UGBA 88 LECTURE 5: INTERFERENCE
AND ATTRITION *ADD TO THIS PLS
THX*
What 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
AND ATTRITION *ADD TO THIS PLS
THX*
What 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
