EU-wide pension stress testing, the influence of valuation

(1)

influence of valuation

Bram Baneke

Master’s Thesis to obtain the degree in Actuarial Science and Mathematical Finance University of Amsterdam

Faculty of Economics and Business Amsterdam School of Economics

Author: Bram Baneke

Student nr: 6131298

Email: baneke9@msn.com

Date: August 31, 2014

Supervisor UvA: Dr. T. Boonen Second reader: Dr. L.J. van Gastel

(2)

(3)

Abstract

The European Insurance and Occupational Pensions Authority (EIOPA) plans to con-duct an EU-wide pension stress test. Specific characteristics of the pension sector in the EU are the variety in risk-sharing mechanisms and the valuation of pension fund across the EU. Two different valuation methods are applied in the EU-wide pension stress test from EIOPA, national valuation and a common theoretical valuation method. A quantitative analysis on aspects of the national valuation is executed, in order to de-termine the influence of national valuation methods on the impact of a stress scenario. For two relevant scenarios, the aspects of the national valuation methods have a large influence on the stress test results and lead to positive stress test results. Another rele-vant scenario results in limit influence of the valuation. Consequently, the scenario for the EU-wide pension stress test could be designed to limit the influence of valuation, as long as the scenario is still relevant for pensions in the EU.

(4)

Preface

I like to thank my supervisors at De Nederlandsche Bank, Bart Wijers and Rick Hoogen-doorn, and all other co-workers for all their knowledge, advice and support. De Ned-erlandsche Bank was a truly inspiring environment for writing my thesis and I like to thank Gisella van Vollenhoven-Eikelenboom for giving me the opportunity. I also like to thank my UvA supervisor, Tim Boonen, for providing sharp comments on my thesis and good supervision in general.

Views expressed in this thesis are my own views and do not necessarily reflect the views of De Nederlandsche Bank.

Amsterdam, 31 August, 2014

(7)

Introduction

This thesis is concerned with designing an wide stress test for pension funds. EU-wide stress tests have already been executed for the banking and insurance sector. In the aftermath of these stress tests, the European Insurance and Occupational Pen-sion Authority (EIOPA) plans to execute an EU-wide penPen-sion stress test. An EU-wide pension stress tests has not been executed before. Therefore, features of stress testing for the EU-wide pension stress test and the problems at hand when designing it are investigated.

The purpose of this thesis is not to actually create a complete EU-wide pension stress test. But rather to investigate what the EU-wide pension stress test could look like. The focus is on the valuation method of the EU-wide pension stress test, but other aspects of stress testing are also investigated.

The EU-wide stress tests for the bank sector and insurance sector already exists. These EU-wide stress tests are considered when designing an EU-wide pension stress test. Hence, these stress tests are studied and possible features that could be adopted in the EU-wide pension stress test are examined. However, there are also parts in the banking and insurance stress test that are not applicable for the pension stress test. The banking and insurance sector are compared to the the pension sector in order to determine which parts should be adopted.

Besides looking at the EU-wide banking and insurance stress tests, national pension stress tests should also be investigated.Impavido(2011) andIonescu and Yermo(2014) already did research on stress testing pension funds. Pension stress tests have been executed on national level (Ionescu and Yermo, 2014), but not on an EU-wide level. Pension stress testing on EU-wide level has consequences for the valuation of the stress test.Impavido (2011) points out that valuation is an essential element of stress testing and varies significant across countries. The EU-wide pension stress test is valued by both national and a common theoretical valuation method (EIOPA,2014c). This is a major difference compared to the EU-wide banking and insurance stress tests. The national valuation methods differ from each other (CEIOPS, 2008). Therefore, the effects and issues of a pension stress test with different national valuation methods are investigated. A complete analysis of the different valuation methods across the EU pension sector is too broad for this thesis. Hence, an analysis on important aspects of the national valuation method is provided. Countries are not compared with each other, but a more general approach is applied by investigating the aspects. It is of course possible to lead the aspects back to the a specific country. However, drawing conclusions on the differences between countries is hard, since not the entire valuation method is examined. The thesis consists of nine chapters. Chapter2 reviews stress testing in general. In this chapter the multiple features of stress testing will be discussed. Chapter 3 discusses pensions across the EU. In Chapter4and Chapter5a comparison between the pension

(8)

2 Bram Baneke — EU-wide pension stress testing

sector and the banking and insurance sector is given. Chapter 4 is about the EU-wide 2014 banking stress test conducted by the European Banking Authority (EBA). Chapter

5 describes the EU-wide 2014 insurance stress test from EIOPA. Chapter 6 provides a description of the model and assumptions for investigating the issues and effects of different valuation methods. The results of this quantitative research is given in Chapter

7. Sensitivities in the quantitative research are discussed in Chapter8. Finally, Chapter

(9)

Stress testing in general

In this chapter stress testing in general is discussed. Although this chapter is named stress testing in general, the stress tests in this chapter and in the thesis are all stress tests from the regulator. First, I discuss the concept of a stress test and why stress tests are conducted. Second, two papers for creating a stress test are discussed in Section2.3. Last, Section 2.4displays the options for a stress test scenario.

2.1 The concept of stress testing

IMF (2012) states that: “stress testing is a technique that measures the vulnerability of a portfolio, an institution, or an entire financial system under different hypothetical events or scenarios”. When making an EU-wide pension stress test, the vulnerability of a portfolio alone is not very interesting. Therefore I focus on measuring the vulnerability of a financial institution or an entire financial system. Following the IMF definition, two important features come to mind. First, what are these hypothetical events or scenarios? Second, how to measure the vulnerability of a financial institution or an entire financial system?

Stress

Stress is a change in the economic environment or in the risk factors of a financial in-stitution. The stress in a stress test is called the scenario. Since stress testing has the word stress in it, the scenario should have an adverse effect on the financial institu-tion. What do these changes in the economic environment or in the risk factors of the financial institution look like? The economic environment of the financial institution could be changed to a different state of the world. In this scenario all the risk factors move in a coherent matter. Another approach is to change one or more relevant risk factors in a non coherent way. This is a more simplified scenario. Only changing one risk factor is called sensitivity testing. Since in that case, the sensitivity to a specific risk factor is investigated. When more than one risk factors are changing at the same time, it is sometimes called scenario testing or scenario analysis. But in order to keep the definitions clear, I will use the term scenario for the stress that is applied in a stress test.

Measuring stress

The second question is how to measure the condition of a financial institution. The answer to this question depends on the type of financial institution. The measure should be relevant for the financial institution in the stress test. A basic concept to measure

(10)

the condition of a financial institution is through the balance sheet of the institution. On the balance sheet of a financial institution are the assets and liabilities. There exist many approaches to value the assets and liabilities. Therefore, assumptions about the valuation are essential in order to determine the assets and liabilities.

Phases of stress testing

In a stress test, there are three basic phases. First, the scenario has to be created. In a stress test from the regulator the scenario is created by the regulator. Stress tests created by the regulator are called regulator stress tests. Stress test can also be created by the financial institution. The EU-wide pension stress test is a regulator stress test and the scenario is created by EIOPA.

Second, the scenario of the stress test should be applied to the financial institution. In a top-down stress test the regulator applies the scenario. For example, the EU-wide pension stress test could be a top-down stress test, so the scenario is applied by EIOPA. In this case EIOPA models will be used to calculate the impact of the scenario. The advantage of the top-down stress test is the comparability of the results, since the regulator will apply the same models to all financial institutions. In a bottom-up stress test, the scenario is applied by the financial institutions. For example, in a bottom-up EU-wide pension stress test the scenario is applied by the pension funds. The advantage of bottom-up stress tests is the use of custom-made models of the financial institutions. These models could be a better fit for the financial institutions.

Third, the results of the stress test have to be interpreted and measures might be taken. The results of the stress test should be interpreted by both the management of the financial institution and the regulator. The management of the financial institutions should interpret the results of the stress test and ask themselves the question whether the results are acceptable. When the results are not acceptable, they should adjust. The regulator also interpret the results and take measures when necessary. The interpretation of the results of the stress test by the regulator is known as the follow-up of the stress test. Sometimes the follow-up of a stress test can be very strict. For example there could be a hurdle rate for the stress test. The follow-up can also be less strict. This is the case when the regulator evaluates the results of the stress test and then decides whether he finds the result acceptable.

2.2 Why use stress testing

Following the definition of Section 2.1, stress testing can be used as a technique that measures the vulnerability of an entire financial system. Such a stress test would give an overview of the impact of the scenario on the financial system. However, testing the entire financial system is almost impossible. Since the scenario would be applied to all the financial institutions in all the districts of the financial system. The scenario might not be relevant for all the types of financial institutions or not relevant in some districts. Therefore, stress tests search for transmission channels to system relevant vulnerabilities. These vulnerabilities can be in the financial system or in the real economy. Stress tests searching for system relevant vulnerabilities are called macro stress tests. Jones et al.

(2004) mention that macro stress tests can be helpful for policymakers in finding relevant system vulnerabilities. In the financial crisis of 2007/08 a lot of systematic risks came to the surface. Hence, the attention of supervisors has shifted towards the search of systematic risks.

Following the definition of Section2.1, a stress test could also measure the vulnerability of an individual financial institution. When a stress test is focused on the ability of a

(11)

financial institution to absorb a stress scenario, the stress test is called a micro stress test. Micro stress tests are focused on the resilience of individual institutions. But the stress tests results can also be aggregated for the total sector. For a financial institution it is important to know the vulnerabilities it is exposed to.IAA(2013) argues that stress testing should be used “to enhance the understanding of if and why a firm is vulnerable to highly uncertain tail risks”. This perspective is likely to be more appealing to the financial institutions.

The objective of the EU-wide pension stress test is to test the resilience of the pension funds to adverse market developments, to test for the systemic relevance of pension funds and to test for interconnectedness of economies and markets. (EIOPA,2014c)

Stress tests as supplements and cross checks

Stress testing can be used to complement and cross check the standard regulatory frameworks. In some regulatory frameworks capital requirements are determined by stressing individual risk factors piece by piece. Therefore, such a framework could also be considered a stress test. The total capital requirement is the sum of all the individual capital requirements added by a correlation matrix. This stress testing can be very useful as a standard capital requirement. However, the impact of two risk factors moving downwards together is only captured by the correlation matrix. In a stress situation the correlation between to risk factors could increase dramatically. A stress test could be executed in order to test the impact of a scenario where multiple risk factors are moving downwards together. Another possibility is the fear for a larger decrease in a risk factor than in the stress of the capital requirements. The effect of this scenario could also be tested in a stress test.

Some regulatory frameworks use stochastic modeling in order to determine the capital requirements. Shaw et al. (2004) mention stochastic modeling produces a distribution of outcomes. A possible capital requirement is a percentile of these outcomes. They also mention stress tests could supplement stochastic modeling since stress testing focuses on the tail events of the probability. The stochastic modeling focuses on the entire distribution, whereas a stress test is able to completely focus on events in the tail of the distribution. Also the correlations could be different in the tails of a distribution. Hence, stochastic modeling could give an unrealistic view of the risks in the tail.Shaw et al. (2004) also argue stress tests produce results more quickly and the results of a stress test are easy to communicate.

Market confidence

When there is a large amount of fear in the market, a stress test could be done to analyze if the fear is rational or not. When the stress in the market is caused by fear for some possible adverse event, the scenario of the stress test should be related to this event. The stress test could produce a rough estimation of the impact of this event. If the results of the stress test are published, the impact of the event will be clear to the market. Therefore, a stress test could enhance the confidence level in the market. The authorities could place an additional capital requirement based on the results of the stress test, when they think the institution should be able to absorb the scenario. This capital requirement is the follow-up of this stress test. This extra capital requirement could create an additional confidence increase of the market in the financial institutions. When no capital requirement is set by the authorities, the authorities could still use the results for risk management purposes. The vulnerabilities of the financial institution are in that case known to the authorities and the institution, which increases the risk awareness.

(12)

2.3 Creating a stress test

Before the scenario is created, the type of the stress test should be clear. The EU-wide pension stress test has both aspects of micro and micro in the objective (EIOPA,

2014c). Two approaches for creating a stress test are of interest. There is a micro stress test approach and a macro stress test approach. The main difference between a micro stress test and a macro stress test is the risks that are stressed. The macro stress test investigates the transmission channels to the financial system or real economy, whereas the micro stress test looks at risks that could impact the financial condition of the institution. As a guideline for the micro stress test,Shaw et al.(2004) is used andJones et al. (2004) for the macro stress test.

Micro perspective

Shaw et al. (2004) mention some steps for designing a stress test. Although this article is focused on designing stress tests by and for insurers, the steps can also be applied on designing micro stress tests in general. Designing a micro stress test can be done as follows. First a baseline scenario to stress against is determined. This baseline could be the current or expected financial condition of the institution. Second, the risks faced by the financial institution should be investigated. For example, longevity could be a risk for pension funds. Third, the key risks should be selected, once all risks are known. Key risks are the largest risks faced by the financial institution. Fourth, these key risks are investigated. The causes and effects of the key risks should be determined. For example, how sensitive is a pension fund to the longevity risk. Fifth, a measure should be chosen. This measure should be able to capture the impact of the risks. Last, a plausible scenario should be identified and the parameter ranges should be quantified. The scenario should be related to the key risks and take the valuation method into account.

Macro perspective

For the macro perspective,Jones et al.(2004) is more applicable. The basic steps here are identifying vulnerabilities, constructing scenarios, balance sheet implementation, second round effects and interpretation. In the first step, the macroeconomic parameters are explored and the parameters that could indicate vulnerabilities are researched. In the second step, data and model to investigate the vulnerabilities from the first step needs to be provided. A macro econometric model could create a coherent scenario. In this scenario, multiple risk factors move together in a coherent way. For example, market values of corporate bonds could be affected by change in the risk free interest rate curve. In the third step, the scenario is applied. This will result in an effect on the balance sheet of the financial institution. In the fourth step the second round effects are investigated. For example, second round effects could be actions from a supervisor to a scenario. In the last step the results of the stress test are interpreted and the supervisor should think about the follow-up of the stress test.

2.4 Options for the scenario

In the view of designing an EU-wide pension stress test, different options for creating a scenario are considered. IAA (2013) discusses many features of stress testing and scenarios. In this section, the essential features are discussed. Essential features in a scenario are: the storyline or the source of stress, the development of the scenario over time, the number of risk factors that are stressed in the scenario and the dependency between the risk factors.

(13)

The source of stress is historical or hypothetical. The historical scenarios already took place. This has both advantages and disadvantages. The advantage is the availability of data for the scenario. The scenario happened in the past and so data from the scenario should be available. Therefore, it should be easier to determine the relevant risk factors, the development and the second round effects of the scenario. It could be difficult to place the historical scenario into the present time. The economic environment in the past might be very different than the one of the present time. When the time lag increases this disadvantage is likely to become larger. The number of historical scenarios is also limited.

When a historical scenario is chosen, one is limited to the scenarios that occurred in the past. However, parts of the historical scenarios can be incorporated in a hypothetical scenario. Hypothetical scenarios are scenarios that never occurred before. Especially in the dynamic financial environment this can be valuable. The development of new financial products increases the need for hypothetical scenarios. The disadvantage of the hypothetical scenario is the lack of data. Creating a hypothetical scenario is much harder and the number of assumptions will increase. The scenario developer should search for similar scenarios or parts of scenarios in the past. For example, there is a large conflict between two countries and there is a threat of trade restrictions. These trade restrictions might be very uncertain and might have never occurred before. However, one might be able to observe data of previous trade restrictions from other countries. The scenario developer could then investigate the effect of those trade restrictions and apply a comparable stress in the hypothetical scenario.

The source of stress should be clearly communicated to create a better understanding of the relevance. In order to do so, the scenario should have a storyline. The storyline can describe certain events that happen in a specific order and develop over a timeline. The storyline can also be very short and contain only one event. The storyline should contain an event relevant for the financial institution, so that the effects of the event on the risk factor of a financial institution can be determined. The effect on these risk factors will be the scenario. The advantage of this way of designing scenarios is the intuitive link from the results of the stress test to the event. When risk factors are stressed without a link to an event, it will be harder to find out when a stress materializes. This might be the reason for the management of the financial institution to challenge the plausibility of the scenario.

A scenario occurs either instantaneous or develops over time. The horizon is the time a scenario develops. More assumptions are needed as the horizon of the scenario increases. The longer the horizon the more uncertain the scenario development becomes. Therefore, one should be careful when the horizon of the stress test is set.

Multiple risks can be stressed in a scenario. The number of risk factors in a scenario should depend on the source of stress. A scenario could also be created without a clear source of stress. In that case, the number of stressed risk factors are arbitrary. All the risk factors could be stressed with a certain level of severity.

Financial institutions should understand the dependency structure of risk factors. This dependency structure is also needed for a stress scenario. The source of stress might affect a risk factor in particular. This risk factor could in turn affect other risk factors. It is essential to know how the risks factors behave after the initial event. Since stress scenarios don’t occur that often, it is hard to analyze the dependency structure of a stress scenario. A macroeconomic model can be used in order to let the risk factors move in a coherent way. However, some risk factors might not be in the macroeconomic model, like longevity. The intuitive way to let the risk factors interact, is through a specific storyline. By doing so, the subjectivity of the dependency structure is clear. More advanced scenario dependency structures and scenario developments can be found inIAA (2013).

(14)

Chapter 3

The EU pension sector

This chapter is concerned with the EU pension sector. In the first section, the pension systems across the EU are discussed. The second section consists of both micro and macro risks faced by a pension fund. In the third section, the valuation in the EU-wide pension stress test is examined. The final section discusses differences in aspects of national valuation methods across the EU.

3.1 Differences within the EU pension sector

The report ofEIOPA(2014b) contains a section that captures the diversity of national occupational pension systems. The report mentions that different pension systems are active across the EU. They also mention that various risk sharing mechanisms are ap-plied on these pension systems. The risks might be owned by the pension fund, the sponsor, the participants of the pension fund or a mix of these. This risk sharing af-fects the security mechanisms. Security mechanisms could be activated when a pension fund is in a weak financial condition. Depending who owns the risk, the security mecha-nisms could be solvency capital, sponsor support, pension protection schemes or benefit adjustments.

Another difference mentioned in EIOPA (2014b) is the regulatory regimes that the pension funds are submitted to. These regulatory regimes contain for example valuation of the pension funds, timespan to recover in case of underfunding and the ability to raise the premium in case of underfunding. In Section3.4, some aspects of the valuation across the EU are discussed in more detail.

Indexation of the pensions benefits varies across the EU (CEIOPS,2008). The objective of pensions is to keep a certain welfare level after retirement. In order to do so, the pension should be corrected for price inflation or wage inflation. An inflation linked indexation can be unconditional, conditional or capped. Including inflation stress in the scenario might be relevant for some countries.

3.2 Risks in pension funds

The EU-wide pension stress test should test the resilience of pension funds to adverse market developments and test the systemic relevance of pension funds (EIOPA,2014c). First the risks for a micro stress test will be investigated. Micro risks should have an adverse effect on the financial condition of a pension fund. Second, the risks for a macro stress test will be explored. These risks should be relevant for the financial system or real economy. Therefore, the transmission channels from the pension fund to the financial system and real economy are of interest.

(15)

Micro risks

Pension funds are exposed to several risks. I follow three steps fromShaw et al. (2004) as mentioned in Section 2.3 to search for key micro risks. The first step is identifying the risk faced by the pension fund. Many studies investigate this subject. In this thesis the risks faced by pension funds across the EU are of interest. The the quantitative impact study (QIS) by EIOPA is focused on pension funds across the EU and so this study is appropriate. Several risks faced by pension funds across the EU are examined in the QIS.

Figure 3.1: Modules and submodules of the Solvency Capital Requirements (SCR) from the QIS (EIOPA, 2012).

In Figure 3.1, the modules and submodules from the Solvency Capital Requirements (SCR) are shown. This modules are the risk factors that are examined in the QIS. In the EU-wide pension stress test these risk factor could be of interest. The SCR can be divided into the basic SCR, the adjustment loss-absorbency and the capital requirements for operational risk. The basic SCR is parted into market risk, health risk, counterparty default risk, pension liability risk and intangible asset risk. The market and pension liability risk factors in turn are parted into risk factors. The market risk is parted into interest rate risk, equity risk, property risk, spread risk, concentration risk, currency risk and counter-cyclical premium risk. The pension liability risk factor is parted into mortality risk, longevity risk, disability morbidity risk, benefit option risk, expenses risk, revision risk and catastrophe risk.

After the risk identification, the next step inShaw et al. (2004) is to select the key risk factors for pension funds. The results of the QIS show which risks are the most relevant for pension funds (EIOPA,2013). The impact of the key risk could be examined in a stress scenario.

In Figure 3.2, the results of the QIS are displayed. The largest risk for all countries that participated in the QIS is market risk. Hence, this risk is a key risk and should be included in the stress test. Another large risk in almost all countries is the pension liability risk. Therefore this risk should also be included in the stress test. The results also show that counterparty default risk is a significant risk in some countries. This

(16)

Figure 3.2: Decomposition of the SCR expressed as a percentage of the liabilities (EIOPA,

2013).

could be related to sponsor credit risk. To conclude, based on these results the stress test should at least have market risk and pension liability risk.

Figure 3.3: Decomposition of the market risk factor expressed as percentage of the lia-bilities (EIOPA, 2013).

Since market risk is a key risk, the split up of this risk factor could be interesting in order to get a better understanding of the pension fund vulnerabilities. Therefore the market risk factor is split up, see Figure 3.3. On average interest rate risk and equity risk are the most significant risks. Other on average less significant risks to consider are currency, spread and property risk. Concentration risk is not very relevant for the pensions in the QIS. To conclude, equity and interest rate risk are the best candidates

(17)

for key risk factors.

Figure 3.4: Decomposition of the pension liability risk factor expressed as percentage of the liabilities (EIOPA, 2013).

Pension liability is also a key risk for pension funds. In Figure3.4, the pension liability risk factor is split up in different risk factors. Within the pension liability risk longevity is the largest risk in all countries. Only in a few countries other pension liability risks play a role. Longevity risk is a classic pension risk. When the participants live longer the pension fund has to provide extra pension benefits for the time the participants life longer. These extra pension benefits are not taking into account in the calculation of the premiums. Therefore, longevity affects the financial position of the pension fund in a negative way. To conclude, longevity is a key risk for pension funds.

The following step inShaw et al. (2004) is to consider the causes and the effects of the selected risks. The focus point of a stress scenario could be the collapse of multiple key risks together. Therefore, it could be interesting to research how the key risks interact. A complete analysis of the relation between the selected key risks is to broad for this thesis. However, market risks are likely to depend on each other. In order to make a coherent scenario, a macroeconomic model should be used.

The dependence between the market risks and the longevity risk is less intuitive. As a starting point one could assume there is no dependence between the two, especially on the short term. However, a longevity shock could have a large impact on insurance companies. These insurance companies could in turn affect the market risks. Or a large market shock could lead to unemployment and less disposable income. Less disposable income could in turn lead to lower living conditions. And this in turn could lead to a lower life expectancy. When the number of births stay constant, longevity will lead to an aging population. Krueger and Ludwig (2007) claim the aging population will lead to lower market returns. To conclude, there is a clear dependency between market risk factors and these risk factors should be modeled in a coherent matter. But the dependency between longevity and the market risk factors is more uncertain.

Besides the risks mentioned in the QIS, other risk factors might as well be relevant for pension funds. Some pensions in the EU provide indexation of the pensions. When the pension benefits have an indexation guarantee, the technical provision of a pension fund increases significant. Therefore, it is important to investigate the differences in

(18)

indexation. Inflation linked indexation adds more risk to the pension fund, since the pension fund has to deal with inflation risk in case of this indexation. An inflation linked indexation might be conditional or capped. This reduces the inflation risk. Including inflation stress in the scenario might be relevant for some countries.

Macro risks

In the macro perspective there is also a need for risk identification. Jones et al.(2004) mentions to narrow the search for specific vulnerabilities in the financial system. Vulner-abilities that are connected with the pension sector are of interest. A good starting point for finding these vulnerabilities is Tower and Impavido(2009). They mention the pen-sion sector can affect the financial system through several channels. There can be effects on the balance sheet of the sponsor, the households, and the government. These effects are most relevant when a pension fund is in bad condition and security mechanisms are activated. For example, a security mechanism could be pension benefit reduction. This mechanism affects the balance sheet of the households. Another security mechanism for pension funds could be receiving support from either the sponsor, a pension protection scheme or the government. The investment activities of the pension funds can also affect the financial system.

Activating a security mechanism will improve the financial condition of a pension fund. But this comes at a cost. In case of sponsor support there might be an effect on the balance sheet of the sponsor. In economic downturns pension funds could face under-funding due to adverse developments. In this case they will ask for sponsor support. When the sponsor is already in trouble due to the economic downturns this might be a problem. Hence, sponsor support could be a transmission channel to the real economy. For a large investor like the Dutch pension sector the investment activities might impact the financial system. Several studies investigate the behavior of the Dutch pension funds.

Bikker et al. (2010) find that the performance of the equity affects the investment behavior of pension funds. Pension funds can rebalance there asset allocation when this is affected by the performance of the assets. When the equity values decrease, a pension fund might buy additional equity in order to keep the portfolio mix constant. The pension fund then invests in an anti-cyclical way. Which is good for the financial stability and the entire financial system.

What does this implicate for the EU-wide pension stress test? The security mechanisms for a pension fund in low financial condition should be investigated. Next the effect of the security mechanisms on the financial system should be examined. Also the investment behavior of a pension fund in a weak financial condition should be investigated.Bikker et al.(2010) andDe Haan and Kakes(2011) find that pension fund investment behavior doesn’t have a negative effect on the financial system.

3.3 Valuation in the EU-wide pension stress test

In the update on the stress test EIOPA announces that national as well as common valuation standard will be applied in the stress test (EIOPA,2014c).

There is no standard valuation method used in the regulation for the pensions across the EU. However, EIOPA is working on a common valuation method, called the holistic balance sheet. This balance sheet controls for security mechanisms such as sponsor support and reduction of pension benefits. The valuation of these security mechanisms can be helpful when determining the effect on the economic system. But they could also cover up the weak condition of a pension fund. Although security mechanisms could be activated and improve the financial condition of a pension fund, they still come at

(19)

a price. This price is for example paid by the sponsor of the pension fund in case of sponsor support. In a stress test the security mechanisms could absorb all the stress. The price paid for activating a security mechanisms should always be considered. The holistic balance sheet is still a theoretical valuation method and is not used in the regulation of pension funds.

National valuation methods are still in place for the regulation of pension funds. There-fore, it makes sense to conduct the stress test under national valuation methods. Im-pavido (2011) mentions the diverse valuation methods applied to pension funds in dif-ferent countries, this is also the case in the EU. Some of these differences in valuation are discussed in the next section.

3.4 Differences national valuation methods in the EU

The valuation method is crucial in measuring the condition of a financial institution. Different valuation methods could result in different impacts of the stress scenario. This would lead to less comparability. For this reason, I investigate the impact of aspects of different valuation methods used across the EU. In this section, I discuss a few important aspects of national valuation methods across the EU.

Asset valuation

There are two options to value the asset side of the balance sheet, market valuation and book valuation. Market valuation has become popular. For example, market consistency is one of the four overarching principles in CEIOPS (2008). In market valuation the assets are valued at market price. Therefore, they can change during the holding period of the asset. Book valuation is also applied in some countries. Book valuation can be applied in order to create additional buffers (CEIOPS,2008). Assets values under book valuation remain unchanged during the holding period of the assets.

Mortality trend

The mortality table is an important part of the valuation method. The mortality table is applied to the accrued pension rights in order to calculate the expected cash flows of the pension fund. The life expectancy of populations across the EU might vary. Hence, the mortality tables are also different. Another difference in the mortality table is not caused by the difference in the life expectancy of participants across the EU. Some mortality tables include a trend over time. A trend changes the mortality probabilities over time, following the expectation of the life expectancy in the future. The trend could provide a better approximation of the mortality probabilities in the future.

Discount rate

The discount rate is one of the most important parameters when calculating the tech-nical provision of pension fund. By applying the discount rate to the cash flows of the pension liabilities, the technical provision can be calculated. Three different discount methods are used across the EU: fixed discount rate, discount rate based on the ex-pected return over the assets and discount curve based on the risk free curve. When a fixed discount rate is applied, the discount rate is just a fixed number. For example, a fixed discount rate could be 4%. The fixed discount rate is applied to discount the all cash flows of all maturities.

(20)

The discount rate can also be based on the expected return over the assets. First the expected return for each asset class has to be determined. Then the expected returns are multiplied by the weight of the asset class in the asset portfolio and summed. That is, the weighted average of expected return of the asset portfolio is calculated. This expected return is applied to discount the cash flows of all maturities. This is also known as a flat rate.

The last discount rate is the risk free term structure. In this method the term structure of risk free assets is applied to discount the cash flows. This term structure can also be constructed by swaps. Notice that this discount method does not have a flat rate for all maturities, but has a different discount rate for different maturities.

(21)

EU-wide bank stress testing

In this chapter, the 2014 EU-wide stress test executed by the European Banking Au-thority (EBA) is described. Features of this stress test could be applied to the EU-wide pension stress test. First, I look at the differences between banks and pension funds relevant for a stress test. Then, the characteristics of the EBA stress test are discussed. Last, I analyze which features of the EBA stress test are applicable for the EU-wide pension stress test.

4.1 Difference between bank and pension funds

There are many differences between banks and pension funds. In this section, I discuss the differences relevant for stress testing. The first difference is the character of the liabilities of banks and pension funds. The liabilities of banks and pension funds have a different character. Where banks have mostly short term liabilities, pension funds are known for their long term liabilities. Short term liabilities like deposits are a typical liability of banks. Pension funds may need to provide participants with a pension benefits in 40 years. Therefore, the liabilities of the pension funds are typically long term. The asset side of the balance sheet also varies for pension funds and banks. Equity and bonds are typical pension assets. Banks might have loans and mortgages as assets. This difference might have an effect on the riskiness of the assets. Especially if a pension fund invests a lot in high grade government bonds, the riskiness of the pension fund liabilities might be lower than the bank assets.

Since banks lend money to each other, the risk of contagion is present. The risk of contagion could be a reason to consider banks system relevant. Pension funds do not lend money to each other. Therefore, the contagion risk for pension funds is assumed to be zero.

Banks are more international institutions compared to pension funds, many banks op-erate in multiple countries. International regulation is needed in order to create a level playing field. International standards for bank regulation started in 1988 with the Basel accord (Hull,2012). The ECB is currently working on the Single Supervisory Mecha-nism (SSM). The SSM will supervise the large European banks. The SSM is an example of the international supervision and regulation of banks. The international Basel III regulatory standard is another example of an international standard for banks. There might still be different regulation between banks in the EU, but it is fair to state bank regulation is very international and thus similar across the EU. This level playing field makes it easier to perform a common stress test as the banks face similar regulation and valuation. In contrast, the pension sector is not as internationally regulated as the bank sector. As mentioned in Chapter 3, the pension regulation varies across the EU.

(22)

4.2 EBA EU-wide bank stress test

In this section, the most important features of the EBA stress test are discussed (EBA,

2014). The measure, the scenarios and the follow-up of the stress test are considered. The EBA uses a bottom-up stress test where they also provide a methodology. In a bottom-up stress test, the EBA designs the stress test scenarios and the banks calculate the impact of the scenarios using their own models. However there is a strict supervision by both the national authority and the ECB on the methodology used by the banks. The banks are challenged on the methodology.

The stress test consists of multiple scenarios. There are three different types of scenar-ios in the stress test: securitization scenarscenar-ios, market scenarscenar-ios and a macroeconomic adverse scenario. The securitization scenarios are not relevant for pension funds. Hence, the focus is on the macroeconomic adverse scenario and the four market scenario. The storyline of the market scenarios is not very clear. The only information about the storyline for the market scenario is the name of the scenario, historical scenario 1, 2, 3 and 4. The name of the scenario suggests the scenario is based on a historical event. However, it is unknown which historical event is mentioned. As a consequence, understanding when this scenario will materialize is difficult. To my opinion this is a missed opportunity for the relevance of the scenarios. The absence of a storyline makes it unable for the management of the banks to link the stress test results to a certain state of the world. Four sources of stress are mentioned in the macroeconomic adverse scenario. Since the cause of the scenario is clear, the management of the banks will be able to link the stress test results to a certain state of the world, which is useful from a risk management perspective.

The four market risk scenarios contain multiple market stresses. There are interest rates stresses, exchange rates stresses, equity stresses, funds stresses, commodities stresses, credit stresses, correlation trading portfolio stresses, basis risk stresses and market liquidity stresses. The macroeconomic adverse scenario contains much more stresses. Macroeconomic parameters are also specified and stressed. For example, the inflation paths are also defined in the macro economic scenario.

The horizon of this stress test is three years. This means the scenario starts in the beginning of 2014 and finishes at the end of 2016. For financial institutions with a short term liabilities, like banks, a three year horizon seems to be reasonable. This horizon also allows the macroeconomic parameters to fully affect the model. Macroeconomic parameters could affect mortgages and deposits on the balance sheet of the bank. The Common Equity Tier 1 ratio is applied for measuring the financial condition of the bank. Common Equity Tier 1 is a specific type of capital. This capital should be able to absorb the stresses from the scenarios. After applying the scenario, the bank should still have a certain amount of capital. The hurdle rates of the stress test are expressed in a ratio of this capital. A bank passes the stress test when the ratio after applying the scenario is at least the hurdle rate ratio.

The stress test assumes a static balance sheet. This implies banks cannot adjust their portfolio in the stress test. Maturing assets and liabilities are replaced by similar assets and liabilities.

The purpose of this stress test is not only to investigate the impact of a scenario, but also to ensure the banks are able to withstand the scenario. This stress test is clearly a part of the regulatory framework since capital requirements are set. The hurdle rates lead to a strict follow-up of the stress test. The banks can either pass or fail the stress test. A bank that fails the stress test should increase their amount of capital. The strict follow-up of the stress test could be explained by banks being system relevant.

(23)

4.3 Possible features for the EU-wide pension stress test

In the set-up of the EU-wide pension stress test some parts of the EBA stress test could be adopted. This section discusses which features of the EBA stress test should be adopted. The bottom-up approach from the EBA stress test can be applied in the EU-wide pension stress test. The bottom-up approach fully utilizes the specific knowledge of the pension funds.

The management of a pension fund might be inclined to preform optimistic management actions in the stress test. The static balance sheet assumption could also be applied to the EU-wide pension stress test, to prevent the influence of management actions. Only the market risks relevant for pension sector should be copied to the EU-wide pension fund. Market risk factors specific for bank assets or liabilities should not be included in the EU-wide pension stress test. For example, bond spread risk could be included in the pension stress test since most pension funds hold bonds.

The horizon of the EBA stress test should not be copied depending on the purpose of the stress test. Only if a short term effect needs to be tested, the horizon of the EBA stress test can be copied. In general the liabilities of a pension fund are longer than the banking sector and thus the horizon of the stress test should also be longer. An alternative could be to apply instantaneous stresses in the EU-wide pension stress test. Here the scenario is implied at the current time and does not develop over time. The follow-up of the EU-wide pension stress test should be less strict than the EBA stress test. First of all, banks are more system relevant than pension funds. The con-tagion risk is much higher in the banking sector. Therefore, they should be conducted to stricter regulation and supervision. Second of all, the international regulation for banks is more developed than the regulation for pension funds. Hence, more interna-tional standards are at hand for the banking sector. These common measures lead to more comparability between banks, which in turn makes it easier to conduct a stress test. Third of all, the pension systems across the EU vary more. Therefore, comparing the pension sectors across the EU is hard and in turn creating a common capital re-quirement is difficult. The purpose of the EU-wide stress test is not to create a capital requirement (EIOPA,2014c).

(24)

Chapter 5

EU-wide insurance stress testing

In this chapter, the EU-wide stress test for insurers conducted by the EIOPA and possible features for the EU-wide pension stress test are discussed. First the differences between insurers and pension funds are reviewed. Second, some of the characteristics of the EIOPA insurance stress test are considered. Third, I specify features of the EIOPA insurance stress test, that could be applied for the EU-wide pension stress test.

5.1 Difference between insurance company and pension

fund

Broeders et al. (2011) provide an overview of the differences between insurance compa-nies and pension funds. They mention the institutional structure, the risk bearing, the contract specifications, the investment policy, the regulation parameters and the policy variables. The differences in the risk bearing, the investment policy and the policy vari-ables are most relevant for a stress test. The regulation parameters are also relevant for the severity of the scenario, but the severity of the scenario is more a political choice. As mentioned in Section 3.1 the risk in pension funds in the EU is borne in multiple ways. Depending on the security mechanism in place, the risk is borne by either the entire pension sector in case of a pension protection scheme, the sponsor in case of sponsor support or the participants in case of pension reduction (Broeders et al.,2011). The risk in insurance companies is borne by the external stakeholders, since they will be affected by the underfunding of an insurance company.

There is also a difference in the indexation of pension funds and life insurance. A typical annuity from a life insurer has no indexation. However, there are also products that include a profit-sharing feature and could have an indexation feature. Most pension systems in the EU provide some kind of indexation. This indexation is linked to either the inflation or the wages. Only half of the indexation across the EU is unconditional (CEIOPS,2008). This difference in indexation implies a different exposure to indexation risk. Where insurance companies face more risk linked to the performance of a certain index, pension funds face more risk on inflation or wages increase.

Broeders et al.(2011) mention a difference in the investment behavior of pension funds and insurance companies. They suggest pension funds have a larger mismatch than insurance companies. Since some pension funds attempt to provide indexation in the pensions, they are more likely to invest in assets that are linked to the pension in-dexation. This difference in investment could be interesting for stress testing, as the institutions have a different exposure due to their investment mix. This might have an impact on the risk factors in the stress test.

(25)

Another difference between the pension and life insurance could be the stability of the new business. Reputation risk is a very relevant risk for insurance companies. Repu-tation risk in insurance is related to the repuRepu-tation of either the insurance sector or the entire financial sector. Reputation damage might occur when a significant insurer or other financial institution fails or has other significant troubles and could result in mass lapse. Pension funds are less sensitive to reputation risks, since participants will always participate in mandatory pensions. Therefore, the pension funds are not exposed to mass lapse risks.

The regulation of the EU insurance companies is entering a new milestone with Solvency II. Solvency II is a risk based supervisory framework for EU insurance companies and should be implemented in 2016. Although the Solvency II framework is not in place yet, the insurance business has a common valuation for the condition of the insurer. As mentioned before, the pension sector in the EU still lacks this common valuation, although the holistic balance sheet is an attempt to a common standard.

5.2 EIOPA EU-wide insurance stress test

One of the objectives of EIOPA is to identify the financial institutions that pose a systemic risk. This is a macroeconomic objective. However, they also want to test the resilience of the financial institutions to adverse developments. This is a micro economic objective. Feautures of the EIOPA insurance stress test are discussed in this section (EIOPA,2014a).

Measure

In the EIOPA insurance stress test the condition is measured by the valuation method from the latest Solvency II regulation. Since Solvency II is an EU-wide regulation, the valuation is equal across countries and so the results of the EIOPA insurance stress test are comparable. Although Solvency II is not completely in force yet, most the insurance companies are likely to be prepared for the Solvency II framework.

Scenarios

The EIOPA insurance stress test consists of multiple scenarios. The scenarios are divided into two different modules, a core module and a low yield module. The core module contains two scenarios, “adverse 1” and “adverse 2”, which affect multiple risk factors relevant for insurance companies. The stresses in this module are applied instantaneous. Therefore, there is no horizon in this module. Both scenarios in the core module consist of market stresses and single factor insurance stresses and are bottom-up stress tests. Notice, EIOPA decided to stress the market stresses and the single factor insurance stresses independent of each other.

The market stresses in the core module affect multiple asset classes and the risk free term structure. Both market scenarios have a hypothetic source of stress. This means the source of the stress is not a historical event. The first scenario starts from the equity market and the second scenario starts from the non-financial corporate bond market. The source of stress is very clear. The equity market scenario has a severe equity shock of 41%, where the non-financial bond scenario has an equity shock of only 21%. On the other hand, the spread shocks are more severe in the non-financial bond scenario. Differences compared to the bank stress test are the insurance factor stresses and the horizon. Insurance companies face different risks than banks. The single factor insur-ance stresses in the core module include 5 non-life catastrophes, provisions deficiency, longevity, mortality and mass lapse stress.

(26)

The second module is the low yield module. This module contains of two scenarios, “low yield 1” and “low yield 2”, where the focus is on a low yield for an extended time. A long term low interest environment could impact a long term investor, like a life insurer. The cash flows for the next 60 years will be examined in this module. However, since the stresses are applied instantaneous, there is no horizon. This module starts with a bottom-up approach where the insurance companies calculate with their own models. Then EIOPA will perform own calculations in a top-down test based on these results. The stress test also contains questionnaires for the investigation on the dynamic responses from the insurance companies to the scenarios.

Follow-up

The public report of the 2014 stress test will not disclose individual results of the participants. From the client point of view this is not very transparent. The client is not able to check the performance of the insurance company under stress. In contrast to the bank stress test, the insurance stress test has no clear hurdle rates. The results of the stress test will be handed to the national regulator and she will consider the results in the evaluation of the insurance company. In this way the stress test serves more as a risk management tool. That is, to investigate the impact of a scenario.

5.3 Possible features for the EU-wide pension stress test

The measure of the condition of insurance companies should not be copied for the pen-sion funds. A penpen-sion fund faces similar risks as insurance companies. However, penpen-sion funds handle underfunding in a different way, since they have security mechanisms to deal with it.

The scenario from the insurance stress test could be applied to pension funds. Most of the market stresses are relevant for pension funds, especially interest rate stresses and equity stresses, since these are key pension risks. In the single factor insurance stresses only the longevity is relevant for the pension funds, since longevity is a key risk for pension funds, see Section3.2. The separation of the market risk and the insurance risk is an option for the pension stress test. By doing so, there is no need for a complicated correlation matrix. The combination of a more general scenario (core module) and a focused scenario (low-yield module) can also be applied in the EU-wide pension stress test.

Although the scenarios from the insurance stress test are relevant for an EU-wide pen-sion stress test, the differences from Section5.1should be taken into account. Indexation and security mechanisms are not captured by the insurance stress test scenarios. The differences in investment behavior requires a different focus in the market stresses. The follow-up from this stress test is applicable for the pension fund stress test. This follow-up implicates the impact of a scenario is investigated in the stress test and na-tional authorities should judge whether the outcome of the stress test is acceptable. As mentioned in Section 4.3, a strict follow-up is not applicable for the EU-wide pension stress test.

(27)

Set-up quantitative analysis

As mentioned in Section3.3, EIOPA plans to apply both national valuation methods and a common valuation method in the EU-wide pension stress test. Hence, a quantitative analysis of the influence of national valuation methods on the impact of the stress test is interesting. The following approach is applied in the quantitative analysis. A fictive pension fund is valued under different valuation methods. Then a stress test is conducted to the fictive pension fund. Since there are no pension scenarios available, two scenarios based on the EIOPA insurance stress test scenarios are conducted. After the scenario is applied, the fictive pension fund is again valued under the different valuation methods. In this chapter, a description of the model and assumptions for investigating the effect and issues of applying national valuation methods in the EU-wide pension stress test is given. In Section 6.1, a description of the fictive pension fund and the measure of the condition of a pension fund is provided. In Section6.2, the scenarios of the quantitative analysis is described. In Section 6.3, the differences in national valuation methods are described and the effect on the stress scenario.

6.1 Model description

In this section, the fictive pension fund and the measure are introduced. This fictive pension fund is based on the Dutch pension funds, since Dutch data was available. The measure for the financial condition of a pension fund is the funding ratio and is determined from the balance sheet of the fictive pension fund. The balance sheet of a fictive pension fund is therefore necessary to determine the impact of a scenario. This section provides a description of the asset and liability side of the balance sheet. And later a description of the funding ratio.

Assumptions assets

The assets of the fictive pension fund are based on the Dutch pension sector. However I simplified the asset mix to only two asset classes, stocks and fixed income. From data on the De Nederlandsche Bank, the average asset mix consists of 40% stocks and 60% of fixed income (DNB,2014). The fixed income can then be divided into government bonds and corporate bonds. The bond portfolio consists of 65% of government bonds and 35% of corporate bonds. The government bonds consist of a mix of Dutch government bonds with different maturities. The corporate bonds consist of AA corporate financial bonds of different maturities. I matched the duration of my bond portfolio with the average duration of the Dutch pension fund (DNB,2014). The asset mix is constant throughout the quantitative analysis.

(28)

Assumptions liabilities

The liabilities consist of the technical provision and the own funds. The technical pro-vision is the amount that a pension fund stores in order to meet the pension obligation. Assumptions on the distribution of the participants are necessary for calculating the technical provision. My fictive pension fund is based on the aggregated data of Dutch pension funds. The participants of the pension fund exists of active workers, sleepers, retirees and surviving dependants. Sleepers are participants who accrued there rights in the past, are not accruing new pension rights and have not reached the retirement age yet. Notice not only the retirement pension, but also the surviving dependant’s pension is included.

In order to calculate the technical provision, the pension obligation has to be valued. In my fictive pension fund a defined benefit pension with no inflation indexation is assumed. In this pension system the benefits are defined and the premium is based on the future benefits. When the mortality probabilities of the participants are taking into account, the pension obligation are converted into a best estimate of the pension cash flows by applying a mortality table. In this thesis, the mortality table from the Dutch Actuarial Society from 2012 is applied. The best estimate of the pension cash flows are discounted by a certain discount rate in order to calculate the present value of these cash flows. Note that the discount rate can be flat or vary over time (this is called a term structure). The present value of these cash flows is the technical provision. The own funds of the fictive pension fund are calculated by reducing the assets by the technical provision. The own funds are negative, if the technical provision is higher than the assets of the pension fund.

The underlying assumptions about the pension obligation is constant in the quantita-tive analysis. That is, the number of workers, sleepers, retirees and surviving dependants stay the same and the accrued pension rights of these participants are constant through-out the analysis. However, valuation of the pension obligation can be different in the quantitative analysis. The mortality table varies by including or excluding a trend. This has an effect of the transformation of the pension obligation to the best estimate of the pension cash flows and in turn to the technical provision. The discount rate also varies in the quantitative analysis, which affects the calculation from best estimate of pension cash flows to technical provision.

The measure of the impact

A measure of the condition of the pension fund is required to determine the impact of the scenario. Impavido(2011) states solvency rules in most cases demand the assets to be higher than the liabilities by some margin. Here liabilities means the technical provision and not the total of the liability side of the balance sheet. This technical provision and the own funds together are the liability side of the balance sheet of the fictive pension fund. Therefore the impact should be measured by the assets relative to the technical provision. A basic measure to do this is the funding ratio. The funding ratio is the measure for the condition of pension funds in the Netherlands. The funding ratio is

FR = assets

technical provision. (6.1)

Here the assets and technical provision are from the fictive pension fund. A high funding ratio indicates a well funded pension fund and vice versa. The scenario could impact the assets and the technical provision of a pension fund and in turn impact the funding

(29)

ratio. The funding ratio is calculated before and after the scenario. The difference in the funding ratio is the impact of the scenario:

∆FR = FR2− FR1. (6.2)

Here ∆FR is the impact of the scenario, FR1 is the pre stress funding ratio and FR2 is

the post stress funding ratio.

6.2 Scenario

To determine the influence of valuation on a scenario, a scenario for the fictive pension fund is needed. The scenarios applied to the fictive pension fund, are based on the “adverse 1” scenarios from the core module of the EIOPA insurance stress test discussed in Section 5.2. This “adverse 1” scenario consists of multiple stresses. In the EIOPA insurance stress test, the market stresses and the single factor insurance stresses are applied separately. This approach is also followed in this thesis. Hence, a single factor insurance stress and a market stress are applied separately to the fictive pension fund. Not all single factor insurance stresses are applied to the fictive pension fund. Since a pension fund is not exposed to mass lapse and non-life stresses, this stresses are not relevant. In Section 3.2, mortality stress is not a key risk. In contrast, longevity is a key risk for pension funds. Therefore longevity stress is the only single factor insurance stress applied. The longevity stress is applied by reducing all the mortality probabilities by 10%, as shown in Table 6.1.

Table 6.1: Longevity stress parameter overview. In the longevity scenario, the mortality probabilities in the valuation of the fictive pension fund are reduced by this parameter.

Stress Stress parameter

Longevity -10%

All relevant stresses from the insurance market stress scenario are applied to the fictive pension fund. Since the assets of the pension fund only consists of equity, Dutch govern-ment bonds and AA corporate financials bonds, only these stresses from the insurance market stress scenario are applied to the fictive pension fund. As shown in Table 6.2, the useful stresses are equity stress, the interest rate stress, the corporate bond stress financials and the sovereign bond stress. The interest rate stress in this table is “stress 1”, which is a stressed term structure provided by EIOPA in the EIOPA insurance stress test.

Table 6.2: Market stress parameters overview. In the market scenario, these parameters are applied to the fictive pension fund.

Stress Stress parameter

Equity -41%

Corporate financials AA bond spread 35 bps

NL bond spread 17 bps

Interest rate stress 1

The bonds in the portfolio of the fictive pension fund have a market value at the reference date of 31 December 2013. Discounting the bond cash flows with the risk free term structure doesn’t lead to the market values. Therefore a bond spread is added to the risk free term structure to create a bond term structure. Discounting the cash flows of

(30)

the bond with the bond term structure results in the market values of the bonds. In Figure6.1, the blue line is the risk free term structure and the red line is the bond term structure.

The value of the bonds changes in the market stress scenario. In the market stress scenario a different term structure is applied to the cash flows of the bond. In the EIOPA insurance stress test, this stress bond term structure is the old bond term structure plus a stress bond spread. This approach is followed when the market scenario is applied to the fictive pension fund. For example, the stress bond spread for Dutch government bond is 17 basis points in the adverse 1 market scenario of the EIOPA insurance stress test, see Table6.2. In this example the stress bond term structure is obtained by applying a parallel shift of 17 basis points to the bond term structure. Notice the stress risk free term structure doesn’t affect the stress bond term structure in this scenario. In Figure

6.1, the yellow line is the stress risk free term structure and the grey line is the stress bond term structure.

Figure 6.1: Example of pre and post stress term structures. The bond term structure is the risk free term structure plus a bond spread. The stress bond term structure is the bond term structure plus a stress bond spread parameter.

6.3 Valuation options in stress assumptions

As mentioned in Section 3.4, there are differences in the valuation of pension funds across the EU. In this model three aspects of the valuation method are varied. The asset valuation is either book or market valuation. The mortality table is with or without a trend. And the discount rate can be fixed, based on expected return over the assets or based on the risk free term structure. Combining these aspects creates twelve possible combinations of the valuation options, as is shown in Table6.3. For example, a valuation option could be: asset valuation on book values, a trend in the mortality table and a fixed discount rate.

These valuation options have an impact on the valuation of the fictive pension fund. In a stress test, the fictive pension fund is first valued before the scenario is applied. This phase is called the pre stress scenario or the baseline scenario. Then the stress scenario is applied and the fictive pension fund is valued after the stress scenario (post stress). Hence, valuation assumptions are essential in a stress test.

EU-wide pension stress testing, the influence of valuation

influence of valuation

Bram Baneke

Contents

Preface

Introduction

Stress testing in general

2.1

The concept of stress testing

2.2

Why use stress testing

2.3

Creating a stress test

2.4

Options for the scenario

Chapter 3

The EU pension sector

3.1

Differences within the EU pension sector

3.2

Risks in pension funds

3.3

Valuation in the EU-wide pension stress test

3.4

Differences national valuation methods in the EU

EU-wide bank stress testing

4.1

Difference between bank and pension funds

4.2

EBA EU-wide bank stress test

4.3

Possible features for the EU-wide pension stress test

Chapter 5

EU-wide insurance stress testing

5.1

Difference between insurance company and pension

fund

5.2

EIOPA EU-wide insurance stress test

5.3

Possible features for the EU-wide pension stress test

Set-up quantitative analysis

6.1

Model description

6.2

Scenario

6.3

Valuation options in stress assumptions