The purpose of this section is to define the dependent variable. As multiple options exist, the main choices are outlined and evaluated. The section ends with the choice of the dependent variable; the risk measure used in the data analysis.
7.1 Assembling risk in one measure
As described in the introductory part, the claim against the Danish banks in general is excessive risk-taking. To test the risk-taking in Danish banks against board composition and structure, a proper risk measure has to be selected, as any measure will be an estimation signal that implies, but does not directly and completely signal, risk-taking.
Multiple options exist, each with their own advantages and shortcomings, which will be evaluated in the following before settling on a risk-measuring method. The choice will depend on the trade-off between accuracy in measuring the individual bank’s risk-taking and the possibility to generally use a measure to infer risk across the sample.
In other words, the trade-off will be one between validity (is the measurement actually testing what should be tested) and (internal consistency-) reliability (can the measurement be appropriately used on all items in the sample) (Bryman, Bell 2003)
General criteria for determining whether a measure is reliable and valid will in this case be that it accurately reflects risk. Therefore, the bank’s liabilities are revisited (fully described in section 3.4): simplistically stated; a bank with more loans than deposits needs to finance this gap in the interbank market. If it cannot, then the bank may raise capital in the stock market, but if that proves unsuccessful then the bank will almost certainly cease to exist. This risk is what the thesis is aimed at describing some of the dynamics behind and therefore, a foundation for choosing a risk measure is that it somehow encapsulates the described mechanisms.
It is acknowledged that other samples may very well yield the need for a different risk measure than the one chosen in this study.
7.2 Stock volatility
While the term ‘risk’ might have different meanings to different groups of people, it is commonly understood as (the possibility of) losing for example an investment. Modern portfolio theorists do not, however, define risk as a likelihood of loss, but as volatility to make use of the market’s combined information and responses to changes in forecast outlooks for a particular company. Volatility, it is argued, translates as shifting outlooks for the business
environment, the business’ risks (credit, market, operational etc.) (Keppler 1990). The finance literature generally supports this claim, and Bertsimas et al. (2004) state that the standard deviation of the return of a portfolio is the predominant measure of risk in finance.
Therefore, this paper would prefer the standard deviation on the individual bank’s stock as its risk measure. However, only 41 of the 67 banks in the sample are listed and traded. Using only those 41 banks would leave the sample less valid; not only because fewer banks would be in the sample, but also because the listed status of the bank may convey information on the bank that would lead to selection bias. For example, if listed banks have more financial directors on the boards than do non-listed banks, the results of the data analysis would suffer from being extrapolated from a smaller sample and because it would have been taken from a biased sample. Therefore, the volatility of the banks’ stocks as measured by standard deviation is not used as this study’s risk measure.
7.3 Absolute growth in loans
Another approach to define risk is to measure risk as the absolute, numeric growth in loans in a given time period. This measure would be deployable for all banks in the sample and thus fulfill the reliability criteria. Also, it is a well-known risk measure: the media frequently cites the absolute growth in loans as an example of ‘reckless’ behavior on the banks’ part, as this number is generally the more extreme when employing risk measures in banking (examples:
(TV2 2008) and (Politikken 2008). The issue with using this risk measure is that it does not account for the absolute risk-level in the bank, because it does not consider deposits at all and because the numeric values can be misleading (not counting inflation, for example).
Therefore, the absolute growth in loans is not used as a risk measure, either, as it is neither a valid description of the actual risk of going into financial distress nor a particularly reliable variable.
7.4 Annual relative increase in loans
Risk-taking could also be measured as the arithmetic mean of the annual percentage increases in loans given by each bank. This measure is used by the Danish Financial Supervisory Authority (Finanstilsynet 2010) (among other parameters) to describe a bank’s riskiness. It therefore presents an obvious choice to pick, but does not solve the issue of not accounting for deposits, which this thesis wants to incorporate. The benefit of using this risk measure is that it accurately depicts the organizational dynamics which occur in the bank when the amount of loans increases, in some banks, rapidly. The circumstantial evidence point in the direction that this has been a contributor to the failure of some Danish banks in the period 2003-2008: The
lawyers commissioned to investigate Roskilde Bank criticize in their report the fact that the credit evaluation on new clients and projects has been lax and sub-standard and attribute this to the aggressive growth strategy of this particular bank. The same scenario is outlined by independent investigators for e.g. Løkken Sparekasse (Skipper-Pedersen, Stenbjerre 2008)
Addressing the dynamic issue of whether growth in loans leads to overburdened case workers in the banks: the research design deployed in this thesis does not yield room for qualitative investigation of the number of credit assessors in each bank in the sample, which is why it is argued that while using the annual percentage growth might be a reliable and valid method, it still does not guarantee on a higher-than-circumstantial basis that annual percentage growth in loans on the short term (6 years in this thesis) is a valid measure of risk-taking. It may very well be, but it need not be. The issue with using this risk measure is also that it still does not account for the absolute risk-level in the bank, because it too does not consider deposits. A suitable example of this is the investigation report on Fionia Bank, which states that: “The Bank’s overall growth in loans has not been particularly high when comparing with banks of the same size”(translated)(Skipper-Pedersen, Stenbjerre 2009). Yet Fionia Bank’s loan/deposit ratio was 138.2 in 2007, well above average.
To exemplify further, consider two banks: one banks lends out ten percent more in year 2 than in year 1, the other lends out fifty percent more. It is impossible to determine which bank has the highest risk profile, because the bank that lend out the most may also be the one which attracted many new depositors and thus had its loans covered by the new deposits, decreasing the risk that its funding would erode.
This does not account for the risk-level from which the increases start – even without new depositors, the bank might already have a huge surplus in deposits. Therefore, the annual relative growth in loans is not seen as the best risk measure, either, as it is neither an accurate description of the actual risk of going into financial distress nor a valid variable for this thesis.
To be certain that an appropriate measure is not rejected, the independent variables are tested on this risk measure in the data analysis section as a control variable similar but less significant results were obtained.
7.5 Growth in the loan/deposit ratio
Accounting for the deposits, yet another risk measure is the percentage-wise growth of loans to deposits. This risk-measure takes into consideration the possible simultaneous increase in deposits and thus does not characterize banks which experience increases in loans as well as
deposits as ‘risky’. The measure is reliable as it is a relative measure – large banks and smaller banks can be indexed together – which counts in its favor.
However, it is not particularly valid, as it does not account for the absolute level of loans to deposits. As a theoretical example, a bank that increases its loan to deposit ratio from 60 to 8033 will be counted as having experienced a 33 % increase in this particular risk measure.
Another bank, which raises its loan to deposit ratio from 160 to 200 will be counted as having experienced a 25 % increase in ‘risk’. This is not an accurate depiction of the risk experienced by the bank, because the second bank – deemed less risky by the measure – is much less liquid than the former example and its funding is much more exposed, all other things equal, to market risk than its competitor in the example.
To further exemplify: the average increase in loans/deposits 2003-200734 is 34.28 %.
However, employing this risk measure seems to place relatively safe banks well above the mean, as witnessed by Lån & Spar Bank, whose growth in the loan/deposit ratio is more than twice the average at 71.86 %. However, their absolute loan-to-deposit ratio remains well under 100 throughout the entire period, which basically classifies the banks as one of the least likely to go into financial distress (Børsen 2010b). It should be noted that some of the percentages correspond well to the reality, as witnessed by Aarhus Lokalbank (121.09 %, 139.56), Amagerbanken (66.40 %, 127.16) and bankTrelleborg (62.60 %, 125.82).
Therefore, this risk measure presents some advantages, but unfortunately some significant drawbacks as well. While accurately depicting the development of the growth in loans relative to deposits, the lack of consideration for the absolute level of the loan to deposit ratio severely limits the applicability of this risk measure.
7.6 The absolute level of loans to deposits
Finally, the absolute level of loans to deposits can be taken as a risk measure. As described above, other researchers in the field have done so recently (Bechmann, Raaballe 2009). The benefit of doing so is - perhaps not surprisingly - that the absolute level of outstanding loans is taken into account and therefore, this measure reflects the bank’s potential exposure to changing market conditions and, subsequently, its liquidity and solvency. This improves the measure’s accuracy. However, the dynamic feature of the relative growth ratio is somewhat lost as the number is inherently an absolute, not a relative, one.
33 Understood as “60 % of the deposits are lend out to 80 % of the deposits are lend out”
34 See section 7.7 Choice of risk measure for a justification of using 2007 and not 2008, the end of the surveyed period
The measure is reliable as it is equal for larger and small banks and all retail banks have a loan/deposit ratio. Moreover, the examples from the researched period seem to correspond well to the suggestion that the absolute level of loans is a relatively accurate description of bank risk.
In order to make the measure increasingly valid, the mean of the loans in the research years can be replaced by the loan/deposit ratio of the end of the boom period. While intuitively less accurate than the mean of all years, the application is in fact well suited to summarize the development in risk-taking in the researched period: when using the peak year of the boom, the risk measure reflects the (absolute) level of loans resulting from the governing bodies’
actions throughout the boom period, which brings with it its own considerations: when did the boom end? Bechmann and Raaballe (2009) use 2008 in their study on bank management pay.
It is argued in this thesis, however, that 2008 is not the most accurate year to capture the result of the boom years:
House prices started stagnating/falling in late 2006 (see section 1.3), stock prices (OMXC20) started falling from the fall 2007 and the default of Bear Stearns and Lehmann Brothers, which sparked an international bank liquidity crisis, happened in 2008. Thus, the loan/deposit ratio at the end of 2007 could be argued to best represent the result of the actions in the boom period – increasing the accuracy of this variable – because the loan/deposit rates at the end of 2008 already account for drastic changes following the liquidity crises as well as new regulations and guarantees for the banking industry by the Danish government, enacted in November 2008. To further exemplify the potential inaccuracy of deploying the loan-deposit ratio at the end of 2008, the case of Roskilde Bank is visited: because the bank’s deposits were already been sold off, the remaining loan-deposit ratio was a result artificially inflated at around 1300; a result of political interfering and not of risk-taking behavior (more than already experienced) on the part of management or the board of directors.
All banks except one that have faced either bankruptcy, been taken over, have entered financial distress or have been publicly speculated to be on the verge of going into distress show loan to deposit ratios at the end of the boom above the average, which in 2007 was a loan/deposit ratio of 123.35:
Table 7.1: Loan/Deposit ratio for selected banks in 2007
Source: {{374 Finanstilsynet 2010}}
* Estimate based on 2006. EIK Banks Loan/Deposit ratio decreased significantly in 2007due to inclusion of a junior loan as a deposit
7.7 Choice of risk measure
The volatility is rejected as it does not apply to all banks. The growth in absolute loans is rejected as it does not account for deposits and thus, does not represent risk as an appropriate measure. The annual percentage increase in loans is considered as it accounts for the organization dynamics that contributed to the failure of banks in the researched period, but it is found that these banks would be measured as relatively risky under the (other) chosen risk measure in any case and it is therefore rejected, because it too is slightly less valid than the final measure. (2009) contemplate using this risk measure, too, in their paper on bank management’s pay. They decide to utilize the absolute loan/deposit ratio, while stating their results would be just about similar with either measure.
The relative growth in loans/deposits is also acknowledged to be accurate in describing some banks’ development in the boom years, but not as a reliable risk measure that can be used generally across the sample. The absolute loan/deposit ratio has its drawbacks in describing the dynamics of the loan-giving (which the annual increase in loans does account for), but this is somewhat circumvented by utilizing the 2007 measure, which is suggested to be the most accurate summary of the actions taken by the banks from the beginning of the research period.
BANK L/D 2007
Bonusbanken 111,1
Lokal Banken Nordsjælland 119,2
Average 123,17
Løkken Sparekasse 126,43
Skælskør Bank 129,5
Morsø Sparekasse 129,9
Sparekassen Faaborg 136,5
Danske Bank 138,2
Fionia Bank 151,3
Forstædernes Bank 151,7
Max Bank 155,7
Forstædernes Bank A/S 163,1
Amagerbanken 165,4
bankTrelleborg 166,5
ebh Bank 183,2
Roskilde Bank 190,2
Aarhus Lokalbank 199,2
EIK Bank *653,5
7.8 Partial conclusion
This section first defined that choosing a risk measure should be one that mitigates the trade-off between (internal consistency) reliability and validity the best way possible. Then, five different risk parameters were evaluated: stock volatility, the absolute growth in loans, the annual relative increase in loans, the growth in the loan/deposit ratio and finally, the absolute level of loans to deposits.
The latter of the five was argued to best serve this particular study’s purpose, as it incorporates the deposits to the bank as well. The year 2007 was chosen to reflect some of the dynamics behind the economic boom up until that year.