CHAPTER 19 Risk-Based Supervision of Pension Funds in the Netherlands Dirk Broeders and Marc Pröpper CONTENTS 19.1 I ntroduction 19.2 Dutch Pension System and Solvency Regulation 19.2.1 Objectives of Prudential Pension Fund Supervision in the Netherlands 9.2.2 Prudential Supervision 19.2.3 Risk and Time 19.2.4 Financial Assessment Framework 19.3 Ma rk-to-Market Valuation 19.3.1 Valuation of Defined Benefit Liabilities 19.3.2 Valuation of Contingent Liabilities 19.3.3 Term Structure of Interest Rates 19.3.4 Valuation of Assets 9.3.5 Contribution Policy 19.4 Risk-Based Solvency Requirements 19 4.1 Standardized Method 19.4.1.1 Interest Rate Risk 19.4.1.2 Equity and Real Estate Risk 19 4.1.3 Currency Risk 9.4.1.4 Commodity Risk 19 4.1.5 Credit Risk 9.4.1.6 Insurance Risk 474 475 478 480 481 484 484 485 486 488 489 490 490 491 492 493 494 494 494 495 473 © 2010 by Taylor and Francis Group, LLC 474 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling 19.4.1.7 Other Risk Categories 19.4.1.8 Overall Capital Charge 9.4.2 Internal Models 19 4.3 Simplified Method 19.4.4 Minimum Capital Requirement 9.4.5 Recovery Plans 19.5 C ontinuity Analysis 19 5.1 Parameters 19.5.2 Risk and Outcome Distributions 19 5.3 Different Functions of the Continuity Analysis 19.5.4 Insight into Indexation 19.6 C oncluding Observations References 507 496 496 497 498 499 500 501 502 503 505 505 507 R isk-ba sed super v ision is being i ncreasingly adopted worldwide As an example of such r isk-based supervision, this chapter reviews the Financial Assessment Framework for pension f unds in t he Netherlands It operates under the Pension Act that was changed significantly in 2007 The main elements of this framework include mark-to-market va luation of a ssets a nd l iabilities, a f ull f unding r equirement, r isk-based so lvency requirements, a nd a co ntinuity a nalysis for assessing t he pension f und’s solvency in the long run Together these building blocks offer a solid foundation for t he incentive-compatible regulation of occupational pensions in the Netherlands 19.1 INTRODUCTION The current financial crisis shows that pension funds have developed a large exposure to risks This exposure entails market risk, interest rate risk, and even liquidity and operational risk, besides more traditional insurance risks like mortality risk and longevity risk It has become evident that the key to managing these risks is an integrated balance sheet approach, considering assets and liabilities simultaneously This is particularly true for defined benefit pension plans that usually run a mismatch between assets and liabilities on their balance sheet The principal responsibility of pension fund trustees, therefore, is to control risks in order to ensure that they are kept within acceptable limits so that the entitlements of members, as agreed between social partners in employment contracts, will be respected Being i mportant financial i nstitutions, pens ion f unds a re sub ject t o government regulation Worldwide, there is an increasing adoption of © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 475 risk-based supervision As an example of such risk-based supervision, this cha pter r eviews t he F inancial A ssessment F ramework f or pens ion funds in the Netherlands This regulatory framework has been discussed in, inter alia, Siegelaer (2005), Nijman (2006), and Brunner et al (2008), although th is c hapter i s th e first t o r eview i t i n f ull s ince i t s be en adopted in the Dutch Pension Act in 2007 The main elements of this solvency regime are as follows The mark-to-market valuation of assets and liabilities i s de scribed i n Section 19.3 This ma rk-to-market a pproach is needed to determine whether a pension fund is currently solvent Section 19.4 de scribes t he r isk-based so lvency r equirements t hat a re i n p lace t o make sure the pension fund is still solvent on a 1-year horizon with a high probability Section 9.5 i ntroduces t he l ong-term co ntinuity a nalysis This instrument assesses the pension fund’s solvency in the long run and the f und’s ab ility t o effectively influence t he financial p osition t hrough the available policy instruments Section 19.6 contains some concluding remarks However, we first describe t he key characteristics of t he Dutch pension system and elaborate on the objectives of prudential pension fund supervision in the Netherlands 19.2 DUTCH PENSION SYSTEM AND SOLVENCY REGULATION The Dutch pension system may be characterized in terms of the usual three layers The first layer is constituted by the state old-age pension, which is financed on a pay-as-you-go basis and provides for a basic income for all citizens of age 65 and over As of 2009, for singles, the gross pension benefit is €1038 a month For couples, if both partners are 65 or over, the gross benefit for each partner is €723 a month It should be noted, however, that these amounts are only paid if a perso n has uninterruptedly lived in the Netherlands f rom t he age of 16 onward Sk ipping t he second layer for a moment, the third layer consists of private saving for retirement This covers tax-favored pension saving, such as life annuities offered by insurance companies The second layer, the main focus of this chapter, comprehends occupational pensions, which are primarily financed by means of contributions by employer and employees It is therefore a capitalized or funded system in which people save for their pensions For most employees, participation in a pension plan is automatically linked to their contract of employment The participation rate among employees is thus close to 100% Typically, the employer pays the bulk of the contributions although the employees’ © 2010 by Taylor and Francis Group, LLC 476 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling part ten ds t o i ncrease The seco nd la yer ser ves t o su pplement t he first Hence, in occupational pension accrual, account is taken of a basic benefit, known as the pension offset This offset is often linked to the state old-age pension benefit For that reason, the supplementary pension is built up on the basis of pensionable earnings, that is, income from labor less the offset If, under a final-pay system, we assume an annual accrual rate of 1.75% of pensionable e arnings, t he result a fter years of ser vice i s a pens ion equal to 70% of final salary Final-pay systems a re ex pensive because of the past service involved Nowadays, some 87% of all active members participate in a career-average scheme Under such a scheme, higher accrual rates are mostly used, of up to more than 2% These schemes go without past service, the pension accrual being indexed to wage growth or inflation instead This indexation is not guaranteed however—it is contingent, and dec ided upon mostly on a y early ba sis, on t he availability of a mple financial resources In over half of all cases, indexation is based on wage growth, usually those in the industry concerned For over 20% of pension plan m embers, pens ion acc rual i s l inked t o t he m ovements i n t he g eneral l evel of consumer prices The so -called collective defined contribution plans are gaining popularity in the Netherlands These plans combine a c areer-average sch eme w ith a fi xed co ntribution r ate f or a n umber o f years This allows corporations to classify a defined benefit scheme a s a defined co ntribution p lan I ndividual defined co ntribution sch emes a re still exceptional in the Netherlands At the end of 2008, the assets held by pension funds totalled €576 billion, while t he ma rked-to-market va lue o f t he l iabilities eq ualed € 606 b illion This translates into a funding ratio of 95%, which is historically low for the Netherlands Figure 19.1 shows the long-term progress of the funding ratio It is closely related to the development of the long-term market interest rate, showing t he i mportance o f ma rket va riables f or a ssessing t he financial well-being of pension funds Figure 19.2 plots the asset allocation over time Notably, in the 1990s, pension f unds i ncreased t heir eq uity ex posure a t t he ex pense o f fixedincome securities Furthermore, within the category of fixed-income investments, private loans have been replaced in large part by traded bonds The increased share of equities and bonds reflects a growing preference for liquid investments This is also reflected in investments in real estate: the percentage invested in has remained fairly stable, but there has been a shift toward i ndirect re al e state (traded p articipations) at t he e xpense of re al estate directly held by the pension fund itself Another trend is investing © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 477 240% 10% 220% 9% 200% 8% 180% 7% 160% 6% 140% 5% 120% 4% 100% 80% 1988 Funding ratio Long-term market interest rate (rhs) 1992 1996 2000 3% 2004 2% 2008 FIGURE 19.1 Long-term de velopment o f f unding r atio a nd l ong-term i nterest rate (From DNB.) 100% 80% 60% Other Real estate Fixed-income securities Equities 40% 20% 0% 1987 1991 1995 1999 2003 2007 FIGURE 19.2 Asset a llocation of Dutch pension f unds over time (Courtesy of Statistics Netherlands, The Hague, the Netherlands.) © 2010 by Taylor and Francis Group, LLC 478 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling in alternatives like commodities, hedge funds, and infrastructure Ther e is also a move toward investing abroad Currently, foreign assets account for some three quarters of the balance sheet total while this was still marginal in the mid-1980s The reason behind this shift is that international diversification is the key to a funded system to fully exploit the benefits of risk sharing Many pens ion f unds a im a t ma intaining a fixed a sset a llocation i n terms o f i nvestment cla sses (strategic a sset a llocation) This rebalancing strategy i mplies t hat cha nges i n t he relative va lue of financial assets g ive rise t o o ffsetting p urchases a nd s ales, so t hat t he r elative w eights i n t he portfolio remain fairly constant However, it is also possible to accommodate va lue cha nges w ithin defined ba ndwidths (tactical a sset a llocation) Pension funds’ rebalancing strategy is described in greater detail in Bikker et al (2007) Apparently, there is some asymmetric behavior Pension funds are eager to rebalance after a period of relative underperformance of equities b ut a llow t heir a sset a llocation t o f ree float after a pe riod o f eq uity outperformance The lion’s share of the accrued assets is administered by pension funds They u sually ve t he l egal f orm o f a f oundation a nd a re g overned b y representatives of employers and employees Being a f oundation, a pens ion fund cannot go bankrupt As such, t he ultimate measure a pension fund can t ake i s t o r educe acc rued ben efits t o r estore i ts financial position In fact, it is a legal requirement to incorporate this possibility in the pension fund’s statutes Pension funds may be linked to a single company or to an entire industry For the latter category, participation is usually mandatory, meaning that companies in certain industries are obliged to take part in the specific industry-wide pension fund At of the end of 2006, 86% of all active members took part in industry-wide pension funds, and only some 13% were members of a co mpany pension f und L ess t han 1% of a ctive member are related to occupational pension funds, which act in the interest of specific occupations like notaries or pharmaceutical chemists 19.2.1 Objectives of Prudential Pension Fund Supervision in the Netherlands Employers a re f ree t o offer a pens ion p lan t o t heir em ployees o r n ot However, if they so, the plan must comply with government regulation This regulation is laid down in a pension act that was changed significantly in 007 A ke y element of t his ac t i s t hat pension f unds must be l egally separated from the company offering the pension arrangement and must © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 479 always be f ully f unded These r equirements p revent a co rporate defa ult from damaging the pension benefits The rules for pension fund supervision are also set in the Pension Act This act identifies De Nederlandsche Bank ( DNB) a s t he r esponsible body f or t he p rudential su pervision o f pension f unds A longside t he D NB, t he N etherlands A uthority f or t he Financial Ma rkets (AFM) supervises t he conduct of t he entire financial market sector including pension funds The objective of prudential pension fund supervision is to offer a high degree of safety to (future) retirees through t he i mposition of st rict supervisory st andards Pension f unds are thereto required to hold a certain solvency margin of additional assets over the marked-to-market value of pension benefits This solvency margin is intended to absorb the risks inherent in the possible changes in the value of assets and liabilities Without a clear statutory requirement to hold such a ma rgin, pension f und’s t rustees may be tem pted to pursue short-term objectives by, say, following a risky investment policy and shift ing the burden o f pos sible n egative co nsequences o n t o y ounger g enerations o r, i n extreme cases, on to society at large However, D utch pens ion f unds c an ex ploit t he po tential o f i ntergenerational risk sharing; see Cui et al (2009) The basis for this solidarity is constituted by mandatory participation in pension schemes Older members can use the young as a safety net if need be, while the younger benefit from t he w ealth o f t he o lder i f t here i s a su fficient la rge su rplus i n t he pension f und This a llows a pens ion f und t o ben efit f rom sha ring r isks across generations if necessary However, in order to safeguard the savings both at present and into the future, the trustees of the fund need to avoid excessive risk taking, since the raison d’être of pension funds (i.e., crossgenerational risk sharing) holds only insofar as future workers continue to find it attractive to participate in the fund As such, i ntergenerational r isk sha ring cannot be st retched i nfinitely In order to demonstrate this, let us assume that there are two generations: current members and future members Intergenerational risk sharing may be compared to financial derivatives The current generation, which exercises discretionary power over the pension assets, has, so to speak, a long position in a put option That is, it has the power to sell its pension commitments t o la ter g enerations a t a g iven p rice The acc eding g eneration of (would-be) members have, so t o say, implicitly written the put option, thereby entering into the obligation to take over the pension fund’s commitments to the older generation should the assets managed by the fund prove insufficient to fund those commitments The crux of the matter is © 2010 by Taylor and Francis Group, LLC 480 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling that such a r isk transfer from the current to future generations carries a price tag This price is the surplus in the pension fund The owner of t he put option, i n t his c ase t he older generation, i nvariably s t he most favorable ex post pos ition I f t he a ssets i n t he pension fund are sufficient to serve both the older and the younger group, the former will see their entitlements (fully) honored If, however, the assets are insufficient to provide for both the older and the younger groups, the older group c an st ill en force f ull pension r ights a nd leave t heir juniors to pick up the tab Thus the youngsters, by definition, find themselves in a suboptimal position, because the pension fund’s money can only be spent once The protection of younger and future members in the pension funds from risk-taking behavior by t he older cohorts requires t he supervision of t he pension fund Another r eason f or p rudential su pervision i s t he absen ce o f ma rket forces Pension funds are not-for-profit organizations and as such not issue equity capital This means t he members i n a pens ion f und a re not only the liability holders but are in effect also its shareholders: the pension fund is a k ind of cooperative However, t he sha res a re not negotiable a s participation is ma ndatory By co nsequence, t he regular control mechanisms found in capital markets are inoperative in the pension industry 19.2.2 Prudential Supervision Prudential regulation is traditionally aimed at (1) defining, (2) overseeing, and (3) enforcing minimum requirements as to the funding and the solvency of the supervised institutions After all, if an institution has full coverage of the technical provisions and sufficient free capital, it is able to absorb setbacks In the absence of adequate reserves, pension funds run t he r isk o f fa ilure a nd ben eficiaries ma y co nsequently se e t heir claims a nd r ights e vaporate i nto t hin a ir P ension f und su pervision therefore centers on the continuity of pension entitlements That raises the question as to when this continuity is endangered One could argue that, for as long as pension plans attract new members, there is a safety net g uaranteeing t he ben efits f or t he o lder m embers T his a rgument relies on the assumption that it must be attractive for younger people to take part in the pension system If younger generations have to pay disproportionately more t han t he a mount required for t he accrual of their o wn pens ions, t hat i s n o l onger a utomatically t he c ase I n t hat event, intergenerational solidarity is no longer assured It is, therefore, of major importance that the distribution of pension contributions and © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 481 benefits among the generations should be kept within a certain bandwidth T his i s a sub tle eq uilibrium t hat ma y e asily be d isturbed b y large f luctuations in capital markets and, hence, needs to be carefully guarded by t he supervisor Ma intaining t he equilibrium is u nderlain by t he a pplication o f t he a ssumption t hat a pens ion f und sh ould be able at any point in time to distribute the claims and rights in money to the beneficiaries, that is, to have full funding at all times This is the core of the Financial Assessment Framework Before moving toward further explaining the Financial Assessment Framework, we first elaborate on a ke y a spect of pens ion f und r isk ma nagement: t he relation between risk and time This link is also relevant for designing incentive compatible regulation 19.2.3 Risk and Time This r elation i s i mportant t o u nderstand a s pens ion f unds u sually t ake a long-term perspective and so o ften d isplay a st rong inclination toward investing in equities There is an intense debate on the relationship between time and risk that is also relevant for the design of incentive compatible regulation We discuss the issue by looking at the characteristics of equity investments versus risk-free investments over time Pension f unds’ la rge st ock ma rket exp osures a re exp lained b y t he assumption that, in the long run, equities are expected to perform better than government bonds Looking ahead, this seems realistic: because equities are r iskier t han government b onds, investors demand hig her returns in co mpensation L ooking a t past p erformance, ho wever, t he picture may look different There are extended periods in the past when equities w ere o utperformed by r isk-free investments One exa mple is the period from 1902 until 1932 when T-bill investments outperformed equities o ver a 30 y ear p eriod d ue t o t he s evere ma rket b reakdown starting from 1929 Most recently a similar picture has emerged for the period 1995–2008 Due to the higher expected returns, the probability of loss on an equity portfolio relative to a risk-free investment declines as the horizon increases This so-called time diversification effect may be readily explained with the help of basic statistical knowledge Suppose the value of a well-diversified equity portfolio S follows a geometric Brownian motion with an instantaneous return µ and instantaneous volatility σ dS = µS dt + σS dW © 2010 by Taylor and Francis Group, LLC 482 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling An amount S invested in a risk-free zero coupon bond will yield an amount S er(T−t) over the time period T − t The probability p that equities outperform this risk-free investment is given by P (S e ( µ− σ2 )(T − t )+σ ε T − t > S er (T −t ) ) = p where ε ∼ N(0, 1) Evaluating this expression yields ⎡ (µ − 12 σ − r )(T − t ) ⎤ p=N⎢ ⎥ σ T −t ⎣ ⎦ where N i s t he c umulative s tandard nor mal d istribution f unction Fr om this relation, it follows that p is an increasing function of time to maturity; see Figure 19.3 Long-term investors use both the higher expected returns and this time diversification effect as arguments to justify their decisions to opt for equity investment They often add t he assumption that returns on equity revert to the mean in the long run Mean reversion would imply that periods of disappointing returns are followed by periods of above-average performance and vice versa; see, for example, Campbell and Viceira (2005) A pension fund, however, must not only consider the benefits attaching to risky investments Its trustees must form an opinion both on the probability that the return on the equity portfolio may fail to keep pace with 100% 0% 80% –20% 60% –40% 40% –60% Probability of return below risk free return Loss given default (rhs) 20% –80% Years 0% –100% 15 22 29 36 43 50 Probability of excess return and loss given default, using µ = 0.08, σ = 0.18, and r = 0.04 FIGURE 19.3 © 2010 by Taylor and Francis Group, LLC 494 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling lessons First, a h igh-risk profi le leaves f unds ex posed to severe downturns i n t he sh ort t urn S econd, t he l ong r un o ver wh ich r isky a ssets are expected to outperform risk-free assets may actually be much longer than that previously expected, even in relation to characteristic long horizon of pension liabilities 19.4.1.3 Currency Risk Currencies e xposures emerg e a s a si de-effect o f f oreign i nvestments Some pension f unds may hedge t heir c urrency r isk by engaging i n c urrency swaps or in forward contracts The capital charge for currency risk (S3) follows from an adverse shock 20% of all currencies against the euro This shock is calibrated using a w eighted average of seven representative currencies including the U.S dollar, British pound, Japanese yen, and the Argentinean peso as a proxy for emerging markets 19.4.1.4 Commodity Risk An important reason for investing in commodities, which is done through index-futures and -options, is risk diversification with other asset classes In recent years, most commodities have shown an upward trend that can be explained by the increased demand from fast-growing economies like China a nd I ndia This s ised i nterest a mong i nstitutional i nvestors Commodity investors have also witnessed short and severe dips, like the recent downturn in oil prices The capital charge for commodity risk (S4) follows f rom a p rice dep reciation o f 30 % The c alibration o f t his sh ock is based on a ba sket of commodities representing world commodities production over a long horizon 19.4.1.5 Credit Risk Credit risk is actually measured as spread risk in the Financial Assessment Framework The capital charge (S5) is derived from the impact on the pension fund’s equity of a general increase of credit spreads by 40%, i ndependent of the time to maturity This formulation is cyclical in the sense that the capital charge is low in benign periods and increases when credit markets de teriorate Idiosyncratic default r isk wa s abst racted u nder t he assumption t hat m ost i nvestments i n c redits a re i n b roadly d iversified portfolios of investment-grade corporate bonds (BBB or better in the S&P classification) This a ssumption a nd t he absen ce o f defa ult r isk ma y n o longer hold true In particular, it could well be t he case that the assumption of exposure to only systematic risk does not hold either for all funds © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 495 In the future, more emphasis will be placed on credit risk assessment as credits have become more popular as an investment class with a risk-return profile between risk-free governments bonds and risky equities 19.4.1.6 Insurance Risk The capital charge for insurance risk (S6) is the only risk category for which the regulator has not set predefined shocks in the standardized method The reason is that each and every fund must have the possibility to determine the actuarial risks corresponding to the specific characteristics of their own population of me mbers a nd b eneficiaries Nonetheless, sufficient demand existed for the DNB to publish standardized tables for the most important constituents of insurance risk: 1-year mortality risk, longevity risk, and the so-called negative stochastic de viations These t hree r isk t ypes a re a ggregated into a capital requirement for insurance risk by assuming full dependence (a correlation of 1) between 1-year mortality risk and the other two categories, and by assuming independence (a correlation of 0) between longevity risks and negative stochastic deviations The t wo la tter r isk c ategories, t hat i s, l ongevity r isks a nd n egative stochastic de viations, ve be en c entral t o ma ny d iscussions abo ut t he affordability o f pens ions u nder t he F inancial A ssessment F ramework They are not 1-year risk categories, but risks spanning the whole lifetime of t he l iabilities, a nd w ere t herefore pa rt o f tech nical p rovisions i n t he early F inancial A ssessment F ramework p roposals L ongevity r isk i s t he risk that the long-term trend in the improvement of survival rates turns out t o be st ronger t han ex pected It i s a s ystematic r isk m eaning e very member add s to t his r isk, a nd t here a re no d iversification effects over a group Negative st ochastic de viations a re a m easure o f t he u ncertainity in t he de termination o f t he tech nical p rovisions They a re r elevant f or smaller pension f unds where t he “ law of la rge numbers” does not apply in full and actuarial estimations carry a significant uncertainity For pension f unds w ith over 1000 members t his r isk quickly decl ines, a nd it i s almost absent for larger funds In fact, since both the 1-year mortality risk and negative stochastic deviations behave as 1/ n , n being the number of members and beneficiaries, large funds are basically only exposed to the longevity risk The capital requirement for longevity risk depends on the age of the members, for example, 10% for the cohort of 30 year old members and 4% for the cohort of 55 year old members The risk of a longerthan-expected lifetime (i.e., viewed from the pension fund perspective) is obviously higher for the young © 2010 by Taylor and Francis Group, LLC 496 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling A compromise was finally found in not picking up the two long-term risk c ategories i n a r isk ma rgin a s pa rt o f t he tech nical p rovisions, b ut as pa rt of t he capital for i nsurance r isks (together w ith 1-year mortality risk) The c apital f or i nsurance r isks i s o ften r elatively s mall f or t ypical funds with an exposure to market and interest rate risks, in part due to the assumed independence from all other risk categories 19.4.1.7 Other Risk Categories The liquidity risk, the concentration risk, and the operational risk are recognized as three separate risk categories However, at the inception of the Financial Assessment Framework they were assigned a 0% capital charge A lesson from the present crisis is that pension funds may be v ulnerable to liquidity risk, for example, related to derivatives collateral management Pension f unds are a lso exposed to asset liquidity risk: during periods of stress block trade prices may be well below recent market prices This risk becomes m ore r elevant f or ma ture pens ion f unds t hat ve a n et c ash outflow than for younger pension funds that are net receivers of pension contributions 19.4.1.8 Overall Capital Charge The overall capital requirement is just one step f rom t he i ndividual r isk capitals per category and follows from applying the prescribed correlation matrix to the individual risk capitals This correlation matrix takes a simple form and assumes all categories are independent (i.e., zero correlation) except for a 0.5 correlation between the capital charge for interest rate risk (S1) on the one hand and for equity and real estate risk (S2) on the other The overall capital charge (S) is therefore defined as S= Si2 + * 0.5 * S1S2 Σ i =1 The calibration of the correlation between S1 and S2 has generated much discussion; se e f or ex ample, N ijman ( 2006) S ome t ake t he l ong-term average as the point of departure and consider a l evel of a ppropriate Others believe that one should look at the correlation in times of stress and argue that 0.5 is too low In times of crisis, a plunge in stock prices often goes hand in hand with a drop in market interest rates as investors demonstrate a flight to quality Also, the assumption of independence across other categories seems to be too strong for the purpose for which © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 497 the c apital is ac tually i ntended W hen c apital is needed t he most, t hat is in times of stress, correlation is likely to be higher than that under normal conditions 19.4.2 Internal Models Instead of using the standardized method, the possibility of developing and employing a n i nternal model ma rked a t rue i nnovation for t he pension fund sector Internal risk modeling allows an independent risk-management function to align the capital requirement much closer to the factual risk profile th an the s tandardized method The su pervisor a lso s t he power to enforce a (partial) internal model if the actual risk profile deviates too much from the assumptions underlying the standardized method, as in the case of, say, a concentrated investment portfolio or a large exposure to idiosyncratic risks The use of an internal model is subject to prior approval by the DNB To obtain such approval, a pension fund must demonstrate that the model is truly embedded in the organization, and that the technical aspects of the model and the data used are adequate Besides being technically adequate, the internal model should therefore also be used in risk management and decision making (in the jargon of validation practitioners this is called the “use test”) The DNB can grant temporary approval for the use of an internal model even if it needs some further improvement However, this applies only if those shortcomings can be fi xed within the next years and are not material enough to prevent responsible use During such a transitional period, the pension fund can rely on prudent assumptions where necessary Even after full swing approval, elements of the internal model may still rely on parts of the standardized method, provided this choice is not motivated by opportunistic motives like regulatory arbitrage and does not harm the spirit of the internal model Reporting requirements for an internal model include a n a nalysis of t he cha nge in t he f unding ratio over t he last year and, once every years, a comparison with the outcome of the standardized method On a m ore technical level, there are basically two requirements for an internal model: the first is that the total capital charge needs to be set at the 97.5% confidence level at a 1-year risk horizon, the second and more challenging requirement is t hat t he model must be st ochastic i n nature The major advantages of this principle-based outline are that the model may be designed along a pension fund’s own definitions of risk categories and that © 2010 by Taylor and Francis Group, LLC 498 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling volatility and correlations may be e stimated more accurately One major disadvantage that may prevent funds from submitting their model is the relatively mild calibration of the standardized method The standardized method is ba sed on several a ssumptions t hat i n practice may u nderestimate risks, like abstracting from idiosyncratic risks, the 0% calibration of risk categories like concentration risk and the independence of risk categories (apart from the correlation of 0.5 between interest rate risks on the one hand and equities and real estate on the other) As a result, internal models have not yet received widespread attention from the sector As a final note on internal models, the idea of what constitutes an internal m odel d iffers ac ross r egimes F or i nstance, i n Ba sel I I t he i nternal ratings-based method (IRB) actually makes use of a predefined methodology: the asymptotic single risk factor method (ASRF) This only allows flexibility to banks to estimate the following underlying parameters: the probability of default (PD), the LGD, and the exposure at default (EAD) In contrast, the internal model under the Financial Assessment Framework offers g reat flexibility to pension f unds, a lbeit t his adva ntage of internal models is currently outweighed by the disadvantages of a r elatively mild calibration of the standardized method and of the practical complexity in the approval process 19.4.3 Simplified Method Next to the standardized method and an internal model, there is a t hird methodology f or de termining t he c apital r equirement This simplified method is introduced from the perspective of proportionality to ease the burden on relatively small and risk-averse pension funds Funds that meet certain criteria would not have to make any calculations if they assume a fi xed capital requirement of 30% of technical provisions, equal to a 130% funding ratio The criteria for approval include a straightforward pension scheme, a r isk-averse i nvestment st rategy, a nd a s imple business model Initial calculations during the development of the supervisory framework indicated that 30% would be the level of required capital for a typical pension fund and this number somehow became the symbol of the Financial Assessment Framework and this simplified methodology It was realized later that the requirement for the average fund profile actually relies heavily on the level of interest rates (all else being equal), and that when rates declined it became closer to 25%, but the requirement for the simplified test remains fixed at 30% to date This was not the reason, however, that no institution felt compelled to turn to this simplified test First of all, few © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 499 funds would meet the criterion of being sufficiently risk-averse, but even if they did, the requirement of 30% would significantly overstate the true risks of t he f und: a per centage a round 10% wou ld b e more appropr iate for an investment portfolio invested in, for example, government bonds The simplified method has, despite its importance in the development of Financial A ssessment F ramework i n p roviding a n a lternative f or s mall and risk-averse funds to the standard model, become redundant in practice The advantages of using the risk-based standard model have set the right i ncentives for pension f unds a nd outweigh t hose of t he simplistic, but conservative, simplified method 19.4.4 Minimum Capital Requirement Next to the capital requirement, pension funds must always at least have the minimum capital available This follows from the fact that the Dutch Pension Act i s ba sed upon t he European d irective on t he ac tivities a nd supervision of i nstitutions for occ upational pensions (2003/41/EC) This directive p rescribes a m inimum c apital r equirement o f a round %–5% of technical provisions In short, pension funds that run investment risk themselves n eed—at t he m inimum—to h old a s m uch c apital a s % of technical provisions and 30% of t he risk capital at death Pension f unds in which the beneficiaries bear all investment risks need to hold only 1% of technical provisions in additional assets, under the condition that asset management costs have been fi xed f or a t l east y ears The f ormer r ule would t ypically a pply t o defined benefit pens ion sch emes, t he la tter t o defined contribution schemes These basic r ules, which a re independent of t he pension f und’s asset mix, lay a floor underneath the risk-based capital requirement discussed before They not create an additional requirement Figure 19.4 shows the supervisory ladder w ith the minimum capital requirement at about 5% of technical provisions and the risk-based capital requirement For an average Dutch pension fund with 50% invested in risky assets (equities, commodities, credits, currencies) and a duration gap between liabilities and fi xed i ncome i nvestments o f 11 y ears, t he o verall c apital cha rge i s approximately 27% of tech nical provisions The required f unding r atio therefore is 127% The ladder also shows the funding level that would be necessary i f a ll f uture i ndexations were to be i ncluded i n t he tech nical provisions This level typically exceeds the capital charge but depends on the indexation target of the pension fund Pension funds are not required to calculate and communicate this real funding ratio The DNB publishes © 2010 by Taylor and Francis Group, LLC 500 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling Free surplus Level needed to fully fund indexation promises Solvency deficit - Max recovery period 15 years * For an average pension fund Funding deficit - Max recovery period year Depends on indexation target 127%* Capital requirement Minimal capital requirement 105% 100% Market value technical provision FIGURE 19.4 Supervisory ladder an estimate of t he real f unding ratio in its annual report and quarterly bulletin Some pension funds follow this practice If the minimum capital requirement or t he c apital requirement i s breached, a pens ion f und needs to fi le a recovery plan at the DNB These plans a re d iscussed i n Section 19.4.5 19.4.5 Recovery Plans As f unding tios o ften sh ow c onsiderable v olatility a p ension f und might fall short of the (minimum) capital requirement If a pension fund suspects t hat a sh ortfall may be a t hand or if t he shortfall actually has occurred, it must inform the DNB immediately of these developments If the (higher) risk-based capital charge is breached, the fund must draw up a l ong-term r ecovery p lan t hat de scribes t he r ealistic m easures t o restore t he appropriate f unding level w ithin 15 years If t he m inimum solvency requirement of 4%–5% is breached, the recovery plan must also include a de scription o f t he r ealistic m easures t o m eet t his m inimum requirement again within the next years Every recovery plan must be based on a recent continuity analysis As the recent sector-wide funding shortfalls, the relevance of this tool, which is the subject of Section 19.5, has increased In case many pension funds together suffer from a shortfall, the Pension Act allows longer recovery periods to be decided upon by the government © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 501 19.5 CONTINUITY ANALYSIS The Financial Assessment Framework aims to provide security for pension fund members, both i n t he short a nd long ter m To de tect pos sible flaws in t he f unding at a n early stage, pension f unds a re obliged to ma ke a so termed continuity analysis once every years It charts the impact of diverging financial market scenarios and of key risks such a s longevity risk over a 15 year horizon The analysis also highlights whether or not there is sufficient scope for adjustment by means of contribution-setting, indexation, or the investment policy, should this be necessary The continuity analysis thus contributes to the assessment of a sustainable financial future for pension funds and consequently to the protection of members’ interests The historical context ex plains t he objective envisaged by t he legislator at t he time the continuity analysis was introduced: “to make pension fund managers aware of possible long-term developments and to compel them to think about t heir reactions t o t hese long-term de velopments,” f rom t he S econd Memorandum of Amendment to the Pension Act, SZW (2004) This clear statement is intended to emphasize that the financial position must be sound today, as well as sustainable in the long term The continuity analysis is an innovation in supervision and will be given f urther sha pe b y pens ion f unds i n t he n ear f uture The legislator has del iberately provided for a n outline only, a nd w ill n ot be i ntroducing further legislation on this point The DNB has published a policy rule containing some principles to be followed when making a regular continuity analysis It also suggested reporting formats for its own assessment of the analyses In short, the model parameterization must be realistic and consistent Use sh ould be made o f a n i nternal m odel t hat t ruly r eflects the fund’s members’ characteristics, pension scheme, contribution policy, indexation policy, and investment policy Vendor models are possible and are often u sed, but the pension fund must be ab le to demonstrate a t rue understanding of the model and its outcomes A continuity analysis provides insight into • The possible development of the long-term financial position • The key factors for the financial position in the long term like contribution rates, indexation policy, interest rates, investment returns, etc • The financial risks and the likelihood of emergency measures (e.g., the development of the solvency ratio, shortfall probabilities, chances of reducing pension rights, or special contributions) © 2010 by Taylor and Francis Group, LLC 502 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling • The application and sufficiency of policy instruments to control longterm financial r isks (contribution policy, i ndexation policy, i nvestment policy) • Expectations on indexation (see Section 19.5.4) • Assumptions r egarding in vestment r eturns, inflation, a nd f und population In general, proper reporting combines the expected development of key parameters and possible adverse developments, for example, for the 2.5th, 25th, 75th, and 97.5th percentile of the distribution of outcomes The pension funds always bears responsibility for the proper functioning of this tool, for one, by making realistic assumptions with regard to future investment returns Only then does the continuity analysis have the potential to contribute to the sustained financial health of the pension system 19.5.1 Parameters The o utcome o f a co ntinuity a nalysis st rongly depen ds o n t he i nput Investment returns and interest rate levels are among the key parameters influencing the funding over time In order not to lose sight of reality, use of these parameters in the continuity analysis is therefore subject to statutory restrictions Legislation identifies a few broad categories for which it stipulates the maximum or minimum expected values that funds may use over the modeled prognosis horizon of 15 years in their continuity analysis These categories and their maximum expected arithmetic returns are fi xed income (4.5%), equities mature markets (9%), private equity (9.5%), equities emerging markets (10%), commodities (8%), and direct real estate (8%) The categories broadly compare to the risk categories discussed under the standardized model for the calculation of the capital requirements In addition, there is a minimum expected value of price and wage inflation of respectively 2% and 3% These returns have been at the center of much debate, illustrating their importance The dependence of the Dutch pension system on uncertain asset returns is one of its key vulnerabilities An important difference to the standardized method, and an acknowledged shortcoming i n practice, is t he absence of a sepa rate c ategory for credits P ension f unds ve i ncreased t heir i nvestments i n i nvestmentgrade credits (BBB or higher in the S&P rating classification) to generate a h igher return t han government bonds, but w ithout r unning t he s ame risks as stocks © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 503 Legislation o ffers pens ion f unds t he pos sibility t o de viate f rom t he restrictions upon prior approval by the supervisor This may be t he case if, for example, volatility, and accompanying expected returns, are higher than (implicitly) that assumed Although meant to be used sparsely and in p ractice a lmost n onexistent d ue t o a lready h igh pa rameterization, it may actually apply more broadly to, for example, the category of credits Importantly, however, with a higher expected return also comes a higher level of volatility in the stochastic continuity analysis A seco nd case for deviating a rises i f t he spec ific s ector or c ompany f und i s gove rned by markedly different levels of inflation than general price or wage inflation So far, this proves seldom the case Pension funds use many more parameters in their model for the continuity analysis than the expected investment returns and expected inflation for which the legislator has set restrictions Other parameters include standard de viations (volatilities) a nd co rrelations, t he (future) de velopment of the pension fund’s population, actuarial parameters on mortality tables, the mortality trend and longevity risk, the initial balance sheet and policy parameters regarding contributions, indexation, and investments The DNB assesses the set of parameters for veracity and consistency This however is a cha llenging task as there are no objective rules for mapping expected investment returns on estimated volatilities and correlations The pa rameters a lso p lay a n i mportant r ole f or pens ion f unds t hat establish t heir cost-effective co ntribution l evel o n t he ba sis o f ex pected returns This emphasizes their importance for real economic decisionmaking, besides t heir use in t he continuity analysis The pension f unds’ reliance in practice on high estimations of future investment returns is undoubtedly a cause for concern and puts constant pressure on the Dutch pension system The exposure to high risks may pay out in the long run, yet result in a significant downturn period in the short run 19.5.2 Risk and Outcome Distributions A co ntinuity a nalysis n eeds t o be st ochastic P erhaps t he m ost co mplicated aspect of modeling is the description of randomness Just as the future is uncertain in reality, the model’s scenarios are not fi xed in advance, but result from random developments The o utput o f a co ntinuity a nalysis t herefore i s n ot a s ingle n umber, b ut a d istribution o f outcomes Consider the following example A pension fund provides for an average pay-based retirement scheme It invests half its assets in equities and © 2010 by Taylor and Francis Group, LLC 504 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling half in bonds On average, equities yield a higher return than bonds, but are a lso acco mpanied b y h igher r isk C urrently t he pens ion f und s a funding ratio (ratio of assets to liabilities) of 125% I n t his ex ample, a prominent role is played by equity returns and interest rate movements With the outcome, it can be determined that the funding ratio is expected to rise to 129% after years and to 145% after 15 years Figure 19.5 shows the possible outcomes of the continuity analysis for the funding ratio after a nd 15 years The funding ratio is shown in the horizontal axis The vertical axis shows the relative frequency of the possible outcomes on the horizontal axis The distribution of positive and negative outcomes relative to the initial funding ratio of 125% clearly widens over time as uncertainty about the outcomes increases In quite a number of cases, the funding ratio soars to over 200% However, in other scenarios the long-term funding ratio remains insufficient to meet the liabilities as it drops far below 100% The underlying dynamics clearly show up as positive and negative risks Therefore, an expectation value only does not suffice Incidentally, in practice pension funds “cap” positive and negative outcomes by reducing or even returning contributions in very good times, and implementing statutorily required recovery plans in bad t imes This should therefore be included in modeling and influence the outcomes 35% t + (years) t + 15 (years) 30% Frequency 25% 20% 15% 10% 5% 0% 40% 60% 80% 100% 120% 140% 160% 180% 200% 220% 240% 260% Funding ratio Histogram of outcomes continuity analysis for funding ratio after and 15 years (example) FIGURE 19.5 © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 505 19.5.3 Different Functions of the Continuity Analysis The “sustainability of the financial position” ser ves as a co rner stone for several concrete functions The primary function is to assess whether the pension f und w ill be ab le t o co ntinue t o m eet i ts l iabilities a nd c apital requirements i n t he l ong r un Unfavorable sc enarios may u nfold wh ere this is no longer the case, such as disappointing investment return or sustained l ow in terest r ates The seco nd f unction i s t o de termine wh ether sufficient adjustment in the funding level can be achieved through contribution setting, indexation, or investment policies The third function is to provide the foundation of a long-term recovery plan The continuity analysis shows what measures are needed to pull a fund with a solvency deficit out of the quagmire within 15 years The fourth function also serves as a foundation namely to confirm wh ether a l ower t han cost-effective contribution is justified Analysis will have to show that, following the contribution cut, the fund still continues to meet the statutory requirements concerning technical provisions and own funds, and that the expectations raised with regard to conditional indexation can be fulfilled The fifth function i s t o provide i nsight i nto t he ex pectations a nd r isks regarding future indexation This important function is discussed in more detail in Section 19.5.4 19.5.4 Insight into Indexation Providing insight into future indexation has gained importance now that pension accrual has become much more dependent on indexation due to the shift from final to average pay schemes After all, in this type of schemes indexation is usually a d iscretionary call by the pension fund’s board of trustees A pens ion f und may have t he a mbition to provide i ndexation, but reviews from year to the next whether it will provide indexation and how much Members cannot derive prior rights from indexation and pension funds are not required to accumulate technical provisions (reserves) for indexation Moreover, if bad times necessitate adjustment, indexation may be r educed B ecause i ndexation i n a n a verage pa y sch eme la rgely determines t he level of t he pens ion benefit to b e re ached at re tirement, pension fund members are well served by realistic communication about indexation ex pectations a nd t he cha nces o f d isappointing i ndexation The P ension A ct co mpels pens ion f unds t o co mmunicate t heir i ndexation intentions annually For t his purpose, t he legislator has developed a so -called i ndexation label t hat pens ion f unds m ust y early co mmunicate to their members It shows the expected indexation over the next 15 years that members might reasonably expect, as well as the indexation in © 2010 by Taylor and Francis Group, LLC 506 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling a downturn situation (i.e., the 95th percentile of the distribution of outcomes) Here, the continuity analysis is a highly useful tool, based as it is on complete modeling of the fund Here, t oo, t he ex ample d iscussed e arlier p roves h elpful The pension fund’s ambition is to fully compensate its members for wage growth of on average 3% per year However, this indexation policy is contingent on the pension f und’s f unding r atio I f t he i nitial f unding r atio ex ceeds 125%, full indexation takes place If it is lower, indexation is reduced, to nil at a funding ratio of 105% Figure 19.6 shows the outcomes of the continuity analysis for indexation The horizontal a xis shows t he f uture i ndexation percentage C olumn “2.4%” i ndicates t he i ndexation of 4% or less (=80% of t he a mbition), but more than 2.1% (=70% of the ambition) The vertical a xis shows the relative frequency (“chances”) of the outcomes on the horizontal axis In scenarios with high investment returns, full indexation will be possible, while i n scenarios w ith d isappointing returns t here w ill be y ears of less than full indexation Column “3%” sh ows t hat ( nearly) f ull i ndexation can take place in 32% of cases The chances of lower indexation (the columns t o t he left) dec rease r apidly Yet t he cha nces of realization of, for example, only half the ambition, that is, 1.5% indexation per year, remain significant For this example, it may be calculated that the expected future 35% 30% Frequency 25% 20% 15% 10% 5% 0% 0.3% 0.6% 0.9% 1.2% 1.5% 1.8% 2.1% Indexation (yearly) 2.4% 2.7% 3.0% FIGURE 19.6 Histogram of outcomes continuity analysis for yearly indexation, ambition is 3% (example) © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 507 indexation will come to 2.1% per y ear I n l ess t han 5% of cas es, f uture indexation amounts to less than 0.6% per year 19.6 CONCLUDING OBSERVATIONS The co rnerstones o f t he D utch F inancial A ssessment F ramework a re t he mark-to-market valuation of assets and liabilities, the risk-based solvency requirements, a nd a l ong-term continuity a nalysis Together t hese building blocks offer a solid foundation for incentive-compatible regulation and supervision The valuation of liabilities follows from the replication principle: in a world without arbitrage opportunities, the mark-to-market value of a pension liability equals the market price of that investment portfolio that generates exactly the required cash flows under all future states of the world The solvency requirement is based on a VaR risk measure a 1-year horizon and a co nfidence l evel o f 97.5% This is rather low compared to Basel II and future Solvency II standards However, it can be motivated by the additional policy instruments pension funds have in the long run These instruments show up in the continuity analysis, one of the key innovations of the Financial Assessment Framework Although still in the start-up phase, over the coming years pension funds will begin making solid continuity analyses Logically, both pension fund and supervisor will experience a learning curve The quality of the analyses will improve over time That is good in itself, but the true benefit will have to come from implementing the insight gained a nd f rom t he su stainability of t he financial p osition i n t he longer term Above all, the continuity analysis is a tool allowing pension funds to gain grip on their dynamics in a situation fraught with uncertainty REFERENCES Bikker, J.A., D.W.G.A Broeders, and J de Dreu 2007 Stock market performance and pension fund investment policy: Rebalancing, free float, or market timing? DNB Working Paper no 154 Bodie, Z 1995 On t he risk of stocks in t he long run Financial Analysts Journal, 51(May–June), 18–22 Brunner, G., R Hinz, and R Rocha 2008 Risk-based supervision of pension funds: A review of international experience and preliminary assessment of the first outcomes In Frontiers in Pension Finance, eds D Broeders, S Eijffinger, and A Houben, pp 165–214 Cheltenham, U.K.: Edward Elgar Publishing Campbell, J.Y and L.M Viceira 2005 The term structure of the risk-return tradeoff Financial Analysts Journal, 61(January–February), 34–44 Cui, J , F de J ong, a nd E P onds 2009 I ntergenerational r isk sha ring wi thin funded pension schemes Working paper available at SSRN: http://ssrn.com/ abstract=989127 © 2010 by Taylor and Francis Group, LLC 508 ◾ Pension Fund Risk Management: Financial and Actuarial Modeling Lo, A.W a nd J Wang 1995 I mplementing o ption p ricing mo dels w hen ass ets returns are predictable The Journal of Finance, 50(1), 87–129 Lukassen, R and M Pröpper 2007 Equity risk at the horizon Life and Pensions, April, 43–47 Nijman, T.E 2006 A standard practice Life and Pensions, April 11, 2006, 34–40 Siegelaer, G.C.M 2005 The Dutch financial assessment framework: A step forward in solvency regulation of pension funds and insurance companies In Solvency II & Risikomanagement, eds H Perlet and H Gündl, pp 595–617 Wiesbaden, Germany: Gabler Verlag SZW, 2004 Second memorandum of amendment to the Pension Act, Ministry of Social Affairs and Employment, Dutch, the Netherlands, www.szw.nl © 2010 by Taylor and Francis Group, LLC ... Supervision of Pension Funds in the Netherlands ◾ 493 This risk has materialized all the more strongly because of the impact on the funding ratios of the steep decline in interest rates during the current... FIGURE 19. 3 © 2010 by Taylor and Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 483 the annual increase of its pension liabilities and particularly on the size of the. .. Francis Group, LLC Risk-Based Supervision of Pension Funds in the Netherlands ◾ 485 exit, marking-to-market is also possible on the basis of the replication principle Thus, in a world without