What Are We Actually Tracing? Process Tracing and the Benefits of Conceptualizing Causal Mechanisms as Systems


Derek Beach

University of Aarhus, Denmark

[**Original document page numbers preserved in brackets for citation purposes. **]

[Start Page 15]

Derek Beach is Professor of Political Science at University of Aarhus, Denmark. He can be reached at derek@ps.au.dk.

1.       Introduction

Most scholars now agree that process tracing, as a distinct social science method, involves tracing causal mechanisms using in-depth case studies. Studying causal mechanisms shifts the analytical attention from causes and outcomes to the causal process that links causes and outcomes together. However, while they are widely used, statements about causal mechanisms are also the least understood type of causal claim in the social sciences.1 This short essay discusses two competing approaches to studying causal mechanisms (a counterfactual- based and systems understanding of mechanisms), arguing that adopting a systems understanding of mechanisms results in research designs that produce actual evidence of the process we are attempting to trace instead of hypothetical ‘what ifs’ or weak comparisons.2

Before we proceed, it is important to note that some scholars use the term causal mechanism to refer to a series of events, or a narrative story, prior to the occurrence of an outcome.3 Describing a series of events can provide a plausible descriptive narrative about what happened, but it does not shed light on the causal question of why things happened. Other scholars like Abell do go a bit further than just tracing events. Abell contends that when studying causal mechanisms, we need to go beyond tracing events to develop narrative structures with action linkages that build on subjective counterfactuals, where we ask actors who participated in a process whether things could have been different at critical junctures of a process.While Abell’s suggestion does point us in the direction of moving beyond just tracing events, he then reduces the scope of research questions that we can study to only those that can be assessed by asking actors themselves whether things could have been different. The two approaches to causal mechanisms discussed in this paper go beyond tracing temporal sequences or subjective counterfactuals in their attempts to ex- plain more explicitly why things happen.

2.       Mechanisms as a series of ‘mini’-counterfactuals and problems with the masking of causal logics

Within the social sciences, the most widespread understanding

[Start Page 16]

of mechanisms is to view them as a form of counterfactual, often described as intervening variables.5 The clearest statement of this position is found in Woodward’s work, where he contends that if a given part of a mechanism did not exist, the next part of the mechanism would also not have existed, thereby enabling us to infer that the preceding part was causally linked to the outcome because its absence produced a difference.6 Mechanisms are viewed as one or a series of ‘mini’-counterfactual claims that are at a lower level than the overall causal claim of a relationship.7 Therefore, in Woodward’s understanding, all mechanisms are sets of counterfactual claims about links in a process. If one part does not work, the rest of the mechanism does not function.

The theoretical logic of the process is masked

The counterfactual-based understanding has several theoretical and methodological disadvantages. First, there is the theoretical problem that the logical causal arguments linking parts of mechanisms together are masked in the counterfactual understanding. When the parts of mechanisms are understood as lower-level counterfactuals, the links between them are typically depicted as causal arrows linking one part to the next without making explicit what it is about one part that links it in a causal sense to the next part. For example, Rosato describes a causal mechanism linking democracy and peace as: Democracy accountability group constraint Peace.8 Yet we are not told anything more about what it is about accountability that actually is linked to group constraint, meaning that the causal logic behind the mechanism is not made explicit: it is “masked” or “black-boxed” such that we aren’t invited to see the causal logic inside the mechanism. The masking of causal links is not problematic in the experimental tradition because it is the actual counterfactual manipulation that enables us to infer causation; we know there is a causal effect because we can measure the difference that the manipulation makes irrespective of whether we actually know anything about how the effect is produced. But this theoretical masking of causal mechanisms becomes a critical weakness when studying causal mechanisms using case studies because we cannot assess whether the causal logic linking parts together actually makes sense, or whether there is empirical evidence of the causal links working in a case.

Methodological problems

In addition to masking the causal logic of mechanisms, theorizing mechanisms as a series of counterfactuals forces us to investigate empirically whether the presence or absence of each part of a mechanism makes a difference in whether or not the process functions to make causal inferences.9 To assess difference-making means that we have to conduct a form of experimental manipulation for each part of the mechanism, investigating the difference that the presence or absence of a part of a mechanism makes for the overall outcome, other things held equal. This can be achieved in principle either through an actual experiment with human manipulation of the research subject, or by implementing a weaker design with a natural or logical experiment.10

What then are the disadvantages of utilizing experiments to study mechanisms? Actual controlled experiments are great for providing evidence of difference-making (whether the presence or absence of a part makes a difference in the function of the mechanism) through the manipulation of the presence or absence of a treatment across cases, enabling strong inferences about causal effects. But to assess parts of a mechanism using actual experiments would involve disaggregating a mechanism into a series of parts and then deploying mini- experiments for each counterfactual claim in our mechanism! There are few social science research questions for which it is relevant to engage in a set of sequential mini-experiments in a meaningful fashion. For example, what kind of experiment could validly model each part of a mechanism separately whereby lobbyists influence policy-makers? Could we create a research situation where we investigate first in an experimental fashion (i.e. holding all other things equal) whether information provided by a lobbyist and not provided in a control case makes the hypothesized difference in policymaker attitudes toward the lobbyist, followed by another controlled experiment where we investigate whether this change of attitude actually produces the hypothesized difference in policy-maker behavior?

If we take the natural experiment route, this requires that we find two real-world cases that are similar in all respects except for the presence and absence of a particular part of a causal mechanism. If we unpack a mechanism into multiple parts (i.e. a series of intervening variables in the counterfactual understanding), a natural experiment would involve trying to find two completely comparable cases for each part: one where the part is present and one in which all other factors are present but the part of the mechanism being assessed. We would then have to repeat this natural experiment for each part of the mechanism! But could this even be achieved? A prominent advocate of counterfactuals admits that “…controlled comparison and matching face the difficulty of finding real-world cases that are identical in all respects but one…”11 But unless we can substantiate that ‘all other things are equal’ except for the presence or absence of a single part of a mechanism, we cannot make a causal inference that the absence of a part made a difference.

Using logical hypotheticals to assess difference-making is a counterfactual-based method where we compare an existing real-world case with a hypothetical “what if” case where all

[Start Page 17]

other things are present except the part of the mechanism whose difference-making we want to assess. The logical argument is then made that if a particular cause had not occurred, the outcome would not have occurred. In effect, a logical counterfactual comparison attempts to approximate a most-similar systems test. To best approximate a most-similar systems test, the most important criterion is the ‘minimal-rewrite’ rule: hypothetical changes should involve only the most minor changes possible to enable us to assess whether the change can logically produce a major change in an outcome.12 But a minimal re-write is a non-achievable ideal in most circumstances, as the absence of all but trivial conditions could have significant knock-on effects for other causal conditions, potentially shifting a number of other conditions in our logical counterfactual situation.13 And logical hypotheticals face the critical challenge of being merely “what ifs” without any actual empirical evidence. Despite many attempts to build a methodology for logical counterfactuals,14 there are no objective empirical truth conditions for assessing a non-existent but possible alternative world.15 Single-case counterfactual comparisons involve comparing an existing real world case with a hypothetical counter- factual case.

At a more fundamental level, even if we could engage in meaningful assessment of the difference that parts of mechanisms make through some form of experiment, assessing the difference that an individual part makes assumes that there is not redundancy embedded in the mechanism. In biological mechanisms there is (luckily) redundancy of key parts of the mechanism, meaning that if we remove one part to see what happens, we still find that the mechanism worked to produce the outcome because another previously unknown part with similar capacities became activated instead.16 There is no rea- son we should not expect similar redundancy in key parts of important social mechanisms. But based on the counterfactual logic, if we found that removing a part had no real effect on the outcome, we would disconfirm the part as being causally relevant, which would be a flawed inference if redundancy exists. In the case of redundancy, the correct inference would be that the part was only one of multiple ways that causal forces could be transferred through the mechanism. Using evidence of difference-making gained from counterfactual-based experimental manipulation, would not enable us to detect redundancy.

An additional problem is that isolating the difference that individual parts make results in a form of atomistic analysis of the workings of mechanisms. Amongst scholars who view mechanisms as systems,17 mechanisms are viewed in a more holistic fashion, and their theories attempt to capture the complex interrelationship between parts, where the effects of individual parts often only manifest themselves fully together with the effects of other parts. This holistic view is seen in Cart- wright’s work, where she writes that, “There are any number of systems whose principles cannot be changed one at a time without either destroying the system or changing it into a system of a different kind.”18

Even if we could utilize experiments, they would still not tell us how a process works.19 An experiment provides us with no direct evidence of how one part is actually linked to the next; all we know is that the absence of a part results in break- down. In the words of Bogen, “How can it make any difference to any of this whether certain things that did not happen would have or might have resulted if other things that did not actually happen had happened?”20

3.       Mechanisms as systems: theoretical and methodological considerations

Proponents of the systems understanding of causal mechanisms contend that “...a mechanism explanation for some happening that perplexes us is explanatory precisely in virtue of its capacity to enable us to understand how the parts of some system actually conspire to produce that happening.”21 The focus in the systems understanding is the dynamic, interactive influence of causes upon outcomes, and in particular how causal forces are transmitted through a series of interlocking parts of a causal mechanism to produce an outcome in actual cases. In the systems understanding, mechanisms are typically described as being comprised of a series of parts com- posed of entities engaging in activities.22

Consider a mechanism as a simple machine. Entities are the toothed wheels in the machine. Entities engage in activities, which are the movements of the toothed wheels. Activities produce change that transmits causal forces through each part of a mechanism.23 The most important aspect of theorizing mechanisms in the systems understanding is therefore capturing explicitly what is inside causal arrows, making explicit the causal logic whereby the activities of one part of a mechanism link it to the next part, attempting to formulate an overall mechanism with “productive continuity” between the cause (or set of causes) and an outcome.24 When there is productive continuity in our theorized causal mechanism, there are no large logical holes in the causal story linking C and O together.25 

[Insert Figure 1 Here]

[Start Page 18]

If a mechanism is represented schematically by C -> [E1 -> E2 -> E3] -> O, then the continuity lies in our description of what the arrows represent, defining very explicitly what activities actually transfer causal forces from one part of a mechanism to the next. A missing arrow or the inability to specify an activity connecting one part to the next leaves an explanatory gap in the productive continuity of the mechanism.26

What the entities and activities are precisely will be dependent on the type of causal explanation we are working with, along with the level at which the mechanism works and the time span of its operation. A theory of a mechanism as a system can be depicted as in Figure 1, where each part of the mechanism between a cause and an outcome is detailed in terms of the entities engaging in activities, focusing in as much detail as possible on how the activities of one part of the mechanism produce the next part of the mechanism. The entities are defined as nouns, whereas the activities are depicted as verbs. As seen in Figure 1, the contextual conditions for the proper operation of a given mechanism should also be made explicit.

Theoretical advantages of unpacking mechanisms as systems

There are not many examples of good theorized mechanisms as systems in the social sciences. Most commonly, theorized mechanisms unpack the causal process slightly, but key parts of the causal story typically remain masked. For example, in Ziblatt’s 2009 article about the sources of electoral fraud in new democracies, he claims to be studying a “capture” causal mechanism that links landholding inequality to electoral fraud in Imperial Germany.27 However, the actual theorized causal mechanism is never fleshed out in enough detail to allow us to either evaluate the underlying theory (the causal logic) or to assess empirically whether there is evidence that confirms the “capture” mechanism. The closest we get to a theoretical description of the causal mechanism is where he writes (italics are the author’s), “They [the landed elites] exert influence indirectly via the capture of rural public officials such as mayors, county commissioners, police officials, and election officials, who in turn are the actors that interfere with free and fair elections. In its most acute form, capture occurs as socioeconomic interests infiltrate the state by using their own personnel to staff the state.”28

While telling us something about the process—that it goes through local officials and in extreme instances might occur through direct staffing—Ziblatt does not detail a causal mechanism that exhibits productive continuity. In particular, he sheds little theoretical light on the causal process whereby landed elites are able to capture local officials. For instance, what types of power resources do landed elites deploy to cap- ture officials? Does capture occur through the use of material resources, such as controlling sources of revenue, or by deploying discursive or social resources? Do landed elites have to actively intervene to capture officials, or do officials anticipate what local officials want? When and why should local officials be responsive? And once captured, what is the process whereby local officials actually engage in electoral fraud? What types of actions do they use? Removal of voters from electoral rolls, pressuring poll officers, and so on?

In this example, the reader is left in the dark about the causal logic linking the parts of the process together. Indeed, very different causal logics could be working in Ziblatt’s theorized mechanism depending on whether the “capture” is theorized to work through material or social resources or through active intervention or anticipation. And at the empirical level, because the links are not made explicit, Ziblatt’s analysis produces small empirical vignettes that insinuate the existence of an underlying mechanism without providing mechanistic evidence that would enable causal inferences to be made. Indeed, how can we claim we have evidence of a process when we are not told what the process is that is being studied?

Working within the systems understanding of mechanisms, we would want to expose the causal logic between each part, making more explicit in particular the activities that are

[Insert Figure 2 Here]

[Start Page 19]

theorized to transfer causal forces from one part to another. For instance, Figure 2 illustrates what the capture mechanism of Ziblatt’s theory could look like if we fleshed out the causal logic linking each part to the next.

The core difference between the capture mechanism used by Ziblatt in his article and the capture mechanism theorized in Figure 2 is that my formulation more explicitly details the causal logic of each part of the causal mechanism, in particular which activities form the causal logic linking the parts together. The subsequent empirical analysis would attempt to assess whether the observable manifestations of the activities of entities for each part were present in a case, enabling us to get closer to actually tracing the empirical fingerprints left by the operation of the parts of a hypothesized causal mechanism.

Early in our research we might not know exactly how one part is linked to the next, thereby resulting in a lack of productive continuity in our explanation. However, this does not mean that we should give up our search for an “unpacked” causal mechanism that exhibits productive continuity. Instead, we can engage in further theoretical work, engaging in a far-ranging search for inspiration from existing theories,29 or engaging in a more inductive, exploratory empirical analysis of a case that attempts to shed light on how particular causal links work.

Methodological advantages (and drawbacks) of understanding mechanisms as systems

In the systems understanding, of causal mechanisms, what we want to trace empirically is whether a process (mechanism) actually played out as we had theoretically expected it to in a given case. Here we are more concerned with what actually took place in the empirical record and develop theories about processes on this basis instead of based on counterfactuals of what “might have” happened if things had been different.30 Instead of engaging in forms of mini-experiments for all the parts of the causal mechanism to assess how a causal process played out, we can assess the observable evidence left by the actual activities of entities in a real world case. The distinction between the type of evidence used to study mechanisms in the counterfactual- and systems understanding can be seen in Figure 3, where the top example is of evidence of difference- making used in a counterfactual, and the bottom example is of mechanistic evidence used in the systems understanding.

The crucial distinction is in the object of the evidence.31 In the systems understanding, mechanistic evidence is used to trace the observable manifestations left by the activities of entities for each part of a mechanism, with the goal of causal explanation, to gain a better understanding of how causal processes actually work. In contrast, in the counterfactual-based understanding, these activities are masked when parts are theorized to be manipulable counterfactuals because they remain masked, depicted merely as causal arrows without telling us what the causal link actually is. The counterfactual-based ac- count in effect transforms the within-case study of causal processes into a cross-case analysis of patterns of variation at a lower level of aggregation, a level at which we lose focus on the process between the cause and outcome.32 The result of transforming a within-case tracing of causal processes into a cross-case assessment of difference-making is that we gain little information about how the process actually played out in a case. But is studying the process between a cause and an outcome not the reason we want to trace a causal mechanism in the first place? This distinction can be seen in Figure 3. The top half illustrates that evidence of difference-making for parts of a mechanism would actually be evidence of the difference between the two “cases” (treatment and control). Evidence of difference-making between the case and the counterfactual control case is the comparison of what happens when part 2

[Insert Figure 3 Here]

[Start Page 20]

 (the second part of the causal mechanism) is present (mechanism continues) and what happens in the control case where part 2 is absent. Difference-making evidence is binary in form, in that either the process broke down (first scenario in Figure 3) or continued without interruption (second scenario). If the process breaks down and we have controlled for other potential causes of breakdown because everything else is the same between the two cases, we can conclude that part 2 has a causal effect and is necessary for the mechanism.

The bottom half of the figure illustrates mechanistic evidence for the same part of the mechanism. Mechanistic evidence would try to assess the observable manifestations left by the activities of the entities for part 2 instead of comparing the causal effect of the presence and absence of a part. When assessing mechanistic evidence, we ask ourselves what observable manifestations the theorized entities engaging in activities might leave in a case. The difficulty with using mechanistic evidence is that it is much more complex and multifaceted than the counterfactual-based difference in the presence or absence of an outcome. Instead, many different types of empirical material can be used as evidence to update our confidence in the functioning of a part of a causal mechanism. Empirical material could be in the form of trace evidence of activities,33 or sequence evidence in the form of expectations about a sequence of activities.34 Evaluating the probative value of evidence of the workings of parts of mechanisms requires careful operationalization of the empirical fingerprints the activities of entities might leave and justifications for what finding or not finding different types of evidence tells us.35

4.       Conclusion: the benefits (and drawbacks) of conceptualizing mechanisms as systems

This essay has argued that a systems understanding of mechanisms has several advantages over the more common counterfactual-based understanding. First, fleshing out the causal links

[Start Page 21]

for each part of a mechanism exposes the underlying causal logics to more detailed scrutiny, resulting in better theories, other things equal. Second, theorizing mechanisms as lower- level counterfactuals implies that we assess them empirically by analyzing the difference that their absence makes for the rest of the causal process. But studying mechanisms as a series of mini-experiments runs into many methodological problems and challenges, the most serious being that it only tells us about the causal effect of the individual parts of a mechanism, and we still do not know how a causal mechanism actually works in reality.

Unpacking mechanisms as systems, where each part is clearly theorized as an entity engaging in particular activities, has theoretical and methodological advantages, but it is by no means a panacea. Indeed, it tells us little, if anything, about the overall causal effect that a cause has on an outcome.36 And on a practical level, collecting mechanistic evidence for each part of a multi-part mechanism is incredibly resource demanding. There is a reason that, despite knowing about causal effects since the 1950s, it was only in the 1990s that medical researchers were able to muster strong mechanistic evidence for each part of the causal mechanism linking smoking and lung cancer.37 But just because it is difficult does not mean that we should not try to unpack mechanisms as systems and trace them empirically when the analytical benefits for our under- standing of how things work are potentially so high.


Abell, Peter. 2004. “Narrative Explanation: An Alternative to Variable-Centered Explanation?” Annual Review of Sociology vol. 30, no. 1: 287–310.

Beach, Derek and Rasmus Brun Pedersen. 2016. Causal Case Studies: Foundations and Guidelines for Comparing, Matching and Tracing. Ann Arbor: University of Michigan Press.

Bennett, Andrew. 2016. “Do New Accounts of Causal Mechanisms offer Practical Advice for Process Tracing?” Qualitative and Multi- Method Research: Newsletter of the American Political Science Association’s QMMR Section vol. 14, no. 1-2.

Bogen, Jim. 2005. “Regularities and causality; generalizations and causal explanations.” Studies in History and Philosophy of Biological and Biomedical Sciences vol. 36: 397–420.

Brady, Henry. 2008. “Causation and Explanation in Social Science.” In The Oxford Handbook of Political Methodology. edited by Janet Box-Steffensmeier, Henry E. Brady, and David Collier. Oxford University Press.

Cartwright, Nancy. 2007. Hunting Causes and Using Them: App- roaches in Philosophy and Economics. Cambridge: Cambridge University Press.

Darden, Lindley. 2002. “Strategies for Discovering Mechanisms: Schema Instantiation, Modular Subassembly, Forward/ Backward Chaining.” Philosophy of Science (Supplement PSA 2000 Part II) 69: S354–365.

Dowe, Phil. 2011. “The causal-process-model theory of mechanisms.” In Causality in the Sciences, edited by Phyllis McKay Illari, Federica Russo and Jon Williamson. Oxford: Oxford University Press, 865- 879.

Fearon, James. 1991. “Counterfactuals and Hypothesis Testing in Political Science.” World Politics vol. 43, no. 2: 169–195.

Gerring, John. 2010. Social Science Methodology. Cambridge: Cambridge University Press.

Goertz, Gary, and Jack S. Levy, eds. 2007. Explaining War and Peace: Case Studies and Necessary Condition Counterfactuals. London: Routledge.

Groff, Ruth. 2011. “Getting past Hume in the philosophy of social science.” In Causality in the Sciences, edited by Phyllis McKay Illari, Federica Russo and Jon Williamson. Oxford: Oxford University Press, 296-316.

Gross, Neil. 2009. “A Pragmatist Theory of Social Mechanisms.”

American Sociological Review vol. 74, no. 3: 358–379.

Hedström, Peter and Petri Ylikoski. 2010. “Causal Mechanisms in the Social Sciences.” Annual Review of Sociology vol. 36: 49–67.

Illari, Phyllis McKay. 2011. “Mechanistic Evidence: Disambiguating the Russo-Williamson Thesis.” International Studies in the Philos- ophy of Science vol. 25, no. 2: 139–157.

Illari, Phyllis McKay and Federica Russo. 2014. Causality. Oxford: Oxford University Press.

Imai, Kosuke, Luke Keele, Dustin Tingley, and Teppei Yamamoto. 2011. “Unpacking the Black Box of Causality: Learning about Causal Mechanisms from Experimental and Observational Studies.” American Political Science Review vol. 105, no. 4: 765–789.

Lebow, Richard Ned. 2000. “What’s So Different about a Counter- factual?” World Politics vol. 52, no. 4: 550–585.

Levy, Jack S. 2015. “Counterfactuals, Causal Inference, and Historical Analysis.” Security Studies vol. 24, no. 3: 378–402.

Machamer, Peter, Lindley Darden, and Carl F. Craver. 2000. “Thinking about Mechanisms.” Philosophy of Science vol. 67, no. 1: 1–25. Machamer, Peter. 2004. “Activities and Causation: The Metaphysics and Epistemology of Mechanisms.” International Studies in the

Philosophy of Science vol. 18, no. 1: 27–39.

Mahoney, James. 2001. “Beyond Correlational Analysis: Recent Innovations in Theory and Method.” Sociological Forum vol. 16, no. 3: 575–593.

Mahoney, James 2012. “The Logic of Process Tracing Tests in the Social Sciences.” Sociological Methods and Research vol. 41, no. 4: 570–597.

Mayntz, Renate. 2004. “Mechanisms in the Analysis of Social Macro- Phenomena.” Philosophy of the Social Sciences vol. 34, no. 2: 237– 259.

Rohlfing, Ingo. 2012. Case Studies and Causal Inference: An Integrative Framework. Houndmills: Palgrave Macmillan.

Rosato, Sebastian. 2003. “The Flawed Logic of Democratic Peace Theory.” American Political Science Review vol. 97, no. 4: 585– 602.

Runhardt, Rosa W. 2015. “Evidence for Causal Mechanisms in Social Science: Recommendations from Woodward’s Manipulability Theory of Causation.” Philosophy of Science vol. 82, no. 5: 1296– 1307.

Russo, Federica and Jon Williamson. 2007. “Interpreting Causality in the Health Sciences.” International Studies in the Philosophy of Science vol. 21, no. 2: 157–170.

Steel, Daniel. 2008. Across the Boundaries: Extrapolation in Biology and Social Science. Oxford: Oxford University Press.

Suganami, Hidemi. 1996. On the Causes of War. Oxford: Clarendon Press.

Tetlock, Philip E. and Aaron Belkin, eds. 1996. Counterfactual Thought Experiments in World Politics. Logical, Methodological, and Psy- chological Perspectives. Princeton: Princeton University Press.

Waldner, David. 2014. “What makes process tracing good? Causal mechanisms, causal inference, and the completeness standard in comparative politics.” In Process Tracing: From Metaphor to Analytic Tool, edited by Andrew Bennett and Jeffrey Checkel. Cambridge: Cambridge University Press, 126–152.

Waldner, David. 2015. “What Makes Process Tracing Good? Causal Mechanisms, Causal Inference, and the Completeness Standard in Comparative Politics.” In Process Tracing: From Metaphor to Analytic Tool, edited by Andrew Bennett and Jeffrey Checkel. New York: Cambridge University Press, 126–152.

Woodward, James. 2003. Making Thing Happen: A Theory of Causal Explanation. New York: Oxford University Press.

Waskan, Jonathan. 2011. “Mechanistic explanation at the limit.”

Synthese vol. 183: 389–408.

Weller, Nicholas and Jeb Barnes. 2014. Finding Pathways: Mixed- Method Research for Studying Causal Mechanisms. Cambridge: Cambridge University Press.

Woodward, James. 2003. Making Things Happen: A Theory of Causal Explanation. Oxford: Oxford University Press.

Ziblatt, Daniel. 2009. “Shaping Democratic Practice and the Causes of Electoral Fraud: The Case of Nineteenth-Century Germany.” American Political Science Review vol. 103, no. 1: 1–21.


1 E.g. Brady 2008; Gerring 2010; Hedström and Ylikoski 2010; Waldner 2014.

2 Rundhardt 2015; Woodward 2003.

3 For examples of this understanding, see Abell 2004; Mahoney 2012, 571; Suganami 1996, 164–168.

4 Abell 2004, 295–296.

5 Gerring 2010; Imai et al 2011; Weller and Barnes 2014.

6 Woodward 2003, 350–358; also Steel 2008; Runhardt 2015.

7 Note that level here does not refer to theoretical levels of analysis, such as macro- versus microlevel theories. Instead, we are talking about the disaggregation of a causal process into smaller constituent parts. The process itself that is disaggregated can be a macrolevel theory (e.g. (e.g. economic development democratization) or a microlevel theory of small group decision-making.

8 Rosato 2003, 585–586.

9 Woodward 2003; Runhardt 2015.

10 Runhardt 2015.

11 Levy 2015, 390.

12 Tetlock and Belkin 1996.

13 Bennett (2016) in this symposium discusses a “Soviet growth” counterfactual. However, a scenario in which perestroika policies actually worked would most likely not be an “other-things-equal” comparison (minimal-rewrite) because other major changes would probably have had to occur for it to actually work (e.g., dramatically increased trade and loans from Western countries).

14 Goertz and Levy 2007; Tetlock and Belkin 1996; Lebow 2000;

Levy 2015; Fearon 1991.

15 Beach and Pedersen 2016.

16 Illari and Russo 2014, 158.

17 For examples, see Machamer 2004; Waskan 2011.

18 Cartwright 2007, 239.

19 Machamer 2004; Waskan 2011; Russo and Williamson 2007; Illari 2011; Dowe 2011.

20 Bogen 2005, 415.

21 Waskan 2011, 393.

22 Machamer, Darden, and Craver 2000; Machamer 2004; Rohlfing

2012, 35–36.

23 Machamer, Darden, and Craver 2000; Machamer 2004.

24 Machamer, Darden and Craver 2000.

25 Machamer, Darden, and Craver 2000, 3; Darden 2002, 283.

26 Machamer, Darden, and Craver 2000, 3. While it can always be argued that one could always unpack the parts of the mechanism even more, thereby heading down into the infinite regress problem, there is a clear pragmatism in the writings of Machamer and others that suggests that we can stop when the causal logic is explicated in a “good enough” fashion to give us a working understanding of the key working parts of the process.

27 Ziblatt 2009, 12–18.

28 Ziblatt 2009, 14.

29 Here formal modelling can be one helpful tool to “game through” potential pathways of mechanisms.

30 Waskan 2011: 394; Gross 2009; Groff 2011; Machamer 2004.

31 Illari 2011.

32 Mayntz 2004: 244–245; Illari 2011.

33 For example: Is there a record of a meeting between the lobbyist and politician?

34 For example, in a rational decision-making process we would expect actors to collect information, evaluate it, and then take a decision that maximizes their utility.

35 For more information, see Beach and Pedersen 2016.

36 Illari 2011, 150; Steel 2008, 68.

37 Russo and Williamson 2007, 162.

Figures and Tables