Revision week

This book is relatively short in comparison with other books in this module because of the deadline for the first assignment. In this book, I have included some suggestions for the second assignment. Please refer to the book regarding the first  assignment for more general advice on how to write your coursework.

Final coursework: relationship with topics already covered

In this chapter, we will discuss some important topics that might be important for the final coursework. I will provide you with some clarifications and suggestions. Similar to the first assignment, you can choose between two options. An option that is based on using STATA and another one that is not.

1. Option 1

1.1 What you have to do

This is the text from the assessment brief:

Using datafrom Compustat or Orbisfor a large sample of banks (at least 50) and over a period of at least 5 years, examine the determinants of bank dividendpayouts:

- Fama and French (2001) hypothesis

- Risk-shifting theory

- Signalling theory

- Life-cycle theory

As proxiesfor dividendpayouts, you can use any (or all) of thefollowing variables:

- Dividend-to-asset ratio

- Dividend-to-equity ratio

- Dividendpayout ratio

Or you can use dividend payout dummy variables asfollows:

- Dividendpay dummy

- Dividend increase dummy

- Dividend decrease dummy

- Dividend omission dummy

In your analysis, you can consider different dimensions/variables, e.g., bank capital, bank size, deposits, and any other,following the relevant literature. Moreover, you can use a variety of specifications (as many as you want, e.g., large banks and small banks or different periods). You should also discuss your results by comparing them with those in the relevant literature. Finally, you should discuss the potential pitfalls of the methodology used (if any).

Clearly, some of the issues covered in Weeks 9- 10 are relevant for this coursework. However, the main problem for this option is that dealing with data on directors in STATA is not very easy. For this reason, I have decided to provide you with some help below.

1.2 Using Compustat/CRSP data and STATA

In this section, I will provide you with details on how to capture the key theories you are required to address in Option 1. You are required to have your dependent variable either continuous (e.g., dividend-to-asset, dividend-to-equity), or binary (e.g., dividend pay dummy, dividend increase/decrease dummy), or both. The important thing is to capture your theories. Now let’s     explain each theory on its own and the way it can be captured.

1) Fama and French (2001) hypothesis states firms with large banks with high- profitability and low-growth rates tend to pay dividends, while smaller banks with low-profit/high-growth tend to retain profits. This means that in order to capture this theory, we need bank size, bank profitability and historical growth rate (e.g., assets growth).

Proftiability can be captured by return on assets (ROA) or return on equity (ROE); size can be captured by the natural logarithm of total assets; and    growth rate can be captured by taking the growth in total assets.

Remember, to calculate the growth rate, we use first difference of the natural log of total assets. That is (Stata code):

qbys (bank_id): gen AGrowth = ln(assets) – ln(l.assets)

Note that we use “qbys” to ensure that we create assets growth by bank.

Below are two papers showing evidence in favour of the Fama and French (2001) hypothesis.

B-Dividend payouts: Evidence from U.S. bank holding companies in the context of the financial crisis

2)  Risk-shifting theory states that dividends may transfer wealth from a bank  that is close-to-default to its owners (i.e., moral hazard). In order to capture this theory, we need to include ZScore. We can calculate it as follows:

ZScore = (return on asset+equity-to-assets)/standard deviation of return on assets

Note that ZScore is also called distance-to-default and it is inversely related to risk, which means that a higher ZScore indicates the banks is     solvent (safer), whereas a lower ZScore indicates insolvency (riskier). Therefore, if we find a significantly negative relationship between ZScore and dividends, it means that the risk-shifting hypothesis/theory holds.

Note that in order to calculate the standard deviation of return on assets we need to calculate it over more than two years to reduce any shocks biasedness. To do so, we use “rangestat” in Stata as follows:

rangestat (sd) ROA, by (bank_id) interval (fyear -2 0)

The above code calculate the standard deviation of ROA on a rolling-    window of 3 years. We can adjust this by changing the interval (fyear -2 0).

After calculating the standard deviation of ROA, we calculate ZScore as follows:

gen ZScore = (ROA+EA)/SD_ROA

(of course I am assuming that ROA and EA are already defined in your data.

Below are a few papers showing evidence in favour and against the risk- shifting hypothesis

A-Moral Hazard, Dividends, and Risk in Banks

B-Payout policy and ownership structure

C-Bank dividends, risk, and regulatory regimes

D-The risk-shifting value of payout

How to describe your results

You should always be consistent and concise when you are writing your results,  whilst ensure that no parts are repetitive. A standard writing style for interpreting a key variable in your study that you want to highlight may include the  relationship between the variable and the dependent variable (positive/negative), the magnitude of the effect, the interpretation of the result, and whether its in  line or contradict other studies. For example:

“The results in Table X shows that the ratio of retained earning exerts a significantly positive effect on dividend payouts. Quantitatively, a one percentage point increase in retained earnings increases banks’ dividends by approximately 0.005 p.p. This result suggests that mature banks distribute larger dividends consistent with the life-cycle theory, but the effect is not quite strong as the magnitude is only marginally differentfrom zero. These results are in line with ABC (20XX) whofinds evidence infavour of the life-cycle theory. ”

Of course, the above paragraph is just a standard paragraph and you may have your own style, but for now start with this style/order and then build your own by keep reading the literature and working on your writing.

2. Option 2

2.1 What you have to do

This is the text from the assessment brief:

Download the annual reports of two banks, “Bank X” and “Bank Y”,for the last 5 years. These two banks must be headquartered in the same country and must be competitors in at least one line ofbusiness (e.g., they engage in “retail banking” activities). Then, compare the two banks in terms of:

a) Overall performance and its main drivers

b) Riskprofile

c) Corporate governance structure

d) Dividendpayout policy (if any)

e) International activities (if any)

f) Funding strategies

g) Hedging strategies (if any)

For some of these items, you can find information on Compustat or Orbis. For  other items, you might have to look for the annual reports of the bank, BoardEx – for c) – or other information on the bank’s website. In the annual reports, you should be able to find even a short section on the economic outlook, and how    this might affect the bank’s profitability and risk. However, you are encouraged to look for other data in other sources. For example, you could download data   on macroeconomic variables, such as interest rates or inflation rates.

NOTE: you do not have to analyse all the variables mentioned below. For some banks, certain items might not be available. You can use numeric variables for a) and b), such as profitability ratios and risk proxies. I suggest that you compare the two banks not only in terms of the average value of these variables but even in terms of evolution over time.

Let’s start now discuss a) more in-depth.

2.2 a) Overall performance and its drivers

The key questions to answer for this section are: Which bank is more profitable? What could be the main reason for a different ROE (e.g., EOA o EM)? Is the         income structure of the two banks similar? For this reason, let’s recap what you     have already said in previous weeks.

As explained in Week 3, the main profitability measures employed in the literature are the Return on Equity (ROE) and the Return on Assets (ROA), which are related to each other because

The ROA can be decomposed even as follows:

The ratio Net Income/Total Revenue is usually called Net Profit Margin, while the ratio of Total Revenue/Average Total Assets is the Asset Turnover or Asset   Utilisation. Thus, the ROE becomes:

We also said that this is the terminology related to ratio analysis for nonfinancial firms. We then discussed ratios that might be relevant for banks, especially retail    banks, for example, the net interest margin, which is the ratio of the net interest      income (interest income minus interest expense) divided by average earning assets. Other related ratios are also available in the material for Week 3 (lecture slides and PDF book).

You can also cite academic papers to back up your arguments, just like for

Coursework 1. For example, the paper “Breaking the Bank? A Probabilistic Assessment of Euro Area Bank Profitability” (document

“jbf2020_ROAdeterminants.pdf”, in the folder “Other material”for Week 3)        examines the impact on ROE and ROE of variables mentioned in Week 3, such as the cost-to-income ratio and the equity-to-total-assets ratio – which is equal to 1   divided by the Equity Multiplier.

You might also discuss the impact of the business model on bank profitability. For example, the ratios loans-to-total-assets and deposits-to-total-assets can be     helpful to understand how important for that bank is the intermediation function. You might also go deeper and analyse the type of loans made by the bank. For     example, does the bank invest mainly in commercial and industrial loans or         mortgages?

For universal banks, investment banking activities, such as underwriting,          market-making, and advisory services (e.g., for M&A), might also be important, as well as insurance and portfolio management activities.

For this reason, I also suggest that you compare the income structure of the two banks in terms net interest margin and non-interest income ratio (which considers  commissions and fees).

Remember that for banks, an important item of the income statement is the loan loss provision (or provisions for loan losses). This is covered in sections 4.3-4.5 of the PDF book for Week 3. In particular, the expense ratio is:

Note that the interest expense depends on the funding strategy of the bank. For    example, insured deposits are, typically, a relatively cheap and stable source of       funding – I will talk more about this in the section related to the funding strategy of the bank (sectionf). You can also measure the ability of the bank to cover their       interest expenses using the so-called interest coverage ratio (Earnings Before        Interest and Taxes divided by Interest Expense).

The PLL, provisions for loans losses, can be a measure of credit risk – which is related to the bullet point b), as mentioned below – but they can be, to some extent, manipulated by banks so that earnings do not fluctuate too much – this practice is   commonly called “earning smoothing” . Thus, other items (e.g., nonperforming        loans) might be better proxies of credit risk.

However, the PLL is still important, and you might compare the average PLL   ratios (PLL/Average Total Assets) of the two banks over the last five years to       reduce the impact of earning smoothing practices. Moreover, rather than using the PLL ratio as a proxy for credit risk – which is the focus on bullet point b), discussed below – you could use it as a proxy for the overall importance of costs  related to lending activities. 1

Suggestions:

-    Plot graphs showing the evolution of the ROE/ROA and their main drivers over time for the two banks. Is the difference in profitability between the two banks likely to increase or decrease? Consider the economic outlook and how it might affect the business model of the two banks.

-    If you make general claims regarding the importance of certain items for     banks’ profitability, try to find some supporting evidence. For example, you could cite academic papers that support a significant relationship between   two or more key ratios that you have examined.

-    It would be helpful to show a diagram with the main activities of the two       banks and (if you can calculate it) the importance of each activity in terms of contribution to income. As an example, check p. 6 ofHSBC’s annual reports,available in ELE in Week 11 “Other Material” , and p. 139                  (“Segmental Analysis”).

Now, let’s talk a bit more about b).

2.2 b) Risk profile

As mentioned in Week 3, banks are subject to different types of risk. The banking literature tends to focus on measures of default risk (such as the Z-score, or other distance-to-default measures), as well as specific types of risk, such as credit risk

(usually proxied by thenon-performing loans ratio).

The Z-score may be calculated according to the following formula:

where SD(ROA) is the standard deviation of ROA over a certain number of       periods. Thus, it should be quite easy for you to calculate the Z-score for both        banks, considering a 5-year period, although you will not be able to measure its      trend over time. If you collect data for ROA and Equity/Total Assets for, say, 1015 years, you can use a 5-year moving window to see how the Z-score changes over    time for the two banks.

You can also examine more specifically different types of risk. In particular, in  the annual report of banks subject to international prudential regulation, the overall

exposure to credit risk, market risk, and operational risk should be reported (see p. 72 ofHSBC’s annual reports,available in ELE Week 11 in “Other Material”). For this reason, you might compare how the exposure to these types of risk has            changed over time and whether a specific type of risk is becoming more/less         important for the two banks or one bank more than for the other.

You can also find some proxies for different types of risk in the annual reports (or even in Compustat/Orbis), which you can use to compare the two banks over the sample period considered:

- Credit risk: in addition to the usual non-performing loans ratio (which you

can proxy using npat/lg in Compustat), you can also examine, if available, information on loan loss provisions, LLP (or provision for loan losses,      PLL), as mentioned above. For banks using IFRS, you can also analyse     specific categories of loans at different stages of the IFRS 9 loan loss        recognition process. For banks that do not follow IFRS 9, this information might not be available. Still, there might be a discussion on different types of non-performing loans and leases, and also information on charge-offs –

that is, payments that are considered uncollectable because the borrower has defaulted. For example, check the Bank of America Annual report on ELE,

“Other material”for Week 11).

- Liquidity risk: you can use proxies for liquid assets, and you can also            consider the impact of stable funding sources, such as customer deposits.      You can do this with data from Compustat/Orbis. For example, you might    use the ratio “Liquid assets/Customer and short-term funding”, as inthis paper,which also uses the ratios “Liquid assets/Interbank liabilities” and      “Liquid assets/Customer deposits” . These ratios are based on data from         Bankscope, but you should find similar times in Orbis if needed. You should also be able to find similar items in the annual reports of the banks, although they might be named differently. For example, the annual report2019 Annual Report - UK bank.pdfcontains data on “Retail deposits” and             “Shortterm borrowings”, which are basically customer deposits and short-    term funding (see p. 257 of the report). Clearly, this type of risk is related to the funding strategies of the bank (see bullet point e) of Option 2). Note that sometimes this type of risk is named “Treasury risk” (see p. 31 ofHSBC’s annual reports).

- Interest rate risk: Interest rate risk is a specific category of market risk.       Some banks might report the sensitivity of their net income to changes in   interest rates in their annual report. As an example, you can check Table 47 of theBank of America Annual report in ELE. In that Section, there is also an explanation of their hedging strategies, which you could use to answer   bullet pointf). Similarly, the annual report2019 Annual Report - UK bank.pdfalso contains a description of the interest rate risk policies (and

hedging) of the bank. If the annual report doesn’t say anything about interest rate risk exposure and management, you can measure the volatility in the net interest income over a certain period. For example, you might use         Compustat or Orbis to download data on interest income and interest expense to calculate the net interest income for 5- 10 years, and then you can calculate the standard deviation of the net interest income over this period.   You can also use quarterly data, if available.

Suggestions:

-    Plot graphs showing the evolution of the key proxies for bank risk. For        example, have the non-performing loans ratios for the two banks converged over the last five years, or have they diverged? What could be the reason for this?

-    As said above, you should make use of data on macroeconomic variables, if available. For example, for interest rate risk, you might want to check data    on the yield curve (see Week 4 material). We have used data from the US in class (availablein this folder), but similar data should be available for other  countries. For example, for the UK, you can go to the website of the Bank of England (click here).

2.2 c) Corporate governance structure

In this section, you should talk about the main features of the corporate               governance structure of the two banks and how this might affect their performance and risk. Thus, the key questions are: Can the corporate governance structure of the two banks explain differences in profitability/risk? Have there been any recent        changes in the board of directors, leading to changes in strategy?

There should be enough information on this in the annual report or on the bank’s website, in particular, in the section on investor relations. For example, for              Barclays PLC, there is a website for investor relations (click here), and in the          section “Who we are” you can find information on the board of directors (click on

“Leadership”). If you cannot find this information, try on BoardEx

(https://wrdswww.wharton.upenn.edu/pages/get-data/boardex/). Remember that you need to be registered on WRDS to access BoardEx, like for Compustat.

In Week 9 we discussed some issues related to corporate governance, and how    they might affect profitability and bank risk. In particular, we know from the           literature that board size might affect bank performance, as well as board                 composition (e.g., board independence and board diversity). We also examined in   class the issue of CEO duality, which occurs when the CEO is also the Chair of the board. The literature on corporate governance is huge and, although the literature    on corporate governance in banks is a bit less developed, you should be able to find plenty of academic papers to guide you.

In addition to board size, independence, and diversity, you could discuss even   executive compensation. For example, on p. 135 ofHSBC’s annual report,you can find information on the amount of compensation to directors, both in cash (salary)

and long-term incentives (e.g., shares). Of course, you can also look for                information in BoardEx, as long as the banks you have chosen are covered by this database.

Suggestions:

- Try to focus on the main features of the corporate governance of the two         banks (e.g., board size, CEO duality), and how they might explain                 differentials in performance/risk. However, do not neglect the importance of roles that might be less common. For example, there is evidence that the

presence of a Chief Risk Officer on the board might help bank performance,

especially during a financial crisis (click here).

2.2 d) Dividend payout policy (if any)

In this section, you should discuss the payout policy of each bank and how the   payout policy might affect their performance. For example, you may discuss the    motivation of each bank’s payout policy, whether they pay dividend to signal their performance or as a risk-shifting behaviour (moral hazard). Therefore, the key       questions are: Can the dividend payout policy of the two banks explain differences in performance (e.g., profitability/risk)? Have there been any recent changes in the payout policy (increase/decrease/omission), leading to changes in strategy and the bank’s future vision?

In Week 9 we discussed some important theories related to dividend payouts and we show that risk and growth can significantly affect dividend payout policy. In     particular, we know from the literature that dividends and earnings can drive each   other (smoothing theory and signalling theory). We also discussed the agency cost  theory, life-cycle theory, risk-shifting theory, Fama and French 2001 theory, and    catering theory. Can you pinpoint the theory(ies) that explain each bank’s dividend

payout policy. The literature on dividend payout contains adequate studies that       would be of help, although the literature on dividend payout in banks is a bit less    developed than non-financial firms, you should be able to find plenty of academic  papers to guide you. Please check Week 9 slides (click here), the last few pages      discuss each dividend theory. An example of where to find dividend details can be found inHSBC’