The Census Bureau wishes to estimate mean household income by state, by taking a random sample of households from each state. It wishes to choose sample sizes such that the estimates for all states have about the same accuracy. Which procedure will come closest to achieving this: selecting an equal percentage of households from each state, or an equal number of households?

1. Equal percentages (0.1% Wyoming households, 0.1% California, etc, etc) or

2. Equal numbers (1000 households from Wyoming, 1000 from California, etc)

Be sure to justify your answer with an explanation.