Stress Testing in Cooperation between Danmarks Nationalbank and Large Danish Banks

Danmarks Nationalbank has performed a bottom-up stress test in cooperation with the six largest Danish banks in order to improve analysis of any vulnerabilities in the Danish banking sector. For the purpose of this exercise, Danmarks Nationalbank set up a baseline scenario and two stress scenarios as well as a number of assumptions to be included in the calculations by all participants. Based on their portfolios and internal models, the participating banks conducted impact analyses for each of the scenarios, and finally Danmarks Nationalbank aggregated the individual results

This chapter describes the stress testing process, including Danmarks Nationalbank's instructions to the banks and the further process of the joint stress test exercise. Finally, the most important general impressions from the ongoing dialogue between the banks and Danmarks Nationalbank in the course of the exercise are described. The aggregate results of the banks' calculations can be found in the chapter entitled Bottom-Up Stress Test.

Introduction

In the autumn of 2009, Danmarks Nationalbank performed a stress test in cooperation with the six largest Danish banks – the joint stress test exercise. Danmarks Nationalbank set up three scenarios for economic development in Denmark – a baseline scenario and two stress scenarios – as well as a number of assumptions to be included in the calculations by all participants. Each bank calculated the impact the economic scenarios would have on its portfolio, and finally Danmarks Nationalbank aggregated the results.

The stress test provides a forward-looking assessment of the ability of the largest Danish banks to absorb shocks to the economy. It is important to bear in mind that stress tests are based on a number of stylised assumptions concerning both the economic development and the development in the banks' financial statements. At the same time, specifying a number of the assumptions to be made by the banks limits the management response options available to the banks in the real world. This means that the banks had only limited opportunity to adapt their strategies to the development. It is also important to keep in mind that the stress test does not provide an exhaustive view of the risk factors faced by the banks, and that the full picture of risk factors varies from bank to bank.

Several advantages can be harvested from performing a stress test in cooperation with the banks to be tested – i.e. a bottom-up stress test – rather than performing the entire stress test centrally – i.e. a top-down stress test. The most important advantage is that the banks make the calculations based on their own models, which they also use for risk management purposes. Hence, the results will better reflect any differences in the quality of each bank's credit books and risk profile. In principle, the calculations also allow for the impact of large exposures and concentration risk in general on write-downs.

Conversely, a number of factors increase the uncertainty concerning bottom-up stress tests compared to top-down stress tests, making it potentially more difficult to interpret the results. The key challenge of bottom-up stress tests is the use of different models to evaluate the banks' portfolios, as each bank uses its own model. This can make data aggregation difficult, as there will be quantitative differences between the results because different models respond differently to different types of stress. The results of a bottom-up stress test should therefore be seen as a supplement rather than as a substitute for a top-down stress test.

An overview of the banks participating in the joint stress test exercise is given below. The procedure and the instructions received by the banks are subsequently explained. The most important impressions from the dialogue between Danmarks Nationalbank and the participating banks are then described, followed by a summary of the results of the exercise.

The participating banks

In the autumn, Danmarks Nationalbank asked the largest Danish banking institutions to participate in a joint stress test exercise. Participation in the exercise was voluntary, but all the invited banks decided to participate. The participating banks are the banks under Danish supervision with a working capital of more than kr. 50 billion, i.e. Danske Bank, FIH Erhvervsbank, Jyske Bank, Nordea Bank Danmark, Nykredit Bank and Sydbank.

Lending by the six participating banks accounts for 80 per cent of the total lending by Danish banking institutions, and their balance sheets likewise account for 85 per cent of the total balance sheet, cf. Chart 16. At the same time, the population covers 43 per cent of lending by mortgage-credit institutes, as Realkredit Danmark and Nordea Kredit, which are subsidiaries of Danske Bank and Nordea Bank Danmark, respectively, are also included in the calculations. On the other hand, Nykredit Realkredit as the parent company in the Nykredit group is not included, as Nykredit Realkredit was not asked to participate in the exercise.

Two of the participating banking institutions belong to large financial groups. These are Nykredit Bank, a subsidiary of Nykredit Realkredit, and Nordea Bank Danmark, a subsidiary of Nordea Bank AB. For Nykredit Bank and Nordea Bank Danmark alike this means that they have traditionally been operating with lower excess capital adequacy than the other banks in the population, as the excess capital adequacy of the groups is placed in the parent company from which it is then allocated to the subsidiaries. As the parent companies are not included in the stress test exercise, it is difficult to assess how the scenarios will impact the groups and the extent to which the parent companies are able to support the subsidiaries.¹

The primary risk factor for the six participating banks is currently assessed to be the development in write-downs. This is reflected in the choice of stress scenarios.

Danske Bank is by far the largest bank in the population, cf. Table 5. At the same time, Danske Bank is the only bank that in its response to the bottom-up stress test explicitly takes into account credit exposure to foreign economies, cf. Chart 17. Just over half the bank's credit exposure is to Danish customers, but it also has significant credit exposures to customers in Finland, Sweden, Ireland, the UK and Norway. Danske Bank's units in Finland, Northern Ireland and Luxembourg are subsidiaries, while the rest are branches.

At the baseline, the six banks have a considerable buffer to absorb write-downs. Four out of six have a Tier 1 ratio of more than 11 per cent, the two exceptions being Nordea Bank Danmark and Nykredit Bank. As previously mentioned, this can reflect their group relations. This results in a considerable buffer in relation to the statutory requirement.

The banks will primarily experience a decline in their excess capital adequacy if they record losses. Current earnings serve as the first bulwark against write-downs. There is a remarkably large difference between the extent to which the banks can write down their lending before the write-downs exceed their current earnings, i.e. profit before write-downs and tax, and begin to eat into their capital. This applies both across banks and over time, cf. Table 6. In 2008, profit before write-downs and tax was lower than before. Conversely, profit before write-downs and tax was relatively high in the 1st half of 2009. It is clear that differences in the current earnings of the banks greatly affect their ability to absorb write-downs in the near future.

The process

The banks' participation in the common stress test exercise was voluntary and based on the premise that rather than publishing the results for individual banks, Danmarks Nationalbank would only publish results in an anonymised form.

The common stress test exercise was subject to a tight schedule. In mid-September 2009, Danmarks Nationalbank issued a set of draft instructions to the six banks. The instructions were to form a common basis for the calculations. The banks were then invited to comment on them before the final instructions were issued at the end of September. The primary objective of the banks' comments was to clarify uncertainties, to identify the needs of individual banks and to point out any bank-specific issues that should be taken into account in order to increase the comparability of the results. As far as possible, and to the extent this would not compromise the principle of the exercise, the comments were incorporated into the final instructions. The stress level of the scenarios was determined by Danmarks Nationalbank.

The results of the calculations were supplied to Danmarks Nationalbank at the end of October. Danmarks Nationalbank, jointly with the Danish Financial Supervisory Authority, subsequently engaged in a dialogue with each of the participating banks to achieve the best possible understanding of what drives the results and to ensure maximum comparability. The results were subsequently aggregated by Danmarks Nationalbank. The participating banks had the opportunity to review the results before publication.

While it was Danmarks Nationalbank's responsibility to set up the economic stress scenarios and aggregate the results, the banks were responsible for the calculations. The fact that the calculations were made decentrally and based on different models makes data aggregation a challenge. It is not given that different models react in the same way to different types of shocks, which is necessary if the results are to be fully comparable. In order to ensure the greatest possible comparability across banks, Danmarks Nationalbank set up a number of assumptions – in addition to the macroeconomic scenarios – to be made by each bank. The most significant assumptions concern lending growth, the composition of exposures and the development in the capital of the banks. Accordingly, the banks' results will reflect both the exercise instructions and circumstances specific to each bank. The advantage of these instructions is that they ensure better comparability of results. The drawback is that they limit the realism of the calculations.

Instructions

The calculations were based on the banks' interim reports for 2009, although some banks also included the actual development in the 3rd quarter. The economic scenarios run until the end of 2011. This means that the banks estimate the development in their profit and balance sheet over a period of 2½ years under the given assumptions. The banks' calculations are made on a consolidated basis, i.e. including subsidiaries. Moreover, the calculations are based on current legislation and do not include expected legislative amendments.

Scenarios

The stress test included three scenarios: a baseline scenario and two stress scenarios. The baseline scenario is the scenario considered by Danmarks Nationalbank at the start of the stress test exercise to be representative of the most likely development in the Danish economy.²The two stress scenarios include a scenario where the Danish economy is exposed to shocks while the development in the international economy corresponds to the development in the baseline scenario, and a scenario exposing both the Danish and the international economy to shocks.

The macroeconomic variables defining the scenarios of the development in the Danish economy are for GDP, unemployment, labour force, money-market interest rates and bond yields, house prices, HICP, imports and exports, business and residential investments, private and public consumption and hourly wages on an annual basis, cf. Table 7. Danmarks Nationalbank also made historical series available to the banks requesting this.

International economic developments in the baseline scenario are assumed to follow the OECD's forecast from June 2009 to end-2010.³Overall guidelines were provided for the way the international economy was to be assumed to develop in order to be consistent with the other parts of the scenarios. However, banks with direct exposure to the development in foreign economies have had a high degree of autonomy in their assessment of the stressed development in those economies.

Assumptions
In addition to the specification of the economic development, a number of assumptions to be included in the calculations by all the banks were specified. The required assumptions may deviate from the assumptions the banks themselves would have made. Hence, the results may provide a different picture of their resilience than the banks themselves would have achieved. To ensure as much comparability as possible it was necessary to specify a number of stylised assumptions, however.

Basically, lending is assumed to be constant over the period concerned so that by the end of 2011, the level of lending of each bank will be the same as in the bank's interim report 2009. Especially in the two stress scenarios, the demand for loans must be expected to reflect the negative economic development. As a result, the banks in the scenarios may look less resilient than they actually are. The banks were therefore allowed to calculate the scenarios under alternative assumptions concerning lending growth. Under the alternative lending assumptions, lending declines by 6 per cent annually in scenario 1 and 8 per cent annually in scenario 2. No alternative lending assumptions were specified for the baseline scenario. All three scenarios assume that the composition of the bank's lending portfolio will remain unchanged during the scenario period.

It was also specified that the value of the collateral pledged as security for the banks' credit exposures must be assumed to follow the development in house prices in the scenarios. The exception was financial assets that must be assumed to follow market developments. The banks' securities portfolios must also be assumed to follow market developments.

It was assumed that raising additional core capital, hybrid core capital and subordinated loan capital is not possible. It was also to be assumed that hybrid core capital and subordinated loan capital would be repaid at the time of an interest step-up, if any.⁴One exception was made to this assumption in that it was assumed to be possible to replace internally issued subordinated loan capital by a similar new loan.

In relation to the capital structures shown in Table 6, the banks were able to recognise in their calculations any changes in their capital bases that were known for certain before the provision of final results. Accordingly, Sydbank was able to recognise kr. 1,286 million as a result of an increase of the share capital and the sale of own shares. Nykredit Bank was able to recognise kr. 800 million as a result of a new capital injection from Nykredit Realkredit and a conversion of further supplementary capital in the amount of kr. 2,400 million into share capital. Finally, Jyske Bank was able to recognise kr. 1,368 million as a result of a capital increase and the sale of own shares.

The banks' appetite for market risk was to be assumed not to change over the period. Operational risk was not included as part of the scenario specification.

In the baseline scenario, it is assumed that the costs of the Financial Stability Company will reach a level so that the banks participating in Bank Rescue Package must pay kr. 10 billion in addition to the guarantee commission of kr. 15 billion. The sector's total costs for the Financial Stability Company will thus amount to kr. 25 billion. The two stress scenarios assume that the banks must pay kr. 20 billion in addition to the guarantee commission. The sector's total costs in connection with Bank Rescue Package thus amount to kr. 35 billion under stress. For the banks that have received capital injections from the Credit Package, the injected capital is assumed to be able to replace funding at 4 per cent p.a. It is assumed that hybrid core capital cannot be converted into share capital unless such a conversion is required. No further public intervention is assumed in any of the scenarios.

The banks were instructed to allow for economic developments in the scenarios when calculating their adequate base capital. This means that the development in the individually calculated capital need should be taken into account in the scenarios.

Output from the banks
The instructions to the banks contained a specification of the output to be provided to Danmarks Nationalbank. The groups with mortgage-credit institutes as subsidiaries were requested to provide the results both including and excluding the mortgage-credit activities. This makes it possible to compare the bank activities of the groups.

The banks were asked to present their expectations of the development in income and costs, write-downs, capital structure, risk-weighted assets and adequate base capital for each scenario. In addition, they must include a breakdown by industry of the write-downs. The banks with significant exposures to other countries than Denmark were asked to specify the geographical distribution of their total write-downs. This is of particular interest for Danske Bank, as just under half of its lending is provided to customers outside Denmark.

The instructions specified that the net operations impact of the bank's write-downs and losses must be reported. The net operations impact is of interest in terms of whether they will have problems meeting the statutory capital requirement. The fact that the final losses may turn out to be considerably lower than what is written down is of no significance if the bank is closed down before that time.

Finally, each bank had to supply a brief but adequate description of the methods used in the scenario calculations. They formed the basis for Danmarks Nationalbank's assessment of the effect the method may have had on the results and thus also formed the basis for the dialogue with the banks.

The banks were able – and indeed encouraged – to state whether the specified assumptions would result in a misleading view of the resilience of the bank concerned. The most questionable assumption was zero lending growth. As mentioned, it was also possible to calculate alternative assumptions concerning lending growth.

Responses, methods and dialogue

One objective of the common stress test exercise was to obtain a more nuanced picture of the resilience of the Danish financial sector to further negative shocks to the economy. This was achieved by presenting the results of the exercise in the chapter Bottom-Up Stress Test. Another objective was to increase the knowledge of the methods used by the banks and to better understand the set-up of economic stress scenarios. The key general issues discussed are described in this section.

Several of the banks' models are based on data from the same publicly available sources. This can be necessary in order to obtain data series that are sufficiently long to be able to model the development of e.g. the probabilities of failures in certain sectors as a result of macroeconomic developments. On the other hand, this may also mean that some of the differences in the banks' results reflect differences in the banks' methods rather than differences in their portfolios. The results may therefore represent actual portfolio differences to a lesser extent than what could be expected.

Two of the six banks chose to supply results including the assumption of unchanged lending and including the alternative lending assumptions where loans and guarantees are reduced in the two stress scenarios. The dialogue with the banks has shown that the banks have very different expectations of how their lending will develop during the period. Some banks expect a natural decline in their lending, especially if the economy is subjected to further stress. Others expect their lending to increase. The differences in the banks' expectations reflect inter alia differences in their views of how economic developments will impact the demand for loans and their strategies in the coming years.

Many of the banks stated that some of the assumptions specified by Danmarks Nationalbank for the stress test calculations gave a less true and fair view of the results than would have been the case if the banks had been given a free choice of methods. The argument is that the view of each bank is distorted when it is forced to use calculation assumptions for the sector that do not match the bank concerned. It is a balancing act to obtain comparable results on the one hand and to work with customised assumptions on the other. It can be argued that it is desirable for banks to present stress test results in their external communication based on both standardised assumptions and the assumptions which, according to the bank, describe the bank's development.

In terms of methods, the technical level of the banks' approach to stress tests of their exposures varies greatly. Obviously, the technical complexity is closely related to their size. The exercise demonstrated that at present banks that are smaller than the participating banks are likely to be unable to perform stress tests based on economic scenarios.

The scenarios set up were read by the banks – and written by Danmarks Nationalbank – as general shocks to the economic development. The dialogue with the banks raised the issue of whether such stress may trigger other more specific risk factors. A case in point might be that further shocks to the world economy may lead to devaluations in some economies that are already vulnerable. Another example might be that the combination of rising interest rates and lack of competitiveness in the scenario with an isolated negative shock to the Danish economy may have a stronger impact on agriculture than is otherwise to be expected. Both examples may lead to larger write-downs than normally warranted by the models.

Finally, the results and the dialogue with the banks demonstrated differences in what is taken into account in the banks' statements of their adequate base capital. These differences must be assumed to be even more pronounced when compared to smaller banking institutions. This may reduce the comparability of the capital needs to be published by banks in connection with the annual reports for 2009 onwards. The capital needs of individual banks contain important information for the market, however.

Summary

Viewed from the perspective of Danmarks Nationalbank, the dialogue concerning the stress test assumptions, scenarios, results and the methods used has been open and constructive. The banks have made a significant effort. In future, Danmarks Nationalbank will extend its dialogue with the Danish Financial Supervisory Authority and the sector in relation to stress testing.

The bottom-up stress test contributes significantly to the assessment of the resilience of the financial sector. But it is important to bear in mind that the results of the exercise are subject to considerable uncertainty. Neither Danmarks Nationalbank's stress test model nor any other type of model can by any means model the complexity of the world. In particular, stress test scenarios will never be realised, as the relevant decision-makers will react if the stress test shows outcomes they wish to avoid. The overall assessment of the resilience of the financial sector is presumably improved by weighting together the development in different indicators and the results of different models.