READING HISTORY BACKWARD

The most egregious example of reading history backward in recent research is arguably the many large-N studies that analyze the period after 1945 but project their findings backward in time in problematic ways, thereby introducing selection bias. For instance, Rasmussen and Pontusson (Reference Rasmussen and Pontusson2017, 28) note how most quantitative studies of the “Ghent effect”—that is, the notion that a Ghent system of unemployment insurance (as opposed to a state-administered system) increases union density—read “history backwards.” The Ghent effect is based on findings from studies that analyze the post-1960 period, which show that countries with Ghent systems have higher union density. This, in turn, has been used to infer that Left parties were behind creating Ghent systems before World War II. However, enlisting new data that go back to 1900, Rasmussen and Pontusson (Reference Rasmussen and Pontusson2017) show that Ghent systems had no effect on unionization rates and that the Ghent effect instead is a result of increasing state subsidization in the decades after 1945. Furthermore, they argue that it was Social Liberal parties that favored Ghent systems, whereas genuine Left parties (i.e., Social Democrats) normally opted for state-administered unemployment insurance before World War II (see the online appendix for two additional examples).

Rasmussen and Pontusson’s (Reference Rasmussen and Pontusson2017) criticism of earlier research on the Ghent effect illustrates how scholars who read history backward become predisposed to overlooking confounding factors, thereby making their explanations too deterministic (Capoccia and Ziblatt Reference Capoccia and Ziblatt2010, 939; Pierson Reference Pierson2000). More particularly, retrospective inquiry risks imputing the causes that explain the advent of institutions from those that explain their later consolidation (Boucoyannis Reference Boucoyannis2015).

We can conceive of this as the explanatory repercussions of reading history backward (figure 1). It often is held that these problems are especially pronounced when political economists do history because they frequently subscribe to either a cyclical view of history (i.e., an emphasis on equilibria) or a teleological view (in which certain institutions create a path of unfettered progress). The first perspective tends to “freeze” history, whereas the second perspective makes history “efficient” (March and Olsen Reference March and Olsen1983). Freezing history leads scholars to ignore bounded change in which institutions are altered in a piecemeal fashion via the interaction between changing and fixed institutional aspects. A good example is what Mahoney and Thelen (Reference Mahoney, Thelen, Mahoney and Thelen2010, 16) refer to as “conversion” or “the changed enactment of existing rules due to their strategic redeployment.” Making history efficient leads scholars to ignore the behavioral realities of often-contingent microprocesses (March and Olsen Reference March and Olsen1983, 737). These notions of history easily rule out counterfactual thinking and instead tend to infuse the analysis with the creeping determinism that also is associated with hindsight bias (Fischhoff Reference Fischhoff1975).

Figure 1 Explanatory Repercussions of Reading History Backward

This shapes the theoretical lens by privileging functional or structural explanations at the expense of more contingent, actor-centered explanations. The previous example, centered on the Ghent effect, is one illustration. The online appendix shows how a larger body of research on the rise and development of medieval representative institutions or parliaments—including previous work by this author—has viewed this outcome as resting on an equilibrium characterized by a relatively even balance of power between monarch and social groups. I demonstrate that by reading history backward, this research has ignored that these institutions were originally introduced by first ecclesiastical and then lay rulers in a much more contingent way, only later to develop into institutions that nobles, clergy, and townspeople could use to constrain rulers (see also Boucoyannis Reference Boucoyannis2015).

However, these explanatory repercussions are not the only problematic aspects of reading history backward. Indeed, retrospective inquiry tends to affect the formulation of the problem under study, the conceptual understanding of the explanandum, and the historical sources that can be enlisted. We can think of this as the conceptual repercussions of reading history backward and, because they have not received nearly as much attention as the explanatory repercussions, they need more elaboration (figure 2).

Figure 2 Conceptual Repercussions of Retrospective Analysis

The first conceptual problem concerns the way that scholars who read history backward pose their research questions. In a genuinely historical analysis, a research question is left open in the sense that it leaves room for the unexpected. In a Popperian vein of thinking, it allows for alternative explanations to be vindicated. Retrospective questions are normally quite narrow—we might even say closed—because they take certain things for granted. This increases the risk that alternative outcomes or causal antecedents are ignored. A common warning signal is when research questions mimic or echo one another across a larger research agenda (Fischer Reference Fischer1970, 24–28).

The second conceptual problem concerns the definition of the phenomenon to be explained. Scholars who delve into history to explain later institutions are apt to project the definition that meaningfully characterizes later institutions back in time, often without recognizing its poor fit with the institutions from which the subsequent ones sprang. We can think of this in terms of a reification of concepts or as conceptual stretching (Sartori Reference Sartori1970). It is related to the explanatory problem of freezing history because it blinds researchers to the possibility that institutions evolved significantly over time—indeed, that institutions sometimes are fundamentally transformed (Mahoney and Thelen Reference Mahoney, Thelen, Mahoney and Thelen2010).

The online appendix illustrates these two problems by showing how many of the social scientists who have recently revisited the origins and development of medieval parliaments have taken the fact that these institutions constrained monarchs as the point of departure. Scholars therefore define representative institutions as “institutions of constraints,” and they ask which factors explain that rulers became constrained by these institutions. This has had two unfortunate consequences. First, it means that the definitions have a poor fit with the early phase of representative institutions, which—as historians emphasize—were institutions called at the whim of the ruler to exercise power rather than institutions that constrained rulers. Second, it has biased the answer in favor of structural explanations centered on the societal balance of power, thereby ignoring the actors’ choices that first made popes and then monarchs summon representative institutions. In other words, this research agenda has frozen history by assuming that the constraining effects of parliament can be projected farther into the past than is warranted by the historical material, thereby ignoring that institutions can change incrementally.

The third conceptual problem concerns the historical data that can be enlisted in the analysis. Whereas the previous work of historians is normally the point of departure when social scientists code large-N datasets and carry out in-depth historical analysis (Møller and Skaaning Reference Møller and Skaaning2018), the conceptual lens decides what type of previous historical data is relevant. By reifying concepts and freezing history, the inquiry allows large-N research designs based on conditional independence and uniformity of variables (Sewell Reference Sewell and McDonald1996). These research designs foreclose the use of most of the qualitative data produced by historians because they are based on enlisting a set of variables coded in a similar—hence, relatively general—way across space and time. Scholars risk committing what has been termed the “quantitative fallacy”—that is, the idea that “facts are important in proportion to their susceptibility to quantification” (Fischer Reference Fischer1970, 90). This slights valuable but often nonquantifiable information about what motivated actors to create institutions and how they actually did so. As shown in the online appendix, this is illustrated by the way the new research on medieval parliaments relies overwhelmingly on quantitative data coded across long periods (i.e., 50- or 100-year intervals) and across relatively aggregate empirical units. Conversely, the abundant qualitative information historians have produced about medieval parliaments has been used only sparingly in these empirical analyses.

READING HISTORY FORWARD

For political scientists, the best remedy against these explanatory and conceptual problems—and therefore against reading history backward—is simple but time-consuming. It consists in an open-minded reading of what historians have written on a particular subject. Open-minded means that we ignore our theoretical priors in general and specific hypotheses in particular and peruse the historical literature—not to find evidence for a preconceived explanation but instead to reveal new aspects of a certain empirical development. This is very much a question of forcing a certain cognitive mindset on ourselves. Berlin’s (Reference Berlin1953) famous distinction between foxes and hedgehogs—or pluralism and monism—comes to mind.

In other words, what is needed is an exploratory review of the historical literature on the subject of interest in which the work of historians is treated not only as a source of facts but also as a source of explanations. This entails a proper reading of this literature, which is why it is time-consuming. This type of reading can be described with a negative: it is the opposite of what Hexter (Reference Hexter and Hexter1979, 241) calls “source-mining”—that is, “the examination of a corpus of writing solely with a view to discovering what it says on a particular matter narrowly defined—going through the indexes and leafing through the pages.” Thus, it requires that we thoroughly immerse ourselves in the historical literature and learn about the agreements and disagreements among historians working on a particular subject, as well as the historical context in which the developments in question occurred (Lustick Reference Lustick1996).

Historians, of course, have their own theoretical and methodological priors; indeed, all history is told from some vantage point, which can be either implicit or explicit, and all historical facts have in some sense been preselected (Lustick Reference Lustick1996; Møller Reference Møller2020; Møller and Skaaning Reference Møller and Skaaning2018). For instance, depending on whether historians work within a Marxist, an Annales, or a third tradition, they are likely to use different concepts and therefore select different evidence—and, in all likelihood, to reach different conclusions (Hexter Reference Hexter and Hexter1979, 61–145). This can be exemplified by contrasting an older historical literature—drawing on Marxism—that analyzed the religious sects during the English Civil War through the analytical lens of proletarian class hostility with more recent work, which jettisons the Marxist perspective and, accordingly, finds much more diversity in the views of the “radical” groups that opposed the royalists in the 1640s and 1650s (Worden Reference Worden2019, 19). The online appendix provides another example of how historians sometimes read history backward, centered on the Sonderweg debate on German history.

However, as Kreuzer (Reference Kreuzerforthcoming, 10) states:

In a word, work by historians is normally more open and forward looking than the historical analysis done by social scientists. Gaddis (Reference Gaddis2002, 56) presents this as the difference between the reductionist models of social scientists, which are needed “to generalize about the past in such a way as to be able to forecast the future,” and the historians’ ecological approach, which “gleefully proliferates variables all the time” (Gaddis Reference Gaddis2002, 58; see also Møller Reference Møller2020, 254–55). This gleeful acceptance of the complexity of the past can be frustrating for political scientists who want to understand average effects of particular variables and who have been trained to avoid overdetermined and/or eclectic explanatory accounts. However, it allows historians to keep an open mind about the possible causes of certain phenomena and the contingency of historical developments, and it is the reason that immersing ourselves in their works is such a good check on retrospection.

Using a metaphor, the previous work of historians comprises a quarry that political scientists can mine to find the hidden gems that their theories and models disguise. After this exploratory phase, they then can use the standard tools of causal analysis to apply the theoretical arguments they have borrowed from historians against new evidence (Coppedge Reference Coppedge2012, ch. 5). In this phase, they also will need to systematically decide which historical accounts are reliable as sources of data given the purpose at hand—and systematically factor in the range of conflicting arguments that often characterize historiography (Lustick Reference Lustick1996; Møller and Skaaning Reference Møller and Skaaning2018).

Thus, the idea is not to substitute the tendency to read history backward with a “barefoot empiricism” that sees historians as presenting consistent, noncontradictory, theoretically coherent, and uncontaminated accounts based on a representative body of data. No serious historian would accept this caricature as a valid view of what they do. However, precisely because political scientists are more constrained by their theories and their more particular expectations, they stand to gain by broadening their perspective through an initial open-minded reading of the previous work of historians.

The online appendix shows that a more forward-looking analysis of the development of representative institutions is possible because medieval historians have already done the difficult work of providing the necessary data; political scientists simply need to process it. This open or forward-looking approach allows scholars to be present at creation, for instance, by asking why first popes and their agents (i.e., papal legates, nuncios, or archbishops) started using proctorial representation, based on Roman Law, at assemblies around 1200 AD and why kings then quickly followed suit (Møller Reference Møller2018). This open approach also enables scholars to correct mistakes that, through an older body of historical work, have made their way into social science, henceforth to be repeated without scholars actually investigating their basis. A good example is the often-repeated observation that the first genuine representative institution was summoned by King Alfonso IX in the realm of Leon in 1188 (Boucoyannis Reference Boucoyannis2015, 315, fn. 86; van Zanden, Buringh, and Bosker Reference van Zanden, Buringh and Bosker2012, 838).Footnote ³ What we know is that Leonese townspeople attended the 1188 assembly (as they would again in 1202 and 1208) and that they in some way had been “chosen from or elected by every city” (et cum electis civibus ex singulis civitatibus) (Reynolds Reference Reynolds and Reynolds2012 [2000], 108). However, it does not follow that they attended as genuine representatives of their town council (concejos). There is, in fact, no evidence that townspeople arrived as proctorial representatives based on Roman Law (Reynolds Reference Reynolds and Reynolds2012 [2000], 108), the defining attribute of medieval representative institutions (see the online appendix). Furthermore, this is entirely implausible given that medieval historians now agree that the invention of proctorial representation at assemblies dates to Innocent III’s and Honorius III’s pontificates in the period 1198–1227 (Møller Reference Møller2018).