“All the nightmare images of ethnic conflict in the twentieth century are here”: Erroneous statistical proofs and the search for ethnic violence in revolutionary Ireland, 1917–1923

In the 1990s, Canadian historian Peter Hart claimed tens of thousands of native southern Irish Protestants experienced something akin to “ethnic cleansing” at the hands of the IRA in the early 1920s. Hart's research revised the “Irish nationalist revolution” (ca. 1917–1923) as an essentially ethnic conflict, and this article re-examines his evidence and methodology. Exaggerating the number of forced migrations, Hart's analysis rests on erroneous statistical proofs which he supported with a gross evidence selection bias. To better understand Hart's revision, his work is compared with Michael A. Bellesiles' Arming America: The Origins of a National Gun Culture (2000), which also exhibited similar but also very different statistical errors. A central argument in this article is that erroneous statistical proofs are best understood as social constructs, where they articulate the prejudices of their host academies. Greater awareness of this problem is needed if it is to be avoided in the future.