Friday, May 31, 2024

Assessing causes in the past (Kreuzer)

Quantitative social scientists have something of a catechism when it comes to providing evidence for causal assertions. If we want to assert that A is a contributing cause to B (for example, living in a neighborhood with many sub-standard housing units is a cause of higher rates of delinquency), we need to conduct a study involving a reasonably large number of cases and then assess whether cases with high-A values are also found to have high B-values. And in order to avoid well-known problems of spurious correlation, we are advised (when possible) to attempt to arrange some kind of experiment -- a field experiment, a natural experiment, or a controlled experiment -- in which the value of A is changed and we observe whether the value of B changes as well. And some quantitative social scientists urge the importance of identifying a possible causal mechanism that would convey influence from A to B.

But this "catechism" gives no credence at all to other forms of causal inference that have been long practiced within the historical social sciences. Mill's methods of similarity and difference offered one such example. Likewise, process-tracing (Bennett), comparative studies of similar cases (Skocpol), case-study methods (Hopkins), and other approaches have been used in comparative historical sociology to formulate and defend hypotheses about causation in history. 

In The Grammar of Time Marcus Kreuzer undertakes to bring these forms of historical-causal reasoning together under the rubric of comparative historical analysis (CHA). He describes the ambitions of CHA in these terms:

Like historians, CHA scholars use the past to formulate research questions, describe complex social processes, and generate new inductive insights. And, like social scientists, they compare those patterns to formulate generalizable and testable theories. (1)

A notable feature of Kreuzer's account is his view that comparative historical analysis permits both exploration (hypothesis formulation) and assessment (hypothesis evaluation). This point corresponds to a distinction of longstanding, the distinction between the context of discovery and the context of justification. Kreuzer takes the view that we can analytically separate the two contexts, but in practice the researcher needs to be involved in both activities. "CHA makes such exploration an integral part of the research process, because studying a constantly changing world requires continuously updating your research questions" (4).

Kreuzer uses the terms "induction" and "patterns" frequently, but it is worth questioning whether historical research supports either term. Induction means discovering persistent regularities; patterns are the regularities that are discovered by induction. But does history really support "induction"? This seems to imply that there is an unchanging underlying order to history, that produces recurring patterns of outcomes. But why should we believe this? In other places Kreuzer emphasis the "chaos" and unpredictability of history, the contingency of historical processes; but this seems to seriously undercut the idea of underlying order that appears to be crucial for an ontological justification of induction. So perhaps the idea of "historical induction" is unfounded.

The title of Kreuzer's book -- The Grammar of Time -- is intriguing, and it requires that we think carefully about what Kreuzer means by the phrase "grammar of time". Kreuzer briefly explains the metaphor in the introduction:

The grammar analogy is meant to highlight several features of CHA. Grammars analyze cultural phenomena – language – that emerged independently of one another in different places. The same goes for CHA. It established itself in different disciplines independent of one another and therefore sub- sumes different traditions that are distinct without necessarily being unique. Grammars also incorporate time to capture change. The conjugation of verbs differentiates degrees of the past and their relationship to the present and the future. And the past perfect tense even makes the past come alive by identify- ing activities that were ongoing in the past rather than having just occurred in the past. Grammars also consider geography, since their rules vary with each language. And etymology, a cognate discipline of grammar, recognizes that language itself is a changing and hence historical phenomenon. Finally, learn- ing grammars is peculiar because it involves understanding more systematic- ally what we already mastered intuitively. It requires paying attention to the scaffold of language, which neither is particularly elegant nor serves many uses after we have learned a language. (2)

However, the features mentioned there do not capture much of what we mean by "grammar" in post-Chomsky linguistics. Making the metaphor of grammar central to this style of analysis suggests that CHA offers the kind of abstract analysis offered by "syntax" (as distinguished from "semantics" or meanings), which in turn suggests that CHA offers an abstract and general way of "parsing" historical moments and changes. A sentence can be grammatically parsed into "noun", "verb", "state-term". And perhaps Kreuzer means to suggest that historical events can be parsed into abstract elements that combine to instantiate a moment of historical change. A second association created by reference to grammar is the idea of a "generative" grammar: the idea that, once we have a correct analysis of a string of words, we can generate a full representation of the state of affairs represented by the string. We might take a further step and follow Chomsky's distinction between "surface grammar" and "depth grammar", and postulate that there is a common structure of events that underlies the apparent diversity of event-types in the historical flow. (This idea is not discussed in the book.) 

An idea that Kreuzer uses frequently is "unfreezing": 

While unfreezing geography and history constitutes the first step in historical thinking, comparisons guide the explorations of this newly unfrozen temporal and geographic terrain. (10) 

It isn't clear what this means, however. It has something to do with putting aside the theoretical categories in terms of which a social scientist characterizes an event. But why "unfreezing"? He contrasts "types of comparison" in these terms in Table 1.1: 

  • Frozen (Thin) (Single or multiple moments without dates)
  • Unfrozen (Thick) (Multiple moments with specific date)

But the point of the contrast is not obvious. Perhaps a clue is offered in "thin" versus "thick"; this seems to be a reference to the idea that one set of observations is contextualized, whereas the second set is abstracted from context. But this does not have much to do with the physical processes of "freezing" when we think of ice and water. So the concepts of "frozen" and "unfrozen" history seem too flimsy to provide clarity about how to think about historical processes, historical change, and historical contingency. And table 2.1 (Varieties of historical time) doesn't do much to clarify the concepts or to demonstrate their meaning and value. The ideas of "solidly frozen", "partially thawed", "fluid, thick", and "fully fluid" don't seem to shed light on different "phases" of historical process, and it isn't clear how they are thought to correspond to the "types of comparison" listed in the final row of the table: "cross-sectional", "contextual", "serial", and "historical".

In Figure 1.1 (CHA's genealogy) Kreuzer suggests that CHA aims at discovering large differentiated processes of change and transformation in world history. 

One of the more interesting chapters is Chapter 5, a detailed review of the evolution of research on proportional representation systems of voting in Europe. Here Kreuzer tries to illustrate the way that a research effort guided by the assumptions of comparative historical analysis might proceed. He holds that this case illustrates the assumptions of "eventful history".

All in all, I'm not sure the metaphor of "grammar" works well for Kreuzer's interpretation of the methodology provided by CHA. Do historical events reflect an underlying syntax, so that we can see the "structural similarity" that exists between the Russian Revolution, the Copernican Revolution, and the Black Lives Matter movement? Do "time" or "history" possess an underlying set of rules that generate all possible outcomes? Surely the answer is no. Compare a technical analysis of the idea of a grammar with McAdam, Tarrow, and Tilly's alternative in Dynamics of Contention. MTT suggest that episodes of contentious politics can be grouped under descriptive categories -- "revolution", "uprising", "riot", "civil war"; but they insist that these categories do not reveal much about the events that they embrace. Instead, MTT argue, it is useful to examine an open-ended set of mechanisms and processes at the meso-level that can be found to recur across the various instances of revolutions or riots; and that the mechanisms of social contention are more explanatory than the high-level categories. This approach is not syntactic; it is more analogous to evolutionary biology, which seeks to identify proximate mechanisms (and one large mechanism) through which populations differentiate into distinct groups.

The book is a welcome contribution to discussions of methodology in historical sociology, and a refreshing alternative to more narrow approaches to the problem of empirically supporting causal claims about historical processes.