Nudged: Behavioural insights in public policy

Nudged: Behavioural insights in public policy

PM&C
Friday, 27 September 2019

Canada-Australia Public Policy Initiative Lecture

Dr David Gruen
Deputy Secretary, Department of the Prime Minister & Cabinet

Image of David Gruen

Introduction 1

Thank you Ian [Shugart] for that kind introduction.

Let me begin by recognising that this lecture is taking place in the traditional territory of the Algonquin people and acknowledge them as the past, present and future caretakers of this land. 

If I had been giving this lecture in Canberra, rather than here in Ottawa, I would have begun by acknowledging the Ngunnawal people, as the traditional owners of the land on which we were meeting, and I would have paid my respects to their elders past, present and emerging.

So the Canadian and Australian acknowledgments of our first peoples are not the same – but they have strong similarities. I mention this because I think it is a fitting example of why we have initiatives like the CAPPI lecture. Australia and Canada are not the same – we are distinct countries with distinct cultures and histories. But we have a lot in common, and we can learn from each other, in the ways we approach topics of mutual interest.

I’m really pleased to be participating in this important policy tradition between Canada and Australia. The Canada-Australia Public Policy Initiative has fostered a culture of genuine exchange and thoughtful discourse between our two countries in the twelve years since its establishment by the Prime Ministers of Australia and Canada.

I will continue the tradition of talking about shared challenges in public policy. Rather than address a particular policy problem, however, I’ll focus on a tool both Canada and Australia have added to their policy toolkits in recent years.

That tool is behavioural insights, an approach to policymaking that draws from psychology, cognitive science and economics to better understand human behaviour, help people to make ‘good’ choices more easily and help improve the effectiveness of public policy interventions.

In my lecture today, I’ll touch on some of the field’s successes and explore what I see as the opportunities for greater impact in the future.

Before I do so, I want to share an experience I had on my trip from Canberra to here. Walking through the international terminal at Sydney airport, from security near the front of the terminal to the gate from which my aircraft was to depart near the back, I was obliged to walk through serried ranks of duty free shops on either side of me – for the best part of 100 meters. I was being nudged by Sydney airport to buy duty free on my way out of the country. There was no compulsion, but there was also no alternative route to my gate. I suspect most of you would have experienced something similar to this.

This is but one example of the general point that the private sector knows a thing or two about behavioural insights, and private firms are not averse to a nudge or two when they judge it in their financial interests. The story of the past decade is a story of governments catching up, and in some cases overtaking, where the private sector has long been.  

The focus of my talk is the development, and increasing sophistication, of public-sector nudges. I’ll begin with an example that shows just how subtle, and sometimes unexpected, human behavioural reactions can be.

Requiring disclosure is a common policy response to conflicts of interest. Many countries require financial advisers to disclose any commissions they would receive from investments they recommend to clients.

The idea, which seems logical enough, is that telling clients about any conflicts should raise their suspicions of advice from conflicted advisers. This, in turn, should provide an incentive for advisers to reduce or eliminate their conflicts of interest so their advice will be trusted.

But it turns out that disclosures of this kind can generate exactly the opposite reaction from clients. Let me elaborate.

Dr Sunita Sah, a psychologist from Cornell University, has found disclosures do indeed cause clients to have less trust in a conflicted adviser. But disclosures can also cause clients to feel social pressure to follow the conflicted advice. Clients can become concerned that refusing to follow the advice would signal distrust in their adviser, with whom they have often formed a personal bond.

Reinforcing this effect, when an adviser discloses that they will receive a personal gain, clients can interpret this as a request that the advice be taken, rather than treating the disclosure as a signal that the advice should be discounted. In this case, the client feels a social obligation to grant a favour to their adviser rather than seek alternative advice that would better serve their financial interests.

These two effects can lead well-meaning disclosures to backfire. In one experiment, Dr Sah found that clients presented with advice they knew to be conflicted were twice as likely to follow that advice as were clients presented with advice with no disclosure of any conflict.[1] So instead of warning people away from conflicted advice, the disclosure increased its uptake.

Needless to say, this is not what policymakers had in mind when they mandated disclosures.

This doesn’t mean we should stop requiring conflicted advisers to disclose their conflicts. Rather – and this is a general rule – it means policymakers should think carefully about the specific design of their policy interventions. It turns out the undesirable effects of disclosure can be reduced if disclosures are made by a third party. Giving clients the opportunity to make decisions in private also reduces the pressure to comply with advice, as does a cooling off period before the decision is finalised.

This example highlights the value of being open to behavioural insights when developing public policy. It also illustrates the importance of real-world testing. Behavioural responses can be surprising, and sometimes hard to predict. They can vary in important ways in different contexts. By testing real-world responses to our interventions, we can reveal the surprises and adjust our policies accordingly.

It is because of results like this one that governments around the world have created behavioural insights units. The challenge for these units has been to take the richer understanding of human behaviour revealed in experiments, and determine what it means for policy. And to make sure we don’t replace one flawed behavioural model with another, the challenge is to test and re-test our behavioural forecasts in different contexts.

Nudges in public policy

Before turning to examples of successful behavioural interventions, I want to spend a little time on the ethics of governments deliberately seeking to influence the choices made by their citizens.

For those with a strong libertarian bent, any attempt by governments to do this is unacceptable. But I think that criticism is too all-encompassing.  It is almost unavoidable that Government policies will influence behaviour, whether or not this is consciously intended.

For example almost invariably, how choices are presented to individuals has a bearing on which choices are made. In situations where a range of choices are available to individuals, the selection of a default option, the ordering of the alternative options and decisions about whether individuals have to opt in or opt out will materially affect the distribution of final choices made.

Importantly, the government interventions we are talking about (‘nudges’) preserve the freedom to choose rather than mandating or outlawing any particular option. What we are talking about is making it easier for people to make choices we have reason to believe will benefit them and/or the wider community.

Cass Sunstein has put together a ‘Bill of Rights’ for nudging — five rules to ensure we are nudging ethically. Sunstein’s five rules are that nudges must:

  1. be consistent with people’s values and interests,
  2. be for legitimate ends,
  3. not violate anyone’s individual rights,
  4. be transparent, and
  5. not take things from people without consent.

These rules seem to me to constitute a reasonable basis on which to judge when government nudges are defensible – indeed desirable.

It was in 2010, under Prime Minister David Cameron, that the United Kingdom became the first country to set up a Behavioural Insights Team (BIT), with the specific goal of incorporating an understanding of human psychology into policy initiatives. For the first four years of its life, it was in the UK Cabinet Office, before becoming a partly privatised joint venture.

The OECD reports that there are now over 200 public sector organisations around the world that have applied behavioural insights to their work – so this is an approach to public policy with enormous appeal.

Canada has been making great strides in applying behavioural insights to the work of government, establishing a central team in the Impact and Innovation Unit in the Privy Council Office in 2015, complemented by teams in several Federal agencies and provincial governments.

In Australia, we’ve adopted a similar approach. The Behavioural Economics Team of the Australian Government, so named so we can use the acronym BETA, sits in my group in the Department of the Prime Minister and Cabinet. The team was explicitly set up to apply behavioural insights to public policy and, equally importantly, to build behavioural insights capability across the Australian Public Service. It sits at the centre of a network of ten behavioural insights teams across the Federal Government, and alongside several state government teams.

I want to take the opportunity of a captive audience to tell you about a few Australian examples of successes in applying nudges to achieve better outcomes for individuals and for the community more broadly.

The first is a project run by BERT, the Behavioural Economics Research Team in the Department of Health. The issue was after-hours care provided by doctors to patients, and the cost of the substantial reimbursements provided by the government to doctors for this after-hours care.

There are two relevant types of after-hours care: urgent and non-urgent. To understand this example, you need to know that the government reimbursement for an urgent after-hours care visit was more than double the reimbursement for a non-urgent visit. And you also need to know that the determination of whether the care provided to patients was urgent or non-urgent was determined solely by the doctors who provided the care. By 2016, the government was faced with dramatic growth in doctors’ claims for urgent, after-hours care, with the total cost having risen by 140 percent over the previous decade.

The health department thought this rise might be explained, at least in part, by doctors incorrectly classifying items as ‘urgent’. So they identified the 1200 doctors with the highest urgent after-hours claims, and ran a randomised control trial. Each doctor received a letter – one of three alternative letters – but which letter any particular doctor received was chosen at random.

One letter compared that doctor’s billing practices to her peers, showing she was claiming the urgent category far more often than others. This drew upon the behavioural insight that individuals are often motivated to change their behaviour when they realise they are out of step with their peers.

The second letter emphasised the consequences of non-compliance, including administrative penalties and legal action. This letter drew upon the behavioural insight that people tend to avoid losses more than they seek equivalent gains.

The third letter was the control. It was a standard bureaucratic compliance letter – which ran to three pages.

Which letter worked best? All three letters successfully reduced claims – but one stood above the rest. The peer comparison letter was more effective than either the standard compliance letter or the loss framing letter. It reduced claims by 24 percent. 

What I like about this trial is that the softer ‘nudging’ approach – simply highlighting how a doctor compared to his/her peers – was more effective than the harder ‘compliance’ approach of threatening sanctions. Both were more effective than the compliance letter that had been the department’s standard.

In the six months after the letters were sent, the 1200 high-claiming doctors reduced their claims by over $AU 11 million (across all three letters), and 18 doctors voluntarily owned up to over $AU 1 million in previous incorrect claims.

This is a great example of how a simple and cheap nudge can yield big dividends.

The second example I want to highlight was led by BETA, the central team I oversee, in partnership with a large Australian bank – Westpac – and the Australian Treasury (similar to Canada’s Department of Finance).

We were interested in customers’ monthly repayments on credit card debts – particularly for those who pay off the minimum amount each month. As I’m sure you know, credit card interest rates are high, and debt can accumulate quickly. And yet there are a large number of customers who pay the minimum repayment, or close to it, each month.

Financial decision-making is one of the richest areas for behavioural insights. We often fail to follow through on our intentions when managing our money. We often put more weight on immediate gratification than our long-term goals. We can be overly optimistic about our future self’s willingness/commitment to pay off debt and stick to a budget. And the minimum repayments on our credit card can act as a psychological anchor. So the minimum repayment can become a reference point from which we make a choice about our repayment, rather than thinking about our ability to pay and the interest costs we could avoid.

We worked with Westpac to identify a sample of 24,000 consumers who had consistently made low repayments for the previous 12 months. I’ll say that again: 24,000 customers. Often behavioural trials can be conducted with a very large number of people – easily enough to generate results with a high degree of confidence in their statistical validity.

We designed a range of different SMS text messages and email reminders. And we randomly allocated these reminders to different customers. Some messages were short, saying simply “Hi John, payment on your Westpac credit card is due next week.” Others were a little more complex and used social norms or loss framing to encourage repayment. In the control group, customers received no reminder message at all.

What was the result? Email reminders had no effect on repayments whatsoever! The text messages, on the other hand, were very effective, increasing credit card repayments by around 28 percent or $AU 134 each on average. What’s more, there was no significant difference in effect between the carefully crafted messages. It was simply the reminder that had the effect, not the specific words.

And these effects persisted. 12 months on, those who received a text message had credit card balances $AU 365 lower than those who had not.

This is another example of how small, low-cost nudges can have a positive impact on people’s lives. And there are plenty more of these examples, from all over the world.  This work can (and should) influence the development of policy responses to a wide range of policy issues – moving away from the often binary choice between ignoring poor outcomes and trying (often unsuccessfully) to regulate them away. 

In Canada, I know the team has trialled a long list of successful nudges, including an intervention that doubled online engagement among Canadian women with careers in the armed forces.

In Denmark, they’ve nudged hospital workers to wash their hands, and businesses to complete their tax returns correctly.

In Singapore, they’ve nudged people to reduce their water usage and take public transport to work.

In France, they’ve nudged people to adopt energy efficient practices, and to undertake physical activity.

These nudges represent worthwhile improvements in government service delivery and public engagement. They also have the added bonus of making the government’s interactions with the public simpler and more user-friendly.

But these ‘first-generation’ nudges apply our new, better understanding of behaviour only at the margins of policies. They’re often seen as a one-off addition to an established program. Or a last resort to improve compliance or address challenges in implementation.

Like other behavioural insights teams around the world, we are now asking the question: what impact can behavioural insights have if we apply them earlier and more comprehensively in the policy process?

 

What's next for behavioural insights?

There are a number of policy areas we believe would benefit from a thorough-going, early and iterative application of behavioural insights. Retirement savings policy, consumer protection and preventive health strategies all require people to embrace behaviours that don’t always come easily. We’re asking them to save for the distant future, understand complex terms and conditions and change their eating and exercise habits. Humans tend to find all these activities challenging.

Retirement income

Insights from behavioural economics are revealing – money-smarts are only one piece of the puzzle in achieving a secure retirement. Cognitive biases – such as present bias, confirmation bias and cognitive overload – all hinder people from reaching their desired financial goals.

Present bias refers to overvaluing the present relative to the future.  While choosing a retirement plan is likely profoundly to influence literally decades of our lives, many of us spend little time – sometimes less than an hour – choosing our plan.

Confirmation bias leads many of us to spend what little time we have set aside to choose a retirement plan looking for one with an investment strategy that supports our existing investing approach.

And cognitive overload occurs when people find it too hard to process a mass of information in order to make decisions. In the context of planning for retirement, it leads many of us to stick with choices we have arrived at by default.

Together, these cognitive biases create a big gap between our intentions and our actions: although people intend to save for their retirement, they often don’t translate that into action. For most people, how much to save, and in what form, are difficult cognitive problems – because of both our limited calculation powers and the apparent enormity of the task.

Partly in response to people’s recognised poor capacity – and inclination – to plan for the distant future, successive Australian governments have mandated a level of compulsory private saving. Similarly in Canada, compulsory saving through the Canadian Pension Plan has been mandated for several decades.[2]

In Australia, the amount of current income mandated to be put aside for retirement has risen gradually since the scheme was put in place in the early 1990s. Currently, Australian employers are required to put aside 9.5 percent of each employee’s base salary into a superannuation fund of the employee’s choosing. This money is then invested on the employee’s behalf until they can access it upon retirement. Complementing this policy is a means-tested public pension, available to people with insufficient personal savings in their retirement.

I suspect the term ‘behavioural insights’ didn’t exist when Australia’s compulsory superannuation system for retirement savings was established in the early 1990s. Compulsory superannuation certainly wasn’t framed explicitly as a behavioural insights intervention. But the system is clearly designed with a behavioural insight in mind: to compensate for people’s well-documented lack of interest in planning for their long-term financial future.

That was a big step forward.

But the job wasn’t done. In regulating superannuation funds in the 1990s, the government adopted the then standard (neoclassical) economics view of human behaviour. Savers were treated as rational economic agents who knew their own business best. The government facilitated consumer choice by requiring superannuation funds to disclose information about investment options. The idea was that people would use the disclosed information, examine their set of options, and do their best to choose where and how to allocate their savings.

One presupposition of this approach is that consumers are sufficiently engaged with their superannuation savings to make these decisions.

Of course, we know that’s often not the case. In 2009, I was a member of a superannuation review panel that examined these questions. At the time, 80 per cent of superannuation fund members were invested in the default fund chosen by their employer. Of that 80 percent, anecdotal evidence suggested that about 20 percent explicitly chose the default option, with the rest making no active choice whatsoever.[3]

We interpreted this as evidence of substantial community disengagement with the superannuation system. It could also be interpreted as evidence of cognitive overload. When complicated decisions are required, people often stick with the status quo and take no decision at all. In that case, the default option becomes very important.

So we set about creating a default option with features that would promote the wellbeing of those who did not actively choose another option.

We called this new default ‘MySuper’. By design, MySuper funds must be simple, cost-effective, and with a diversified portfolio of investments. After the reform became law in 2013, employers were required to offer a MySuper fund, into which contributions were defaulted for those employees who did not choose their own superannuation fund. Trustees of MySuper funds must comply with a number of requirements including types of fees and reporting obligations – all designed to safeguard the retirement savings of those who remain disengaged from the system.

Importantly – and as I explained earlier, this is a feature of nudging interventions – there is no compulsion. There is a wide range of other funds available that are not required to comply with the MySuper principles, and employees are free to choose them if they wish. The crucial point is that all the default funds – funds that employees are defaulted into if they make no active choice – are MySuper funds and must comply with government-mandated minimum standards.

That was another big step forward.

27 years after Australia’s superannuation system was created, the government continues to look for ways to improve it to better accommodate real-world behaviour.

The system has generated a retirement nest egg for Australians of $AU 2.9 trillion[4] or $2.6 trillion Canadian dollars. It has eased the pressure on the public pension system and will provide a comfortable retirement for most Australians.

It certainly is a success, notwithstanding some remaining challenges. The creation, and refinement, of Australia’s modern retirement income system, is an example of the kind of fundamental reform that follows from a more sophisticated appreciation of real-world human behaviour.

In 2019, however, the behavioural insights toolkit — including a collection of catalogued behavioural patterns and a framework for real-world testing and iteration — can help us shortcut the drawn out policy process I’ve just outlined.

Early in the policy process, we can now start with a richer understanding of human behaviour and build this into the design. We can create space at the outset for faster testing and iteration in policy implementation. We can measure and adapt to the realities of people’s behavioural responses to our policy frameworks as they’re implemented.

Consumer protection

Another clear policy opportunity is consumer protection. In the same way we know that most people don’t adequately plan for their retirement when left to their own devices, we also know most people don’t read terms and conditions.

And you don’t need to be an expert in consumer behaviour to understand why: we are continuously flooded with information. We engage with ever-expanding channels of inter-personal communication and choose from a dizzying array of goods and services in every facet of our lives.

In many cases the terms and conditions are long and hard to decipher. Paypal’s terms and conditions are longer than Hamlet and certainly less enjoyable to read.

In this context, where the vast majority of people don’t read them, it’s clear disclosure requirements are not successfully informing consumers.

As with superannuation, the regulatory framework that protected consumer purchases often relied on a neoclassical economics principle: ensuring the information is available for consumers to make the optimal choice. But we now know this is not enough to generate good outcomes for consumers.

As well as not reading terms and conditions, consumers are heavily influenced by what Richard Thaler calls “supposedly irrelevant factors”. Consumers are swayed by factors like the framing of costs and benefits, the number of choices available and the timing of when choices are presented to them.

These factors combine to create sub-optimal outcomes. Studies in the UK have found that over 9 million households could save over £300 a year by switching their energy providers, and that mobile phone customers are overpaying for mobile phone contracts by £355 million a year.

There is a role for behaviourally-informed regulation to promote better outcomes for consumers and protect them from retailers who seek to profit from these biases.

We can find some advice on how to do this from the UK Behavioural Insights Team, who has published a set of steps it recommends for regulators to strengthen consumer protection.

The first is to incorporate consumer outcomes into the criteria for a well-functioning market. For example, regulators could measure the percentage and distribution of consumers getting a ‘bad deal’ in key markets by looking at how many of them could access a better deal if they switched. They could also examine consumer comprehension and consumer satisfaction.

The second step recommended by the UK team is to use data to actively and directly inform consumers about market performance, and identify behavioural market failures.

Third, regulators can design remedies to overcome identified behavioural market failures. For example, they can promote simple heuristics to consumers like ‘you should compare three quotes before a major purchase’. They can mandate timely, smart disclosures. And they can create low friction complaints mechanisms.

Finally, regulators can test whether these remedies are actually leading to better outcomes for consumers.

This kind of approach represents a significant shift for most economies. The Australian Government has started down this path with the introduction of the consumer data right. The idea here is consumers will be able to access their personal data from banks, electricity providers and others, and transfer this data (with appropriate safeguards) to brokers who can present them with tailored alternative offers from competitors.

In Australia, the consumer data right will be first applied to the banking sector, along the same lines as the system in the UK. And it will then be rolled out to energy (like retail electricity supply plans), telecommunications (like mobile phone plans) and other sectors.

This will make it easier for consumers to compare and switch products. Consumers will be presented with specific information about the savings available to them if they change providers. This makes the savings more salient, helping to overcome present bias. And the brokers remove some of the friction costs in searching for a better deal and switching providers.

Australia is also starting to think about behaviourally-informed regulation in other consumer protection contexts. For example, the Government has agreed to implement a deferred sales model for add-on insurance. This is seeking to improve the situation when retailers can employ pressure tactics to sell clients insurance when they are purchasing something else, like a car or a flight. The problem is these insurance products tend to have poor claims ratios and consumers are not engaged or well-informed when making the purchasing decision.

Add-on insurance is also significantly more expensive than counterpart products sold in the standalone market. In one example, travel insurance was almost three times the price when purchased as add-on insurance than as standalone insurance — from the same underwriter.

It’s clear there are a number of cognitive biases at play in the purchase of add-on insurance. Consumers may be performing mental accounting, where they assign travel insurance to the ‘flights’ bucket of spending, rather than their ‘insurance’ bucket, which is likely to make them less price sensitive. They may also be susceptible to the ‘halo effect’, where they may interpret the friendliness of a salesperson as a signal they are trustworthy and therefore selling a worthwhile insurance product.

The Australian Government is planning to pass new legislation to prevent retailers from selling this insurance on the spot. Instead, clients will need to defer their decision to purchase it for a few days, when they have some distance from the high-pressure environment. Consumers will have the opportunity to more carefully consider their options and potentially explore alternatives from other providers.

And this approach has been successful. A UK trial of deferred sales for add-on car insurance found the deferral period reduced sales of the add-on insurance and marginally increased standalone insurance sales. In the main, consumers decide, on reflection, not to go ahead with the purchase.[5]

In this discussion, I’ve outlined two policy areas where behavioural insights can fundamentally reshape the foundation of our approach. And I expect we will see behavioural insights being applied in this way more often in the future: developing smarter more behaviourally informed regulation and practices; delivering more effective public information campaigns; and helping improve the efficiency of markets.

And looking ahead, data science and machine learning also offer some fascinating possibilities for the future of behavioural insights – helping to identify more subtle behavioural patterns and make better predictions.

The ability to train machines to analyse vast integrated digital datasets is going to make an enormous difference to government service delivery.

As Harvard academics Ben Buchanan and Taylor Miller put it: 

“Machine learning can spot cancer. It can translate complex texts. Drive cars. Beat the best human in the world at one of the most complex games ever invented. Devise alien-like designs to create more efficient physical structures. Save energy.”

This is an emerging field in applied behavioural science but it opens up huge opportunities for delivering more personalised policies and services.

Conclusion

Today I’ve given you a taste of some of the successes of ‘nudges’ in strengthening the implementation of policy initiatives. And I’ve looked ahead and described some of the opportunities available to us as policymakers to take advantage of the far richer understanding of human behaviour we now have.

Behavioural insights can make our policy frameworks more effective, more user-friendly and cheaper to operate. Importantly, they achieve this by supporting people to follow through on their intentions, helping them lead more fulfilling lives. It’s an exciting time to be working in public policy.

 

[1] I am grateful to Amelia Johnston and Anne-Line Giudicelli for much help with the preparation of this lecture, and to Jenny Wilkinson for helpful comments on an earlier draft

[2] Sah, Loewenstein and Cain (2018) Insinuation anxiety: concern that advice rejection will signal distrust after conflict of interest disclosures.

[3]  Although there are significant differences between the Canadian and Australian approaches to retirement saving, both have a significant level of compulsory saving.

[5] APRA (June 2019) Quarterly Superannuation Performance

[6] Treasury (9 September 2019) Reforms to the sale of add-on insurance products: proposal paper