What will automation mean for wages and income inequality?

What will automation mean for wages and income inequality?

In this podcast, we examine how technology has affected employment and incomes in manufacturing and other sectors and whether automation could widen the gap between high- and low-income jobs.

In this episode for the McKinsey Global Institute’s New World of Work podcast, Sree Ramaswamy and Anu Madgavkar, both partners at the McKinsey Global Institute, look at what automation could do to workers’ wages, and how that could, in turn, further widen the gap between the rich and the poor. The conversation also explores how smart policies could prevent unemployment and wage friction in the future.

What will automation mean for wages and income inequality?

Podcast transcript:

Peter Gumbel: Hello, my name is Peter Gumbel, and welcome to the McKinsey Global Institute’s latest podcast on the new world of work. Today, we’re going to be talking about some of the income inequality and other issues related to automation, but not just to automation; the broader issues in society of people, especially in middle-wage jobs, who are being affected by today’s global economic trends.

To talk about them, we have two partners from the McKinsey Global Institute: Anu Madgavkar, who is a partner based in Mumbai, and Sree Ramaswamy, who’s a partner based in Washington, DC. Thanks both for being here.

I would like to start with you, Anu, because you have done a lot of work around the issue of income inequality and came out with a landmark report in 2016, Poorer than their parents. Could you talk initially about what you have been seeing, in terms of the pressures on people, especially middle-income workers, mainly in advanced economies?

Anu Madgavkar: When most people think about income equality, there are two predominant lenses that they apply. One is just the fact that there are a lot of people in poverty, and there is an absolute issue of, “Do you have enough?” And then there is this issue about not being as well-off to the same extent as richer people; that is, the gap between richer and poorer people widening.

While both these are important, we looked at yet another trend that we think is even more significant, which is that for large shares of the population in the advanced economies, there has really been no positive movement or no sense of progress in terms of where their incomes have gone over the last one to two decades. When we looked at the data across the US as well as a set of European economies, we found that for as many as two-thirds of households at the level of market income, which is what they earn mainly from wages, incomes were flat or falling (Exhibit 1). Even after taking into account transfers that they received from their governments, there was still something like 25 percent of the population that didn’t progress and, in fact, had fallen behind.

Poorer than their parents? A new perspective on income inequality

So, this is a very significant trend, because apart from shaping living standards and access to opportunity, we also found that it influences behaviors and attitudes about what people expect and how they view the ability of the state or how they view issues like globalization, foreign trade, and immigration. A lot of these attitudes are colored by this phenomenon of just not having made progress.

Peter Gumbel: What had you found to be the real reasons for this phenomenon? Was it mainly the recession and the global financial crisis? Or were there broader issues that were really driving this income trend?

Anu Madgavkar: The recession did play a role. But what we found is that the recession actually unpeeled and revealed a set of underlying trends that were driving this phenomenon. These trends differed based on the countries we looked at. In some of the countries in Europe where aging is important, or where the structure of the household is changing, you see smaller households, older people, and fewer earning members per household. Those demographic shifts had begun to accumulate in countries like Italy, leading to this notion that household income wasn’t advancing.

In the US, demographics really weren’t the issue. There was some unemployment, but it was more the issue that the share of wage income and the overall pool of whatever economic growth there was, that share of labor income was falling. And it was also the case that for many less-skilled people, they were not able to either find work, or they were just working fewer and fewer hours. There wasn’t enough work to go around, and then there weren’t enough wages coming out of that work. So, these two factors were the most important in the US.

Peter Gumbel: When you looked at the falling wage share, could you identify the extent to which automation was really driving this? Or were there other factors?

Anu Madgavkar: It wasn’t easy to disaggregate, but what we did find is that it was highly correlated with educational levels and skill levels. We did find that mid-skilled and lower-skilled people were the most affected. There was enough evidence to suggest that skill biased technology. Technological absorption and adoption by all these businesses has, over a period of time, softened wages and reduced relative bargaining power of less-skilled people. That appears to have been a phenomenon that had a lot to do with this issue.

Peter Gumbel: Sree, just turning to you now. You have just finished up a report on manufacturing in America. Are you seeing the same tendencies in terms of manufacturing? Is there very intense pressure on these low-skill and middle-skill jobs?

Sree Ramaswamy: Yes. We do see the same sorts of pressures. One of the things that we did in the Making it in America research was to pick up almost where Anu’s work left off, talking about the labor share of the economy, the demand of national income that goes to workers. And you do see the decline, and it’s accelerated in the last ten years or so because of the recession.

But these are long-run structural trends that have been going on for 20 to 25 years. One of the things we did was to look at how much different sectors of the economy contribute to that decline in workers’ share of GDP, and we found that manufacturing accounts for about somewhere between 65 to 70 percent of that decline in labor share and GDP over the last 25 years or so. That is a pretty significant amount, considering that manufacturing itself only employs about 9 to 10 percent of the labor in the US.

What explains that amount of contribution? It really is a factor of how manufacturing in its heyday used to be a pretty significant driver of middle-income growth. What we’ve seen over the last 25 years is both a reduction in the size of manufacturing in the economy and a reduction of labor within manufacturing. The issues that Anu was talking about in terms of the technological change and the skill-biased displacement, you do see that in the income that’s going to certain types of workers within manufacturing. But we also found a broader issue, which was the decline of manufacturing overall in the US economy. Not so much as a share of the US economy; as economies industrialize, people spend more on healthcare and education. That’s certainly true. What we looked at was the absolute amount of output in the US manufacturing sector, and we found that in absolute terms, the output growth in the US manufacturing sector has been slowing for the last 25 years, very different from a sector like agriculture, for instance, where we’re producing twice as much output now as we did 20 years ago. In manufacturing, we’re producing almost about the same output as we did 20 years ago.

Peter Gumbel: It’s quite interesting to note that you focus on both manufacturing and agriculture there, because in our new report on automation, we make the point that in agriculture, for example, employment did drop very, very sharply, and in manufacturing more recently, it’s also dropped very sharply and these seem to be part of a natural economic trend, as sectors shift. But do you think that this manufacturing decline is irreversible? Or is there is there something that could stop, or at least slow, the decline?

Sree Ramaswamy: Certainly, in the US context, the absolute decline in manufacturing is something that is not inevitable or should be accepted as inevitable or irreversible. Manufacturing used to be a major job driver, 50 to 60 years ago (Exhibit 2). It used to account for 30 to 35 percent of employment in the US. We are not going to go back to those days. In the US or in other advanced economies, manufacturing is not going to be this mass driver of low-skill or middle-skill jobs the way it used to be.

History shows that technology has created large employment and sector shifts, but also creates new jobs.

But there are other reasons that manufacturing matters to an economy, especially an economy like the United States. One of those is the fact that even though it accounts for 9 or 10 percent of employment, it drives 35 percent of productivity growth. It drives about 60 percent of exports. It drives about 70 percent of all the private-sector R&D that goes on in this country. If you think about productivity or exports or R&D, those are elements that mark how competitive a nation is in the global economy. And you see that manufacturing disproportionately contributes to those factors. That’s why we think it makes sense to look at manufacturing, the decline that has happened, and try to understand how much of it is a natural consequence of an economy maturing, and how much of it was born by policy choices or other choices that were made by firms that could be reversed.

Peter Gumbel: Do you have a sense of the role of automation and technology more broadly in this decline? Or is it more a case of there are a whole slew of factors that are bringing this phenomenon to bear?

Sree Ramaswamy: Broadly, if you compare the US manufacturing sector to the manufacturing sector in countries like Germany or Korea or Japan, you find that those countries are far more automated than the US is. And a lot of that is because of the industry mix. When you look within the manufacturing sector, there only two industries, automobiles and electronics, that tend to drive the vast majority of industrial robots. Within those industries, certainly you find that the US automobile industry is no less or no more automated than the German automobile industry. It’s just that the US manufacturing sector has a whole lot of other industries—food and beverage processing, chemicals, for instance—that tend to be less automated. As a whole, the US manufacturing sector is far less automated than the German or the Japanese or the Korean manufacturing sector.

Peter Gumbel: What are the options available to help low- and middle-wage workers, not only in manufacturing, but more broadly in the economy? Where are the new jobs? What are the skill requirements for them? What will they have to do? And what will policy makers have to do to help them manage this transition?

“The big challenge is not that the jobs won’t exist. The big challenge is how do you make the transition very fast and as friction-free as possible.”

Anu Madgavkar: The thing about automation is that it’s hollowing out types of work which are very predictable, whether that is the so-called higher-skilled knowledge work—office-related routine work—or on the more physical side, very predictable work, maybe in—on factory assembly lines. The jobs of the future at various skill categories will exist that are fundamentally less predictable, where the environment is not something that you can know, where human judgment and cognitive ability at whatever skill level still matters (see interactive). There might actually be a lot of this low-skill kind of work in sectors such as construction, for example, or in sectors such as landscape gardening, for example.


For higher-skill workers, it might be a lot around creative, managerial, more interaction-oriented types of work. The jobs of the future will exist. The big challenge is not that the jobs won’t exist. The big challenge is how do you make the transition very fast and as friction-free as possible.

There are a multitude of factors that actually make transitions quite difficult between types of jobs. People don’t have the skills. People don’t have the means to figure out where the job demand lies. They don’t have the ability to actually physically move from where they live to where they need to work and therefore, working with both employers on the firm side but also with communities, with groups, with the educational system to make many of these friction-causing issues to be dealt with is going to be very, very important in the future.

Peter Gumbel: And are there any precedents out there for helping people through these transitions?

Anu Madgavkar: There are precedents, and there are lessons that can be learned from smaller-scale initiatives, experiments, pilots, or even larger systemic policies that have been implemented. For example, Germany uses a model of work sharing, which said that when work is either getting automated or falling off due to temporary demand slowdowns, you actually take whatever work has to get done and spread it out across a larger number of people. That way people are not out of work for any point in time and keep their skills alive in that context, and don’t drop out completely.

The apprenticeship system can streamline a formal education for young people, along with on-the-job training, and allows them to move quite seamlessly and become part of a workforce, even as they complete their formal education. All of this in the European model has actually worked well because these are almost trilateral partnerships between the government, the firms or the companies, and then the workforce. Labor unions, trilateral partnerships, cooperation, and ways of working together to bring solutions to bear can work.

We have seen smaller examples of things like colleges, which have worked very closely with industry associations to have more clarity around what skills are in demand, say, in the auto industry, and to create those slivers of skill certifications that allow somebody to zoom in on acquiring a marketable skill and then go, going to the market with that, rather than spending a whole amount of time trying to acquire a whole bunch of skills, which may not be relevant.

Peter Gumbel: Sree, from your perspective in manufacturing, what are the actions that can be taken to help communities which have been affected by the layoffs and displacement?

Sree Ramaswamy: One of the things that the research shows quite clearly is a divergence, almost a bifurcation, in the prospects of different types of industries and different types of companies in dealing with these pressures in manufacturing. For instance, the largest multinational firms have done quite well. They’ve bolstered very solid revenue growth over the last 20 to 25 years. Their returns on capital and their profit margins tend to be quite good, or even higher than some of their global competitors.

A minute with the McKinsey Global Institute
MGI partner Sree Ramaswamy discusses how within US manufacturing, a few sectors and large firms continue to see growth, but small and midsize firms are falling behind.

It’s the small and mid-sized firms that have really kind of borne the brunt of this issue in the US. And you see that in some other indicators, for instance, the widening trade deficit in the US, even in R&D-intensive industries, where you would expect a competitive advantage. You see it in the aging plants, in the aging equipment, in terms of smaller and mid-sized firms.

When you think about what you can do to help these communities, firms, and workers, there are a couple of points that come up. This notion of the speed of the technological change is coinciding with a slowdown in the ability of workers and firms to move to new markets, to move to new geographies, to move to new skill levels. There’s a mismatch between the speed at which we need to operate and the speed at which we are operating. Getting to some of those structural issues, what is holding back labor mobility, what is holding back skill mobility, trying to address those in a large-scale way, you do see a thousand flowers blooming. You see a lot of experiments, a lot of pilots going on, but at some point, you have to start asking the question, “What is working and how do I scale that up?” And those questions, at this point, are not being asked.

The other aspect that I would call out is that one of the challenges with dealing with technological change is that if you have an income problem and you have a demand problem, then you don’t have markets for new products and services and therefore you don’t have investment. One of the things that we have to try to figure out is how to get long-term capital to make investments in those workers or in those plants and allow those firms to start competing again, to start capturing global market share.

Peter Gumbel: And have we seen some examples where that has been achieved?

Sree Ramaswamy: When you look at the global landscape, you do see a lot of other countries where provincial governments, family investors, retirement investors—the pension funds—they do take some pretty long-term stakes in manufacturing firms. In the manufacturing sector, having a quarterly mind-set or even a three-year mind-set is not really enough, because some of these technology investments take so long, and these capital upgrades take so long even to finish making the upgrade, let alone start realizing the efficiency gains from those upgrades. You need to look at five to seven to sometimes a ten-year horizon. And you do see examples from Asia and parts of Europe where you see long-term investors starting to shore up that base.

Peter Gumbel: So it sounds like both of you are essentially issuing a call to action. As you look forward, what would you say are the most important policy responses to this issue?

Anu Madgavkar: This whole issue of inequality and its manifestation of not advancing or not making progress is too big of a problem to have a silver-bullet solution. It’s too big of a problem of recent manifestation, because we only started to see it in the last decade, and therefore I think there isn’t enough evidence base to say that, “This is what works.” We are operating essentially in an environment where we have a huge problem and we don’t really know what works. Therefore, what is really important is to first take stock of a whole range of interventions. There are interventions that will tackle the problem from the skilling perspective, which is the kind of discussion we just had. But then there are also things that will improve labor mobility or things that then get into transfers, to try and make incomes grow by smarter transfers, or transfers that are better targeted at people who really need them. And then there is a set of interventions just to boost investment and demand, as Sree said.

It is going to be important for both policy makers and the corporate sector to think about what types of interventions they really want to try out and experiment with. See what works, socialize, and build more partnerships around, but it is going to be a process of trying, learning, and building alliances, including with local communities. That’s really going to be the only way forward.

Sree Ramaswamy: I would agree with that. One of the things one of the temptations that we see is to try to fight battles that are 20 or 30 years old, or try to recreate what used to be 25 years ago. And the reality is that the technology has moved on, the markets have moved on, and you really can’t recreate what happened, what used to be. It’s much more about identifying the new opportunities. And there are opportunities out there. There is a lot of market growth in the huge numbers of new consumers joining the global marketplace. There are amazing opportunities for innovation and productivity growth. It’s much more about how to get the coordination and the scale that’s needed to take some of these ideas that are promising and get these coalitions, as Anu was saying. And they may be regional coalitions, they may be coalitions that cross the sector boundaries, and getting that long-term orientation back into the sector.

Peter Gumbel: So we really need to reinvent the future?

Sree Ramaswamy: We really need to invent the future. Yes, that’s right.

Peter Gumbel: Thank you very much, Anu and Sree. That was a podcast on the New World of Work, the latest in our series by the McKinsey Global Institute, and the guests today were Anu Madgavkar, a MGI partner based in Mumbai, and Sree Ramaswamy, who is a MGI partner based in Washington. Thank you both.

To listen to more episodes from this podcast, subscribe to the New World of Work podcast on Google Play, iTunes, or Stitcher.

About the author(s)

Anu Madgavkar and Sree Ramaswamy are partners of the McKinsey Global Institute, where Peter Gumbel is a senior editor.
More from the McKinsey Global Institute

An executive’s guide to AI

Article - McKinsey Quarterly

Why digital strategies fail

Article - McKinsey Quarterly

Strategy to beat the odds

Interactive - McKinsey Quarterly

Five Fifty: Strategy vs execution