Exercise

eGulf Auctions Experiment: Noncompliance

In a previous exercise, we found that sellers on eGulf who posted more than 10 pictures of their WePhones during auctions had higher sales prices. However, we never examined whether noncompliance could have confounded our results. Specifically, was there some systematic difference between the sellers who chose to post more than 10 pictures of their WePhones and those who chose to post less than 10 pictures?

If an outcome variable is associated with systematic difference between compliers and noncompliers, our observed treatment effect may be spurious. Therefore, let's test whether compliance was associated with any traits among the sellers, and whether such traits were associated with final sales prices. With the dataset eGulf:

Instructions

100 XP
  • 1) Test whether sellers who were offered treatment and opted into the treatment had different prior average feedback scores than sellers who were offered treatment but did not opt into treatment (tracked with Seller_Offered_Treatment, Seller_Opted_Into_Treatment, and Seller_Feedback_Score)
  • 2) Estimate how sellers' prior average feedback scores were correlated with their recent auction's final sales price (Final_Sales_Price).