Unit 10 Dummy Variables: Does One Line Fit Everyone?

Every explanatory variable so far has been a number. Years of schooling, classes attended, capital per worker — quantities that can be larger or smaller, and for which “a one-unit increase” means something.

Much of what economists want to compare is not like that. Rural and urban. Treated and untreated. Before and after the reform. These are categories, not quantities, and there is no sense in which one is one unit more than another.

Twice already we have handled such a variable by writing 0 and 1 and moving on. It is time to ask what that actually did — and then to ask a question the technique makes possible.

Every regression in this book has assumed that the same relationship holds for everyone. One return to schooling for every worker. One attendance effect for every student. One production function for every firm.

That assumption is rarely stated aloud, but every regression makes it unless we tell it otherwise.

This unit shows how to put groups into a regression, how to let a relationship differ between them, and how to test whether it does.