The data confirm the expected role of population and income. Early- and middle-adopting DMAs had, on average, five times larger populations and 24% larger per capita incomes than late- adopting DMAs. After controlling for log population and income, however, differences between early and late adopters appearmuch more idiosyncratic. Indeed, in regressions controlling for log pop- ulation and income, F-tests show no statistically significant re- lationship between television adoption category and percent high school educated, median age, or percent nonwhite at the DMA level. (See the online appendix to this paper for details.) All of the models we estimate below will control for DMA-level log popula- tion and income, so the parameters.
