Defaults have become increasingly popular, especially among regulators, who have realised that they can counteract behavioural biases to improve both individual and societal outcomes. However, to use defaults effectively we need to understand the conditions in which they are appropriate and what constitutes a good default.
When Sunstein and Thaler released their book Nudge, regulators around the world sat up and listened. Nudge addressed the issue of how to get individuals to make the right decisions about issues that have significant long-term impacts for their lives. Their suggestions about applying choice architecture and libertarian paternalism provided some much-needed guidance. In fact, in the United Kingdom, regulators went so far recently as to set up their own ‘Nudge Unit’. Locally, National Treasury in particular, is exploring the possibility of using default solutions to address everything from preservation to annuity selection and investment strategies.
Defaults first became popular when defined contribution (DC) funds cameinto existence. Many early defaults simply translated accepted conventions (like pensionable pay percentages) directly from defined benefit (DB) schemes where individual members had limited choices because the benefits design was set by the employer and trustees.
Originally, defaults existed primarily as a back-stop for when individuals didn’t make their own decisions. As such, in those early days, many trustees simply opted for the most conservative options on the grounds that these would appear the safest and least controversial. It would take some time to understand the extent of the unintended consequences of such 'safe' decisions.
Defaults help individuals to make the right decisions about issues that have significant long-term impacts for their lives.
Gradually trustees and regulators became more circumspect about what would constitute a responsible default option. But the real breakthrough in thinking about the power of the default model only came about when academics like Sunstein and Thaler started to introduce behavioural economics principles into their consideration of default structures. This has led to the creation of an entirely different beast. Far from being a stop-gap for when people fail to make decisions, defaults are now understood to provide a structure for steering them to the right ones. Essentially these defaults rely on research that individuals’ observed choices are less about their preferences and more about how choices are presented to them. In many cases, choices can appear so overwhelming that individuals simply retreat from them.
With these kinds of defaults, the idea is to distil the best advice from experts into a series of pre-assigned choices that maximise the chances of achieving a particular outcome. Individuals can still make different choices if they want, but if they prefer not to engage, don’t have any strong preferences or would rather trust an expert to make their decision for them, they can follow the default which has now been properly thought out.
One example of this new approach to default construction can be seen in national programmes for organ donation. In many countries, such as South Africa, if you die, you only become an organ donor if you have actively chosen this option in advance. But in other countries, including Austria, Belgium and Sweden, anyone who dies is automatically assumed to be an organ donor unless they elect to opt out of that default. The graph below shows the dramatic difference in outcomes of these two different default structures.
What accounts for such a large disparity in results? It’s unlikely that moral preferences are this divergent. Instead, what this illustrates is how dramatically defaults influence choice.
Default structures seem to influence choice in three important ways:
So, defaults can be very powerful and can be used to significantly improve individual well-being and societal outcomes. But they are not always applicable. We need to appreciate when to use them and how.
Defaults can be very powerful and can be used to significantly improve individual well-being and societal outcomes.
Defaults work best when we have a clear idea of what decision is best for everyone2. For instance, in organ donation, we know that everyone benefits if everyone is a donor. But not all decisions are like this. People are different and when those differences change the optimal outcomes, then one-size-fits-all defaults don’t work.
Typically, defaults are also used in circumstances where experts have more insight into the correct decision than individuals3. For some decisions, individuals have strong preferences and know enough to be able to better identify the best solution for them than an expert would. For instance, when you choose where to live, you are normally the best person to identify what will work for you.
Decisions with clear gatekeepers also lend themselves to defaults. There is a whole range of decisions where the government is a gatekeeper and either sets a default or constrains what choices individuals can make. Even the tax incentives to make contributions towards retirement are a form of nudging.
Considering this, it should be clear that many decisions in employee benefits lend themselves to defaults. While individuals may have little idea of how much to save for retirement, how much risk cover they need, how to invest – particularly given the long-term views required – trustees, with the help of expert advisers, often have enough information to better answer these questions.
Within the categories of employee benefits that we have identified – retirement funding, risk benefits, healthcare benefits, and financial education – the first two lend themselves to defaults, to varying extents. Where medical scheme membership is a condition of employment, employers will typically offer one or more schemes of choice which employees can choose between. In addition, some employers may restrict the choice of benefit options within the chosen medical scheme. Although this is a type of default, it is not an individualised solution. Experts can definitely help individuals to make the right choice, but they need a lot of information from the individual to do so, and so healthcare benefits don't lend themselves to defaults. As far as financial education is concerned, it is used in situations in which individuals have to face decisions alone and so defaults would not be entirely relevant. In 'Failure to launch', we will explore how employers and trustees can effectively implement financial education programmes.
A good default must work to satisfy the needs of the largest representative body of fund members – not the needs of the trustees or the service providers.
A good default should reflect what would be considered the best advice of the fund’s experts for that population.
A good default should maximise the probability of meeting a fund’s needs and targets.
Defaults should seem intuitively sensible to minimise opting out (de-selection) by members.
A good default should be cost-effective.
All this means that a fund’s set of defaults should represent the most optimal path for the majority of individuals in the fund to reach their targets, in the absence of additional information. Get the default wrong and the outcomes could be crippling for members. Get the default right, and the risk of potential harm should be mitigated to some extent. This means that the exercise of setting an optimal default demands as much insight into the composition of the population being serviced, and the appropriate targets for that population, as it does in defining a theoretically correct strategy.
An alternative to a one-size-fits-all default is a smart default. In a smart default, experts are still often in the best position to identify the right solution, but they need more information about the individual to get it right.
In some cases, this is a demographic variable which is readily available to the fund. For instance, many funds apply a life-stage approach for their investment strategy default. This is a smart default4 of the most basic kind because it adjusts the asset allocation of the individual as they approach retirement. It is based on a single variable – such as years until retirement or age – which then generates different answers for different individuals without the need for engagement.
One of the fair criticisms of life-stage investment strategies, though, is that they don’t take into account the total savings reflected in a member’s account at the point they start to de-risk, or perhaps their specific choice of how they want to convert that savings into a post-retirement income stream. By simply adding these two small pieces of additional information, we now have an even ’smarter default’ that provides members with individualised solutions over the course of their membership. As defaults get ‘smarter’, they are able to incorporate an increasing number of variables to reach the correct answer.
Sometimes, smart defaults require an individual to engage to a limited extent, or they could be improved through limited engagement5. This is when an expert needs key information to which they don't have easy access to make a decision – for instance, what are an individual’s priorities for retirement. In this case, they could use advice or interactive tools to present a limited range of choices, along with a series of questions whose answers will guide an individual to the correct decision.
Other areas where smart defaults are applicable in this book include the approach discussed in 'The heart of the matter' where we alter the mix of risk benefits and retirement funding depending on member age, and our discussion in 'The journey: Not just the end game' on individualised asset allocation.
Defaults can prove to be useful tools in setting contribution rates. Experts are likely to have better insight than individuals into what an appropriate contribution rate is to maximise the probability of reaching the target. We talk about targets in 'What’s the point'.
Given the current low return environment, funds may find that the default contribution rate ends up being more than many members can stomach. In fact, Alexander Forbes Research & Product Development found that a 25-year-old new fund entrant, who plans to retire at the age of 65 with a replacement ratio of 75%, will need to contribute 17% of his pay towards retirement savings. This rate may be more than many individuals are comfortable with. This means that maximising the probability of meeting the target (Principle 3) may clash with either meeting an individual’s other needs (Principle 1) or maximising the number of individuals using the default (Principle 4).
A 25-year-old new fund entrant, who plans to retire at the age of 65 with a replacement ratio of 75%, will need to contribute 17%, of their salary.
Choosing the contribution rate default when it starts to infringe on expectations of take-home pay is a key challenge.
Getting people to choose the contribution rate default when it starts to infringe on their expectations of their take-home pay is a key challenge. When rates are set too high, individuals are likely to opt out6. When they are too low, individuals are unlikely to meet their targets7. A key concern for the UK’s auto-enrolment programme is that all employees are enrolled at a 3% contribution rate, which is unlikely to make a significant difference in increasing retirement incomes. Because defaults are so often seen as advice, setting a low default can result in individuals saving less than they would otherwise have done8.
But there is an alternative called auto-escalation, which academics originally proposed as the Save More Tomorrow (SMarT) programme that could provide a less painful way of achieving what is required9.
Auto-escalation slowly adjusts contribution rates to the level required as individuals receive salary increases. This means it improves their chances of a decent retirement outcome without disrupting consumption. We can implement auto-escalation in a few ways. In its original conception in the SMarT programme, contribution increases were linked to annual increases. Each year when an individual receives their salary increase, some percentage of the increase (say 1–2% of a 7% increase) would be redirected towards saving. This means that while contributions toward retirement are increasing, so is take-home pay, making a smaller negative impact on members.
Since it has been rolled out more broadly, it has often been used as a default where all new employees start at a defined contribution rate, with a predetermined increment applied to increase the rate each year, until they reach a cap. In this case, the increase in rate is not related to the annual salary increase as such, but applied at a specific increment for all employees.
If we consider a fund in which individuals committed that part of their salary increase that was above inflation to their retirement fund contribution for three years and then held that contribution rate constant until retirement, we would see the average replacement ratio increase from 40% to 82% at the end of that three-year period.
A final way in which we can apply auto-escalation is on an age basis, where we set default contribution rates for different age bands, with older ages having higher default contribution rates.
In this final route where all new employees follow the same step-wise increases in contributions based on their age, a few potential pitfalls have been identified. Extensive US experience suggests that we should be wary of10:
Any of these could result in individuals saving less than is necessary or even less than they would have without auto-escalation.
The starting contribution rate under an auto-escalation structure could be set by age. In standard auto-escalation, any new employee, whether 25 or 40 years old, would start at the minimum contribution rate.
In South Africa, inflation complicates the setting of the increment. An aggressive increment, such as 2%, increases the chances of real take-home pay falling, while small increments may fall short of the contributions required.
In the case of individual need, research suggests that individuals with lower incomes tend to be less likely to opt out of a default – in other words, make a choice different from the default11. Yet, it could be these very individuals who can least afford higher contribution rates, and have the greatest need for take-home pay. So we need to ensure that individuals receive proper advice and communication around the impact of such defaults before they are implemented.
Another challenge with auto-escalation is in industries or companies with high employee turnover. If employees change jobs frequently and the firms that they join all use auto-escalation, they may be perpetually enrolled at the minimum contribution rate. Or worse, they may transfer into a fund without auto-escalation and the net effect of a low starting contribution rate with one employer together with an only average contribution rate with the next employer could also result in the individual being underfunded.
To avoid this, the starting contribution rate under an auto-escalation structure could be set by age. In standard auto-escalation, any new employee, whether 25 or 40 years old, would start at the minimum contribution rate. If the strategy is set around age, then a 25-year-old would start at the same rate as a model 25-year-old enrolled in the strategy, say 10%, while the 40-year-old would start at the same rate as a model 40-year-old, say 18%.
Given that as people age, their salaries tend to rise in real terms (up to a point), this still retains the benefits of auto-escalation. If the increase in contribution rate is structured to be a fixed annual increment, it does raise significant risk of take-home pay falling at some point. Because it is a default, individuals could choose to opt out of the structure at this point. However, as this is more likely at older ages when individuals become more aware of retirement, employees may be willing to stick with the strategy.
Despite its shortcomings, auto-escalation may work well in bargaining council funds, where members may remain within a single fund over their lifetime even though they may not always have the same employer.
Legislation will likely require boards of trustees to select a default option for fund members who annuitise their fund credit at retirement. This default could be offered within the fund or outside the fund; it could be a life annuity or a living annuity. Unfortunately, research shows that no single default is likely to meet the needs of the majority of any fund’s population12.
What makes this particularly challenging is that funds will have to set a default and individuals are likely to take this as advice13. For this reason, trustees will need an explicit strategy to engage individuals in the decision of choosing a default. This could include providing advice, providing simple engagement tools that direct them to choose smart defaults or providing just-in-time education to help them make the decisions themselves. We discuss communication techniques in 'STOP THE PRESSES! WE NEED TO TALK'.
Trustees need to consider what types of annuities to offer as well as how to provide the appropriate advice to support decision making. In keeping with our framework, some of the key steps involved will be:
We have already discussed in 'The heart of the matter' how to do a broad analysis of needs and wants within the fund. In addition, trustees can also analyse recent decisions by retirees as well as the profile of upcoming retirees. This analysis should help trustees to make sure they have the appropriate menu of options available and an advice framework, if needed.
Let’s look at an example of such an analysis. The analysis is based on a large fund in the retail and wholesale industry. The graph below shows the annuitisation decisions of recent retirees by size of their benefit and the proportion of retirees selecting each annuitisation option.
In our example here, the evidence suggests that members with significant fund credits are purchasing living annuities. By contrast, members with low fund credits (or potentially low incomes) are simply withdrawing cash. This gives trustees some sense of what type of default annuity will be the most attractive to members. It also allows trustees to check if decisions appear to be prudent. If many retirees appear to be making particular kinds of poor decisions, then they can incorporate this into the advice framework.
We can also carry out an analysis of upcoming retirees. This involves looking at those employees who are within seven years of their normal retirement age.
A look at the projected fund credits of upcoming retirees can help to shed some light on the likely decisions that they will make when they reach retirement. The graph below serves as an example of how projected fund credits may look.
A final, but perhaps more crude, assessment is to look at the spread of salaries of upcoming members. In some instances, annual salaries have been used as a proxy for the financial literacy levels of members. Although this may be a less precise approximation, it could supply a basic guideline.
Living annuities, for example, require a significant level of financial understanding (and financial advice) to set an appropriate investment strategy and drawdown rate, and to understand that the pension is not guaranteed for life. A higher level of financial sophistication is required to understand the risks and lack of predictability of the pension and appreciate the benefits of a living annuity.
Inflation-linked annuities are easy to understand as they provide pension increases equal to inflation for the rest of the retiree’s life. These annuities are therefore predictable (in real terms) and offer the ultimate form of financial security. With-profit annuities fall somewhere between living annuities and inflation-linked annuities. Pension increases are linked to an underlying investment return but are guaranteed never to be negative (in other words, the pension will never reduce) and are guaranteed for life.
For our example fund, the graph below shows that we have a skewed salary picture. This can complicate things, and may indicate that a single default annuity option may not be suitable for all members. The implied variability in financial literacy is self-evident.
These types of analyses will help trustees and their advisers to identify what kinds of annuities to offer, whether there are enough retirees and whether trustees and employers have a big enough governance budget appetite to make an in-fund solution feasible and what level of advice they’ll need. They may also provide sufficient information for a smart default to be structured where members’ defaults are contingent on this kind of information.
By limiting the questions to which annuities are likely to be best for the population, trustees can design tailored communication and education with this in mind. And as time goes on, the trustees can monitor the take-up and establish whether their approach is working.
Regulators have seized on the ability of defaults to influence behaviour as a powerful way to improve decision making across a range of contexts. National Treasury has made them a prominent feature of retirement reform across many retirement decisions. For this reason, trustees need to understand both the power of defaults and how to apply them correctly. They need to be based on a solid grasp of members’ needs, appropriate targets, the best possible advice and ongoing monitoring to ensure trustees channel the desired behaviour.
1 Johnson and Goldstein (2003)
2 Carrol, Choi, Laibson, Madrian and Metrick (2009)
3 Carrol et al (2009)
4 Fernandes, Lynch and Netemeyer (2014)
6 Beshears, Choi, Laibson and Madrian (2010)
7 Choi, Laibson, Madrian & Metrick (2004)
8 Choi et al (2004)
9 Thaler and Benartzi (2004)
10 Van Derhei (2012)
11 Beshears, Choi, Laibson and Madrian (2010)
12 Butler, Hu and Kloppers (2012)
13 Brochetti, Dee, Huffman and Magenheim (2011)