Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

BSE · Topics in Applied Econometrics for Public Policy · 2022/23

Assignment 2 · Second Part

Deadline:  Thursday  15 of June at  2 p.m.   (before the  TA  session).   To be submitted in groups of 4.

This problem set is based on Stevenson and Wolfers (2006), who estimate how uni- lateral divorce changed family violence and whether the option provided by unilateral divorce reduced suicide and spousal homicide. To estimate the effect the authors exploit the variation occuring from different timing of divorce law reforms across the United States.  Unilateral divorce both potentially increases the likelihood that a domestic vio- lence relationship ends and acts to transfer bargaining power toward the abused, thereby potentially stopping the abuse in extant relationships. In states that introduced unilateral divorce they find a 8– 16 percent decline in female suicide, roughly a 30 percent decline in domestic violence for both men and women, and a 10 percent decline in females murdered by their partners.

Data. Divorce example.dta contains a subsample of the data used in their paper.

1. Estimate the static TWFE coefficient to measure the effect of law reform on suicidal rates. What is the coefficient of interest recovering?

2. Estimate the Bacon decomposition.  In the light of the results, complement your previous answer.

3. Estimate a TWFE excluding:  (i) already-treated, (ii) never-treated, (iii) both. In- terpret your results.

It is well known that the coefficients of the static TWFE diff-in-diff are contaminated. We will now try three different ways to estimate consistent coefficients of the actual effects.

1. Use the Callway Sant’Anna 2021 method to get consistent estimates of the treatment effects.   Interpret the coefficients.   Do it both conditional and unconditional on covariates.

2. Use the method proposed by Sun and Abraham 2021 to obtain the dynamic spec- ification (a.k.a.  event study).  You can balance your panel by just considering 5 pre-treatment periods and 11 post-treatment periods.  What are the coefficients reporting? Why is it better in comparison to the ”baseline” event study?

3. Finally, apply the imputation method proposed by Boruasky et al. 2021 to estimate the event study specification.

4. Comment the main differences of three methods and under which conditions they are the optimal choice.