Causation

What Is Causation?

Causation is a cause-and-effect relationship in which one event, action, or condition helps produce a change in another. It asks what would happen if a suspected cause changed, not merely whether two things occur together.

The cause is the factor thought to influence an outcome, and the effect is the resulting outcome. Causal claims guide action in medicine, cities, business, and courts. A causal claim says changing one factor would likely change another under stated conditions. For related concepts, see the glossary of terms.

How Causation Works: Cause, Effect, Mechanism, and Context

Causal reasoning tests whether an apparent relationship is real.

1. Define the outcome, such as missed appointments.
2. State the suspected cause and expected effect.
3. Describe a plausible mechanism, or how the cause could produce the outcome.
4. Compare outcomes with and without the suspected cause using a fair comparison group.
5. Check alternatives, including group differences, outside events, and measurement errors.
6. Test whether the result holds in the relevant setting and population.

Causation vs. Correlation

Correlation means two variables change together. Causation means changing one contributes to a change in the other. Correlation can identify questions worth investigating, but cannot answer them alone.

A hidden third factor is a confounder, and reverse causation occurs when the supposed effect influences the supposed cause. The Australian Bureau of Statistics explanation of correlation and causation likewise cautions that association alone cannot prove cause and effect.

Key Types of Causation

Most outcomes have several causes. These categories clarify causal claims.

Causal Inference and Causal Analysis

Causal inference draws a justified conclusion about whether a cause affects an outcome. Causal analysis includes framing the question, mapping pathways, selecting evidence, testing alternatives, and interpreting results.

Its central idea is counterfactual thinking: what would have happened if the suspected cause had differed? Because that alternative is usually unobservable, researchers use comparable groups or repeated observations. Causal diagrams can map causes, effects, and confounders before analysis. Pre-specifying a hypothesis and method can reduce selective reporting. For example, a school meal program analysis might compare attendance before and after implementation with a similar school without the program.

Methods Used to Establish Causation

No method automatically proves causation. The best design depends on ethics, feasibility, data, and the decision at stake.

• Randomized controlled experiments randomly assign participants to conditions. Randomization helps balance known and unknown confounders.
• Natural experiments examine policy, geography, timing, or other outside changes that approximate random assignment.
• Longitudinal studies follow people or systems over time, helping establish whether cause came before effect.
• Quasi-experimental designs use structured comparisons, such as before-and-after changes with a comparison group.
• Controlled observational studies adjust for measured differences, but cannot remove an unmeasured confounder.
• Triangulation compares methods and data sources. Agreement increases confidence.

A Simple Example of Causal Thinking

Suppose a clinic wants to know whether text reminders increase appointment attendance. The outcome is attendance, and the suspected cause is receiving a reminder. A strong test would randomly send reminders to some eligible patients while others receive usual scheduling communication, then compare attendance rates.

The clinic should consider travel, appointment type, language preference, and phone access. A plausible mechanism is that messages help patients remember or reschedule. If attendance is consistently higher in the reminder group, the conclusion is that reminders increased attendance for this population and delivery method, not that every individual attended because of a text.

Benefits of Understanding Causation

Causal thinking improves decisions by focusing on actions that can change outcomes.

• It helps prioritize interventions instead of superficial patterns.
• It supports prevention by identifying factors that raise risk.
• It distinguishes useful changes from activities that merely coincide with improvement.
• It strengthens accountability by separating plausible contribution from unsupported blame.
• It improves assessment of claims in advertising, news, policy, and research.

Limits and Challenges of Causal Claims

Even careful research has limits. Responsible conclusions name them.

• Experiments may be unethical or impractical, especially when a suspected cause could harm people.
• Incomplete records and measurement error can distort comparisons.
• Small, biased, or unrepresentative samples may not apply elsewhere.
• Hidden confounders can remain after statistical adjustment.
• Feedback loops can make causes and effects influence each other over time.
• Population-level effects do not predict every individual outcome.

Common Pitfalls When Assessing Cause and Effect

Simple checks can prevent overconfident conclusions.

• Do not assume that because one event happened first, it caused the next.
• Do not rely on memorable examples while ignoring base rates and the full comparison group.
• Do not treat a plausible mechanism as proof without outcome evidence.
• Ask whether reverse causation could explain the pattern.
• Remember that adjustment helps, but does not guarantee all confounding is removed.
• Ask: compared with what, what else changed, who was studied, and how certain is the estimate?

Causation in Law and Philosophy

In law, factual causation often uses the but-for test: would the harm have occurred but for the defendant's conduct? Legal or proximate causation also asks whether the connection is sufficiently direct for liability. An intervening event can weaken or break that connection. Legal responsibility differs from scientific explanation, which may describe many contributing factors without assigning blame.

Aristotle offered four causes: material cause, what something is made from; formal cause, its structure; efficient cause, the agent or process producing it; and final cause, its purpose. Modern causal inference focuses on interventions, mechanisms, and observable outcomes. Causal reasoning is also useful in app design and development, where teams test whether a design change improves a defined user outcome.

A Practical Checklist for Evaluating a Causal Claim

Use this sequence before acting on a claim that one thing causes another.

1. State the claim specifically.
2. Identify the proposed cause, outcome, population, and time frame.
3. Check whether the cause occurred before the outcome.
4. Look for a credible comparison group or counterfactual.
5. Consider confounders, reverse causation, and selection bias.
6. Assess whether the proposed mechanism and evidence fit together.
7. Decide whether the finding applies to the situation at hand.
8. Communicate the conclusion with appropriate confidence and limits.