Analytical sociology is based on the idea of mechanism-based account of explanation. According to analytical sociologists, a fully satisfactory social scientific explanation requires that the causal mechanisms be specified. The central aim of science is to understand phenomena, and this is precisely what mechanisms provide. If we take understanding to be an ability to answer what-if questions, the contribution of the mechanism-based information becomes apparent. A simple causal claim tells us about counterfactual dependency: It tells us what would have happened if the cause had been different. The mechanism tells us why the counterfactual dependency holds and ties the relata of the counterfactual to the knowledge about entities and relations underlying it. The account of a causal mechanism integrates an isolated piece of causal knowledge with a much larger body of knowledge and helps us to answer many natural follow-up questions about the conditions under which the causal dependency holds: For example, what are the necessary background conditions and what are the possible intervening factors that have to be absent for the effect to be present? In this way the mechanism expands our ability to answer what-if questions, i.e., it deepens our understanding.

While mechanism-based account of explanation requires that explanations are causal, it is not wedded to a specific theory of causation. However, the mechanism perspective sets some important constraints for an acceptable theory of causation. 

1. The idea of productive causal activities associated with the mechanism perspective implies a commitment to the locality of causal processes: Whether A is a cause of B depends on facts about spatiotemporally restricted causal process, not on what would happen in other similar situations. This means that theories that attempt to define causality in terms of regularities (such as Hume’s constant conjunction theory and many probabilistic theories of causation) are not compatible with mechanism-based theories. This does not mean that the supporters of the mechanism-based perspective would have to disregard regularities as an important source of evidence about causal relations.
2. The attempts to define causation in terms of mechanisms faces some tricky problems. First, most characterizations of mechanisms employ causal notions, so a definition of causation in terms of mechanisms would be circular. Second, such an account would face a thorny question about the causal powers of fundamental (physical) entities. If causal relations at the fundamental level are not mechanical, the definition is false, and if they are mechanical, we end up with an infinite regress, which is regarded by many as an unhappy consequence.
3. Some supporters of mechanism-based explanations find Wesley Salmon’s conserved quantities theory of causal processes intuitively appealing. However, Salmon’s theory has turned out to be a disappointment. In addition to many technical and philosophical problems associated with the theory, it seems quite ill suited to provide a foundational account of a mechanism-based explanation. First, it is very difficult to see how the theory could be applied at all in biological and social sciences (or even some parts of physics), where explanations do not attempt to track spatiotemporally continuous processes such as transfer of energy or momentum. Second, Salmon’s approach is marred by the fact that it does not include any considerations of explanatory relevance. As a consequence, the same counterexamples that have been raised against the covering-law account can be raised against Salmon’s approach.
4. A manipulation account of causation can be combined with the mechanism-based account of explanation. For example, in Woodward’s account, causal claims track relations of counterfactual dependency. They tell us what would have happened to the effect if the cause had been subject to a surgical intervention that would not have affected any other part of the causal structure. One of the novelties of Woodward’s theory is its account of causal generalizations in terms of invariances. According to Woodward, the explanatory qualities of a generalization are determined by its ability to tell us about the counterfactual consequences of possible interventions, not by the properties traditionally associated with laws of nature. Woodward’s account supplements the mechanism-based account by providing an account of explanatory relevance and by making sense of causal production. However, it does not preempt the importance of the mechanism-based perspective. The mechanism approach differs from Woodward’s approach by its emphasis on the importance of opening up black boxes and making explicit the causal cogs and wheels through which effects are brought about.

The idea of mechanism also plays important role in causal inference. Especially in nonexperimental contexts, the ideas about possible causal mechanisms often have a crucial role to play in distinguishing true causal relations from spurious correlations. Mechanisms help in causal inference in two ways. The knowledge that there is a mechanism through which X influences Y supports the inference that X is a cause of Y. In addition, the absence of a plausible mechanism linking X to Y gives us a good reason to be suspicious of the relation being a causal one. The knowledge of mechanisms also has an important role in extrapolation of causal findings from one setting to another: the assumption about the similarity of causal mechanisms is crucial for making inferences from one setting or population to another. However, mechanisms are not some sort of magic bullet for causal inference. The problem often is not the absence of possible mechanisms, but how to discriminate between a number of potential mechanisms. To avoid lazy mechanism-based storytelling, the mechanism scheme must be made explicit and detailed, and its assumptions must be supported by relevant empirical evidence.

Methods and analytical sociology

The idea of mechanism-based explanation is a metatheoretical idea that can be combined with many research methods. Thus it can utilize both quantitative data (like register-based statistics, surveys, etc.) and qualitative data (like ethnographic observations or interviews). However, two general observations can be made:

1. The ambitious goals of mechanism-based explanation usually requires multiple sources of evidence. This highlights the role of theory in the integration of different sorts of evidence. Causal inferences are theoretical inferences, not simple outputs of data analysis.
2. A consistent mechanism-based approach requires that evidential value of all data is based on the understanding of causal processes that have produced it. Thus we have a mechanistic “theory” of data generation in addition of mechanism-based account of the phenomenon to be explained.