ALM practice no guarantee for successful
funding ratio management
Preliminary results from the fifty largest Dutch pension funds for the period 2001-2010.
By J.G.G.W. ten Have
Abstract
This thesis is a study based on qualitative data of the fifty largest Dutch pension funds in the period 2001–2010, testing the use of Asset Liability Management (ALM) at these funds. This paper focuses on the value created by ALM and in addition relates asset allocation and interest rate risk management to current funding ratios. Preliminary results show counterintuitive results regarding the use of ALM, indicating that ALM is not a proper tool for increasing performance at pension funds. Further investigation shows that, although applying ALM, the two largest Dutch pension funds, ABP and Zorg & Welzijn, took nevertheless too much risk in their asset allocation and were possibly speculating on their interest rate risk before the financial crisis in 2007.
MSc. BA Finance Thesis
Supervisor: Prof. dr. R.A.H. van der Meer 2nd supervisor: dr. B.A. Boonstra
2
Table of content
I. Introduction………...……….3
1.1 Problem statement………...………..3
1.2 Outline……….………4
II. Background and literature review………...………4
2.1 Basics pension funds……….………5
2.2 Factors that influence pension funds………...………7
2.3 Theoretical Underpinnings………13
2.3.1 Scale of pension funds………14
2.3.2 Asset allocation………...………15
2.3.3 Asset Liability Management for pension funds……….17
2.3.4 Investment strategies……….……….………19
2.4.5 Summary………...………20
III. Methodology and Data description……….………..…………21
3.1 Hypothesis………..………..…………21 3.2 Methodology……….………22 3.3 Data description………..………26 IV. Results……….………...………28 4.1 ALM-strategy………28 4.1.1 Value ALM-strategies……….………28
4.1.2 Effect ALM on funding ratios………..…….………30
4.2 Asset allocation ………...………30
4.2.1 Three largest pension funds ……….………31
4.2.2 Positive outliers………...………32
4.2.3 Negative outliers……….………33
4.3 Interest rate risk pension funds……….…………34
4.3.1 ALM-strategies and interest rate risk………..…………...………34
4.3.2 Interest rate risk……….…………36
4.4 Pension sponsors……….………37
4.4.1 Deposits and withdrawals……….………37
V. Conclusion and discussion……….……….………38
References……….………42
Appendix A, the fifty largest pension funds………45
3
I. Introduction
Last few years pension funds1 are part of daily news. Newspaper headlines announce; ‘pension funds have to discount their liabilities2’, ‘rising interest rates give vent to pension
funds3’, ‘funding ratios of pension funds are decreasing 4’. What are the reasons for these alarming headlines? Bad management? Is it the low interest rate, which is used to discount pension liabilities? Or did pension funds take too much risk despite their Asset Liability Management-models5?
1.1 Problem statement
Pension funds are an essential vehicle for securing future payments. Pensions affect almost the whole population in the Netherlands, what makes it subject to everybody’s concern. In order to get grip on the policies of pension funds ‘De Nederlandsche Bank’ formally ‘de Pensioen- en Verzekeringskamer’ requires pension funds since 2001 to execute an ALM-study. Management of assets involves decisions on the strategic portfolio mix as well as liabilities of pension funds with the goal to secure future and current pension payments (Haneveld, Streutker and van der Vlerk, 2005). The benefits of ALM are rather obvious. It provides an understanding of the funds overall position in terms of its obligations and gives a complete view of managing assets and liabilities (Romanyuk, 2010). In addition, ALM gives pension funds the ability to better quantify risks, prepare for future uncertainties and ideally to lead to a higher efficiency level and a better performance expressed in the funding ratio6. However, the implementation and execution of ALM has its challenges. Firstly, each fund has its own targets, risk preferences and constraints what makes it hard to develop an optimization algorithm that would fit to all pension funds. Secondly, long term strategic decisions are hard to predict and may not be readily available to pension funds. Since most pension funds are momentarily underfunded and have funding ratios that are below the norm7, it can be
1 Oxford Dictionary: A pension fund is defined as a fund from which pensions are paid, accumulated from contributions from employers, employees, or both.
2
http://www.telegraaf.nl, February 6th 2013, Kortingen pensioenfondsen onvermijdelijk. 3http://www.telegraaf.nl, January 8th 2013, Oplopende rente lucht voor pensioenfondsen. 4http://www.telegraaf.nl, February 4th 2013, Dekkingsgraad pensioenfondsen weer gedaald.
5 Plantinga (2005) defines an Asset Liability Management-model as a model of the assets and liabilities of a financial institution that facilitates decision-making with respect to asset allocation and the properties of the liabilities.
6 http://www.dnb.nl, September 20th, 2012, Increase in average funding ratio since end-June. The funding ratio is defined as the ratio between the available assets and pension obligation to pensioners, which indicates if a pension funds is able to meet its future promised payment at all times.
4 suggested that ALM is too complex and does not contributes to a better performance. For that reason, the objective of this thesis is to investigate the relation between ALM and performance at pension funds. The main research question of this master thesis is:
To what extend does Asset Liability Management creates value for pension funds?
This thesis will scrutinize the use of ALM of the fifty largest Dutch pension funds for each individual year for the period 2001–2010, where the fifty largest pension funds represent almost 90% of the total assets of pension funds8.
1.2 Outline
Before this thesis goes further into the actual research, section II gives an overview of research contribution on pension fund. This section elaborates on the principles of the Dutch pension system, the optimal scale for pension funds, how pension funds developed their asset allocation through time and current developments in regulation and funding ratios. Section III describes the methodology used and gives a description of the data. Section IV goes further into the results, which elaborates on the value of ALM-strategies, the asset allocation of pension funds and gives an overview how interest rates play an important role in pension funds behavior. Section V, which is simultaneously the last section, includes the conclusion and discussion.
II. Background and literature review
This section will discuss the relevant literature concerning pension funds and describes the basis of the Dutch pension system, discusses the proper scale for pension funds, gives an impression how pension funds put their portfolios together through the years and gives an overview of old and new regulations. Furthermore, this section elaborates on the issues funding ratios, interest rates, different sorts of risk and the asset allocation management process.
states that if the cover factor of a pension fund is not above the minimum required norm of 105%, the fund has to enforce the announced curtailments within the statutory period for recovery.
5 2.1 Basics pension funds
Pension funds are nowadays a common and indispensable vehicle for the working class and retirees in both public and private organizations. In essence, pensions ensure after reaching a certain age, disability or death a periodic payment which replaces the former salary of the employee (Drijver, 2005). What forms the basis for the existence of pension funds is explained by solidarity between different generations and participants. Some will benefit more from their pension savings than others due to the different age that people live. This benefit is translated into a higher amount of money that the longer living group will receive than they have actually saved. Among solidarity to society pension funds have an important purpose. As most pension funds have a large number of participants, individual risk can be reduced due collective risk sharing.
The Dutch pension system is based on a pay-as-you-go financed state pension, also known as AOW, which provides an elementary uniform pension to older residents at a level that is linked to the minimum wage (Van Rooij, Siegmann & Vlaar, 2008). On top of the AOW more than 90% of the employees saves compulsory for their retirement through their employer, which is defined as Defined Benefit (DB). Employee solidarity has a central position and is given shape in collective schemes, applied by industry wide and company pension funds (Van der Lecq and Steenbeek, 2007). All other private pension arrangements are defined as Defined Contribution (DC), which are used to modify the pension portfolio with personal preferences and is particularly vital for individuals that are self-employed and who lack work-related pension provisions (Van Rooij, Siegmann & Vlaar, 2008). According to Tower Watson9 the Dutch pension system was characterized as DB, but is shifting to DC. This shift requires a growing awareness for current DC arrangements, which will not fit in the current system. Bovenberg and Nijman (2009) argue in their paper that stand-alone collective pension schemes, in which participants share risk among themselves, are an attractive alternative between corporate DB schemes and individual DC pension plans. Traditional occupational DB schemes are not maintainable because neither the benefactors nor the young and future members can provide expensive guarantees to the retirees. They conclude that in the coming years further reforming will be needed to adapt Dutch collective pension schemes for an aging population and a changing transitional labor market with different needs.
The Dutch pension system consists out of 454 pension funds on December 31th, 201110. These funds represent around 807 billion Euro of assets, which is equal to 133% of
6 GDP, ranking the Netherlands among the highest of the world11. These 454 pension funds can be divided into four main categories. The first group consists out of all funds that are related to a single company (i.e. Shell, ING, Philips and Heineken). This group of employees work all in the same company and participation in the pension program is primarily mandatory. The second group consists out of all funds that are related to companies in the same industry (i.e. ABP, Zorg en Welzijn and PMT). All participants in this group are employed in a company in the same industry and are mandatory to participate in the pension program. The third group of pension funds consists out of participants who work individually and have the same occupation (i.e. medical specialists, general practitioners and physiotherapists). The participants in this group do not have an employee - employer relation and do not always have to participate mandatory in pension programs. The fourth group consists out of a few saving funds for companies and one pension fund, the notarial pension fund, that falls under a special law. Table 1 gives an overview of the number of pension funds end 2001 and end 2011 for all of the four categories as described above. Figure 1 shows that the total number of pension funds is decreasing steadily through the years due to mergers.
Table 1: Number of pension funds on December 31th 2001 and December 31th 2011 in the Netherlands.
2001 2011 Difference in
percentage
Pension funds related to a single company 843 359 -57%
Pension funds related to the same industry 100 77 -23%
Pension funds related to individuals with the same occupation 11 12 9%
Other pension funds 9 6 -33%
Total 963 454 -53%
Source: DNB
Figure 1: Total number of pension funds in the Netherlands for the period 1997–2012.
Source: DNB
11 http://www.towerswatson.com, January 2012, Global Pension Assets Study 2012. 0 200 400 600 800 1000 1200
7 2.2 Factors that influence pension funds
Research has highlighted the most important factors that influence the behavior and strategic planning of pension funds. The following section will discuss the subsequent factors in detail; an aging population, rules & codes, the funding ratio, interest rates and different sorts of risk (figure 2).
Figure 2: Factors that influence the behavior and strategic planning of pension funds.
The first factor that influences the behavior and strategic planning of pension funds is aging. In the Netherlands a rapid aging of the Dutch population is concerning pension (Whiteford and Whitehouse, 2006). The Dutch Central Bureau of Statistics (CBS)12 explains the aging of the Dutch population by a lower birth rate, a lower mortality rate and the baby boom after World War II. In the next 30 years the amount of people above the 65 years old will double. Due to aging of the population, costs of the AOW13, are expected to increase from 4.7% of GDP in 2006 to 8.8% of GDP in 2040 (Ewijk, van Draper, ter Rele, Westerhout and Donders, 2006).
The second factor that influences the behavior and strategic planning of pension funds are rules and codes. Ponds and van Riel (2007) argue that there was barely any form of supervision for pension funds after World War II. In the seventies, many pension funds were underfunded and solvency requirements did not have any effect on pension funds. As of January 2007 pension funds need to adapt to the ‘Financieel Toetsingskader’ (FTK, Financial
12 http://www.cbs.nl, December 17th, 2010, Key figures of the population forecasts 2010-2060. 13 Van Rooij, Siegmann & Vlaar (2008): AOW is the first layer of the Dutch pension system.
8 Assessment Framework), which is a framework of supervision rules introduced by the formerly Pensioen- en Verzekeringskamer. The framework was introduced to improve supervision of pension funds into its financial positions and future developments on the short and medium long term. The main argument for these rules are to make the pension funds more transparent. The FTK is developed to be in line with international market valuation of assets and liabilities, and was a precursor of the European risk based rules.
As part of this regime a new valuation foundation was introduced, where investments are valued at market level. Van Rooij, Siegmann & Vlaar (2008) argue that the impact of valuation on the basis of market prices is relatively small if fixed discount rates are used for conditional rights14. They conclude that if market valuation is used for both unconditional15 and conditional rights, the volatility of the contribution increases significantly. As a solution to this problem they suggest to increase the duration of assets, which will reduce the mismatch between assets and liabilities. However, it is not clear if this option is available for pension funds due to the limited supply of long-term bonds. Vlaar (2005) explains that when extending the duration of bonds, the duration gap is increased in a fixed rate environment and decreases it under the market.
He scrutinizes to what extent optimal investment policy by Dutch pension funds is affected by modifications in regulation. He concludes that a complete market valuation method results in higher costs for DB pensions relative to a fixed discount rate method. The only two exceptions are funds with durations of bonds with more than 22 years and funds that systematically overestimate future returns while at the same time have no incentive to shorten the duration of their bonds. Vlaar (2005) explains his finding by the mean reversion in interest rates. Premiums are highest under market valuation when projected returns are lowest. Under the FTK the optimal duration of bonds in portfolio is much higher than before the introduction of the regulations. The gain of extending the duration is maximized under market valuation, where costs are also low for high durations under actuarial accounting.
An important measure as part of the FTK is that future cash flows are discounted against current forward rate structures16. The liabilities always have to be covered by
14 Bikker and Vlaar (2007): Because indexation of pension benefits is only awarded if the financial position of the fund allows it, the Dutch system is hybrid rather than pure DB. It combines a DB nominal pension with Targeted Benefit indexation. Although in this system the employees bear some risk, the term conditional DB is more apposite than DC, since contributions are not fixed whereas (maximum) benefits are clearly defined. 15 Bikker and Vlaar (2007): Indexation is unconditional, the solvency tests for this fund are applied to the real funding ratio, based on a discount rate equal to the actual real market interest rate.
9 investments, taking into account the minimum amount of required equity17. Pension funds are required to have enough equity, where in 97.5%18 of the cases it is prevented that within a period of one year the fund has a shortage of technical resources. Another measure of the FTK is that a fund that is not solvable and has a reserve deficit, has to present a long-term recovery program.
The third factor that influences the behavior and strategic planning of pension funds is the funding ratio, which is the ratio between the available assets and pension obligation to pensioners19. Table 2 shows the amount of pension funds per quarter that had in 2007 a funding ratio under the 105% norm20, have a funding ratio between the 105-130% or have a funding ratio above the 130%.
Table 2: Average estimated funding ratio for each quarter of all Dutch pension funds. Average estimated funding ratio (%) Number of pension funds with a funding ratio < 105% Number of pension funds with a funding ratio 105-130% Number of pension funds with a funding ratio > 130% 2007K1 141% - 195 252 2007K2 152% - 110 337 2007K3 150% 1 132 313 2007K4 144% 2 151 283 2008K1 132% 7 256 166 2008K2 136% 10 232 175 2008K3 121% 53 292 73 2008K4 95% 292 92 25 2009K1 92% 314 65 20 2009K2 102% 213 152 27 2009K3 109% 108 243 37 2009K4 109% 106 233 35 2010K1 108% 123 211 28 2010K2 100% 233 111 17 2010K3 99% 246 95 17 2010K4 107% 171 157 19 2011K1 112% 104 207 28 2011K2 111% 102 211 24 2011K3 94% 276 44 11 2011K4 98% 227 84 13 Source: DNB
The table shows an increasing trend in pension funds that have a funding ratio below the required 105% norm until the second quarter of 2009. This same trend is noticeable starting in
apply to different maturities. This reflects differences in the yield demanded on investments with different horizons.
17 http://www.dnb.nl, January 1 2007. Financieel Toetsingskader (FTK) voor pensioenfondsen.
18 http://www.pensioenfederatie.nl, May 30th 2012, Uitwerking van het nieuwe Financieel toetsingskader (FTK). The 97.5% is distracted from the objective to realise an actual pension. Absolute guarantees concerning pension liabilities are not realistic, therefore the government decided to require a minimum .
19 http://www.dnb.nl, September 20th 2012, Increase in average funding ratio since end-June.
10 the first quarter of 2010. Of all pension funds, 227 pension funds have a funding ratio below the on December 31th 2011. De Nederlandsche Bank requires pension funds to have a funding ratio of at least 105%, which means that pension funds should be able to meet their payment promises at all times. When the funding ratio of a pension fund falls under the 105% norm the fund has the obligation to recover to the required percentage within the recover period21. There are two options for pension funds to increase their funding ratio. Firstly, discount pension payments. Secondly, let future generations pay a higher pension fee.
The fourth factor that influences the behavior and strategic planning of pension funds is the interest rate. An important reason why a significant amount of the pension funds have a funding ratio below the required norm is the low interest rate22. The interest rate is used to discount pension liabilities. Pension funds need to build up larger stakes of assets to be able to fulfill future payments, resulting in critical situations. Table 3 shows that every quarter the quartile interest rate on the basis of the 10 years Dutch Government bond rate, which is knows as the risk-free rate, drops steadily over the years projected. To mitigate the pressure for pension funds to build up larger stakes of assets De Nederlandsche Bank has introduced several measures. Early 2012, DNB announced23 that for ultimo 2011 the average interest rate of the last three months can be used for discounting liabilities instead of the normally used monthly interest rate, which resulted in an average rise of the funding ratio of 4%. More recently DNB introduced a package of pension measurements, with as main objective to spread negative effects of the package over active participants and pensioners24. As part of the announced package, premiums do no longer need to be part of financial recovery when a pension fund has a cover shortage starting in 2014.
Another important measure is the Ultimate Forward Rate (UFR), which attempts to limit unrealistic attrition of the interest rate curve on the long term25. The UFR is effective since September 30th 2012 and has its immediate effect on the funding ratio of pension funds. The UFR is a notorious measure that is point of discussion. For ‘zero’ interest rates with a maturity longer than twenty years the underlying 1-years forward interest rates are
21 http://www.dnb.nl, January 6th 2012, Measures taken by DNB to reduce uncertainty about pension funds. 22
http://www.pensioenfederatie.nl, September 24th 2012, Overwegend evenwichtig pensioenpakket met gemiste kansen.
23 http://www.dnb.nl, January 6th 2012, Measures taken by DNB to reduce uncertainty about pension funds. 24 http://www.pensioenfederatie.nl, September 24th 2012, Overwegend evenwichtig pensioenpakket met gemiste kansen.
11 extrapolated to the UFR26, which is fixed at 4.2%. The use of a rate that is calculated on the basis of a longer period, 20-60 years, is distressing27. The last sixty-year are characterized as a period of economic growth and it is unknown if this period is representative for the future.
Table 3: Quarterly interest rates on the basis of the 10 Years Dutch Government bond rate expressed in percentages.
Year/ Quarter Average interest rate Year/ Quarter Average interest rate
2000K1 5.60 2006K3 3.89 2000K2 5.40 2006K4 3.79 2000K3 5.39 2007K1 4.03 2000K4 5.23 2007K2 4.38 2001K1 4.89 2007K3 4.43 2001K2 5.14 2007K4 4.31 2001K3 5.04 2008K1 4.05 2001K4 4.75 2008K2 4.43 2002K1 5.08 2008K3 4.48 2002K2 5.24 2008K4 3.95 2002K3 4.74 2009K1 3.74 2002K4 4.50 2009K2 3.86 2003K1 4.06 2009K3 3.65 2003K2 3.95 2009K4 3.50 2003K3 4.14 2010K1 3.40 2003K4 4.33 2010K2 3.07 2004K1 4.07 2010K3 2.65 2004K2 4.31 2010K4 2.85 2004K3 4.18 2011K1 3.35 2004K4 3.81 2011K2 3.44 2005K1 3.60 2011K3 2.72 2005K2 3.30 2011K4 2.43 2005K3 3.22 2012K1 2.23 2005K4 3.37 2012K2 2.05 2006K1 3.49 2000-2010 4.16 2006K2 3.95 2000-2012K2 3.99 Source: DNB
The fifth factor that influences the behavior and strategic planning of pension funds are several sorts of risk. As part of the FTK there are six groups of risk subdivided (Table 4), namely: interest rate risk (S1), business value risk (S2), currency risk (S3), commodity risk (S4), credit risk (S5) and insurance risk (S6)28.
26 http://www.dnb.nl, October 3th 2012, UFR method for calculating the term structure of interest. The extrapolation of this forward rate consists of the weighted average of the perceived forward rate on the basis of the three months average swap curve and the UFR. For interest term structures until a maximum of sixty years the zero interest rates are calculated on the basis of swaps, where the fixed interest rate is swapped against the 6-months EURIBOR. The adapted 1-year forward interest rates starting with a maturity of sixty years are constant and are equal to the UFR.
12
Table 4: Description sorts of risks relative for pension funds.
Sorts of risk Description
Interest rate risk (S1) Mismatch of assets and liabilities on the balance sheet Business value risk (S2) Sensitivity for fluctuations on exchange markets
Currency risk (S3) Value of investment changes due to a change in the exchange rate Commodity risk (S4) The value of the investment declines in time
Credit risk (S5) Change of the average credit spread, resulting in a revaluation of investments Insurance risk (S6) Covering process risk, Negative Stochastic Deviation and mortality trend
uncertainty to a certain degree
Interest rate risk or S1 can be seen as one of the most important risks for pension funds since most pension funds have a longer obligation on their asset side of the balance sheet compared to their liability side29. Due to this mismatch on the balance sheet a pension fund experiences an interest rate risk. When the interest rates are going down, the value of liabilities grows faster than the value of assets. This is firstly explained by the larger impact of interest rate changes on the longer term. Secondly, pension funds experience interest rate risk when there is not a perfect correlation between the short- and long-term interest rates. This is explained by the change of the value of the liabilities as a consequence of an interest rate change, which is not equal to the change of value of the assets. Thirdly, the volatility of the short- and long-term interest rate is not equal. Even when there would be a perfect correlation, a change in the volatility would result in a different change of the value of assets compared to the liabilities.
Business value risk is denoted as S2, which represents the value of stocks and real estate30. This risk influences investments that are sensitive to fluctuations of the stock exchange list. Not only stocks and indirect real estate are part of these investments, also direct real estate, investments in not exchange listed companies, convertibles, equity notes, total return swaps and some derivatives are included.
Currency risk (S3) is defined as the risk that the value of an investment in a foreign currency aggravates in reaction to a change in foreign currency rates31. The possibility that the exchange rate of a foreign currency is getting worse compared to the euro means a risk for the pension fund, since liabilities are in Euros and assets are partly in foreign currencies. This risk influences the direct exposure of the pension fund in a currency as well as in investments like stocks that are valued in a different currency than the Euro.
Pension funds that invest in commodities (S4) carry the risk that the value of
13 investments changes32. Commodities are an investment category where pension funds mostly invest in index-futures and options. Examples of commodities are: energy, industrial metals, expensive metals, agriculture and livestock. A pension fund faces commodity risk when investments are sensitive for changes in the commodity prices. More specifically this applies to all long and short positions in commodities, indices of commodities and derivatives in commodities.
Credit risk, denoted as S5, is the risk that a pension fund experiences when the average credit spread changes33.The consequence of a changing average credit spread is that the value of for instance corporate bonds changes too. Funds carry the risk that the weighted average credit spread of the credit portfolio increases. A reason for an increase is related to a negative development of economic growth, which results in a devaluation of investments. Credit risk is reflected in the interest rate margin of credits or credit spread.
Insurance risk (S6) is not regulated by DNB34. Pension funds are free to define the size and the standard model themselves. S6 consists out of process risk, Negative Stochastic Deviation and Mortality Trend Uncertainty.
When the risk elements S1 to S6 are calculated, the total risk is calculated using the following square root formula:
√ (I) where St is the total risk and S1 to S6 the different forms of risk (Westrik, 2012). The
correlation between the interest rate risk and equity risk is 0.5, with = 0.5. This correlation represents the probability that these risks occur simultaneously.
2.3 Theoretical Underpinnings
The previous paragraph gave an overview of all important topics and characteristics of pensions and pension funds. The literature shows that more pension funds are merging every year, this section looks into the advantages and disadvantages for pension funds getting more substantial. The literature shows that half of the pension funds had a funding ratio below the 105% norm on December 31th 2011, the next section looks into the asset allocation of pension funds searching for literature giving other explanations than the low interest rate.
14 2.3.1 Scale of pension funds
The advantage of collectivity or risk solidarity is by very nature not open to the question (Steenbeek and van der Lecq, 2007). However, the desirable scale of such collectivity arrangements is subject to discussion. It is only possible to guarantee solidarity through agreements that clarifies rights and obligations. Organization of such guarantees in the form of pension funds comes with administrative- and investments costs (Bateman and Mitchell, 2004). Managing costs are important because they reduce the rate of return on investments and consequently raise the cost for retirees. Bateman, Kingston and Piggot (2001) report in their paper that an increase in annual operating costs with 1% of pension fund assets indicates a cumulated reduction of 27% of eventual pension benefits. Between 1998 and 2004 around 300 Dutch pension funds (25%) merged or was sold due to the relative high costs of pension plans (Bikker and de Drue, 2006). However, nowadays many pension funds still operate inefficient. Pension funds can become more efficient by employing economies of scales to decrease administrative costs. Bikker and de Drue (2006) argue that the administration of DC plans are less costly compared to DB plans, which is firstly explained by the limited availability of options and secondly by the collective nature of Dutch pension funds.
In the Netherlands, pension funds do not incur extraordinary marketing costs and costs of training in risk awareness (Bikker and de Drue, 2006). Furthermore, they conclude that outsourcing of the administration only raises costs, which is clarified to underreporting of costs of pension funds. Small funds outsource most frequently. However, outsourcing of liabilities and investments risk appears to reduce the administration costs. When looking at the administrative costs for pension funds with a higher proportion of clients, it seems that more investments result in higher administration costs. Bikker and de Drue (2006) conclude that economies of scale dominate the strong diffusion in administrative as well in investments costs of pension funds.
Bikker, Broeders and de Dreu (2007) argue that in the short term, outperformance of equities over bonds and other investments automatically results in a higher actual equity allocation, since pension funds do not continuously rebalance their investments. They find that in the medium term, outperformance of equities leads to an increase of strategic equity allocation. These findings suggest that investment strategies of pension funds are partly driven by the cyclical performance of the stock market.
15 where smaller pension funds are more prone to adjust for pension contributions. An explanation that Bikker, Broeders and de Dreu (2007) give for these findings is that large pension funds tend to be less risk averse compared to small funds and managers enjoy more freedom concerning implementation of market timing strategies. De Dreu and Bikker (2009) suggest that a further consolidation of the Dutch pension sector could contribute to improved sophistication of the investment policies of pension funds. Smaller funds invest particularly in less risky sophisticated and less complex other assets classes, while large funds do the opposite and rely on their asset allocation strategies. An explanation for the lower risk levels at small pension funds could be dedicated to the limited availability of sophistication with respect to the believed optimal asset allocation.
Biker and de Drue (2006) and James, Smalhout and Vittas (2005) come up with two possible causes for economies of scale in the investment of pension funds assets. Firstly, they suggest that a larger fund is able to spread its fixed costs over more assets, which are likely to increase less than variable costs. Secondly, larger pension funds have more bargain power which results in more competitive deals. Davis, Grob and de Haan (2007) support the findings that small pension funds have higher costs. In addition, Bikker and de Drue (2006) find that industry-wide pension funds are considerably more efficient compared to the company funds and other funds. The element that these pension funds need to transfer pension rights less frequently and that they use relative straightforward pension schemes under the corresponding collective labor agreement explains this. Bikker and de Drue (2006) conclude that the compulsory funds are most efficient within the industry pension funds because they do not need to compete for clients.
2.3.2 Asset allocation
16
Table 5: Asset allocation in percentages for all Dutch pension funds for the years 1983, 1993 and 2000.
Year Stocks Bonds Real estate Cash Other Total
1983 16 71 13 0 0 100
1993 29 58 13 0 0 100
2000 41 34 11 1 1335 100
Source: Haneveld, 1998
Table 6 provides an overview of the average asset allocation of all Dutch pension fund over a more recent period. As can be seen pension funds started to trade bonds for higher risk investments like stocks, which is in line with the data of Haneveld (1998).
Table 6: Asset allocation in percentages for all Dutch pension funds period 2001-2010.
Year Stocks Fixed-interest
securities Real estate Other
2001 42.55 36.44 10.94 10.08 2002 36.04 42.59 11.24 10.13 2003 40.78 40.51 10.34 8.38 2004 40.97 41.62 10.01 7.40 2005 42.21 40.86 9.87 7.06 2006 42.52 40.35 10.81 6.32 2007 40.66 43.02 10.44 5.88 2008 30.78 52.15 10.94 6.13 2009 36.28 48.99 9.83 4.89 2010 35.85 45.29 10.31 8.55 Source: DNB
The change in the asset allocation of pension funds generated a higher return, but increased the risk of pension funds. Table 7 gives an overview of the average performance of all Dutch pension funds per quarter starting in 2007. A separate overview of the fifty largest pension funds is not available at DNB, but when taking in consideration that 90% of the assets is owned by this group it gives a representative indication. Stocks show a volatile performance, with a maximum average decrease of 22.67% in the fourth quarter of 2008 and a maximum average increase of stocks prices of 16.33% in the second quarter of 2009.
17
Table 7: Performance per asset class per quarter in percentages for all Dutch pension funds.
Year/ Quarter Results real estate (%) Results stocks (%) Results fixed-interest securities (%) Results other (%) Total results (%) 2007Q1 3.96 2.31 -0.06 1.25 1.55 2007Q2 -1.57 6.85 -1.97 0.88 1.64 2007Q3 0.73 -1.08 1.29 12.48 1.34 2007Q4 -3.02 -3.48 0.47 7.77 -0.70 2008Q1 -1.10 -14.43 0.06 16.73 -4.09 2008Q2 -2.66 -0.93 -1.64 6.14 -0.72 2008Q3 1.27 -10.70 1.97 -15.48 -4.59 2008Q4 -13.32 -22.67 2.70 -13.49 -8.60 2009Q1 -5.85 -6.73 0.41 -10.62 -3.74 2009Q2 3.07 16.33 3.30 -0.05 6.98 2009Q3 5.51 14.22 5.79 12.04 9.19 2009Q4 0.44 6.42 1.56 3.77 3.04 2010Q1 3.40 6.96 4.81 1.50 5.09 2010Q2 -0.13 -4.23 4.69 -3.38 0.30 2010Q3 3.98 4.46 4.45 5.89 6.72 2010Q4 3.86 9.73 -5.76 4.97 -0.43 2011Q1 1.20 0.03 -2.48 3.02 -0.44 2011Q2 1.59 -0.75 1.62 -3.00 0.76 2011Q3 -3.56 -11.75 12.08 -4.32 2.11 2011Q4 3.66 8.16 2.89 7.20 4.90 Source: DNB
2.3.3 Asset Liability Management for pension funds
An ALM-model is a model that financial institutions, like pension funds, currently use in their decision-making process with respect to the allocation of asset and the properties of liabilities (Plantinga, 2005). The board of a pension fund has the obligation to find acceptable policies that guarantee solvency, which means that the fund is able to fulfill all promised payments in the long run (Drijver, 2005). The solvency measured at a certain moment in time is called the funding ratio, as described before. The funding ratio changes over time, mainly due to fluctuation in assets and liabilities. In order to prevent that the funding ratio falls below 1 and pension funds are underfunded, a pension fund needs to actively manage its asset portfolio.
18 decisions to abandon return maximization if a goal strikes the employee retirement income security act. In practice, surplus optimization or portfolio perseverance require the acceptance that accounting classification of liabilities are a correct reflection of real corporate liability.
Van der Meer, Plantinga, Salomons & Boonstra (2011) describe how ALM has developed within a financial institution, which will helps to understand the basics of this strategy. They distinguish four different phases in the development of ALM. The first phase is characterized as there is no formal coordination between decisions concerning assets and liabilities. The portfolio mainly consists out of bonds that are kept until they are fully matured. Another implicit assumption is a constant and inactive view on the interest rate and the composition of the portfolio. In this phase, investors tend to ignore coordination of decisions with respect to the asset and liability portfolios. For pension funds the link between assets and liabilities is made implicitly by means of actuarial assumptions. In particular, the actuarial discount rate has a dominant role. Pension funds pick an actuarial discount rate that is low enough to ensure that the portfolio of assets is able to exceed it.
In the second phase, the focus lays on the development of the asset portfolio independent of the liabilities. This phase in the development of an ALM format is also known as the phase where financial institutions become aware that taking more risk can add value to the asset portfolio. Portfolio theory offers investors the ability to actively manage the investment portfolio in minimizing risk and optimizing return and to choose for portfolios that minimizes risk. In this phase liabilities are still ignored in the composition and management of the portfolio.
The third phase is led by investment decisions that aim to control risks that have originated in the liability portfolio, most institutional investors are in this phase. The third phase of ALM is characterized by matching assets with liabilities. Both risk and return on assets and liabilities are considered in the asset portfolio, which include the correlation between assets and liabilities. The so-called matching approach is the most common used approach in this phase of ALM. This approach matches assets and liabilities by creating an assets portfolio with similar cash flow characteristics as liabilities. Another method is the surplus optimization approach, where the expected utility of surplus is maximized.
19 A dynamic model is assumed being a model in which an approach can be assessed regarding its anticipated decisions. Reactions on future developments are taken into account when the result of a study using a dynamic model concerns a decision rule.
2.3.4 Investment strategies
Since 2007 the ALM-strategy is point of discussion. Goorbergh, Molenaar, Steenbeek and Vlaar (2011) find that the current model does not take sufficient extreme economic changes into account. They developed a risk model to better address events like the current financial crisis. Van der Lans (2011) included the regular ALM-model ‘Crisis for Assets Returns’, in his research. He finds evidence that current ALM-models do not give a realistic risk indication of investments.
Bikker, Broeders and de Dreu (2007) debate in their study the optimal asset allocation of pension funds. When a fund is choosing portfolio weights that are different from the strategic asset allocation, it is called tactical asset allocation or market timing. Market timing refers to taking short-term risk, by buying or selling the underlying securities or derivatives. However, the potential availability for making an extra return through market timing is limited, which is indicated as fundamental law of active management36. In 90% of the cases variability in pension funds returns is explained by strategic asset allocation and less than 5% by market timing. Blake, Bruce and Timmermann (1999) find evidence that market timing causes an average loss between the 20 and 66 basis point per year, suggesting that pension funds are unsuccessful in exploiting market timing.
Bikker, Broeders and de Dreu (2007) find that investments strategies of pension funds are partly driven by cyclical performance of stock markets. Pension funds tend to base investment choices on recent market developments, rather than on long-term trends. What specifies this finding is that equity reallocation of pension funds is higher after underperformance of equity investments rather than on outperformance. They report that in 16% of the cases rebalancing took place after positive equity returns compared to 48% after negative equity returns, which is no anti-cyclical behavior and is not part of ALM. In addition to ALM-research, De Dreu and Bikker (2009) conclude that asset allocation does not follow directly from optimization of ALM-strategies but is determined by human judgment.
20 2.3.5 Summary
The advantage of collectivity or risk sharing is not questionable, however the desirable scale of such collectivity arrangements is subject to discussion. Pension funds have several important costs; administrative-, investment-, managing-, operating- and marketing-costs. Pension funds could especially benefit more from economies of scales for the administrative costs. An 1% increase in annual operating costs indicates a cumulated reduction of 27% of eventual pension benefits, suggesting that mergers will pay off.
Previous research suggests that larger funds tend to adapt their equity allocation on the basis of actual equity returns relatively more than smaller funds. This is explained by a less risk averse investment policy and more freedom concerning market timing strategies for managers, which is important to monitor in the future. Conclusively, the literature suggests further consolidation of the Dutch pension sector could contribute to improved sophistication of the investment policies of pension funds. In addition, the literature describes that compulsory industry wide pension funds are most efficient, which is explained by the element that these funds need to transfer pension rights less frequently and use relative straightforward pension schemes under the corresponding collective labor agreement. The compulsory funds are most efficient within the industry pension funds because they do not need to compete for clients.
In the last 60 years pension funds started to use more active strategies like ALM for managing risks and to generate higher returns. Pension funds traded their more conservative assets like bonds for higher risk assets like stocks and started to invest in derivatives like option- and future contracts. The statistics show that the return on stocks was very volatile, with a maximum average decrease of 22.67% and a maximum average increase of 16.33% in only one quarter indicating the high risk profile of stocks.
21 through tactical asset allocation is limited due to the simple explanation that an asset manager is not able to forecast future market directions always correctly.
Pension funds are tend to base investment choices on recent market developments, rather than on the long-term. What specifies this finding is that equity reallocation is higher after underperformance of equity investments rather than on outperformance, which is no anti-cyclical behavior and is not part of ALM. In addition, ALM-research finds that asset allocation does not follow directly from optimization of ALM-strategies but is determined by human judgment.
III. Methodology and Data description
This section gives firstly an overview of the formulated hypothesis. Secondly, the methodology describes the statistical relationships that are going to be tested using the dataset. Thirdly, the section data description gives an overview of the data collected and gives a summary of the descriptive statistics.
3.1 Hypothesis
Part I of this master thesis has given an extensive overview of pension funds and its characteristics. In the last few decades pension funds started to use more active strategies like ALM to manage risks and to generate higher returns. Currently, ALM is the most common used strategy by pension funds. However, several papers find that current ALM-models have their shortcomings and do not take sufficient extreme economic changes into account (Goorbergh, Molenaar, Steenbeek and Vlaar, 2011), (Van der Lans, 2011). In addition, the literature describes that pension funds show no anti-cyclical investment behavior, which is not part of ALM. How is this possible, is ALM the proper tool for managing assets and liabilities at pension funds. Therefore, the following hypothesis tests if Dutch pension funds that use ALM have a better funding ratio than Dutch pension funds that do not use ALM, the following hypothesis is formulated;
22 Last decade the amount of pension funds in the Netherlands decreased by almost half, where pension funds are prone to benefit from collective risk sharing by merging. The literature suggests a further consolidation of the Dutch pension sector, however the desirable scale of such consolidations has its influence on the behavior of pension funds. A benefit of larger pension funds compared to smaller pension funds is a more active position on financial markets to manage financial positions (Kakes, 2006). The smaller pension funds have a limited availability of sophistication with respect to the optimal asset allocation (Dreu and Bikker, 2009). As part of this change in scale, pension funds traded their more conservative assets like bonds for higher risk assets like stocks and started to invest in derivatives like option- and future contracts (Drijver, 2005).
But how did pension funds manage their interest rate risk. One of the most important risks which pension funds face is interest rate risk. This interest rate risk is linked to the funding ratio, since the interest rate is used to discount pension liabilities37. Interest rates are momentarily historically low, strengthen the effect on a negative development of funding ratios. After the start of the financial crisis in 2007 the average funding ratio of Dutch pension funds decreased significantly. Half of the pension funds had a funding ratio below the 105% norm on December 31th 201138. How could this happen, since ALM-studies should optimize the asset allocation and the amount of risks hedged. Therefore, the following hypothesis tests the relation between the use of ALM in terms of covering interest rate risk and the funding ratio;
H2: The outcome of ALM-studies covering 50% or more interest rate risk has a positive influence on the funding ratio.
3.2 Methodology
Research is defined as the systematic investigation into and study of materials and sources in order to establish facts and reach new conclusions39. The most common known types of classifying and characterizing research are quantitative research and qualitative research. Quantitative research methods were originally developed in the natural sciences to study natural phenomena (Myers. 2009). Quantitative research emphasizes numbers that represent values and levels of theoretical constructs and concepts and interpretation of the numbers is
37 http://www.dnb.nl, January 1 2007, Standard interest rate risk model.
23 viewed as strong scientific evidence of how a phenomenon works. Qualitative research methods were developed in the social sciences to enable research to study social and cultural phenomena. Qualitative data sources include observation and participant observation, interviews and questionnaires, documents and texts, the researcher’s impressions and reactions. Since the data is collected from annual reports and form the basis for this thesis, the data is qualified as qualitative data.
To test these hypotheses a self-generated qualitative dataset of the fifty largest pension funds is used, which is based on the annual reports of each fund. To measure the actual performance of pension funds, the funding ratio of each individual pension fund for each individual year is used. This approach seems to be the most obvious, since the funding ratio reflects total performance best in terms of asset allocation. In order to generate a proper benchmark for the funding ratio of each individual fund, the average funding ratio of the fifty largest pension funds is calculated. The equation (II) calculates the average funding ratio of the number of pension funds ̅ per year using the following formula:
̅ ∑ (II)
where ∑ indicates the sum of the funding ratios for that year and n the number of pension funds. The average deviation (III) between the fifty largest pension funds and the pension funds that used ALM per year is denoted as:
̅ ̅ (III)
where ̅ is the average funding ratio of the fifty largest pension funds and ̅ is the average funding ratio of all pension funds that did use ALM in that year. The formula of the average deviation (IV) between the fifty largest pension funds and the pension funds that did not use ALM per year is denoted as follows:
24 where ̅ is the average funding ratio of the fifty largest pension funds and ̅ is the average funding ratio of all pension funds that did not use ALM in that year. For calculating the average deviation (V) of each individual fund the following equation is used:
∑ ∑
(V)
where ∑ is the average funding ratio of the fifty largest pension funds for the period 2001 until 2010 and ∑ is the average funding ratio of the individual fund for the period 2001 until 2010.
In order to check if the relation between the use of ALM and size at nominal level is statistical significant, table 8 is used where the frequencies of the two variables are compared with each other. The year 2004 is important, since this year shows the most observations that pension funds did explicitly not use ALM. For analyzing the data, the chi-square test is used which is the most elementary method. As denoted in table 9 the chi-square test shows in 2004 a c2 -value of 0.97557, which indicates that 0.32330 or 32.33% of the cases is determined by coincidence. This percentage is relatively high and shows that there is a chance of 67.67%, which is not statistical significant, that there is an association between the size of pension funds and the use of ALM in 2004.
Table 8: observed frequencies ALM use in 2004 (Oij).
Use of ALM No use of ALM Total
25 largest pension funds 11 2 13
26 – 50 largest pension funds 6 3 9
Total 17 5 22
Table 9: statistical overview ALM use in 2004 (C2).
Row Column Oij Eij Oij-Eij (Oij-Eij)2 (Oij-Eij)2/Eij C2 1 1 11 10.04545 0.95455 0.91116 0.09070 1 2 2 2.95455 -0.95455 0.91116 0.30839 2 1 6 6.95455 -0.95455 0.91116 0.13102 2 2 3 2.04545 0.95455 0.91116 0.44545 Sum 22 22 0 0.97557 0.32330
25 financial crisis started. Again, the chi-square test is conducted which resulted in a c2 value of 3.10454, indicating that 0.07807 or 7.807% of the cases is determined by coincidence (table 11). This percentage is statistical significant at 90% indicating that there is a positive relation between the size of pension funds and the use of ALM in 2007. With a reliability of 90% significance the critical value of chi-quadrat, with one degree of freedom, is 2.70554. This is larger than 3.10454 indicating that the relation between the size of a pension funds and the use of ALM is reliable at a 99% significance level.
Table 10: observed frequencies ALM use in 2007 (Oij).
Use of ALM No use of ALM Total
25 largest pension funds 23 1 24
26 – 50 largest pension funds 24 1 25
47 2 49
Table 11: statistical overview ALM use in 2007 (C2).
Row Column Oij Eij Oij-Eij (Oij-Eij)2 (Oij-Eij)2/Eij C2 1 1 23 23.02041 -0.02041 0.00042 0.00002 1 2 1 0.97959 0.02041 0.00042 0.00043 2 1 24 23.97959 0.02041 0.00042 0.00002 2 2 1 0.20408 0.79592 0.63349 3.10408 Sum 49 48 0 3.10454 0.07807
In order to check if the relation between covering interest rate risk for 50% or more and size at nominal level is statistical significant, table 12 is used where the frequencies of the two variables are compared with each other. Again the chi-square test is used, which is the most elementary method. As denoted in table 13 the chi-square test shows in 2007 a c2-value of 560.16417, which indicates that 0.00000 or 0% of the cases is determined by coincidence. This percentage shows that there is a chance of 100%, which is statistical significant, that there is a positive relation between covering 50% or more interest rate risk at pension funds and size in 2007.
Table 12: observed frequencies covering interest rate risk in 2007 (Oij).
Covered 50% or > interest rate risk
Covered less than 50% interest rate risk
Total
25 largest pension funds 14 10 24
26 – 50 largest pension funds 15 10 25
26
Table 13: statistical overview ALM use in 2007 (C2).
Row Column Oij Eij Oij-Eij (Oij-Eij)2 (Oij-Eij)2/Eij C2 1 1 14 23.02041 -9.02041 81.36776 3.53459 1 2 10 0.97959 9.02041 81.36776 83.06293 2 1 15 23.97959 -8.97959 80.63307 3.36257 2 2 10 0.20408 9.79592 95.96002 470.20408 Sum 49 48 0 560.16417 0.00000 3.3 Data description
The Dutch pension system consists out of 454 pension funds on December 31th 2011 and is by almost 90% of total assets represented by the fifty largest pension funds40. This thesis therefore focuses of the fifty largest pension funds in the Netherlands. The fifty largest pension funds are defined by ‘Het Financieel Dagblad, November 11th 2011’, which selected these pension funds on the basis of their total assets (Appendix A). This paper focuses on the period 2001-2010, since there are insufficient annual reports available of the period before 2001. At the time of generating the dataset, the 2011 annual reports were not available. As a criteria for selecting if a pension fund is using an ALM-strategy, the annual reports of each individual pension fund for each separated year is used.
The available annual reports were send to me after personal contact with most of the fifty pension funds and were additionally found in the database of ‘The Chamber of Commerce’41. The annual reports form the basis for generating a dataset, dividing the individual years for each fund in separated sheets (Appendix B- K). For each of these years the asset allocation mix denoted in the annual reports is used, subdivided in three categories; fixed-interest securities, business values and other investments. The category fixed-interest securities consists out of the percentage bonds and other low risk products that the fund had on its balance-sheet in that specific year. The category business values consist out of the percentage; stocks, real estate, private equity investments, commodities, infrastructure and alternatives that the pension funds had on its balance-sheet. The category other investments consist out of the percentage; hedge funds, global tactical asset allocation, derivatives, liquidities and remaining on its balance-sheet.
In order to test if pension funds that invested in an ALM-strategy generated a higher funding ratio than pension funds that did not, two groups for each separated year were
40 http://www.kps.nl, October 31th 2012, First pensions: ‘’Communiceer relatief rendement’’.
41The Chamber of Commerce is a Dutch non-profit organization that aims to help entrepreneurs to do better
27 generated for the period 2001-2010. The first group contains all fifty pension funds that used an strategy and the second group contains the pension funds that did not use an ALM-strategy in that specific year. These pension funds are categorized in one of these two groups, only if it is explicitly denoted in the annual reports that they did or did not use an ALM-strategy. For the last three years most annual reports are available, which will result in higher quality data concerning these years. A notable characteristic of the data is the specific asset allocation in the annual reports of the pension funds regarding 2008, 2009 and 2010. To test if the size of pension funds is related to the risk appetite in terms of asset allocation, the data of the percentage fixed-interest securities of the fifty largest pensions for the period 2001-2010 is collected from the annual reports.
To test how interest rate risk relates to the use of ALM at the fifty largest pension funds and to test if the size of pension funds is related to a higher risk appetite, the data regarding interest rate risk is collected from the 2007 annual reports of each individual pension fund. The funds are separated in covering 50% or more interest rate risk and covering less than 50% interest rate risk, this is because a clear distinction could be made using annual reports. In the annual reports some funds give an exact percentage covered, but most pension funds only report covering interest rate risk partly or for the greater part interpreting that these words respectively suggest a percentage under or above the 50% interest rate risk covered. For comparison, the funding ratios of June 30th 201242 are collected to test the relation between the use of ALM, covering interest rate risk and the funding ratio.
The descriptive statistics of the funding ratio show that the mean, reported in row two of table 14, varies between 102.56% in 2008 and 148.08% in 2007. The standard deviation, which is denoted in row three, shows there is a high level of variation between the funding ratios of each year, which lays between 0.1056 and 0.2110.
Table 14: Descriptive statistics funding ratio.
Funding ratio 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Mean 130.55% 114.29% 120.33% 125.66% 128.93% 139.39% 148.08% 102.56% 113.92% 111.46% Standard deviation 0.1332 0.1969 0.2110 0.1967 0.1913 0.1907 0.1823 0.1125 0.1056 0.1027 Median 128.10% 112.00% 117.20% 122.00% 125.00% 135.00% 144.35% 101.00% 111.65% 109.00% Minimum 105.50% 93.00% 99.00% 104.00% 107.30% 115.60% 117.10% 80.00% 98.10% 95.10% Maximum 165.00% 212.50% 232.40% 234.30% 235.80% 234.90% 209.50% 135.80% 154.30% 146.50% Number of observations 37 39 40 47 48 48 48 49 50 50
28 The descriptive statistics of the fixed-interest security show that the mean, denoted in row two of table 15, varies between 42.52% in the year 2001 and 50.82% in the year 2008. The standard deviation denoted in row three shows little variation and lays between 0.1003 and 0.1283.
Table 15: Descriptive statistics fixed-interest security.
IV. Results
4.1 ALM-strategy
Currently, ALM-strategies are the most common used strategies by pension funds. However, several papers debate about its shortcomings (Goorbergh, Molenaar, Steenbeek and Vlaar, 2011), (Van der Lans, 2011). The next section elaborates on the value added by ALM-strategies for pension funds.
4.1.1 Value ALM-strategies
Table 16 shows the average funding ratio for all fifty pension funds and the average funding ratios with and without the use of ALM-strategies for each separated year. The data shows that the number of observations without the use of ALM is very concise. An explanation for the limited amount of observations is the lacking reporting in annual reports. When a pension fund is using ALM it is clearly denoted in the annual report, which will result in the most significant outcomes. Table 16 shows that the average funding ratio for pension that used an ALM-strategy is lower than the average of all fifty pension funds. In addition, figure 5 visualizes that pension funds that did not use an ALM-strategy performed better than pension funds that did use ALM. What amplifies these findings is that when advancing the average funding ratio of the fifty largest pension funds with one year, since the use of ALM possibly has its effect on the funding ratio one year later, the results are virtually identical for the years
29 before the financial crisis. This is striking because the use of ALM gives pension funds the ability to better quantify risks, prepare for future uncertainties and therefore should generate a higher performance expressed in the funding ratio, however this is not the case (Romanyuk, 2010). The results found are firstly explained by the literature suggesting that current ALM-models do not take sufficient economic changes into account (Goorbergh, Molenaar, Steenbeek and Vlaar, 2011). A second explanation for these counterintuitive results is the possibility that pension funds do not use the actual outcome of ALM-studies, giving their own interpretation for managing assets and covering risks. However, for taking actual conclusion concerning this second explanation the availability of data insufficient.
Table 16: Funding ratios with and without ALM-strategies for the period 2001-2010.
Year Average funding ratio all 50 (%) Observations with ALM Average funding ratio with ALM (%)
Observations without ALM Average funding ratio without ALM (%) N 2001 130.55 7 127.47 4 134.33 11 2002 114.29 9 107.39 4 145.80 13 2003 120.33 12 114.06 4 151.63 16 2004 125.66 17 120.31 5 150.10 22 2005 128.93 34 125.53 3 140.20 37 2006 139.39 41 139.21 1 137.70 42 2007 148.08 43 147.41 2 197.70 45 2008 102.56 45 102.04 1 101.50 46 2009 113.92 47 114.09 1 105.70 48 2010 111.46 50 111.46 0 - 50
Figure 5: Average funding ratios of pension funds with the use of ALM and without the use of ALM for the period 2001-2010. 0.00% 20.00% 40.00% 60.00% 80.00% 100.00% 120.00% 140.00% 160.00% 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
30 4.1.2 Effect ALM on funding ratio
In order an explanation if the use of an ALM-study has a negative effect on funding ratios, the results of 2004 and 2007 are compared with the funds funding ratios of June 30th 201243. Table 17 displays the number of pension funds that did use ALM, did not use ALM and did not explicitly denote in their annual report if they used ALM in combination with the amount of observations with a funding ratio below the 105% norm. After comparing ALM-studies of the different funds with recent funding ratios, 2004 shows the highest amount of pension funds that did not use ALM. The results show that 80% of the pension funds in 2004, that did not use an ALM-study, suffered a liquidity problem at the end of June 2012. In contrast, 76% of the pension funds that did use an ALM-strategy experienced a funding ratio below the required norm of 105%. For the funds that do not explicitly denote in their annual reports if they used ALM, the results show that 64% of the funds is facing problems at the moment. Funds that did use an ALM-study in 2004 did perform slightly better compared to funds that did not, however the difference is negligible. In 2007 only two funds do not report the use of an ALM-strategy explicitly, however do not have a funding ratio below the 105% norm on 2012. Since these results are negligible, a preliminary conclusion can be drawn. The use of ALM is possibly related to a lower funding ratio and current models are not the proper tool for increasing performance at pension funds and do not add value. However, due to the conflicting results further research is crucial.
Table 17: The use of ALM-studies in 2004 and 2007 compared with funding ratios of end June 2012. Number of observations
<105% on 30-06-12
Percentage <105% on 30-06-12
No use of an ALM-study in 2004 4 out of 5 80%
Use of an ALM-study in 2004 13 out of 17 76%
Not explicitly given if ALM-study is used in 2004 18 out of 28 64%
Use of an ALM-study in 2007 34 out of 47 72%
Not explicitly given if ALM-study is used in 2007 0 out of 2 0%
4.2 Asset allocation
The literature describes that pension funds are getting larger and traded there more conservative assets for higher risk assets like stocks and started to invest in derivatives (Drijver, 2005), (DNB44). But did pension funds employ a correct asset allocation? The next section looks into the asset allocation of the largest, best performing and worst performing pension funds in comparison to the average of the fifty largest Dutch pension funds.
31 4.2.1 Three largest Dutch pension fund
Table 18 displays the funding ratios and the percentage fixed-interest securities of the three largest pension funds in the Netherlands, which are: ABP, Zorg & Welzijn and PMT Metaal en Techniek. The results show that the funding ratios of all three pension funds plummeted after 2007, which is due to the financial crisis and is used as a starting point. The results show that pension funds invest on average every year a smaller part of their total assets in fixed-interest securities, which are considered most safe. Pension funds took more risk every year applying no anti-cyclical investment policy, which is in line with the literature (Bikker, Broeders and de Dreu, 2007). These pension funds did not forecast the financial crisis, what putted them in perilous positions when stock prices plummeted.
Table 18: Funding ratios and percentage fixed-interest securities three largest pension funds 2001-2010.
The results show that the three largest pension funds invested before the financial crisis a lower percentage in fixed-interest securities compared to the average of the largest fifty pension funds. Figure 6 shows that these higher risk portfolio´s resulted in a significant lower funding ratio after the start of the crisis. The higher risk taking can partly be explained by the fact that the three largest pension funds already had funding ratios that were lower than the average funding ratio of all funds. These funds tried to make the fund more solvable, which is consisted with earlier literature and can be used as a starting point (Davis, Grob and de Haan, 2007). Another explanation that the literature advocates is that smaller funds invest particularly in less risky sophisticated and complex other asset classes, while large funds do the opposite and rely on their asset allocation strategies (De Dreu and Bikker, 2009).
Average all 50 funds ABP Zorg & Welzijn PMT
