Causal mechanisms

The central tenet of causal realism is a thesis about causal mechanisms or causal powers. We can only assert that there is a causal relationship between X and Y if we can offer a credible hypothesis of the sort of underlying mechanism that might connect X to the occurrence of Y. The sociologist Mats Ekström puts the view this way: “the essence of causal analysis is ... the elucidation of the processes that generate the objects, events, and actions we seek to explain” (Ekstrom 1992, p. 115). Authors who have urged the centrality of causal mechanisms for both explanatory and purposes include Nancy Cartwright (Nature's Capacities and Their Measurements), Jon Elster (Explaining Social Behavior: More Nuts and Bolts for the Social Sciences), Rom Harré (Causal Powers), and Wesley Salmon (Scientific Explanation and the Causal Structure of the World). (Hedstrom and Swedberg's collection, Social Mechanisms: An Analytical Approach to Social Theory, is a useful source. An important advocate for a realist interpretation of science is Roy Bhaskar's A Realist Theory of Science.)

Nancy Cartwright is one of the most original voices within contemporary philosophy of science. Cartwright places real causal mechanisms at the center of her account of scientific knowledge. As she and John Dupré put the point, “things and events have causal capacities: in virtue of the properties they possess, they have the power to bring about other events or states” (Dupré and Cartwright 1988). Cartwright argues, for the natural sciences, that the concept of a real causal connection among a set of events is more fundamental than the concept of a law of nature. And most fundamentally, she argues that identifying causal relations requires substantive theories of the causal powers (capacities, in her language) that govern the entities in question. Causal relations cannot be directly inferred from facts about association among variables. As she puts the point, “No reduction of generic causation to regularities is possible” (Nature's Capacities and Their Measurements, p. 90). The importance of this idea for sociological research is profound; it confirms the notion shared by many researchers that attribution of social causation depends inherently on the formulation of good, middle-level theories about the real causal properties of various social forces and entities.

What is a causal mechanism? Consider this formulation: a causal mechanism is a sequence of events, conditions, and processes leading from the explanans to the explanandum (Varieties Of Social Explanation, p. 15). A causal relation exists between X and Y if and only if there is a set of causal mechanisms that connect X to Y. This is an ontological premise, asserting that causal mechanisms are real and are the legitimate object of scientific investigation.

Aage Sørensen summarizes a causal realist position for sociology in these words: “Sociological ideas are best reintroduced into quantitative sociological research by focusing on specifying the mechanisms by which change is brought about in social processes” (Sørensen 1998, p. 264). He argues that sociology requires better integration of theory and evidence. Central to an adequate explanatory theory, however, is the specification of the mechanism that is hypothesized to underlie a given set of observations. “Developing theoretical ideas about social processes is to specify some concept of what brings about a certain outcome—a change in political regimes, a new job, an increase in corporate performance, … The development of the conceptualization of change amounts to proposing a mechanism for a social process” (239-240). Sørensen makes the critical point that one cannot select a statistical model for analysis of a set of data without first asking the question, what in the nature of the mechanisms we wish to postulate to link the influences of some variables with others? Rather, it is necessary to have a hypothesis of the mechanisms that link the variables before we can arrive at a justified estimate of the relative importance of the causal variables in bringing about the outcome.

The general nature of the mechanisms that underlie sociological causation has been very much the subject of debate. Two broad approaches may be identified: agent-based models and social influence models. The former follow the strategy of aggregating the results of individual-level choices into macro-level outcomes; the latter attempt to identify the factors that work behind the backs of agents to influence their choices. (Sørensen refers to these as “pull” and “push” models; Sørensen, 1998.) Thomas Schelling’s apt title Micromotives and Macrobehavior captures the logic of the former approach, and his work profoundly illustrates the sometimes highly unpredictable results of the interactions of locally rational behavior. Jon Elster has also shed light on the ways in which the tools of rational choice theory support the construction of largescale sociological explanations (The Cement of Society: A Survey of Social Order). The second approach (the “push” approach) attempts to identify socially salient influences such as race, gender, educational status, and to provide detailed accounts of how these factors influence or constrain individual trajectories—thereby affecting sociological outcomes.

Emphasis on causal mechanisms for adequate social explanation has several salutary effects on sociological method. It takes us away from uncritical reliance on uncritical statistical models. But it also may take us away from excessive emphasis on large-scale classification of events into revolutions, democracies, or religions, and toward more specific analysis of the processes and features that serve to discriminate among instances of large social categories. Charles Tilly emphasizes this point in his arguments for causal narratives in comparative sociology (Tilly 1995). He writes, “I am arguing that regularities in political life are very broad, indeed transhistorical, but do not operate in the form of recurrent structures and processes at a large scale. They consist of recurrent causes which in different circumstances and sequences compound into highly variable but nonetheless explicable effects” (Tilly 1995, p. 1601).

Citations

1. Dupré, John, and Nancy Cartwright. 1988. Probability and Causality: Why Hume and Indeterminism Don't Mix. Nous 22:521-536.
2. Ekstrom, Mats. 1992. Causal explanation of social action: The Contribution of Max Weber and of Critical Realism to a Generative View of Causal Explanation in the Social Sciences. Acta Sociologica 35 (2):107(16).
3. Sørensen, Aage B. 1998. Theoretical mechanisms and the empirical study of social processes. In Social Mechanisms: An Analytical Approach to Social Theory, edited by P. Hedström and R. Swedberg.
4. Tilly, Charles. 1995. To Explain Political Processes. American Journal of Sociology.

What social science can do

Quite a few postings here emphasize the limits of social science knowledge. Prediction of the behavior of large social wholes is difficult to impossible. There are few strong regularities among social phenomena. Social entities and processes are heterogeneous, plastic, and path-dependent. So the question arises: what can the social sciences do that takes them beyond the realm of description and reportage of the blooming, buzzing confusion of social comings and goings, to something that is more explanatory and generalizable?

I think there is an answer to this, and it has to do with identifying mid-level mechanisms and processes that recur in roughly similar ways in a range of different social settings. The social sciences can identify a fairly large number of these sorts of recurring mechanisms. For example --

• public goods problems
• political entrepreneurship
• principal-agent problems
• features of ethnic or religious group mobilization
• market mechanisms and failures
• rent-seeking behavior
• the social psychology associated with small groups
• the moral emotions of family and kinship
• the dynamics of a transport network
• the communications characteristics of medium-size social networks
• the psychology and circumstances of solidarity

Further, the social sciences can attempt to discover the circumstances at the level of individual agents that make these mechanisms robust across social settings. They can model the dynamics and features of aggregation that they possess. And they can attempt to discover the workings of such mechanisms in particular social and historical settings, and work towards explanations of particular features of these events based on their theories of the properties of the mechanisms. Finally, they can attempt to find rigorous ways of attempting to model the effects of aggregating multiple mechanisms in a particular setting.

What this comes down to is the view that the main theoretical and generalizing contribution that the social sciences can make is the discovery and analysis of a wise range of recurring social mechanisms grounded in features of human agency and common institutional and material settings. They can help to constitute a rich tool box for social explanation. And, in a weak and fallible way, they can lay the basis for some limited social generalizations -- for example, "In circumstances where a group of independent individuals make private decisions about their actions, the public goods shared by the group will be under-provided."

This approach affords a degree of explanatory capacity and generalization to the social sciences. What it does not underwrite is the ability to offer general, comprehensive theories about any complex kind of phenomenon -- cities, schools, revolutions. And it does not provide a foundation for confidence about large predictions about the future behavior of complex social wholes.

Safety as a social effect

Some organizations pose large safety issues for the public because of the technologies and processes they encompass. Industrial factories, chemical and nuclear plants, farms, mines, and aviation all represent sectors where safety issues are critically important because of the inherent risks of the processes they involve. However, "safety" is not primarily a technological characteristic; instead, it is an aggregate outcome that depends as much on the social organization and management of the processes involved as it does on the technologies they employ. (See an earlier posting on technology failure.)

We can define safety by relating it to the concept of "harmful incident". A harmful incident is an occurrence that leads to injury or death of one or more persons. Safety is a relative concept, in that it involves analysis and comparison of the frequencies of harmful incidents relative to some measure of the volume of activity. If the claim is made that interstate highways are safer than county roads, this amounts to the assertion that there are fewer accidents per vehicle-mile on the former than the latter. If it is held that commercial aviation is safer than automobile transportation, this amounts to the claim that there are fewer harms per passenger-mile in air travel than auto travel. And if it is observed that the computer assembly industry is safer than the mining industry, this can be understood to mean that there are fewer harms per person-day in the one sector than the other. (We might give a parallel analysis of the concept of a healthy workplace.)

This analysis highlights two dimensions of industrial safety: the inherent capacity for creating harms associated with the technology and processes in use (heavy machinery, blasting, and uncertain tunnel stability in mining, in contrast to a computer and a red pencil on the editorial offices of a newspaper), and the processes and systems that are in place to guard against harm. The first set of factors is roughly "technological," while the second set is social and organizational.

Variations in safety records across industries and across sites within a given industry provide an excellent tool for analyzing the effects of various institutional arrangements. It is often possible to pinpoint a crucial difference in organization -- supervision, training, internal procedures, inspection protocols, etc. -- that can account for a high accident rate in one factory and a low rate in an otherwise similar factory in a different state.

One of the most important findings of safety engineering is that organization and culture play critical roles in enhancing the safety characteristics of a given activity -- that is to say, safety is strongly influenced by social factors that define and organize the behaviors of workers, users, or managers. (See Charles Perrow, Normal Accidents: Living with High-Risk Technologies and Nancy Leveson, Safeware: System Safety and Computers, for a couple of excellent treatments of the sociological dimensions of safety.)

This isn't to say that only social factors can influence safety performance within an activity or industry. In fact, a central effort by safety engineers involves modifying the technology or process so as to remove the source of harm completely -- what we might call "passive" safety. So, for example, if it is possible to design a nuclear reactor in such a way that a loss of coolant leads automatically to shutdown of the fission reaction, then we have designed out of the system the possibility of catastrophic meltdown and escape of radioactive material. This might be called "design for soft landings".

However, most safety experts agree that the social and organizational characteristics of the dangerous activity are the most common causes of bad safety performance. Poor supervision and inspection of maintenance operations leads to mechanical failures, potentially harming workers or the public. A workplace culture that discourages disclosure of unsafe conditions makes the likelihood of accidental harm much greater. A communications system that permits ambiguous or unclear messages to occur can lead to air crashes and wrong-site surgeries.

This brings us at last to the point of this posting: the observation that safety data in a variety of industries and locations permit us to probe organizational features and their effects with quite a bit of precision. This is a place where institutions and organizations make a big difference in observable outcomes; safety is a consequence of a specific combination of technology, behaviors, and organizational practices. This is a good opportunity for combining comparative and statistical research methods in support of causal inquiry, and it invites us to probe for the social mechanisms that underlie the patterns of high or low safety performance that we discover.

Consider one example. Suppose we are interested in discovering some of the determinants of safety records in deep mining operations. We might approach the question from several points of view.

• We might select five mines with "best in class" safety records and compare them in detail with five "worst in class" mines. Are there organizational or techology features that distinguish the cases?
• We might do the large-N version of this study: examine a sample of mines from "best in class" and "worst in class" and test whether there are observed features that explain the differences in safety records. (For example, we may find that 75% of the former group but only 10% of the latter group are subject to frequent unannounced safety inspection. This supports the notion that inspections enhance safety.)
• We might compare national records for mine safety--say, Poland and Britain. We might then attempt to identify the general characteristics that describe mines in the two countries and attempt to explain observed differences in safety records on the basis of these characteristics. Possible candidates might include degree of regulatory authority, capital investment per mine, workers per mine, ...
• We might form a hypothesis about a factor that should be expected to enhance safety -- a company-endorsed safety education program, let's say -- and then randomly assign a group of mines to "treated" and "untreated" groups and compare safety records. (This is a quasi-experiment; see an earlier posting for a discussion of this mode of reasoning.) If we find that the treated group differs significantly in average safety performance, this supports the claim that the treatment is causally relevant to the safety outcome.

Investigations along these lines can establish an empirical basis for judging that one or more organizational features A, B, C have consequences for safety performance. In order to be confident in these judgments, however, we need to supplement the empirical analysis with a theory of the mechanisms through which features like A, B, C influence behavior in such a way as to make accidents more or less likely.

Safety, then, seems to be a good area of investigation for researchers within the general framework of the new institutionalism, because the effects of institutional and organizational differences emerge as observable differences in the rates of accidents in comparable industrial settings. (See Mary Brinton and Victor Nee, The New Institutionalism in Sociology, for a collection of essays on this approach.)

Heterogeneity of the social

I think heterogeneity is a very basic characteristic of the domain of the social. And I think this makes a big difference for how we should attempt to study the social world "scientifically". What sorts of things am I thinking about here?

Let's start with some semantics. A heterogeneous group of things is the contrary of a homogeneous group, and we can define homogeneity as "a group of fundamentally similar units or samples". A homogeneous body may consist of a group of units with identical properties, or it may be a smooth mixture of different things, consisting of a similar composition at many levels of scale. A fruitcake is non-homogeneous, in that distinct volumes may include just cake or a mix of cake and dried cherries, or cake and the occasional walnut. The properties of fruitcake depend on which sample we encounter. A well mixed volume of oil and vinegar, by contrast, is homogeneous in a specific sense: the properties of each sample volume are the same as any other. The basic claim about the heterogeneity of the social comes down to this: at many levels of scale we continue to find a diversity of social things and processes at work. Society is more similar to fruitcake than cheesecake.

Heterogeneity makes a difference because one of the central goals of positivist science is to discover strong regularities among classes of phenomena, and regularities appear to presuppose homogeneity of the things over which the regularities are thought to obtain. So to observe that social phenomena are deeply heterogeneous at many levels of scale, is to cast fundamental doubt on the goal of discovering strong social regularities.

Let's consider some of the forms of heterogeneity that the social world illustrates.

First is the heterogeneity of social causes and influences. Social events are commonly the result of a variety of different kinds of causes that come together in highly contingent conjunctions. A revolution may be caused by a protracted drought, a harsh system of land tenure, a new ideology of peasant solidarity, a communications system that conveys messages to the rural poor, and an unexpected spar within the rulers -- all coming together at a moment in time. And this range of causal factors, in turn, shows up in the background of a very heterogeneous set of effects. (A transportation network, for example, may play a causal role in the occurrence of an epidemic, the spread of radical ideas, and a long, slow process of urban settlement.) The causes of an event are a mixed group of dissimilar influences with different dynamics and temporalities, and the effects of a given causal factor are also a mixed and dissimilar group.

Second is the heterogeneity that can be discovered within social categories of things -- cities, religions, electoral democracies, social movements. Think of the diversity within Islam documented so well by Clifford Geertz (Islam Observed: Religious Development in Morocco and Indonesia); the diversity at multiple levels that exists among great cities like Beijing, New York, Geneva, and Rio (institutions, demography, ethnic groups, economic characteristics, administrative roles, ...); the institutional variety that exists in the electoral democracies of India, France, and Argentina; or the wild diversity across the social movements of the right.

Third is the heterogeneity that can be discovered across and within social groups. It is not the case that all Kansans think alike -- and this is true for whatever descriptors we might choose in order to achieve greater homogeneity (evangelical Kansans, urban evangelical Kansans, ...). There are always interesting gradients within any social group. Likewise, there is great variation in the nature of ordinary, lived experience -- for middle-class French families celebrating quatorze Juillet, for Californians celebrating July 4, and for Brazilians enjoying Dia da Independência on September 7.

A fourth form of heterogeneity takes us within the agent herself, when we note the variety of motives, moral frameworks, emotions, and modes of agency on the basis of which people act. This is one of the weaknesses of doctrinaire rational choice theory or dogmatic Marxism, the analytical assumption of a single dimension of motivation and reasoning. Instead, it is visible that one person acts for a variety of motives at a given time, persons shift their motives over time, and members of groups differ in terms of their motivational structure as well. So there is heterogeneity of motives and agency within the agent.

These dimensions of heterogeneity make the point: the social world is an ensemble, a dynamic mixture, and an ongoing interaction of forces, agents, structures, and mentalities. Social outcomes emerge from this heterogeneous and dynamic mixture, and the quest for general laws is deeply quixotic.

Where does the heterogeneity principle take us? It suggests an explanatory strategy: instead of looking for laws of whole categories of events and things, rather than searching for simple answers to questions like "why do revolutions occur?", we might instead look to a "concatenation" strategy. That is, we might simply acknowledge the fact of molar heterogeneity and look instead for some of the different processes and things in play in a given item of interest, and the build up a theory of the whole as a concatenation of the particulars of the parts.

Significantly, this strategy takes us to several fruitful ideas that already have some currency.

First is the idea of looking for microfoundations for observed social processes; (Microfoundations, Methods, and Causation: On the Philosophy of the Social Sciences). Here the idea is that higher-level social processes, causes, and events, need to be placed within the context of an account of the agent-level institutions and circumstances that convey those processes.

Second is the method of causal mechanisms advocated by McAdam, Tarrow, and Tilly, and discussed frequently here (Dynamics of Contention (Cambridge Studies in Contentious Politics)). Put simply, the approach recommends that we explain an outcome as the contingent result of the concatenation of a set of independent causal mechanisms (escalation, intra-group competition, repression, ...).

And third is the theory of "assemblages", recommended by Nick from accursedshare and derived from some of the theories of Gilles Deleuze. (Manuel Delanda describes this theory in A New Philosophy of Society: Assemblage Theory And Social Complexity.)

Each of these ideas gives expression to the important truth of the heterogeneity principle: that social outcomes are the aggregate result of a number of lower-level processes and institutions that give rise to them, and that social outcomes are contingent results of interaction and concatenation of these lower-level processes.

Quasi-experimental data?

Stan Lieberson is one of a group of sociologists for whom I have great respect when it comes to intelligent thinking about social science methodology. His 1985 book, Making It Count: The Improvement of Social Research and Theory, is a good example of some of this thinking about the foundations of social science knowledge, and I also admire A Matter of Taste: How Names, Fashions, and Culture Change in the way it offers a genuinely novel topic and method of approach.

Lieberson urges us to consider "a different way of thinking about the rigorous study of society implied by the phrase 'science of society'" instead of simply assuming that social science should resemble natural science (3-4). His particular object of criticism in this book is the tendency of quantitative social scientists to use the logic of experiments to characterize the data they study.

An experiment is an attempt to measure the causal effects of one factor X on another factor Z by isolating a domain of phenomena -- holding constant all other causal factors -- and systematically varying one causal factor to observe the effect this factor has on an outcome of interest. The basic assumption is that an outcome is the joint effect of a set of (as yet unknown) causal conditions:

C1 & C2 & ... & Cn cause Z,

where we do not yet know the contents of the list Ci. We consider the hypothesis that Cm is one of the causes of Z. We design an experimental environment in which we are able to hold constant all the potentially relevant causal conditions we can think of (thereby holding fixed Ci), and we systematically vary the presence or absence of Cm and observe the state of the outcome Z. If Z varies appropriately with the presence or absence of Cm, we tentatively conclude that Cm is one of the causes of Z.

In cases where individual differences among samples or subjects may affect the outcome, or where the causal processes in question are probabilistic rather than deterministic, experimentation requires treating populations rather than individuals and assuring randomization of subjects across "treatment" and "no-treatment" groups. This involves selecting a number of subjects, randomly assigning them to controlled conditions in which all other potential causal factors are held constant, exposing one set of subjects to the treatment X while withholding the treatment from the other group, and measuring the outcome variable in the two groups. If there is a significant difference in the mean value of the outcome variable between the treatment group and the control group, then we can tentatively conclude that X causes Z and perhaps estimate the magnitude of the effect. Take tomato yields per square meter (Z) as affected by fertilizer X: plants in the control group are subjected to a standard set of growing conditions, while the treatment group receives these conditions plus the measured dose of X. We then measure the quantity produced by the two plots and estimate the effect of X. The key ideas here are causal powers, random assignment, control, and single-factor treatment.

However, Lieberson insists that most social data are not collected under experimental conditions. It is normally not possible to randomly assign individuals to groups and then observe the effects of interventions. Likewise, it is not possible to systematically control the factors that are present or absent for different groups of subjects. If we want to know whether "presence of hate speech on radio broadcasts" causes "situations of ethnic conflict" to progress to "situations of ethnic violence" -- we don't have the option of identifying a treatment group and a control group of current situations of ethnic conflict, and then examine whether the treatment with "hate speech on radio broadcasts" increases the incidence of ethnic violence in the treatment group relative to the control group. And it is fallacious to reason about non-experimental data using the assumptions developed for analysis of experiments. This fallacy involves making "assumptions that appear to be matters of convenience but in reality generate analyses that are completely off the mark" (6).

Suppose we want to investigate whether being a student athlete affects academic performance in college. In order to treat this topic experimentally we would need to select a random group of newly admitted students; randomly assign one group of individuals to athletic programs and the other group to a non-athletic regime; and measure the academic performance of each individual after a period of time. Let's say that GPA is the performance measure and that we find that the athlete group has a mean GPA of 3.1 while the non-athlete group has an average of 2.8. This would be an experimental confirmation of the hypothesis that "participation in athletics improves academic performance."

However, this thought experiment demonstrates the common problem about social data: it is not possible to perform this experiment. Rather, students decide for themselves whether they want to compete in athletics, and their individual characteristics will determine whether they will succeed. Instead, we have to work with the social realities that exist; and this means identifying a group of students who have chosen to participate in athletics; comparing them with a "comparable" group of students who have chosen not to participate in athletics; and measuring the academic performance of the two groups. But here we have to confront two crucial problems: selectivity and the logic of "controlling" for extraneous factors.

Selectivity comes in when we consider that the same factors that lead a college student to participate in athletics may also influence his/her academic performance; so measuring the difference between the two groups may only measure the effects of this selective difference between membership in the groups -- not the effect of the experience of participating in athletics on academic performance. In order to correct for selectivity, the researcher may attempt to control for potentially influential differences between the two groups; so he/she may attempt to control for family factors, socio-economic status, performance in secondary school, and a set of psycho-social variables. "Controlling" in this context means selecting sub-groups within the two populations that are statistically similar with respect to the variables to be controlled for. Group A and Group B have approximately the same distribution of family characteristics, parental income, and high school GPA; the individuals in the two groups are "substantially similar". We have "controlled" for these potentially relevant causal factors -- so any observed differences between academic performance across the two groups can be attributed to the treatment, "participation in athletics."

But Lieberson makes a critical point about this approach: there is commonly unmeasured selectivity within the control variables themselves -- crudely, students with the same family characteristics, parental income, and high school GPA who have selected athletics may nonetheless be different from those who have not selected athletics, in ways that influence academic performance. As Lieberson puts the point, "quasi-experimental research almost inevitably runs into a profound selectivity issue" (41).

There is lots more careful, rigorous analysis of social-science reasoning in the book. Lieberson crosses over between statistical methodology and philosophy of social science in a very useful way, and what is most fundamental is his insistence that we need to substantially rethink the assumptions we make in assigning causal influence on the basis of social variation.

More on what can be explained

A previous posting argued that most social facts don't admit of social explanation because they are too fundamentally conjunctural or too boringly ordinary. Let's extend this thought by considering what sorts of social facts do admit of explanation.

One obvious category is the example of a perplexing mid-range social regularity. Why do used cars usually sell for less than their "real" value? Because of the asymmetry of information between buyer and seller (the market for lemons). Why does ethnic conflict turn violent more commonly in circumstances where the institutions of civil society are weak? Because weak civil institutions undermine trust between distinct intermixed groups. (Here is a posting summarizing some recent thinking on this connection between civil society and ethnic violence.) Why do collectivized farms usually witness lower labor productivity than privately owned farms? Because of some common features of labor management and supervision that are usually a part of collective farm practices but not of private farm practices, that are likely to result in individual effort that is of lower quality or intensity than the private alternative. (This is sometimes referred to as the "easy-rider" problem.)

This category encompasses cases of non-trivial regularity. These are all examples of what I call "phenomenal" social regularities. They are not manifestations of some underlying set of social laws, but rather the common results of a set of mechanisms or processes in a range of cases (article). In each case the explanation proceeds by identfying a common but non-obvious mechanism or structural feature that produces the observed outcome.

Another important category of social phenomena demanding explanation is the large, complex social occurrence. (William Sewell calls these "events" -- historical occurrences that are "irreversible, contingent, uneven, discontinuous and transformational" (link). ) Here I have in mind things such as the great Pullman strike of 1894, the defeat of France in the Franco-Prussian War in 1870, the civil war in Lebanon in the 1980s, the Rwandan genocide, or the selection of alternating over direct current electricity transmission systems. In each case we want to know why the event occurred -- the event is important and obscure -- and it is credible that there may be a small number of social mechanisms and circumstances that can be discovered and that brought about the event.

So explanation of these singular but extended events takes the form of a causal narrative that turns on a small number of important causal factors or mechanisms. The burden of explanation is to discover the mid-level social processes and mechanisms that caused the outcome to occur. And a feature of generalization comes into this account as well, but at a different point in the story: to say that P caused O in the circumstances is also to imply that P would have similar effects in similar circumstances in other settings.

So the idea of a social regularity comes into this discussion twice. First, we are often led to ask for an explanation when we observe a curious regularity -- instances of similar behaviors or outcomes in separate cases. "Why does this weak regularity obtain across independent cases?" And second, the type of explanation highlighted here is a causal explanation, which implies assertions about counterfactual regularities. "The outcome occurs because of the regular causal powers of such-and-so a causal mechanism." A factor that possesses the causal power to help bring about a certain kind of effect plainly plays a role in statements like this: "whenever P occurs in circumstances substantially similar to C, O is likely to occur." And this is a statement of an idealized regularity. This in turn lends some support for the idea that explanation and the discovery of a level of social generalization are linked -- but not in the way that the nomological-deductive model of explanation would imply.

(See Varieties of Social Explanation for other perspectives on these questions.)

Social science history and historical social science

Social science methods and historical explanation seem to come together in several different ways; what can we say about the differences of approach between “history using the tools of the social sciences” and “social science research that pays close attention to history”?

E. P. Thompson treats the making of the English working class. His work is multi-faceted. He gives treatment of workingmen’s organizations and publications; churches and pastors; riots and chants; petitions to parliament; and much else. The story is historical in several respects: it provides an account of change over time and it engages in detailed and fine-grained description of specific circumstances in the past. Is Thompson attempting to explain something? Perhaps it is more accurate to say that his aim is to describe this extended, multi-location, multi-group process of “making”, along with some sense of the circumstances and features of agency that brought this “class” into being. And he goes out of his way to emphasize the contingency of the story that he tells: this “class” could have taken a very different shape, depending on altered circumstances and agency along the way. His is as much like the work of a biographer, detailing the development of personality, the contingencies of personal history, the formation of character, and the actions of the mature person.

Charles Tilly treats the development of contentious politics in France over three centuries. His account too is “historical”: it describes the development and diversity of contentious politics in France through revolution and periods of quiet. His account too is attentive to difference; he emphasizes the many ways in which French contentious “underclass” politics varied across time and across region. The politics of workers in Paris were quite different from those of the winemakers of the Vendée. But Tilly’s account is deliberately sociological and theoretical. The goal of his study is to discover causes; to test a few theoretical hypotheses about mobilization; and to use the “data” of French working class history as a basis for testing and evaluating sociological theory.

Each of these examples is a major intellectual contribution; each contributes to our historical understanding; each focuses on a historically situated working class. But the two oeuvres have substantial differences of orientation and feel. One is explicitly theoretical in its goals; the other is nuanced and descriptive. One aims at arriving at explanations; the other is interested in providing a qualitative understanding of the experience of ordinary men and women of the 18th and 19th centuries in rural England. One is historical social science, while the other is social science history.

So it is an important question within the philosophy of history, to articulate the difference between these two configurations of “social science” and “history.” How are the two genres distinguished? Are they differences of style, each embodying a complex of narrative and explanatory values? Are they at opposing ends of some sort of spectrum, ranging from descriptive to explanatory or concrete to abstract? Or are they actually logically different in some way—perhaps along the lines of the distinction between three conceptions of time described by William Sewell?

Perhaps most extremely, would we be right to consider excluding Tilly’s work from the domain of the “historical” and place it instead within the domain of social science, distinguished from other varieties of social science primarily by the fact that the data upon which it depends are facts about the past? In other words, is it possible to suggest that “historical social science” is not a variety of historical writing at all?

How might we characterize some of the differences between these two bodies of writing about the past? Do they constitute different paradigms, research frameworks, or forms of historical practice? Do they embody different complexes of assumptions about what to emphasize, what the standards of rigor are, what is required by way of description, detail, and fact; what is intended by way of explanation and understanding; the role that interpretation of the lived experience of agents plays; and so on?

Comparative historical social science is a particular instance of historical social science. There is a well-developed contemporary literature on the conceptual and methodological issues raised by comparative historical social science. And the participants in this literature generally seem to come down on the side of the “social science” conclusions rather than the “historical explanations” side of the debate. The goal of comparative social science is to assess causation, and to use knowledge of concrete historical cases as a source of evidence for evaluating causal theories. Examples include the explanation of social revolution (Theda Skocpol), the explanation of social contention (Charles Tilly), the explanation of economic development (R. Bin Wong, Philip Huang), the explanation of labor union politics (Howard Kimmeldorf).

Now let us turn the lens in the other direction and ask, in what ways do the contents of social science knowledge aid in the construction of historical knowledge? What is the role of theory and causal hypothesis in paradigm examples of historical knowledge? Virtually all historians would first insist: “Historical research cannot take the form of application of social science theory to the data. Rather, the historian’s task is to discover the particular and the grain of the materials in front of him. History is not the unfolding of theoretical premises and good historical knowledge does not result from deducing consequences from general social science theories.” That being conceded: are there forms of historical inquiry and knowledge that are importantly and rationally assisted by social science theory?

One variant of historical writing where social science theory is apparently pertinent is in the “causal narrative”. Historians are well served by appealing to social science theories of causal mechanisms in order to explain the transitions that they identify in their causal narratives. This is a logical point. And yet, it is strikingly difficult to find examples of leading historians who make use of social science theory in this way. Philip Huang is an example of a professional historian who makes substantial use of social science theory and concepts; Simon Schama is an example of a historian who is averse to this use. More commonly, the authors who provide causal narratives informed by social science theory are themselves sociologists or other social scientists (Skocpol, Tilly, Wolf, Paige).

It seems from some of these scattered observations, that there is indeed a significant difference between social science history and historical social science. The explanatory goals appear to be different, and the methods of reasoning and standards of rigor and adequacy seem to be distinct as well. So the question of how the disciplinary differences fit together is one that demands continued scrutiny.

Are there patterns of economic development?

There is an old-fashioned and discredited theory that holds that there are only a small number of development trajectories. Crudely, Western Europe's experience -- agricultural modernization, handicraft manufacture, population growth, urbanization, and large-scale mass manufacturing -- is the paradigm and "normal" case, and different processes in other countries are deviations or abnormalities. This is the approach economic historians once took towards Asian economic development; it is substantially refuted by Bin Wong (China Transformed: Historical Change and the Limits of European Experience) and Ken Pomeranz (The Great Divergence: China, Europe, and the Making of the Modern World Economy.).

A somewhat better approach postulates that there are alternative pathways of development, and that English, Italian, Indian, Chinese, and Brazilian historical experiences of development all illustrate different trajectories. Charles Sabel and Jonathan Zeitlin explore this idea (World of Possibilities: Flexibility and Mass Production in Western Industrialization). This approach emphasizes path dependence and the salience of institutions in economic development. Thus Robert Brenner maintains that it was differences in the particulars of the social-property relations governing farming that explained English transformation and French stagnation (The Brenner Debate: Agrarian Class Structure and Economic Development in Pre-industrial Europe; see also a short descriptive essay, The Brenner Debate).

But other historians have pushed contingency and variation even deeper. So Pomeranz argues against a nation-based model of development. He argues that China's processes of development were very different in different regions, north and south, east and west. So instead of analyzing "China," he picks out one large macro-region, the lower Yangzi region, as the unit possessing enough integration to possess a distinctive pattern of development. Essentially, this is to say that the complex of institutions, crops, population dynamics, and urban patterns are unified but distinct in north China and southeast China, and that each constitutes a system of production with its own dynamics. So this serves to disaggregate China into several important and different regions.

So, with all this disaggregation and differentiation of economic development, let's ask the question again: are there patterns of economic development? Or is every region, city, or state sui generis?

Here is what seems plausible to me. The best hope we have for generalizations about economic development is not at the level of wholes -- regions or nations. Rather, what we can hope to do is to discover a number of recurring processes and mechanisms -- political, demographic, technology, institutional, and economic -- that can be identified and studied in multiple historical cases. In this category of recurring processes and mechanisms, I would include "proto-industrialization," "scissors crisis," "high level equilibrium trap," "state fiscal crisis," and "rapid urban growth" -- along with dozens of other comparable social and economic processes. These are mid-level social processes and mechanisms that correspond to specific opportunities or situations of persons and groups in a developing society, and they can arguably occur in historically separate cases. And actors will adjust their behavior in relation to these processes in their particular settings, to pursue their goals. Finally, some of these processes will aggregate in particular historical settings -- often in novel ways -- to give rise to a particular historical trajectory. (Notice that this is methodologically very similar to the picture that McAdam, Tarrow and Tilly paint about the possibility of generalizations about contentious politics; Dynamics of Contention.)

Is sociology analogous to epidemiology?

Quantitative sociology attempts, among other things, to establish causal connections between large social factors (race, socio-economic status, residential status) and social outcomes of interest (rates of delinquency). Is this type of inquiry analogous in any way to the use of large disease databases to attempt to identify risk factors? In other words, is there a useful analogy between sociology and epidemiology?

Suppose that the divorce rate for all American men is 30%. Suppose the rate for New York City males with income greater than $200K is 60%. We might want to draw the inference that something about being a high-income male resident of New York causes a higher risk of divorce for these persons. And we might want to justify this inference by noticing that it is similar to a parallel statistical finding relating smoking to lung cancer. So sociology is similar to epidemiology. Certain factors can be demonstrated to cause an elevated risk of a certain kind of outcome. There are "risk factors" for social outcomes such as divorce, delinquency, or drug use.

Is this a valid analogy? I think it is not. Epidemiological reasoning depends upon one additional step -- a background set of assumptions about the ontology and etiology of disease. A given disease is a specific physiological condition within a complex system of cells and biochemical processes. We may assume that each of these physiological abnormalities is caused by some specific combination of external and internal factors through specific causal mechanisms. So the causal pathways of normal functioning are discrete and well-defined, and so are the mechanisms that cause disruption of these normal causal pathways. Within the framework of these guiding assumptions, the task of the statistics of epidemiology is to help sort out which factors are causally associated with the disease. The key, though, is that we can be confident that there is a small number of discrete causal mechanisms that link the factor to the disease.

The case is quite different in the social world. Social processes are not similar to physiological processes, and social outcomes are not similar to diseases. In each case the failure of parallel derives from the fact that there are not unique and physiologically specific causal systems at work. Cellular reproduction has a specific biochemistry. Cancerous reproduction is a specific deviation from these cellular processes. And specific physical circumstances cause these deviations.

Now think about the social world. A process like "urbanization" is not a homogeneous social process. Rather, it is a heterogeneous mix of social developments and events; and these components are different in different times and places. And outcomes that might be considered the social equivalent of disease -- a rising murder rate, for example -- is a also composite of many distinct social happenings and processes. So social systems and outcomes lack the simple, discrete causal uniformity that is a crucial part of epidemiological reasoning.

This is not to say that there are not underlying causal mechanisms whose workings bring about a sharp increase in, say, the population's murder rate. Rather, it is to say that there are numerous, heterogeneous and cross-cutting such mechanisms. So the resultant social outcome is simply the contingent residue of the multiple middle-level processes that were in play in the relevant time period. And the discovery that "X, Y, Z factors are correlated with a rise in the incidence of O" isn't causally irrelevant. But the effects of these factors must be understood as working through their influence on the many mid-level causal mechanisms.

Are there discrete social mechanisms?

McAdam, Tarrow and Tilly direct our attention to the level of the concrete social mechanisms that recur in many instances of social contention (Dynamics of Contention). They specifically refer to escalation, radicalization, brokerage, and repression as examples of social mechanisms that produce the same effects in the same circumstances, and that concatenate into historical processes and events. To this list I would add my own examples -- free-rider problems, norm diffusion, and communications networks.

I wholeheartedly endorse the idea that social explanations need to proceed on the basis of an analysis of underlying social mechanisms. But can this program be carried out in a Mendeleev sort of way -- try to discover a "table of elements" of causal mechanisms that aggregate into "molecules" of social contention?

The closer I look at the argument, the more concerned I become about the discreteness and elementality of the items MTT offer as examples. Take brokerage -- isn't this really an umbrella term that encompasses a number of different kinds of negotiation and alliance-formation? So brokerage isn't analogous to "expansion of ice during freezing" -- a clear example of a physical causal mechanism that is homogeneous across physical settings. Brokerage is rather a "family-resemblance" term that captures a number of different instances of collective behavior and agency.

If we find this line of thought somewhat persuasive, it suggests that we need to locate the causal connectedness among social settings at an even deeper micro-level. It is the situation of "agents with interests, identities, networks, allies, and repertoires" that constitutes the causal nexus of social causation on contention -- not a set of frozen mid-level groups of behaviors such as brokerage or radicalization. Instead, these mid-level concepts are descriptive terms that allow us to single out some broadly similar components of social contention.

Or in another vocabulary: the level at which we find real causal connections in the social world is the level of the socially situated and socially constituted individual in interaction with other individuals -- the perspective of methodological localism (Levels of the Social). This doesn't undermine causal realism -- but it does undermine the idea that there are meso-level "causal mechanisms" such as brokerage that really recur across instances.

Processes versus structures

Comparative historical sociology seeks to provide an answer to this type of question: what causes certain kinds of large historical outcomes? And it proceeds, often, on the basis of controlled comparison of a small number of cases. Theda Skocpol's classic book, States and Social Revolutions: A Comparative Analysis of France, Russia and China, is a good example of the approach. So far so good. But what kinds of causes do CHS researchers typically look for? The method of comparison is often described in terms of Mill's methods of similarity and difference. Find cases with variation in the outcome to be explained and similar/different causal conditions; and then seek out patterns of co-variation that suggest that certain factors are necessary and/or sufficient conditions for the outcome to be explained. These factors are then said to cause the outcome. (Mill's approach to social research is described in Fred Wilson's entry on Mill in the Stanford Encyclopedia of Philosophy.)

This way of formulating the approach has fairly strong ontological presuppositions. Basically, it assumes that social causes are large, pervasive factors that obtain or fail to obtain in the multiple cases. For example, in explaining revolution the investigator might identify food crisis, population density, weak state institutions, and war as potential causal factors, and then compare the cases with respect to the variance of these factors. The comparative method assumes that large social units (societies, regions, social groups) are the operative units, and their causal properties are largescale, pervasive social conditions.

But what if our view of social causation is focused at a more disaggregated level -- not at the level of gross social conditions and structures, but at the level of lower-level processes and mechanisms? What if we thought that the action is really taking place at the level of the contingent unfolding of social processes at more local levels? This ontology wouldn't imply that the large social factors just mentioned are not part of the true causal story. But it would cast serious doubt on the expectation that there will be neat patterns of covariance across cases identified solely at this level. And yet this is exactly what Mill's methods require.

The turn to concrete social mechanisms as the unit of social analysis suggests that it is most fruitful to seek out explanations of outcomes as the "concatenation" of a set of common social mechanisms (Social Mechanisms: An Analytical Approach to Social Theory). And this implies that the traditional comparative method is not likely to succeed; there won't be a neat pattern of co-variation at the level of macro-characteristics and structures. So what is the alternative?

We might say that a credible alternative, still falling within a broad definition of "comparative historical sociology", is this: select a number of cases for detailed study. Uncover in some detail the processes and factors that appear to have led to the outcome (process-tracing). And arrive at generalizations by discovering that certain mechanisms or processes recur across multiple cases, and that large structural factors interact with these processes in recurring ways.

This is the approach that McAdam, Tarrow, and Tilly take in Dynamics of Contention. And it would appear to me that this approach permits an appropriate marriage between social ontology and social science methodology. The methodology is suited to the ontological insight that social phenomena are composed of lower-level social mechanisms and processes, and the outcomes are the contingent and path-dependent result of the concatenation of these mechanisms. There are no "laws of revolution" (or war, or civil strife); rather, there are a large number of social mechanisms that can be observed in many instances of these large social outcomes. These mechanisms can be rigorously analyzed, and we can explain outcomes (for example, the success of the Bolshevik revolutionary strategy) as the result of a concatenation of various of these mechanisms.

How does regional economic development work?

Countries, states, regions, and cities are interested in stimulating economic development in their jurisdictions. Various possible strategies are often mentioned:

• encourage entrepreneurship
• improve the talent base
• enhance the attractiveness of the region to outsiders with creative talents
• create a legal, fiscal, and regulatory environment that encourages new businesses
• create larger pools of venture capital
• attract out-of-region businesses through regional business-attraction centers
• encourage research and development in local universities
• facilitate the movement of inventions from the lab to the business plan

The primary question here is a causal one: what strategies actually work, and how would we use empirical research to evaluate alternative policy interventions by governmental and non-governmental organizations? How would we decide which policies to invest in?

In order to begin to answer this question, we need to get a little more specific about what we mean by "regional economic development." I'll just stipulate an answer to this:

Regional economic development entails the creation of new businesses and expansion of existing businesses, in a way that expands the total number of jobs and results in a rising average wage.

So regional economic development aims at creating more employment and a rising standard of living in the region, and it seeks to do this through causing expansion of profitable business activity in the region. And, in order to create higher-paid jobs, the businesses created or expanded need to be on the high-value-added end of the spectrum; this often means skill- or knowledge-intensive industries. New jobs in low-productivity manufacturing or service businesses will not increase the average wage.

(The last statement isn't necessarily true in a region with persistent unemployment. A new low-productivity factory with a low average wage will raise the average standard of living if it significantly reduces unemployment. However, it seems perverse to have the goal of stimulating the growth of low-productivity sectors and businesses rather than higher-wage opportunities.)

There are several ideas that arise from these simple statements and questions. One concerns the limits that exist on the ability of state or local governments to actually influence the rate of growth of business activity, jobs, and wages at all. It is an open question whether a state or region that has done an optimal job of aligning its policies and investments will actually have a higher probability of growth in these outcomes than one that has no strategies, bad strategies, or poorly aligned strategies. This is because the decisions made by investors and entrepreneurs are influenced by many other factors besides the policy tools available to the municipal or state governments. And the success or failure of business choices is determined by events outside the policy arena. The best way of capturing this thought is to recognize that the causal influence of good policies may still be small relative to other non-policy factors.

Second, though, we might have strong theoretical reasons for thinking that one policy choice is likely to have larger effects on the desired outcome than another. If we can demonstrate a theoretical basis in economics or organizational theory, for example, for concluding that factors X, Y, and Z are favorable for producing the outcomes we want, even if they are not decisive, then it is logical that we should try to identify the most influential of these factors; design a coherent package of policies that enhance these factors (i.e. avoid combinations that are self-defeating), and make the effort possible to implement this package of policies. That is, there may be a basis in social science theory for judging that policy X should have a positive effect on the probability of desired outcome Y. The reasoning may be economic (businesses will have an incentive to locate in regions where they can have a high confidence of recruiting a qualified workforce) or perhaps logistical (businesses will choose locations where transport is convenient) or fiscal (firms will take careful account of the total cost of doing business in Michigan versus California or North Carolina). But if we can demonstrate on theoretical grounds that X should positively influence Y, then we have reason to implement X even if we don’t have direct empirical evidence of X’s effectiveness.

This is the role of theory in justifying the choice of a policy package: a set of antecedent theoretical reasons for believing that this set of policies will work to increase the outcomes we are interested in achieving. There is another more empirical basis for assessing alternative policies, based on comparison of cases: look at a number of cases in which X is present or not present, and measure the status of Y. From a social science point of view, this inquiry is feasible but difficult. If we select six cities for comparison and find that the cities with a high percentage of college grads have high growth while those with low college grad percentages have low growth -- does that demonstrate that "talent causes economic growth"? Not exactly; the observed correlation could be spurious, accidental, collateral, or reversed. And likewise with the large-N version of the study. Suppose we find either that Y is positively correlated with X or that Y is not correlated with X, based on the values of X and Y over a large number of cases. Does the first finding demonstrate that “X causes Y”, and does the second case demonstrate that “X does not cause Y”? Neither conclusion is justified on the basis of just these facts. The non-correlation finding may be the result of a genuine causal influence masked by a number of disturbing influences; and the positive correlation may be the result of a common cause that is influencing both X and Y. So, as econometricians and epidemiologists know very well, the design of empirical studies that sort out the causal impact of various treatments is challenging.

In addition to these issues having to do with our ability to assess the causal weight of various policy treatments, we also have to consider the ratio of costs and benefits associated with various bundles of policies. Policy makers are forced to arrive at some estimate of the relative costs and benefits of the available policy options in order to make sensible choices among them. But here too there are the same problems of estimation: it is difficult to estimate either the probability of success of a given policy or the net value of the success of the policy.

The upshot of this short reflection, if there is one, is that it is very difficult to arrive at solid, precise, and rigorous estimates of either the economic impact of various policy choices or of their likelihood of success. And yet the policymaker does not have the luxury of indecision; economic development is a crucially important component of the wellbeing of a region's population. So it seems unavoidable that the best we can do is to assess policies on the basis of their likely contribution to economic growth, based on available economic and social theories, and be ready to finetune our choices as experience indicates.

Micro cultures?

Is there such a thing as a "micro"-culture -- a culture that is somewhat distinctive of a particular community in a specific time and place, and different from the culture of similar communities in other places?

The sorts of communities I'm thinking of might include sports teams, university faculties, union locals, church congregations, street gangs, or rural villages. Is it possible that Village X is a friendlier place than Village Y? Or that Local W is more militant than Local Z? Or that Gang S is less willing to use violence against younger brothers of rivals than Gang T is? And is it possible that these differences are real, persistent, and self-reinforcing through mechanisms that constitute a basis for dynamic cultural continuity?

Take the Boston Celtics. The players and coaches are all recruited in a national and international labor market; they generally are well acquainted with each other through their high-profile college basketball careers, camps, and leagues; and the NBA spends a lot of money in cultivating a national NBA culture rather than a micro-culture for particular teams. So you might imagine that every team would be simply a generic expression of national basketball culture. And yet it's possible that the Boston Celtics are a distinctive team, in terms of a number of characteristics. They might have more of a team ethic than, say, the Los Angeles Lakers; they might have some traditions that play out in practices and games; they might have a lore about the Celtics greats of the past -- Havliceck, Bird, Dennis Johnson, and they might conceivably have a style of play that persists through changes of coaches and players. So here the question is whether the national culture of NBA basketball is the key, or whether there are important differences in behavior, values, practices, and style that persist for specific teams over time. This speculation raises two sorts of questions: How would we attempt to determine whether there are such differences? And what sorts of social mechanisms would serve to preserve and reproduce such cultural specifics?

The basketball version of this story is pretty speculative on my part -- few of us have the close acquaintance with professional sports to have an opinion on the subject of local team cultures. But university faculties -- that's a different story. Here I feel more confident in asserting that there are indeed significant differences of culture across institutions, even institutions that are otherwise very similar. Among select liberal arts colleges, for example, there are wide variations on different campuses about the value of "university citizenship" (service on college committees, for example); the expectations and duties associated with teaching undergraduate students (high commitment versus low commitment to spending time with students; a felt obligation to provide meaningful commentary on student writing assignments); and probably differences around the standards of what is acceptable in gender relations (for example, the acceptability of romantic relationships between faculty and students or senior faculty and junior faculty). A little less tangibly -- it seems to me that there are visible differences on different college faculties concerning the depth and extent of social relationships among faculty members. Some faculties are tight-knit, with a lot of socializing independent from official functions; and others are more distant, with few serious friendships among faculty members. Some faculties give an impression of welcoming newcomers; others give a more standoffish or even unfriendly impression. (As a young faculty member on the job market years ago I was very aware of the reputations that different philosophy departments had for the way in which they conducted interviews with prospective assistant professors. One department was legendary for positioning the members of the search committee in a semi-circle surrounding the candidate, so that the candidate couldn't actually see everyone at once.)

One way of rephrasing the question here is this: is the climate and culture of a place (a university or a basketball team) just the net result of the personalities and idiosyncracies of the group of people who happen to have been recruited into the group; or does the culture of the group have a certain normative persistence, capable of transmission to new arrivals? The example of university faculties seems to support the second interpretation. And in fact it seems possible to identify some of the social mechanisms through which this transmission occurs: for example, imitation, coaching and mentoring, social discipline, and the transmission of narratives of "who we are at University X or College Y". The behaviors that are valorized through existing group members' practices and observations serve as a prescriptive model for the behaviors of more junior people. And of course, these processes of transmission are imperfect and malleable. The culture of a place changes over a period of several decades.

So I am tempted to think that small communities do in fact make their own cultures over an extended period of time, and that these cultures -- values, practices, modes of inter-personal interaction, experiences of commitment -- have some degree of durability. They do in fact succeed in changing the new recruits that come into the community, often enough to allow the mix of values and experiences that characterize the micro-culture to persist across multiple generations.

Paired comparisons

Sidney Tarrow is a gifted and prolific student of comparative politics. (Listen to my interview with Professor Tarrow.) He has spent much of his career trying to understand social movements, contentious politics, and the causes of differences in political behavior across national settings. And one of his special contributions is his ability to think clearly about the methods that social scientists use.

Tarrow attaches a lot of weight to the idea of "paired comparisons" as a method of research and discovery: Locate a few cases that are broadly similar in many respects but different in a way that is important, interesting, or surprising. Then examine the cases in greater detail to attempt to discover what explains the difference between the two cases. (One of his early books that employs this method is From center to periphery: Alternative models of national-local policy impact and an application to France and Italy.)

Nothing special turns on "pairs" here; what Tarrow is describing is really the logic of small-N comparative research. The point about the broad similarity that is the basis for choosing the cases follows from the logic of causation: if we presuppose that the outcome P is caused by some set of antecedent social and political conditions and we know that C1 and C2 have different outcomes -- then the more factors we can "control" for by finding cases in which these factors are constant, the better. This is so, because it demonstrates that none of the constant factors in the two cases are the cause of variation in outcome. And this limits our investigation of possible causes to the factors in which the cases differ.

If this sounds like Mill's methods of similarity and difference, that's appropriate -- the logic is the same, so far as I can see. Here is Mill's method of difference:

A B C D -E => P
A B -C D -E => -P

And in this case -- making the heroic assumption that A,B,C,D,E exhaust all possible causes of P, and that the cause of P is deterministic rather than probabilistic -- then we can infer that the presence of C causes P.

This reasoning doesn't take us to a valid conclusion to the effect that C is the only factor that is causally relevant to the occurrence of P; it is possible, for example, that there is a third case along these lines:

-A B -C D -E => -P

This would demonstrate that A is a necessary condition for the occurrence of P; withhold A and P disappears. And each of the other factors might also play a role as a necessary condition. So it would be necessary to observe as many as 32 cases (2^5) in order to sort out the status of A through E as either necessary or sufficient conditions for the occurrence of P. (The logic of this kind of causal reasoning is explored more fully in my essay, "An Experiment in Causal Reasoning," which is also published in Microfoundations, Methods, and Causation.)

But I don't think that Tarrow is intending to advance the method of paired comparison as a formal method of causal inference, along the lines of inductive or deductive logic. Instead, I think he is making the plausible point that this method should be understood as a part of an intelligent research strategy. Social processes are complex. We are interested in explaining variation across cases. And we probably have the best likelihood of discovering important causal relationships if we can reduce the number of moving parts (the other kinds of variation that occur across the cases).

Tarrow gives an example of the application of the method of paired comparisions in the context of his early study of the political fortunes of the Italian Communist Party (PCI) in the south of Italy. In this case the paired comparison involves northern Italy and southern Italy. Both are subject to the same national political structures; both populations speak Italian; both populations have an Italian national identity. However, the PCI was fairly successful in mobilizing support and winning elections based on a militant political program in the north, and was persistently unsuccessful in doing these things in the south. What explains the difference?

As Tarrow explains his reasoning, his expectation in conducting the research was a "structural" one. He expected that there would be large structural factors in post-war Italy -- features of economic and political institutions -- that would explain the difference in outcome for PCI political activism. And there were indeed large structural differences in social organization in the two regions. Northern Italy possessed an economy in which industrial labor played a key role and constituted a substantial part of the population. Southern Italy was agrarian and backward, with a large percentage of exploited peasants and only a small percentage of industrial workers.

But, very significantly, Tarrow now believes that these "structural" expectations are probably too "macro" to serve as the basis of social explanation. Instead, he favors the importance of looking at the dynamics of social processes and the specific causal mechanisms that can be discovered in particular social-historical settings. This means looking for causal factors that work at a more strategic and meso level. In terms of the southern Italian PCI outcome that he was interested in explaining thirty years ago -- he now believes that the causal mechanism of "brokerage" would have shed much more light on the political outcomes that were of interest in Italy. (This is the heart of the approach that he takes, along with Doug McAdam and Chuck Tilly, in Dynamics of Contention.)

This finding doesn't invalidate the heuristic of paired comparisons. But it probably does invalidate the expectation that we might discover large "structure-structure" patterns of causation through such comparisons. Instead, what the method facilitates is a more focused research effort on the part of the comparativist, in the context of which he/she can search out the lower-level causal mechanisms and processes that are at work in the various settings under study.

Charting inequality

Inequality is a central and familiar topic for sociological research (as it is for social and political philosophy). But it is worth probing what kinds of inequality may exist in society and what kinds of explanations might serve to account for these various social processes and outcomes. What do we want to know about social inequalities?

Most evidently we can observe inequalities across individuals and groups with respect to the level of attainment of various important social goods: income, wealth, home ownership, health status, educational attainment, and employment, to name several. It is trivial to observe that there are differences across individuals with respect to these goods -- John has higher income and better health than Phil. What is less trivial is the discovery that members of a group, defined in terms of one or more socially significant properties, show differences with respect to these outcomes: inequalities of income between men and women, inequalities of incidence of diabetes across white and black adults, differences in educational attainment across rural, suburban, and urban residence, and so forth. Once we have identified inequalities like these across social groups, we want to have a causal explanation of the differences. What are the social processes through which these differences in outcome arise?

A related kind of inequality is somewhat less tangible. It has to do with unobservable social properties like status, power, or privilege. We can't directly measure this person's total social power or that person's social status. But we can arrive at comparative judgments about the relative level of status and power for various individuals, and we can likewise make comparative judgments about these qualities as exhibited by various social groups. Once again the question of social causation arises: what are the social processes that give rise to different levels of power, status, or privilege for various social groups?

In each instance we want to know what the social processes and mechanisms are that proliferate outcomes differentially across groups defined by such characteristics as race, gender, income status, etc. What factors and processes cause the occurrence of inequalities with respect to a social good across social groups defined by a socially salient property (race, gender, age)? If there is a measured difference with respect to the social property across two groups, there must be a causal factor that distinguishes the groups. A natural hypothesis is that human talents and personalities are randomly distributed across all human beings. On this assumption, we should expect that there will be no differences of outcome across social groups that are purely based on differences of talent, since by hypothesis there is no difference in the distribution of talents across randomly selected groups. So if there are observed differences in outcome, we need to find a social process that would explain the differences across groups. There must be a causal factor that explains the difference.

We might say that there are only a few basic possibilities. Members of the groups might possess personal characteristics differentially that causally produce the good. This might be the result of filtering or differential recruitment into the group. Second, the mechanisms assigning the good to individuals might discriminate on the basis of the property or a correlated property. Third, membership in the group might give members differential access to resources or opportunities that are themselves productive of the good.

So far we have formulated the problem as one of explaining inequalities across groups defined by some other feature. But there is a separate issue about inequalities that sociologists address. Suppose we notice that the “rich” are gaining a higher percentage of income over time and the “poor” a falling percentage. These two groups are defined with respect to the very property with which they are measurably unequal. Formulated from this perspective, the question is this: what are the causal factors possessed by the group of “rich” that accounts for their high income, and conversely for the “poor”? Upon investigation we might find that the two groups are different in a variety of ways: amount of higher education, place of residence, race, age. This might lead us to ask the question, are some of these factors causal with respect to income; are other factors collateral effects of high income; and are yet others simply the non-causal correlates of the truly causal factors?

This suggests two angles of approach on the question of inequalities across groups. One is to single out the two populations at opposite ends of a particular spectrum of difference and examine their various characteristics. The other is to single out a socially salient property (race, for example) and investigate whether there are differences with respect to the social good (income). The first cut raises the question, what factors distinguish the high-achieving and low-achieving groups that might explain the observed differences; the second asks, why is the property causally relevant to the distribution of the good?