Widening wealth gaps provoke civil war.

Thus, the heart of the MPF model is the relative wage and the wealth pump that it powers. When the relative wage declines, it leads to both immiseration and elite overproduction. Both, as we now know, are the most important drivers of social and political instability. However, outbreaks of instability—violent anti-government demonstrations and strikes, urban riots, terrorism, rural uprisings, and, if things really go to hell, state collapse and full-blown civil war—are the results of individual people’s actions. How does the model connect the structural drivers to people’s motivations? It assumes that the key role in all such events is played by extremists, those who have been radicalized and primed for aggression. When such radicals are few in relation to the rest of the population, they offer no serious threat to the stability of the regime, because they are easily isolated and suppressed by the police. But if they are many, they start coalescing into extremist organizations, which can credibly challenge the ruling class. The number of radicals in proportion to the total population, therefore, is a key variable that the MPF model needs to track.

The process of radicalization works like a disease that, as it spreads, changes people’s behaviors and makes them act in violent ways. The compartment of the MPF engine connecting structural drivers to unrest thus needs to model this dynamic of social contagion. It is quite similar to equations used by epidemiologists—for example, in forecasting the dynamics of COVID outbreaks.

The model keeps track of three kinds of individuals. The first is the “naive” type, corresponding to the susceptibles in the epidemiological framework. This is the class into which individuals are placed when they become adults. (The model tracks only individuals who are active adults; children and the retired elderly are not modeled, as they are assumed to have no effect on the dynamics.) Naive individuals can become “radicalized” by being exposed to individuals of the radical type (corresponding to infectious individuals in a disease model). The more radicals there are in the population, the higher the chance that a naive individual will catch the “virus of radicalism.”[3]

When a high proportion of the population is radicalized, sociopolitical instability is also high. Riots are easily triggered and readily spread; terrorist and revolutionary groups thrive and receive support from massive numbers of sympathizers; and the society is highly vulnerable to an outbreak of civil war. However, the relationship between the degree of radicalization and the overall level of political violence (measured, for example, by the number of people killed) is nonlinear. As the proportion of radicals in the population grows, it becomes increasingly easy for them to link up and organize, potentially leading to an explosive growth of revolutionary parties. There is also a threshold effect. As long as the power of revolutionary groups is less than the power of the state’s coercive apparatus, the overall level of violence can be suppressed to a low level. But should the balance shift in favor of radicals, the forces of the regime can suddenly implode, as we saw in numerous examples of state collapse in the previous chapter.

So far we’ve been speaking of “radicals” as though they are an interest group. But that’s not the right way to think about it. In reality, radicals usually don’t all belong to a single radical party. During periods of high political instability, there are many issues dividing the population and the elites. (We discussed this fragmentation of the ideological landscape in chapter 4.) Thus, there are many factions of radicals, each motivated by a different ideology and warring with other factions. Some become left-wing extremists, and others join right-wing organizations; still others become ethnic or religious extremists. And even within the right, as well as within the left, radical groups are splintered and may be more focused on internecine fighting than on fighting the opposite end of the ideological spectrum.

Generally speaking, an outbreak of political violence is dynamically similar to a wildfire or an earthquake. A single spark can start a prairie fire, as Mao famously said. But most sparks start little fires that go out before they manage to spread into conflagrations. Others grow to mid-size blazes. Only very few sparks start wildfires that sweep through all of the prairie. Complexity scientists have devoted a lot of attention to such processes in which the statistical distribution of the event size obeys a “power law.” Whether we quantify these processes in square kilometers of prairie burned down or the strength of an earthquake on the Richter scale—or the severity of political violence events as measured in the number of people killed—they all have the same type of dynamics. [4] In a prairie fire, whether the initial fire caused by a spark spreads depends on how much combustible material is in reach, and whether fire can jump from one patch of dried grass to another. In a revolution, whether the initial revolt against the regime spreads depends on the number of radicals (analogous to combustible material) and how well they are connected, or how fast they can expand their networks of rebellion. As a result of such autocatalytic, self-driving dynamics, an initially small event can unexpectedly blow up into a rare, large-scale disaster—a “black swan” or a “dragon king.”

Because the relationship between the radicalization quotient and the resulting scale of political violence is governed by a power law, normal statistics (e.g., the average level of violence) don’t work very well, and the MPF model captures possible outcomes by assessing the probability of really severe events, such as the American Civil War or the Taiping Rebellion. Such extreme events may not be highly probable, but we need to worry about them, simply because they have the potential to cause unimaginable human misery. Is a 10 percent probability of a second American civil war high or low? Put it in personal terms: Would you take a bet that could cause your personal extinction with a probability of 10 percent? I would not, even if offered a huge reward. You need to be alive to enjoy the prize, no matter what size it is.

Let’s get back to the MPF model. An additional element in the model is that a naive individual can become radicalized not only through contacts with other radicals but also through exposure to violence resulting from radical actions. For example, someone whose relative or friend was killed in a terrorist act perpetrated by right-wing extremists might join a left-wing revolutionary group. This second route to radicalization is also a kind of social contagion (but mediated by violence instead of radical ideology).

The third type of individual in the model, in addition to naïfs and radicals, is the “moderate” (corresponding to “recovered” in epidemiological models). This group comprises former radicals who have become disenchanted with radicalism and violence and have concluded that members of the society need to pull together and overcome their differences. The moderates differ from the naïfs in that they value peace and order above all, and they work actively to bring it about. In other words, naive individuals don’t have an active political program, radicals work actively to increase instability, and moderates work actively to dampen things down.

In summary, new radicals are created when naïfs encounter the already radicalized ones or are exposed to violence. The more radicals there are (and therefore the higher the rate of violence), the more likely that a naive individual becomes radicalized. However, moderates also have a role: “infection” by radicalism declines as moderates increase in number and exert their moderating, instability-suppressing influence.

The numbers of radicals, however, don’t grow without limit. As the level of violence increases, some radicals turn away from extremism and convert into moderates. The probability of a radical becoming disgusted with radicalism and turning into a moderate increases with the overall level of violence, but with a time delay, as high levels of political violence do not instantly translate into a social mood of revulsion against violence and a desire for internal peace. Violence acts in a cumulative fashion; several years of high instability, or even outright civil war, have to pass before the majority of the population begins to earnestly yearn for order.

The social contagion module of the MPF engine thus keeps track of the radicalization and moderation processes. Now it needs to connect to the dynamics of structural drivers of instability. This is done through the Political Stress Index (PSI), which combines the strength of immiseration and elite overproduction. [5] Popular immiseration is measured by inverse relative income (median family income divided by GDP per capita). Thus, when typical incomes fail to increase with economic growth, this factor causes the PSI to increase. Intraelite overproduction/competition is measured by the number of elites (including elite aspirants) in relation to the total population. The PSI “tunes” the probability that a naive individual becomes radicalized. When structural conditions result in high social pressures for instability, radical ideas fall on fertile soil and readily take root. Alternatively, if the PSI is low, an encounter between a naïf and a radical (or a naïf experiencing a political violence event) is unlikely to result in radicalization.

Now that we have the MPF engine, let’s use it to investigate the possible trajectories that the American social system could take beyond the 2020s. Keep in mind that this is a model (even a prototype) and its predictions should be taken with some degree of skepticism. The goal is not to predict the future but to use the model to understand how possible actions may shape different futures. The MPF engine is a kind of “morality tale,” like the story of a kind girl and an unkind girl, a narrative motif that is present in hundreds of traditional societies.

We start the engine in 1960 and first run it for the sixty years for which the history is fixed. From the MPF point of view, the most important trend during this time is the decline in the relative wage. The decline in the relative wage turns on the wealth pump, and elite numbers begin to increase in an accelerating manner. By 2020, both immiseration and elite overproduction, and thus the PSI, reach very high levels. The radicalization curve, tracking the number of radicals, which had been staying flat near zero, starts to grow after 2010 and literally explodes during the 2020s. So does political violence. At some point during the 2020s, the model predicts, instability becomes so high that it starts cutting down the elite numbers. Remember that MPF is a model, which means that it abstracts reality into mathematical equations. But in real life, instability causing declines in elite numbers is not abstract at all. Think about what happened in America as a result of the Civil War, when huge numbers of Southern elite men were killed on the battlefield and the rest were deprived of their elite status.

In the model, the cataclysm of the 2020s reduces elite numbers, which results in a decline in the PSI. Additionally, high levels of violence accelerate the transition of most radicals into moderates. The radicalization curve falls as precipitously as it rose and at some point after 2030 hits the minimum. Because radicals are the ones who drive violence, instability also declines. The social system regains its stability. But in this inertial scenario, the root cause of instability—the wealth pump—continues to operate. Gradually, elite numbers begin increasing. Meanwhile, the moderates who suppressed the 2020s peak of violence are slowly retiring and dying off. The uneasy peace holds for the next generation (twenty-five to thirty years), but then there is a repeat of the 2020s fifty years later. [6]

The inertial scenario thus predicts a rather grim future: an outbreak of serious violence during the 2020s and, if nothing is done to shut down the pump, a repeat every fifty to sixty years. What are the alternatives?