SMM248 Statistics for Finance 2022
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
MSc Corporate Finance
MSc Finance
Alternative Timed Assessment
SMM248
Statistics for Finance
2022
QUESTION 1
The excel file CAPM1.xlsx contains quarterly data covering 30 years (T=120 quarterly observations) for the excess returns of IVECO plc (yt ; iveco), the excess returns of the market portfolio (x2t; maTket) , IVECO’s sales (x3t; sales) , and IVECO’s debt ( x4t; debt) . Both sales and debt are measured in thousands of US dollars. A researcher estimates the following empirical asset pricing model which can be considered as an extension of the CAPM.
yt = F1 + F2x2t + F3 log(x3t) + F4 log(x4t) + et (1)
where log denotes the natural logarithm.
(a) Define the concepts of sensitivity and elasticity of a random variable y with
respect to another random variable x. What is the sensitivity of IVECO excess returns to the market excess returns? And the elasticity? What is the sensitivity of IVECO excess returns to debt? And the elasticity? Discuss potential reasons why the researcher might have formulated the model using logarithmic values of sales and debt instead of raw sales and debt.
(10 marks)
(b) A salient financial event has occurred in the first quarter of year 10th of the
sample period and this could have altered the relationship between IVECO’s excess returns and some of the above variables. Conduct a test at the 5% significance level of the hypothesis that the model suffers from a structural break. Write the null and alternative hypotheses. What is the name of the test? What test statistic is used? What is the outcome of the test? Draw a graph that shows the probability distribution of the test statistic, the sample value of the test statistic, the 5% critical value and the p-value.
(10 marks)
(c) Inspired by finance theory, a researcher conjectures that the marginal effect of sales on IVEC0 excess returns is not constant but instead depends on its level of debt. Write down: (i) the re-formulation of regression model that is needed to test for this effect using sales and debt in levels (that is, the logarithmic transformation no longer appears in the model); (ii) the expression of the marginal effect of debt on IVEC0 excess returns (iii) the null hypothesis and the alternative hypothesis for this test; (iv) the name and expression of the test statistic, and its probability distribution. Does the data support this conjecture? (10 marks)
(d) Conduct two different tests to assess whether the model is correctly specified (model diagnostics): (i) RESET test, (ii) Breusch-Godfrey LM test with maximum lag 4 quarters. Write down the null and alternative hypothesis for each test, and discuss the test outcome. Plot the correlogram using maximum lag order 12, what does it suggest?
(10 marks)
(e) Conduct a test to assess whether the sales and debt factors altogether add
significant explanatory power to the baseline CAPM? Write down the null and alternative hypothesis of the test, and represent in a graph the sample value of the test statistic, the 5% critical value and p-value. Discuss the test
outcome. (10 marks)
(Total: 50 marks)
QUESTION 2
A researcher has information on US banks, pre- and post-financial crisis. Some banks had large investments in subprime mortgages just before the crisis started, while others did not invest in subprime. She wants to study the impact of the financial crisis on the share prices of banks that had large investments in subprime mortgages, compared to those that did not. She collects monthly data for N=40 banks over 24 months, 12 preceding the collapse of Lehman Brothers and 12 after this event. The following OLS regression results are obtained for two models:
Dependent variable: log of Share Price
Regressors: (1) (2)
Subprime Postcrisis |
0.103 (0.081) 0.159 |
0.101 (0.073) 0.138 |
Subprime*Postcrisis Five Other Controls |
(0.061) -0.053 (0.021) No |
(0.059) -0.055 (0.019) Yes |
Observations R- |
600 0.120 |
600 0.243 |
Note: Standard errors in parentheses. The regressions include a constant term, which is not reported. “Subprime” and “Postcrisis” are dummy variables. “Subprime” equals 1 if the bank engaged in subprime mortgage lending, 0 else. “Postcrisis” equals 1 in post-crisis period, 0 else.
(a) Write down the regression equation that corresponds to the regression
results in column (1), labelling the coefficients as F0 (intercept), F1 , etc. Given the regression that the researcher estimated first in column (1), list some reasons why the researcher might have been motivated to estimate
the second regression in column (2)? (3 marks)
(b) The R-squared of the model estimated in column (2) is larger than that of
the model estimated in column (1). What does this suggest? (4 marks)
(c) What is the expected differential log share price pre-crisis of treatment banks (those that invested subprime) versus baseline or control banks (those that did not invest subprime). (7 marks)
(d) The differences-in-differences (diff-in-diff) estimator in this context measures the difference between the expected differential log share price pre-crisis of treatment (subprime investing) banks versus control banks (no subprime investing) and the expected differential log share price post-crisis of treatment versus control banks. Which parameter in the model represented in column (1) represents this diff-in-diff estimator? Interpret the sign and statistical significance (or lack thereof) of the diff-in-diff estimator.
(11 marks)
(Total: 25 marks)
QUESTION 3
The recruitment firm ALTA plc has been commissioned to conduct a study of the factors that are most likely to influence the probability that an MSc in Finance graduate student at the University of London receives a first-job offer from a top firm. ALTA categorizes any firm listed in the FTSE100 index at any point in time as a “top firm”, and it assumes for simplicity that all students that receive a first-job offer from a top firm accept it. ALTA has access to a historical database of students (alumni) that contains the following information for each: graduation mark (if “First class” degree, 1, otherwise 0; first_grad) dummy, top job placement (if “top firm”, 1, else, 0; top_job), gender (male 1, female 0; gender), and age at the time of graduation (age). A logit
regression is estimated with these variables for N = 125 students.
The estimation results by maximum likelihood are as follows:
Dependent variable: top_job Method: Binary logit (quadratic hill climbing) Sample(adjusted): 1 125 Included observations: 125 after adjusting endpoints Convergence achieved after 7 iterations |
|||||
Variable |
Coefficient |
Std . Error |
-Statistic |
Prob . |
|
C |
7.62037 |
2.86775 |
2.65726 |
0.0079 |
|
first_grad |
0.45339 |
0.13939 |
3.25270 |
0.0011 |
|
gender |
0.02364 |
0.01639 |
1.44184 |
0.1550 |
|
age |
-0.24657 |
0.24445 |
-1.005194 |
0.3148 |
|
LR statistic (3 d.f.) Probability (LR stat) |
47.83854 2.30E-10 |
McFadden R-squared 0.585978 |
(a) Calculate the probability that a male graduate who is 23 years old at the
time of graduation and is awarded a first-class MSc degree receives a first- job offer from a top firm .
(5 marks)
(b) Conduct a one-sided test for the hypothesis that male graduates are more
likely to receive a first-job offer from a top firm ceteris paribus (Note: use a one-sided test). Write down the null and alternative hypothesis. Draw a graph of the probability distribution of the test statistic used that contains the sample value of the test statistic, the 5% critical value and the p-value.
(5 marks)
(c) Test the null hypothesis that the probability of receiving a top-firm job offer is unrelated to the MSc degree award, gender and age. Write down the null and alternative hypotheses and discuss the outcome. (5 marks)
(d) Why is it not appropriate in this context to use the conventional coefficient of determination R2 ? Interpret the McFadden pseudo-R2 value reported in the estimation output, what does it measure? Why is the McFadden measure called “pseudo” R2 ?
(5 marks)
(e) How would you modify the model so that it is possible to test the conjecture
that the marginal effect of a one unit increase in age on the probability of receiving a top-job offer is different for males and females. Write down the
model equation. (5 marks)
(Total: 25 marks)
2022-08-20