Optimal Pension Design for Heterogeneous Individuals with Behavioural Biases
Freedom of choice and flexibility in pension plans may improve welfare, because it allows individuals to pursue welfare by means of their fitted plan: high or low retirement savings, risky or safe retirement investments, an annuity or (partly) a lump sum. But, individuals are not always acting rational and this may induce ‘wrong’ choices and utility losses. Present bias, for example, may lead to individuals retiring too early, insufficient retirement savings and too large lump sums compared to annuitization. Furthermore, according to prospect theory, people overreact to small probability events, and underreact to large probability events (this bias is called probability weighting). By estimating preference parameters this project results in a better understanding of optimal saving behavior and optimal retirement planning for heterogeneous agents in the Canadian population. In combination with an examination of present bias and probability weighting, this leads to valuable support for pension design and choice architecture.
The expected outcomes have direct applications for the pension industry. Firstly, practitioners learn about the distribution of risk aversion and time preferences, while behavioral biases are taken into account. This is important for the design and evaluation of ‘optimal’ pensions. Second, the distribution of present-bias and the length of present bias has direct implications for the design of choice architecture (if possible even for different groups of individuals). What would be optimal choice boundaries and what is the optimal timing of offering choices (without people being prone to present-bias). Thirdly, our measurement of probability weighting has direct implications for the construction of portfolios.
We propose as methodology a survey experiment among a sample of the Canadian population that includes money-earlier-or-later questions based on multiple time-delays to identify three issues. Firstly, we identify the degree of present bias in the Canadian population. Secondly, we identify the duration of the “present” with the aid of multiple time-delays – this is a gap in the scientific literature and leads to innovative practical advises for effective choice architecture. Thirdly, we take probability weighting into account by including questions with low risk conditions and high-risk conditions.
This research provides insights into the distribution and stability of risk aversion and time preferences, where behavioral biases are taken into account. The results will also show the size, variation, stability and length of the present-bias, and the bias that people have in the interpretation of probabilities. Together they are important for the design of pension plans and choice architecture.
Marike Knoef is Professor of Empirical Micro-Economics at the Department of Economics of Leiden University.