This cross-post from mypersonal blogexplains why I think democracy will probably recede within the next several centuries, supposing people are still around.
The key points are that: (1) Up until the past couple centuries, nearly all states have been dictatorships. (2) There are many examples of system-wide social trends, including the rise of democracy in Ancient Greece, that have lasted for a couple centuries and then been reversed. (3) If certain popular theories about democratization are right, then widespread automation would negate recent economic changes that have allowed democracy to flourish.
This prediction might have some implications for what people who are trying to improve the future should do today (although I’m not sure what these implications are). It might also have some implications for how we should imagine the future more broadly. For example, it might give us stronger reasons to doubt that future generations will take inclusive approaches to any consequential decisions they face.[1]
Introduction
There’s a strange new trend that’s been sweeping the world. In recent centuries, you may have noticed, it has become more and more common for people to choose their own leaders. Five thousand years after states first emerged, democracy has been taking off in a big way.
The average state’s level of democracy over the past two hundred years. States with sub-zero scores are more autocratic than democratic.[2]
If you follow politics, then you’ve probably already heard a lot about democracy. Still, though, a quick definition might be useful. In a proper democracy, the state’s most important figures are at least indirectly chosen through elections. A large portion of the people ruled by the state are allowed to vote, these votes are counted more-or-less accurately and more-or-less equally, and there’s no truly serious funny business.[3]
Proper democracies are something new. For most of the past five thousand years, dictatorship has been the standard model for states. We don’t know much about the first state, Uruk, but the most common theory is that it was a theocracy ruled by a small priestly class. Monarchy emerged a bit later, spread across the broader Near East, and then stuck around in one form or another for thousands of years.
Many archeologists suspectthese little bowlswere used to ration out grain to people doing forced labor. They are also by far the most common artifact found around Uruk, which is often taken as an ominous sign.
In other parts of the world, small states with noteworthy democratic elements have emerged from time to time. Certain small states in Greece, as the most famous example, were borderline-proper democracies for a couple hundred years. However, if there was any trend at all, then the trend was toward more consistent and complete dictatorship. States with noteworthy democratic elements tended to lose these elements over time, as they either expanded or fell under the influence of larger states.[4] No sensible person living one thousand years ago would have predicted the recent democratic surge.
It’s natural to wonder: Will this rise in democracy last? Or will democracy turn out to be only a passing fad—something like the Ice Bucket Challenge of regime types?[5]
Let’s suppose, to be more specific, that one thousand years from now people and states still at least kind of exist. How surprised should we be if democracy is no more common then than it was in the year 1000AD?
An Outside View
One way to approach this question is to think hard about history, political science, economics, the future of technology, and all that. Another way to approach the question is just to look at the long-run trend.
The trend, again, is roughly this: Democracy was very rare for about five thousand years, then became common over the course of about two hundred years. There was also a period of major backsliding in the middle of its recent rise.
At first glance, this pattern is not very reassuring. Two hundred years of history normally says very little. There are plenty of examples of social trends that reversed themselves after only a couple hundred years. The famous trend toward greater democracy in Ancient Greece, as an especially pointed example, pretty much burned out after a few centuries.
As a very different example, consider the rise of witch-hunting in early modern Europe. Toward the beginning of the sixteenth century, the practice took off very quickly. Tens or even hundreds of thousands of people were killed. By the early eighteenth century, though, the practice had essentially vanished again. Over a couple hundred years, the cultural wave rose and then broke.
Around 1500AD, Europeans started doing a lot of this. A couple centuries later, they thought better of it.
Some trends have survived for longer without reversals. For example, after nearly two thousand years, the spread of Islam and Christianity still have not been reversed. Some aspects of Ancient Egyptian and Near Eastern religion also seem to have survived for thousands of years. Certain very foundational trends, like the rise of states itself, also show no signs of reversing.
For about three thousand years, mummification was an important cultural practice in Egypt. Then nearly everyone became Christian.
Ultimately, if all we knew about the spread of democracy is that there has been a two-century trend, then I think that any strongly optimistic take would be a mistake. Democracy could keep spreading and then stick around. We know, though, that centuries-long social trends often reverse themselves.[6] We also know that dictatorship has been the standard mode of government for nearly all of recorded history.
Why So Much Democracy, All of a Sudden?
Of course, there are theories about why democracy has taken off. We probably shouldn’t trust any of them too much, since they are very hard to confirm. Still, so long as we keep their limits in mind, they can help us make slightly less blind predictions about the future of democracy.
One popular group of theories points the finger at industrialization. The Industrial Revolution began at roughly the same time as the rise of democracy. Looking across different countries, there is also a clear statistical link between industrialization and democratization. It is natural to wonder whether modern economic conditions are somehow more conducive to democracy than pre-modern conditions were.
Daron Acemoglu and James Robinson (A&R) have developed one influential economic framework for thinking about democratization. They argue that elites have a natural resistance to democracy, because democratization involves giving up power and subjecting themselves to policies favored by commoners. This resistance can decline if elites become less afraid of the policies that commoners support. It can also decline if elites become more afraid of revolts by disaffected commoners.[7][8]
A&R suggest that industrialization helps to harmonize the policy preferences of common people and elites. In pre-industrial economies, the wealth of elites tends to be tied up in huge tracts of land. They allow or force farmers to work the land, while taking a large portion of what these farmers produce. Elites essentially do not work, beyond their military service, or contribute much to economic growth. Income polarization is unsurprisingly extreme. In this economic context, it is natural for ordinary people to take an interest in radical land redistribution. Demands for land reform have been a cornerstone of populist movements in agrarian societies.
The Roman tribuneGaius Gracchusadvocated for land reform and the expansion of voting rights. The etching of him running for his life, a bit later on, was only available in a small image size.
In an industrial society, by contrast, it is more difficult and disruptive to confiscate the sources of elite wealth. Capital is both less divisible and more mobile than land. For example, forcibly dividing ownership of a large car factory is messier and more likely to reduce its output than forcibly dividing up a tract of land. If radical distribution is on the table, then car manufacturers can also threaten to build their factories elsewhere or to simply not build them at all. Landowners cannot really make an equivalent threat. In addition, in an industrialized society, levels of income polarization tend to be lower. These factors may make ordinary people less likely to demand extreme levels of redistribution.
Furthermore, elites in industrialized societies should be less bothered by moderate redistribution. Economic development tends to increase the value of having an educated and entrepreneurial workforce. Popular redistributive programs such as tax-funded primary schools therefore begin to benefit elites too. Industrial elites ultimately have less reason to worry about the policies that voters will demand.
At the same, these elites have more reason to worry about what will happen if they refuse democratic reforms. For one thing, industrialization tends to be associated with urbanization. Urban populations likely find it easier to coordinate and form up into crowds that can credibly threaten centers of power. For another thing, the early stages of industrialization seem to be associated with a greater reliance on mass armies. Armed and mobilized commoners are in a better position to threaten elites. They can also exert pressure simply by resisting conscription unless they receive political concessions. Finally, political turmoil of any kind may cause more economic damage in industrialized societies than in agricultural societies. Lengthy and complex industrial supply chains allow the damage to compound.
In summary, compared to landed aristocrats, industrial elites have less to fear from democracy and more to fear from standing in its way. The Industrial Revolution has helped democracy to spread by softening elite resistance.[9]
There are a lot of historical cases that this story can’t explain on its own.[10] There are also some obviously important factors, especially cultural factors, that the general A&R framework ignores. I do suspect, though, that the story captures some respectable portion of the truth.[11]
Automation and Democracy
At first glance, some aspects of the above story are reassuring. Unless something goes horribly wrong, like full-on nuclear war or worse-than-anyone-thought climate change, most developed states are very unlikely to deindustrialize in the coming centuries. We should expect a lot of the recent economic changes that have supported democracy to stick.
On the other hand, we should not imagine that the economies of developed states will stay frozen in amber. Changes will keep coming — and some of these changes might push states back toward dictatorship. I feel especially nervous about the long-run impact of automation.
Most AI researchers believe that automated systems will eventually be able to perform all of the same tasks that people can. According to one survey, the average researcher even believes that this milestone will probably be reached within the next century-and-a-half. I’m unsure whether this particular estimate is reasonable. I do agree, though, that something like complete automation will probably become possible eventually.[12] I expect human labor to lose nearly all of its value once it becomes cheaper and more effective to simply use machines to get things done.[13]
If human labor loses its value, then this strikes me as bad news for democracy. It would seem to eliminate most of the incentive for elites to accept democratization. If the suppression of protests can be automated, using systems that are more effective than humans, then public revolt would become much less threatening. Striking and resisting conscription would also, obviously, become totally obsolete as a method of applying public pressure. Elites would have weaker incentives to make any concessions at all.
Furthermore, it seems to me, elites would have more to lose from democracy. In a world without work, the two central forms of income would be welfare payments and passive income from investments. We should not be surprised if there is a huge lower class that depends on welfare payments to survive. Demands for radical redistribution might then become more common, for a few reasons: income polarization would be higher, redistribution would be simpler and less disruptive, and wealth levels would become extremely transparently divorced from merit. More moderate forms of redistribution would also stop benefitting elites at all. The resources and freedoms of common people would simply have no connection to economic growth.[14][15]
The Valley of Democracy
One perspective on our place in history is that we are living in the little valley between industrialization and widespread automation. This valley has just the right conditions for democracy to flourish. When we leave the valley, though, we will once again be entering territory where democracy can hardly grow at all.
Of course, maybe whatever economic changes are coming will be fine for democracy. No one can make confident predictions here. Still, if I had to bet, I would bet against democracy. If we put aside economic theories and simply note the extreme rareness of democracy, throughout nearly the entire history of states, then a major decline in democracy seems more likely than not. If we take economic theories into account, then the analysis becomes much murkier. To my mind, though, this analysis is not reassuring. We should mourn democracy if it dies, but we shouldn’t be surprised.[16]
Thank you, especially, to Allan Dafoe for conversations on the political impact of AI.
Disclaimer: I haven’t actually read all that much about democracy. I would not be stunned if parts of this post are silly or misleading in ways that experts could easily identify.
The democracy ratings used to produce this graph are taken from the PolityV dataset. Although the dataset is widely used by political scientists, and the broad historical picture it paints is probably roughly right, some its individual ratings are admittedly bizarre. For example, according to the Polity team, the United States is currently less democratic than it was in 1845.
“Proper democracy” isn’t a standard term. It is loosely equivalent, though, to what David Stasavage calls “modern democracy.” In his book The Decline and Rise of Democracy, Stavage mainly uses the concept of “modern democracy” to draw a contrast with “early democracy.” An early democracy is a complex society that does not qualify as a modern democracy, but that still places some significant constraints on leaders and still allows for some significant public participation. Participation can take a number of forms and may emphasize discussion more than it emphasizes elections. Opportunities to participate are also usually much more limited for common people than for elites.
The Roman Republic is one famous state that qualifies as an “early democracy.” In general, in the premodern world, early democracies were not extremely rare, although they were most likely to arise in places where states were small and weak.
Over time, early democracy persisted in some societies, but it died out in many others. It did so as societies grew in scale; it also did so as rulers acquired new ways of monitoring production; it did so finally when people found it hard to exit to new areas. It is for all these reasons that the title of this book refers first to a decline in early democracy and then to the rise of modern democracy.
People have been debating the robustness of the rise of democracy for a long time. My impression is that, among proponents of democracy, the dominant intellectual mood has ranged from quasi-religious faith in the trend (in response to early progress and Enlightenment-era interpretations of history) to grave doubt (in response to the rise of fascism and communism) to incautious optimism (in response to the end of the Cold War) to gnawing worry (in response to the past twenty years).
Here are some long quotes that give a taste of intellectual thought in each era.
Whithersoever we turn our eyes we shall witness the same continual revolution throughout the whole of Christendom. The various occurrences of national existence have everywhere turned to the advantage of democracy; all men have aided it by their exertions: those who have intentionally labored in its cause, and those who have served it unwittingly; those who have fought for it and those who have declared themselves its opponents, have all been driven along in the same track, have all labored to one end, some ignorantly and some unwillingly; all have been blind instruments in the hands of God.…
The whole book which is here offered to the public has been written under the impression of a kind of religious dread produced in the author’s mind by the contemplation of so irresistible a revolution, which has advanced for centuries in spite of such amazing obstacles, and which is still proceeding in the midst of the ruins it has made. It is not necessary that God himself should speak in order to disclose to us the unquestionable signs of His will; we can discern them in the habitual course of nature, and in the invariable tendency of events: I know, without a special revelation, that the planets move in the orbits traced by the Creator’s finger. If the men of our time were led by attentive observation and by sincere reflection to acknowledge that the gradual and progressive development of social equality is at once the past and future of their history, this solitary truth would confer the sacred character of a Divine decree upon the change. To attempt to check democracy would be in that case to resist the will of God; and the nations would then be constrained to make the best of the social lot awarded to them by Providence.
Democracy may, after all, turn out to have been a historical accident, a brief parenthesis that is closing before our eyes.
If so, in its modern sense of a form of society reconciling governmental efficiency with legitimacy, authority with individual freedoms, democracy will have lasted a little over two centuries, to judge by the speed at which the forces bent on its destruction are growing. And, really, only a tiny minority of the human race will have experienced it. In both time and space, democracy fills a very small corner. The span of roughly two-hundred years applies only to the few countries where it first appeared, still very incomplete, at the end of the 18th century. Most of the other countries in which democracy exists adopted it under a century ago, under half a century ago, in some cases less than a decade ago.
Democracy probably could endure if it were the only type of political organization in the world. But it is not basically structured to defend itself against outside enemies seeking its annihiliation, especially since the latest and the most dangerous of these external enemies–Communism–parades as democracy perfected when it is in fact the absolute negation of democracy, the current and complete model of totalitarianism.
[B]arring some sudden decline in the rate of technological change, and on the (outrageous) assumption that war will somehow be eliminated during the next half-century, it is possible to predict that…democracy will be universal. Each revolutionary autocracy, as it reshuffles the family structure and pushes toward industrialization, will sow the seeds of its own destruction, and democratization will gradually engulf it.
We might, of course, rue the day. A world of mass democracies may well prove homogenized and ugly. It is perhaps beyond human social capacity to maximize both equality and understanding on the one hand, diversity on the other. Faced with this dilemma, however, many people are willing to sacrifice quaintness to social justice, and we might conclude by remarking that just as Marx, in proclaiming the inevitability of communism, did not hesitate to give some assistance to the wheels of fate, so our thesis that democracy represents the social system of the electronic era should not bar these persons from giving a little push here and there to the inevitable.
From Anne Applebaum’s Twilight of Democracy (2020):
It is possible that we are already living through the twilight of democracy; that our civilization may already be heading for anarchy or tyranny, as the ancient philosophers and America’s founders once feared; that a new generation of clercs, the advocates of illiberal or authoritarian ideas, will come to power in the twenty-first century, just as they did in the twentieth; that their visions of the world, born of resentment, anger, or deep, messianic dreams, could triumph. Maybe new information technology will continue to undermine consensus, divide people further, and increase polarization until only violence can determine who rules. Maybe fear of disease will create fear of freedom.
Or maybe the coronavirus will inspire a new sense of global solidarity. Maybe we will renew and modernize our institutions. Maybe international cooperation will expand after the entire world has had the same set of experiences at the same time: lockdown, quarantine, fear of infection, fear of death. Maybe scientists around the world will find new ways to collaborate, above and beyond politics. Maybe the reality of illness and death will teach people to be suspicious of hucksters, liars, and purveyors of disinformation.
Maddeningly, we have to accept that both futures are possible. No political victory is ever permanent, no definition of “the nation” is guaranteed to last, and no elite of any kind, whether so-called “populist” or so-called “liberal” or so-called “aristocratic,” rules forever. The history of ancient Egypt looks, from a great distance in time, like a monotonous story of interchangeable pharaohs. But on closer examination, it includes periods of cultural lightness and eras of despotic gloom. Our history will someday look that way too.
I think this whipsaw pattern suggests that people sometimes give too much weight to recent events and sub-trends when thinking about the future of democracy. For example, this past year, public intellectuals who have written about the future of democracy have tended to dwell a lot on the impact of the coronavirus pandemic. If someone writes a new history of democracy in 2100, though, I’m skeptical that the pandemic will actually show up in a major way. Similarly, I think we probably shouldn’t read too much into reports that democracy has recently been experiencing a global recession. Although the past decade should make us feel somewhat more worried, our overall predictions for the world shouldn’t change very much.
You could argue, on the other hand, that it’s unfair to draw an analogy between the modern rise of democracy and premodern trends that burned out after a few centuries. The pace of economic and technological change is much faster in modern times than it was before the Industrial Revolution. If a political trend can survive for two centuries in the modern world, then this may be more impressive than surviving for a couple centuries when the background pace of change is low.
On the other other hand, if change will be just as fast (or even faster) in the coming centuries, then maybe this gives us an opposing reason to expect existing trends to burn out quickly. I think these two considerations might balance each other out.
An important nuance to A&R’s theoretical framework, which I’ve glossed over, is the idea that democractization is also a solution to a commitment problem faced by elites. You might think elites could placate commoners simply by making policy concessions, while still retaining their dictatorial powers. However, A&J point out that promises of future concessions can be difficult to trust. Simply stepping out onto your balcony and anncouning “I’ll never screw you people over again” probably won’t do much much to resolve a crisis. Actively reducing your ability to screw common people over in the future, by moving from dictatorship to democracy, might produce better results.
One alternative framework for explaining democratization is the elite competition framework. This framework emphasizes conflict between political elites and economic elites with limited political power. Ben Ansell and David Samuels argue that economic elites have an interest in supporting democratization as a way of protecting their wealth from political elites. They correctly believe that they have less to fear from pro-redistribution voters than they have to fear from political elites with extractive tendencies. Industrialization tends to support democratization, according to this theory, because it creates a large new class of wealthy people with limited political authority.
If we want to explain why India is a long-standing democracy and China is still a dictatorship, for example, then talking about industrialization is not going to do us much good. Clearly many other factors matter too.
The general idea that economic change can be a major driver of political change is also supported by the Neolithic Revolution. The low population densities of hunter-gatherers, along with their inability to store large food surpluses, made it nearly impossible for states to emerge for most of human history. When people transitioned to sendentary agricultural production, in several different parts of the world, this transition tended to be followed by the emergence of states. The causal link in this case is very clear.
The Neolithic Revolution is also an interesting case to consider, because it shows us that economic change does not always push in the direction of greater liberty. Hunter-gatherer groups tend to be quite egalitarian, often take consensus-oriented approaches to major decisions, and rarely hold slaves. The rise of states therefore meant the rise of inequality, social hierarchy, and slavery.
When thinking about the future of automation, we shouldn’t lose sight of the fact that the human brain is just a physical object. It was produced through a physical process, without even a blueprint or guiding plan. At first glance, then, there is no obvious reason to think that engineered objects will never be able to match its functionality. We also should keep in mind that the entire field of computer science is less than a century old. Furthermore, for most of this time, computer scientists lacked access to anything close to the amount of computing power used by human brains. The fact that we have not yet achieved anything close to full automation is not very telling.
Complete automation is importantly different than partial automation. When only some jobs are automated away, there will often an increase in demand for workers to perform jobs that haven’t yet been automated. This dynamic has allowed average wages to rise despite the tremendous amount of automation that has followed the Industrial Revolution. The dynamic should break down, though, if we eventually approach full automation. See this paper and this paper for discussions that touch on this point.
All of my analysis of the impact of complete automation assumes that people still at least kind of exist and call the shots politically. If AI systems call the shots by this point in history, on the other hand, then the appropriate analysis becomes very different.
More abstractly, I suppose, the one really necessary assumption behind my analysis is that there is some kind of thing that has political power but no longer has a major role to play in production. These things could in principle be normal humans, really heavily biologically modified humans, brain emulations, or any other sci-fi staple. The less human the future is, though, the less I trust my analysis or even the general concept of democracy to be applicable. It’s also conceivable that there will be only be a short window of time between complete automation and the world becoming a pretty much unrecognizably post-human fever dream. In that case, my analysis of the impact of complete automation might only apply to a brief (but still potentially consequential) moment in history.
I can also see at least a couple ways that complete automation could bolster democracy. First, by dramatically increasing economic output, complete automation could reduce elite fears about redistribution. If there are strongly diminishing returns on pie, then, when the size of a pie grows, the difference between the value of eating the whole pie and the value of eating half-the-pie shrinks. In a subjective sense, then, elites in a wealthy world may have less to lose from wealth redistribution.
Second, automation might make it easier to “lock in” democratic insitutions. If you can automate much of the work involved in enforcing the law, counting votes correctly, and so on, then it may become much easier to safeguard these institutions against future meddling. On the other hand, by the same logic, automation might also make it easier to “lock in” dictatorial institutions.
I feel conflicted about exactly how likely the survival of democracy is. Still, I think it’s a good habit to attach numbers to your predictions, so I’ll try to say something more precise.
Let’s suppose that one thousand years from now individual people still at least sort of exist, with a population of at least one million, and still largely govern themselves. I think, then, that there is something like a 4-in-5 chance that the portion of people living under a proper democracy will be substantially lower than it is today. For reference, depending on how you draw the line, between ten and fifty percent of people currently live under a proper democracy.
The main thing driving my prediction is the observation that democracy is very rare historically. If the recent wave of democratization is mostly a cultural phenomenon, which doesn’t reflect some deeper form of economic determinism, then historical analogies suggest that we should expect the wave to crest eventually. We’ll probably regress back toward the cultural baseline sometime in the next thousand years. If the wave does reflect a kind of economic determinism, on the other hand, then it’s not clear what we should predict. There are some at least vaguely plausible stories we can tell in which automation seriously harms democracy. These particular stories might turn out to be speculative nonsense, of course, but even if we dismiss them as such, I don’t think we can trust future economic changes to support democracy. At best, I think, we could suppose that the odds of a net positive impact and the odds of a net negative impact are roughly even.
[[EDIT: After some reflection, 4-in-5 feels too high to me. For a mix of reasons, some raised in the comment section, I’ll now give 65% as my mostly made-up credence.]]
Is Democracy a Fad?
This cross-post from my personal blog explains why I think democracy will probably recede within the next several centuries, supposing people are still around.
The key points are that: (1) Up until the past couple centuries, nearly all states have been dictatorships. (2) There are many examples of system-wide social trends, including the rise of democracy in Ancient Greece, that have lasted for a couple centuries and then been reversed. (3) If certain popular theories about democratization are right, then widespread automation would negate recent economic changes that have allowed democracy to flourish.
This prediction might have some implications for what people who are trying to improve the future should do today (although I’m not sure what these implications are). It might also have some implications for how we should imagine the future more broadly. For example, it might give us stronger reasons to doubt that future generations will take inclusive approaches to any consequential decisions they face.[1]
Introduction
There’s a strange new trend that’s been sweeping the world. In recent centuries, you may have noticed, it has become more and more common for people to choose their own leaders. Five thousand years after states first emerged, democracy has been taking off in a big way.
The average state’s level of democracy over the past two hundred years. States with sub-zero scores are more autocratic than democratic.[2]
If you follow politics, then you’ve probably already heard a lot about democracy. Still, though, a quick definition might be useful. In a proper democracy, the state’s most important figures are at least indirectly chosen through elections. A large portion of the people ruled by the state are allowed to vote, these votes are counted more-or-less accurately and more-or-less equally, and there’s no truly serious funny business.[3]
Proper democracies are something new. For most of the past five thousand years, dictatorship has been the standard model for states. We don’t know much about the first state, Uruk, but the most common theory is that it was a theocracy ruled by a small priestly class. Monarchy emerged a bit later, spread across the broader Near East, and then stuck around in one form or another for thousands of years.
Many archeologists suspect these little bowls were used to ration out grain to people doing forced labor. They are also by far the most common artifact found around Uruk, which is often taken as an ominous sign.
In other parts of the world, small states with noteworthy democratic elements have emerged from time to time. Certain small states in Greece, as the most famous example, were borderline-proper democracies for a couple hundred years. However, if there was any trend at all, then the trend was toward more consistent and complete dictatorship. States with noteworthy democratic elements tended to lose these elements over time, as they either expanded or fell under the influence of larger states.[4] No sensible person living one thousand years ago would have predicted the recent democratic surge.
It’s natural to wonder: Will this rise in democracy last? Or will democracy turn out to be only a passing fad—something like the Ice Bucket Challenge of regime types?[5]
Let’s suppose, to be more specific, that one thousand years from now people and states still at least kind of exist. How surprised should we be if democracy is no more common then than it was in the year 1000AD?
An Outside View
One way to approach this question is to think hard about history, political science, economics, the future of technology, and all that. Another way to approach the question is just to look at the long-run trend.
The trend, again, is roughly this: Democracy was very rare for about five thousand years, then became common over the course of about two hundred years. There was also a period of major backsliding in the middle of its recent rise.
At first glance, this pattern is not very reassuring. Two hundred years of history normally says very little. There are plenty of examples of social trends that reversed themselves after only a couple hundred years. The famous trend toward greater democracy in Ancient Greece, as an especially pointed example, pretty much burned out after a few centuries.
As a very different example, consider the rise of witch-hunting in early modern Europe. Toward the beginning of the sixteenth century, the practice took off very quickly. Tens or even hundreds of thousands of people were killed. By the early eighteenth century, though, the practice had essentially vanished again. Over a couple hundred years, the cultural wave rose and then broke.
Around 1500AD, Europeans started doing a lot of this. A couple centuries later, they thought better of it.
Some trends have survived for longer without reversals. For example, after nearly two thousand years, the spread of Islam and Christianity still have not been reversed. Some aspects of Ancient Egyptian and Near Eastern religion also seem to have survived for thousands of years. Certain very foundational trends, like the rise of states itself, also show no signs of reversing.
For about three thousand years, mummification was an important cultural practice in Egypt. Then nearly everyone became Christian.
Ultimately, if all we knew about the spread of democracy is that there has been a two-century trend, then I think that any strongly optimistic take would be a mistake. Democracy could keep spreading and then stick around. We know, though, that centuries-long social trends often reverse themselves.[6] We also know that dictatorship has been the standard mode of government for nearly all of recorded history.
Why So Much Democracy, All of a Sudden?
Of course, there are theories about why democracy has taken off. We probably shouldn’t trust any of them too much, since they are very hard to confirm. Still, so long as we keep their limits in mind, they can help us make slightly less blind predictions about the future of democracy.
One popular group of theories points the finger at industrialization. The Industrial Revolution began at roughly the same time as the rise of democracy. Looking across different countries, there is also a clear statistical link between industrialization and democratization. It is natural to wonder whether modern economic conditions are somehow more conducive to democracy than pre-modern conditions were.
Daron Acemoglu and James Robinson (A&R) have developed one influential economic framework for thinking about democratization. They argue that elites have a natural resistance to democracy, because democratization involves giving up power and subjecting themselves to policies favored by commoners. This resistance can decline if elites become less afraid of the policies that commoners support. It can also decline if elites become more afraid of revolts by disaffected commoners.[7][8]
A&R suggest that industrialization helps to harmonize the policy preferences of common people and elites. In pre-industrial economies, the wealth of elites tends to be tied up in huge tracts of land. They allow or force farmers to work the land, while taking a large portion of what these farmers produce. Elites essentially do not work, beyond their military service, or contribute much to economic growth. Income polarization is unsurprisingly extreme. In this economic context, it is natural for ordinary people to take an interest in radical land redistribution. Demands for land reform have been a cornerstone of populist movements in agrarian societies.
The Roman tribune Gaius Gracchus advocated for land reform and the expansion of voting rights. The etching of him running for his life, a bit later on, was only available in a small image size.
In an industrial society, by contrast, it is more difficult and disruptive to confiscate the sources of elite wealth. Capital is both less divisible and more mobile than land. For example, forcibly dividing ownership of a large car factory is messier and more likely to reduce its output than forcibly dividing up a tract of land. If radical distribution is on the table, then car manufacturers can also threaten to build their factories elsewhere or to simply not build them at all. Landowners cannot really make an equivalent threat. In addition, in an industrialized society, levels of income polarization tend to be lower. These factors may make ordinary people less likely to demand extreme levels of redistribution.
Furthermore, elites in industrialized societies should be less bothered by moderate redistribution. Economic development tends to increase the value of having an educated and entrepreneurial workforce. Popular redistributive programs such as tax-funded primary schools therefore begin to benefit elites too. Industrial elites ultimately have less reason to worry about the policies that voters will demand.
At the same, these elites have more reason to worry about what will happen if they refuse democratic reforms. For one thing, industrialization tends to be associated with urbanization. Urban populations likely find it easier to coordinate and form up into crowds that can credibly threaten centers of power. For another thing, the early stages of industrialization seem to be associated with a greater reliance on mass armies. Armed and mobilized commoners are in a better position to threaten elites. They can also exert pressure simply by resisting conscription unless they receive political concessions. Finally, political turmoil of any kind may cause more economic damage in industrialized societies than in agricultural societies. Lengthy and complex industrial supply chains allow the damage to compound.
In summary, compared to landed aristocrats, industrial elites have less to fear from democracy and more to fear from standing in its way. The Industrial Revolution has helped democracy to spread by softening elite resistance.[9]
There are a lot of historical cases that this story can’t explain on its own.[10] There are also some obviously important factors, especially cultural factors, that the general A&R framework ignores. I do suspect, though, that the story captures some respectable portion of the truth.[11]
Automation and Democracy
At first glance, some aspects of the above story are reassuring. Unless something goes horribly wrong, like full-on nuclear war or worse-than-anyone-thought climate change, most developed states are very unlikely to deindustrialize in the coming centuries. We should expect a lot of the recent economic changes that have supported democracy to stick.
On the other hand, we should not imagine that the economies of developed states will stay frozen in amber. Changes will keep coming — and some of these changes might push states back toward dictatorship. I feel especially nervous about the long-run impact of automation.
Most AI researchers believe that automated systems will eventually be able to perform all of the same tasks that people can. According to one survey, the average researcher even believes that this milestone will probably be reached within the next century-and-a-half. I’m unsure whether this particular estimate is reasonable. I do agree, though, that something like complete automation will probably become possible eventually.[12] I expect human labor to lose nearly all of its value once it becomes cheaper and more effective to simply use machines to get things done.[13]
If human labor loses its value, then this strikes me as bad news for democracy. It would seem to eliminate most of the incentive for elites to accept democratization. If the suppression of protests can be automated, using systems that are more effective than humans, then public revolt would become much less threatening. Striking and resisting conscription would also, obviously, become totally obsolete as a method of applying public pressure. Elites would have weaker incentives to make any concessions at all.
Furthermore, it seems to me, elites would have more to lose from democracy. In a world without work, the two central forms of income would be welfare payments and passive income from investments. We should not be surprised if there is a huge lower class that depends on welfare payments to survive. Demands for radical redistribution might then become more common, for a few reasons: income polarization would be higher, redistribution would be simpler and less disruptive, and wealth levels would become extremely transparently divorced from merit. More moderate forms of redistribution would also stop benefitting elites at all. The resources and freedoms of common people would simply have no connection to economic growth.[14][15]
The Valley of Democracy
One perspective on our place in history is that we are living in the little valley between industrialization and widespread automation. This valley has just the right conditions for democracy to flourish. When we leave the valley, though, we will once again be entering territory where democracy can hardly grow at all.
Of course, maybe whatever economic changes are coming will be fine for democracy. No one can make confident predictions here. Still, if I had to bet, I would bet against democracy. If we put aside economic theories and simply note the extreme rareness of democracy, throughout nearly the entire history of states, then a major decline in democracy seems more likely than not. If we take economic theories into account, then the analysis becomes much murkier. To my mind, though, this analysis is not reassuring. We should mourn democracy if it dies, but we shouldn’t be surprised.[16]
Thank you, especially, to Allan Dafoe for conversations on the political impact of AI.
Disclaimer: I haven’t actually read all that much about democracy. I would not be stunned if parts of this post are silly or misleading in ways that experts could easily identify.
The democracy ratings used to produce this graph are taken from the PolityV dataset. Although the dataset is widely used by political scientists, and the broad historical picture it paints is probably roughly right, some its individual ratings are admittedly bizarre. For example, according to the Polity team, the United States is currently less democratic than it was in 1845.
“Proper democracy” isn’t a standard term. It is loosely equivalent, though, to what David Stasavage calls “modern democracy.” In his book The Decline and Rise of Democracy, Stavage mainly uses the concept of “modern democracy” to draw a contrast with “early democracy.” An early democracy is a complex society that does not qualify as a modern democracy, but that still places some significant constraints on leaders and still allows for some significant public participation. Participation can take a number of forms and may emphasize discussion more than it emphasizes elections. Opportunities to participate are also usually much more limited for common people than for elites.
The Roman Republic is one famous state that qualifies as an “early democracy.” In general, in the premodern world, early democracies were not extremely rare, although they were most likely to arise in places where states were small and weak.
From The Decline and Rise of Democracy:
People have been debating the robustness of the rise of democracy for a long time. My impression is that, among proponents of democracy, the dominant intellectual mood has ranged from quasi-religious faith in the trend (in response to early progress and Enlightenment-era interpretations of history) to grave doubt (in response to the rise of fascism and communism) to incautious optimism (in response to the end of the Cold War) to gnawing worry (in response to the past twenty years).
Here are some long quotes that give a taste of intellectual thought in each era.
From Alexis de Tocqueville’s Democracy in America (1831):
From Jean-Francois Revel’s in “Can the Democracies Survive?” (1984):
From Philip Slater and Warren Bennis’s “Democracy is Inevitable (1990):
From Anne Applebaum’s Twilight of Democracy (2020):
I think this whipsaw pattern suggests that people sometimes give too much weight to recent events and sub-trends when thinking about the future of democracy. For example, this past year, public intellectuals who have written about the future of democracy have tended to dwell a lot on the impact of the coronavirus pandemic. If someone writes a new history of democracy in 2100, though, I’m skeptical that the pandemic will actually show up in a major way. Similarly, I think we probably shouldn’t read too much into reports that democracy has recently been experiencing a global recession. Although the past decade should make us feel somewhat more worried, our overall predictions for the world shouldn’t change very much.
You could argue, on the other hand, that it’s unfair to draw an analogy between the modern rise of democracy and premodern trends that burned out after a few centuries. The pace of economic and technological change is much faster in modern times than it was before the Industrial Revolution. If a political trend can survive for two centuries in the modern world, then this may be more impressive than surviving for a couple centuries when the background pace of change is low.
On the other other hand, if change will be just as fast (or even faster) in the coming centuries, then maybe this gives us an opposing reason to expect existing trends to burn out quickly. I think these two considerations might balance each other out.
An important nuance to A&R’s theoretical framework, which I’ve glossed over, is the idea that democractization is also a solution to a commitment problem faced by elites. You might think elites could placate commoners simply by making policy concessions, while still retaining their dictatorial powers. However, A&J point out that promises of future concessions can be difficult to trust. Simply stepping out onto your balcony and anncouning “I’ll never screw you people over again” probably won’t do much much to resolve a crisis. Actively reducing your ability to screw common people over in the future, by moving from dictatorship to democracy, might produce better results.
One alternative framework for explaining democratization is the elite competition framework. This framework emphasizes conflict between political elites and economic elites with limited political power. Ben Ansell and David Samuels argue that economic elites have an interest in supporting democratization as a way of protecting their wealth from political elites. They correctly believe that they have less to fear from pro-redistribution voters than they have to fear from political elites with extractive tendencies. Industrialization tends to support democratization, according to this theory, because it creates a large new class of wealthy people with limited political authority.
We can break down the A&R story down into three high-level pieces, which each seem pretty plausible to me:
First: The prevalence of democracy depends (in part) on how strenuously elites oppose it.
Second: The level of elite resistance depends (in part) on the costs and benefits that democratization poses for them.
Third: Industrialization has tended to raise these benefits and lower these costs.
If we want to explain why India is a long-standing democracy and China is still a dictatorship, for example, then talking about industrialization is not going to do us much good. Clearly many other factors matter too.
The general idea that economic change can be a major driver of political change is also supported by the Neolithic Revolution. The low population densities of hunter-gatherers, along with their inability to store large food surpluses, made it nearly impossible for states to emerge for most of human history. When people transitioned to sendentary agricultural production, in several different parts of the world, this transition tended to be followed by the emergence of states. The causal link in this case is very clear.
The Neolithic Revolution is also an interesting case to consider, because it shows us that economic change does not always push in the direction of greater liberty. Hunter-gatherer groups tend to be quite egalitarian, often take consensus-oriented approaches to major decisions, and rarely hold slaves. The rise of states therefore meant the rise of inequality, social hierarchy, and slavery.
When thinking about the future of automation, we shouldn’t lose sight of the fact that the human brain is just a physical object. It was produced through a physical process, without even a blueprint or guiding plan. At first glance, then, there is no obvious reason to think that engineered objects will never be able to match its functionality. We also should keep in mind that the entire field of computer science is less than a century old. Furthermore, for most of this time, computer scientists lacked access to anything close to the amount of computing power used by human brains. The fact that we have not yet achieved anything close to full automation is not very telling.
Complete automation is importantly different than partial automation. When only some jobs are automated away, there will often an increase in demand for workers to perform jobs that haven’t yet been automated. This dynamic has allowed average wages to rise despite the tremendous amount of automation that has followed the Industrial Revolution. The dynamic should break down, though, if we eventually approach full automation. See this paper and this paper for discussions that touch on this point.
All of my analysis of the impact of complete automation assumes that people still at least kind of exist and call the shots politically. If AI systems call the shots by this point in history, on the other hand, then the appropriate analysis becomes very different.
More abstractly, I suppose, the one really necessary assumption behind my analysis is that there is some kind of thing that has political power but no longer has a major role to play in production. These things could in principle be normal humans, really heavily biologically modified humans, brain emulations, or any other sci-fi staple. The less human the future is, though, the less I trust my analysis or even the general concept of democracy to be applicable. It’s also conceivable that there will be only be a short window of time between complete automation and the world becoming a pretty much unrecognizably post-human fever dream. In that case, my analysis of the impact of complete automation might only apply to a brief (but still potentially consequential) moment in history.
I can also see at least a couple ways that complete automation could bolster democracy. First, by dramatically increasing economic output, complete automation could reduce elite fears about redistribution. If there are strongly diminishing returns on pie, then, when the size of a pie grows, the difference between the value of eating the whole pie and the value of eating half-the-pie shrinks. In a subjective sense, then, elites in a wealthy world may have less to lose from wealth redistribution.
Second, automation might make it easier to “lock in” democratic insitutions. If you can automate much of the work involved in enforcing the law, counting votes correctly, and so on, then it may become much easier to safeguard these institutions against future meddling. On the other hand, by the same logic, automation might also make it easier to “lock in” dictatorial institutions.
I feel conflicted about exactly how likely the survival of democracy is. Still, I think it’s a good habit to attach numbers to your predictions, so I’ll try to say something more precise.
Let’s suppose that one thousand years from now individual people still at least sort of exist, with a population of at least one million, and still largely govern themselves. I think, then, that there is something like a 4-in-5 chance that the portion of people living under a proper democracy will be substantially lower than it is today. For reference, depending on how you draw the line, between ten and fifty percent of people currently live under a proper democracy.
The main thing driving my prediction is the observation that democracy is very rare historically. If the recent wave of democratization is mostly a cultural phenomenon, which doesn’t reflect some deeper form of economic determinism, then historical analogies suggest that we should expect the wave to crest eventually. We’ll probably regress back toward the cultural baseline sometime in the next thousand years. If the wave does reflect a kind of economic determinism, on the other hand, then it’s not clear what we should predict. There are some at least vaguely plausible stories we can tell in which automation seriously harms democracy. These particular stories might turn out to be speculative nonsense, of course, but even if we dismiss them as such, I don’t think we can trust future economic changes to support democracy. At best, I think, we could suppose that the odds of a net positive impact and the odds of a net negative impact are roughly even.
[[EDIT: After some reflection, 4-in-5 feels too high to me. For a mix of reasons, some raised in the comment section, I’ll now give 65% as my mostly made-up credence.]]