Master Thesis
A Study on Risk Drivers in Dutch Pension Funds’ Asset Allocation
Master Thesis MSc Finance
University of Groningen, Faculty of Economics and Business July 29, 2015 Erik-Jan Nieboer Student Number: S2573830 Bloemstraat 31 9712 LB Groningen Tel.: +31 (0)622212858 E-mail: h.j.nieboer@student.rug.nl Supervisor: S. Drijver Abstract
This thesis studies the relationship between the equity allocation of pension funds and characteristics of active participants and the board. Those characteristics are the
wealth, age, income and gender distribution of active participants and board characteristics such as the age of the board members and the presence of a woman within the board, of funds from 2008 until 2013. The analysis is conducted using a regression of the equity allocation on these variables. This study partially proves that pension funds do account for the preferences of their participants. Another finding is that the asset allocation to equity decreases with the age of the board members, and
1. Introduction
The pension fund system in The Netherlands scores as one of the best in the world, as it is appointed as the most adequate, and second-best in sustainability1. Also the size of the funds is significant, at the end of 2013, total wealth in Dutch pension funds accounted to 1,020 billion euros2, almost twice as much as the total savings of Dutch households3. Historically, pension benefits of retirees have been indexed yearly keep up with inflation, as the pensions were well funded and returns on investments were outstanding. However, since the credit crunch of 2008, many pension funds have cut indexation, hurting retired participants and older participants most. Some funds raised pension contributions, weighing especially on the active participants. The Dutch government has increased the retirement age, partly because of these events.
The institutional organization of the pension system in the Netherlands is structured as a three-pillar system. The first pillar provides a basic pension linked to the statutory minimum wage for all inhabitants that reach the retirement age, the second pillar consists of pensions managed by pension funds and the third pillar consists of tax-deferred personal savings. The Dutch pension funds fall within the second pillar.
Dutch pension funds are mostly defined benefit plans (DB), but defined contribution (DC) funds are on the rise. DB plans guarantee a monthly payoff after retirement, based on the average salary over ones lifetimes or final salary, where the payoff in a DC plan is based on the contributions made by the employee, the employer or both. Any gains, losses, income and expenses are allocated to this account and it does not guarantee any payoffs. This transformation from DB to DC plans gives way to transfer risks from the funds to the participants, but should in exchange give the participants more input in the organization of their DC pension funds, a DC plan should therefore account for participants’ preferences, including, but not limited to, risk-aversion, lifetime expectancy, preferable pay-out, bequest motive and portfolio composition, as argued and illustrated in numerical examples by Konicz and Mulvey (2015). Where Konicz and Mulvey (2015) argue that a defined contribution plans should be constructed according to the preferences of their
1
Melbourne Mercer Global Pension Index, October 2013.
participants. Honda (2012) goes even further and argues that the total risk attitude of a pension fund should be the aggregate of all the participants. In this view, especially since the Dutch participants have limited choice-ability, it is interesting to study if Dutch pension funds exame and assimilate the risk preferences of their participants within the DB portfolios.
Dutch employees have little choice in deciding which pension fund fits their interests and preferences best, especially regarding the DB plans, since they are automatically enrolled in the pension funds their employer has appointed, or is appointed to. Kool, Prast and Van Rooij (2007) show that even if Dutch pension fund participants had more choice-ability, they would choose portfolios that are more risk averse than their present pension fund portfolios, but, if they could choose, the majority would not. This weakens the argument that the portfolios should consider this aggregate risk attitude, since the participants their selves, prefer to hand it over to the pension fund. The main reason people do not want to take matters in own hand is that they consider their selves not financially educated and sophistically enough. Even when offered financial advice to increase their knowledge, most would still reject the choice to manage their own pensions. When people are thoroughly dissatisfied with their pension fund, and they do change, according to Henken and Van Dalen (2015) this decision to change is based on irrational choices. Participants switch to another pension fund, insurer or bank out of dissatisfaction, without trusting their new pension manager to do better in terms of risk preferences or returns. Arguing that participants of pension funds rather have fund managers to deal with their portfolios raises the question if pension funds hold into account the preferences of their participants, and for what reason. Evidence that pension funds are not only dealing with optimizing returns, but are also accounting for preferences of their participants, can be seen in the corporate social responsible investments of the funds. Pension funds were one of the pioneers in responsible investments (Cox and Schneider, 2010), something that can’t be related to financial motives alone.
who has reached the retirement age and receives a monthly grant from the pension fund. The focus of this study will be on the active participants, since the non-active participants in most cases become active participants in another fund and transfer their accrued pensions, when allowed4, to their new pension fund. Therefore I expect they will not be actively involved in funds they were previously enrolled in.
Next to the influence of participants’ preferences on pension funds, the composition of the board of pension funds can exercise an influence. As argued by Ntim and Osei (2011), company performance is affected by composition of the board. The characteristics of individual executives and the board heterogeneity are determents of board behaviour as Kiel and Nicholson (2003) conclude in a study to Austrialian firms, the relationship between firm performance, board composition and size and firm size is positive.
This findings and theories raises the following main question which I will try to answer in this study: “Do personal characteristics of participants and board members
of pension funds impact risk-taking?”
In present literature, characteristics of individuals have been studied with regard to risk attitudes, for individuals as investors, as well as on a personal level. Some of these characterstics have been studied to explore the impact on the behavior within pension funds, but not in the scope this study will take. I will study the impact of personal preferences of the active participants, as well as the board members on the pension fund portfolios, to extend the empirical literature to the influence of personal preferences on risk attitudes within pension funds.
In this study, the end of year allocation to equity will be used as the relevant risk measure of a pension fund, since this is a clearly defined measure of risk, next to that, it is accurate and perceived in most literature as a good measure for risky assets. The data is collected through Dutch pension funds’ annual reports and the Orbis database on board composition. Data spans from 2008 to 2013 and covers 15 pension funds. The sample consists of demographics characterstics of the active participants and the board composition of Dutch pension funds.
The results conclude that there is some evidence that implies that personal characterstics of active participants and the composition of the board influence risk taking within pension funds. The economic interpretation of the results are in line
4 Only when the new and the previous pension fund have a funding ratio of at least 100%, transmitting
with present emperical studied, but since the results differ in economic and statistic significance throughout the difference models, arguably due to presence of multicollinearity, the results should be interpreted with caution.
2. Literature review
This section starts with breaking down personal characteristics of investors and natural persons to risk attitudes as observed in the literature. After that, the observed characteristics will be structured in subsections to perform an extensive study on the individual characteristics to determine the expectations towards risk, which then will be formed into individual hypotheses per characteristic.
Risk preferences are extensively surveyed within the present literature, the focus of this paper will be on four distinctive characteristics that influence the risk aversion of the participants. These characteristics are age, gender, wealth and income. The literature will point a direction of the influence of these characteristics on risk aversion. Hypotheses are formed, using the outcome of the literature review.
2.1. Wealth
The influence of wealth on the risk aversion of investors and individuals was first observed by Von Neumann and Morgenstern (1943). They show in a set-up of gambles that wealthy individuals are willing to take on more risks. Which implies that risk aversion is negatively related to the level of wealth. These results are further developed extensively within the literature and there is a great pool of proof that wealth is indeed negatively related to risk aversion. Important impactful results have been confirmed by a variety of studies that show that wealth is indeed negatively related with risk aversion (Hallahan, Faff and McKenzie, 2003; Cohn, Lewellen, Lease and Schlarbaum, 1975; Cass and Stiglitz, 1972; Friedman, 1974; Pratt, 1964, Riley and Chow, 1992). in the study of Bucciol and Miniaci (2015) to household portfolio risk, a skewed distribution of risk tolerance is found, which is increasing with wealth, and many households bearing little risk. Nonetheless most of the studies take a different approach to wealth, different samples, and different tests. The results are conclusive, implying that risk aversion decreases with wealth. Risk aversion tends to be countercyclical, Sousa (2015) finds that shocks to wealth are positively related to the allocation in risky assets. Stretching this result to pension funds, the assumption is made that when a positive (negative) shock in wealth occurs, they would invest more (less) to risky assets proportional. The total wealth of a pension fund can be seen as the total wealth within the fund.
should be positive. Although there is no clear reason, except for diversification for pension funds to increase their risk allocation to risky assets when they are wealthier, since their main objective is to have enough assets to pay future pensions. This argument leads to the forming of the following hypothesis:
Hypothesis 10: There is no effect of the wealth of participants on the asset allocation to risky assets of pension funds.
Hypothesis 11: There is an effect of the wealth of participants on the asset allocation to risky assets of pension funds.
2.2. Age
invested to risky assets to free disposable assets to make a comparison with other studies in that field, such as studies conducted by Cohn, Lewellen, Lease and Schlarbaum (1975) and Morin and Suarez (1983). Hallahan, Faff and McKenzie (2003) explore the relation of risk tolerance and demographic factors and find a non-linear relation between age and risk tolerance, the risk tolerance increases with age to the age of around 35, after this age, the risk tolerances decreases. The risk tolerance is affected by the interaction between wealth and age, but when accounting for this interaction, the same results stay intact. An important difference between this investigation and previous studies, is that risk tolerance is measure psychometrically and not in a measure of allocation to certain (risky) assets.
Human capital in the sense of finance is the sum of future income of an individual. In the study of portfolio optimization in the presence of human capital, the literature points to decreasing the proportion to risky assets with age in a proposed world with a riskless asset and no borrowing. Young workers have a great portfolio of human capital and therefore can diversify away the risk of equity exposure with their human capital, since the correlation between these two is low. Therefore equity exposure should decrease with age and human capital (Bodie, Merton and Samuelson, 1992; Cocco, Gomes and Maenhout, 2005; Love, 2013). This confirmed results in behavioural psychology that younger people take more risks (Sharpe, Wang and, 2011).
A cross-sectional approach to study the Finnish pension funds in 2002, found a negative relationship between the average age of employees and the percentage of equity in the portfolio of the pension fund of -1,7%. This coefficient implies that when the average age of the employees rises with one year, the allocation to equity decreases with 1,7% (Alesto and Puttonen, 2005). That confirmed their hypothesis that pension funds with younger participants have longer investment horizons and thus hold more risky assets than funds with older participants. This result is confirmed by a study done of Bikker, Broeders, Hollanders and Ponds (2012) in a study to lifecycle investments in Dutch pension funds and show that the allocation to risky assets decreases with the age of the participants. When age increases, the human capital of such individual shrinks, leaving less time to make up for shocks.
Hypothesis 20: There is no effect of the age of participants on the asset allocation to risky assets of pension funds.
Hypothesis 21: There is an effect of the age of participants on the asset allocation to risky assets of pension funds.
2.3. Income
Income is positively correlated to wealth, human capital and age. The higher the income of individuals, the more likely it is that the individual has the possibility to accumulate wealth over their lifetime. As human capital is the expected income from an individual over their remaining lifetime, income can be a good proxy for expected income over one’s lifetime. Human capital and the level of wealth both have the same expected positive signs on the allocation to risky assets, therefore the first intuition is to expect that income also will have a positive effect on the risk allocation to equity of investors. This effect has been studied extensively in literature. The study of Riley & Chow (1992) shows a positive relationship between income and allocation to risky assets, they therefore determine that risk-taking is positively related to income. Individuals can become risk-takers when they were risk-avoiders when incomes increases in an environment of uncertainty, this is shown by Chang (2008) in a study were people were given the choice of insurance in an uncertain situation. In a household portfolio study by Kapteyn and Teppa (2011), their results confirm that income is a positively related on the risky asset share of a households’ asset allocation. These results can be extrapolated to investors, where several empirical studies have shown that risk aversion decreases with income (Grable and Lytton, 1998; Hallahan, Faff and McKenzie, 2003; Scooley and Worden, 1996)
The relationship between income and the risk tolerance in pension funds is not described in recent literature. A reason for this is that, although higher income individuals will take more risk in their own investments, pension funds would not have a reason to do that.
Hypothesis 30: There is no effect of the income of participants on the asset allocation to risky assets of pension funds.
2.4. Gender
In a survey that measures the expectation of investors about future equity returns, Dominitz and Manski (2007) find that men are expecting higher returns than women, which underlines their assumption that women are more risk averse than men. Arano, Parker and Terry (2010) find that women are more risk averse than men in pension portfolio decisions. These factors would influence the allocation of the pension funds’ assets should they account for the preferences of their participants.
Barber and Odean (2001) study the investment portfolios of 35,000 households. They hypothesize that men are more overconfident then women, and thereby are more risk tolerant, which is confirmed in their results, that points out that men make 45% more trades than women, thereby concluding that men are more confident in uncertain situations.
Sebai (2014) found similar evidence in a study of more then 2,000 customers of a Tunisian broker. After splitting the sample between men and women, they observe statistically and economically significant, that women are more risk averse than man and invest their resources more conservatively. In Pension fund analysis, several studies find evidence that women allocate less to equity in their pension portfolio (Bernasek and Shwiff, 2001; Bajtelsmit, Bernasek, and Jianakoples, 1999). Although the literature clearly points to women being less risk averse then men, a reason that pension funds still could appoint more equity to women, is their life expectancy, since women tend to live longer and therefore possess more human capital.
Since the goal of the pension fund is to provide enough money to pay the pensions of the participants, the gender distribution should not have an impact on the amount of risky assets of the pension fund, therefore the following hypothesis comes to order:
Hypothesis 40: There is no effect of the gender of participants on the asset allocation to risky assets of pension funds.
2.5. Board effects
The results of Berger, Kick and Schaeck (2014) suggest that the younger the board of directors is, the more risk the board takes. This result is in line with a study of Platt and Platt (2012) who find that the average age of the executives of firms that go bankrupt, is significantly lower than firms that do not. Johnson, Schnatterly and Hill (2013) synthesize the available literature on board composition demographics and argue that this effect is due to the valuable experience of older executives. This effect in corporate firms will likely have less of an impact on boards of pension funds, since the strategic asset allocation is determined by the asset liability management study, done by external advisors and therefore I argue that the asset allocation should be independent of the composition of the board.
To test if this effect of age on risk taking, by board of pension funds is also present, the following hypothesis is constructed.
Hypothesis 50: There is no effect of age distribution of the board on the asset allocation to risky assets of pension funds.
Hypothesis 51: There is an effect of age distribution of the board on the asset allocation to risky assets of pension funds.
Since board members of pension funds have the mandate regarding the risk management of the pension fund, the demographics of the board, and by that, the risk preferences will influence the investments decisions made by the board.
To test if this also accounts for Dutch pension funds, the following hypothesis will be tested.
Hypothesis 60: There is no effect of gender distribution of the board on the asset allocation to risky assets of pension funds.
3. Data and Methodology
This section discusses the methodology and the data that is used to study the relationship of active participants’ and pension fund board members preferences to pension funds
3.1. Methodology
The methodology used in this paper is similar as the most prominent one used in the literature to study risk attitudes of investors, namely the asset allocation to equity of total investments (Schooley, 1996; Riley and Chow, 1992; Honda, 2012).
The end of the year allocation to equity is measured as the total value of equity investments as stated in the annual report as percentage of the total investments. This measure does not account for other assets categories, such as real estate, hedge funds, private equity, nor does it take into account pension funds’ measures to reduce the level of risk, using for example put options or derivatives to hedge risks. Taking these measures in account would more adequately measure the real risk, but the data and disclosure of pension funds makes the allocation to equity the most accurate measure for the amount of risk. Since pension funds do not rebalance their portfolio very often (Bikker, Broeder and De Drue, 2010), the real allocation will deviate from the strategic allocation. Nonetheless the real asset allocation to equity at the end of the year is the proxy for the risk of the pension funds’ portfolio because it represents the real risk that is taken by the pension fund instead of the admired.
𝑒𝑛𝑑 𝑜𝑓 𝑦𝑒𝑎𝑟 𝑒𝑞𝑢𝑖𝑡𝑦 𝑠ℎ𝑎𝑟𝑒𝑖,𝑡 = 𝛼 + 𝛽1 𝑤𝑒𝑎𝑙𝑡ℎ𝑖,𝑡+ 𝛽2 𝑎𝑔𝑒𝑖,𝑡+ 𝛽3𝑖𝑛𝑐𝑜𝑚𝑒 𝑖,𝑡+ 𝛾1 𝑖𝑛𝑐𝑜𝑚𝑒 𝑐𝑎𝑝 𝑖,𝑡 + 𝛽4 𝑔𝑒𝑛𝑑𝑒𝑟𝑖,𝑡+ 𝛽5𝑎𝑔𝑒 𝑏𝑜𝑎𝑟𝑑𝑚𝑒𝑚𝑏𝑒𝑟𝑠𝑖 + 𝛾2 𝑤𝑜𝑚𝑒𝑛 𝑏𝑜𝑎𝑟𝑑𝑟𝑜𝑜𝑚𝑖+ 𝛽6 𝑠ℎ𝑎𝑟𝑒 𝑓𝑜𝑟𝑚𝑒𝑟𝑖,𝑡 + 𝛽7 𝑠ℎ𝑎𝑟𝑒 𝑟𝑒𝑡𝑖𝑟𝑒𝑑𝑖,𝑡+ 𝛽8 𝑓𝑢𝑛𝑑𝑖𝑛𝑔 𝑟𝑎𝑡𝑖𝑜𝑖,𝑡 + 𝛽9 𝑓𝑢𝑛𝑑𝑖𝑛𝑔 𝑟𝑎𝑡𝑖𝑜2 𝑖,𝑡+ 𝛽10log(𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡𝑠) + ∈𝑖,𝑡 (1)
The coefficient β is estimated per variable to assess the influence and significance of the variable on the end of the year equity allocation. The i represents the pension fund and t the year in which measurement took place.
First, two variables will be discussed, since discussing them afterwards could raise questions when reading the explanation of the key variables. Since this paper addresses the question to the influence of active participants, the influence of former and retired participants on equity allocation has to be accounted for. Therefore two variables, denoted as share former and share retired are included to account for these effects of former and retired participants of the pension fund. The variables are constructed as the share of former participants to total participants and retired participants to total participants, respectively.
The dependent variable for wealth is the provision for pension liabilities for the active participants, divided by the number of active participants. This is the most accurate measure to account for the wealth of the active participants of the pension fund, since it is the actual wealth the pension funds has reserved for the participants’ pension. Besides it is a value that is split in the balance sheet to account for the three groups of participants, so that the wealth per active participant is calculated more accurately, than dividing the total wealth by the participants, which could have been the other measure for wealth.
The second hypothesis is about the relationship of age on equity allocation, this is included in the model through the variable age. This variable comprises the average age of the active participants within the pension fund as the statistic for age.
information that is available. This variable income is a function depending on three factors, first, the total in a year paid premium. Next to the paid premium a dependent factor is the franchise, which is the income over which no pension premium is paid. In the sample: the average franchise is 13.200 euro. The third factor is the premium percentage. Only paying active participants are included in the calculation, since occupationally participants are exempt from premium payments. Hence, the formula used to estimate the average income per active participant is shown in equation 2.
𝑖𝑛𝑐𝑜𝑚𝑒 𝑝𝑒𝑟 𝑎𝑐𝑡𝑖𝑣𝑒 𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡 = (
𝑝𝑎𝑖𝑑 𝑝𝑟𝑒𝑚𝑖𝑢𝑚 𝑝𝑟𝑒𝑚𝑖𝑢𝑚 𝑝𝑒𝑟𝑐𝑒𝑛𝑡𝑎𝑔𝑒)
𝑝𝑎𝑦𝑖𝑛𝑔 𝑎𝑐𝑡𝑖𝑣𝑒 𝑝𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡𝑠+ 𝑓𝑟𝑎𝑛𝑐ℎ𝑖𝑠𝑒 (2) Another factor is the maximum pensionable salary within the defined benefit plans, in some pension funds there is maximum salary over which premiums are paid. Above this income, participants can choose to participate in other regimes that are mostly defined contribution plans or a combination of a DC and DB arrangements. Premiums for the defined benefit plan are paid up to this maximum salary, which has a mean value of 52.000 euro within the sample. Since the maximum pensionable salary doesn’t come up in this formula, the calculated incomes are downwards biased because it omits high incomes above the maximum pensionable salary. To account for this effect, a dummy variable, 𝑖𝑛𝑐𝑜𝑚𝑒 𝑐𝑎𝑝 is introduced, which takes the value of 1 when the there is a cap, and a value of 0 when not.
To test if there is a relation between gender and the asset allocation of pension funds, gender is included in the regression model. A variable is constructed that consists percentage of men of the total active participants, calculated by dividing the year-end active men by the total of active participants in the fund, the variable is denoted as gender.
Estimating the relationship of the age in the boardroom, the variable age
board members is constructed. It consists of the average age of the board members of
the pension fund per year.
influence of women being present on the board, and the fraction. This is in line with previous empirical studies on the influence of women in boards on risk attitudes.
The funding ratio is the ratio between the total assets and the provision for pension liabilities. It is no measure of wealth since it accounts for the total assets as comparison to the liabilities, a very wealthy fund, can for that reason have a low funding ratio. Another factor of investing with regards to the level of the funding ratio is the incentive of pension funds. As stated earlier, the main objective of a pension fund is to optimize their investments, in such a way that the liabilities can be paid in the future. The Dutch Central Bank sets regulations on the calculation of the funding ratio and has guidelines on the minimum funding ratio required, which is 105%. When funds have sufficient funds to pay future liabilities, they can choose to increase the payments to pensioners to remain the level of purchasing power or decrease the premiums paid by the participants, and perhaps even a combination. They also can decrease the level of risk, since they are unlikely to fall underneath required ratios. Although this is true, the funding ratio as calculated by the Dutch Central Bank standard, reflects the liabilities, as were the pensions to be paid without reckoning for future loss of purchasing power. The ambition of pension funds is to pay future pensions and indexing those to cope with the effects of inflation. The funding ratio that accounts for future inflation is often called the real funding ratio, with a ratio of 100% implying that future pensions can be paid and indexed. That implies that pension funds will not only aim at the 105% required by the Dutch Central Bank, so that when the 105% funding ratio is accomplished, the pension funds will not directly change the allocation to risky assets, but when the real funding ratio reaches the full funding of 100%, a decrease in risky assets can be expected, since all liabilities including future indexation is covered within the assets of the fund. So after a certain height, a decline in the allocation to equity to the funding ratio is expected.
The same effect of decreasing the proportion of risky assets will be present when funds are underfunded. The regulations for pension funds forbids the pension funds to take on much risks when the funding ratio is low, which means that the funding ratio will also have an effect on the level of risky assets within pension funds. This implies a parabolic shape in the level of risky assets to the funding ratio.
risk while the future pensions, including indexations is already covered would have less additional benefits. This implies a parabolic shape of the funding ratio on the allocation to equity, in the model this is denoted by a linear and a quadratic term of the funding ratio, denoted as funding ratio and funding ratio2 respectively.
The more the participants the fund has, the more funds tend to invest in equity, that result is found by Bikker and De Drue (2009). Therefore the variable size in participants in added to the regression. To deal with great size variations among pension funds and to avoid heteroskedasticity, the logarithm of the number of participants is taken.
Finally, εi,t is the variable that denotes the error term.
3.1. Data sample
The dataset is obtained to study the relationship between active participants’ and board members’ preferences on risk attitudes and the risk attitude of pension funds. The data is collected from Dutch pension funds that are in the top 100 largest pension funds in The Netherlands.
Comparing the sample to the complete population of Dutch pension funds, the sample consist of mainly the biggest pension funds, therefore any inferences that will be made from the sample, are skewed to the biggest pension funds, this is expressed as a limitation in this study. Table 1 provides an overview and some summary statistics of the sample, and compares it to the average Dutch pension fund. The Dutch pension funds environment is skewed, with a small proportion of funds, holding the most assets and participants. As shown in table 1, this is also what the sample represents, where the sample is being representative for relatively large funds, but neglecting the smaller pension funds. Total participants includes participants that are present in more than one fund, as people can be enrolled in more than one fund.
Table 2 presents descriptive statistics, with end of year allocation to equity, wealth of actives, age of actives, pensionable income, gender distribution as key variables regarding the participants, and the average of the board room members and the gender distribution within the board as key variables regarding the board effects.
From the descriptive statistics in table 2 it can be concluded that there are some large variations between pension funds. The average allocation to equity is 28.37% of total assets, with a median of 29.05%. Within the sample, the minimum equity allocation is only 13.39%, whereas the maximum is more than three times that number at 42.54%. Just as the equity allocation, other variables show reasonable variations among pension funds. Wealth varies from € 3,646 to € 261,179, a large variation, mainly due to the characteristic. With a mean value of € 89,741 and a median wealth of € 77,626 it is skewed to the right. Age of the active participants
Table 1
Sample characteristics and sample representativeness
This table present some summary statistics of the sample used in this study and makes a comparison to all Dutch pension funds over the same period from 2008 until 2013, The characteristics are averaged over these 6 years. This table therefore presents the representativeness of the overall to complete Dutch pension sector. Data on the Dutch pension sector is collected through the Dutch Central Bank (DNB) website.
Characterstic Sample Dutch pension funds
Number of funds 15 525
Number of participants total 939,234,346 17,721,447
Participants per fund 62,615 33,755
Wealth of funds total in euro 674,220 million 805,795 million
varies from 33.94 to 48.20, which gives the impression that there are no very old pension funds included in the sample, since in that case the average age should be more in the direction of the retirement age. The average pensionable income is € 38,792, close to the median value of € 38,764. Since the average share of men in pension funds is 0.60, and the median share of men is 0.70, the sample is skewed to the left regarding the share of men in the funds. The sample consists of pension funds with predominantly men, an observation consistent, but a little bit less then the average participation in pension funds in the Netherlands, in which men make up 58 per cent of the total participation5. In 80 per cent of the boards of pension funds, at least one woman is present, whereas in the other 20 per cent, only men occupy the board.
Table 3 shows the correlation matrix of the independent variables, for each variable the correlation to the other variables is displayed. There is presence of correlation between the independent variables. There is positive correlation between the average age of the actives and wealth, since participants accumulate their wealth over the years. Wealth is also highly correlated with the share of former participants (negatively) and the share of retirees (positive) and the total participants. Age and income are highly correlated, wherefore the same reasoning can be used as regarding the age and wealth, since employees’ salary tends to increase with age. It is also highly negatively correlated with share of former participants, when people are leaving for another job, the share of former increases whereas the average age decreases. Income is negatively correlated with an income cap, because when a cap is in place, the pensionable salary will be biased downwards because it omits high incomes. The dummy variable for women in the board is correlated with the other dummy variable, the cap for income, and is negatively correlated with income.
Table 2 Descriptive statistics
This table presents descriptive statistics for the pension funds in the sample, a total of 15 pension funds from 2008 to 2013. The table presents descriptive statistics for the variables in the model of equation 1. Statistics are fund-year mean, median, standard deviation and the minimum and maximum value within the sample. Detailed description of all variables can be found in the methodology section.
Variable Mean Median Standard Deviation Minimum Maximum
Equity allocation in percentage of total assets (%) 28.37 29.05 7.53 13.39 42.54
Wealth (per active participant, in thousands Euro) 89.74 77.62 63.22 3.65 261.2
Age of actives 42.72 43.65 3.41 33.94 48.20
Pensionable income (in euro) 38,792 38,764 13,544 15,652 74,220
Dummy: income cap 0.86 1 0.34 0 1
Gender (share of men) 0.64 0.70 0.23 0.15 0.91
Age of board members 55.14 56.72 4.03 44.90 60.67
Dummy: women in boardroom 0.80 1 0.40 0 1
Share former participants 0.43 0.43 0.16 0.14 0.72
Share retired participants 0.27 0.23 0.17 0.03 0.56
Funding ratio (%) 105.41 104.90 9.59 84.70 128.60
Table 3 Correlation matrix
This table presents correlation between the variables used in this study. There is high correlation present between some of the variables. Wealth refers to the wealth reserved for the active participants within a fund. It is highly correlated with the average age of the actives, since participants accumulate the wealth over the years. Wealth is also highly correlated with the share of former participants (negatively) and the share of retirees (positive) and the total participants. Age and income are highly correlated, wherefore the same reasoning is in place, as employees’ salary tends to increase with age. It is also highly negatively correlated with share of former participants, when people are leaving for another job at higher ages, the share of former increases whereas the average age decreases. Income is negatively correlated with an income cap, when a cap is in place, the pensionable salary will be biased downwards because it omits high incomes. The correlation between the share of former and the share of retires can be explained by the same reasoning as wealth and age. The share of former participants is negatively related to the share of retirees, since those to are related by definition. Concluding there is a presence of multicollinearity between the independent variables.
Wealth Age Income
Dummy : income cap Gender Age (board) Dummy: Women in board Share former Share retired Funding ratio Participa nts Wealth 1 0.8257 0.7333 -0.2763 0.2726 -0.1049 -0.0334 -0.7347 0.5470 0.3473 -0.4395 Age of actives 1 0.7781 -0.2823 0.2405 0.0579 -0.0353 -0.7799 0.5607 0.1698 -0.2373 Income 1 -0.7404 0.2842 0.0085 -0.4255 -0.5781 0.6049 0.2675 -0.4087
Dummy: income cap 1 -0.1606 0.0349 0.6666 0.2896 -0.5550 -0.1492 0.3034
Gender (share of men) 1 0.0343 -0.3799 -0.0659 0.5374 -0.1167 -0.5356
Age of board members 1 -0.1542 -0.3112 0.1476 -0.0074 0.3898
Dummy: women in
boardroom 1 0.0663 -0.2380 0.1499 0.1474
Share former participants 1 -0.6580 -0.2496 -0.0292
Share retired participants 1 0.1775 -0.3080
Funding ratio 1 -0.4155
4. Results
In this section empirical results are presented. The descriptive statistics and correlation matrix are discussed in the previous part, just as the consequences of the presence of multicollinearity, which influences the use of the models. This section illustrates the coefficient estimates of the sample.
4.1. Regression estimations
In this section, the models are estimated. The model estimates the average equity allocation in percentages per annum per fund. An ordinary least squares (OLS) model is used to test how the allocation reacts to the different characteristics of the fund. Table 4 presents the regression estimates for the different models. Model 1 in the table estimates the results for the all the variables, even though some of these variables are highly correlated, as stated in the previous section. Model 2 drops the variables accounting for the board characteristics, and model 3 drops the variables depicting the participants’ characteristics. The models 4 to 7 all drop the share of former and share of retired participants, since those variables have a high correlation with the key variables depending on the participants’. Where model 4 only drops the share of former and share of retired variables, model 5, model 6, and model 7 drop two of the four key variables regarding the participants, to deal with the multicollinearity. The models predict some statistically significant variables in the different models.
with previous a study that finds that allocation to equity or risky assets in general, increases with wealth of participants. (Bikker, Broeders, Hollands and Ponds, 2012).
The coefficient of average age of active to the allocation to equity, is not significant in the full model, nor in the model that only accounts for preferences of active participants (model 2), nor in the model that only accounts for the age of active participants. however in model 4, the average age of active participants is significant in a statistical manner. It predicts a coefficient of -1.15, implying that with increasing age, the average allocation to equity would decrease with 1.15% if age increases by one year. This finding is in line with previous studies to age dependent investing within pension funds. (Alesto and Puttonen, 2005; Bikker, Broeder, Hollands and Ponds, 2012). But, since there is mentioning of multicollinearity in the sample, and the coefficient is only significant in a model with the presence of two highly correlated variables, and looses significance in a model that drops these variables, this result should be handled with caution.
income is negatively related to risk aversion (Riley and Chow, 1992; Kapteyn and Teppa, 2011).
The coefficient for gender, denoted in this model as the share of men, is significant in four of the six models, and has a negative relationship varying between -8.4 per cent and 15.2 per cent. This would imply that, with the share of men increasing 0.01, or 1 per cent, the allocation to equity actually decreases with 8.4 per cent. This is clearly also economically significant since the share of men in the sample varies between 0.15 and 0.91. These results are conflicting the general view in the literature of risk attitude between men and women, where men are considered less risk averse then women, and more overconfident. Even in pension fund portfolios, where Beransek and Shwiff (2001) find that funds with more women allocate less to equity. The two models, in which no significant coefficient is estimated, is the full model 1, and the model focusing on wealth, model 6.
Turning to the relation of board members on the equity allocation, the model predicts some interesting results, the first variable is the age of the board on the equity allocation, this variable is included in six of the seven models, and is significant in all the six models. The relationship between the age of the board members and the equity allocation is negative in all, but one model, with coefficient ranging from +0.74 to -0.64, although the correlation between dependent variables regarding the active participants and the age of board members is low, dropping the active participants’ age and income variable in the model, causes the relationship between age of board members and the dependent variable to be positive. However, since most models predict a statistically and economic significant negative sign, the results indicates a negative relationship. The results are in line with evidence found in a previous study by Berger, Kick and Schaeck (2014) that found that younger teams take more risk and are responsible for more bankrupts than older boards.
The share of former participants tends to be positively related to the amount of risk taking, which is in line with the idea that those are non-retired participants, and the pension fund is still taking risk to accumulate wealth for their pensions. This relates to the share of retired participants having a negative relationship, whereas the wealth of retirees is mostly converted into a lifetime annuity, therefore the wealth of the retiree is still within the pension funds, although mostly none of these funds are invested in equity. The funding ratio of a pension funds tends to have a positive relation to risk taking, and confirms the parabolic shape of the funding ratio to the equity allocation. However, the funding ratio is only statistically in two of the models. In general, the relation of equity allocation to the logarithm of total participants is positive, although not statistically significant in six of the seven models.
Table 4
Regression estimates
This table provides estimation results from the model in equation (1) that uses an OLS to estimate the end of year equity allocation in percentages of pension funds, based on a set of pension fund characteristics and control variables. Model (1) is the full equation, including all variables, model (2) only accounts for the effects of the active participants, including associated (control) variables. Model (3) only accounts for the effects of board characteristics. The rest of the models deals with causes of multicollinearity, and therefore drops variables that are highly correlated, as presented in table 2. Coefficient estimates for the different models are denoted behind the variables with the standard error within parentheses, *** denotes significance at the 1% level, ** denotes significance at the 5% level and * denotes significance at the 10% level. The adjusted R2 is stated to present how well the model fits the data.
Model (1) Model (2) Model (3) Model (4) Model (5) Model (6) Model (7)
α -107.646* (60.298) -137.746** (62.542) 36.503*** (9.928) -34.328 (63.489) 52.035 (67.642) 5.254 (6.754) -3.204 (6.326) Wealth per active participant (in
thousands euro) 0.055* (0.029) 0.060* (0.031) 0.023 (0.025) 0.0194 (0.019) Age of actives -0.674 (0.554) -0.110 (0.586) -1.155*** (0.436) -0.246 (0.266)
Pensionable income (in
thousands euro) 0.083 (0.192) 0.015 (0.209) 0.350*** (0.162) 0.096 (0.083)
Dummy: income cap 0.545
(6.994) 11.998* (6.63) 19.904*** (5.146) 15.201*** (4.004)
Gender (share of men) 2.935
(6.436) -15.196*** (4.358) -13.295*** (3.922) -8.408** (4.069) -6.155 (4.179) -12.572*** (3.742)
Age of board members -0.370*
(0.213) -0.509*** (0.182) -0.507** (0.208) -0.519** (0.235) 0.738*** (0.236) -0.636*** (0.195)
Dummy: women in boardroom 9.144***
(2.908) 11.152*** (1.696) 2.930 (2.340) 9.046*** (1.812) 8.706*** (1.291) 1.570 (2.397)
Share former participants -7.098
(11.913)
19.173* (10.212)
Share retired participants -31.939***
5. Discussion
This research is aimed at the causal relationship between the risk taking of pension funds, in the thought of allocation to risky assets, and the perceived risk preferences of participants based on their characteristics. It differs with previous studies, as this study includes more than one risk characteristic, and tries to identify if the board and active participants’ characteristics are determents of risk taking within pension funds.
The analysis is conducted by estimating the coefficients using an ordinary least squares (OLS) regression on four active participant’ characteristics, being wealth, age, income and gender. The regressions take into account two characteristics of the board, being the average age of the board members, and the presence of at least one woman in the board.
The conducted analysis presented some interesting results. However, some caution interpreting this results is just, since none of these results could be proved in every conducted model.
The relationship between board characteristics and allocation to equity are studied, the first factor is how the average age of the board affects the allocation to equity. I show that the average age within the board is negatively related to the amount of equity, with an average estimated value of -0,5%. A reaction could be that this can be explained because the average age of the board is correlated to the average age of the active participants, but this is not the case. Although these results have found before in other studies (Berg, Kick and Schaeck, 2014), it is surprising that personal characteristics of the board members, influence the allocation decisions, especially since the pension fund by definition is an organization that should be governed to safely accumulate the pension of its participants. When at least one women is on the board of a pension fund, compared to a board that is fully occupied by men, has a positive relation with the amount of risk taking that is statistically significant at levels from 8.7 to 11.1 per cent, a result that was also present in the study of Berger, Kick and Schaek (2014), but is at least as surprising as the previous finding, for the same reasons.
This study gives an insight in how pension funds are dealing with the characteristics of their participants in some models, although not statistically significant in every model, the direction and economic interpretation of the coefficients, shows some interesting results regarding this phenomenon. This does not directly let me reject the null hypotheses, but does fuel the argument that pension funds risk attitudes, should, and most likely are considering, the aggregate attitude of all it’s participants (Honda, 2012). Regarding the relation on board characteristics on the risk taking of pension funds, this study shows a clear direct of decreasing risk with increasing board members’ age, and increasing risk within the present of women.
References
Adams, R., Ferreira, D., 2009. Women in the boardroom and their impact on governance and performance. Journal of Financial Economics 94, 291-309.
Arano, K., Parker, C., Terry, R., 2010. Gender-based risk aversion and retirement asset allocation. Economic inquiry 48, 147-155.
Allesto, N., Puttonen, V., 2006. Asset allocation in Finnish pension funds. Journal of Pension Economics and Finance 5, 27-44.
Ameriks, J.A., Zeldens, S.P., 2004. How do household portfolio shares vary with age? Working paper, Colombia University.
Bajtelsmit, V. L., Bernasek, A., Jianakoples, N. A., 1999. Gender differences in defined contribution pension decisions. Financial Services Review 8, 1-10.
Barber, B. M. and T. Odean, 2001, Boys will be boys: gender, overconfidence, and common stock investment. Quarterly Journal of Economics, 116 (1), 261-292.
Berger, A.N., Kick, T., Schaeck, K., 2014. Executive board composition and bank risk taking. Journal of Corporate Finance 28, 48-65.
Bernasek, A., Shwiff, S., 2001. Gender, risk and retirement. Journal of Economic Issues 15 (2), 345-356.
Bikker, J.A., De Drue, J., 2009. Operating cost of pension schemes: The impact of scale, governance and plan design. Journal of Pension Economics and Finance 8, 63-89.
Bikker, J. A., D. W. G. A. Broeders, and J. De Dreu, 2010, Stock market performance and pension fund investment policy: Rebalancing, free float, or market timing? International Journal of Central Banking 6, 53-79.
Bikker, J. A., Broeders, D. W., Hollanders, D. A., Ponds, E. H., 2012. Pension funds' asset allocation and participant age: A test of the life-cycle model. The Journal Of Risk and Insurance 79 (3), 595-618.
Bodie, Z., Merton, R.C., Samuelson, W.F., 1992. Labor supply flexibility and portfolio choice in a life cycle model. Journal of Economic Dynamics and Control 16, 427-449.
Bucciol, A., Miniaci, R., 2015. Household Portfolio Risk. Review of Finance 19 (2), 739-783.
Chang, Y., 2008. Risk avoidance and risk taking under uncertainty: a graphical analysis. American Economist, 73-85.
Cocco, J.F., Gomes, F.J., Maenhout, P.J., 2005. Consumption and portfolio choice over the lifecycle. Review of Financial Studies 18, 491-533.
Cohn, R.A., Lewellen W.G., Lease, R.G., Schlarbaum, G.G., 1975. Individual investor risk aversion and investment portfolio composition. The Journal of Finance 30 (2), 605-620.
Cox, P., & Schneider, M. (2010). Is Corporate social performance a criterion in the overseas investment strategy of U.S. pension plans? An empirical examination. Business & Society 49, 252–289
Dominitz, J., and C. F. Manski, 2007. Expected equity returns and portfolio choices: evidence from the health and retirement study. Journal of the European Economic Association 5 (2-3), 369–379.
Friedman, B., 1974. Risk aversion and the consumer choice of health insurance option. Review of Economics and Statistics 56, 209-214.
Grable, J.E., Lytton, R.H., 1998. Investor risk tolerance: testing the efficiency of demographics as differentiating and classifying factors. Financial Counsel and Planning 9, 61-74.
Hallahan, T., Faff, R., McKenzie, M., 2003. An exploratory investigation of the relation between risk tolerance scores and demographic characterstics. Journal of Multinational Financial Management 13, 482-502.
Henken, K., Van Dalen, H., 2015. De dubbelhartige pensioendeelnemer, over vertrouwen, keuzevrijheid en keuzes in pensioenopbouw. Netspar NEA Papers 58. 1-58.
Johnsonn, S.G., Schnatterly, K., Hill, A.D., 2013. Board composition beyond independence: Social capital, human capital and independence. Journal of Management 39, 232-262.
Honda, T., 2012. Dynamic optimal pension fund portfolios when risk preferences are heterogeneous among pension participants. International Review of Finance 12 (3), 329-355.
Kapteijn, A., Teppa, F., 2011. Subjective measures of risk aversion, fixed costs, and portfolio choice. Journal of Economic Psychology 32, 564-580.
Kool, C. J., Prast, H. M., Van Rooij, M. C., 2007. Risk-return preferences in the pension domain: Are people able to choose? Journal of Public Economics 91, 701-722.
Konicz, A.K., Mulvey, J.M., 2015. Optimal savings management for individuals with defined contribution pension plans. European Journal of Operational Research 243. 233-247.
Love, D.A., 2013. Optimal rules of thumb for consumption and portfolio choice. The Economic Journal 125, 923-961.
Neumann, J. Von, Morgenstern, O., 1944. Theory of Games and Economic Behavior. Princeton University Press, Princeton.
Morin, R.A., Suarez, F., 1983. Risk aversion revisited. Journal of Finance 38 (4), 1201-1216.
Ntim, C.G., Osei, K.A., 2011. The impact of corporate board meetings on corporate performance in South Africa, African Review of Economics and Finance 2, 83-103. Platt, H., Platt, M., 2012. Corporate board attributes and bankruptcy. Journal of Business Research 65, 1139-1143.
Pratt, J.W., 1964. Risk aversion in the small and in the large. Econometrica 32, 122-136.
Riley, W., & Chow, V., 1992. Asset allocation and individual risk aversion. Financial Analysts Journal 48 (6), 32-37.
Schooley, D.K., Worden, D.D., 1996. Risk aversion measures: comparing attitudes and asset allocation. Financial Services Review 5, 87-99.
Sebai, S., 2014. Further evidence on gender differences and their impact on risk aversion. Journal of Business Studies Quarterly, 308-319.
Sharpe, D. L., Wang, F., Yao, R., 2011. Decomposing the age effect on risk tolerance. The Journal of Socio-Economics 40, 879-887.
Sousa, R.M., 2015. What is the impact of wealth shocks on asset allocation? Quantitive Finance 15 (3), 493-508.
Vander Weyer, 2011. More women in the boardroom? Never mind the equality, watch the performance. The Spectator, 29.
