What Would Justify a Causal Claim?

Introduction

Students who attend a study group may do better on an examination. The observation is useful, but a further question remains: what would happen if more students attended? A causal claim crosses from the pattern observed to the consequences of a change. To justify that crossing, we need a specified alternative and a comparison that separates the change from other influences.

This syllabus begins by distinguishing scientific questions and defining the unobserved alternative outcome. It then examines what assigning people to different conditions contributes, when adjustment is useful, and how an observational study can make its intended comparison explicit. Studies of changes not assigned by the researchers connect those arguments with development in an actual community. The final units distinguish effects from mechanisms and ask how approaches with different weaknesses can strengthen an inference together.

By the end, you should be able to state the alternatives a causal claim compares, explain which assumptions a design needs, identify a plausible alternative explanation and propose evidence that would bear on it. The aim is a defensible argument about a particular question, with its limits visible. The name of a method cannot supply that argument on its own.

Scope and omissions

The nine core articles span six units and 112 full-work pages. This is an intermediate conceptual path through causal identification, with formal arguments, diagrams and design comparisons. Understanding what the calculations assume matters more here than being able to implement every estimator. Manuscript and bundled access copies have different page counts, explained in their access notes.

The examples connect education, development, psychological mechanisms and epidemiology. Each setting has its own populations, measures, timing and ethical constraints. Income support in one rural community does not stand for every change in poverty, and health-method examples do not supply treatment advice. Consent, feasibility, burdens and potential harms constrain which interventions can be studied.

This is not exhaustive training in regression discontinuity, instrumental variables, interrupted time series, longitudinal estimation, sensitivity analysis or causal discovery. Optional works extend some of those questions. Identifying an effect in one study remains distinct from explaining every part of a process or generalizing it to new people and settings.

Synthesis

Causal reasoning asks what would differ under alternative conditions, for particular people over a particular period. Since we cannot observe both alternatives for the same case at the same time, the design must justify the comparison we use. Random assignment, adjustment, and natural experiments offer different routes, each with conditions and vulnerabilities that belong in the argument.

Keep three further distinctions visible. A well-identified average effect need not describe every individual. Evidence that an intervention works does not automatically identify its mechanism. Agreement among studies is most useful when their questions are comparable and their important sources of bias differ. Disagreement can expose either an inferential weakness or a genuine difference in the effects being estimated.

Return to the study-group example. Specify the intervention, comparator, population, outcome and time. Sketch an assignment-based design and an observational alternative. For each, name a serious threat, the evidence that could reduce it, and one assumption that would remain. Then propose a mechanism and a further comparison capable of testing it. Explain what you would learn from agreement across the approaches and what disagreement would lead you to investigate.

To examine how these design arguments connect to psychological explanation, explore How Do We Build and Test Psychological Theories?. For environmental change across development, continue with How Do Contexts Shape Developmental Pathways?.

For effect magnitude, uncertainty and the assumptions behind statistical summaries, continue with What Can Statistical Evidence Tell Us?.