Analysts at the pharmacy chain CVS are trying to understand whether coupons work for the purpose of increasing sales. (In practice, coupons are a promise for a future discount. A sample CVS coupon is shown in the image below.) To this end, these analysts have assembled a dataset of historical transactions which includes customer-level information on whether they have received a coupon and/or made purchases from the store. The dataset includes the two following variables in addition to a customer identifier: Couponit: whether customer i received a CVS coupon in over the week prior to week t (=1 if yes, =0 if no). PurchasedValue,t: total amount purchased (in dollars) by customer i from some CVS store during week t (net of all discounts). With this dataset, these analysts estimate the following regression: PurchasedValueit = Bo + Bi Couponit Their results indicate that B1 = -3 (p-value < 0.01). Based on this evidence, should the company stop distributing coupons? What feasible alternative empirical approach would you consider to addressing the question? Explain your rationales