Consider the following fixed effects model:

log(Salesit)=0+1log(Advertisingit)+ai+D2017+uit

t=2016, 2017

where ai is a fixed effect for each state, D2017 is a dummy that takes the value 1 in year 2017 and zero in year 2016, and sales and advertising are measured in millions. We obtain an estimate of 1 of 0.5.

1.What is the approximate rate of return to advertising (assume there is no endogeneity)? Write a sentence using the estimated effect.

2.Suppose that advertising expenditures and sales are higher in states with larger land area (i.e., the size of the state is positively correlated to advertising and sales). Will this cause a bias in the estimated effect? Explain.

3.Suppose using the same data we estimate the following model in differences:

log(Salesi2017)-log(Salesi2016)=0+1 (log(Advertisingi2017)-log(Advertisingi2016))+ui2017-ui2016

Explain whether or not you are able to tell what the estimate of 1 is?