Q1

Please concisely answer the following:

a. When starting an empirical exercise in Economics and Finance, we need to

produce a table of basic statistics for the variables used in the study. Briefly
describe the major purpose(s) of doing so.
[4 marks]
b. The fifth assumption of the Classical Linear Regression Model (CLRM) is the
normal distribution of the residuals. Briefly discuss the major problem(s) if the
assumption is violated.
[4 marks]
c. If you are studying New Zealand-listed companies (based on annual data) from
NZ 40 index from 2019 to 2022, describe how will you consider the normality
assumption in the progress of the study and explain why.
[4 marks]
[Total: 12 marks]
Q2
In theory we have a model which explains a company’s expected stock return:
Stock Return = α+β1*RiskFactor1 + β2*RiskFactor2* + β3*RiskFactor3 +ε

Now we have run a time series OLS regression and our main interest is to test if the
variable of RiskFactor1 has a beta value being greater than 1. The regression software
produced these results in the default printout: β1 equals -1.389, Standard Error=0.344,
T-value equals 4.360, and P-values 0.0233 (Degree of Freedom=1389)

2a. Describe in detail the null hypothesis and the alternative hypothesis. [4 marks]

2b. Describe the statistical reasoning of the test. (Your answer should include the required
assumptions so that there will be a specific statistic distribution of the estimated
coefficient and the decision rule.) [4 marks]

2c. Provide your testing result and conclusive statement of this hypothesis testing (under a
conventional significance level). [4 marks]

[Total: 12 marks]

Q3
Suppose we have a regression model:

For White’s general test for heteroskedasticity, we run the auxiliary regression:
uˆt2 = 1 +  2 x2t +  3 x3t +  4 x22t +  5 x32t +  6 x2t x3t + vt

3a. Describe the contents of the hypothesis tested in the auxiliary regression
[4 marks]
3b. If the residual variance is proportional to the value of a variable of X3, which is not a
variable in the regression model. Describe what might be the problem if you use this
White test to identify the heteroskedasticity problem.
[4 marks]

3c. Provide some ideas or suggestions to overcome the problem? (Explain why)
[5 marks]

[Total: 13 marks]

Q4.
We have a linear regression model: y= α + ß1 x1 + ß2 x2 + ε

4a. After we have collected data, we find that the correlation coefficient between
x1 and x2 is exactly negative 1. Discuss what you can do to deal with it.
[4 marks]

4b. If the correlation coefficient is actually 0.62, describe what you should do next
in the regression process.
[4 marks]
4c. What if the actual correlation coefficient between x2 and the residual ε is 0.42,
describe what you should do to identify it and deal with it in the regression
process.
[5 marks]
[Total: 13 marks]
Q5.
5a. The table below shows the out-of-sample performance of the credit scoring model.

The customer defaults the The customer does not

loan default the loan
Model predicts the customer 30 120
will default
Model predicts the customer 100 750
will not default the loan

Calculate the model accuracy rate. Note that the accuracy rate is the ratio between the
number of corrected predictions to the total number of predictions. Discuss whether this
model performs better or worse than the ‘naïve’ predictive model. The naïve model makes a
prediction based on the aggregate default proportion of overall data only. In other words, the
naïve model predicts that all applicants do not default on the loan.
[4 marks]

5b. Discuss the advantages and weaknesses of the logit model compared to the OLS
linear probability model.
[5 marks]

5c. Calculate the two probabilities of loan approval when your salary is 90,000 dollars
per year based on the two models below.

Logit model: Y = 0.55 + 0.11*log(X)

OLS linear probability model: Y = 0.40 + 0.03*log(X)

Where Y is a binary variable that takes a value of 1 if the loan is approved and 0 otherwise; X
is the annual salary in dollars; log(.) is a natural log function.
Note that the cumulative probability of the logit function is 1/(1+exp(-Z)), where Z is the
quantile value.
[5 marks]
[Total: 14 marks]
Q6.
6a. Air pollution can adversely affect cognitive ability. You are researching whether
an increase in air pollution can increase the analyst forecast error. If you are using the two-
stage-least-square (2SLS) approach and you want to use the amount of rainfall as an
instrumental variable, carefully discuss whether this instrumental variable is appropriate.
[4 marks]

6b. Below is the results from Brogaard, Li, and Xia (2017) Panel A of Table 5:
Difference-in-differences analysis of the effect of stock liquidity on default risk. Carefully
discuss the meaning of the “Pre-match (1)” and the “Post-match (2)”. What can we conclude
from Panel A of Table 5?

[5 marks]

6c. Discuss the implication of the regression estimation when the “dependent
variable” has a measurement error problem.
[3 marks]
[Total: 12 marks]
Q7.
7a. The time-series model of stock return, Rt, is the AR(1) model with the estimated
coefficients as follows: Rt = 0.89Rt-1 + ut. Check the model for stationarity. Show the work to
support the conclusion.
[3 marks]
7b. Discuss when should the Autoregressive model (AR) is more appropriate than the
Moving Average (MA) model and when should the MA model is more suitable than the AR
model.
[4 marks]
7c. Let Y1t, Y2t, Y3t are endogenous variables and X1t, X2t, X3t are exogeneous variables.
Given the system of equations below, write down the reduced form equations and determine
whether the system of equations is identified based on the order condition.

Equation 1: Y1t = a0 + a1Y3t + a2X3t + u1t

Equation 2: Y2t = b0 + b1Y1t + b2Y2t + b3X1t + b4X2t + u2t
Equation 3: Y3t = c0 + c1Y2t + c2X2t + c3X3t + u3t
[5 marks]
[Total: 12 marks]

Q8.
8a. Discuss the situation when the fixed effect panel regression is the most appropriate
when it is compared with the pooled OLS regression and the random effect regression. What
test should be done? What to do when the testing hypothesis is rejected or is not rejected?
[4 marks]

8b. Based on the Vector Auto-Regressive model (VAR), carefully explain when and
how should we use the impulse response function (IRF). Also, discuss when should we use
the variance decomposition analysis.
[4 marks]

8c. Carefully explain how to calculate the earning announcement CAAR(-1,+1) which
is the cumulative average abnormal return surrounding the period one day before and after
the earning announcement. This CAAR(-1,+1) is based on all 500 firms in the S&P500 index.
[4 marks]
[Total: 12 marks]