Modern China: Trends in Stability

A Structural-Demographic Analysis

As someone who has studied both Fourier transforms and data science, I greatly enjoyed reading Peter Turchin’s Ages of Discord, which applies lots of math to historical demographic data and political instability trends to predict that over the next 30+ years the United States will have more political instability than we have seen for the previous 50 years. Earlier works (including his longer Secular Cycles) applied his analysis to historical cases including much of Europe, China, and some of the Middle East, but Ages of Discord is the first to apply a mathematical application of the theory (called the structural-demographic theory) to a current state. When he first made predictions in 2010 (and followed with a paper in 2013), his claim that political instability (protests, riots, and domestic terrorism) would increase over the next 10 years was merely academic, but after he turned out to be right, more people have taken notice.

First, a very short summary for anyone who has not read Ages of Discord: Turchin claims that empires and large nations go through cycles of instability, with integrative periods (where states are stable and populations increase) and disintegrative periods (where revolutions, civil wars, and general political violence are more common). Integrative periods are then sub-divided into expansion, where populations increase; and stagflation, where populations stagnate but the number of elites continues increasing. Disintegrative periods are broken into crisis, where the elites fight with each other and the population falls as a result of wars or plagues; and depression, where the population bottoms out but elite numbers keep falling. At some point, there are few enough elites that they can stop fighting over resources and focus on mutually beneficial work. Think of it as a mash-up of Malthus and Marx, plus a lot more math.

For the short summary for the United States, you can read Turchin’s in-progress summary.

In general, historians take a low view of anyone who claims that there are universal forces guiding history, as so many past claims have quickly failed to predict future behavior (or even past behavior: better historical work usually shows the original claim only worked for the examples used). Also, even Turchin thinks that popular articles overstate his predictions: he only said that protests and riots were more likely during a 10-year period, but did not specify an exact number, date, or reason. A lot of confusion seems to come from the difference between popular theories, which claim to explain everything, and data science, which sees even a partial prediction as useful. For example, if Netflix can predict which movie you will like 50% of the time, that is a grand success even though the wrong movie is suggested half the time.

In general, Turchin would prefer that the quality of his analysis is measured on historical data (which it was designed for) and not on predictions for the future. However, most of the public interest is in predictions of the future, so we are left wondering whether he was just lucky. Unfortunately, the best time to make a bold prediction was in 2010, and his predictions for any time soon are that this is an unpredictable time but with more instability than usual. That is valuable to know, but much less exciting than the 2010 prediction.

Since the structural-demographic theory doesn’t have any bold predictions to make for changes in the United States right now, is there anywhere else we can make some predictions and test the theory? (We won’t know if the predictions are right for a while, but predictions are fun.) For the structural-demographic theory to be useful, states need to be big enough that their future will mostly be decided by internal forces (elections and protests) instead of external forces (invasion or political interference). That means the first stop should be the country with the second-largest GDP in the world: China. Turchin has only briefly commented on it, so here is a basic analysis for anyone who is curious like me.

China was one of the original test cases for the structural-demographic theory, as the last 600 years had two cycles that match the predictions very well. You can see the population pattern below for China, where the population grows fastest while a dynasty has consolidated power, but tends to fall or stay constant when there is instability at the end of a dynasty. Data below are only census data, but there is also a likely population fall around 1850.

Population of China with dynasties listed at the bottom. Data are from, but you could read a much longer discussion at Korotayev et al., 2006. Note that the y-axis is on a log scale so earlier trends are still visible.

For anyone not familiar with recent Chinese history, a very brief summary: after the Qing dynasty was in power for over 200 years, there started to be more civil war and civil unrest, combined with losses against foreign powers. It lost legitimacy in the eyes of the people and the army, and was replaced by the Republic of China (ROC) in 1912. However, the ROC almost constantly faced civil war from one group or another, and it finally lost to the Chinese Communist Party (CCP) in 1949. The CCP has been in power ever since.

Now back to the structural-demographic theory: Turchin originally applied the theory to pre-industrial societies where food production limits population in a Malthusian way, but he applied it to the modern United States in Ages of Discord, so I’ll try to follow those methods here.

To understand the political violence patterns in a particular country, we want to measure three things: population well-being, elite overproduction, and state stability. First, I’ll go over these to find where China is in the cycle, and then I’ll delve more into predicting political violence.

Population Well-Being

Revolutions and civil wars require lots of people, so elites need the common people to join in. Taking part in a revolution is dangerous, so happy commoners try to avoid them. Unhappy populations, however, are more willing to overthrow the government in hopes that the next one will be better.

Two measures of un/happiness are worth considering and have data available: inequality and height.

In general, people report their happiness about their income based on comparison to both their parents and others in their country, so comparative poverty is more important than absolute poverty. The Communist Party originally derived much of its support from promises to take land from the rich and give it to the poor. Inequality in China has been increasing at least since the government switch from state planning to capitalism in 1978, but recent government efforts to end poverty have helped prevent inequality from reaching levels in the United States.

Income inequality in China vs USA.

The data above show income inequality, which is a more common measure than Turchin’s choice of relative wages. The labor share of income follows a similar pattern, but with a smaller range of values. Recent data on the minimum wage show that it was increasing linearly up until the last couple years when minimum wage increases stopped rising by as much.

Inequality is a good relative measure, but we can also look at direct effects of poverty by looking at average heights. When life is good for the average person, their children tend to eat well and get taller. Nutrition around the world has increased over the last 200 years, but we can still see variations around that trend. The average citizen height in the United States hit a minimum in 1890, while Chinese citizen did not have their average height start increasing until the generation born in the 1930’s.


The chart above ends at 1980 and is the average for all men (data are better for men than women due to military records), but other sources can give us more recent data. The United States has had average height stop increasing since 1980, but Chinese citizens continue to grow taller.

Chinese heights from “Height and body-mass index trajectories of school-aged children and adolescents from 1985 to 2019 in 200 countries and territories: a pooled analysis of 2181 population-based studies with 65 million participants.” NCD Risk Factor Collaboration (NCD-RisC). Lancet 2020; 396: 1511–24.

Age at first marriage is also a predictor of well-being (marriage at younger ages in correlated with high well-being), but there is a worldwide long-term trend toward later marriage that is also changing marriage patterns. Without data going back before 1980, it is hard to tell how much of the increasing marriage age (from 22 in 1980 to 25 in 2020) is due to short-term trends. However, it is clear that the marriage age in China has risen.

In sum, life for the average Chinese citizen is still getting better even as inequality increases, so while inequality is increasing, it is likely overshadowed by how much better life is for current Chinese citizens compared to their parents.

Elite Overproduction

As inequality increases, more and more of the newly rich want to translate economic power into political power by joining the elite. While many historical societies nicely labeled their political elite (nobles, etc.), the United States does not. This causes problems for Turchin (he eventually settles on the number of lawyers as a good proxy, since most US senators were previously lawyers), but Communist Party membership nicely marks Chinese citizens as political elites. Only 6% of the population is a member of the Party, so most Chinese citizens are not members. This is a larger “elite” group than we would like, and Party members are not always richer than non-Party members, but it is hard to define a sub-group that is as easily measurable.

With the Party’s control over much of the economy, the economic benefits of party membership have become greater in recent years. President Xi Jinping has realized the danger of too many citizens joining the party for economic reasons, and has recently made it much harder to gain membership.

Fewer Applicants and Lower Acceptance Rate in Xi Era

In historical societies, efforts to reduce the number of elites have resulted in increased competition to gain those positions, which eventually creates a class of failed elite aspirants who end up starting revolutions and civil wars. However, Xi has matched the increase in selectivity for party membership with a crackdown on corruption, which makes it less profitable to be a party member. This may explain the reduced number of applicants in the last few years. If so, wealthier Chinese citizens are not yet feeling extreme pressures to gain political power, as shown by the lower rate of applications.

State stability

The simplest way to measure state stability is via debt: a country that cannot afford to pay its army cannot maintain control of internal conflict. Even in the absence of wars (wars have been much less common in the last 50 years than for most of history), state finances can be used to create jobs to placate elites or commoners. While economic analyses often look at total debt including corporations inside China, for our analysis we are focused only on debt held by the government.

Currently, government debt is rising, but it is only at 60% of GDP, which is far less than the 125% of the United States. With much of the OECD at a debt ratio of 100% of GDP, China still has room for more debt before it becomes a problem.

China’s debt to GDP ratio from

Secular Cycle Progress

China has been growing in international power since the Chinese Communist Revolution, and it continues to grow. The lack of internal instability places it clearly in the an integrative period, but the question is whether it is still in a period of expansion or if it has moved to stagflation.

On one hand, the increase in inequality, rising marriage age, rising numbers of elites, and rising debt over the last 20 years suggests that China has moved into stagflation. However, the continued increases in height show that much is still well for the general population, so it must be early in the stagflation period. It does appear that China is moving in the same direction as the United States, but at a slower pace. Unlike in the United States where internal unity has not been a concern until recently, the Chinese Communist Party has been very focused on stability, and it seems to be working for now. Government efforts to prevent inequality and reduce the number of elites (and elite aspirants) are currently having the intended effect. (Remember, stability is not necessarily good, being a dissident in China is very dangerous, and minorities face extreme persecution.)

The population of China is expected to peak around 2030, which will introduce new challenges for the Party. While population decline is historically associated with worsening conditions for the common citizen, the Chinese population decline will have been primarily caused by the one-child policy (now two-child policy). This creates a situation that has few historical precedents but is shared by many rich countries: the economic situation of most Chinese citizens will probably continue to get better for many years, but elites will be left with fewer workers. China’s GDP per capita continues to catch up with its neighbors in Taiwan, South Korea, and Japan, but at some point in the next 20 years the catch-up growth will stop and China will be left with a lower rich-world growth rate.

Once overall economic growth slows, it will become harder for elites to keep increasing their wealth without reducing economic opportunities for the common citizens. Then the state will have to start choosing between the elites and the common people, so at least one group will start supporting the government less. That said, this is a long ways off. Let’s do some more short-term analysis to look at more detail on Chinese stability.

Political Instability

In addition to charting cycles of instability, Turchin’s actual prediction was based on calculating a Political Stress Indicator (PSI) that includes things like the proportion of the population between ages 20 and 29 as those are the people most likely to take part in political extremism. Here, I follow the same math for China.

Mass Mobilization Potential

For Turchin’s prediction of political instability, he calculates a Mass Mobilization Potential (MMP) using relative wage, urbanization rate, and the proportion of the population between the ages of 20 and 29, and multiplies them together. For relative wage, he uses the average of farm and manufacturing wages divided by GDP per capita. Higher scores mean that common people are more likely to support political change. These are shown below for China.

Urbanization rate, proportion of male population between the ages of 20 and 29, and inverse labor share of income, all scaled with 1980 as 1. Numbers past 2020 are projections.

Here we see an increase in the MMP driven primarily by the massive increases in urbanization. China has many residency restrictions that make it harder to receive full benefits when people migrate to cities, and those may be designed to prevent the concentration of people who could take part in mass movements.

Elite Mobilization Potential

To calculate the willingness of elites to take part in political change, Turchin looks at the average elite income. This is a focus on economic elites rather than political elites, which is harder because there is not clear definition of an economic elite (a millionaire? a billionaire?). Turchin ends up just assuming that the number of economic elites increases when the labor share of income is low, as that means the returns to capital are higher. I ran his calculation, with some extrapolation before 1985 and after 2015.

Scaled number of Chinese elites and their average income over time, with projection after 2020 assuming the same relative wage as 2015.

Given the way relative wage is defined here (focused on agriculture and manufacturing), the large increase in elite income in the 90’s reflects both an decrease in the labor share of income and an increase in the number of highly educated citizens working in skilled professions.

While my limited data makes this projection sensitive to starting assumptions, the trend of small increases in inter-elite competition by Chinese economic elites over the last 20 years matches anecdotal reports of increased competition between college-educated citizens in cities.

State Fiscal Distress

To calculate State Fiscal Distress (SFD), Turchin multiplies the debt ratio by the inverse of trust in government. Unfortunately, the Chinese government makes it hard for survey companies to measure trust, so finding trends in survey data would require significant expertise in the subject. Instead, trust seems to remain high, although there are reasons to question the data. I will just assume it is a constant given the lack of evidence against this.

Political Instability

Assuming all previous factors are equally important, we multiply them together.

Mass Mobilization Potential, Elite Mobilization Potential, State Fiscal Distress, and the Political Stress Indicator for China. Numbers past 2020 are projections.

At first, this appears to show that political stress is increasing greatly in China! However, we should remember that these numbers are relative, and I started at a low-stress point. The steady increase is driven primarily by increasing urbanization and debt, but both started from very low numbers. Chinese debt today is similar to the debt of the United States before the 2008 financial crisis. Also, China had a very rural population in 1980, and has now reached urbanization levels the United States had in 1950.

Below, I offset the USA’s PSI by 30 years an scale the USA’s PSI so that the minima for both countries line up.

Political Stress Indicator for China and the USA, with the USA offset by 30 years (so the data begin in 1950).

When comparing to the United States, we see that China had very low political stress in 1980, so it is still far from reaching the levels we see in the United States. What China will look like in 2050 is much less predictable. The Chinese government has decisions to make: will it keep encouraging more citizens to believe they can become economic elites, or will it raise taxes on the richest citizens to discourage them from becoming richer? The decisions made today will set the course for China’s political future many decades from now.

What have I learned from doing all of this? The biggest challenge has been that easily-accessible economic data tends to be recent data, and trying to see trends over 100 years requires significant work making sure that data can be consistently used over a long time period. Finding data between 1950 and 1980 in China is particularly difficult due to communist economic policies. Even with good data, every country has its own quirks, so predictions will always be suggestive rather than absolute.

Postscript: Bi-generational Cycles

In addition to the long (hundreds of years) secular cycles, political instability tends to have peaks approximately every other generation. In the United States, Turchin identifies those peaks as the Civil War in 1865, labor protest violence around 1920, civil rights riots in 1969, and now 2020/21. In China, these patterns are not as clear. It would be convenient to pick the 1945–49 Chinese Communist Revolution and the 1989 Tiananmen Square protests and predict the next major protest around 2035–40 (which would be unlikely to produce much change due to united elites). However, the largest previous major instability events include the Taiping Rebellion (1851–1864) and the Warlord Era (1916–1928), which do not make for a consistent pattern. Someone with access to Chinese historical data would need to check the last 100 years to see if there is a pattern for recent history.

I am not an expert on Chinese economic and demographic history, or on the inner workings of the Communist party. Please comment with any suggested reading from experts on those subjects!

Software Engineer at StreamSets / UC Berkeley MIMS grad / data viz

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store