1. On January 1, 2017, the City of Philadelphia introduced a tax on sugar-sweetened beverages and sugar-substitute beverages. A telephone survey of 1777 people in Philadelphia and in nearby cities (Trenton, Camden, and Wilmington) obtained data on whether those people drank bottled water on a daily basis in the month prior to the tax being introduced (December 2016) and in the month following the tax being introduced (January 2017). In December, 46.5% of Philadelphians and 43.9% of non-Philadelphians reported being daily drinkers of bottled water. In January, 50.2% of Philadelphians and 37.2% of non-Philadelphians reported being daily drinkers of bottled water.

1. Estimate the effect of the tax on Philadelphians’ frequency of daily bottled water consumption using the difference in differences estimator. State the estimated effect in words – don’t just provide a number. (Show your numerical work for partial credit in case of computational error.)

2. What OLS regression using the 1777 individual observations would obtain the same numerical estimate of the effect as does the difference-in-differences estimator in (i). Define each variable that appears in the regression.

3. Indicate which coefficient’s OLS estimate or what combination of OLS coefficient estimates from the regression given in (ii) would equal the difference-in-differences estimator.

4. If the assumptions underlying the difference-in-differences estimator’s unbiasedness were satisfied, would a regression with added demographic variables for the individuals be any better than the difference-in-differences estimator in (iii)? VERY briefly explain.

If the assumptions underlying the difference-in-differences estimator’s unbiasedness were not satisfied, would a regression with added demographic variables for the individuals be preferable to the difference-in-differences estimator in (iii)? VERY briefly explain.