EC3380 Econometrics 2: Microeconometrics 2018/19
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
EC3380
Summer Examinations 2018/19
Econometrics 2: Microeconometrics
1. Define wi = 1 if individual i received a treatment, and wi = 0 otherwise. According to the potential outcomes framework, y1i is the outcome of individual i if she/he gets the treatment and y0i is the outcome of individual i if she/he does not get the treatment. |
|||||
Consider the difference: |
|||||
D = B(y1i~wi = 1) _ B(y0i~wi = 0) |
|||||
where B is the expectation operator. |
|||||
(a) Can B(y1i~wi = 1) and B(y0i~wi = 0) be observed from the data? (3 marks) |
|||||
(b) What is the interpretation of the counterfactual term B(y0i~wi = 1)? Is it observed |
|||||
directly from the data? (3 marks) |
|||||
(c) Decompose D into the effect of the treatment on the treated and selection bias. |
|||||
(5 marks) |
|||||
(d) How does random assignment solves the selection bias? (3 marks) |
Preamble to Questions 2-6: Tackling climate change is one of the most important issues of our
times. In the following questions, you will be asked to consider how microeconometric methods may
be applied to evaluate the effects of climate change, as well as to evaluate the effects of policies that governments worldwide are implementing to preserve forests and the nature.
2. Between July and November 2018, wildfires engulfed the state of California in the United States. It is known that wildfires dramatically increase the number of air pollutants. In fact, an article in the New York times on November 18th 2018 entitled “Air Quality in California: Devastating Fires Lead to a New Danger” mentioned that:
“The wildfires that have laid waste to vast parts of California are presenting
In this question, you will assess the suitability of the Differences-in-Differences method to
analyse the effects of wildfires on health.
The outcome variable yit is the number of patients admitted to the hospitals in county i at time t due to respiratory diseases, typically caused by air pollutants. The explanatory variable xit is equal to 1 if county i experienced wild fires at time t, and zero otherwise.
(a) Write the differences-in-differences regression. Note that different counties may experience wildfires at different points in time. (5 marks)
(b) Spell out the main identification assumption of the Differences-in-Differences method.
(4 marks)
(c) How could you verify if the identification assumption above is likely to hold? In particular, which regression model could you implement to verify if the data supports that assumption? (4 marks)
(d) What alternative regression models could you implement if the data does not support the main identification assumption of the Differences-in-Differences method? (3 marks)
(e) Pollution from wildfires travel through air, and may affect neighbouring counties to those that experienced wildfires. This is referred to as “spillover effect” since the control units suffer partially the effect of the treated units. Are spillover effects a threat to identification of treatment effects? If so, what would be the direction of the bias?
(3 marks)
3. Standard or classical standard errors assume that eit ~ N(0, σ2 ), where N (µ, σ2 ) is the Normal distribution with mean µ and variance σ 2 , and independent across units. Under heteroskedasticity, the variance of the error σ is allowed to vary with i, that is,
eit ~ N(0, σi(2)), but it is still assumed as independent across units. In EC338 you have seen a third option: clustered standard errors.
(a) Describe what clustered standard errors are, and how it generalizes the hypothesis that errors are independent across units. (5 marks)
(b) Describe how hypothesis testing can be affected if the units suffer common shocks but the researcher utilizes the classical standard errors instead. (3 marks)
(c) Should clustering be implemented in the forest fire application in Question 2? More specifically, which level of clustering could be appropriate? Justify your answer. (Hint: there is more than one answer to this question. Any answer will be accepted as long as appropriately justified.) (3 marks)
2022-05-25