UGBA 88 LECTURE 5 INTERFERENCE AND ATTRITION fully solved graded A+ 2023/2024

UGBA 88 LECTURE 5 INTERFERENCE AND ATTRITIONWhat does interference do to the interpretation of treatment/control group comparisons? - correct answer It complicates the interpretation What can you use to account for interference? - correct answer Cluster randomization What does ATTRITION do to treatment and control groups that were comparable at the beginning of an experiment? - correct answer It makes it so that the treatment and control groups may not be comparable--they may be systematically different What is the main assumption we make In the potential outcomes framework (and potential outcomes and causal inference)? - correct answer No interference What does NO interference mean about the treatment applied to one unit? What does no interference mean about potential outcomes? - correct answer The treatment applied to that unit doesn't affect the outcomes of other units; it means that potential outcomes depend only on their OWN treatment status (eg: Y1i depends only on the Di=1) What is spillover? Example? - correct answer (actually measuring the) Indirect causal effect of other's treatment status on an individual's outcome; Allowing some employees to WFH may demoralize or motivate other workers, therefore changing their outcome (productivity) How is it that cluster randomization can account for interference? - correct answer When you randomize into clusters rather than individual units, spillovers are taken away because everyone in the cluster is in the same treatment group, and therefore are unaware of other treatments, so their outcomes are not affected by others' treatment status. What kind of experiments are used to estimate spillovers? - correct answer Hybrid experiment designs What is a naïve experiment? - correct answer An experiment where there's an outcome of interest, a treatment, and a control, and people are randomly assigned treatment or control. What is there to worry about when running a naïve experiment? - correct answer Spillovers (the treatment effect of one person affecting the outcome of another) What do you have to do when you do cluster randomization (eg: giving one treatment status to an entire department and another treatment status to another department) to eliminate spillover/interference? Example? - correct answer You have to define a new causal question; eg: What is the causal effect of giving SUBSETS (as opposed to individuals) of people in a department an incentive on enrollment rates What is a hybrid research design in the context of the letter/incentive/no letter no incentive to sign up for the retirement plan? - correct answer In a hybrid design, there are multiple treatment and control departments, however, half of the treatment departments get the incentive, and the other half do not, but still know about event, then the control group doesn't know about it or get any incentive. How do you measure the spillover effect in the hybrid research design? What will this difference tell us about? - correct answer Compare the unincentivized employees in the treatment group's enrollment in the retirement plan to the control group's enrollment; this difference will tell us about the role of information--the more informed treatment employees probably will have a higher attendance and enrollment rate (can only be because of the increased information and knowledge about the event) What is a balance check? - correct answer Checking before an experiment which parts of your treatment, treatment with no incentive, and control groups are the same and which are different. What results would a naïve experiment see, and what results does a cluster randomized experiment? - correct answer Naïve sees the results for the incentivized treatment group and the unincentivized treatment group, which have no appreciable difference because of spillovers; Cluster sees both incentivized and unincentivized treatment groups as well as the CONTROL group--> the former 2 having a big gap between them and the control group. What is attrition? - correct answer When people drop out of an experiment/study.