Humankind has made huge strides in technology. Big data and data analytics, artificial intelligence, machine learning and deep learning are being used in various daily applications and industries. But when it comes to retirement schemes, we appear to be stuck in the past.
The leading solution that private retirement planners have come up with is a spectrum of well-diversified risk/return investment portfolios drawn from 1950s financial technology, or at best a series of target date funds, where risk-taking follows a predefined ‘glidepath’. A recent innovation has been the enabling of the same solutions at lower cost using technology, or robo advisers. Surely we can do better.
Asia is unique in that many countries in this region have mandatory national retirement schemes. Herein lies the benefit. A wide cross-section of society participating in a national scheme not only brings the power of large-scale asset accumulation as a tour de force when negotiating in the public interest, it also provides the ability to risk-pool across retirement cohorts. But herein also lies the problem. The financial industry’s definition of cohorts is narrow, usually using risk tolerance, age or time to retirement as parameters.
Let’s visit the basics. There are two phases during one’s lifecycle: the accumulation phase, which is before retirement, and the decumulation phase, which is in retirement.
The priority in the accumulation phase should be on accumulating savings so as to target the income needed in retirement, with a focus on growth in future annuity value. When approaching retirement and post-retirement, the lifecycle risks are about mitigating uncertainty around the affordable, in-retirement income so that the accumulated savings can support the rest of the retiree’s life.
Consequently, the retirement fund’s target income investment strategy during the accumulation phase should be consistent with the way a variable deferred life annuity product is managed. It would be a bespoke function of one’s age, correlation of future wage income to the returns of all assets in place, personal profile, habits, and risk and loss tolerance, among other things.
One’s target income in retirement or during the decumulation phase determines the type of life annuity product generated. Mitigating uncertainty around future income generation is therefore an important consideration.
It should be obvious that both phases of the lifecycle are simply joint saving, investing and planning decisions linked by the same target life annuity product – quite unlike the way financial planners think or provide solutions for retirement currently.
These ideas are not new. Nobel laureate Robert C. Merton wrote the theory behind lifecycle finance in the 1980s and articulated it for finance practitioners in his article, “Thoughts on the Future: Theory and Practice in Investment Management”, which was published in the Financial Analysts Journal in 2003.
Traditionally, the defined-benefit (DB) plan of a corporation or government would have taken on the responsibility of providing employees a set of life annuity payments upon retirement, which is usually a function of their years of service, age and last-drawn salary. In theory, it’s a sound principal that allows retirees to maintain some semblance of pre-retirement purchasing power, even during retirement.
However, due to lack of focus, operating budgets and investment expertise, DB plans ended up making promises they couldn’t keep. Now the decision-making responsibility has shifted to the individual through defined-contribution (DC) plans. This means the burden of retirement planning falls on those who are not savvy in such matters. It’s like asking patients to perform brain surgery on themselves while fully awake.
The industry needs to be awakened from its fee-induced coma. Why not establish the necessary infrastructure for big data analytics, deep learning and investment science to create a target income-focused retirement plan that is personalised, bespoke or customised over one’s lifecycle that is also low cost from its full use of technology?
As a start, those in the accumulation phase of their lifecycle should be stratified into financial economics-driven representative cohorts that transcend age and risk tolerance so that appropriate target income investment profiles can be created for each cohort while taking advantage of risk-pooling arrangements. Representative retirement cohorts could be created using:
A mathematical optimisation method called stochastic dynamic programming can then be used to maximise monthly (future) target income at the point of retirement. Given the discussion so far, this approach is more realistic and useful, as compared to mean-variance optimisation that maximises risk-adjusted end of period wealth, which our industry is fond of using.
Consider the following optimal portfolio risk allocation simulation, which a well-trained financial engineer can perform using the backward induction method from dynamic programming. For convenience, we model a median Singaporean degree-holder from age 25 to 65, and assume he saves 10% of his gross monthly salary in his private retirement fund, on top of his nationally mandated Central Provident Fund (CPF) contribution. The simulation assumes no management or transaction fees.
The blue bands in the chart illustrate the potential deferred annuity income values in retirement, commencing at age 65, from his private retirement fund. The median outcome of an immediate annuity at age 65 corresponds to approximately S$3,300 (US$2,369) of inflation-adjusted retirement income per month for 20 years – assuming a life expectancy of 85 years – which isn’t that bad an outcome. This would be on top of his CPF LIFE annuity payout, which currently pays between S$600 to S$2,100 per month for life to Singaporeans, depending on one’s CPF balance at retirement and choice of LIFE annuity payout.
The red bands illustrate the retirement portfolio’s optimal risk allocation to risky and riskless assets; some may find this graph similar to the glidepath from target date funds. The risky asset is represented by a broadly diversified global equity index, i.e., MSCI World, and the riskless asset is represented by a laddered portfolio of inflation-protected government bonds, with a duration profile matching the retirement income liabilities. Note that both outcomes and risk allocations have uncertainty – or confidence interval – bands associated with them. This is due to varying market returns, and the fact that our simulation model dynamically adjusts for that variation.
Participants in retirement plans, on the other hand, should not be bothered with the complex financial engineering involved to get the above results, be it dynamic programming, Bellman equations or backward induction; rather, they should care more about the outcomes.
If the purpose of saving and investing for retirement is to maintain an adequate standard of living in retirement as one ages, which includes meeting healthcare costs, the investment strategy should focus on achieving an inflation-protected target income stream for life.
In summary, the next generation of shockproofed and timeproofed retirement solutions must aim to help participants achieve a unique stream of target retirement income that is sufficient to maintain their standard of living, meet in-retirement healthcare costs, protected from the ravages of inflation and designed to last throughout their retirement years. It should also minimise the likelihood of the financial retirement industry not meeting those three fundamental objectives.
Given the complexity in the big data analytics, artificial intelligence and financial engineering behind the design and implementation of such bespoke retirement products at the national level, it would require the best of the investment management industry, academia and policymakers with the right mindset to come together in a public-private partnership format to structure the right solutions for our retirement. Investors deserve nothing less.
* Joseph Cherian is practice professor of finance at the National University of Singapore Business School. Ong Shien Jin is professor of practice at the Asia School of Business in Malaysia.
Our thanks to Lim Chuan Poh, former chairman of A*STAR, for first proposing and discussing this idea with us.