Image by Steve Johnson via Unsplash

The closing panel of our recent Behavioural Economics and Public Policy conference included leading politicians and policy makers sharing their views on behavioural insights (BI) and government. The full audio of the panel is available here. This two-part blog piece presents some of the most interesting issues, approaches and questions considered by the panel, extracted from the transcript. Find Part 1 here.


  • Chair: Miranda Stewart, University of Melbourne.
  • Jennie Granger PSM, former Director General, Her Majesty’s Revenue and Customs (HMRC).
  • David Gruen, Deputy Secretary, Prime Minister and Cabinet.
  • Andrew Leigh, Shadow Assistant Treasurer and Federal Member for Fenner, ACT.
  • Jane Mitchell, Behavioural Insights Team, Australian Taxation Office (ATO).

Miranda Stewart: Jennie Granger was formerly Director General at Her Majesty’s Revenue and Customs (HMRC) Customer Compliance in the UK, and Deputy Commissioner in the ATO. Jennie oversaw the implementation of a process in HMRC called ‘Promote, prevent, respond’, which combines behavioural ideas with an enforcement response.

Jennie Granger

Former Director General at the Her Majesty’s Revenue and Customs (HMRC), UK

My voice in the room on BI debates is, in (Behavioural Insights Team Australian Principal Advisor) Alex Gyani’s words, the ‘sponsor-implementer’ person in the organisation. Building on his success factors, he spelt ‘SCALE’, I would like to add an ‘R’ to that: Resilience. This is really important for sponsor-implementers.

First, as Jane said: Revenue organisations are huge. HMRC at the time I joined in 2012 had 60,000 staff, but it is running a system with 45 million individuals and businesses in it. We have all of the complexity of the community, all of the complexity of businesses, engaged with the organisation. The core question is: how do you influence the taxpayer base to continue to pay their taxes, or to get back on track if they deliberately strayed off, or exited?

Social cost is one of the really big things that drive all of us. If we can make it simple and you will do it, you are more likely to do it again and stay engaged. If you think we are being too heavy handed if you just made a mistake, it actually escalates bad behaviour. So there is a tremendous value around BI approaches, but one of the really big issues is that the administrator has to have multilayered strategies so that those who do not do the right thing are properly dealt with. This is important from a fairness perspective, to reinforce that. That is lesson 101 in my strategy for all of you.

I want to emphasize, and this is quite important for thinking about randomised trials and how long research takes, is that the tax world, the community and business are constantly changing and adapting. This means that anything that the administrator does as a mutation is not just about the particular group it is aimed at, it also has effects in other parts of the system.

I would say also, that we have been using what you call behavioural insights for decades. This approach is really important and it is really based on smart data. The core skill for any effective revenue organisation is the ability to model data, understand the insights from it, develop new insights from it, and be able to implement strategies to influence negative trends or reinforce positive trends. In lots of ways, we are about influencing the culture of our nations. I shall give you four examples.

The first example shows that ‘ah-ha!’ moment, a surprising result, when you start to see things differently.

When I was a Deputy Commissioner in Personal Tax in the ATO, at the turn of this century, we could not understand why we had people who had been good tax return filers suddenly drop off the cliff and stop filing. We found this unusual because –as you all know– most can get a refund, so you actually queue at the door to get your tax return filed. We did some modelling and found that men in their thirties and forties were the ones who suddenly had stopped complying. We had to dig deeper, because we did not have good social analysis at that time to understand what had happened to cause that. And the answer was that their relationship had broken up. Why was that important? Well, they may have fallen to pieces a little bit of course. But more importantly, their female partners typically did all the paperwork of the household. Some of them had not even realised that it was time to do their tax return.

Why do I tell you that story? It is important because it is about what we can do proactively. We had already gone down the path of how we might segment the massive taxpayer base in various ways. This taught us that it is important to think about lifecycle or business cycle milestones, and to start to think about what you can do to support taxpayers going through that transition. That is where I started following up with how you can get really interesting and different insights, that tell you something about why people are not doing what you think they ought to do.

My second example is the well-known peer pressure example, where HMRC did a trial with letters, to see if they could convince taxpayers who were reluctant to pay their tax debt to pay it. That experiment happened right before I arrived at HMRC, which was an enthusiastic early adopter of these techniques, and willing to pilot that. I will just say two things about this example.

First, we found that trying to generalise why you should pay your taxes did not work. It is not that you should not do that, because it is always a good idea to keep putting forward to the community, more broadly, why paying taxes is an important thing to do. But, the more local we could get, the more effective that strategy was. If we could truthfully say, ‘everybody else on you street has paid’, there were really good effects on taxpayer compliance. Second, we found that compliance following these trials dropped off very quickly for a very simple reason. We did not reinforce it by contacting people who ignored it. As soon as we started to do that, the response started to lift again because it was a cue to, ‘this time, you must pay attention’.

The next example is the so-called ‘Spooky Eyes’ or ‘Risky Business’ campaigns in the UK, and whether or not they were successful. This was the first such campaign that really caught the media’s attention, both in the UK and internationally. We were struggling with the hidden economy, as revenue administrations always do, but there was a perception that it was starting to grow. One of the things to note about the UK is that they measure their tax gap every year. HMRC had done lots of research on this and identified lots of factors about why it was growing, why people engage in it. We had one overwhelming factor in our research and that was: the people engaged in the hidden economy did not believe they could get caught. They did not believe HMRC had the skills to be able to detect those involved. They thought, once we had found them, that we had the skills to then be able to investigate them and be able to do whatever we needed to do, but they really did not think they would get caught.

The ‘Eyes’ campaign looked like a national campaign; we had posters on bus stops, train routes, and in supermarkets, at shopping centres, at bank ATMs. It actually was not everywhere. What we did is put the posters on routes that were most frequented, and locations most frequented, by those who were likely to engage in the hidden economy. We were implementing a principle of localising the response. The goal was to send the message that we now have lots of data and we are able to detect people, and it would be better if you came forward first, before we came and found you. We found that this campaign passed the ‘Pub Test’ – people in the pub were talking about it: ‘I did it; I came forward and I brought a reasonable outcome’, or ‘I did not and this is what happened to me’. Targeted task forces were put in the field so that they would increase presence to enforce the tax law. Does it work? So far, more than 50,000 taxpayers have come forward through that process.

The ‘Eyes’ campaign itself is gone, but the communications strategy is not. There was a combination of ‘come in, we will help you’ or ‘if you do not, face the consequences’. The tax debt did go down, marginally. We were hoping for containment of the problem, preventing growth.

My final example is called ‘Stop the profit’. This was a policy change initiative. When I arrived in the UK, as had been the case in Australia, a big unresolved issue was mass-marketed tax schemes that had been acquired by thousands of taxpayers, which had not been settled, despite the fact that in the matters that had gone to court HMRC won in over 80 per cent of these cases. This was important because there were five or six billion pounds of revenue at stake, and very poor public perception on the issue.

We identified that the reason people were not coming forward to settle the scheme matters was that they had a cash-flow advantage; even when they ultimately had to pay tax, they still had the advantage of the time value of money. We initiated a policy change which was to require payment of the assessed tax upfront. If you did not –if you chose not to, so it was not an absolute– then you would pay a penalty rate that was much higher.

Lessons from the journey include:

  1. Smart data exploitation skills are essential. You have got to know what your ability is to manipulate that data. Ministers will ask, ‘what about this? What about that? And could you just do this?’ If you have got to go back and say, ‘I am sorry, we are going to have to do another survey, it will take six months’, you have blown it. You do not want to be asking these questions at the last minute.
  1. Multi-expert teams to solve wicked problems. Most of the issues I have encountered in the last 10 years are multidimensional. They are not contained in the tax system only, and they are multidimensional within the tax system. Often, we will have a range of experts from other Departments. We will also have social psychologists, data scientists, communications experts, IT experts. It is a mistake to bring them all in at the end; you need to build that straight away.
  1. Base analysis on control groups, as already discussed in the BI literature and on this panel.
  1. Play the behaviour, not the motivation. You cannot measure motivation, you can measure behaviour. It is necessary to continually refresh and reset, not just because behaviour changes, but also because the taxpaying population changes. One third of small businesses turn over every three years. That means there is a third that is new and a third have gone. You are not dealing with the same groups.
  1. Be publicly transparent. There is a great story about how I was accused of ripping up the Magna Carta, in my role in HMRC. I can tell you some other time.

Miranda Stewart: We now turn back to the broader context of government. It was noteworthy to hear Jennie talk of multidimensional problems that arise across government, not just in the revenue agency. Maybe this is a whole of government question that we could put to David Gruen, as a Deputy Secretary of the Department of Prime Minister and Cabinet.

Another question is the issue of promoting experimentation, while reassuring citizens about the kind of certainty in their lives and their data.

David Gruen

Deputy Secretary, Prime Minister and Cabinet

I am not sure whether the fact that the two blokes on the panel came up behind this [podium] while the two women panellists did not tells you something. I suspect it is not statistically significant, but nevertheless I thought I would at least draw your attention to that. I want to talk to you a little bit about the behavioural economics team of the Australian Government, which goes by the acronym BETA, which many of you would know. Some of the people in the audience work in BETA, so they definitely know about it.

One of the interesting things about BETA is that it was not funded directly by Government. It has been funded by Departments who, as you would be aware, do not have a lot of spare money laying around. Departments have on the broad decided that this is something worth putting some money aside for. As Jane said, I think it may have been that the ATO was the first entity in the Australian Government to use BI. Since 2011, a couple of other Departments set up BI teams including Human Services, Health, and the Department of the Environment and Energy.

The idea of setting up a central unit within PM&C is more recent. BETA was initially funded by 19 partner agencies across the Commonwealth Government. It was a small team, at first. There are lots of governments that pay lip service to BI and then you discover there are two and a half people working on it. It is now the case that BETA has 30 people working in it, so it is a substantial team.

When it was set up, BETA committed to achieving a rate of return of 10 to 1. When I was told this, I thought –‘Good luck’. That is a lot, right? Ten dollars of saving to the Government for every dollar spent. So far, some areas are more obviously prospective than others. Tax is obviously one such area. And of course, the gains are not supposed to be only savings to the government. You hope that the trials are designed to have wider benefits, not just saving money for the government. The aim of a metric is to hold people to account and to see how we go against that metric.

Another major issue, which may not worry academics so much, but certainly worries policymakers, is convincing people to scale things up. You can run a trial and have very strong results, but to the extent that the BI team owns the trial, convincing the Department to put their resources aside and actually scale it up, which is where the big gains will come from, is not as straightforward as it might sound.

BETA has been a conduit for smart people with academic backgrounds to come into the public service. My attitude towards the public service is that most people outside the public service do not have much idea what we do, and do not think it is very interesting. It is only when you get inside that you find that it is genuinely fascinating. For those of you who are sceptical, do not take my word for it –ask (Professor Emeritus) Bob Gregory. Bob says most of the good ideas he had came from working in the public service and being asked questions by people who were grappling with issues that academics had not thought about, because they had not happened yet. They did not have any data, something is going on out there and governments are asking questions about ‘what do we do?’

We have attracted quite a few people like that into the public service. And once we have got them, we show them just how interesting public policy is. And sometimes it sticks.

We have a ‘publish by default’ rule for BETA trials. As you might imagine if you know anything about government, that is challenging. So far, so good, but that has something to do with how you choose your trials. No government wants to be told that their pet intervention does not work. So that is tricky.

One trial that is kind of cool is the antimicrobial resistance trial, where we got the Chief Medical Officer to write letters to the GPs who were in the top 30 per cent in prescribing antibiotics. We did it by area. Prescribing rates are different depending on the socioeconomic nature of where you are prescribing: the data shows that low socioeconomic areas tend to have higher prescribing rates than high socioeconomic areas. The best designed letter reduced prescribing rates over the next six months by 12 per cent, relative to the control. The trial is still going, so we will find out whether the effect lasts or whether it is disappearing. The trial has all sorts of neat features. A bigger effect was produced for the people in the top half of that 30 per cent, relative to the people between 15 and 30 per cent. The savings on the Pharmaceutical Benefits Scheme are large. And of course, there is a very big positive spillover. Antimicrobial resistance is a big deal, so to the extent that you can reduce unneeded prescribing of antibiotics, there is a public benefit.

Another trial, which is not yet published is aimed at credit card use. BETA has teamed up with the Treasury and a credit card provider –a big bank– to engage with credit card customers who pay the minimum balance every month. Everybody would be aware that the interest rate charged on credit cards is enormous, and it is almost certainly not in people’s best interest to pay the minimum amount if they can pay more. An SMS sent to people just before they get their credit card bill raises the amount they pay back, on average, by one quarter. That is just a preliminary result, but it is a very big effect.

Doing randomised trials is part of the work, but one of the things that BETA is increasingly going to do is think about markets where BI can tell you something even if you cannot do a randomised trial. My favourite example is markets where it is in the producers’ interest to make their offer as complicated as possible, so the consumers have no idea what they are doing. Examples are mobile phone plans, or the offers that the retail sector provides on electricity.

In their recent report, the Australian Competition and Consumer Commission (ACCC) said that electricity retailers ‘have made pricing structures confusing and have developed a practice of discounting which is opaque and is not comparable across the market’. We know why they have done that: because the last thing they want you to do is actually understand the nature of the offer, because then you are likely to get the best deal. In the modern world there are quite a lot of markets like these. Without necessarily running a randomised trial, it is going to be BI that allows you to think about what are the right interventions.

Clearly, Homo economicus (or ‘economic man’) has no problem with this. Homo economicus just calculates it all and works out the best plan. Well, good luck to them. But the rest of us just do whatever. Thinking from a behavioural point of view, how do you intervene in this market in a way that is light touch, but improves the chances that consumers actually get good deals? It might be Trip Advisor style. You can imagine a lot of potential things you may do where people teach each other what works, but leaving it just as it is, is almost certainly not Pareto optimal.

Miranda Stewart: This closing panel gave us an opportunity to hear from leaders from policy and political arenas. As observed by Alex Gyani in his keynote speech, an interesting question about BI in government is not, ‘how can we make people do stuff?’, but ‘why are they not doing it already?’. That reframing of the question nudges researchers and policymakers to think differently about what government does. The answer might not always be psychological; it might be legal, or structural, or economics, leading to different policy and administrative approaches.

Adam Oliver (Associate Professor at the London School of Economics and Political Science) in his keynote asked whether behavioural insights are changing the whole field of economics, or whether there is a specific field of BI and the rest of economic analysis remains unchanged. Can we say the same thing about public policy? Do behavioural insights, and behavioural research, change the whole way we think about policy, and about what government can and cannot do? Or it is just one tool in a policy toolkit? Our speakers have raised pertinent issues and arguments about the multi-dimensional, interdisciplinary and practical issues for policy makers and administrative agencies, to bring BI usefully into government for the future. Thanks very much to all our panellists.


Further reading

Behavioural Insights and Public Policy: A Discussion – Part 1, by Maria Sandoval Guzman

Nudging Businesses to Pay Their Taxes: Does Timing Matter?, by Christian Gillitzer and Mathias Sinning

Leave a comment

Your email address will not be published. Required fields are marked *