Another Clever Proxy for Quantitative History

Peter Turchin


Join 36.9K other subscribers

When I became interested in what eventually became Cliodynamics, initially I thought that I would just play with some mathematical models of historical processes, because I had read many times before that there is very little quantitative data in history. In fact, the opposite is true. It turned out that history has massive amounts of data with which we can test our theories, and this empirical corpus continues to grow. The main source of new quantitative data is the clever use of “proxies” — indirect quantitative indicators of various processes of interest in historical dynamics. I see articles publishing such new data almost weekly.

The latest one is Linking European building activity with plague history by a team of dendrochronologists based in many labs across Central and Northern Europe, led by Fredrik Ljunqvist. Dendrochronology is the method of dating tree rings, and it allows us to pinpoint the date when a tree was felled down to a year. Ljungqvist’s group went around Europe taking cores from wooden beams in old houses and then dated them. So far they have collected nearly 50,000 such dates for building construction.

The inscription on this house in Lübeck claims it was built in 1535. I wonder, would a dendrochronological date agree? (Photo by the author)

Because new houses are usually built to accommodate extra population, the distribution of dated building years provides us with a great quantitative proxy for population increases (see panel A in the chart below):

From Figure 1 of Ljungkvist et al. 2018

One thing to keep in mind is that because old houses are constantly destroyed, the probability that a 13th century house would survive to the present is much smaller than such probability for a 17th century house. Thus, the felling dates data should be detrended. If one does so, then two great oscillations become even more apparent, with slowdowns in building new houses reflecting the two periods of general crisis in Europe: the Late Medieval crisis of the 14th and 15th century, and the General Crisis of the Seventeenth Century. You can read more about these cyclic crises in our book Secular Cycles.

The second chart (panel B) in the figure shows the distribution of plague outbreaks, recorded in historical sources, from a well-known compilation based on the pioneering work by the French demographer Jean Nöel Biraben in the 1970s. As the article title indicates, the main factor, potentially explaining building slowdowns, which Ljunqvist et al looked at, is mortality resulting from epidemics. Here I must temper my praise for the article with some criticism. While the data themselves are wonderful, I am not sure that relating them to plague outbreaks is the most interesting thing one could do with them.

An old house in Toulouse (photo by the author)

As the authors themselves acknowledge, European population, and thus building activity, started to decline well before the Black Death of 1347-1352. The causes of this decline are actually well known, and are due to worsening structural-demographic conditions in Western Europe. Furthermore, and even more interestingly, population in most countries did not start growing immediately after the cessation of plague outbreaks. In our book Secular Cycles we discuss the possible reasons for late medieval England and France, and come to the conclusion that the factor that held back population growth was incessant socio-political instability (the Hundred Years War in France, and the Wars of the Roses in England).

Ljungqvist and co-authors propose a similar explanation for the Seventeenth Century population decline in Germany (the Thirty Years War), but they treat it as a singular event. In fact, the Thirty Years War was part of the General Crisis of the Seventeenth Century that affected most of Eurasia, from England to Japan.

Some years ago Walter Scheidel and I proposed that we can use the frequency of coin hoards as a quantitative proxy for internal warfare (see our article, Turchin P, Scheidel W. 2009. Coin Hoards Speak of Population Declines in Ancient Rome. PNAS 106: 17276-17279.). One could use a similar approach in the analysis of the data collected by Ljungqvist et al. As an example, here’s what coin hoard data for Bohemia (modern Czech Republic) look like:

You can clearly see civil wars associated with the Late Medieval and the Seventeenth Century crises (the Hussite Wars and the Thirty Years War). Unfortunately, Czech building dates are not yet publicly available. It will be very interesting to see whether coin hoard data provide a better predictor for the cessation of building activity, compared to the plague data.

Fredrik Ljungqvist has graciously shared with me data on another region, Central and Northern Europe (data were collected by Hans Hubert Leuschner of University of Goettingen). The coin hoard data I have for this region has fairly crude resolution (50-year intervals), but here’s what happens if we plot these two proxies together:

The blue curve is detrended building activity, and the red bars are coin hoard per half-century. It would be better to resolve coin hoards at shorter time intervals (which is something the Seshat project will soon do), but you get the rough idea—peaks of coin hoards are associated with troughs in building activity.

More generally, there are many more variables whose effects could be investigated. We provide a lot of such proxies in Secular Cycles. For example, we show the dynamics of temple or church building activity for several secular cycles (these buildings can be well dated from historical records). It would be interesting to see whether building activity by the elites and the state (who are primarily responsible for temples and churches) is paralleled by building activity driven by more humble classes.

In any case, despite my critique, it is a very welcome development that we have a new proxy for historical dynamics. We may be at the point where traditional sources, such as historical chronicles, are being mined out, but I expect that quantitative historical data will continue to flow in, thanks to new clever ways of looking at history quantitatively.

An old house in Aarhus, Denmark

Note added 17.XII.2018. Fredrik Ljungqvist clarifies: The description that ‘Ljungqvist’s group went around Europe taking cores from wooden beams in old houses and then dated them’ does not feel fully accurate as we compiled data already available from decades of archaeological dendro-dating work in Central Europe and reused it for our study. (All the data contributors were included as co-authors.) We did not do any new dating work for this study.”


Notify of
Most Voted
Newest Oldest
Inline Feedbacks
View all comments
Clinton Reilly

Fascinating. When developing historical military war-gaming models I have found that conventional historians often say that there is limited quantitative data available. However there is a huge amount if you look. I wonder if conventional historians lack quantitative training and so do not ‘see’ the numbers.

Chris Morris

Very interesting. The relationship between coin hoards and house build may be complicated by the fact that there are two sorts of house build – self-build and build by hired expert. The former doesn’t really require coins in the economy at all (and no coins = no hoards) whereas the latter requires considerable monetarisation to work. High class houses, such as in the photos, are the ones most likely to survive and they are almost certainly not self-build. I suppose what I’m driving at is that the strength of the relationship between hoarding and housebuilding will vary with the nature of the economy (and probably urbanisation) as well as population and war

al loomis

we are going to see a new coin-hoarding period soon, if meteorology is an effective proxy.

Loren Petrich

Reminds me of our host’s proxies for United States patriotism in “Ages of Discord”. That seems like something difficult to quantify, but he did it.

The first is which people states’ counties are named after. In the colonial era, it was often British notables, in the early 19th cy., it was often national notables like George Washington and Thomas Jefferson and Benjamin Franklin, while in the late 19th cy., such notables became less common name sources. This source of data petered out early in the 20th cy., however.

The second is counts of visits to nationalistically significant sites like George Washington’s Mt. Vernon estate and the Statue of Liberty.


What conclusions can you draw from that?

Loren Petrich

Patriotism has risen and fallen along other social indicators, and our host Peter Turchin has used several in “Ages of Discord”. He concludes that the US was in an integrative phase in the early nineteenth century, and high patriotism is consistent with the other indicators. However, the US was in a disintegrative phase in the late nineteenth century, and low patriotism is consistent with that. The US was in an integrative phase again in the mid twentieth century, and visits to national monuments were high back then. However, those visits have declined since then, in agreement with social indicators pointing toward a disintegrative phase.

Loren Petrich

Another aspect of integrative vs. disintegrative phases is territory change — integrative phases sometimes have acquisitions and disintegrative phases losses. That pattern partially holds true for the US also. When it became independent, the US had only the Appalachians, territory east of those mountains, and territory near the Great Lakes. In 1803, the US acquired the Louisiana Purchase from France. In 1818, some land in the northern Midwest from Britain, also ceding a bit of land there. In 1819, Florida and some other territory from Spain. In 1845, the Republic of Texas. In 1846, the Oregon Territory. In 1848, California and nearby, conquered from Mexico. In 1853, the Gadsden Purchase, a little bit of land from Mexico.

So the US acquired all of its contiguous real estate around the history-cycle peak in the 1820’s, though much of it did not get turned into full-scale states until some decades later.

The dates are rather discontinuous, with a big gap that was roughly 1820 – 1845. In this time, there was a revolt in Canada, and the US did not use it as a pretext for “liberating” Canada. Also in that time, Texas became independent from Mexico. Historian Peter Klingberg has proposed that the US has foreign-policy cycles, alternating between extroversion, willingness to acquire territory and go on military adventures abroad, and introversion, aversion to those activities, each phase lasting roughly one generation. So in the early 19th cy., the US alternated extroversion – introversion – extroversion. Each phase is self-limiting, I think. Introverted phases end from challenges abroad, and extroverted phases end from burnout from big wars.

Roger Cooper

This cartoon on XKCD has some relevance,

J. Daniel

The proxies noted here and many of the ones in Ages of Discord are really clever. Anyone know what Yale Law School tuition has been doing in the last few years?

Studies of advancement in technological performance over time have traditionally relied on quantitative data (e.g. Moore’s law, Wright’s law, Carlson curves, etc.) if only because there is no other way to do it.


Question is what Yale Law (or undergrad) tuition proxies as many do not pay the full rate (roughly half do at the Ivies/equivalents; more than half for privates that are kinda Ivy-backups for kids who’s families can pay but who couldn’t actually get in to an Ivy, and less than half pretty much everywhere else).

J. Daniel

Undergraduate tuition is consistent with your point. I’m not so sure about law school tuition. What may vary over time and thus detract in some measure from Yale Law School tuition as a proxy is the availability of tuition loans.


The super-elite law schools (probably only YLS, HLS, and SLS by now) offer fin aid (59% of YLS’s student body receives fin aid grants: while everybody else offers merit scholarships.

Clinton Reilly

This discussion about proxies is very informative! There is a concept called “Tactical Posture” used in military simulations. It is intended to be a measure of whether a particular commander prefers an aggressive posture (i.e. moves towards the enemy and attacks) or a defensive posture (i.e. remains stationary in relation to the enemy) or a retiring posture (i.e. retreats from the enemy). It can be assigned as a subjective assessment and can be obvious in many cases. For example, in the American Civil War “Buffalo Hood” repeatedly attacks and is chosen for attacking roles as was the case at Gettysburg. However perhaps the degree of aggression can be quantified to make the assignment of the posture more objective. Just a thought. Any comments?

  1. Home
  2. /
  3. Cliodynamica
  4. /
  5. Regular Posts
  6. /
  7. Another Clever Proxy for...

© Peter Turchin 2023 All rights reserved

Privacy Policy