2. Anti-capital punishment activists often point to research showing no statistical relationship between murder rates and execution rates. Consider the model
mrdrteit = αt + β1 execit + β2 unemit + ci + uit
where mrdrteit is the per-capita murder rate in the state I at time t; αt is shorthand for a vector of time dummies; execit is the number of executions in a state I in the current year t plus the last two years, and unemit is the unemployment rate in the state I at time t.
(a) Assuming that ci does not equal 0 and x = {exec, unem} is strictly exogenous, how would you decide which linear panel data estimator to use?
(b) Do you have any concerns about relying on the asymptotic properties of Î²Ë for this estimation? Why or why not?
(c) Under what circumstance would exec fail to be strictly exogenous conditional on ci?
2. Anti-capital punishment activists often point to research showing no statistical relationship between murder rates and execution rates. Consider the model
mrdrteit = αt + β1 execit + β2 unemit + ci + uit
where mrdrteit is the per-capita murder rate in the state I at time t; αt is shorthand for a vector of time dummies; execit is the number of executions in a state I in the current year t plus the last two years, and unemit is the unemployment rate in the state I at time t.
(a) Assuming that ci does not equal 0 and x = {exec, unem} is strictly exogenous, how would you decide which linear panel data estimator to use?
(b) Do you have any concerns about relying on the asymptotic properties of Î²Ë for this estimation? Why or why not?
(c) Under what circumstance would exec fail to be strictly exogenous conditional on ci?
