Correlation, Causation, and Confounding Variables

Comparativists work with data from many countries at once: GDP, literacy rates, freedom scores, oil exports, voter turnout. When you plot two of these against each other and see a clear pattern, it's tempting to say one causes the other. But a correlation only tells you that two things tend to rise and fall together across the countries you looked at. It says nothing about which one, if either, is doing the causing.

The danger is the confounding variable: a hidden third factor that independently affects both things you're measuring, creating a correlation between them that has no direct causal link at all. Oil wealth and authoritarianism often appear together, but weak pre-existing institutions can produce both outcomes separately, oil dependence on one side and a slide toward dictatorship on the other, without oil itself forcing the political outcome. The correlation is real. The story that oil wealth directly causes dictatorship is not established by that chart.

This is why comparativists separate empirical claims, statements about what is actually happening that can be checked against evidence, from normative claims, judgments about what is good or bad. A scatterplot can support an empirical claim like higher income countries tend to score higher on freedom indexes. It cannot tell you whether freedom is a better goal than growth. Splitting data into subgroups, say rich countries versus poor countries, is one way researchers probe a correlation further: if a relationship holds in both groups, it's sturdier; if it vanishes in one group, the original correlation may have been driven by wealth all along, not by the variable you first suspected.