QUESTION 4 Suppose a manufacturing rm has two factories (Factory 1 and Factory 2), and a single production process (Process A) that is used in both factories. A new process (Process B) is developed that potentially reduces production costs. To test whether Process B is less costly than Process A, an experiment is designed where: 1. Within each Factory, products are assigned randomly to Process A or Process B. 2. Production costs for each product are recorded. Note that resources (i.e. materials, workers, equipment) are n_ot reassigned across factories. Let Yi be the cost of producing product i, let Xi be 1 if Process B is used to produce i and 0 if Process A is used, and let Wi be 1 if product i is produced in Factory 1 and 0 if it is produced in Factory 2. In a regression of Yi on X, it is advisable to: O a. Exclude Wi as products are randomly assigned and including Wi would increase standard errors. 0 b' Exclude Wi as it is uncorrelated with Xi O C' Include Wi as E(ui | Xi) :9 0, but E(Xi | Wi) = 0 Q d' Include Wi as E(ui | Xi) :9 0, but E(ui | Xi, Wi) = E(ui | W) QUESTION 5 Continue to use the example from Question 4. Suppose each product is randomly assigned to a process by a computer program, but some products get reassigned on the factory oor (for practical reasons). Let Zi denote the original assignment and Xi the actual process used to produce i. In a regression of Yi on Xi and Wi, OLS is: O a. Potentially biased because Wi should not be included 0 b' Potentially biased, but an IV regression using Zi as an instrument can be used to obtain a consistent estimator O C' Unbiased because the products were randomly assigned in the beginning 0 d' Unbiased as long as Zi is also included as a control variable