The Comparative Prescription in “Studying Public Policy”

Of Simeon's various methodological injunctions, one of the most forceful is that “policy study needs to be comparative” (Reference Simeon1976: 550, 551) Simeon concludes, “case studies can be a very valuable tool: but they must be comparative” (580, italics added). The direction suggested by Simeon was largely consistent with the conventional understanding of the comparative method as it existed at the time, its underlying logic of causality and its search for linkages between independent and dependent variables. In this approach, time played a distinctly secondary role.

As Simeon notes, the most striking characteristic of the existing literature was the concentration on single-case studies. He notes that such works can be useful in falsifying existing theories and suggesting new hypotheses to be applied and tested in further studies (551). He argues, however, that “few of these potential benefits have been realized” noting the following problems.

Individual case studies tend to be isolated and unique, each looking at different issues, using different methods, and asking different questions. This makes comparison extremely difficult. Cumulative knowledge and theory cannot simply grow automatically by piling case studies on top of each other. Case studies have also a tendency not to focus on the “normal” but on the unique, exotic or important, so insights gained from them may actually be misleading. Moreover, in focusing on a specific decision or piece of legislation, case studies tend to ignore those issues or alternatives which simply do not come up for debate. It is easy to get submerged in the minutiae of the issue itself, and therefore to miss what might be much broader factors influencing the outcome. (551)

Simeon argues that these problems might be “quite easily overcome” and offers three suggested strategies (551). In addition to applying different models or lenses to a single case (as Graham Allison did in Essence of Decision, Reference Allison1971), the remaining two suggestions are essentially calls for use of the comparative method either through programmes of co-ordinated case studies in which “similar questions, frameworks, and methods” are applied to “carefully selected issues of different sorts” or through explicit comparisons within a single study (551). Regarding the latter, he notes that “much is to be gained by very simple comparisons and by the selection of cases which offer particular promise of illuminating wider aspects of policy” (551).

The purpose of the comparative method for Simeon is to develop cumulative generalizations and theories and “systematically to link some set of independent explanatory variables with some dependent ones” (552). Consistent with this overarching aim, the structure of Simeon's article is to first present a detailed discussion of how to best “conceptualize the dependent variable” (556) in which he laudably suggests developing dependent variables that capture the scope, means and distributive impact of public policies (559). The following section of the article then outlines various explanations which he refers to throughout as independent variables. His ultimate conclusion is that “policy studies will advance to the extent that clear and explicit links between dependent and independent variables are established” (580).

In this, Simeon's prescriptions were consistent with, and did not fundamentally challenge, the conventional comparative method as understood at the time: “testing causal inferences largely through inspection of covariance across cases between a few explanatory variables of theoretical interest and the outcome to be explained” (Hall, Reference Hall, Mahoney and Rueschemeyer2003: 389). The underlying logic is described by Hall as follows: “The key point is that….the comparative method is essentially correlational. It bases inference about causal relations on covariation between a dependent variable and a small set of independent variables, and inspection of the cases is used primarily to determine the presence or value of such variables in them” (380). In this endeavour, the unfolding of time does not play a preeminent or privileged role.

Further insight into Simeon's approach can be gleaned from the studies critiqued in “Studying Public Policy” such as Bryden's Old Age Pensions and Policy-Making in Canada which Simeon characterizes as “among the best of such studies” (Reference Simeon1976, 551, fn. 9).Footnote ¹ Foreshadowing many of Simeon's prescriptions, Bryden's study of pensions examines the expansion of government intervention into realms formerly outside the acceptable range of state intervention (the scope of policy), the form this expansion took (the means of policy), and why these policies redistributed income among lower income groups rather than across them (the distributional impact of public policy), capturing all three of dimensions of policy later suggested by Simeon (Reference Simeon1976: 559–66). In explaining these outcomes, Bryden combines the environment (demands generated by urbanization and industrialization), power (the balance between dominant economic interests versus emerging pro-redistributive interests), ideas (the “market ethos”), institutions (especially the multi-party system), the policy process which itself is argued to have had an independent effect and, finally, feedback effects and how “outcomes affect inputs—both the nature of future demands and the level of support” (Bryden, 1974: 13).Footnote ² Thus, Bryden links pension policy with the three most vital elements of political science as identified by Simeon: power, conflict and ideology (Reference Simeon1976: 550). While Bryden and Simeon share a concern with what would come to be some of the central building blocks of historical institutionalism, such as power and the role of ideas, Bryden's work deviates from Simeon's methodological prescriptions in one single striking aspect: Bryden's deliberate and explicit adoption of the case study approach. Bryden was well aware of criticisms of the case study method of the kind later outlined by Simeon and presages them (1974: 5–6).

In turn, Bryden expressed his own skepticism of the use of comparative studies over single-case studies. First, he evinced concern in regard to comparative studies where such studies “obscure qualitative differences in policies” (1974: 7).Footnote ³ This contrasts with Simeon's criterion, regarding the dimensions of policy to be appropriately explained, that “we should, at least in principle, be able to measure them” (Reference Simeon1976: 557). This distinction would come to be mirrored in a broader divide in historical institutionalist methodology described by Fioretos and colleagues: “While historical institutional research has retained a particular affinity for qualitative methods associated with historiography and process-tracing, researchers have come to embrace a wider array of methods,” including more quantitatively oriented approaches (Reference Fioretos, Falleti, Sheingate, Fioretos, Falleti and Sheingate2016: 18).

Perhaps most importantly, Bryden's central concerns regarding the comparative approach relate to the historical unfolding of the process of policy development over time—an element of Bryden's work which remained unacknowledged in Simeon's brief reference to Bryden's study. In a passage that could easily have been written by a contemporary historical institutionalist, Bryden argues, “There is no substitute for the case study in one vital aspect of policy study. Policy making is a historical process, not merely in the narrow sense that any policy output has a time dimension even when viewed in isolation from its antecedents, but also in the broader sense that it has antecedents: it is an outcropping of a historical development” (1974: 6). It is this explicit recognition of the importance of time and history in Bryden—less fully appreciated or explicitly developed in Simeon's article—which most fully foreshadows the subsequent development of historical institutionalism. At the same time, Bryden's work also foreshadows a predisposition toward single-case and outcome studies that would, as argued below, emerge and persist in historical institutional analysis and against which Simeon's arguments in favour of a comparative methodology stand in stark contrast.

The Concept of Time in Historical Institutionalism and in “Studying Public Policy”

Arguably, the rise to prominence of historical institutionalism and its explicit and systematic focus on the unfolding of processes over time has been one of the most important developments in policy studies since Simeon wrote. One of the most obvious examples of a concept capturing the unfolding of processes over time is path dependence resulting from “the existence of self- reinforcing feedback processes” (Pierson, Reference Pierson2004: 11). Although historical institutionalism is the branch of institutionalism that “can be associated with the idea of path dependency” (Lecours, 2005: 16), a robust literature has developed proposing that more attention be given to reactive sequences. (Jacobs and Weaver, Reference Jacobs and Weaver2015) More important than these specific concepts, though, is the centrality of the concept of time which they illustrate.

Thinking in terms of processes unfolding over time has at least three distinct dimensions: feedback effects, timing and conjunctures, and sequence and ordering. In regard to path dependence (an instance of positive feedback effects), the methodological implications of this conceptualization are twofold. First, relatively small changes at one point in time may result in major shifts at later points in time. Second, “key causes” may be “temporally removed from their continuing effects” as “some initial event or process generates a particular outcome, which is then reproduced through time even though the original generating event or process does not recur” (Pierson, Reference Pierson2004: 15–16, 45, italics in original). Crucially, the relationship between casual variables and their effects is different in the initial stages of the process (when those variables trigger the causal process) than later in the sequence when those initial causal variables may even be absent. As Pierson highlights, “This is very different from the more typical search for invariant relationships among factors, in which the analyst assumes that if adding x to a setting causes y, then the removal of x should remove y as well” (46). Just as it is increasingly widely recognized that spatially distinct cases may violate the assumption that causal variables have relatively consistent effects across cases, the same situation may obtain over time with causal variables having different effects over time due to sequencing or other complex effects such as “tipping points,” “diminishing effects,” or positive feedback. Put differently, the causal relationship among variables shifts over the temporal course of the process and is different at the end than it was at the beginning. The crucial methodological point is that this militates against bisecting a causal chain at any given point in time in order to try to determine which independent variables at that point in time explain the dependent variable at that point in time. The central problem is that the value of the dependent variable at any given point in time may depend on the value of an independent variable at some other earlier point in time.

Time also matters for two additional reasons captured in the concepts of conjunctures or timing (interaction effects between distinct causal processes occurring at the same time) and sequencing or the order in which particular events take place (Pierson, Reference Pierson2004: 15). As an example, one conjuncture of particular importance in Bryden's account of the development of public pensions in Canada was the intersection of pension development with the “external” event of the Great Depression. In Canada, means-tested pensions were already in existence when the Depression occurred so that the resulting political pressures were channelled into demands for expansion of these programmes rather than into demands for programmes for for all aged people as occurred in the US where no programme was yet in place (183).

Similarly important is sequencing, with the order in which events or processes occur being seen to be of fundamental importance in determining outcomes. Sequencing matters because the causal relationship between different causal variables changes over time and is different at distinct points in the sequence. If the study of public policy is characterized by the challenge of multifinality (the same combination of causal variables leading to divergent outcomes), one reasonable possible cause is difference in the order or sequence in which those causal factors combine. For example, the political possibilities were different under a sequence in which contributory pensions and means-tested pensions were implemented simultaneously as in the US, in comparison with Canada in which a universal plan emerged first and a contributory plan was later grafted on: “in both cases, policies implemented at given points in time were among the determinants of subsequent designs” (Bryden, 1974: 7).

In response to the challenges posed both by feedback and sequencing, some propose simply treating a dependent variable in T1 as an independent variable in T2. However, this fails to recognize that historical institutionalism and related concepts such as path dependence represent a fundamentally different causal logic than that which underpins the conventional comparative approach. As Thelen and Mahoney note, “The temporal revolution in political science reflected a broad chorus reacting to the deficits of viewing politics in cross-sectional, one-off, snapshot ways” (Reference Thelen, Mahoney, Mahoney and Thelen2015: 25). Pierson crisply notes the problem of “adopt[ing] a cross-sectional approach to studying what should be understood as a long-term causal chain” (Reference Pierson2004: 3). As Hall argues, “Theories of path dependence explicitly draw our attention to the importance of history [and] militate against analyses into which past developments are simply imported as an independent variable because they imply that the causal impact of such developments depends on where they are located within the historical chain” (Reference Hall, Mahoney and Rueschemeyer2003: 385–86). As Sanders notes, “The central assumption of historical institutionalism is that it is more enlightening to study human political interactions…sequentially, as life is lived, rather than to take a snapshot of those interactions at only one point in time” (Reference Sanders, Rhodes, Binder and Rockman2008: 39). Creating a series of synchronic snapshots does not adequately address these concerns. It is subject to the problem outlined above that the value of the dependent variable in any given bisected slice of time may depend on the value of independent variables at some earlier point in time.

One might consider Simeon's own references to time in light of this more recent reconceptualization of the role of time in the study of public policy. At the most basic level, Simeon suggests the potential contribution of comparative study at different points in time: “policy study needs to be comparative across both space and time” (Reference Simeon1976: 550). Here, comparison across time is treated as analogous to comparison across space: “comparison—across time, units within a nation and between nations—again provides the tool for exploring some of these dimensions” (555). This is suggestive of a search for the type of relationship among variables similar to that which motivates comparative analysis across political units. It is also subject to the problems with examining discrete slices of time as outlined above.

At several points, Simeon also explicitly calls for examinations of policy over time rather than across points in time: “We need to look at the broad evolution of patterns of policy over long periods within countries, provinces and other units, in the ways they deal with similar problems as a first step towards the primary goal of explaining the differences” (550- 51, italics added) Furthermore, he argues that “case studies are not enough. We need longitudinal studies of the evolution of policy over long periods, and we need to take studies of culture, voting and the like and try to formulate hypotheses by which they might be related to policy” (580). In these instances, it remains open to debate whether policy remains simply the dependent variable with longer spans of time allowing for better achieving what he himself refers to as the “first step”: describing and defining policy as the dependent variableFootnote ⁴ In this formulation, evolution of policy over time is a central characteristic of what is to be explained but is not itself a crucial element of the explanation.

In an implicit recognition of the role of feedback (although Simeon never uses the term), Simeon notes that institutions themselves may be seen as policies and, as such, they “are both dependent variables, reflecting earlier decisions, and independent factors, conditioning the future play of political forces” (575). However, this passing note is made only in reference to institutions rather than being generalized to other explanatory factors and the continued insistence on the dichotomy of dependent and independent variables maintains the appearance of the atemporal nature of these causal relationships. While Simeon notes that institutions may be treated as both dependent and independent variables, it seems a stretch to interpret this as a methodological call for “comparative dynamics.” Rather, in regard to attempting to assess the impact of institutions on policy, he himself explicitly argues that “the most fruitful approach will probably be to conduct comparative studies of similar issues across units with clear institutional variations” (575). In this, Simeon's prescriptions appear to not be as fully appreciative of the types of challenges which the dynamic interaction of explanatory factors over time can pose for the utility of the comparative method in theory development and theory testing as almost certainly would be the case had he been writing today.Footnote ⁵

The Role of the Comparative Method in Historical Institutionalism

Proponents of both the broader historical institutionalist approach and its more specific variants such as comparative historical analysis continue to wrestle with the issue of the fit between an ontological focus on the operation of intertemporal causal mechanisms and the methodological choice between single-case and outcome studies versus the comparative method privileged by Simeon (Fioretos et al., Reference Fioretos, Falleti, Sheingate, Fioretos, Falleti and Sheingate2016; Mahoney and Thelen, Reference Mahoney, Thelen, Mahoney and Thelen2015). Adherents of these schools have not shifted away from a generalizing intent. As Pierson argues “most social scientists remain interested in developing at least limited generalizations—arguments that can ‘travel’ in some form beyond a specific time and place” (Reference Pierson2004: 6). Nevertheless, the ontology of intertemporal causal mechanisms is sometimes seen to augur in favour of single-outcome studies analogously to Noël's argument that “when” context appears important, idiographic studies provide the best road to knowledge” (Reference Noël2014: 662). In terms of the broader literature at the most general level, there remains a strong ongoing tension between a focus on the historical development of policy and comparative analysis.

In regard to the broader historical institutionalist school, this tension is clearly captured in a comparison of three of the most authoritative reviews of the state of the art each taking place a decade or more apart (Fioretos et al., Reference Fioretos, Falleti, Sheingate, Fioretos, Falleti and Sheingate2016; Pierson and Skocpol, Reference Pierson, Skocpol, Katznelson and Milner2002; Thelen and Steinmo, Reference Thelen, Steinmo, Steinmo, Thelen and Longstreth1992). According to the nascent historical institutionalist critique launched by Thelen and Steinmo, the early new institutionalism, however useful in terms of “illuminating cross-national differences,” tended toward “the study of comparative statics,” that is, explaining “different policy outcomes in different countries with reference to their respective (stable) institutional configurations” (Reference Thelen, Steinmo, Steinmo, Thelen and Longstreth1992: 14). This opened the approach to the critique that it represented a version of institutional determinism (14). Rather, according to Thelen and Steinmo, what was needed were more dynamic models intended to capture the interplay of causal factors over time (15). In turn, this required examination “both across countries and over time” (16). Of the seven substantive chapters comprising their collection, all but two were explicitly cross-nationally comparative.

A decade later, the overview of historical institutionalism provided by Pierson and Skocpol gave considerably less emphasis to the comparative approach as a hallmark of historical institutionalism. Pierson and Skocpol identify three central characteristics of historical-institutionalist scholarship in political science: first, this scholarship addresses “big, substantive questions,” takes time “seriously,” and pays attention to context and configurations, that is, examining the combined effects of institutions and processes rather than just one institution or process at a time (Reference Pierson, Skocpol, Katznelson and Milner2002: 695–96). In this rendition, the emphasis on time appears to have almost completely displaced an emphasis on comparison. In their discussion of historical institutionalism as a “major research strategy,” they note that seemingly non-comparative studies may “juxtapose time periods, regions, and policy sectors, turning what appear to be one or a few national instances into settings for many carefully compared cases” (715). However, they push this logic further to include single-case or single-outcome studies: “even within what appear to be single-case studies, empirical observations have often been multiplied by formulating and testing hypotheses about the mechanisms that connect causes to effects” (715). Thus, they conclude that “some [studies] are explicitly comparative, while others analyze trends within just one macrocontext” (694).

Further illustrating this tension, Fioretos and colleagues note, in their 2016 state-of-the-art examination of historical institutionalism, that “early historical institutionalists” were “methodologically committed to in-depth study of events and cases” and that they “favoured methods of agreement and difference among a small number of cases to identify the causal role of institutions” (Reference Fioretos, Falleti, Sheingate, Fioretos, Falleti and Sheingate2016: 11). Thus, it was the combination of historiographical approaches (including process tracing) and comparative research designs that guided engagement with empirical materials. While they note that “it was within comparative politics that historical institutionalism first emerged as a distinct approach to study the effects of institutions on politics” (21), this places historical institutionalism as a method within a subfield of political science (comparative politics) rather than characterizing comparative methodology as central to the historical institutionalist approach. Interestingly, they note that “historiographical modes of inquiry, counterfactual analysis and process-tracing…have remained hallmarks of the tradition” (9). The comparative method which they attribute to early historical institutionalists is conspicuously absent from this list of central modes of inquiry.

In their discussion of developments in historical institutionalism, including recent “methodological refinements” resulting from the challenges of equifinality (different paths leading to the same outcomes) and the challenges resulting from overdetermination (multiple historical events that could potentially be said to have caused a given outcome), the primary adjustment in regard to the former has been “to deepen and refine their use of qualitative methods to leverage historical archives” (17) and, in regard to the latter, to “refine how they study sequences to better adjudicate which events are proximate in causing political outcomes” (17–18). Interestingly, they do not even consider the possibility of harnessing the comparative method to addressing these very real methodological challenges.Footnote ⁶

Of course, there are a number of works of exceptional scholarship in the historical institutionalist tradition which combine historical and explicitly comparative approaches (Prasad, Reference Prasad2006; Sheingate, Reference Sheingate2001; Swenson, Reference Swenson2002; Thelen, Reference Thelen2004). As discussed below, the same may be said of comparative historical analysis, American political development and Canadian public policy. That said, the comparative imperative in these areas appears no more powerful now than at the outset of these various research programmes, and part of the reason appears to be the emergence of newly dominant ontologies which emphasize historical and case study methodologies in contrast with the comparative method.

Thus, a similar tension exists in the field of American political development which, as Robert Lieberman argues, is increasingly diverging from comparative politics and is becoming the last “respectable” field of area studies in American political science.Footnote ⁷ The roots of this divergence, he argues, results from the highlighting of American exceptionalism (a foundational point of American political development) combined with a single-country methodological focus by which American exceptionalism “becomes, analytically speaking, a self-fulfilling prophecy.” In contrast, he argues that the study of American political development should be thought of as inherently comparative and that “it is, at least, implicitly comparative and, at its best, explicitly so.” He argues that close careful studies of several countries contributes to understanding what is general about political development, what is nationally specific and what helps in developing theories about how politics works.

An explicitly comparative approach appeared to be asserting itself in historical institutionalism as a distinct subfield under the rubric of “comparative historical analysis.” As Mahoney and Rueschemeyer argue, “all comparative historical works fit comfortably within the field of historical institutionalism, but historical institutionalist works that are not explicitly engaged in systematic comparison do not fall within the field of comparative historical analysis” (Reference Mahoney, Rueschemeyer, Mahoney and Rueschemeyer2004: 11). Systematic comparison was a prerequisite for works in this genre, although Mahoney and Rueschemeyer noted that such comparisons could be between nation-states, formal subnational units or informal subnational regions, supranational regions or organizations, different socially constructed groups within a nation-state as well as different periods of time (Reference Mahoney, Rueschemeyer, Mahoney and Rueschemeyer2004: 14). However, not even a decade later, the comparative element of comparative-historical analysis would fade. Mahoney and Thelen issued a “fresh programmatic statement about comparative-historical analysis” (Reference Mahoney, Thelen, Mahoney and Thelen2015: xv) intended to update their initial programmatic statement of 2004. The commitment to relying on “systematic and contextualized comparison of similar and contrasting cases” (Thelen, Reference Thelen2004: 13) was, by 2015, replaced by a commitment to “problem-driven case-based research” (Thelen and Mahoney, Reference Thelen, Mahoney, Mahoney and Thelen2015: 5). A cursory examination of the appendix provided by Thelen and Mahoney, which lists “recent award-winning CHA [Comparative Historical Analysis] books in political science, 2000–2014,” includes works which are historical but, in no meaningful sense, comparative.Footnote ⁸

There are a number of important contributions by Canadian scholars to these broad discussions of new institutionalism and its historical institutionalist variant (Lecours, 2005; Smith, Reference Smith and Lecours2005). Moreover, there have been a number of important Canadian contributions regarding the nature of policy change over time and methodological approaches to understanding such change (see, especially, Howlett, Reference Howlett2009; Howlett and Cashore, Reference Howlett and Cashore2009; Howlett and Rayner, Reference Howlett and Rayner2006; Rayner, Reference Rayner2009). Given the specific relevance of “Studying Public Policy” to the Canadian context, there is also the question of whether the portrayal outlined above appropriately applies to the Canadian public policy literature on which Simeon's article has had such an impact. Turgeon argues, albeit on the basis of rather slim empirical evidence, that comparison is “increasingly a method of choice for policy analysis” (Reference Turgeon, Turgeon, Papillon, Wallner and White2014: 7) although the most recent of the handful of works cited in support of this claim is now almost a decade old. Certainly, a number of works by Canadian scholars have indeed been both comparative and historical (Boychuk, Reference Boychuk1998, Reference Boychuk2008; Maioni, Reference Maioni1998) including impressive recent contributions (Boothe, Reference Boothe2015; Haddow, Reference Haddow2015; Olive, Reference Olive2014; Wallner, Reference Wallner2014; White, Reference White2016). However, Noël suggestively, if not skeptically, raises the rhetorical question: “Are we so far, then, along the comparative turn? Do we compare naturally and more fruitfully than did our predecessors?” (Reference Noël2014: 653). In terms of the Canadian study of public policy, Noël argues that conclusions in this regard are not easy to draw and, at the very least, require significant further empirical assessment. There is, however, little systematic evidence to suggest that comparison has risen to methodological pre-eminence or that single-case and single-outcome studies have systematically become increasingly rare.

Combining Temporality and the Logic of Comparison

To the degree that the ontology of causal mechanisms unfolding over time can be seen to augur in the direction of single-case and single-outcome studies, the comparative method poses its own challenges to historical institutionalist analyses that are not comparative. Following the same logic that drove Simeon to favour of comparative approaches, assertions regarding either the uniqueness or the generality of processes unfolding over time can only be empirically assessed through comparison.

One example can be developed using the concept of path dependence. Often portrayed very literally, it is possible to think about a process being “launched” on a particular path and later being “on”—or even “locked into” —a specific path or as having taken a different “path” than otherwise might have been the case. However, path dependence is simply a concept. What is it specifically that the conceptual rendering of path actually represents? As Pierson notes, “Without careful attention to the identification of the mechanisms at work, analyses of path dependence can easily become descriptions of what happened rather than explanations for why it happened” (Reference Pierson2004: 49). Furthermore, without clearly specifying the mechanisms that reinforce a particular path or trajectory, path dependence arguments “degenerate into little more than a description of stability” (49). Béland and Hacker capture the essence of the problem neatly: “Of course, policy development always follows some sequence” (Reference Béland and Hacker2004: 52). Following a similar logic, do Vale distinguishes between tracing processes (an inductive procedure in which there is always a process that can be traced) and “process-tracing” (a deductive procedure in which the researcher looks for evidence of a “series of theoretically predicted intermediate steps” (68), that is, following some sequence is not the same as being on a path in a conceptually meaningful sense.

One possible way of thinking about the concept of a path is that it represents a given and specific relationship among causal variables. Being on a path and remaining on a path can be taken to mean that a given relationship between causal variables continues to obtain. Being on a different path implies that a different set of causal relationships exists. Differences between cases on distinct paths are not simply the result of variation in the values of independent variables but fundamental differences in the underlying relationships among variables. “Path-breaking” or “path-shifting” change need not be a radical rupture; rather, some change initiates a shift in the relationship among causal variables such that those factors interact in new and different ways than they did prior to the shift.

If the definition of path distinctiveness is precisely that a distinctive specific relationship between causal variables obtains, the challenge is how to establish this empirically. How do we know that path dependence is significant, by what methods do we measure it, and against what standards do we weigh the results? One possibility is to establish a shift or shifts over time within a single context. If there is a shift over time such that one can demonstrate that a new configuration among causal variables has come into play in a given example, the existence of distinct paths before and after an event (or in one period relative to another) can be represented as having resulted from path-breaking change. However, if there is no path-breaking change within a single context study, how does one know or demonstrate that a path (a set of relationships of sufficient causal weight to keep the case from veering off the path) even exists? How does one demonstrate or identify the forces which reinforce a given relationship between causal factors or empirically assess which type of events or developments might disrupt the relationship?

In many circumstances, the existence of processes conforming to the type of causal logic which underpins historical institutionalist and path dependence approaches cannot be established except comparatively. Comparative analysis of the relationship between casual factors may, in many cases, be required for the meaningful analysis of causal processes over time within individual settings. The central research endeavour becomes investigating not only the “links between dependent and independent variables” (Simeon, Reference Simeon1976: 580), but rather whether there is similarity or variation in the relationship between causal variables across cases such that different cases over time (or even an individual case across different periods of time) represent a different type of relationship among these factors. As Peter Hall argues, “Instead of thinking about political explanation as a matter of identifying a short list of variables that might impinge on an outcome, we should also be thinking about how these variables interact with one another within specific contexts to form distinctive patterns of politics across space and time” (Reference Hall2016: 44). In this vision, paths become analogous to the distinct cells of a typology in which different causal relationships obtain for all cases within each cell—much in the style of George and Bennett's “typological theorizing” (Reference George and Bennett2005: ch. 11). In typological theories, causal generalizations are limited and contingent. The central focus of empirical investigation becomes scope conditions: what factors determine the scope of cases for which a particular constellation of causal relationships exist. Extending this analogy to intertemporal analysis, the focus becomes what factors account for instances in which individual cases shift between cells over time such that a certain causal constellation which obtained in one period of time no longer obtains in another.

Path dependence analysis, in turn, should establish that a specific unit is distinctive in terms of the causal relationship existing between variables over a given period compared to some other period of time or some other case. The analysis would then turn to the question of what keeps the case in question on that path over time and/or what kind of disturbances have moved it (or might move it) to another path over time. By definition, such analysis must be undertaken over time as what is to be explained (the relationship between various causal variables) itself can only be established over some period of time. At the same time, it necessarily would also be comparative. The problem of identifying durable shifts as opposed to ordinary variation “calls us back to the roots of the comparative tradition.”Footnote ⁹

The value of applying the comparative method in light of a recognition of the importance of intertemporal causal mechanisms is further accentuated by two of the major challenges faced by conventional comparative analysis to which historical institutionalism is the ostensible solution: equifinality (in which different paths lead to the same outcome) and multifinality (in which similar causal mechanisms generate divergent outcomes.) As do Vale puts it, “the comparative framework must either systematically compare different processes across countries to explain how these paths led to similar outcomes, or contrast similar processes that have been changed in their unfolding, thus leading to different outcomes” (Reference Do Vale2015: 64). While this position is clearly based on the recognition of the importance of causal processes that unfold over time, it insists that a comparative methodology is a requirement for adequately interrogating these processes.

Further in this vein, Streeck argues that the ontology itself can be conceived as essentially comparative. As he summarizes succinctly, the “particular ontology of the social world” which animates comparative historical analysis is as follows: “Difference as a present fact is seen on a background of similarity as a past possibility, contingently suppressed by specifiable causes identifiable by comparative causal analysis” (Reference Streeck, Mahoney and Thelen2015: 265). In this vision, history remains key: “In comparative-historical analysis, present differences…are assumed to have been caused by identifiable events or conditions in a historical past, long enough ago not to be contemporary with the effects to be causally explained by them” (265).Footnote ¹⁰ Thus, in Streeck's particular variant, “comparative-historical analysis, in summary, is concerned with relatively stable, lasting, non-incidental differences between social entities whose origins lie far enough back in time to require uncovering by systematic historical research” (266). Streeck argues that “Comparative-historical analysis, unlike a historiography of events…deals with genuine otherness among otherwise similar and therefore comparable societies—one could also say: with multilinearity as distinguished from unilinearity in the development of societies” (272). As he notes, this “deep otherness of otherwise similar social structures” calls for “explanation by different histories” (273). In this formulation, those structures and histories cannot be adequately understood except by reference to comparison and combining comparative approaches with consideration of development over time.