hasselhoff) and his a heavily modified, artificially intelligent Pontiac Firebird. ‘Kitt’ was a car that could drive itself, engage in thoughtful and articulate conversations, carry out missions and (of course) come up with solutions to complex problems! A car that was very far from what was technologically possible in the 80s – and this was part of its charm.
Today this technology is becoming reality. Google, Tesla and most major automakers are testing self-driving cars with many advanced features like Kitt’s – if not better. They may not fire rockets, but they can drive themselves; they can search the internet; they can answer questions in a language of your choice; and they can be potentially integrated with a number of other technologies (such as car sharing apps) to revolutionise the way we own and use our cars. It will take years until we are able to purchase and use a self-driving car – but it appears very likely that this technology is going to become roadworthy within our lifetime.
Artificial intelligence (AI) is already becoming part of our life. You can buy a robotic vacuum cleaner online for less than a £1000. You can get gadgets like Amazon’s Alexa, that can help you automate your supermarket shopping, for instance. If you are Saudi, you can boast that you are compatriots with a humanoid: Saudi Arabia was the first country to grant citizenship to Sophia, an impressive humanoid and apparently a notorious conversationalist who does not miss an opportunity to address a large audience – and it has done so numerous times already in technology fairs, national congresses – even the UN Assembly! Sophia is the first robot to be honoured with a UN title!
What will be the impact of such technologies on labour markets? If cars can drive themselves, what is going to happen to the taxi drivers? Or the domestic housekeepers – who may find themselves increasingly displaced by cleaning robots. Or warehouse workers who may find themselves displaced by delivery bots (did you know that Alibaba, the Chinese equivalent of eBay, owns a warehouse where most of the work is carried out by robots?). There is no doubt that labour markets are bound to change. But should we (the human labour force) be worried about it? Acemoglu et al (2017) think that we should:
Using a model in which robots compete against human labor in the production of different tasks, we show that robots may reduce employment and wages […] According to our estimates, one more robot per thousand workers reduces the employment to population ratio by about 0.18–0.34 percentage points and wages by 0.25–0.5 per cent.
Automation is likely to affect unskilled workers more than skilled ones, as unskilled jobs are the easiest ones to automate. This could have widespread social implications, as it might widen the divide between the poor (who are more likely to have unskilled jobs) and the affluent (who are more likely to own AI technologies). As mentioned in a recent Boston Consulting Group report (see below):
The future of work is likely to involve large structural changes to the labour market and potentially a net loss of jobs, mostly in routine occupations. An estimated 15 million UK jobs could be at risk of automation, with 63 per cent of all jobs impacted to a medium or large extent.
On the other hand, the adoption of automation is likely to result in higher efficiency, huge productivity gains and less waste. Automation will enable us to use the resources that we have in the most efficient way – and this is bound to result in wealth creation. It will also push human workers away from manual, routine jobs – and it will force them to acquire skills and engage in creative thinking. One thing is for certain: labour markets are changing and they are changing fast!
- What do you think is going to be the effect of automation on labour market participation in the future? Why?
- Using the Solow growth model, explain how automation is likely to affect economic growth and capital returns.
- In the context of the answer you gave to question 2, explain how human capital accumulation may affect the ability of workers to benefit from automation.
 Daron Acemoglu and Pascual Restrepo, Robots and Jobs: Evidence from US Labor Markets, NBER Working Paper No. 23285 (March 2017)