The Persistence of Historical Persistence
Today I’m going to talk about an annual review that Tom Pepinsky and I are doing, on historical persistence! We have a draft up on SSRN (comments welcome!), but I’ll discuss some of the highlights here — Broadstreet readers might be particularly interested in the recent publication statistics for historical persistence papers.
Tom and I were tasked to review recent quantitative literature on historical persistence, which we define as “causal effects that (1) operate over time scales of a decade or more, and (2) explain spatial variation in political, economic, or social outcomes.” That decade marker is fairly arbitrary, but the point here is that we are focusing on studies that look at relationships between variables over longer periods of time. Our scope for this review is also on causal inference and quantitative research designs (though note there are many other approaches that can be successfully used to study the past).
We join a growing list of great reviews on this topic — including but not limited to pieces by Nunn and Cantoni and Yuchtman, or Caicedo’s chapter out in the Handbook of Historical Economics on ‘historical econometrics.’ There’s also a great ARPS piece by Simpser, Slater, and Wittenberg on historical legacies, that I recently discovered.
Notably, we collected publication data for the last decade on historical persistence papers. We surveyed the top journals in political science and economics (see paper draft for a full list), between 2000 and 2021, finding the articles that looked at causal relationships that operated for over a decade. (We have more scope conditions, but importantly we excluded panel data that didn’t estimate long run relationships, and also excluded studies on local effects of structural historical features.)
What did we find? Figure 1 has the time trend. The number of historical persistence papers in political science is rising, after the spike in economics. (Note the short term fluctuations could be explained by a number of factors, that we don’t try to explain).

We also look at the temporal scope of these studies — in our data, 50% of causal arguments operate at decade-long time scales, and 47% at century-long timescales. We even have supposed causal effects over millennia, though this is not as common.
Figure 2 shows geographic coverage; Europe and Germany take the prize, though a substantial proportion of persistence articles are also global. But the main takeaway from this figure is there are many opportunities to contribute to this literature!

In the piece we also talk about methods, and methodological challenges (including the tragic tradeoff between omitted variable bias and post-treatment bias). We’re still revising this section (thanks for the useful comments, David!), but there’s an especially interesting discussion of spatial analyses as relates to historical persistence. This includes the fact that spatial autocorrelation could be biasing estimates, but also the fact that many papers do, in fact, correct for this.
That’s a snapshot for now, while we keep working (and take a much needed summer break). Stay tuned!