Instrumental variables are used in statistical analysis to address problems of confounding or endogeneity, especially in regression models. An instrumental variable is associated with the independent variable of interest but affects the dependent variable only indirectly, through that independent variable. By using an appropriate instrument, researchers can estimate causal relationships more reliably, even when there are unmeasured confounders.
For example, economist David Card (1995) used proximity to a college as an instrumental variable to study the effect of education on income. People living closer to a college were more likely to attend, but location itself was assumed not to directly influence their later earnings. This approach allowed Card to isolate the causal impact of education on wages.
« Back to Glossary Index