The new Zhang et al. paper on links between climate and political crisis in pre-industrial Europe is fascinating (see ScienceNow story here). I was conflicted about whether to cover it; the sweeping conclusions and reliance on statistical signal-searching made me nervous, but it obviously deserved attention.
In the end I just didn’t have time to do the research justice, and went with a more tractable PNAS paper on recent human evolution instead. But historian/policy expert/civil unrest wonk Jack Goldstone did send a superb analysis, presented here in their entirety:
It is well established that waves of socio-economic factors — long changes in trend in population, prices, real wages — occurred in Europe from 1000 to 1900 AD, and that such waves also seem related to the incidence of wars and revolutions. In particular, the periods 1320-1420, 1620-1660, 1760-1810, and 1848-1860 were notable for high levels of national and international conflicts, and each came at the end of a period of sustained increases in population and prices and decline in real wages.
People have long guessed that climate changes were an even more fundamental factor driving these waves, and the distinguished historian Geoffrey Parker is writing a book making that claim.
The difficulty in establishing that guess always lay in getting precise dates for climate change that could be compared with changes in the other, more reliable series of population, prices, and wages, for which we have widespread annual observations. Climate data from tree rings and temperature series were too varied and too vague to identify clear turning points.
The authors of this paper seem to have overcome that by gathering and averaging enough data to identify precise turning points, focusing on a “cold phase of 1560-1660”. But I have to ask — why was the cold phase exactly 100 years? Did that come out of the data, or was it selected?
If you look at their data on European Temperature, temps start falling in the early 1500s; the decline in temp accelerates from 1560 onwards, but then begins a slow upturn around 1600, returning to the 1560 levels by 1660. So if we examine a period of declining temperatures we could identify the period 1500—1560 as the decline stage, and 1600-1660 as the recovery stage — and have to ask why conflict was much higher in the recovery stage than in the decline stage. The authors eliminate that question by focusing on the “cold phase” of temperatures from the 1560 accelerated decline all the way until the 1560 level is regained in 1660. But why do that? Why not focus on the the true “Cold Phase” when temperatures remained quite low from 1580 to 1640? IN short, there is some arbitrariness to selecting a climate “phase.”
The authors make a very good effort to overcome this with statistical tests of lag relations between climate and other causal variables that lead to conflict. They try to endogenize everything (e.g. make everything depend on climate). But that doesn’t quite work.
In particular, they have a problem with their first moderate period, from 1500 to 1560. If the change in temperature with the onset of the cold period marked a historic break, then the lines for agricultural production and prices should have shown some kind of break too, if with a lag. But that’s missing. IN fact, the lines for prices, output, and population show a steady trend from 1500 to 1640 (red and black lines in graph C, red line in Graph H). These curves all move faster toward the end of the cold period, but there is no change in trend at all with the onset of the cold period; the trend change only occurs near or at the end of the cold period.
The authors explain this with a story that continued population growth into the cold period caused pressure on prices which led to epidemics and then wars which in turn pushed population downwards and caused a reversal of price and output pressures.
But I don’t buy it because prices were already rising well before the cold period. And these series can also be explained by starting with population and tracing the consequences to population changes. Indeed — look closely at the last graph, graph H, and the black line (population growth). It shows a sharp increase in the rate of population growth in the period from 1580 to 1620, well into the Cold Period. Why was this? I can’t see any reason, nor the authors giving any, for the jump in growth rates 20 years after the Cold Phase hits.
IT seems to be an exogenous jump in population. But that jump is what drives all the other peaks, as they track those decadal movements more than the broad movements in temperature.
So I would say the authors have compiled a lot of good evidence that seems to implicate climate. But they are dealing with only two broad waves, and a lot of coincidence and a few decisions about where to place cut points in the series can drive apparent correlations.
I would ask them to test their theory — which is now simply designed to prove climate is the causal factor — by comparing their results against models that use other variables, such as population, or prices, as their prime causal factor. Only if the climate model gives superior results would I be willing to say that they have made their case. And the way I look at their date it seems the population growth factor moved mostly on its own, and had a more decisive impact on the other variables over this period.”
Yaneer Bar-Yam at the New England Complex Systems Institute, whose own recent work has touched on recent climate/unrest links, also commented:
They show that temperature variations between 1500 and 1800 in Europe are related to changes in agricultural production, food supply, food prices, disease, wars, migrations, population size. All of this makes sense because when food supply is affected, the entire society is affected and it is shifted toward violence due to desperation. It is not Malthus’ concept of exploding populations, it is the supply reduction version of the problem: when food availability goes down societies enter crises. This is the historical version of the study we did on the effect of current food prices on riots and revolutions. Their study shows for a hundred years around 1600 food scarcity was driven by weather.