An idle Sunday afternoon thought: In historical work, we frequently come upon things that almost happened. That is to say, of the set of all things that did not happen, some seem much more plausible than others. We see traces in the archives of an idea that was never publicly stated, a speech that was written but not delivered, a movement that almost got off the ground but couldn’t find quite enough funding, or lost a key organizer. And so on. On some level, I think we privilege these “almost happened” events over those that are much more implausible. Why?
I think the answer has something in common with “regression discontinuity designs.” I’m no expert on RDDs, but the basic idea is quite simple: if there is some meaningful cutoff between two groups, then it makes sense to compare cases that are quite close to the cutoff to try to estimate the effect of being selected into the group. For example, if students are selected for special education based on their scores on a standardized IQ test, we can compare students just below and just above the threshold to see if being placed in the classes has a significant change on the outcome (on the assumption that there is very little different between students with, say, an IQ of 79 and 81 beyond their placement in the class).
Historical counterfactuals based on events that “almost happened” seem somewhat analogous. We can’t actually get data on the outcome of interest – in historical analysis, we almost never have many cases that differ only along one variable – but we can at least reasonably speculate (“simulate”?), how would this small change have affected a later event? By restricting ourselves to events that almost happened, we try to get at the importance of an event somewhat separated from all the other contingent factors surrounding it. In other words, if we can say “if this small and random event had occurred, X and not Y would have occurred at time 1, and that small change would plausibly have had a big consequence at time 2” then we are making a case for the importance of event Y.
A bit less abstractly, I’ll give an example from my own work. I study the history of national income statistics (GNP, GDP, etc.). In the early period of official national income statistics (the 1920s-1940s), lots of authors debated precisely what to include in the statistics and how to include different goods, services, intangibles, etc. Eventually, most of these debates were settled into the modern standards (the UN System of National Accounts being a key document), and some debates were re-opened in the 1970s. Two of the most criticized omissions were unpaid housework (cooking, cleaning, childcare, etc. primarily performed by women) and the environment (the value of clean air, water, and so on).
What’s interesting from this perspective is that women’s work was “almost” included in the accounts in the 1920s-1930s, while the environment was nowhere to be found. In prominent texts (a 1921 NBER report, the first official Department of Commerce statistics in 1934, etc.), women’s work was discussed, and estimates were even produced (but not included in the final totals). No such estimates existed for the environment. And so, in some sense, it becomes much more plausible to ask, what would have happened if early national income statisticians had included women’s work? How important was this omission? How much would development economics have changed (say)? Or, how much would this inclusion have delegitimized the statistics? Asking those questions of the environment requires a much bigger leap – we have to change more things at the beginning of the story, because no one in 1921 or 1934 was pushing for their inclusion, and no estimates were made that could simply have been added to the final totals.
In some sense, this whole post is getting at similar issues to those discussed in Emigh’s methodological work on “Negative Case Methodology” – trying to answer questions of the form, “Why didn’t X occur?” E.g. why is there no large, visible social movement around finance in the United States?* For Emigh (if I’m remembering correctly), studying the “almost happened” gives us leverage on understanding why an outcome did not obtain. We study the small movements that did form but failed to get traction in order to understand the forces that prevented a larger movement from emerging. My discussion above flips Emigh: instead of focusing on the “almost happened” to figure out why it was “almost,” I want to ask, “What difference would it have made it if had happened?”
Enough of that. Back to the books.
* Greta Krippner’s new, and not yet published work, addresses this question.