DC Defaults & Heterogeneous Preferences

Transcription

1 DC Defaults & Heterogeneous Preferences Jori Arts MSc Thesis

2 Master Thesis DC Defaults & Heterogeneous Preferences Author: Jori Gabriël Arts Supervisors: prof. dr. B. Melenberg (TiU) prof. dr. E.H.M. Ponds (APG) Second reader: prof. dr. B.J.M. Werker (TiU) A thesis submitted in partial fulfilment of the requirements for the degree of Master of Science in Finance Tilburg School of Economics and Management Netspar August 21, 2015

3 Abstract In this thesis we research the welfare effects of a uniform default asset allocation strategy for an individual defined contribution scheme, in coherence with the Dutch first and second pillar retirement benefits. Due to individualization, preferences between pension plan participants have become increasingly more heterogeneous, thus there is no evident solution to a uniform default that applies to all participants. We find that by accounting for additional sources of pension income, such as the Dutch first pillar state pension, heterogeneity in labor income leads to a wide variety of preferred asset allocation strategies. If a single uniform default were to be offered, participants may suffer substantial welfare losses. i

4 Acknowledgements I would like to express my sincere gratitude to my supervisors prof. dr. Bertrand Melenberg and prof. dr. Eduard Ponds for their insightful comments, feedback, and support. I wish to thank my colleagues and fellow interns at APG for the enjoyable time I ve had during my internship. In addition, I would like to thank all participants who attended my presentations at APG and the Junior Pension Day for their valuable comments. Finally, a thank you to my parents for their unconditional support throughout my studies. ii

5 Contents Abstract Acknowledgements Contents i ii iii 1 Introduction Background Problem formulation and research questions Relevance Research approach Structure Literature review Trend from defined benefit towards defined contribution Plan characteristics Employer-driven factors Employee-driven factors Beyond the reallocation of risk Implications of individual choice Saving Investment Annuitization The need for defaults Behavioral issues and bounded rationality Default design and heterogeneous preferences Model & methodology Labor income Pension income AOW Collective defined contribution Individual defined contribution iii

6 3.3 Welfare evaluation Scenario set Research design & simulations Summary statistics Default asset allocation strategies Heterogeneous participant profiles Wage levels Wage growth Results Preferred life cycle asset allocation strategies Replacement rate analysis Comparison of fixed, age rule, and life cycle asset allocation strategies Consumption loss of a default Sensitivity analysis Risk and time preference Life expectancy Subsistence level of consumption Conclusion Summary Discussion and future research References 60 iv

7 Chapter 1 Introduction 1.1 Background Over the past several years defined benefit plans have gradually shifted towards defined contribution plans in the occupational pension systems across the world. In some countries, the majority of assets in the private sector occupational pension plans are invested in DC plans. The shift in plan structure is mainly explained by two factors, employeeand employer-driven. On the one hand, the risk of pension underfunding that has been highlighted during the recent financial crisis and its persistence due to declining long-term interests have reduced the incentives for employers to offer DB plans. Traditional defined benefit plans are characterized by promised retirement benefits: promises that plan sponsors can no longer make. On the other hand, employees wish to have more choice and control over their pension, which an individual defined contribution plan would allow. The transition from DB to DC pension plans thus implies a shift in investment risk from plan sponsor to plan participant: a reallocation of risk within the financial system. Participants are therefore becoming increasingly exposed to financial market risk as their retirement income becomes dependent on investment performance rather than a promise, leading to greater variability in potential outcomes (Broadbent, Palumbo, & Woodman, 2006). Even though retirement income is subject to greater variability in DC plans, individual choice and control over one s pension are becoming increasingly more important, as voiced in the national pension dialogue organized by the Dutch government. Individualization of pension schemes introduces various aspects of choice, both in the accumulation and decumulation phase, such as risk profile, contribution and indexation policy, retirement age, and risk sharing. However, allowing such individual choices within the Dutch collective pension system may lead to adverse selection and strategic behavior which undermines the foundation of the collective system (Dellaert & Ponds, 2015). A (partial) shift from a collective to an individual system is required. Various prototypes of individualized schemes are present, of which two are most com- 1

8 mon: collective defined contribution (CDC) and individual defined contribution (IDC), the latter being the most individualized of sorts. Individual DC schemes are characterized by a personal retirement account with the highest form of choice complexity; it offers individuals the freedom to choose and to implement their own contribution and investment policy. In reality however, early experiences with DC plans have made evident that most participants choose highly suboptimal saving and investment strategies. For this reason, defaults have been introduced: contributions are invested according to a default life cycle asset allocation. Research has shown that most people simply follow the given defaults due to a variety of reasons such as financial illiteracy and passive choice (Choi, Laibson, Madrian, & Metrick, 2004; Benartzi, Peleg, & Thaler, 2007; Lusardi & Mitchell, 2007a; Beshears, Choi, Laibson, & Madrian, 2009). People often choose not to choose. There may be a case of omission-commission bias: an active choice results in more regret if things turn out wrong, than a passive choice recommended by experts, as argued by Potters & Prast (2009). In its recent study, the European Insurance and Occupational Pensions Authority (EIOPA, 2015) questions the offering of too many choices that members of occupational DC schemes have available. Participants often have no intention to deviate from a default investment strategy. EIOPA highlights the importance of supporting members in making adequate choices; they find the average participant lacks time, motivation, and financial knowledge to make optimal decisions. A default is often too generalized and not suitable for the majority of participants in a scheme, and in some countries the only mandatory option. The researchers therefore recommend national governments and supervisors to work on legislation to improve the relation between investment policy and participant characteristics; the Netherlands and U.K. are currently in the process of doing so. It can be concluded that in case of a complex choice architecture, most people follow a default life cycle asset allocation; however, what is the impact of such a default when taking into account heterogeneous preferences? As discussed above, most defined contribution pension plans employ a life cycle asset allocation strategy as their investment policy. Life cycle theory pertains to the consumption and savings behavior of individuals over their lifetime. Individuals generally face two main decisions in their financial planning over the life cycle: the consumption and investment decision, as outlined by Bovenberg, Koijen, Nijman, & Teulings (2007). Through the consumption decision, individuals decide their consumption smoothing pattern over time. Through the investment decision, individuals decide how to invest the pension premiums in a variety of financial assets so as to smooth consumption across various contingencies that may arise in the future. Theory suggests an age-dependent investment strategy, where younger participants invest a larger fraction of their financial wealth in risky assets than the elderly. This finding is based on the idea of human capital being a risk-free asset: the discounted value of future labor income. As a participant grows older, human capital is slowly depreciated, and as such should the fraction of financial capital invested in risky assets be reduced (Merton, 1971; Bodie, Merton, & Samuelson, 1992; 2

9 Campbell & Viceira, 2002). A default life cycle asset allocation allows for heterogeneity based on age between participants; besides age however, various other heterogeneous factors are present that are not accounted for. Main factors studied in this thesis are different possible career paths and wage levels of the individual. Other factors that can be thought of are flexible career paths like dismissal, illness or voluntary work interruptions, home equity and mortgage position, inheritance, and various lifestyle goals. These factors influence the outcome of the pension portfolio and play an important role in the preferred DC life cycle asset mix that are not taken into account when the plan s default is determined. Ideally, the DC pension should be a tailor-made product that invests premiums based on an individual s characteristics, assets and liabilities, and risk appetite over the lifetime. The lack of these factors in the default investment strategy might lead to welfare losses for a heterogeneous group of participants, which is precisely the aim of this thesis to research. 1.2 Problem formulation and research questions The main problem formulation of this thesis is: What is the welfare loss of a uniform DC default life cycle asset allocation strategy for a heterogeneous group of participants? In order to answer this question, the following research questions will be investigated: How do heterogeneous characteristics such as age, labor income, and risk appetite affect the preferred asset allocation strategy over the lifetime? How does the division of total pension income into first and second pillar retirement benefits affect the preferred asset allocation strategy over the lifetime? 1.3 Relevance With the possible introduction of individual components in the Dutch occupational pension system, this research will be highly relevant from a practical viewpoint. The findings of this research will be particularly useful for pension funds in determining the life cycle asset allocation strategy that participants will follow by default in an individual defined contribution scheme, if they do not make an active choice. In the coming years, pension funds will be tasked with designing an adequate default option that matches the background of the participant base, aiming to minimize the welfare loss in the aggregate. This research will offer insights into the relevant heterogeneous factors of participants and to what extent they have an effect on the preferred life cycle asset allocation of their individual DC pension. Pension funds can use these insights to design a default that aims to appeal to the majority of their participant group, which in currently existing DC 3

10 schemes is severely lacking. Apart from practical relevance, the research conducted in this thesis will be a valuable addition to previous studies on DC defaults for two main reasons. Firstly, we study defaults in the context of the Dutch pension system as a whole, taking into account both first and second pillar benefits, whereas previous studies have mainly focused on just individual schemes, such as 401(k) plans found in the U.S.. Secondly, heterogeneity plays a major role in this research. Due to individualization, preferences between participants have become increasingly more heterogeneous, thus there is no evident solution to an adequate default design. 1.4 Research approach In this thesis the extent to which individual characteristics of plan participants lead to different life cycle asset allocation strategies, in deviation from the default choice, will be analyzed. Firstly, a literature review on the topic of DC defaults and individual welfare is conducted. This analysis is supported by an empirical analysis researching the technical aspects of life cycle investment strategies and the associated welfare effects for individuals. In order to measure welfare effects between different types of life cycle asset allocation strategies, that is, following a type of default, the utility of individuals is researched. Individuals aim to maximize utility over the lifetime, which is the weighted sum of future expected utility at each point in time. Similarly to Bovenberg et al. (2007), it is assumed that utility at a point in time depends only on consumption at that time. Individual preferences feature positive constant relative risk aversion. This implies that individuals value additional consumption less at larger levels of consumption, in terms of utility; the negative elasticity of marginal utility with respect to the level of consumption is known as the coefficient of relative risk aversion, which is constant. Positive risk aversion implies that individuals prefer a smooth pattern of consumption over time or across contingencies; highly risk-averse individuals prefer a very stable stream of consumption. In order to perform an empirical analysis, data on the trajectory of labor income during the working life, and pension income during retirement is simulated for a Dutch pension plan participant. In this study, pension income consists of a total of three parts, split between two pillars in the Dutch pension system. In the first pillar, a participant receives a flat-rate state pension (AOW) related to minimum wages. In the second pillar, for one part a career average collective defined contribution pension is accrued, a type of DB pension with the exception that accrued rights can be both positively and negatively adjusted depending on the plan s funding ratio. For the other part, a participant accumulates assets in an individual defined contribution plan, for which an annuity will be purchased at retirement age. The aim is to evaluate various default designs for the individual plan for a heterogeneous group of participants. Firstly, the effect of heterogeneous factors such as wage growth and wage levels on the preferred default design will be researched. Secondly, a comparison will be made between the preferred default designs of heterogeneous partic- 4

11 ipants and a single uniform default applied to the entire group of participants, analyzing the welfare effects. Preferences will be captured by a constant relative risk aversion utility function and welfare effects will be computed by determining the certainty equivalent for each default design. 1.5 Structure This thesis is structured as follows. Chapter 2 provides a detailed literature review of prior studies on the topic of defined contribution pension schemes and default options. First, the international trend from defined benefit towards defined contribution schemes is explained by a number of factors. Second, the implications of individual choice are discussed; the suboptimal decisions individuals make if left with the freedom to choose their own savings and investment policy. The third section outlines the need for default options in defined contribution schemes with freedom of choice. Behavioral biases and bounded rationality that lead individuals to opt for a default are studied. The chapter is concluded with a discussion on the importance of matching the default design with the heterogeneous preferences of participants. Chapter 3 covers the model and methodology used in the empirical analysis. First, income streams during the accumulation and decumulation phase are described. Next, the criteria for welfare evaluation of the proposed default investment strategies are defined. Lastly, the risk model used to generate economic scenarios for the key variables of the analysis is specified. Chapter 4 outlines the research design of the thesis. In the first section, summary statistics of the simulations are presented. The subsequent section outlines the various types of default asset allocation strategies that are considered, such as fixed, age rule, and life cycle strategies. Finally, heterogeneous participant profiles are constructed for which the default strategies will be evaluated. Chapter 5 discusses the results of the empirical analysis. For each participant profile, the preferred default asset allocation strategy is derived; the findings are supported by an extensive analysis of replacement rates, which provides a greater insight into the effects. The subsequent sections focus on the individual welfare loss resulting from suboptimal choices by participants, and the heterogeneity amongst participants. Lastly, a sensitivity analysis is performed, and an extension of the utility framework is researched. In chapter 6, the thesis is concluded with a summary of the analysis, a discussion of assumptions and limitations of the thesis, and topics for future research. 5

12 Chapter 2 Literature review 2.1 Trend from defined benefit towards defined contribution Traditionally, defined benefit pension plans have made up a vast share of the occupational pension systems across the world; however over the past decades, defined contribution plans have gradually become more dominant in many countries and now account for the majority of invested assets in the private sector occupational pension plans (Broadbent et al., 2006). The shift in plan structure implies a reallocation of investment risk, and can be attributed to a number of factors, driven by both employers and employees Plan characteristics In defined benefit plans, participants are guaranteed to receive retirement benefits determined by years of service and wage history, either based on an employee s career average or final salary. The benefit is defined in the sense that the benefit formula is known in advance, and generally benefits must be paid regardless of the value of financial assets in the employer s pension plan, thus plan sponsors are tied to a guarantee. This type of plan stands in sharp contrast to defined contribution schemes where each employee has an account into which the employer and employee make regular contributions. Employees are often given a choice as to how the contributions are to be invested. In principle, contributions may be invested in a wide variety of assets, though most plans limit the choice of the asset mix to a conservative portfolio of stocks and bonds, varying in risk characteristics. Retirement benefits of the individual depend on the amount of contributions made and the accumulation of investment returns on the financial assets; the employee thus bears all investment risk. Valuation of both plans differ, as described by Bodie, Marcus, & Merton (1988): Whereas the DC framework focuses on the value of the assets currently endowing a retirement account, the DB plan focuses on the flow of benefits which the individual will receive upon retirement. (p. 141). 6

13 As Munnell, Muldoon, & Aubry (2008) outline, the transition from defined benefit to defined contribution implies a shift in financial market risk from plan sponsor to plan participant: a reallocation of risk within the financial system. Participants are therefore becoming increasingly exposed to financial markets as their retirement income becomes dependent on investment performance rather than a guarantee. The reallocation of risk will be discussed more extensively in the subsequent sections Employer-driven factors The shift from defined benefit to defined contribution plans can be mainly attributed to the wishes of employers to reduce the risk of pension underfunding. Defined benefit plans accumulate significant funding obligations for employers as employees earn entitlements to future retirement benefits as they work. As has been highlighted during the recent financial crisis, defined benefit plans can easily become underfunded because of a decline in the value of a fund s invested assets on the one hand, and an increase in the value of a fund s liabilities as a result of declining long-term interest rates on the other hand. Munnell et al. (2008) have researched the impact of the financial crisis on plan sponsors of state and local pension plans in the U.S., recording funding ratios of as low as 87 percent in 2007 and 65 percent in October 2008, if assets were valued at market. Since long-term interest rates continue to decline, in order to recover from such low funding ratios, funds crucially depend on the evolution of equities, exposing plan sponsors to considerable market risk. In that sense DB funds have had a doubly negative impact by the crisis, in that not only portfolio returns have been negative, but in many cases the near risk-free discount rates used for valuing liabilities have fallen, leading to an unexpected rise in obligations, as noted by Pino & Yermo (2010). Due to the large impact of the financial crisis, employers have come to realize that they are tied to guarantees that they can no longer make, nor are willing to make. This has led to an accelerated transition to defined contribution plans, where market risk is transferred to participants. Apart from the wish of employers to reallocate market risk to employees, the regulatory environment has also caused employers to opt for defined contribution plans. Particularly pension regulation concerning defined benefit funds has become increasingly complex, therefore costly to administer for the employer. An example is the U.S. Employee Retirement Income Security Act of 1974 (ERISA), a major pension reform, imposing minimum standards for almost every aspect of private pensions, such as funding, coverage, reporting, and disclosures (Clark, Munnell, & Orszag, 2006). Also in the U.K., as Davis (2004) describes, closure of most private DB schemes to new entrants was the consequence of an increase in the overall burden of regulation since the mid-1980s. By 2003, 80 percent of private DB funds had closed to new members, relative to 41 percent being closed in Though as Forman (1999) notes, there are economies of scale present in pension plans; it can be relatively inexpensive to administer a large DB plan. 7

14 In any event however, DC plans are relatively simple to administer and the concept of such plans, i.e., an individual savings account, is easier to communicate to participants Employee-driven factors Even though employees are faced with an increasing exposure to financial markets as a result of the transition to defined contribution plans, several employee-driven factors have accelerated this shift; DC pension plans do have positive aspects for employees. First and foremost, employees wish to have more freedom of choice and be in control of their pension, which defined contribution plans allow. More freedom of choice allows participants to design a pension product that is tailor-made to the individual s preferences. Due to individualization, preferences between participants have become increasingly more heterogeneous. This is most evident for human capital; flexible career paths and wage levels have resulted in a very diverse participant base. Yet in defined benefit schemes, all participants contributions are pooled and invested according to a strategic asset allocation strategy. Due to this one size fits all approach of DB schemes, freedom of choice is severely limited. On the contrary, defined contribution schemes allow for individual retirement accounts where the investment strategy can be tailored towards the individual. Even though DC schemes offer the individual the freedom to implement the optimal investment strategy according to their own preferences, there is one critical remark. In reality, individuals often make irrational decisions when freedom of choice and pensions are concerned (Munnell & Sunden, 2004). Apart from inadequate pension savings due to myopia, individuals generally make suboptimal investment decisions that do not contribute to the pension they pursue at retirement. Moreover, the average participant does not even have the motivation nor ambition to extensively research the accrual of his or her pension portfolio (van Els, van Rooij, & Schuit, 2006). Secondly, changes in industrial structure and labor force composition have given rise to an increasingly mobile workforce. Due to the nature of defined benefit plans, employees are often not able to transfer their accrued entitlements across employers, which penalizes mobile workers since benefits based on formulae using job tenure and earnings accrue value disproportionally in the later years of employment; benefit accruals increase significantly the closer an employee gets to retirement. This backloading of benefits implies that employees that frequently switch jobs give up a large portion of their potential pension benefits under DB plans (Gustman & Steinmeier, 1993). In a study conducted by Blake (2000), accrual losses from a DB plan for a typical U.K. worker that changed jobs at the average level of six times during their career amounted to 25 to 30 percent of the full benefit they would have received if remained with the same employer for an entire working life. In terms of portability and accrual, defined contribution plans are to be preferred, avoiding accrual losses and provide mobile workers with a more flexible way of saving for an adequate pension. 8

15 2.1.4 Beyond the reallocation of risk As has been mentioned before, a shift in plan structure from DB to DC implies a reallocation of market risk from employer to employee. In return, the employee receives more freedom of choice by the use of individual retirement accounts, and is not harmed by frequent job changes. Though beyond the transfer of market risk, what benefits of collectivity are foregone and what drawbacks are abandoned by opting for individual schemes? Collective pension plans allow for intergenerational risk sharing, in two ways. Firstly, systematic risks such as labor and financial market shocks can cause large swings in the value of one s pension. By smoothing such risks over multiple generations, the effect of adverse shocks on the current generation are limited, and thus welfare enhancing. Secondly, risks that are either not traded on financial markets or only in limited capacity, such as macro longevity risk and inflation risk, can also be smoothed instead of borne by one generation (van Ewijk, Lever, Bonenkamp, & Mehlkopf, 2014). Capital markets however, provide no means of sharing risks with future generations, the unborn. Collective pension funds, in their role as long-lasting financial institutions, can facilitate intergenerational smoothing of systematic risks with mandatory participation rules (Gollier, 2008; Cui, Jong, & Ponds, 2011). Thus, participants in individual pension plans bear the entire risk of economic shocks within their generation, whereas participants in collective pension plans share part of the risk with future generations. The effect on individual welfare is however ambiguous; on the one hand, intergenerational risk sharing improves welfare since losses are not entirely incurred on the current generation; on the other hand, collective pension contracts imply a suboptimal allocation of consumption across time and states of nature, and lack age-dependent policies (Westerhout, Bonenkamp, & Broer, 2014). Thus, collective pension plans can also reduce welfare. Such plans apply age-indepen- /dent policies in setting contributions and benefits, adjusting these parameters in response to adverse shocks without distinguishing between different generations, whereas individual plans apply age-dependent policies. Furthermore, policies on contribution and indexation are not matched in collective plans, as would be required by the principle of consumption smoothing, creating disparity between working and retired generations. In addition, individual plans apply age-dependent investment policies based on total wealth, reducing equity exposure as the participant grows older, whereas collective plans typically adopt investment strategies based on their financial wealth, independent of participant age. All in all, these imperfections of collectivity lead to a suboptimal allocation of consumption, reducing individual welfare. As Westerhout et al. (2014) find, collective pension schemes may be either superior or inferior to individual schemes, depending on the way the schemes are designed. Collective schemes could improve by introducing individual elements to their policies for investment, contribution and indexation. This would reduce the welfare losses of collectivity whilst reaping the benefits of intergenerational 9

16 risk sharing. Individualization of pension schemes introduces various aspects of choice, both in the accumulation and decumulation phase, as extensively discussed. However, allowing such individual choices within collective pension systems may lead to adverse selection and strategic behavior which undermines the foundation of the collective system (Dellaert & Ponds, 2015). A (partial) shift from a collective to an individual system is required. 2.2 Implications of individual choice The introduction of individual components in the occupational pension system leaves critical decisions concerning contributions and investments in the hands of employees (Bernheim, 1996). Individuals suffer substantial welfare losses if they lack the expertise to implement optimal savings and investment strategies in order to maximize their welfare (Bovenberg et al., 2007). In this section, we discuss the suboptimal decisions individuals make if left with the freedom to choose their own contribution and investment strategy in the accumulation phase, and the choice whether to annuitize one s wealth in the decumulation phase Saving For consumption to be smoothed over the life time, an adequate contribution policy should be determined by the individual. By optimally balancing between consumption and saving at each point in time over an individual s life, a suitable pension can be accumulated that smoothly transitions from consumption during the working life to consumption during retirement. In their study on 401(k) pension plans, Choi, Laibson, Madrian, & Metrick (2002) find that two-thirds of employees believe that they are saving too little and that one-third of these self-proclaimed undersavers intend to raise their rate of saving in the next two months. Employees who report that they save too little actually do have low saving rates. However, almost none of the employees that intend to raise their savings in the next two months actually do so. Procrastination is an important factor underlying this behavior (Madrian & Shea, 2001). People want to save for their retirement, but lack the capacity to carry out their intention to do so. There are several different reasons why individuals may procrastinate their saving decisions. It may be rational to stick with the status quo if the costs of gathering and evaluating the information needed to make a saving decision exceed the short-term benefits. The costs associated with decision making are especially high for financially illiterate individuals. At least two sources of complexity are involved in making an optimal saving decision. Firstly, employees must choose the fraction of salary to contribute to their pension and how that contribution should be allocated between available funds. An additional source of complexity, for some employees, is how to evaluate the many different 10

17 saving options, they may simply not know the potential outcomes of their decisions. Increasing the complexity of a decision making task leads to procrastination (Tversky & Shafir, 1992; Shafir, Simonson, & Tversky, 1993). Thus, individuals rather stick to their initial savings rate than to alter it, which is highly suboptimal. In order to stimulate adequate retirement saving, Thaler & Benartzi (2004) propose an automatic savings program, Save More Tomorrow (SMarT), committing people in advance to allocate a portion of their future salary increases toward retirement savings. The results from its first implementation suggest that a vast majority remained in the program, with an average savings rate increase from 3.5 percent to 13.6 percent over the course of 40 months. These findings highlight the great importance of default schemes in individual pension plans, which will be extensively discussed later Investment Apart from decisions related to the contribution policy, employees are tasked with a second major decision in individual schemes, namely how the contributions are to be invested, the investment policy. Literature on asset allocation indicates that individuals seem to make significant mistakes. Firstly, individuals make decisions that seem to be based on naive notions of diversification. Participants of 401(k) plans display a tendency to split their contributions evenly amongst available funds or portfolios, irrespective of the type of investment that is offered, the 1/n heuristic. The proportion of the assets participants invest in stocks depends strongly on the proportion of stock funds in the plan. Participants thus use naive diversification strategies that are heavily influenced by the menu offered by their plan. Using a database of 170 retirement savings plans, Benartzi & Thaler (2001) find that approximately 62 percent of funds offered are equity investments and that the fraction of assets held by participants in these 170 plans is exceptionally close to 62 percent as well. This raises high concerns for plan designers: what is the right mix of equity and fixedincome funds to offer? If the plan offers many equity funds the employees might invest too aggressively, which is recommended for young workers, but not for older ones. Then, should the plan offer different age-dependent funds? Another case of naive diversification is found by Choi et al. (2002), that of company stock. Approximately half of all 401(k) plans, by assets, offer employees the opportunity to invest in company stock. Some employers even require that the matching contribution is to be held in the company s own stock, at least for a period of time. Employees are generally tempted to invest their discretionary contributions in company stock due to a familiarity bias, not realizing the risk of such an investment. Since company stock is highly volatile due to idiosyncratic risk, risk that is firm-specific and theoretically not compensated in terms of return, and highly correlated with the labor income of the employee, allocating contributions in company stock is a very poor diversification strategy. Apart from naive diversification strategies, 11

18 individuals also tend to naively extrapolate past returns of company stock. Benartzi (2001) finds that current contributions to company stock are greatly influenced by the returns booked by that stock over the preceding 10 years, even though past returns are uncorrelated with future returns. Additional evidence is supported by Patel, Zeckhauser, & Hendricks (1991), reporting that purchases of mutual funds are unduly influenced by recent good performance, even though persistence of performance is absent. Secondly, individuals that do actively manage their pension investments tend to mistime their equity exposure, found in a study by Benartzi & Thaler (2007). Throughout the 1990s, participants were continuously increasing their allocation to equity in anticipation of a large upswing in equity markets, both in terms of contributions and account balances held. In their study, the researchers calculate the mean allocations to equities from 1992 through 2002 for new participants, and find an allocation of 58 percent of assets in 1992, and a staggering 74 percent in 2000 right before the collapse of the dot-com bubble. The next two years, equity allocations fell back to 54 percent. Effectively, participants had tremendously increased their equity exposure as markets peaked and reduced their equity exposure as markets were at their lows; the market timing of participants was exactly wrong. Clearly, the average participant does not have superior information to realize returns in excess of the market, even though the participant might think so due to overconfidence in their own abilities and estimates. On the contrary, the vast majority of individuals does not actively manage their pension, but rather sets an initial allocation and leaves it unchanged for longer periods of time, as concluded by Agnew, Balduzzi, & Sunden (2003): 87 percent of the annual number of trades in the studied panel are zero and only 7 percent of the observations exceed one, over a four year period. Thirdly, individuals tend to treat accumulated assets differently from future contributions, what is called mental accounting. Mental accounting refers to the implicit methods individuals use to code and evaluate investments and other financial outcomes (Kahneman & Tversky, 1984; Thaler, 1985). In retirement savings plans, participants are inclined to use separate mental accounts for assets they have already accumulated in the plan, and for amounts they have not yet contributed. The propensity to adjust the allocation of accumulated assets is much lower than that of future contributions. An explanation for this phenomenon is posited by Benartzi & Thaler (2007). Participants may fear the potential regret of reallocating their assets and observing the new investment decision underperforming the original allocation. Concerning future contributions, however, a reference point has not been set so there is less potential to experience any regret due to changes. This type of behavior leads to suboptimal investment, in the sense that financial wealth is split into two components and invested separately, which contrasts sharply with the traditional life cycle model which treats the labelling of wealth as irrelevant since wealth is regarded as mutually interchangeable in a perfect capital market. That is, the marginal propensity to consume should be the same for both mental accounts (Shefrin & Thaler, 1988). Apart from investing accumulated assets and future contributions differ- 12

19 ently, Ameriks & Zeldes (2004) also find evidence that participants adjust the allocation of their assets and contributions in different ways; only 27 percent of the participants studied reallocated their accumulated assets, whereas 53 percent reallocated their future contributions over the sample period Annuitization Some retirement savings plans, like the 401(k) plans in the U.S., allow participants to either take out their accumulated pension assets as a lump-sum payment or to purchase an annuity. A single life annuity is an insurance product that pays out a periodic amount with a first payment in the year in which the holder reaches retirement age, for as long as the holder is alive. The main appeal of an annuity is that it offers retirees the opportunity to insure against micro-longevity risk, the risk for an individual to get older than the average life expectancy, resulting in outliving one s assets (Brown, Mitchell, Poterba, & Warshawsky, 2001). Another risk associated with lump-sum pensions is underconsumption risk, the risk that one will die with too much wealth left unconsumed. In absence of bequest motives, underconsumption represents a risk since unconsumed wealth is a foregone consumption opportunity from which one derives utility. Since annuities pay a guaranteed stream of income until the holder s death, one is sure to fully consume over the remaining life time. In addition to the insurance role of annuities, annuities pay a rate of return to survivors, net of administration costs, that is greater than the return on conventional assets of matching financial risk (Davidoff, Brown, & Diamond, 2005). This mortality credit is the result from the redistribution of assets of those who die early to those that are still alive. Considering the fact that annuitizing one s wealth insures against longevity and underconsumption risk whilst providing a higher return than conventional assets, why do individuals make the suboptimal decision not to purchase an annuity? Individuals can achieve substantial welfare gains if they eliminate uncertainty around their retired life, yet the overall annuity market is very thin and of limited size (Inkmann, Lopes, & Michaelides, 2007). In academia, this is known as the annuity puzzle. We will briefly discuss some underlying factors influencing the behavior of participants reaching retirement age. First of all, the major asset in the portfolio of most retired individuals is social security which is in itself an annuity (Mitchell & Moore, 1998). Since social security, think of the AOW in the Dutch first pillar, provides a minimum level of income for the entirety of an individual s life, retirees are already somewhat protected against longevity risk, therefore the benefits of additional annuitization in the second pillar will be reduced. By taking a lump-sum pension, individuals have the choice to consume their wealth freely, and even participate in equity markets post-retirement age to potentially increase their pension. Additionally, annuity prices may be unattractive due to price deviations from actuarially fair value, driven by adverse selection of participants. Annuity providers need to adjust 13

20 prices to reflect the fact that individuals who choose to annuitize their wealth generally have a longer life expectancy than the average individual on which the fair value is based. In their pricing analysis, Mitchell, Poterba, Warshawksy, & Brown (1999) have found that while administrative costs account for a three to five percent reduction in annuity payouts, adverse selection is responsible for reduction of eight to twelve percent. The argument of adverse selection naturally explains the argument of health uncertainty; as an individual acquires additional information about their expected remaining lifetime as they grow older, one may refrain from purchasing an annuity if the individual faces significant uncertainty about future health care costs (Brown & Warshawsky, 2004). Finally, if retirees have a strong wish to leave a bequest, full annuitization is suboptimal; under standard assumptions, purchasing an annuity with a value of only 60 percent of ones wealth at retirement age would be optimal (Yaari, 1965; Walliser, 2001). 2.3 The need for defaults Individual defined contribution pension plans offer participants the freedom to choose and implement a saving and investment strategy that is tailored to individual preferences. Participants decide how much to save and how to allocate their savings over different funds in order to maximize their lifetime utility, depending on their age, risk aversion, and human capital, amongst others. In reality however, most people do not choose optimal saving and investment strategies, as discussed extensively in the previous section. For this reason, default options have been designed for participants that lack the capacity or willingness to make such decisions about their pension. In this section, we discuss the behavioral biases and bounded rationality that lead participants to opt for a default, and study the impact of heterogeneous preferences on the design of default options Behavioral issues and bounded rationality Beshears, Choi, Laibson, & Madrian (2008) differentiate between revealed preferences tastes that rationalize an individual s observed actions and normative preferences representing an individual s actual interests. In many cases, the assumption that both are identical is violated due to behavioral biases and bounded rationality. In the context of pensions, the default effect will be explained. Due to a complex choice architecture embedded in individual schemes, most people follow the given default life cycle asset allocation, they choose not to choose. Iyengar, Jiang, & Kamenica (2006) find that the presence of more rather than fewer options makes decision makers more likely to decide against choosing, even when the choice of opting out of a default has negative consequences for their well-being. That is, the presence of more choice is associated with lower chooser confidence. Schwarz & Kliban (2004) conclude that when individuals are faced with too much choice, they are discouraged 14

21 from choosing anything, or, if forced to choose, simply pick something based on rules of thumb, which in turn leads to suboptimal outcomes. The fact that people tend to avoid complex choices has to do with anticipated regret. An active choice results in more regret if the outcome is negative, than a passive choice (Potters & Prast, 2009). As Choi et al. (2002) state: At any point in time employees are likely to do whatever requires the least current effort: employees often follow the path of least resistance. (p. 70). Apart from the dilemma of complex choice in individual schemes, often participants simply do not wish to choose, the path of least resistance implies a passive choice. Bodie & Prast (2011) offer some insight in the matter. Participants may erroneously believe that a default is the recommended choice by plan designers, or that they may interpret the default as a social norm approved of by others. In any case, deviating from the default requires an act of substantial effort on behalf of the participants, one that is often procrastinated. Procrastination is especially likely to arise when the marginal cost of delaying is small, leading to large delays and high cumulative costs (Madrian & Shea, 2001). By delaying active choice, participants remain with the default option. Even if participants wish to make well-thought-out decisions, many simply do not understand the mechanics of retirement saving. In a study by Lusardi & Mitchell (2007b), it is found that financial illiteracy is widespread: young and older people appear to be underinformed about financial concepts like retirement planning, mortgages, and other decisions. Being unfamiliar with even the most basic economic concepts to make sensible saving and investment decisions has serious implications for building up an adequate pension. Under a complex choice architecture, participants need financial education to understand the outcomes of their decisions, which can be costly and infeasible on a large scale. Defaults provide a way for plan designers to avoid costs associated with pension education and communication by acting as a paternalistic institution, protecting the participants against behavioral biases. We can conclude that there is a discrepancy between participants preferences for increased choice and their actual reactions to the provision of choice. Although freedom of choice is perceived as a desirable feature in individual pension schemes, in most cases the provision of choice either inhibits the likelihood to make a choice or greatly affects one s well-being after a choice is made (Botti & Iyengar, 2006). As a consequence of this behavior, many participants opt for the default offered by the scheme, which in itself defeats the purpose of a self-directed investment account; the paradox of choice Default design and heterogeneous preferences Given that the vast majority of participants in individual defined contribution schemes remain with the default option (Choi et al., 2002; Beshears et al., 2009), it is of utmost importance that the strategy of the default reflects the preferences of the participant base. In academia, the consensus is that a life cycle strategy is desirable: an age-dependent 15

22 investment strategy, where younger participants invest a larger fraction of their financial capital in risky assets than the elderly, based on the idea of human capital being a risk-free asset: the discounted value of future labor earnings. As a participant grows older, human capital is slowly depreciated, and as such should the fraction of financial capital invested in risky assets be reduced (Merton, 1971; Bodie et al., 1992; Campbell & Viceira, 2002). Such a default design allows for heterogeneity based on age between participants, besides age however, various other heterogeneous preferences are present that influence the optimal saving and investment decision. Participants differ in their human capital by wage level and growth, which in turn determines the relative importance of the first pillar pension benefit. Apart from that, people differ in their preferences for risk taking, as measured by risk aversion level; a highly risk-averse participant may not want to allocate too high of a fraction of their contributions to risky investments. Besides these more observable characteristics, one can think of home equity and mortgage position amongst other sources of personal wealth. In an ideal setting, the individual DC pension should be a tailor-made retirement account that invests contributions optimally according to the participant s preferences; the default should be designed to take all these preferences of the participants into account. In reality however, the defaults being offered are often too generalized and not suitable for the majority of participants due to increasing heterogeneity between participants. In a recent report, EIOPA (2015) highlights the importance of matching the preferences of participants with the investment policy of defaults in defined contribution plans. The researchers therefore recommend national governments and supervisors to work on legislation to improve the relation between investment policy and participant characteristics; the Netherlands and U.K. are currently in the process of doing so. These developments pressure pension funds to design adequate default options that match with the background of the participant base, a difficult task with a tremendous impact on the economic welfare of participants. In the subsequent empirical analysis, we will explore individual defined contribution default designs in coherence with first and second pillar collective benefits for the Dutch case, by identifying how heterogeneous factors affect preferred life cycle allocations and analyzing the welfare effects of following a uniform default set by the pension fund versus the preferred choice. 16