6. Kuznets Cycles (Upright or Inverted U-curves), a.k.a. “Things Get Worse Before They Get Better”
A classic cycle curve that all foresight professionals need to understand is the Kuznets curve. It was first developed by economist Simon Kuznets in the 1950s, and proposed that as economies develop, they first typically increase in economic inequality (get more socially divided), then finally decrease in inequality (get more socially coherent) once a sufficient average income is attained, tracing out an open, U-shaped (or inverted U-shaped) growth dynamic. The curve is controversial, as some economists report seeing it and others don’t. It also doesn’t always operate, as we’ll discuss below. But I believe that evidence will show it is a very common pattern in the way most economies develop, most of the time.
Inequality change is not the only kind of Kuznets curve. In general, any self-correcting (homeostatic) dynamic where some social process gets worse before it gets better, while another controlling social variable gets progressively better, can be called a Kuznets curve. Such curves are found in many processes of social change. I think they are as ubiquitous and as important as all the other curves we have discussed, but they are much less widely known at present.
More specifically, a Kuznets curve is any curve where the independent and controlling variable, plotted on the X-axis, is new wealth or freedom or learned experience or some other positive social variable, often presented in a per-capita average measure. As this variable continues to grow in magnitude, a socially dependent variable, plotted on the Y-axis, moves through three general stages. In the first stage, the dependent social variable gets worse. In the second stage, it “bottoms out” (or “tops out”, depending on how you’ve oriented the Y-axis) and may stay bad for the entire stage. In the third stage, the social variable finally improves, often getting better than its best in the first stage.
6A. Social Inequality Kuznets Curve (Inverted U-curves)
The classic Kuznets curve describes what happens with various social inequalities (income, assets, political representation) as developing nations become developed nations, and in a lesser-known variation, what happens as powerful new economic productivity technology emerges in society. Various inequalities rise sharply at first, then max out, then decrease again.
For the income inequality Kuznets curve, the proposed controlling variable is often average income or income per capita. In the first phase of the curve, as new productive technology is introduced into underdeveloped, low income per capita environments, wealthy owners of capital capture the great majority of the new value technology creates relative the rest of society, and income inequality increases. In the second phase, as the new technologically-created social wealth keeps increasing, wages and entrepreneurship throughout society rise broadly and fast enough for inequality growth to stop, though inequality remains at or near its peak. In the third phase, as total social wealth goes even higher, elites are no longer able to make superior returns with the (now old) technology, middle-class entrepreneurship keeps growing, and newly wealthy average citizens vote themselves new entitlements and redistributions, and grow business regulations, and inequality goes down again. This increasingly becomes an overregulated, underproductive situation that is ripe for new technology and social innovation (deregulation, etc.) to come in and start the curve going again.
If new technology were to stop entering this equation, the curve might have a perfect inverted U, like the cartoon version above. But as new productive technologies emerge, a Kuznets curve may start again for each of them, with capital flowing fastest first to the owners of that new technology, and starting another cycle anew. Seeing Kuznets curves in the short run can be harder, and they are clearer when you look over the long term, with large numbers of societies, as we do below.
Lindert and Williamson in “Growth, Equality, and History“, in Explorations in Economic History (Oct 1985), pp. 341-377 offer a great set of composite curves (above) showing a measure of the general trend in income inequality since the 1600s in England, and from the 1900s to the 1970s in industrialized countries in Europe and the US. Looking first at the leftmost curve, we see that the Industrial Revolution, which began in 1750 in England, at first led to great income inequality (Kuznets phase one). But by 1800, industrial towns were emerging all over England, and by 1850 most English workers lived and worked in those towns, and England had become the “workshop of the world”. Inequality stayed high during this period (Kuznets phase two) but it didn’t go higher, as so many workers were now directly benefiting from the industrial system. Over the next century (Kuznets phase three, 1860s to 1960s) we saw a significant decline in income inequality in Europe and the US, as social entitlements, tax redistribution, social enterprises and regulation all rose, while no truly disruptive technologies that weren’t simply refinements of industrial technologies emerged to start the cycle over again.
But look next at the United States curve, and you will see that it did a “U-turn” in the 1960s, after falling sharply in inequality for at least thirty years beginning in the 1930s. Income inequality started rising rapidly in the US around this time, and in other leading countries shortly after, and has been growing strongly ever since, accelerating sharply during the deregulation eras of the 1980s and continuing today, as Thomas Piketty charts in Capital in the Twenty-First Century (2014).
I would argue that the developed world since the 1960s has entered Kuznets phase one of the Information Revolution, a major new technological revolution that is once again creating incredible new wealth, and transferring it overwhelmingly at first to the owners of capital and to corporations at much faster rates than to the rest of society in the developed world. Read The Second Machine Age (2014), and its discussion of technological unemployment and our jobless recovery, to see several good arguments and data for that perspective. Accelerating automation and cyclic recession create conditions that allow periodic job shedding in the U.S., and in phase one of the Kuznets curve, new jobs for those displaced aren’t replaced by holders of capital at anything like their original rates.
As with the Industrial Revolution, I expect the US and other developed nations will eventually enter a phase two of this curve, where inequality finally tops out (stops getting worse). After some number of decades in that phase, we will turn it to phase three, where inequality goes on a long downswing again, as we saw occur in most countries in the first half of the twentieth century.
In coming years, I bet evidence-based economics will show us that there is an optimum income inquality distribution in human social systems. We want inequality that incentivizes work and innovation, but not so great that corruption, cronyism, and class-based structures dominate. I think the reported CEO to average worker pay differentials of 300-500X that we find in top private and public firms in the US and rapidly developing nations are very likely corrupting, while ratios of 30-50X, like we find more frequently in public European companies, are likely innovation-promoting. But beyond such obvious insights, we’d all like to know exactly where income distribution and firm economic concentration should be to be healthy, and how innovation and Kuznets processes will temporarily disrupt this healthy optimum.
If there is an optimum distribution, that would mean it is developmental. Many developmental processes have a normal or lognormal distribution. Many natural income distributions are lognormal, as statistician Roger Thatcher notes here. I suspect phase one of the Kuznets curve takes the affected industries and economies out of a lognormal distribution, phase two is maximally disruptive, and phase three finally restores it. For example, as the statisticians at Gapminder.org show, the world as a whole has been moving away from the two hump, OECD vs. developing nations income distribution of the 1970s and earlier (picture left), toward a fully lognormal distribution expected in the 2020s and beyond (picture right). This log normalization, as well as global mean income convergence data, argue to me that our global economic system, is on the way to a more fair and natural income distribution driven by increasing global integration due to the Information Revolution (which is still quite young), away from a Kuznets disruption caused by the Industrial Revolution.
Clearly the Kuznets curve as a concept can have significant impact on our long-term foresight, if we accept that it is operating in some relevant context.
6B. Environmental Degradation Kuznets Curve (Inverted U-curves)
After income inequality, another area where scholars have found and specifically proposed a Kuznets curve by name is in measures of environmental quality. Here it is called the Environmental Kuznets curve. It charts the common outcome that as cities develop, at first their air and water quality gets worse (phase one), then bottoms out as regulation and pollution-curbing behavior starts to gain broad traction (phase two), and then, as average social wealth keeps growing, cities add punitive fines and pay for costly environmental cleanup activities (phase three) and the environment starts to rebound, though it may never get all the way back to its pristine state.
The figure at right shows environmental Kuznets curves for two groups of countries in the 1980s. Sulphur dioxide air pollution at first got worse, then as countries reached a GDP per capita near $8K, air pollution topped out, and then it got progressively better. The green line shows countries with strong protections for property rights, and the red line countries with weaker protections. Both followed the Kuznets curve, suggesting it is a more fundamental process than capitalism and property rights, though property rights may moderate the severity of the curve.
The number and quality of trees in a country as it develops is another example of an environmental Kuznets curve. Do you think there are more trees in the US today than in 1900? The forestry service estimates there are now more. The nadir of US tree coverage apparently occurred in the early 20th century. Since the 1920s, increasingly aggressive forest management practices, and a decline in planted acreage with the rise of the green revolution and corporate farming have led us to plant vast numbers of new trees. Tough the wood represented by those trees is at present much less than what was logged, some estimates are that there are now as many trees as existed in the US in 1800. Will we ever get back to where we were in 1700? A Kuznets curve is clearly running here, in every country rich enough to have restorative soil and water management and tree husbandry programs.
6C. Social Stability Kuznets Curve (U-curves)
Political scientist Ian Bremmer has also popularized a social stability Kuznets curve, though he didn’t label it as such. He observed that as previously closed, autocratic societies are liberalized (opened up), and citizens get an increasing number of new freedoms of action and information per capita (the X-axis variable), at first social stability goes down (instability grows, phase one), then instability maxes out (phase two), then finally the society gets more stable and self-policing, eventually becoming even more stable than before it was opened. See Bremmer’s insightful book, The J-Curve: A New Way to Understand Why Nations Rise and Fall (2007), for more. Fareed Zakaria’s The Future of Freedom (2007) is another good book in this vein. The latter explores differences between liberal and illiberal democracies and how they cycle irregularly between those two states.
Some scholars have taken to calling these kinds of curves J-curves, and a Wikipedia page on J-curves offers more good examples of them in finance, balance of trade, and other domains. But that’s a poor name. U-curves is best, as that’s what they look like. J-curves is a term best used for superexponential, hockey stick growth, which looks exactly like a capital J on a non-log scale. Notice that when the Y-axis variable is a positive variable (stability) rather than a negative one (inequality, crime, etc.) the U-curve is not inverted but upright, with the second half rising higher than the first half, but often not very much higher, before the next system shock occurs, and things again get initially worse before they get better, tracing out another U.
Bremmer doesn’t cite his curve as another kind of Kuznets curve, because this area of scholarship is still underdeveloped, and it isn’t yet widely recognized that there is general class of self-improving systems curves, Kuznets curves, that are just as important as S-curves. But these curves will eventually become mainstream knowledge among academics, analysts, strategists, policymakers, and foresight professionals. In the meantime, you can use them yourself, and wait for others to catch up. Whenever Kuznets curves operate, progress happens via the “evolutionary two step”, which is one step backward, two steps forward, again and again. Almost every powerful new addition of technological capacity to a complex system, for example, oxygen production on the primitive Earth, or the first factories in Manchester, will usually makes things worse for a bit before they get better. That’s very valuable knowledge.
For a sense of the impact such curves might have on future policy, imagine how much different the recent US interventions in Afghanistan and Iraq might have been, had our political and military leaders expected and believed in Bremmer’s U-curve. If we believed growing average per capita freedoms really do create more stability over time, we might have provided aggressively subsidized cellphones, low-power television, audio and video players, newspapers, and other communications technology in Iraq, rather than keeping them expensive tools of the wealthy, as they remained until recent years (400K mobile phone subscribers in 2003, and 21 million in 2011). Such new informational openness and communications empowerment will certainly cause new problems in the short run in autocratic cultures. Even with US-led coalition forces monitoring the network and granularly denying privileges to bad actors (kill switches on phones, a bill now being legislated in California), transient new instabilities would have emerged in the second phase.
But in the Bremmer model, this new openness would have led inevitably to even greater social stability after a few more years. With accelerated communication, transparency, and information exchange in more open countries, local problems and rulebreakers of all types are more swiftly reported, not to a central authority, but to all the local and regional authorities on the network. The new stability may have been multipolar (many regional ethnic and religious stability groups) rather than unipolar, (a stronger central state), but if that’s what emerges, that might well be the best solution. That is exactly what we are seeing try to emerge in Iraq with ISIS today, in spite of all our efforts to support the central state.
Political leaders can always facilitate the emergence of a multi-state system in any country, if social conditions favor it. Long-oppressed people crave and are sometimes willing to die for the right to self-determination. Wherever civil violence or lack of governance exceeds a certain threshold, the emergence of Balkanized states with self-governance but no standing armies could be offered as referenda in any country, as proposed in Leslie Gelb and Joe Biden’s Three State Partition Proposal for Iraq in 2006, and as realized by the UN’s creation of North Sudan and South Sudan in 2011. We need only adopt the strategy.
New openness, communications access, freedom and transparency always cause new problems at first in previously restricted societies. This is one reason central authorities resist them. Another reason they resist is the threat to their power. But in the third generation, according to this reasonably validated and common-sense model, even greater stability is ultimately created, bottom-up, than existed before those new citizen freedoms (a form of wealth) existed.
6D. Technology Experience Kuznets Curve (U-curves)
The last Kuznets curve we will propose is what, in 2002, I called a “fourth law” of technology, one of a number of rules of thumb that may outline general features of how technology and society interact. Several such lists of empirical observations for technology behavior have been proposed. If interested, see my other proposed general laws of technology here, and history of technology scholar Melvin Kransberg’s six laws of technology here. My proposed fourth law of technology is: “The first generation of any technology is often dehumanizing. The second generation is generally ambivalent to humanity. The third generation, with luck, becomes net humanizing.”
With a little reflection, we can find evidence for this behavior in technological systems at every scale, and recognize it as a technology experience Kuznets curve. The per capita value on the X-axis is the experience that citizens get with the technology, either cumulatively or as an annual average. We can observe this law with respect to the beneficial effects of political civilization on humanity (our first generation was the age of monarchy, slavery, and perpetual state warfare), with industrialization (our first generation was the polluted, dehumanizing, child labor utilizing factory), with automobiles (our first generation uses dirty fossil fuels, and originally had few safety features), with televisions (our first generation are noninteractive, are poor substitutes for community participation, and they separate and de-educate us, see Robert Putnam’s Bowling Alone (2001) for details), with calculators (our first generation made us innumerate and caused us to lose mental calculation skills even when we desired to retain them), with computers (our first generation are expensive and have terrible interfaces and are restricted to an educated technological elite), with the internet (our first generation is virus-ridden, scam-infested, and far too anonymous), with video games (our first generation made us asocial, and oriented us almost exclusively to childish pursuits) with cell phones (our first generation increase motor vehicle accidents as they require too much human attention), and many other technological systems.
It is a constant challenge to the designers and users of any technology to seek ways to minimize the duration and extent of the negative externalities we so often see with any new technological deployment. Yet even with our best intentions, our first uses of new technological freedoms often take us backward, for a bit. Those who would criticize a technology as dehumanizing and unacceptable would do well to recognize the operation of this curve, and consider ways to accelerate the transition to the third generation.
Fortunately, the faster and more intelligent our technology becomes, the greater the social standard we can hold it to, and the sooner we can move it from dehumanization and disruption to enhancement in its net effect. A recent example is takeback legislation (cradle-to-cradle design and recycling of manufactured goods) a third generation of manufacturing that has increased the sustainability of European manufacturers without significantly impacting their competitiveness. There are good arguments that sustainable takeback programs would have been impossible in a world without supply chain automation, recycling automation, and other technological advances, but there is a time when such advances become affordable, and it is incumbent upon us to recognize when that time has arrived, and to advocate for the next generation to emerge.
Only in the third generation of these technologies, when they have strong intelligence built into them, do we finally see their net humanizing benefits. Cities become smart and sustainable, calculators teach us to be numerate in appropriate life contexts, cellphones become wearable, voice-driven, and smart enough to turn themselves off when we try to text while driving (let’s hope), television becomes participatory and interest- and community-based (internet TV, coming soon), wearable nonfiction video games and simulation platforms emerge which are always-on, socializing and educational, etc. A big picture overview of this process for web technologies in general, as they move from one-way to two-way, conversation-centric models, is Harri and Henry Oinas-Kukkonen’s Humanizing the Web (2013).
Many natural systems, such as body response to weight training and to pharmaceutical drugs, seem to work through an “evolutionary two step”, too. For example, the acne drug Accutane almost always makes acne worse at first. Also, one achieves better results with athletic training by taking a general (and somewhat randomised) approach of “one step back, two steps forward” in terms of training intensity, instead of aiming for a linear increase.