Variables & Causal Mechanisms — Identify independent/dependent variables and causal mechanisms

The systematic study of causation in the social sciences did not emerge overnight. For centuries, political philosophers from Aristotle to Montesquieu speculated about what causes governments to flourish or collapse, but their reasoning relied on anecdote, analogy, and normative argument rather than on rigorous empirical design. The intellectual breakthrough came when scholars recognized that causal claims required explicit identification of the factor doing the causing (the independent variable), the outcome being affected (the dependent variable), and a plausible story about how one produces the other — the causal mechanism. Tracing this intellectual trajectory helps us appreciate why modern political science demands all three elements before accepting any explanatory claim.

This historical arc reveals a persistent question at the heart of political science methodology: how do we move from observing that two things go together to claiming that one actually produces the other? The answer requires clarity about which variable is the cause, which is the effect, and what process links them. The remainder of this lesson builds that clarity from the ground up.

Before constructing any research design, a political scientist must specify the core elements of a causal argument. These elements function together: an independent variable (IV) is the factor whose variation a researcher believes produces change in the outcome; a dependent variable (DV) is the outcome whose variation the researcher seeks to explain; and a causal mechanism is the process, sequence of events, or pathway through which the IV transmits its effect to the DV. A fourth concept — the confounding variable — refers to an uncontrolled factor that correlates with both IV and DV, threatening the validity of any causal claim. Understanding all four is essential to evaluating and designing empirical research.

A causal argument in political science can be represented as a directed diagram in which arrows denote the hypothesized direction of influence. The following diagram illustrates a research hypothesis about economic conditions and populist voting, including the mediating mechanism of institutional trust and a potential confounding variable.

Several features of this diagram deserve emphasis. First, the solid arrows represent the hypothesized direction of causation — they flow from left to right, from cause to effect. Second, the mediating variable (declining trust) appears within the causal chain; it is part of the mechanism, not external to it. Third, the confounding variable (media framing) is depicted with dashed lines because it represents an alternative explanation that, if unaddressed, could invalidate the researcher's claim. Every well-designed study must either control for potential confounders or argue persuasively why they are unlikely to operate. The capacity to distinguish these roles — cause, effect, mechanism, confounder — is the foundation of causal reasoning in political science.

Establishing that X and Y covary is a necessary but insufficient step toward causal inference. A correlation between two variables could reflect genuine causation, reverse causation, or a confounding factor. Causal mechanisms provide the theoretical glue that transforms a statistical association into a credible explanation. A mechanism specifies the intervening steps — the actors, decisions, institutional rules, or psychological processes — through which a change in the IV transmits an effect to the DV. As political methodologist James Mahoney notes, mechanisms turn 'black-box' explanations into transparent, testable stories.

Three Criteria for Causal Claims

Following the tradition running from Hume through modern methodology, most political scientists require three criteria before accepting a causal claim. The first criterion is covariation: the IV and DV must vary together in the predicted direction. If we hypothesize that higher education spending increases democratic stability, we must observe that countries spending more on education also tend to be more stable. The second criterion is temporal precedence: the cause must precede the effect in time. A change in education spending must occur before any observed change in democratic stability. The third criterion is the elimination of alternative explanations: confounders, reverse causation, and other threats to internal validity must be ruled out, either through research design (such as randomization) or through theoretical argument and statistical controls.

Qualitative vs. Quantitative Approaches to Mechanisms

Causal mechanisms can be investigated through both quantitative and qualitative methods, and the choice between them has significant implications for research design. Quantitative researchers often test mechanisms via mediation analysis, decomposing the total effect of X on Y into a direct component and an indirect component passing through one or more mediators. Qualitative researchers, by contrast, tend to use process tracing — detailed within-case analysis that follows the causal chain step by step, looking for observable evidence at each link. Both approaches seek the same goal: to open the 'black box' between X and Y and demonstrate that the theorized process actually operates. In practice, the strongest research programs triangulate across methods, using large-N analysis to establish covariation and case studies to verify the mechanism.

Real political phenomena rarely involve a single IV affecting a single DV through a single mechanism. Research designs must account for several types of variables that occupy different positions in the causal structure. Understanding these roles is essential for specifying an accurate model and avoiding misattribution of effects.

Suppose a researcher reads news reports suggesting that countries with more women in parliament tend to spend more on social welfare. She wants to move from this observation to a testable causal argument. Let us walk through the steps required to identify the variables, specify a mechanism, and anticipate threats to causal inference.

Different research traditions approach variables and causal mechanisms in distinct ways, each with characteristic strengths and limitations. The following table compares the three major approaches students of political science will encounter: large-N statistical analysis, experimental design, and qualitative case-study methods.

The variable-centered framework introduced in this lesson provides the vocabulary for causal reasoning, but advanced research methods courses build on this foundation with more formalized frameworks. The most influential is the potential outcomes framework (also called the Rubin Causal Model), which defines causation in terms of counterfactuals: the causal effect of X on Y for a given unit is the difference between the outcome that unit would experience under treatment (X = 1) and the outcome it would experience under control (X = 0). Since we can never observe both states for the same unit at the same time — the fundamental problem of causal inference — all empirical strategies (randomization, matching, instrumental variables) are ultimately attempts to approximate this unobservable counterfactual comparison.

Understanding the basic logic of IVs, DVs, and causal mechanisms prepares you for this more rigorous formalization. When you encounter the potential outcomes framework in an advanced course, you will recognize that it is simply making precise the same ideas you have learned here: the cause is the treatment assignment (IV), the outcome is the potential outcome (DV), and the mechanism is the process that translates treatment into outcome. The concepts in this lesson are not simplifications to be discarded — they are the conceptual scaffolding upon which all advanced causal inference is built.