How Machine Learning Will Change Cliometrics

Machine learning (ML), it is sometimes claimed, will change the world. In this chapter, we argue that ML does offer great potential for advancing the field of cliometrics. We first attempt to demystify ML by describing its most widely used methods and showing how these are often natural extensions of traditional approaches reshaped to take advantage of increases in computer power. We then proceed to discuss applications of ML in existing cliometric research. These fall into three categories. First, ML provides a tool for improving solutions to preexisting empirical problems, especially in causal identification. Second, ML can produce new representations of data that facilitate a fresh way of looking at the world with conventional cliometric techniques. Third, those new representations naturally lead to a quantitative approach to the inductive generation of new facts and theories. We suggest that this third category offers the most likely route by which ML will change the way historical research is done. To buttress this conclusion, we reflect on our own experience when employing ML methods in the study of English legal and cultural history. We discuss how ML allowed us to reimagine the way in which we conduct our research. We conjecture that ML will not only expand both the scope and breadth of cliometric research but also realign its orientation with what has always been one of the goals of historical research – to describe the ebb and flow of history.