MSc Accounting and Finance
Dr. Wenjie Ding
Lecturer in Finance
B28 Aberconway Building
Objectives of the coursework
❑ You demonstrate the ability to conduct empirical analysis
❑ With reasonable structure and proper referencing styles.
❑ You are able to use EVIEWS for data analysis.
❑ You show understanding of the theories behind the empirical tests.
❑ You are able to interpret the findings in relation to the hypotheses and theories.
❑ You are able to present the analysis and results in an effective, concise and neat way.
The coursework has three parts
Part 1 (40%):
Testing the CAPM and Fama-French three-factor model using time-series
regressions (US data)
Part 2 (40%):
Testing the determinants of audit fees in the UK
Part 3 (20%):
❑ Please cover Parts 1 and 2 equally.
❑ Plan carefully and use wisely the space (2500-3500 words).
Requirements For Parts 1 and 2
❑ Explain the theories and models being tested.
❑ State clearly and explain the null and alternative hypotheses.
❑ Explain and summarize your data,
❑ e.g., what industry portfolio you use, how many months, means, median, stdev, etc.
❑ Estimate regressions and present the findings
❑ Create your own tables, do not use screenshots for tables.
❑ Interpret the findings in relation to the hypotheses and theories.
❑ Perform diagnostic tests to evaluate goodness of fit of your regressions.
❑ Perform analysis of residuals (it is required!)
❑ Explain the possible bias.
❑ Demonstrate how these potential bias could be corrected.
❑ Compare the results before and after the adjustment.
❑ Any other data, tests, or information that you feel are interesting and relevant.
Requirements For Part 3
❑ Finally, you are required to critically compare quantitative research with
qualitative research, in terms of philosophical stances, relationship
between theory and data; epistemological considerations; and
❑ Before the evaluation, you must first define the concepts (i.e., quantitative
research and qualitative research) with reference to the relevant literature
on research methods.
❑ For higher marks, you need to give some good examples where
quantitative/qualitative research methods are more suitable.
❑ Part 3 should not exceed 800 words.
Presentation style and format
❑ While there is no strict requirement of a presentation style and format
❑ A good starting point: 12pt, Times new roman, 2x or 1.5x line spacing, and justified text.
❑ The key is to report the findings in an effective, efficient and neat manner.
❑ Follow academic journal conventions, e.g., table format, structure of the
analysis, writing style, etc.
❑ See my examples of presentation styles at the end of this PPT.
Presentation style and format - Tables
❑ Create your own table
❑ Do not just copy and paste the EVIEWS output to the word document.
❑ Insert the tables in the main text
❑ Label your tables, e.g., Table 1. Summary statistics.
❑ EVIEWS output gives many statistics.
❑ Only report those relevant to your hypothesis testing and the empirical analysis.
❑ You must decide what to include and how to organize the table.
❑ Text within tables do not count towards the final word count (minus them).
❑ Put more statistics into one table to save space!
❑ Use ***, ** and * to denote statistical significance (Important!).
Borrowed from Baek et al. (2004)
❑ You could report different regressions in different columns.
❑ The ***, ** and * superscripts denote statistical significance at 1%, 5%, and 10% respectively.
❑ Must report coefficients. T-statistics or p-values are reported in () or  under each coefficient.
Borrowed from Cooper et al. (2003)
❑ T-statistics are reported in ( ) under each coefficient.
❑ Regressions with different dependent variables are reported in different columns.
❑ Focus on how academic research tend to format their tables and how the table are
named: Table 4 with a title that explains what this table is about.
Borrowed from Galema et al. (2008)
❑ T-statistics reported in new columns.
❑ Different panels for different sets of
❑ Two regressions reported in each
panel, as columns (1) and (2).
❑ Think about including the Adjusted
R-squared, F-statistics, or other tests,
such as analysis of residuals?
Check list for Part 1:
1. Have you explained the CAPM and FF-3 model with reference to the literature?
2. Have you explained your data sample? e.g., which industry portfolio are you using? What is the
sample period? Reported summary statistics?
3. Have you stated your hypotheses and explained your prediction according to CAPM and the FF-
4. Have you created your own tables to report your results?
5. Have you interpreted your regression results correctly?
6. Have you discussed about diagnostic tests?
7. Have you checked the residuals for heteroscedasticity and autocorrelation? Have you explained
what they are and how would they affect your results?
8. If you have detected these problems, have you estimated regressions using adjusted standard
errors or other adjustments?
9. Have you reported the adjusted results and compare before and after adjustment?
10. How does the adjustment affect your results? Do your main conclusion remain similar after the
Check list for Part 2:
1. Have you explained the Audit Fee model? Have you explained why it is important to study audit fees? Have you
explained what the main variables are with reference to the literature?
2. Have you explained your data sample? Reported and explained your summary statistics?
3. Have you stated your hypotheses and explained your rationale for such prediction with the support of the literature?
4. The hypotheses for the Audit Fee model can be given using a sentence, e.g.:
❑ H1: Firm size is significantly and positively related to audit fees.
5. Have you created your own tables to report your results?
6. Have you interpreted your regression results correctly? (See Worksheet 2 of the workshop), e.g., the coefficient of firm
size is positively related to audit fees, significant at the 1% level.
7. Have you discussed about diagnostic tests? How much variations in audit fees are explained by your model (look at
8. Have you checked the residuals for heteroscedasticity? Have you explained what they are and how would they affect your
9. If you have detected these problems, have you estimated regressions using adjusted standard errors or other
10. Have you reported the adjusted results and compare before and after adjustment?
11. How does the adjustment affect your results? Do your main conclusion remain similar after the adjustment?
12. Would there be a problem of multicollinearity between the independent variables? E.g., between log assets and log sales?
How do you deal with it?
13. Can you estimate different model specifications? Start from a smaller model (maybe only firm size?) and then adding
more independent variables?
Potential weaknesses or problems
❑ The explanation of the findings are too repetitive and mechanical.
❑ Students not relating the audit fee variables to the literature.
❑ No summary statistics or relevant discussions on the data used.
❑ Have not explained what industry portfolio is used.
❑ Did not perform analysis of residuals.
❑ Did not perform tests to adjust for heteroscedasticity or autocorrelation.
❑ Used bullet points or numbered lists!
Hints for getting higher marks
❑ If you aim for distinction or higher marks, you must spend extra effort and do
extra analysis or tests.
❑ Any other data, tests or information that you feel are relevant to the empirical
❑ Additional tests of residuals? And why?
❑ Additional tests of some related theories?
❑ Extra analysis using different data?
❑ What else can you do with EVIEWS?
CAPM - Some more tests you may consider?
❑ Are there any other macroeconomic variables that may
explain portfolio returns in addition to the CAPM and
❑ Corporate bond yield spreads as a measure of market default risk
❑ Interbank rate spreads, e.g., LIBOR-OIS as a measure of funding illiquidity.
❑ Monthly changes in US CPI
❑ Monthly changes in US non-farm payroll, as a measure of employment
❑ Monthly changes in US GDP
❑ Monthly changes in US/EUR Exchange rates
❑ Some more?
Data can be found here: http://research.stlouisfed.org/fred2/
CAPM - Some more tests you may consider?
❑ Subsample analysis? Does the model work better in a particular
sample? E.g., before 1996 and after 1996? Any structural break in the
data, using Chow test of structural breaks? How does the break affect
the performance of regression?
❑ Control for outliers (extreme negative returns) during the financial
crisis. Create a dummy variables called crisis, which denote 1 during
the financial crisis and 0 otherwise and use that as an independent
❑ Would the regression model perform better?
❑ When were the financial crises? http://www.nber.org/cycles.html
❑ Use other empirical models to fit to the portfolio returns, e.g., GARCH
models, AR, MA, ARMA models? And why?
Audit Fees – some other tests you may consider?
❑ Can we measure the relationship between auditor choice and other
firm characteristics? Yes – use a regression
❑ But – the LHS variable includes only 1’s and 0’s
❑ This type of regression is known as a Logit Regression
❑ Fitted values measure the probability of a firm using a Big 4 firm as an
❑ βi’s measure the effect of the variable on the probability of a firm
choosing to use a Big 4 auditor
Read this for an easy introduction of Logit regression: