Report from the McKinsey Technical Institute

The world is in need of a new engine of GDP growth. Shifting demographics in many countries brought on by aging populations and declining birthrates are reducing the share of working-age populations and creating an economic growth gap: in the not-so-distant future, without an acceleration in productivity growth, there will not be enough workers for countries to meet their aspirations for growth in GDP per capita.

In this context, automation could be a significant opportunity. Our estimates suggest it could help serve as a new productivity engine for the global economy, bridging that economic growth gap. Automation could raise global productivity by as much as 0.8 to 1.4 percent annually. While that will not be enough to ensure all countries meet their per capita GDP growth aspirations, it will make a major contribution toward that goal.

In order for this growth to take place, people will need to keep working—alongside the robots that will help provide the productivity boost. Even with automation, a deficit of labor is more likely than a surplus. Yet the adoption of automation will change the nature of work, and the public debate about it often focuses on the prospect that it could lead to very large-scale unemployment. Such fears are not new: already back in 1966, a US government commission noted concerns that technological change “would in the future not only cause increasing unemployment, but that eventually it would eliminate all but a few jobs, with the major portion of what we now call work being performed automatically by machine.” 79

In fact, the large-scale shifts in employment that automation will enable are of a similar order of magnitude to the long-term technology-enabled shift in the developed countries’ workforces away from agriculture in the 20th century. That movement did not result in long-term mass unemployment because it was accompanied by the creation of new types of work not foreseen at the time. We cannot definitively say whether things will be different this time. But our analysis does show that automation will fundamentally alter the workplace, requiring all workers to cohabit extensively with technology and reshaping the corporate landscape.

THE PRODUCTIVITY BOOST FROM AUTOMATION COULD BRIDGE A LOOMING ECONOMIC GROWTH GAP

GDP growth was exceptionally brisk over the past half century, driven by the twin engines of employment growth and rising productivity. However, declining birthrates and the trend toward aging in many advanced and some emerging economies mean that peak employment will occur in most countries within 50 years. The workforce in Japan is already shrinking in size, and the total number of workers in China will start to decline within a decade. This expected decline in the share of working-age population will place the onus for future economic growth far more heavily on productivity gains. Employment growth of 1.7 percent annually between 1964 and 2014 in the G19 countries and Nigeria is set to fall to just 0.3 percent per year. 80

Prior MGI research has shown that even if productivity growth maintains its 1.8 percent annual rate of the past half century, the rate of GDP growth will fall by as much as 40 percent over the next 50 years. On a per capita basis, the GDP growth decline is about 19 percent (Exhibit 23). In order to compensate for slower employment growth, productivity would need to grow at a rate of 3.3 percent annually, or 80 percent faster than it has grown over the past half century. 81


79. Technology and the American economy, Report of the National Commission on Technology, Automation and Economic Progress, US Department of Health, Education and Welfare, February 1966.
80. The global productivity challenge created by waning employment growth is detailed in our report Global growth: Can productivity save the day in an aging world? McKinsey Global Institute, January 2015. The G19 countries are the G20 minus the European Union: Argentina, Australia, Brazil, Canada, China, France, Germany, India, Indonesia, Italy, Japan, Mexico, Russia, Saudi Arabia, South Africa, South Korea, Turkey, the United Kingdom, and the United States. Including Nigeria, these 20 countries generate more than 80 percent of global GDP.
81. Global growth: Can productivity save the day in an aging world? McKinsey Global Institute, January 2015. Our estimate of employment growth’s contribution to GDP growth in this report differs slightly from this earlier research, as we have assumed productivity measured in each country, rather than a global average.


Exhibit 23

Without an acceleration in productivity growth, demographic trends will cut GDP growth by nearly half, causing a decline in historic GDP per capita growth rate.

Exhibit 23

NOTE: Numbers may not sum due to rounding.

SOURCE: The Conference Board Total Economy database; United Nations Population Division; McKinsey Global Institute analysis


The size of the workforce over the next 50 years is too small to maintain current per capita GDP growth without accelerating productivity growth

An economic growth gap has opened up as a result of the shrinking workforce. The number of full-time equivalents needed just to maintain the current GDP per capita over the next 50 years is larger than the number of workers who will be available, given demographic trends in most countries. We estimate this gap in economic output as being about 130 million full-time equivalents (FTEs) for the G19 countries and Nigeria alone. (In other words, the economic output equivalent to an additional 130 million full-time workers would be needed to maintain current GDP per capita, assuming no productivity gains.) If countries are to achieve more ambitious longer-term growth in line in line with their development and the aspirations of their citizens goals (that is, growth in GDP per capita), the gap would be considerably larger—the economic output of about 6.7 billion FTEs by 2065 (Exhibit 24).

Exhibit 24

Demographic trends are creating a pressing need for an acceleration in productivity growth.

Exhibit 24

1. Additional full-time equivalents (FTEs) needed to achieve growth target.
NOTE: Numbers may not sum due to rounding.

SOURCE: The Conference Board Total Economy database; United Nations Population Division; McKinsey Global Institute analysis


For the purposes of our analysis, we based our country-level GDP projections on McKinsey’s proprietary Global Growth Model. This model projects a global GDP growth rate of 2.9 percent, resulting in an annual productivity growth of 2.8 percent. For advanced economies overall, the projections in the McKinsey model result in a compound annual growth of GDP per capita to 2030 of 1.4 percent, and 3.4 percent for emerging economies.

For the G19 countries and Nigeria that we discuss in this report, the model projects GDP growth of 2.7 percent, resulting in GDP per capita growth of 2 percent to 2030. 82

The growth gap between the projected growth and growth that must be provided by productivity increases is most pronounced in fast-growing countries such as China, India, Indonesia, and Nigeria, but it is also prevalent in Germany and Japan and other countries that are already experiencing a slowdown or decline in the working-age population. Furthermore, productivity growth has slowed in many countries. 83

Exhibit 25 illustrates the need for continued productivity growth, even just to maintain GDP per capita (an admittedly unsatisfactory outcome) in both the medium term (by 2030) and in the long term (by 2065) for the 20 largest economies in the world. Without productivity growth, aging countries with older demographics simply would not have enough workers needed to maintain GDP per capita. Younger countries have more than enough workers to maintain GDP per capita.

Automation can help bridge the projected growth gap by compensating for the slowdown in workforce growth

Our analysis of the automation adoption scenarios suggests that automation could help bridge the projected growth gap caused by a deficit of full-time equivalents worldwide. Automation alone will not be sufficient to achieve long-term target growth across the world, given the decline in the working-age population and the need for high productivity to achieve that target. Especially in fast-growing countries, other measures to boost productivity will be needed. However, notably, the productivity gains from automation could suffice to at least maintain today’s GDP per capita.

Our methodology takes into account only labor substitution gains. Other performance gains—in the form of improved quality, fewer breakdowns, greater safety, and so on—would come on top of this overall productivity boost. We also assume that human labor displaced by automation would rejoin the workforce and be as productive as it was in 2014, that is, new demand for labor will be created. In some ways, this is a conservative assumption, given that if automation produces productivity gains, we assume displaced labor reenters the workforce at a lower level of productivity than the level of labor productivity at the time the displacement occurs. Others could argue that the activities performed by workers who are displaced by automation could have lower levels of economic output than the activities that had been taken over by machines. In any case, it is vital that there be new demand for labor displaced by automation.


82. McKinsey & Company’s proprietary Global Growth Model provides complete time-series data for more than 150 concepts and 110 countries over 30 years. It incorporates more than a dozen major international databases from such institutions as the United Nations, the World Bank, the International Monetary Fund, and the Bank for International Settlements. The structure of the model emphasizes the drivers of economic growth, including demographic factors, education, energy supply, physical capital, and some determinants of total factor productivity. It captures the long-term effects of urbanization and industrialization, as well as the impact of sociopolitical institutions, especially on finance and governance. Because business-cycle fluctuations affect growth in the short term, the model also links trade and international capital flows, credit and asset markets, and the monetary relationships that determine inflation, interest rates, and exchange rates. See the technical appendix for more details. The growth model shows a compound annual growth rate for GDP and GDP per capita for 2015 to 2030 by country as follows: China 5.1 percent GDP, 4.9 percent GDP per capita; France 1.0 percent and 0.6 percent; Germany 1.3 percent and 1.6 percent; India 5.8 percent and 4.8 percent; Italy 0.5 percent for both GDP and GDP per capita; Japan 1.0 percent and 1.3 percent, United Kingdom 1.5 percent and 1.0 percent; United States 2.2 percent and 1.4 percent.
83. Global growth: Can productivity save the day in an aging world? McKinsey Global Institute, January 2015.


Exhibit 25

Aging countries have the greatest need for productivity growth just to maintain GDP per capita.

Exhibit 25

1. Countries on borderline of demographic categorization between aging and growing populations (under 1 million FTE surplus/deficit in 2065) categorized according to demographic statistics and grouping of similar regional economies.

SOURCE: The Conference Board Total Economy database; International Labour Organisation; United Nations Population Division; Statista; McKinsey Global Institute analysis


Only considering the labor substitution effects on current work activities, and assuming that displaced labor reenters the workforce at 2014 levels of productivity, we estimate that, by 2065, the productivity enabled by automation could potentially increase economic growth by 0.8 percent to 1.4 percent annually, the equivalent of 1.1 billion to 2.3 billion FTEs. (Exhibit 26).

Exhibit 26

Globally, automation could become a significant economic growth engine as employment growth wanes.

Exhibit 26

1. Additional full-time equivalents (FTEs) needed to achieve growth target.
NOTE: Numbers may not sum due to rounding.

SOURCE: The Conference Board Total Economy database; United Nations Population Division; McKinsey Global Institute analysis


Exhibit 27 shows what the effect of automation could be on the gap between growth targets and economic output by country for the G19 and Nigeria. We discuss individual countries and groupings of countries in more detail later in this chapter.

Exhibit 27

Modeled effects of automation on the gap between growth targets and economic output by 2030.

Exhibit 27

NOTE: Surplus/deficits calculated as gap to projected GDP in 2030 as a percentage of 2014 GDP by using population projections; base case (no automation) assumes 2014 productivity. Assumes no other productivity growth than that provided by automation.

SOURCE: The Conference Board Total Economy database; International Labour Organization; United Nations Population Division; McKinsey Global Institute analysis


While our estimates of the productivity boost from automation are substantial, they are of an order of magnitude comparable to major technologies that have been introduced in the past two centuries. For example, between 1850 and 1910, the steam engine has been estimated to have enabled productivity growth of 0.3 percent per annum. Analyses of the introduction of robots in manufacturing and IT estimate that they have accounted for annual productivity increases of 0.4 percent and 0.6 percent, respectively (Exhibit 28). 84 One difference is that automation of current work activities as we have analyzed it encompasses multiple technologies, not just one.

Exhibit 28

Automation of existing activities could increase productivity at magnitudes similar to other major technologies.

Exhibit 28

NOTE: We include multiple technologies in our analysis of “automation,” so these technologies are not entirely comparable, but meant to provide an order of magnitude comparison.

SOURCE: Nicholas Crafts, “Steam as a general purpose technology: A growth accounting perspective,” Economic Journal, volume 114, issue 495, April 2004; Mary O’Mahony and Marcel P. Timmer, “Output, input, and productivity measures at the industry level: The EU KLEMS database,” Economic Journal, volume 119, issue 538, June 2009; Georg Graetz and Guy Michaels, Robots at work, Centre for Economic Performance discussion paper 1335, March 2015; McKinsey Global Institute analysis


The productivity potential of automation is a multiplier of other productivity levers. Economies that have relatively low productivity can also accelerate productivity growth through other means, for example by adopting best practices from other countries with high productivity, in order to reap the full benefits of automation’s potential and achieve a faster growth trajectory. Previous MGI research has estimated that about three-quarters of potential productivity improvements could come from the broader adoption of best practices and technologies, as companies catch up with sector leaders. The remaining one-quarter would come from technological, operational, and business innovations that go beyond best practices and push the frontier of the world’s GDP potential. Business leaders and policy makers can encourage this acceleration of productivity by removing barriers to competition in service sectors, investing in physical and digital infrastructure, exploiting data to identify transformational improvement opportunities, opening economies to crossborder flows, and crafting a regulatory environment that fosters increased productivity and innovation. 85


84. Nicholas Crafts, “Steam as a general purpose technology: A growth accounting perspective,” Economic Journal, volume 114, issue 495, April 2004; Mary O’Mahony and Marcel P. Timmer, “Output, input, and productivity measures at the industry level: The EU KLEMS Database,” Economic Journal, volume 119, issue 538, June 2009; George Graetz and Guy Michaels, Robots at work, Centre for Economic Performance discussion paper number 1335, March 2015.

85. Global growth: Can productivity save the day in an aging world? McKinsey Global Institute, January 2015.


Exhibit 29

In the United States, automation can help achieve projected GDP per capita growth.

Exhibit 29

NOTE: The “projected GDP per capita” scenario for the United States uses projections from McKinsey’s Global Growth model, with GDP compound annual growth rate (CAGR) for 2015–65 of 1.9%, resulting in a productivity CAGR of 1.5%. The “maintain current GDP per capita” scenario assumes GDP will grow at the same rate as population (0.5% CAGR for 2015–65), resulting in a productivity CAGR of 0.1%. See technical appendix for details.

SOURCE: The Conference Board Total Economy database; International Labour Organisation; United Nations Population Division; McKinsey Global Institute analysis.


Emerging economies with aging workforces will get a productivity boost from automation but will also need to find additional sources to maintain their growth trajectory

This category includes Argentina, Brazil, China, and Russia, which all face economic growth gaps as a result of projected declines in the working population. Their first distinguishing factor is that their current productivity is not sufficient to support GDP per capita over the long run. For these economies, automation could provide the productivity injection necessary just to maintain GDP per capita. However, to achieve a faster growth trajectory that is more commensurate with their developmental aspirations (GDP growth of 4.1 percent and GDP per capita growth of 3.8 percent), these countries would need to supplement automation with additional sources of productivity, such as process transformation and other technologies.

This grouping is not monolithic, and there are some divergences based on demographic and economic growth differences. Argentina, Brazil, and Russia, for example, are projected to have lower GDP per capita growth than China. Argentina and Brazil have younger populations than China and Russia; the median age is about seven years lower. They also have significantly faster-growing populations.

While these economies could receive a strong productivity boost from automation, their wage levels are lower than in the advanced economies we discussed above, and their overall adoption of automation may be slower as a result, since the business case for adoption may be less compelling. As we have seen, in these countries it could take longer for the cost of automation solutions, especially if they involve hardware, to make automation feasible when compared to the costs of human labor.

Policy measures including encouraging increased competition and the development of a high standard of technological literacy in the population at large could help speed the process of automation adoption. China, for example, already has the highest rate in the world for technology-enabled payment platforms; a recent report by market research firm Nielsen found that 86 percent of Chinese paid for online purchases with digital payment systems, double the global average. 86

A second distinguishing factor for these economies that will affect the impact of automation is that their overall productivity levels tend to be relatively low compared with those of advanced economies. To capture the full multiplier effects from automation, and achieve their aspirations for a continued fast growth trajectory, these countries will need to supplement automation with other levers to enhance productivity.

Exhibit 30 shows the example of China from this group of fast-growing emerging economies. China’s population is aging rapidly, which means that in the longer term its working-age population will peak as early as 2024 and could shrink by one-fifth. 87 Within the next decade, its workforce will be short of the equivalent of about 600 million workers to attain the projected growth if it does not drive productivity improvements. Early adoption of automation could lessen this gap by about 100 million FTEs, but the country still faces a likely shortfall.


86. China maintains robust e-commerce growth, Nielsen, March 2016
87. Global growth: Can productivity save the day in an aging world? McKinsey Global Institute, January 2015.


Exhibit 30

Automation can contribute to productivity growth in China, but its high projected GDP per capita growth requires additional productivity levers.

Exhibit 30

NOTE: The “projected GDP per capita” scenario for China uses projections from McKinsey’s Global Growth model, with GDP compound annual growth rate (CAGR) for 2015–65 of 4.0%, resulting in a productivity CAGR of 4.5%. The “maintain current GDP per capita” scenario assumes GDP will grow at the same rate as population (-0.2% CAGR for 2015–65), resulting in a productivity CAGR of 0.3%. See technical appendix for details.

SOURCE: The Conference Board Total Economy database; International Labour Organisation; United Nations Population Division; McKinsey Global Institute analysis.


In Exhibit 31, we highlight Brazil as another example in this group. The model shows that early adoption of automation could allow Brazil to meet its medium term GDP per capita growth expectations. But finding additional levers to accelerate Brazil’s productivity growth would be beneficial in both the medium term, as the majority of the adoption scenarios show automation not providing sufficient economic growth to meet projected GDP per capita growth, and in the long term.

Exhibit 31

Automation of existing activities could help Brazil meet its medium-term GDP per capita growth aspirations, but additional productivity acceleration could be required.

Exhibit 31

NOTE: The “projected GDP per capita” scenario for Brazil uses projections from McKinsey’s Global Growth model, with GDP compound annual growth rate (CAGR) for 2015–65 of 3.3%, resulting in a productivity CAGR of 3.2%. The “maintain current GDP per capita” scenario assumes GDP will grow at the same rate as population (0.2% CAGR for 2015–65), resulting in a productivity CAGR of 0.1%. See technical appendix for details.

SOURCE: The Conference Board Total Economy database; International Labour Organisation; United Nations Population Division; McKinsey Global Institute analysis


Emerging economies with younger populations will get a boost from automation but will need other productivity gains to ensure sufficient longerterm growth

The third grouping we identify are emerging economies with younger populations. These include India, Indonesia, Mexico, Nigeria, Saudi Arabia, South Africa, and Turkey. Saudi Arabia is something of an anomaly in this group because of its high wages adjusted for purchasing power parity (see Box 6, “With a combination of strong population growth and high wages, Saudi Arabia is atypical”). These countries all have aspirations for high longterm growth, in order to lift living standards for the rising population. At the same time, these countries have strong population growth rates. The ratio of working to total population will peak in the 2050s (except in Turkey and Saudi Arabia, where it will peak early), ensuring they have the necessary workforce to maintain GDP per capita. Even so, in these countries, automation on its own will not suffice to meet the growth aspirations, and other productivity levers will be needed. Exhibit 33 shows how the productivity impact of automation could play out in Nigeria, which has the highest population growth rate of the group. Still, to meet the projected GDP per capita growth, it faces a shortfall of ten million workers in 15 years. Early adoption of automation could close this gap.

Box 6. With a combination of strong population growth and high wages, Saudi Arabia is atypical

Saudi Arabia and some of its neighbors in the Middle East are experiencing relatively slow growth, but they have both high wages, measured at purchasing power parity, and a growing workforce. Saudi Arabia, for example, has a demographic bulge: more than half the Kingdom’s population is younger than 25, and by 2030 the number of Saudis aged 15 years and over will likely increase by about six million. Based on historical trends in participation, this upcoming demographic bulge could almost double the size of the Saudi labor force. 1 Its productivity growth of 0.8 percent between 2003 and 2013 was well below the average for its G20 peers. 2


1. Saudi Arabia beyond oil: The investment and productivity transformation, McKinsey Global Institute, December 2015.
2. Ibid.


Automation could provide a considerable productivity boost to these countries, enough to meet GDP per capita growth aspirations. Moreover, automation will be economically feasible in these economies rapidly, because of their relatively high wage levels. The challenge for governments in these countries will be to create additional human jobs to employ the large cohort of young people who will reach working age in the near future. However, the ratio of working-age population to total population will peak in the 2030s, reversing the previous demographic dividend and making continued productivity gains more important (Exhibit 32).

Exhibit 32

Saudi Arabia’s relatively high wage rates could drive early adoption, and its relatively young demographics put a premium on creating jobs.

Exhibit 32

NOTE: The “projected GDP per capita” scenario for Saudi Arabia uses projections from McKinsey’s Global Growth model, with GDP compound annual growth rate (CAGR) for 2015–65 of 2.2%, resulting in a productivity CAGR of 1.6%. The “maintain current GDP per capita” scenario assumes GDP will grow at the same rate as population (0.6% CAGR for 2015–65), resulting in a productivity CAGR of 0.1%. See technical appendix for details.

SOURCE: The Conference Board Total Economy database; International Labour Organisation; United Nations Population Division; McKinsey Global Institute analysis.


 

Exhibit 33

In Nigeria, automation can contribute substantially to economic growth, but adding other means of productivity growth will likely be required to meet expectations.

Exhibit 33

NOTE: The “projected GDP per capita” scenario for Nigeria uses projections from McKinsey’s Global Growth model, with GDP compound annual growth rate (CAGR) for 2015–65 of 5.3%, resulting in a productivity CAGR of 2.4%. The “maintain current GDP per capita” scenario assumes GDP will grow at the same rate as population (2.4% CAGR for 2015–65), resulting in a productivity CAGR of -0.5%. See technical appendix for details.

SOURCE: The Conference Board Total Economy database; International Labour Organisation; United Nations Population Division; McKinsey Global Institute analysis


AUTOMATION’S IMPACT ON EMPLOYMENT COULD BE OF THE SAME ORDER OF MAGNITUDE AS PREVIOUS MAJOR STRUCTURAL ECONOMIC SHIFTS

A recurring question about automation is its effect on employment. Many forecasters paint a sometimes dire picture of what the adoption of automation could do for jobs, particularly blue-collar jobs. The World Economic Forum, for example, has predicted that more than five millions jobs could be lost to robots in 15 major developed and emerging economies over the next five years. 88

The advent of large-scale automation in the workplace will undoubtedly alter the nature of the workplace, and the nature of work itself, as machines increasingly take over activities that were hitherto the domain of human workers. But to some extent, this is an old story. Telephone operators in the 1950s needed to physically connect switchboard plugs, but these jobs no longer exist today, as no physical capabilities are required to connect calls. Telex and telegraph operators are a dying profession. The typing pool that was a mainstay of office life in the 1950s and 1960s has not survived. In their place have come myriad new jobs born of the technological developments, from call center employees to IT help desk personnel. Personal assistants no longer take dictation, but monitor email.

Such anecdotal evidence is underscored by macroeconomic data. Positive gains have been reported in both productivity and employment in the United States in more than two-thirds of the years since 1929 despite the rapid onward march of technological development. 89 One-third of new jobs created in the United States in the past 25 years did not exist, or barely existed, 25 years ago. 90 Moreover, every three months about 6 percent of US jobs are destroyed by shrinking or closing businesses, while a slightly larger percentage of jobs are added. 91 This is not just a US phenomenon; findings from other countries indicate that these are global trends. For example, a detailed analysis of the French economy by McKinsey’s French office, published in 2011, showed that while the internet had destroyed 500,000 jobs in France in the previous 15 years, it had created 1.2 million others, a net addition of 700,000, or 2.4 jobs created for every job destroyed. 92

Will this pattern continue with automation, or could things be different this time? (See Box 7, “Is this time different?”). Certainly the scale and potential scope of the work activities that have the potential to be automated are very substantial indeed. Our model contemplates the possibility that hundreds of millions of workers will have to shift the activities they are paid to do.

Yet even this sort of large-scale structural economic shift over such a long period of time is not unprecedented. In the United States, for example, the share of farm employment fell from 40 percent in 1900 to 2 percent in 2000, while the share of manufacturing employment fell from 25 percent in 1950 to less than 10 percent in 2010 (Exhibit 34). 93 In both cases, the jobs that disappeared were offset by new ones that were created, although what those new jobs would be could not be ascertained at the time.

Whatever the net impact on employment, the nature of work will change with automation. There will be tighter integration between humans and machines than there is today, and this will increase overall efficiency, since machines will be more accurate with the activities that they take on. This will free up humans to perform tasks that use higher-level capabilities, especially those that require social and emotional ones.


88. The future of jobs: Employment, skills, and workforce strategy for the fourth Industrial Revolution, World Economic Forum, January 2016. Other forecasters have made specific predictions about large scale job losses related to automation or painted scenarios of a world without work. See Jeremy Bowles, The computerization of European jobs, Bruegel, July 2014; Martin Ford, Rise of the robots: Technology and the threat of a jobless future, Basic Books, 2015; and Carl Benedikt Frey and Michael A. Osborne, The future of employment: How susceptible are jobs to computerisation? Oxford Martin School, September 17, 2013.
89. Growth and renewal in the United States: Retooling America’s economic engine, McKinsey Global Institute, February 2011.
90. Global growth: Can productivity save the day in an aging world? McKinsey Global Institute, January 2015.
91. Artificial intelligence, automation, and the economy, Executive Office of the President, December 2016.
92. Impact d’internet sur l’économie française: Comment internet transforme notre pays (The internet’s impact on the French economy: How the internet is transforming our country), McKinsey & Company, March 2011.
93. Stanley Lebergott, “Labor force and employment 1800–1960,” in Output, employment, and productivity in the United States after 1800, Dorothy S. Brady, ed., NBER, 1966; World Bank data; Mack Ott, “The growing share of services in the US economy—degeneration or evolution?” Federal Reserve Bank of St. Louis Review, June/July 1987.


Box 7: Is this time different

Automation of human work activities has been occurring for at least the past two centuries, and while these technologies have automated a wide range of activities people had formerly been paid to do, new activities, occupations and jobs have been created. This has offset what would have become a situation of mass unemployment, had new demands for human labor not been produced.

But is this time different? Are we reaching a point where the pace and/or types of automating work activities will outstrip the global economy’s ability to create new activities and jobs for people to be paid to do?

In some ways, this is an “evergreen” issue, in that these questions and concerns have accompanied the adoption of automation through history. Beyond the time of the Luddites in 19th-century Britain, John Maynard Keynes wrote about the “new disease” of technological unemployment during the Great Depression. 1 Over the years, many have speculated about decreasing demand for human labor as machines automate more work, including Keynes, in the same article from 1930 quoted above, and the 1966 US Commission on Technology, Automation and Economic Progress. But while the workweek has declined from working 10-18 hours per day, six days per week, during the Industrial Revolution to about eight hours per day, five days per week in the mid-20th century, it has not declined substantially since that point in many developed countries.

Some argue that there are factors suggesting that this time is different. Eric Brynjolfsson and Andrew McAfee describe an inflection point between the first machine age, based on the automation of physical tasks through mechanization, and a second machine age, based on the automation of cognitive tasks through digital technologies. 2 Digital technologies’ basic capabilities, including computing power, storage capacity, and communications throughput, appear to be developing exponentially. For example, Moore’s Law suggests that the computing power that can be purchased for $1 doubles roughly every two years. Chroniclers of “exponential technologies” such as Ray Kurzweil extrapolate out to a time when a computer will have the computing power of a human brain, and beyond, potentially pointing to a future in which, combined with the appropriate software, an “artificial general intelligence” could be created that rivals that of human beings. Others point to a dimension of human work described by the economists Daron Acemoglu and David Autor: between routine and nonroutine work. 3 While much of the work that had historically been automated was routine (for example, what we describe in this report as predictable physical activities and the collection and processing of data), many of the examples of newer automation technologies that we find remarkable automate what we would have described as non-routine work. That includes driving cars in busy streets, or diagnosing disease.

However, the labor market has until now always adapted to the replacement of jobs with capital, with price effects tending to balance the forces of automation and creating new complex tasks for people to be paid to do. The December 2016 US White House report on artificial intelligence and automation states: “Recent research suggests that the effects of AI on the labor market in the near term will continue the trend that computerization and communication innovations have driven in recent decades… The economy has repeatedly proven itself capable of handling this scale of change, although it would depend on how rapidly the changes happen and how concentrated the losses are in specific occupations that are hard to shift from.” 4 To support some of the assumptions in our modeling analyses, we argue that many of the factors affecting the pace and extent of automation adoption, such as the engineering of solutions to specific problems, and particularly the non-technical organizational change management, regulatory and acceptance dynamics around technology adoption, have not changed.

However, changing other assumptions in our model could lead to significantly different outcomes. For instance, our assumptions on the time required to develop capabilities and integrate and customize them into solutions that solve specific problems assumes that these activities will primarily be performed by people. But in a world where machines can teach themselves, these time frames could be considerably different from our assumptions.

We also only analyze the automation of current work activities at the performance levels at which humans currently perform them. If machines can perform these tasks at significantly higher levels, and/or other new productive activities, then productivity growth could accelerate even faster than we modeled. Similarly, we modeled workers displaced by automation returning to the workforce at the productivity of workers in 2014; if they return at higher levels of productivity (for example, at the average level of productivity of workers at the time they are displaced), then overall productivity growth will also accelerate. Could this lead to a surplus of labor? This would depend on the economy’s ability to create new and more things to pay people to do.

Is this time different? We can’t definitely say—but the question is a familiar one.


1. John Maynard Keynes, “Economic possibilities for our grandchildren,” in Essays in Persuasion, Macmillan, 1933. The essay is available online at www.econ.yale.edu/smith/econ116a/keynes1.pdf.
2. Erik Brynjolfsson and Andrew McAfee, The second machine age: Work, progress, and prosperity in a time of brilliant technologies, W. W. Norton & Company, 2014.
3. Daron Acemoglu and David H. Autor, “Skills, tasks, and technologies: Implications for employment and earnings,” in Handbook of Labor Economics, volume 4B, David Card and Orley Ashenfelter, eds., Elsevier, 2011.
4. Artificial intelligence, automation, and the economy, Executive Office of the President, December 2016.


Exhibit 34

Employment in agriculture has fallen from 40 percent in 1900 to less than 2 percent today.

Exhibit 34

SOURCE: Stanley Lebergott, “Labor force and employment 1800–1960,” in Output, employment, and productivity in the United States after 1800, Dorothy S. Brady, ed., NBER, 1966; World Data Bank, World Bank Group; FRED: Economic Research, Federal Reserve Bank of St. Louis; Mack Ott, “The growing share of services in the US economy—degeneration or evolution?” Federal Reserve Bank of St. Louis Review, June/July 1987; McKinsey Global Institute analysis


AUTOMATION COULD CHALLENGE SOME CONVENTIONAL BELIEFS ABOUT THE GLOBAL ECONOMY AND PATHS TO PROSPERITY

The advances in automation we have outlined and their potential impact on national economies could upend some prevailing models of development and challenge ideas about globalization. Among other possible repercussions, it could skewer conventional wisdom about the economic advantages of having a high birthrate. Low-cost labor may lose its edge as an essential development tool for emerging economies, as costs of automation fall. And these economies, in turn, will have new opportunities to leapfrog into higher-value manufacturing and services—including IT—that will enable them to compete head-on with advanced economies. Automation could also accelerate the diminishing of trade of physical goods that began a decade ago, as digital cross-border flows consolidate their pre-eminence.

Automation could reverse a demographic dividend for economies

The country scenarios we have outlined above could overturn ideas about the relationship between economic growth and population growth. The declining fertility rates and aging trends that have taken hold in a broad range of countries including China, Germany, Italy, and Russia have long been viewed as harbingers of weaker economic growth in the future— and a major policy challenge. The aging and shrinking of the workforce is unprecedented in modern history, and one of its consequences is that the number of retirees will likely grow more than twice as fast as the labor pool, leaving fewer workers to support the elderly. 94 Conversely, countries with a high birthrate and a growing working-age population were viewed as having an opportunity to achieve more rapid GDP growth, as they could build future growth on the twin pillars of productivity and employment growth.

Automation could help change that scenario. Countries experiencing population declines or stagnation will be able to make use of it to help maintain living standards even as the labor force wanes. Meanwhile, countries with high birthrates and a significant growth in the working-age population may have to worry about where new jobs will come from in a new machine age. This is especially the case for high-income and low-growth countries, including some in the Middle East.

The stakes are high because of the sheer numbers of people who have lifted their living standards as a result of countries following the classic developmental path—and the millions more who still hope to do so. In India, the impact of automation on employment could affect working hours of the equivalent of 220 million people; in China, that rises to 380 million fulltime equivalents.

Chinese companies still have low levels of automation overall, although they are moving to ramp up. There are only 36 robots per 10,000 Chinese manufacturing workers, about half the average of all advanced economies and about one-fifth the US level. 95 Auto factories are less than 30 percent as automated as US plants, and food processing is only about 12 percent as automated as US food processing. 96 This gap reflects the cost of Chinese manufacturing labor, which has risen but remains low by the standards of advanced economies. The average manufacturing worker makes about 10 percent of the average US manufacturing wage, for example. Our research finds that most Chinese manufacturers are not yet able to realize the maximal value from robots due to a production process that is less than optimal. Chinese companies may adopt a hybrid model that mixes the speed and precision of automation with the flexibility of human labor.

Low-cost labor: No longer a development panacea?

Starting already in the 1880s in Japan, country after country around the world including South Korea, Taiwan, and most notably China, has followed a familiar pattern of development. A combination of low wage agriculture and manufacturing—at times often backed by protectionist policies to encourage import substitution and boost exports—creates jobs and swells household income. As workers become more productive and households more prosperous, manufacturing moves up the value chain, producing higher quality products. Country dwellers flock to cities to join this industrialization wave, creating urban pockets of consumers with disposable income that helps generate greater prosperity. This trend has been driven by a huge influx of 1.2 billion people joining the global labor market between 1980 and 2010, and it has brought millions out of poverty. In 1990, about 23 percent of the world’s population belonged to the “consuming class,” by which we mean that they earned more than $10 per day. In 2010, that share had risen to 36 percent of the global population, and we project it will exceed 50 percent by 2025. 97

That labor-intensive economic development model is still largely intact and is being followed by countries from India to South Africa. For countries still stuck in poverty, including in sub-Saharan Africa, it remains the obvious path to prosperity. But in the new world that is taking shape, low-cost labor may lose some of its edge as an essential developmental tool for countries, as automation drives down the cost of manufacturing globally. Indeed, research by the Harvard economist Dani Rodrik suggests that a “premature deindustrialization” is already taking place in some emerging economies, although he ascribes that more to trade and globalization than to technological progress. 98

As we have seen, the type of activities most susceptible to automation includes physical activity or operating machinery in a predictable setting—in other words, highly routine work. A number of occupations in manufacturing sectors that use low-cost labor fit into this category, such as sewing machine operators, who have a 98 percent automation potential. Agriculture, too, has a high automation potential because much of its activity is both physical and predictable, and thus replaceable by machines. The speed of adoption of automated technology depends in part on the size of the firms. For example, in India, where much farming is on a small-scale family subsistence model, changing to larger-sized farms would sharply raise the automation potential.

For advanced economies that have lost manufacturing jobs over the past decades because of competition from lower-cost labor and the buildup of supply chains elsewhere, the drive to automation may end or even reverse the outflow. The wage gap between advanced and emerging economies is wider in manufacturing than for other sectors. At the same time, companies in advanced economies often have a greater ability to fund the capital expenditure that is needed to build highly automated manufacturing plants.

That does not mean jobs will flow back in large numbers, since only a very small part of manufacturing value added is driven by labor cost, and any “re-onshoring” that does take place will likely happen in highly automated plants. Moreover, for companies in advanced economies, the business case for automation is not simple: hardware and software can be costly, and integrating them successfully is laborious. But machines also have some clear advantages: they can run continuously and do not require large human resources departments. Operations are easier to manage if they are next door, which makes oversight and shipping easier than if operations were halfway around the world. Politically and from a public relations standpoint, too, it can be advantageous to be seen as a local stalwart.

New opportunities for higher-value manufacturing and services through automation

Growing use of automation across sectors and within sectors raises the prospect that some countries could leapfrog in the future to become active in industries where they are now weak or have little presence and either no infrastructure base or one that is aging and technically obsolete. For example, Saudi Arabia and Iran have aspirations to build up an automotive industry to serve not just the domestic market but also the wider Middle East region. They both have domestic supplies of raw materials, such as iron ore and bauxite, and plentiful energy; deploying automation technologies could help them leapfrog into state-of-the-art manufacturing. In Russia, where the number of employees will likely decrease by 30 percent over the next half century as a result of a declining birthrate, automation could compensate for the smaller workforce and revitalize growth in manufacturing sectors. New technologies and integrated solutions might reduce capital expenditure, benefiting emerging economies with limited resources. They also potentially provide the ability for emerging economies to create economies of scale.

Emerging economies could gain an edge in some service sectors, including health care and social assistance, IT, and professional, scientific, and technical services. That is because the gap in wages between countries for these services is less pronounced than it is for manufacturing, and the cost of the software needed to compete on a global scale is much lower.

Automation could accelerate the diminishing of physical trade

What we now can see as the heyday of globalization, the 20-year period that began in the mid-to late 1980s, the global flows of goods, services, and finance grew rapidly, outpacing GDP growth. Since the global financial recession in 2008 and slow recovery, however, that rapid expansion has stopped in its tracks. Growth in global goods trade has flattened, trade in services has posted only modest growth, and financial flows have fallen sharply. At the same time, data flows have soared, with cross-border bandwidth growing 45 times larger since 2005. Data flows now account for a larger impact on of global GDP than does global trade in goods. 99

Automation could to some extent push this trend further, as companies rely less on traditional shipping methods and more on digital transactions and exchanges. For example, 3D printing, if widely adopted by global manufacturers, could reduce global trade volumes as more products are “printed” where they are consumed. There are already examples of this at work, such as GE Aviation, which is beginning to use 3D printing to produce fuel nozzles for its new Leap engine.100 A fuel nozzle made the traditional way consists of 20 components, with a supply chain that spans countries. 3D printing allows the company to produce best-quality nozzles in one piece, at one location, eliminating the need to ship intermediate parts across borders.

At the same time, the rise of digital platforms such as Alibaba, Amazon, and eBay changes the economics of doing business across borders, bringing down the cost of international interactions and transactions. These platforms create markets and user communities with global scale, providing businesses in advanced and emerging economies alike with a huge base of potential customers and effective ways to reach them. Companies based in developing countries can overcome constraints in their local markets and connect with global customers, suppliers, financing, and talent far more easily than they ever could. 101


95. The China effect on global innovation, McKinsey Global institute, October 2015.
96. Ibid.
97. Global growth: Can productivity save the day in an aging world? McKinsey Global Institute, January 2015.
98. Dani Rodrik, Premature deindustrialization, NBER working paper number 20935, February 2015
99. Digital globalization: The new era of global flows, McKinsey Global Institute, February 2016.
100. Ibid.
101. Playing to win: The new global competition for corporate profits, McKinsey Global Institute, September 2015.


Automation can quickly become a new engine for the global economy at a time when the working-age population in numerous countries is stagnant or falling and productivity growth is struggling to compensate. Whatever their economic structure, wage levels, growth aspirations, or demographic trends, countries around the world could benefit from adopting automation to maintain living standards and help meet long-term growth aspirations. The rapid development and growing adoption of automation technologies will create myriad new opportunities even as they likely disrupt the world of work and challenge long-held conventions about the global economy and paths to prosperity. In order to make the most of the potential offered by automation and, at the same time, manage its consequences on companies, national economies, and workers around the world, policy makers, business leaders, and men and women everywhere will need to think through the implications that these new technologies will bring and prepare for significant changes. In our final chapter, we discuss how stakeholders around the world can position themselves to benefit fully from automation’s potential while avoiding its pitfalls.