Causal Arguments: Establishing That X Actually Causes Y
Causal arguments claim that one thing causes another. They are central to policy debates ('this policy will cause that outcome'), scientific reasoning, and everyday decision-making. But establishing causation is far harder than most people realize, and causal arguments are among the most frequently flawed in public discourse.
The Criteria for Causation
Temporal precedence: the cause must precede the effect. This seems obvious, but it is violated more often than you might think. People often observe two phenomena occurring together and assume one causes the other without verifying which came first.
Covariation: the cause and effect must be correlated -- when the cause is present, the effect tends to occur; when absent, the effect tends not to occur. Covariation is necessary but not sufficient for causation, as we discussed in the correlation vs. causation article.
Elimination of alternatives: other possible explanations for the observed relationship must be ruled out. This is the hardest criterion to satisfy. There may be confounding variables, reverse causation, or coincidence. The gold standard for eliminating alternatives is the randomized controlled experiment.
Mechanisms and Causal Chains
A causal argument is stronger when you can identify the mechanism by which the cause produces the effect. 'Smoking causes cancer' became more convincing when the specific mechanisms (tar damaging lung cells, carcinogens causing mutations) were identified, not just the statistical correlation.
Causal chains link a cause to an effect through intermediate steps. 'Cutting taxes will grow the economy because lower taxes leave more money for consumers, who spend it, creating demand for goods, which causes businesses to hire more workers.' Each link in the chain is a separate causal claim that needs its own support.
Long causal chains are inherently weaker than short ones because the probability of each link being correct must be multiplied. If each link has a 90% chance of being correct, a chain of five links has only a 59% chance of being entirely correct (0.9^5). This is why causal arguments about distant effects should be treated with more skepticism.
Common Causal Fallacies
Post hoc ergo propter hoc ('after this, therefore because of this') assumes that because B followed A, A caused B. 'I wore my lucky socks and we won the game, so the socks caused us to win.' Temporal sequence does not establish causation.
Single cause fallacy assumes that an effect has only one cause when it may have multiple. Most real-world phenomena have multiple contributing causes. Poverty is not caused by a single factor; it results from the interaction of education, opportunity, policy, health, and many other factors.
Oversimplified causation ignores moderating and mediating variables. 'Education causes higher income' ignores that the relationship is moderated by field of study, type of institution, social connections, local economy, and many other factors. A complete causal account acknowledges this complexity.
- •Establishing causation requires temporal precedence, covariation, and elimination of alternatives.
- •Identifying the mechanism of causation strengthens causal arguments significantly.
- •Long causal chains are weaker than short ones because uncertainty compounds at each step.
- •Common causal fallacies include post hoc reasoning and single-cause thinking.
- •Most real-world effects have multiple causes -- acknowledge this complexity.