Luohan Academy

Causes, Consequences, and Solutions of Equality Issues in the Digital Era

Event materials

  • Transcript


Michael Spence:

Thank you, Ginger. And thank you to Luohan Academy for allowing me to talk. So, I'm going to start my stopwatch, so I shut myself off before you do. Long asked me to talk a little bit about this subject from a global point of view. And so let me try to do that. I want to start, however, by sharing a perspective that I have, that some of you may not agree with. Trends in inequality, both positive and negative have been and are influenced by many things. Technology breakthroughs, including the cascade of digital ones we’re now experiencing, structural transformations associated with growth and development, economic and social policies of a variety of kinds, including especially tax and investment in human capital, institutional arrangements and laws, and beneath all that of course social norms.

In the end, the conclusion I draw from this is that inequality is not really just a technological issue. It's a constrained social choice. The broader subject is, from my point of view, political economy. Let me turn to sort of long-term patterns as a kind of preliminary to the discussion where we dive in deeply on a number of aspects of this. Globally inequality has been falling pretty steadily in the  entire post war period. This is the result of a phenomenon that I and others call convergence. And it represents a sharp break from the previous 200 years before World War II, where the developing countries in this more recent period started to grow initially, somewhat slowly, but steadily and eventually more quickly than the advanced countries.

None of this is surprising economically to those who study it, but basically they do this by importing and adapting technology. And now digital technology and then increasingly developing their own and also by leveraging the accessible demand and increasingly open, until recently, global economy. And that produces a high potential productivity growth in the tradable sector that spills over into the rest of the economy, creating resources that are then invested in tangible and intangible assets that allow the domestic sectors to become an additional powerful growth and employment engines.

This global convergence strand is still working, but I want to flag an issue I'll come back to in a minute, which is there are now serious questions about whether the current collection of low-income countries with more than a billion people in them will actually somehow get trapped and not be participants in this convergence process.

If we turn now to within country inequality patterns, within the developing countries by almost all standard measures of inequality, inequality declined or equality increased for roughly 30 years after World War II. And then that pattern reversed very substantially actually, and inequality has been rising since then. There is considerable variation across countries and levels and rates of change in these measures. And that reflects, again, differences in social choices and responses to the underlying technological and market pressures. Contributing to this pattern of rising inequality was globalization at scale, particularly in the last few decades. That's probably run its course. Labor saving and skilled bias, technological advancement, and digital technologies are factors in both of those things and increasingly major players in both.

Within the low-income developing countries, think of them as late or slow starters in the growth and development process, labor intensive manufacturing as a source of comparative advantage is losing its power as an employment engine. This is because of the breakthroughs in artificial intelligence and advanced robotics that I’m sure we'll hear more about. As the share of labor cost declines, labor costs become increasingly less decisive with respect to location. And so at this point, it's not clear what's going to replace it. The development economists like Danny Rodrik think of this as a very major question, contracting growth and development strategy.

Within the emerging economies, think of these as middle and middle to high-income countries where most of the world's population now lives, inequality levels and trends vary really quite widely. There is a tendency for growth and development processes to produce widening inequality. As we know, from the work of Kuznets and W Arthur Lewis, at least for a while. That is true even if everybody is gaining albeit at different rates. This is generally regarded, at least in high growth countries as a relatively benign form of inequality, provided it doesn't get too extreme. That's said that's only one dimension. Digital technology has and will play a powerful role in making growth patterns inclusive. The Luohan Academy's first major study documented this for China, but then you can see the same things going on in India and increasingly in a wide range of middle-income countries. These are countries that have left the labor intensive sectors behind either by exporting them to lower income countries, or by making them digitally capital-intensive thanks again to AI and advanced robotics. One thing I think it's worth noting and is not well studied yet, is there has been an explosion of entrepreneurial activity globally, especially among digital and internet companies, you will find unicorns growing almost like weeds on a global basis. It's true the significant batches of them are in the United States, China, and Europe, but the remainder are growing very quickly. I attribute this to talent, low entry barriers associated with digital platforms, and digital ecosystems, low capital requirements, platform-centered ecosystems, and an interesting development, which is a kind of global expansion of the entire supportive ecosystem constructs, including financing that enable this development. I view this very positively. Let me turn now to digital transformations briefly. There are many dimensions of these transformations and much of the discussion, at least in the developed world  is conducted through the lens of work and employment and skills. That's important, but there are other dimensions such as access to markets and services that are important. We actually have automation not only in things that directly relate to jobs, but we have automation in the creation of markets, in expanding access to them, in improving their efficiency, enclosing informational gaps that have an exclusionary effect for subsets of the population frequently poor or low access subsets of the population.

And this pattern, which I think is really important, goes well beyond the economy, narrowly defined to things like health, the delivery of primary care to low access populations and to education.

A third lens for thinking about digital transformations has to do with big structural changes in the economy. I don’t want to go into this, but a classic example that we’re right in the middle of this traditional retail being adapting to and being disrupted by the digital technology is associated with that with e-commerce and that of course accelerated dramatically in the pandemic.

If I had time, I would talk a little bit about something I discovered, which is Marchetti’s constant , but I won’t. Maybe we can talk about it later. To me, terminology is important. So I’m going to talk because I find that words like autonomy, automation, mechanization and so on are used a little bit too interchangeably in some of the literature. Mechanization is old. We saw it in its most powerful form in the industrial revolution. This was really mainly what we would now call machine augmentation. Machines dug trenches, but they didn’t do it by themselves. Blue collar jobs were eliminated. It was disruptive. It did require the acquisition of new skills.

All of those things are common to the current situation. But after difficult transitions, it drove astonishing increases, these waves of technology and mechanization, in productivity and incomes. But these machines were not autonomous. Or to put it slightly flamboyantly, the only programmable computers in those days, in the pre-digital era, were people. Does this automatically produce rising measured inequality? The answer is not necessarily. It produces a rising inequality only if there's a positive correlation between the winners, the so-called winners and the underlying income and wealth distributions. The effect really depends on where the impacts land and how long they last. And I think that being able to distinguish between transitory effects, as opposed to sort of things that might be very long term or permanent, is important. I put in my notes here. The concept of losers in these transitions is very much dependent on the time horizons. Loser is not a permanent condition.

Now let me talk very briefly about the kind of modern digital transformation. Automation implies at least to be the ability of machines to act autonomously. This is genuinely new and puts us in somewhat uncharted territory, because it is a product of the digital technologies. It's new, not because it writes people out of the script, but it does it in a different way in what we don't really know how to call it, but in the information processing, control, coordination, decision and transactions, dimensions of the economy, we have mechanization and automation for the first time. Or put differently, through the lens of work, the white-collar layer of the economy was largely devoid of machines in the pre-digital era.

I think of automation as one of the big opportunities and challenges, digital machines replace human beings. And I’m going to talk about two versions of automation. It's just the way I think about it. It's probably more subtle ways to do it. Digital machines replace humans in automation one in what are called routine or codifiable tasks. This has been well studied by David Autor and colleagues. It basically affects jobs that we know how to do, but we also know precisely how we do them, so we can code it. I think the main thing that's been discovered is that this kind of automation, which I call ‘automation one’, did affect predominantly middle-income jobs, and it's produced a pattern again, well documented by David and others of job and income polarization, at least on a transitional basis. And as an interesting aside it didn't generate a lot of employment. ‘Automation two’ is machine learning plus automation one. This has produced a dramatic increase in the range of tasks and activities that could be automated, because it includes sequences of tasks, some of which cannot be coded and it also includes tasks that human beings can't do it all.

We will, I think, almost surely see further rounds of automation and augmentation. This will occur across the entire economy to think of it as a collection of super powerful general purpose technologies. That said we're not very far along and, certainly in terms of implementation, data from MGI (McKinsey Global Institute) and others suggest that, the penetration rates across sectors in the economy, pick any economy, vary enormously and widely. I was very struck, but there's lots of examples of these technologies, but the one that struck me was Deep Mind and its recent announcements that it has developed algorithms and machine deep learning algorithms that are capable of reasonably active, accurately predicting the 3-D structure proteins from the amino acid sequence cell itself.

That I believe just listening to the scientist is an enormously powerful productivity increasing tool for biomedical science and its applications to drugs and so on. So let me just conclude. Is this going to produce rising inequality? I think to me, it's too soon to tell. It's certainly not clear that this more powerful kind of automation in this layer of the economy will operate as the previous one did disproportionately on middle-income jobs. But we'll have to see. I'll just finish with this. I think it's almost certain to produce a burst of productivity growth and reverse most of the recent trends we saw. It will show up well outside the bounds of conventional measures of economic performance in health care, biomedical science, the energy transition, in education, and in the inclusive growth pattern.

I'm pretty optimistic especially for societies that invest heavily in human capital in parallel, as fast as they can with the digital technologies. Thanks for listening to me.


For more information, please visit Luohan Academy's youtube channel: Luohan Academy


to leave a comment