QUESTION 2 - DOMESTIC VIOLENCE (a) In general, why is randomization useful? (4 points) (b) Other than partial compliance, spillovers, evaluation-driven effects, and attrition, provide several downsides of the randomization approach. (4 points) (c) There has been much debate about whether hardline responses to domestic violence (i.e., arrests) are productive. What makes it hard to provide convincing evidence on the impact of different response strategies on re-offense rates? (4 points) Suppose you conduct a randomized experiment to evaluate the impact of arrests (treatment) as opposed to advising or separation (control) on re-offense rates. You provide randomly shuffled color-coded forms to officers. If the color of the form that is being used for a particular case is pink, then the suggested strategy is to arrest; if the color of the form that is being used for a particular case is yellow, then the suggested strategy is to advise; if the color of the form that is being used for a particular case is orange, then the suggested strategy is to separate). (d) Provide a brief anecdote or story to explain what always-takers, never-takers, compliers, and defiers are in the context of this experiment. Only use language that a policymaker who has not taken statistics can understand. Do not use words like "counterfactual or treat or control. (12 points) Table 1. Assigned and delivered treatments in spousal assault cases. Delivered treatment Coddled Assigned treatment Arrest Advise Separate Total Arrest Advise Separate Total 98.9 (91) 17.6 (19) 22.8 (26) 43.4 (136) 0.0 (0) 77.8 (84) 4.4 (5) 28.3 (89) 1.1 (1) 4.6 (5) 72.8 (83) 28.3 (89) 29.3 (92) 34.4 (108) 36.3 (114) 100.0 (314) Note: Sample sizes are in parentheses. (e) Assuming no defiers, what are the shares of always-takers, never-takers, and compliers? Show your calculations. (12 points) (f) Suppose you estimate the LATE from this randomized experiment. What are some reasons why the LATE might be larger than the population ATE? What are some reasons why the LATE might be smaller than the population ATE? (8 points) QUESTION 2 - DOMESTIC VIOLENCE (a) In general, why is randomization useful? (4 points) (b) Other than partial compliance, spillovers, evaluation-driven effects, and attrition, provide several downsides of the randomization approach. (4 points) (c) There has been much debate about whether hardline responses to domestic violence (i.e., arrests) are productive. What makes it hard to provide convincing evidence on the impact of different response strategies on re-offense rates? (4 points) Suppose you conduct a randomized experiment to evaluate the impact of arrests (treatment) as opposed to advising or separation (control) on re-offense rates. You provide randomly shuffled color-coded forms to officers. If the color of the form that is being used for a particular case is pink, then the suggested strategy is to arrest; if the color of the form that is being used for a particular case is yellow, then the suggested strategy is to advise; if the color of the form that is being used for a particular case is orange, then the suggested strategy is to separate). (d) Provide a brief anecdote or story to explain what always-takers, never-takers, compliers, and defiers are in the context of this experiment. Only use language that a policymaker who has not taken statistics can understand. Do not use words like "counterfactual or treat or control. (12 points) Table 1. Assigned and delivered treatments in spousal assault cases. Delivered treatment Coddled Assigned treatment Arrest Advise Separate Total Arrest Advise Separate Total 98.9 (91) 17.6 (19) 22.8 (26) 43.4 (136) 0.0 (0) 77.8 (84) 4.4 (5) 28.3 (89) 1.1 (1) 4.6 (5) 72.8 (83) 28.3 (89) 29.3 (92) 34.4 (108) 36.3 (114) 100.0 (314) Note: Sample sizes are in parentheses. (e) Assuming no defiers, what are the shares of always-takers, never-takers, and compliers? Show your calculations. (12 points) (f) Suppose you estimate the LATE from this randomized experiment. What are some reasons why the LATE might be larger than the population ATE? What are some reasons why the LATE might be smaller than the population ATE? (8 points)