The Journal of Public Policy

The Implications of Behavioral Economics for Pension Plan Design


By David Hollanders, University of Amsterdam

For all their varieties, a common feature of pension reforms in the last decades has been the shifting of risks from employers to employees. Closely accompanying this shift is a discussion of the desirability to shift control to employees as well, that is, to increase individual choice in pension plans at the possible expense of collective solidarity and risk sharing. This discussion would better be informed by behavioral economic research that has focused on pension plan participants’ choice behavior. Three main conclusions are presented here. Together, they suggest that a majority of people are unable or unmotivated, or both, to choose. The implication for pension plan design is that a well-designed default option should be available.


Often people react to a choice-problem differently depending on how it is presented. This is called a framing effect if non-essential changes in presentation—not affecting the underlying choice in substance—leads to substantial differences in outcomes. Recent empirical research has focused on framing effects in the pension domain. It could be argued a priori that framing effects do not occur in those domains that involve large monetary stakes and that consequently ‘penalize’ framing effects dearly. This assumption is however not confirmed by recent research. The main outcome, on the contrary, is that framing effects abound in the pension domain.

Many people avoid choices that are (perceived as) extreme. This leads to framing effects, as can be illustrated with a simple experiment taken from Benartzi and Thaler (2002). Consider portfolios A, B and C, denoted in the left-hand side of Table 1. The portfolios (only) differ with respect to the fraction invested in equity (considered risky) and bonds (relatively safe). Portfolio A is thus the riskiest portfolio and portfolio C the least risky portfolio. Participants of the experiment were asked to order portfolios A, B and C according to their preferences. A large majority of 70.8% of participants had a preference order such that portfolio B was preferred over portfolio C.

Table 1             

Portfolio A: 80% bonds, 20% sharesPortfolio B: 60% bonds, 40% shares

Portfolio C: 40% bonds, 60% shares

Portfolio B: 60% bonds, 40% sharesPortfolio C: 40% bonds, 60% shares

Portfolio D: 20% bonds, 80% shares

In another round, (other) participants were asked to order portfolios B, C and D. These results are presented in the right-hand side of Table 1. Compared to the earlier choice-problem, portfolios B and C are unaltered, portfolio A is removed and portfolio D is added. Note that, in the ordering on the left-hand side, portfolio B is the middle choice (in terms of riskiness and in terms of presentation), and, in the ordering of the right-hand side, portfolio C is the middle choice. Consistency subscribes that people do not change their preferences between B and C when alternative options vary.

Presented with this changed setup, a minority of 46.2% prefers B over C, however. This corresponds with an absolute decrease of 24.6 percentage points and a relative decrease of 34.7% compared to the first presentation; these changes are substantial and statistically significant. So a non-essential change in presentation led to a reversal in preferences. This outcome can be interpreted as extremeness aversion, as the portfolio that is presented as the middle-option is preferred by most participants, independent of its exact features. Teppa and van Rooij (2012) show similar framing effects.


Financial knowledge of pension fund participants has been the topic foci of recent empirical research. The main outcome of this literature shows that a majority of people have no—or at most limited—financial knowledge. For example, Lusardi and van Rooij (2010) evaluate financial knowledge with a questionnaire containing eleven questions. Two examples follow (all questions in Lusardi and van Rooij (2010), page 31):

1) Buying a company fund usually provides a safer return than a stock mutual fund. True or false?

2) If the interest rate falls, what should happen to bond prices: rise/fall/stay the same/none of the above?

A minority of 5% of respondents answered all questions correct, whereas 7.6% did not answer a single question correct. Almost a quarter of participants (24.3%) answered at most three or fewer questions correct, and 52.5% answered at most six questions or fewer correct. A quarter of respondents (25.4%) answered at least nine questions correct. Lusardi and van Rooij (2010) conclude: “Dutch households lack sufficient knowledge to make sound portfolio decisions.”


Economic choices are sometimes postponed or avoided all together. The pension domain is not an exception. A well-documented example is given by an experience in Sweden (see Cronqvist and Thaler, 2004). After a pension reform in 2002, citizens were to invest part of their pension assets themselves. They had to choose from a list of preselected funds containing 456 funds. After an intensive campaign, 66.9% of participants chose a fund different from the default fund. (Note: No distinction could be made between not choosing and choosing the default fund.) The percentage of (entering) participants choosing actively dropped to 8.4% a year later in 2003. Together, these findings suggest that (i) only an intense campaign—like that held in 2002—can induce a majority of people to choose, (ii) absent a campaign, a large majority does not choose, (iii) with a campaign, a substantial minority still does not choose. Furthermore, the selected portfolios tended to be relatively risky, costly and performed worse ex post.

The Swedish experience is consistent with other findings, as reported by van Rooij et al. (2007) and Benartzi and Thaler (2007). The former authors report, for example, a result found by Ameriks and Zeldes (2004) that about 50% of investors did not change their portfolios over a period of ten years (and 14% made one change).



Together, the findings suggest that a majority of people are not able to choose, do not actually choose when given the choice or make irrational (i.e., inconsistent) choices. And, that leaves out other possible biases as anchoring effects, choice overload and herd behavior. Added to that, education and information do not seem to adequately address the issues. Willis (2011) proposes that education will be marginally helpful at best as a result of the low current knowledge-level, the complexity of financial decisions, the speed of change of financial products, the lack of interest among consumers, and the ability of the financial industry to outmaneuver education. Willis concludes that effective financial education should be (but hardly can be) “extensive, intensive, frequent, mandatory, and provided at the point of decision-making, in a one-on-one setting, with the content personalized for each consumer.” Bodie and Prast (2011) state: “information and education seem at best to influence intensions only.”

This has two implications. First, it does not follow that there should not be any choice; in fact, there is on all accounts a non-neglectable minority that wants to. Van Rooij at al. (2011) document that, in 2010, 23% of a representative Dutch survey wanted to invest their own pension assets (whereas 63% did not want to do so, and 14% “did not know” if they wished to). Second, a carefully designed default option should be available in every pension plan. At the end of the day, there are too many people who choose not to choose. And, given the complexity of pensions and the resulting search costs, that does not even need to be a bad choice.

This post is based on research for the research project Freedom and Pension reforms conducted at the Amsterdam Institute for Advanced labour Studies (AIAS), University of Amsterdam.



Benartzi, S. and R.H. Thaler (2007), Heuristics and biases in Retirement Savings Behavior, Journal of Economic Perspectives, 21(3), pp. 81-104.

Benartzi, S. and R.H. Thaler (2002), How much Is Investor Autonomy Worth?, Journal of Finance, 57, 1593-1616.

Bodie, Z.and H. Prast (2011), Rational pensions for irrational people, Netspar discussion paper, 76.

Cronqvist, H. and R.H. Thaler (2004), Design Choices in Privatized Social-Security Systems: Learning from the Swedish Experience, American Economic Review 94(2), pp. 424-428.

Kooreman, P. and H. Prast (2010), What does behavioral economics mean for policy? Challenges to savings and health policies in the Netherlands, de Economist 158(2), pp. 101-122.

Lusardi, A. and M. van Rooij (2010), Financial Literacy: Evidence and Implications for Consumer Education, Netspar Panel Paper 16.

Prast, H., F. Teppa and A. Smits (2012), Is information overrated? Evidence from the Pension Domain, DNB Working paper No. 360/December 2012.

Rooij, M.C.J. van, C.J.M. Kool and H.M. Prast (2007), Risk-return preferences in the pension domain: are people able to choose?, Journal of Public Economics 91, pp. 701-722.

Rooij, M. van, H. Prast and A. Smits (2011), Gedragsreacties van deelnemers op nieuwe contracten, ESB, 96(4625S), pp. 60-65.

Teppa, F. and M. van Rooij (2012), Are Retirement Decisions Vulnerable to Framing Effects? Empirical Evidence from NL and the US, DNB Working paper No. 366.

Willis, L.E. (2011), The Financial Education Fallacy, American Economic Review, 101(3), 429-434.