EC 303: Empirical Economic Analysis
Empirical Economic Analysis
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: THEend8_
EC 303: Empirical Economic Analysis
Due Thursday November 11, 11:59 pm.
1 Theoretical Problems
Problem 1 (10 points). Suppose X has a Uniform distribution with support [a, b]. Explain how you
estimate θ = (a, b) by method of moments (MoM).
Problem 2 (10 points). Suppose Xi ∼ Gamma(α, β). To see what gamma distribution is please refer
to page 195 of your textbook. What roles do these 2 parameters play? How do they change the shape of
distribution? Explain how you estimate θ = (α, β) by method of moments (MoM). (Hint: Use pdf of gamma
distribution to find the first and second moments and then construct sample analogs)
Problem 3 (20 points, 5 points each). Suppose we have the following estimator σˆ2 = c
∑n
i=1(Xi−X)2,
where c ∈ R is a constant that may depend on the sample size n.
a. Express the expected value of σˆ2 as a function of c, σ2, and sample size.
– Hint: Use the fact that
∑
(Xi −X)2 =
∑
X2i − (
∑
Xi)
2
n , and that E[Y
2] = V (Y ) + E[Y ]2.
b. What does the value of c need to be for this estimator to be unbiased?
c. If instead we use c′ = 1n , what would the bias of the estimator be? Does this biased estimator tend to
over- or under-estimate the true value σ2? How does the bias change as n increases?
d. Taking as given that V[σˆ2] = c2σ4[2(n− 1)], what value of c minimizes the MSE? What does this tell
you about your estimator?
Problem 4 (15 points, 5 points each). Work on the problems from the textbook chapter 7: 10, 16, 24.
All questions will be graded.
Problem 5 (20 points, 10 points each). Explain how you estimate λ and ν2:
a. Suppose X has a Poisson distribution with parameter λ. Explain how you estimate λ by using first
moment and then by using second moment. Are they the same?
b. Let {Y1, Y2, ..., Yn} denote an i.i.d. sample from a population with a F (ν1, ν2) distribution where ν1
and ν2 are the numerator and denominator degrees of freedom, respectively. If
E(Yi) = ν2/(ν2 − 2).
Derive an estimator of ν2 using the method of moments.
2 Empirical Problems
Problem 2: Comparing performances of estimators. In this problem, you are going to compare per-
formances of three estimators with stata simulations.
1
a. (5 points) Suppose true data generating process is normal with N(6, 3). Consider the following three estimators:
sample mean θˆ1 = X¯, sample median θˆ2 = X˜, and θˆ3 =
n+1
2n X
(n) where X(n) is the largest order
statistic i.e. maximum in the sample. Pick sample size (n) and number of samples or simulation
size (m). For the values you picked, how do these three estimators compare in terms of variances?
Which estimator is more variable? Which estimator is less variable? Rank them from the best to the
worst. Interpret your results. How do results change when you increase/decrease n and m? Interpret.
(Hint: In order to answer this question, you need to generate random samples of size n from the given
distribution and repeat m times.)
b. (5 points) Repeat the same process above when true data generating process is Uniform with U [0, 2]
c. (5 points) Is there any estimator that outperforms others in terms of variability (lower variance) in both cases?
Is there a single estimator that stands out in both cases?
d. (5 points) Now, repeat the same procedure in parts a and b with a fourth estimator θˆ4. Use 10 percent trimmed-
mean (ignore top and bottom 5 percent of the distribution) as an estimator. Compare performance
of 4 estimators with normal and uniform data generating processes. Rank them from the best to the
worst.
e. (5 points) Is θˆ4 more robust estimator? Why? How does it perform compared to best estimators in normal and
uniform cases?