Review: Ages of Discord

Ages of Discord is an unusual history book. If you grew up reading Asimov's Foundation series, you would have come across an interesting idea, Psychohistory, which is that with large enough numbers, you could predict the future by examining the large scale interaction of people, applying equations, and then you could affect the future by planning exactly around that.

Ages of Discord is precisely that book, except Peter Turchin doesn't seem to have read Foundation, and so chooses to call his study of history Cliodynamics instead. Regardless of what it's called, does Turchin's version work? Effectively, what Turchin does is to create a simplified model of society, dividing society into the haves and havenots (the elite, and the worker population are the terms he uses). He postulates that relative wages fall as the worker population increases, but because the have-nots have no political power, what creates social turmoil is when the relative wages fall for the elites as the number of people striving for joining the elite reduces the possibility that any given elite-educated person can actually join the elite. Because capitalism has essentially winner-take-all outcomes, you get situations where the elites fight over the scraps and that creates ages of discord.

After considering and dismissing several alternative explanations, they come to the conclusion that immigration results in a larger proportion of population who are both poor and cannot vote. This facilitates the move to the right and away from redistributive policies, which then causes income inequality to rise... My explanation of the observed association is based on the Structural-Demographic Theory: (1) labor oversupply (proxied by immigration) leads to (2) elite overproduction (proxied by wealth inequality) and heightened intraelite competition that, in turn, results in (3) elite fragmentation (proxied by political polarization).(pg. 94)

The mathematical models Turchin uses are straightforward, and with it he produces graphs and charts, including ways of measuring general population welfare indirectly that might surprise you:

average heights for native-born Americans, separately for men and women and for whites and blacks. Prior to 1970, despite some divergence among different segments of the population, the overall pattern was vigorous advance, resulting in gains of about 5cm across the board. After 1970, however, this vigorous growth regime was transformed into one of stagnation and even decline. The timing of the break point is somewhat difficult to determine, because adult height can be affected by environmental conditions at any point during the first two decades of life. I follow the practice established in Chapter 3 and plot data not by the year of birth, but by the year when individuals reached age 10, the midpoint of the growth period. Using this convention we see that the overall average (averaging over both gender and race, while weighting by the number of observations in each category) peaks during the early 1970s (Figure 11.1b). Comparing the dynamics of this index to real wages we observe that ups and downs in the average height tend to precede the ups and downs in the real wage by another 5–10 years (in addition to the shift of 10 years, resulting from plotting the heights data by the data of age = 10y). In other words, when the average population height is plotted not by the date of birth, but by the date of reaching the age of 15–20 years, there is a large degree of parallelism between the fluctuations of the real wage and average stature. A possible explanation is that the level of wages experienced by the parent generation has a most direct effect on the biological wellbeing of their children when they are going through their adolescent growth spurt. If this explanation is correct, then we expect that average stature will again decline as a result of decreasing real wages after 2003 and especially following the Great Recession. FIGURE 11.1 Changes in average height, 1925–1995. (a) Average heights of white and black men (left scale), and white and black women (right scale). Data source: (Komlos 2010). Data are plotted by year of reaching age 10. (b) A weighted mean (averaging over gender and race) plotted together with real wages for unskilled labor. Average height is plotted by the year of birth (top scale) which is shifted by 15 years with respect to the calendar year (bottom scale). (pg. 201)

 When you read the book, you can start to see why immigration is such a hot-button topic:

There are several reasons why labor supply grew faster than overall population growth, of which two appear to be most important. One big factor is immigration. In 2011 the total American work force was 153 million, of which 24.4 million workers were foreign-born (this number includes both legal and illegal immigrants). The proportion of foreign-born in the labor force is currently around 16 percent (compared with five percent 40 years ago). The second factor was the increasing number of women working. In the 1970s only 40 percent of women were in the labor force; today this proportion is close to 60 percent. If the labor participation rate of native women (so that we don’t double-count foreign-born women in the labor force) stayed at its 1970s level, today there would be 20 million fewer workers—an effect of nearly the same magnitude as that of immigration. (pg. 225)

 So what does Turchin predict with all his fancy models? There I'm afraid the book falls short. He ends the book simplying saying that what his graphs predict is that the 2020s will be an extremely turbulent decade, and saying that given what we know about the inputs to the model, we might be able to do something about it. That's pretty lame-duck dodging the question stuff. What would be actionable would be if he applied the same models to other societies to see whether the same Cliodynamics apply, so that if you wanted to you could at least consider an escape plan to a society that's not about to have civil-war levels of death and destruction.

Nonetheless, the book has interesting models, lots of interesting data, and has ideas worth considering, even if it doesn't provide anything actionable. It's dense reading, but recommended.