Economics
Fixed Effects
Quick fact
Fixed effects can be implemented simply by subtracting each entity's average over time from every observation—a technique called 'within transformation' or 'demeaning'.
Why this is interesting
Imagine comparing two schools: one with high test scores, one with low. Are the differences due to teaching quality, or just the kinds of students who enroll? Fixed effects cut through that confusion.
Read the full explanation
Understanding Fixed Effects
When analyzing data that tracks the same entities (like schools, companies, or countries) over multiple time periods, you often worry that something unique to each entity—unmeasured but constant over time—could bias your results. For example, a school's neighborhood culture might affect both student enrollment and test scores, but never change during the study. Fixed effects solves this by using only the changes within each entity across time. Think of it as comparing each school to itself at different moments: any improvement or decline is attributed to variables that actually changed, not to fixed characteristics.
A deeper explanation
Mechanically, fixed effects models include a separate intercept (dummy variable) for each entity, absorbing all time-invariant differences. Alternatively, the 'within estimator' subtracts each entity's mean from every variable, removing those constant factors. The key insight is that only variation over time within an entity is used to estimate coefficients. This eliminates bias from unobserved confounders that are constant over time—but it cannot control for time-varying unobservables. Fixed effects are crucial in economics, political science, and sociology for estimating causal effects from non-experimental panel data, making them a foundational tool in modern empirical research.