Using percentages in the face of uncertainty
In this post I talk about how historians could potentially improve their efforts to ensure their audience understands the level of uncertainty we have about historical events. I use examples from James C. Scott’s book Against the Grain, but these problems aren’t specific to him or this book.
There are many unknowns in history, from the exact dates and details of events to the major causes of historical trends. Filling in gaps in our understanding with educated guesswork is the job of a historian. But I think these theories would be even more useful if historians were very clear about the level of uncertainty in their theories, preferably by putting percentages on their estimates for how likely they think their theories are to be true.
For example, Scott offers several explanations for why there are so many abandoned settlements in Mesopotamia dating to ~2000-700 BCE: disease, environmental degradation, bad harvests leading to abandonment by the working class, war. All these answers are compelling, but he doesn’t offer any way to think about which is most important. Do we think disease explains 20% of the abandoned settlements? Or 80%? How much more important is abandonment by working class vs. war? The explanations are all plausible, but “plausibility” is not “100% correct,” and the problem is that I don’t know if in this case “plausible” means 90% or 10% likely to be correct.
Here’s another example: We aren’t certain why people started planting crops in the first place (around ~10,000 BCE). Scott considers two commonly-given explanations (people wanted to be able to store their food longer, and people wanted to be able to do work now and reap the benefits later) and says “neither … are remotely plausible” due to the large amount of work agriculture takes. But he argues people might have taken up agriculture because it was extremely easy in floodplains — this is the type of work that a hunter-gatherer “might take up.”
Should we take this to mean that the floodplain explanation is 100% of the reason that people start agriculture? Probably not. But we aren’t sure if 1) other explanations aren’t given because Scott thinks that floodplain agriculture is ~90% likely to be the reason that people started agriculture and so the other options aren’t worth mentioning since they only make up 10%, or 2) floodplain agriculture is actually only 30% likely to explain the adoption of agriculture, but the others aren’t relevant to the discussion at this point in the book.
I’m fine with either explanation 1) or 2), but it’s not knowing which is the case that bothers me. And I don’t mean this as a critique of James C. Scott in particular — Against the Grain is a very good book. But I think historians (definitely popular historians) often do their audience a disservice by not providing a framework to understand how much confidence we should put in their theories.
This applies to things outside of history too (e.g., see the Good Judgement Project, and apparently the economist Larry Summers required that people give numbers to their uncertainty when in the White House), and you can catch me starting to put percentages on my uncertainties going forward.