[et_pb_section admin_label=”section”][et_pb_row admin_label=”Row”][et_pb_column type=”4_4″][et_pb_text admin_label=”Text” background_layout=”light” text_orientation=”left” use_border_color=”off” border_color=”#ffffff” border_style=”solid”]
This is a guest blog post by Simeon J. Newman who is a doctoral candidate in 
sociology at the University of Michigan with interests in political, historical and urban sociology, social theory, and the philosophy of the social sciences. His dissertation analyzes the politics of urban expansion and citizenship in Latin America since the middle of the 20th century. He’s also interested in temporality in social science explanation and has studied the state and civil society in the context of radical political change. His work has been supported by the Social Science Research Council and the National Science Foundation. Simeon is an alumni of the Philosophy of Social Sciences Seminar 2015 and he is currently developing his research and field work with the Critical Realism Network Graduate Student Working Group.
[/et_pb_text][/et_pb_column][/et_pb_row][et_pb_row admin_label=”row”][et_pb_column type=”4_4″][et_pb_text admin_label=”Text” background_layout=”light” text_orientation=”left” use_border_color=”off” border_color=”#ffffff” border_style=”solid”]
John Stuart Mill’s method of agreement and difference is probably the dominant approach in contemporary explanatory comparative and historical social science [i]. Based on comparisons across cases and an inferential logic concerned with identifying necessary and sufficient conditions [ii], the method has a lot to recommend it. Yet, it seems to have developed so splendidly that its limitations–long apparent to its critics–are becoming apparent even to its core proponents.
The Millian method is routinely included in comparative-historical and 
research design graduate-level sociology courses. In these settings, many students like me come to realize that the basic assumption underpinning the Millian method–a nominal-variables social ontology–leads Millians to downplay important causal factors. This stems from its “degrees of freedom problem”: If a Millian analyzes three cases, she will have (N-1=2) two degrees of freedom; she will be forced to argue that all three cases are alike in all important ways except two; but she will be free to argue that one or both of two factors are the causally important ones. In this way, the method’s epistemological strictures dictate its ontological assumptions rather than ontology dictating the adequacy of the method [iii].
James Mahoney is perhaps the Millian method’s foremost proponent. Earlier this year he delivered a talk at the University of Michigan (Mahoney and Barrenechea 2016) [iv] in which he discussed the importance of causal accounts based on concatenations of multiple determinative factors responsible for outcomes of interest. While he had touched on this in earlier written work (see especially Mahoney 2008), he forayed further into this issue during his talk in a way that suggests that he may be moving towards approach(s) that critical realists advocate (e.g., Gorski 2008; Steinmetz 1998, 2003, 2014). To make my case, I will focus in the rest of this post on the themes of causation, credence and fallibilism, and the place of theory in social science inquiry as they feature in some of Mahoney’s published work and in his talk, in prominent Millian research, and in critical realism.
Causation
In his earlier work, Mahoney noted that, in its pure form, the Millian method is based on the assumption that “causes exhibit . . . invariant relationships with outcomes within a specified domain of cases” (Mahoney 2003:341, 340; see also Mahoney 2008:420). It is on the basis of such deterministic ontological assumptions that Millians draw their conclusions. Skocpol (1979) argues that, within certain scope conditions, if you have a rigid agrarian class structure and the state enters into crisis, then you will have a social revolution; these two factors are determinative of social revolution. Yet, of course, it is reasonable to doubt that the world works so deterministically. Indeed, Mahoney discusses “probabilistic causality” for this very reason. In proper Millian fashion, though, he considers probabilistic causation to be a subtype of necessary-and-sufficient causation (Mahoney 2003:355, 368; Mahoney 2008:417, 425-29).[v]
In his lecture, Mahoney supplemented his account of causation with J. L. Mackie’s (1965) concept of an “INUS” condition. Mackie defines an INUS condition as “an insufficient but necessary part of a condition which is itself unnecessary but sufficient for the result” (1965:245), or, as Mahoney puts it, INUS conditions are “individual components of a package of factors that is sufficient (but not necessary) for an outcome” (Mahoney and Barrenechea 2016:8-9). [vi] The general point of INUS-based causal accounts is to point to what Mackie calls “elliptical or gappy causal laws” by showing not that X is the cause of Y but rather that X causes Y (Mackie 1965:252-53). The invocation of sufficiency is clearly meant to render INUS conditions in terms intelligible to those familiar with necessary-and-sufficient causal inference. But INUS conditions, be they in Mackie’s conceptualization or in Mahoney’s use, point to the limitations of Millian conceptions of causation. INUS conditions are neither necessary nor sufficient causes; they don’t fit into the group of factors that can be held constant across cases nor the group that varies between them in variables-based causal assessment. What kind(s) of causation do INUS conditions point to?
By embracing INUS conditions, Mahoney seems to actually advocate a 
non-Millian conceptualization of causation that is quite similar to what critical realists call combined and conjoint causation. [vii] On this conception of causation, variables don’t cause outcomes–combinations of causal factors do. Rather than identifying one (or a few) factor(s) that is (are) constant across all of the cases in which the outcome of interest occurs and that is (are) absent in all cases in which the outcome of interest doesn’t occur, what is important in this approach is how factors concatenate and–as concatenations–produce outcomes. The implications are considerable: unlike with the Millian-variables approach, different sets of factors whose elements are mutually exclusive could in principle concatenate to produce the same outcome.The point of empirical research that departs from such a conceptualization would, therefore, not be to identify variables (Xs) that are (always, as in covering-law approaches; only sometimes, as in approaches that stress scope conditions; or stochastically, as in probabilistic approaches) associated with an outcome of interest (Y); it would instead be to identify constellations of factors (ABCDEF, GHIJKL, ACKOPQR) that are.
Had she taken an approach based on conceptions of INUS conditions or combined and conjoint causation, Skocpol might have found that the French Revolution was the result of (A) a rigid class structure, (B) an old regime’s attempt to squeeze more taxes out of the peasantry and (C) their political restlessness, (D) the growth of a bourgeois class and (E) the ideologies of popular sovereignty that its members promulgated, and, finally, (F) decisive and assertive leaders; she might have found that the Russian Revolution was the outcome of the concatenation of (G) the coercive conscription of the peasantry, (H) the rapid formation of an industrial working class (I) whose members were thrown together and forced to fraternize in some of the largest factories in the world, (J) the inability or unwillingness of successive political elites to end the country’s involvement in the First World War, (K) the rising prominence of class-specific political organizing and identification, and (L) the presence of a highly-disciplined and ideologically-committed socialist organization that was (M) willing and able to carry out a surgical insurrection and then (N) hold onto power; and she might have found that the Chinese Revolution was the product of a concatenation of some of these factors (ACK) along with others (OPQR). The important thing for accounts of combined-and-conjoint causation is that the constellation of factors conceived of as jointly causal is not necessarily commensurate across cases, as it must be in the Millian and other deterministic variables-based approaches. [viii]
Credence and fallibilism
How do we know that the set of factors that we identify was responsible for the outcome of interest? Using the Millian approach, the idea is to select cases that allow the researcher to compare candidate causal variables vis-à-vis the outcome variable of interest, and, assuming a pattern of co-variation is discovered, infer the causal importance of candidate causal variables. For example, in Skocpol’s account, since there was no crisis of the state and no social revolution in England, but there was a crisis of the state and social revolution in France, Russia, and China, she attributed causal status to crises of the state.
Of course, if necessary-and-sufficient causal assessment is inappropriate, such causal inference is not possible. How, then, do we proceed? How do we know that the concatenation of factors ABCDEF caused the French Revolution? In his talk, Mahoney advocated uncertainty regarding inferences. Thus, I suspect that he would answer that we can’t know, in the strong sense of the word (i.e., we can’t arrive at justified true belief), whether factors ABCDEF caused the French Revolution. The expectation for this kind of research should, it appears, instead be for researchers to draw the most-plausible conclusions.
This implies a degree of fallibilism which is not present in Millian logical inference. The idea appears to be to proceed on the assumption that the point of this kind of social scientific research is to offer plausible conclusions on the basis of a sober conceptualization of the object of inquiry, an appropriate epistemology, and a sound research design and execution. These happen to be precisely the expectations that critical realists have maintained are reasonable of social science. That is, given that the social sciences’ object of inquiry is an open-system which prevents social researchers from isolating single explanatory factors, critical realists since Bhaskar (1975, 1979) have argued that the point of social scientific research must be to identify causal “mechanisms” (or generative causal structures, or whatever else you want to call them) that are plausibly responsible for the outcome observed, knowing full well that these conclusions are fallible. [ix] How does one come to such plausible-but-fallible conclusions? This brings us to theory.
Theory
On a straightforward Millian approach, theory could be used in one or both of two ways. First, it could be used for conceptual or epistemological purposes, for example, to develop criteria according to which knowledge of a crisis of the state can be based (see Skocpol 1979). Or, second, it could be used for propositional or empirical purposes, for example, to formulate the hypothesis that the timing of exposure to colonialism determines subsequent development outcomes (see Mahoney 2010) in order to be able to go out and collect empirical data usable for testing the hypothesis. [x] Yet, in his presentation, Mahoney stated that he preferred an approach that employs generalizations “to make sense of facts.” The precise meaning of this phrase remained unspecified. But, insofar as he was advocating the use of theory to comprehend causation–how causes (“to make sense of”) are related to empirical observations (“facts”)–Mahoney’s perspective appears to depart from a straightforward Millian approach, but still lacks a resolution to the question of the place of theory in this endeavor.
Again, critical realism seems to offer a solution to those pushing beyond the limits of the Millian method. For critical realism, theory is necessary to comprehend the ontological aspects of the outcome of interest. Critical realists often conceive of theory as a menu of causal factors (causal mechanisms, generative causal structures, tendencies, etc.), some subset of which may have been responsible for the outcome of interest in a given study. For critical realists, the point is to figure out which such factors were most-plausibly concatenated to produce the outcome of interest; this is where critical realism excels. Based as it is on a stratified ontology in which it is assumed that there are underlying (“real”) causes that are in certain instances inaccessible to direct empirical verification, critical realists have argued that social researchers should touch on what must have been the case for the observed outcome Y to have happened as it did. Doing so is what critical realists call “retroduction” or (borrowing from pragmatism) “abduction.” Mahoney didn’t suggest this approach; but it seems possible that retroduction could be of help to researchers who wish to best pursue the kind of social scientific inquiry that he advocated.
End notes
[i] Pioneered in the social sciences by Barrington Moore (1966), perfected by Theda Skocpol (1979), promulgated by Skocpol and Margaret Somers (1980), and then picked up by countless others, the Millian approach attained such dominance in the field of comparative-historical political social science that Jeffery Paige (1999:789ff) organized his review of the methodological and substantive literature according to various responses to the question, “what’s wrong with Theda Skocpol?”
[ii] Advocated as being central to the scientific method by the philosopher John Stuart Mill, the Millian approach is concerned with identifying the factors that are common and different in any given situation, from which we can build inferences about the necessary and/or sufficient conditions that explain an event. Accordingly, for Mill:
“If two or more instances of the phenomenon under investigation have only one circumstance in common, the circumstance in which alone all the instances agree, is the cause (or effect) of the given phenomenon” — John Stuart Mill, A System of Logic, Vol. 1. 1843. p. 454.
“If an instance in which the phenomenon under investigation occurs, and an instance in which it does not occur, have every circumstance save one in common, that one occurring only in the former; the circumstance in which alone the two instances differ, is the effect, or cause, or a necessary part of the cause, of the phenomenon. — John Stuart Mill, A System of Logic, Vol. 1. 1843. p. 455.
[iii] Skocpol (1979) famously argued that there were only two fundamental causes for social revolutions: (i) a rigid agrarian class structure and (ii) a crisis of the state. Having only three cases, she had two degrees of freedom; had she examined four cases, she may have concluded that another factor–perhaps ideology, industrialization, leadership, or some other factor–was an important cause.[iv] Titled “The Logic of Counterfactual Analysis in Historical Explanation,” the talk was delivered to the University of Michigan Department of Political Science on 15 April 2016.[v] Like Mahoney, I will not treat probabilistic causality separately in this post.[vi] In published work, Mahoney (2008) tries to reconcile INUS conceptions of causation with the necessary-and-sufficient logic of the Millian method, even though Mackie himself (1965:245-52) offered INUS conditions as a way to move beyond the shortcomings of Millian conceptions of causation.[vii] Ragin (1989) advocates something similar that is still within the variables paradigm, but a comparison between his approach and critical realism’s is beyond the scope of this post.[viii] My rendering of combined-and-conjoint causation may appear similar to one or more of Mahoney’s varieties of “within-case analysis” in which “analysts can . . . show that a relationship is causal despite the fact that a cross-case nominal comparison reveals one or more cases in which scores on the explanatory and outcome variables deviate from a general pattern” (2003:362). The difference between within-case analysis as Mahoney recommends and combined-and-conjoint causation as critical realists recommend has to do with how explanations themselves are conceptualized. For Mahoney and other (neo)Humeans, only empirical observations are considered causally important; for critical realists, the idea is that after identifying empirical patterns the researcher must go one step further by employing theory to comprehend the causal account (see below).[ix] Assuming research on the object of inquiry already exists, fallibilism implies that the point of new research is, in part, to identify explanations that are better than those previously offered in the literature (see Gorski 2004).[x] I am assuming–although this assumption is not at all uncontroversial–that the Millian approach is not simply and purely inductive.
References
Bhaskar, Roy. [1975] 2008. A Realist Theory of Science. London: Routledge
Bhaskar, Roy. [1979] 1998. The Possibility of Naturalism: A Philosophical Critique of the Contemporary Human Sciences. New York: Routledge
Gorski, Philip S. 2004. “The Poverty of Deductivism: A Constructive Realist Model of Sociological Explanation.” Sociological Methodology 34(1):1-33.
Gorski, Philip S. 2008. “The ECPRES Model: A Critical Realist Approach to Causal Mechanisms in the Social Sciences.“ pp. 147-194 in Björn Wittrock and Peter Hedström eds., The Frontiers of Sociology, Leiden: Brill, 2008.
Mackie, J. L. 1965. “Causes and Conditions.” American Philosophical Quarterly 2(4):245-264.
Mahoney, James. 2008. “Toward a Unified Theory of Causality.” Comparative Political Studies 41(4-5):412-436.
Mahoney, James. 2010. Colonialism and Postcolonial Development: Spanish America in Comparative Perspective. Cambridge and New York: Cambridge University Press.
Mahoney, James and Rodrigo Barrenechea. 2016. “The Logic of Counterfactual Analysis in Historical Explanation.” Unpublished Paper presented to the University of Michigan Department of Political Science, 15 April.
Moore, Jr., Barrington. 1966. Social Origins of Dictatorship and Democracy: Lord and Peasant in the Making of the Modern World. Boston: Beacon Press.
Paige, Jeffrey M. 1999. “Conjuncture, Comparison, and Conditional Theory in Macrosocial Inquiry.” American Journal of Sociology 105(3):781-800.
Ragin, Charles C. 1989. The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. Berkeley ; London: University of California Press.
Skocpol, Theda. 1979. States and Social Revolutions: A Comparative Analysis of France, Russia, and China. Cambridge and New York: Cambridge University Press.
Skocpol, Theda and Margaret Somers. 1980. “The Uses of Comparative History in Macrosocial Inquiry.” Comparative Studies in Society and History 22(2):174-197.
Steinmetz, George. 1998. “Critical Realism and Historical Sociology. A Review Article.” Comparative Studies in Society and History 40(1):170-186.
Steinmetz, George. 2003. “Odious Comparisons: Incommensurability, the Case Study, and “Small N’s in Sociology.” Sociological Theory 22(3): 371-400.
Steinmetz, George. 2014. “Comparative History and its Critics: A Genealogy and a Possible Solution.” Pp. 412-436 in A Companion to Global Historical Thought, edited by Prasenjit Duara, Viren Murthy and Andrew Sartori. London: Blackwell..
[/et_pb_text][/et_pb_column][/et_pb_row][/et_pb_section]