The spread of democracy across the globe is one of the most remarkable stories of the last two centuries. That may be easy to miss amid the current malaise over democratic backsliding. Russia and China grow more repressive by the day, India has fallen for Prime Minister Narendra Modi’s ethnic majoritarianism, and throughout Europe and the Americas right-wing populists seem to be chipping away at political institutions.

But from a longer-term perspective, the broadening footprint of free government is more striking. As late as 1900, the world contained just five electoral democracies. ¹ By 2022, there were 90. From Armenia to Zambia, kings, juntas, and strongmen have lost ground to governments that face at least some accountability to voters.

One of the greatest challenges for political science is to explain how this came about. There are two main schools of thought — but both leave questions unanswered.

One view attributes political change to deep historical causes. For Karl Marx, the driving force is economic development. As technologies of production advance, one class rises, another falls. Democracy prevails when the bourgeoisie wrests power from a weakened aristocracy. In the 1950s and 1960s, “modernization theorists” such as Seymour Martin Lipset revived the idea that economic progress fuels political change. Others echoed Marx’s emphasis on class-based revolutions — either, like him, associating democracy with an ascendant bourgeoisie ² or instead linking it to mass movements of workers. ³ This image of popular uprisings also permeates media. Democracy, journalists suggest, is born in public squares, when “the people” explode into a “velvet,” “color,” or “Facebook” revolution.

But if economic development is the key to democratization, why have some countries — like India — become democratic while remaining poor and agricultural, while others — like Singapore — remained authoritarian as they grew rich? Clearly, there is no income threshold at which free government appears automatically. And if the driving force is revolution, why do autocratic elites not see this coming and react accordingly? Even if they cannot freeze their advantage, why not settle with the rising groups on some institutional compromise rather than wait to be swept onto the dustheap of history? Why not share power before they lose it all? 

The second school of thought picks up here. In this view, rulers are not swept away by irresistible forces — they choose to democratize because they think doing so is in their interest. They might think this for several reasons. Perhaps they reform precisely to forestall a revolution that would strip them of both power and wealth. ⁴ A more open government might be the change needed to keep things from changing. Autocrats facing military threats from abroad might give men the vote to motivate them to defend the country. ⁵ Or, if rulers are split into factions, one of these might try to outflank another by enfranchising its own potential supporters. ⁶  

In fact, one can think of many reasons a ruling group might deliberately choose to share some of its power. Political science journals are full of ideas along these lines, often formalized in abstract, game-theoretic models. Democracy emerges as the equilibrium — the outcome that results when all actors choose their best strategies against those of the others. 

Both approaches have a satisfying scholarly gravitas. In one, democracy is the result of powerful historical forces. In the other, it’s a rational choice, a conclusion derived from postulates with a bit of rigorous math. Both feel like what a social science explanation should feel like — logical, weighty, serious.

But are they right? What if democracy is not the automatic outcome of deep structural factors? What if it’s not even the result of self-interested calculation? What if, instead, autocrats just slip up?

***

About 10 years ago, I started reading about all the episodes in which democracy has emerged since 1800. (Depending on whose definition and list one uses, there are from 51 to 183 of these.) I sought out a variety of sources ⁷ with the hunch that things were often not as neat as these scenarios implied. My goal was to establish exactly what had happened in each of these episodes — and to see to what extent the details would fit the main arguments put forth by political scientists. A few cases I knew well — such as the collapse of Soviet power in the late 1980s and early 1990s — resembled neither an inevitable revolution nor a well-planned, rational process. They reeked of chaotic improvisation and serial blunders.

In trespassing on the terrain of historians, I was not doing anything radically new for a political scientist. “Qualitative” approaches — involving the interpretation of individual cases or narrow comparisons — have always appealed to some as alternatives to the “quantitative” methods that use statistics to interrogate large datasets. These days, “mixed methods” that combine elements of both are particularly fashionable. What was a little unusual — if not reckless on my part — was to attempt a systematic, qualitative comparison of so many cases. Any deadlines I set for myself soon disappeared in the rear-view mirror. 

Of course, historians do not always agree on how to interpret any given case, and the record can be difficult to unpack. Writers of history have their own political beliefs and worldviews. Newspapers and magazines target particular readerships. Memoirs and interviews may be distorted by self-justification, and while private diaries are less likely to be self-serving, they may be partial or episodic. For each case, I used multiple sources, often following leads suggested by the materials themselves. To check my work, a research assistant reclassified a random sample of the cases from scratch; our level of agreement was high.

I compiled a synopsis of each episode. I looked for observable implications of the various deliberate-choice theories of democratization. Did the political transition follow protests, strikes, or other mass actions? Over what issues? Did democratization occur around a time of war, civil war, or significant threat of these? Did the reformers think newly enfranchised citizens would vote for them? 

The more I read, the stronger my hunch grew. There were cases that plausibly fit each of the rationales I had distilled from the explanatory literature. Just not that many.

***

Of course, history is not short of revolutions — some of which have led to freer government. On various occasions, a popular uprising has overthrown a dictator, opening the way to broader participation. And in some of these cases, it’s hard to think of steps the leader could have taken that would have saved him. As the Great Depression tanked Ecuador’s economy in the 1930s, mass street protests and strikes forced the fraudulently elected president, Juan de Dios Martínez Mera, to resign, leading to an unusually free election. (The flicker of democracy was short-lived: Martínez’s successor was removed by the military after declaring himself a dictator.)

There are also cases that look like unforced, deliberate choices. The argument that elites share power to forestall revolution fits South African President Frederik Willem de Klerk’s surrender of apartheid in 1994. At that point, fear of unrest was a powerful motivator. Italian Prime Minister Giovanni Giolitti’s decision to broaden the franchise in 1912 demonstrates the role of war. As he sent conscripts to conquer a colony in Libya, it became “impossible to deny them the vote.” ⁸ And in the U.S., voting rules have often been shaped by partisan competition.  

So, yes — democracy does sometimes emerge through irresistible revolutions or the deliberate decisions of incumbents. Yet, from close up, these explanations look a lot less powerful than expected. Both often clash with the historical facts. Indeed, when I tallied up the cases, less than a third appeared to fit any of the main deliberate choice arguments, and in only a handful of episodes were rulers swept away by mass uprisings despite making optimal decisions. Far more often — up to three quarters of the time — change came because the ruler made a mistake.  

Many revolutions are far from inevitable. More often than not, rulers simply miscalculate. In 1848, after Parisian mobs stormed the Tuileries Palace, France’s King Louis Philippe I had to shave off his whiskers and flee to England in disguise. A ragbag provisional government then called an election with close to universal suffrage. This was hardly unavoidable. In fact, the pressure only escalated to street protests after the government banned a banquet organized by the opposition — and then again after troops shot dead several dozen protesters. As rioting spread, Adolphe Thiers, who Louis Philippe had appointed to form a new government, begged the king to approve a franchise reform that would have increased suffrage from about 0.5% to just 0.8% of the population — and likely returned calm to the streets. Even the reformers whose banquet was banned were asking only for 1%! Yet the king refused point-blank. ⁹

The revolution caught everyone by surprise. Alexis de Tocqueville, who served in the provisional government, wrote later that he had heard all three of Louis Philippe’s closest advisors describe the events as “a pure accident.” The revolutionaries were, Tocqueville wrote, “as astonished by their victory as the defeated were by their loss.” ¹⁰  

Not only do many revolutions seem avoidable, they rarely look like the passing of a baton from one class to another. Ecuador’s Martínez was overthrown not because he represented any class in particular but because he tried to assume dictatorial powers. After 1848, Marx himself had to do a lot of creative thinking to explain the class character of what had happened in France. His essay The Eighteenth Brumaire of Louis Bonaparte is a brilliant attempt to rescue his theory from the assault of complicated reality.  

As for the deliberate-choice accounts, they occasionally fit, but more often they seem like tortured post hoc reconstructions whose superficial plausibility often evaporates when one digs deeper into the evidence. Take the argument that rulers choose democracy to assure the population of future income redistribution to the poor, thus preempting revolution — a theory advanced by Daron Acemoglu and James Robinson in an influential 2006 book. ¹¹ Consistent with this, I found that historical accounts mention popular mobilization in at least 75% of democratization cases. So far so good. But does such mobilization lead incumbents to compromise? In almost half of those 75% of cases, rulers did not strike deals with protesters — they were overthrown by them à la Louis Philippe. It was their successors who liberalized. And in more than 20% of the transitions catalyzed by mobilization, the protests were not about redistribution but over issues like military defeat, election fraud, or ethnic identity. Even when the grievances were about redistribution, democratic rebels did not always want more of it. Sometimes the rebels were middle-class citizens angry at a left-wing dictator’s expropriations. Once one considers all key elements of the argument, fewer than 10% of democratizations still fit. 

Far from rational optimizers, many leaders seemed disoriented and confused, blundering about in moments of chaos. Louis Philippe, in Tocqueville’s words, was like a man awakened at night by an earthquake and “knocked flat before he had understood.” ¹² Or take Mikhail Gorbachev’s political innovations of 1990: When a journalist asked if he was moving left or right, he admitted that he was “going around in circles.” ¹³

Political reform often feels to those involved like “a leap in the dark,” the phrase often attributed to Lord Derby to describe Benjamin Disraeli’s 1867 Reform Act, which enfranchised part of the working class for the first time. Historians of democracy keep coming back to the image. It turns up in accounts of Otto von Bismarck’s embrace of universal male suffrage in 1871, and in those of the reforms of Francisco Franco’s heirs in 1970s Spain. Italy’s Giolitti took a “jump in the dark” in 1912, while Poland’s authorities made a “leap into the dark” after World War I. ¹⁴ All in all, a lot of leaping going on. 

Myopia and miscalculation are far more obvious in contemporaneous accounts than careful planning. Even when not swept from power unexpectedly, leaders often drifted into reform because they found themselves cornered by their own previous moves. Democracy emerged as a result of their errors. Indeed, in the vast majority of historical cases, the evidence fits a scenario I call “democracy by mistake.” 

***

What is a mistake? By this, I mean a nonoptimal choice. Rulers usually prioritize staying in power. If one action has a 20% probability of saving their regime, and another has only a 10% chance, choosing the second will generally be a mistake. 

Not all choices with undesirable outcomes qualify. A losing gamble may nevertheless have been optimal ex ante. Nor, when all options are bad, is it a mistake to choose the “lesser evil.” In the example above, the best option has only a 20% chance of success — hardly attractive odds, but still better than the alternative. Of course, some rulers may prioritize something other than political survival — say, avoiding bloodshed. If such leaders pursue that objective at the expense of their survival prospects, that is a deliberate choice, not a mistake. 

“Democratization by mistake” occurs if a country becomes democratic as a result of one or more mistakes by its authoritarian rulers. Of course, not every choice made by the ruler has to be suboptimal; it takes just one significant action in the chain. Nor does that action need to be the last. A weakened dictator may quite rationally choose to step down, just as a chess player resigns once he sees there is no way to avoid a checkmate. The question is whether earlier errors brought the dictator face-to-face with defeat. And, needless to say, not every error by an incumbent results in democracy. Even those that trigger the fall of a dictator sometimes produce another autocrat. Here, those deep historical factors come back into play. Mistakes only lead to free government if underlying conditions — economic development, a favorable global environment — support such a change.

Certain types of mistakes appear repeatedly. Quite a few involve the faulty use of sticks and carrots to control the population. Judging what mix of tools fits a given context is extremely hard. Rulers may end up sliding down the “slippery slope” if concessions empower opponents to demand even more. Sudan’s General Ibrahim Abboud thought that authorizing campus debates in 1964 would allow students to let off steam; a few weeks later, he was a private citizen, shopping in the souq. ¹⁵ Yet, rejecting all concessions means missing opportunities to divide and conquer — as Louis Philippe learned the hard way. Repression can be either too weak to deter or excessive and poorly targeted, provoking backlash. 

In other cases, dictators mismanage relations within their own regime — unnecessarily alienating the armed forces or security services, antagonizing allies like business or the church, or empowering secretly disloyal agents. Then there are mistakes fueled by hubris. A surprising number of strongmen believe their own hype and call elections only to do poorly, revealing themselves to be both unpopular and inept, unable to manage even a dodgy vote. Chile’s General Augusto Pinochet was “stunned and enraged” when he lost a plebiscite on the future of his regime. Reporters record him haranguing his colleagues as “traitors and liars!” ¹⁶ (Some, by contrast, falsify so blatantly that it sparks unrest. After government candidates in Guatemala’s 1944 congressional election won 4,000 more votes than had been cast, disgusted army officers staged a coup. ¹⁷ ) Other victims of hubris — e.g., Argentina’s General Leopoldo Galtieri — start avoidable wars that do not go their way. ¹⁸ I found mistakes played a role in about 74% of the cases. 

How to be sure I had not just missed the logic that rendered a puzzling action rational? Some cases were undeniably tricky and took weeks of work, often resulting in answers I labeled as “low confidence.” But many were not so hard. As dominos fell across Eastern Europe in late 1989, Romania’s narcissistic dictator Nicolae Ceaușescu held a rally and had national TV broadcast it around the country. In the middle of the usual stilted speech, the cheering turned to booing, and Romanians watched a shadow of fear cross their leader’s face. Within days, videos were circulating of his summary execution. Many things about the regime’s fall can be debated, but Ceaușescu did not have to broadcast live. Or consider the decision of Haiti’s Jean-Claude “Baby Doc” Duvalier to televise his wife’s lavish “May Ball” across the island, a decision he surely regretted when riots broke out. ¹⁹  

Sometimes the ruler himself later admits to errors. “We made a mistake,” said Turkey’s General Kenan Evren after losing power. ²⁰ Gorbachev, asked about his missteps in 2011, listed five. ²¹ When rulers quickly reverse policies that had proved destabilizing, that offers prima facie evidence of error. Associates often warn incumbents they are blundering and urge a change of course. They could be wrong, but the alternatives they propose merit consideration. In other cases, it is pretty clear the ruler had incomplete or inaccurate information. Some are overthrown by coups they could have prevented with a little advance warning. Others, like Panama’s José Antonio Remón Cantera, let their guard down and are assassinated — usually not a deliberate choice. 

No account of history is immune to misinterpretations. Still, confidence in my conclusions is strengthened by the major redundancy I found in rulers’ errors. Many were serial blunderers. Among those judged to have made at least one mistake, about three quarters made two or more and 40% made three or more. Even if I were wrong on multiple details, the significance of mistakes would still survive. 

All this trawling through history suggests a rather different view of how democracy spread across the globe. Underlying factors like economic development still explain whether or not a country is primed to jump to democracy. (In poorer countries, dictators’ mistakes are more likely to produce another dictator.) But the triggers that determine exactly when change occurs are far more haphazard and specific than usually thought. From this perspective, history becomes both less predictable and more interesting. Deep structural forces matter, but so do the vagaries of human drama. 

Once one sees the role of mistakes in this sphere, other aspects of history also start to look less orderly. Many of the same errors — the same hubris, slippery slopes, and misjudgments of people — show up in breakdowns of democracy. Mistakes also appear in reforms of the electoral system — which often lead to defeat at the polls for the reforming party — and in the global spread of human rights treaties — which various dictators signed, only to be prosecuted later for violating them. 

***

The reactions to my paper, which came out in the American Political Science Review three years ago, were interesting. ²² It was a quick hit among opposition sympathizers in several authoritarian states like Russia and Belarus, whose dictators had proved hard to dislodge over the years. ²³ The idea that regimes can crumble after political blunders was taken by readers as a reason for hope. The regimes’ supporters were less enthusiastic. The paper even turned up as the theme of debate one evening on one of Russia’s Kremlin-friendly TV talk shows — certainly the first time its shock jocks looked to the APSR for material. ²⁴ An assortment of loyal blowhards lined up to attack the argument. 

Among Western political science colleagues, some liked the detailed historical examples and the catalog of official missteps I’d compiled. But I sensed a certain resistance. People were not quite sure what to do with an argument that departed from both the view of history as a vector of deep structural forces and the paradigm of (always) rational choice. 

What lay behind some of my colleagues’ ambivalence? Why are mistakes so hard to fit into many political scientists’ models of the world?

One point of view, quite common, is that introducing irrationality makes explaining behavior too easy. One can always call a puzzling action irrational. And since irrationality is difficult to fit into a predictive framework, such claims are hard to falsify. To play the irrationality card is to reject the analytic discipline that makes rationalistic explanations powerful. It is a sin against parsimony, a kind of social science nihilism. 

I think this is a misconception. In fact, it is not easy to explain an action as irrational. To do so, one must first define what the rational choice in the circumstances would have been. Then one must seek historical evidence about the actor’s state of mind. One has to do the deductive work that a rational-choice adept does — and then a lot of empirical sleuthing in addition. 

Parsimony is, indeed, attractive if combined with accurate empirical claims. But parsimony itself seems too easy if the claims are not true. Einstein did not mince words on this: “the supreme goal of all theory,” he wrote, “is to make the irreducible basic elements as simple and as few as possible without having to surrender the adequate representation of a single datum of experience.” A single datum!

One can fail to make the case that mistakes matter if the empirical part is not convincing. The difficulty of getting inside the actor’s head causes some to reject this approach outright as infeasible, but that ignores the fact that almost all social science involves some theory of mind. Historians do this all the time. It is hard, not impossible. 

One way to avoid the difficulties of probing individual psychology is to focus on the (very) big picture. In fields such as macroeconomics, one can bypass the detail and explore emergent properties of the system produced by interactions of countless actors. When many players each make random errors that offset each other, we can safely ignore them. Similarly, why worry about individual mistakes if arbitrageurs have incentives to “correct” them, profiting from irrational bidders, and in the process rebalancing the market?

Unfortunately, these points do not help here. When historical outcomes turn on the individual actions of a few leaders, there is no averaging or arbitraging the errors away. They are a feature of historical analysis, not a bug. 

A second objection to emphasizing mistakes is that it requires one to think about counterfactuals — specifically, the path that would have been optimal but was not taken. Of course, one cannot observe what did not happen. Some have clung to the notion that we can explain history exclusively in terms of observations, without considering alternative worlds. ²⁵ But counterfactuals are built into the way social scientists think about causality these days. ²⁶ And it is hard to tell a story about what happened without making claims about what caused what. Rather than rejecting counterfactuals, other scholars have turned to evaluating their quality. ²⁷ The most reasonable ones are those that require only minimal “rewriting” of history. This problem may seem overwhelming in the abstract, but it requires no extravagant leap to imagine Ceaușescu waiting until after his rally to authorize its broadcast. 

Although I disagree with them, there are reasonable methodological arguments for restricting attention to deliberate, rational choices. But more obscure psychological factors may also contribute. For one thing, it feels counterintuitive to suppose that good things (like democracy) follow from bad things (like mistakes). Historians call this the fallacy of identity — “the assumption that a cause must somehow resemble its effect.” ²⁸ Similarly, we suppose that a historically significant event like democratization must have a historically significant cause, rather than resulting from a blunder. Satisfying accounts make the world feel less absurd. A focus on accidents violates “the norm of reporting history as a good story, with all the relevant details neatly accounted for.” ²⁹ Finally, there is the historian’s occupational hazard, “hindsight bias,” which makes what actually happened seem inevitable. Mistakes — except in Greek tragedies — are usually thought to be avoidable. All these predispositions incline us to scrub historical accounts of the uncertainty, meaninglessness, and randomness that infuse real life. 

The danger is that we end up believing neat stories rather than engaging with messy truths. The way forward is surely to combine the rational and irrational into more sophisticated frameworks. At times, political scientists seem to think the choice must be between two extreme positions — either only study those actions for which a fully rational logic can be found or reject all a priori assumptions and generalizations. It does not have to be one or the other. Even in economics, the home turf of simple models, the move recently has been toward incorporating blind spots and psychological biases into the mathematical equations. ³⁰ To assume that science cannot handle this degree of complexity would itself be…a mistake.