BETA Podcast: Interview with Michael Sanders

Elaine Ung:

Hello and welcome to another episode in Beta's podcast series. Beta is the behavioural economics team of the Australian government, and the team sits in the federal department of the Prime Minister and Cabinet. Beta's mission is to advance the wellbeing of Australians by applying behavioural insights to public policy and administration.

Hi, my name's Elaine and I'm part of the team. In this podcast series, we interview a whole range of behavioural economics and behavioural insights, academics, practitioners, and policy makers. This episode is actually part of our mini BX series, where we interviewed some of the speakers at the 2018 behavioural exchange conference or BX 2018. There, speakers travelled from all around the world to share the work they've been doing in the behavioural economics and insights field.

In this episode I interview Michael Sanders, who is the chief scientist and director of research evaluation and social action at the behavioural insights team, also known as BIT. As chief scientist Michael and his team supports BIT with the design of randomised controlled trials. He's also part of BIT's ventures team working on BIT's online platform like test and build. I hope you enjoy.

Hello. I'm at the International Convention Centre in Sydney at the Behavioural Exchange Conference, BX 2018. I have with me Michael Sanders. He's the chief scientist of the Behavioural Insights Team. Welcome. Could you please give us a bit of background, so who you are and where you come from?

Michael Sanders:

Sure. So I am the chief scientist at BIT as you said, and that means that I run or coordinate our research and evaluation operation globally. So that's our design and running ... I've done running of randomised control trials, a lot of our work in social capital, our machine learning and data science work, our qualitative research team. Pretty much anything that comes under those auspices. In London, in New York, in Sydney, in Wellington and in Singapore.

Elaine Ung:

Well, so lots of broad work happening there. And how did you become interested in behavioural economics?

Michael Sanders:

So, I graduated from my economics undergraduate in 2008. which you know was a very, very long time ago now. But it was also at the same time as while we were being taught that people are rational and people respond to incentives and there can't be runs on British banks, because the system of incentives is set up such that that just can't happen. As we were being taught that outside in the real world, the financial system was collapsing and there was a run on northern rock.

I had this very visceral reaction to, on the one hand, this is the way the world is supposed to be according to economics and this is the way the world is actually behaving. It's completely different from one another. And so I sort of started casting around for alternative explanations of the world and that sort of landed me on behavioural economics. And so I then went on to do a PhD in behavioural economics and Philanthropy and charitable giving. And I was about 18 months into that when an opportunity to come to the behavioural insights team for three months on comment came up to work on charitable giving research. So I applied for that. Got it. And then six and a half years later here I am still hearing David Halpern keeps paying my salary, so no one's fired me, so I thought I should carry on doing the work.

Elaine Ung:

Wow. That's a great journey so far. And you've spoken about some of the broad work that you do. Could you share with us what you're working on right now?

Michael Sanders:

Sure. There are a few things I think I'm really excited about. The first of which is our work in children's social care and machine learning. So Children's social care is an area where we're talking about in the UK, tens or hundreds of thousands of the most vulnerable young people, and an area where there isn't a lot of good empirical evidence. And what we're able to do using modern machine learning techniques like topic modelling and gradient boosted trees is really get into a ... We have a lot of information in textual case nodes. As you start shining the light of empiricism into that space. We're able to predict, for example, which cases are likely to escalate to become more serious and more serious action need to be taken up to six months in advance. And that means we can better support families and children to sort of avoid those very negative outcomes.

I'm really excited about that. Another thing that I'm really excited about is in what we're doing in social connectedness. This is based on a trial we ran last year with the National Citizen Service, which is a youth volunteering programme for 17 year olds. We did a trial last year with 6,000 people where four weeks before the summer programme, which is ... It's a great programme, but some people don't turn up. Those are typically people from ethnic minorities or from lower income families. And so the programme wants to boost their turn up rate. So what we did was we paired people up beforehand for weeks, before we say, Hey Elaine, Michael, your buddies now. Why don't you chat on this online platform that we built at BIT for you to communicate on? So we'd chat and we'd say, hi, what are you doing? What are you looking forward to?

We send them behavioural prompts every week to say, oh, you should be packing about now. Why don't you talk to Michael about the things that he's packing? Make sure you don't forget your trainers or your sneakers or whatever you say, in Australia for the shoes you wear for running. And so what we found there was that that reduced the dropout rate for young people from ... Well, people in general by about 25%, which is a pretty huge effect. What's really fascinating and which went actually against my prior expectation at the beginning was that who you're buddied with really matters, but not in the way that I thought it would. So we thought of ... If you pair somebody from a low income family with somebody else from a low income family, they're going in with the premade friend, they got something in common. It helps bridge that social divide.

And so that's going to be more effective than pairing somebody from a low income family with somebody from a more affluent background. Actually that's just not the case. What we find is the effects are much larger, much more pronounced people who get paired with someone who is different from them. So a person from a low income family pair somebody from a more affluent family. That we are bridging that social divide, you using the programme in a safe environment where people can leave when they want to, where it's all anonymous. That is really sort of we're seeing twice or three times as large effects when we buddy people who are people who are dissimilar to them.

And so this is something which we think could be hugely impactful moving forward. So we're testing this out now in a much larger scale study with 100,000 young people. So that's quite terrifying for us. We're also trying it with graduate programmes with large employers because again, we see people from lower income families or ethnic minorities who like to show up to those, even if they get the places. And then at universities as well, where we see exactly the same thing. So this is something that I'm really excited that over the next couple of years to see really come to fruition.

Elaine Ung:

Wow, that sounds fantastic. And you mentioned impact and I can see that it has a lot of potential for great social impact as well. So that's really fantastic. We've all heard of those trials that BIT has done on, for example, tax and stuff like that. But could you share with us another example of your work or of bits work that demonstrates the potential impact of behavioural economics?

Michael Sanders:

Sure. So one of my favourite studies that I've been part of in the time I've been at BIT is one, which we did a few years ago, working with young people who are ... They get really good grades at GCSE, is the highest stakes exams you take when you're 16. they're in the top 20% nationally. They're from lower income families and they go to schools where typically people don't go on to universities or they don't apply to good or what we consider to be 'elite' universities. Part of that, again, it's about social distance. If my mental image of somebody who goes to Oxford or Cambridge or the LSE is somebody's posh, white, maybe he wears the bow tie someone a bit like me, I guess, then maybe I'm less likely to apply to that institution. So what we did was we found in the undergraduate cohort at the University of Bristol, two young people, so Rachel and Ben, who are from that kind of background and went on to to go to Bristol.

We got them to write letters essentially to their past self, saying, Hey, I was in your shoes once, I got good grades. I was thinking about going to X university or not going, but thankfully I had one teacher who really inspired me and now I'm at Bristol. I love it. Oh, and what's more, you can get all these financial support and these better universities actually really, really want people like you to be here because they really value social mix. So we've got them to write those letters. Then we then got the department for education who we collaborated with on this, to print up basically 11,000 copies of them and we sent them out to the kids who were in that space at the moment, they're 16, 17 years old.

They're just about thinking to go to university and then they get a letter from Ben or Rachel.

So for trial control group getting their letter. One group gets a letter from Ben, one group gets a letter from Rachel, and one group has letter from both Rachel and from Ben. So they're spaced out a bit. It's not two letters on the same day. So that's what we did. Just fast forward three years to get the actual results through and what we see is applications to be accepted by actually accepting that place. So basically going to these elite universities, the Russell Group in the UK, and we see that the rate of that goes from 8.5% to 11.4%, which is absolutely huge increase. And you know, as Kassar said in his talks, like that's number, but the people there is 332 young people a year extra from lower income families who are going to go to those elite universities. This is important for a couple of reasons.

Firstly, it's great to have more young people from lower income families go to those universities. They're likely to get better jobs, they are likely to get paid more. It's a piece of social mobility action. But we also know that people who do jobs like ours in central government, civil services are drawn from those elite universities and there's more selective universities. And that means that hopefully some of those 332 people will end up doing jobs a bit like ours. And they'll end up in central government. And to the extent that we think that the central government is going to function better if it better represents the people who it's supposed to be serving, then we can hopefully see like even more powerful, longer term consequences.

Elaine Ung:

That's really amazing. What a fantastic project to have worked on. Do you have a favourite behavioural economics concept or a favourite behavioural bias?

Michael Sanders:

A favourite behavioural bias? Should I have a favourite behavioural biases? Biases are bad, right? So the one I'm writing I'm with Susanna Hume, my dear friend and colleague at BIT and at King's College London, a book about social influence. So I think I would say that the class of things I'm most interested in in social influence. At the moment we have this sort of perception that social influence is bad, right? So you've got Cambridge Analytica, you've got Facebook, our elections are being manipulated. We're being put into echo chambers by these things. Things like Instagram, which basically feeds off of our social motivations are really bad for our self-esteem and our body image, for example. So the current malaise about social influences is a bad thing. However, if you look at the behavioural insights literature and a lot of the work that's been done by teams like us at BIT and you have Beta and other people, is social implements can be harnessed for such enormous good.

I've read a good example of that other than the ones I've already spoken about, is some work done by my doctoral student BB Grout, where we get people to nominate. People are getting an educational context. They nominate people who can support them in their education and then we send them messages every week. This could be somebody who ... it's definitely somebody who's already in their life, somebody who already cares about them. The kind of person who, if they said, I'm feeling ill, can you come around and like just look after me a bit. Or Oh, I've got really important theatre performance or football game or cricket match. Depending on your preferences. Then this kind of person would go along to that.

But that kind of social connectedness, social capital is being directed towards you other than education. So we think it could be reapplied in perhaps a better way. And so what we do is we buddy people up and then they get these messages over the course of the year. And what we see as this huge impact on GCSE grades, which can deal with people's ability to progress in life. One of the things we are seeing so often is that social influence can be such an enormous force for good if we think about it and if we apply it correctly. And that's one of the things that behavioural insights teams around the world who've really been doing and something I'm really keen to see continue.

Elaine Ung:

I think that's right. We see a lot of the different behavioural biases can be used for good and bad, and it really does depend on the way you frame things. And that's partly where a lot of the ethics come in, because we need to make sure that we are actually doing, we need to know the outcome that we want, we're seeking and take that approach from the very beginning. So that's really important. Thank you for sharing. Could you share how you apply behavioural economics in your everyday life?

Michael Sanders:

That's a difficult question. So I'm a rational automata. So I don't need to hear economics. Surely, I'm an economist several times over. No. So I think the main thing is I'm so susceptible to plant applying fallacy. So I will say if you asked me to do X, Y, or Z. I'll say yes to absolutely everything. And so I have to recognise the fact that time in the future is not an infinite resource. And so I have a lot ... I do a lot more stuff around trying to motivate my own planning, having decision rules now.

Particularly now that I have a young child, I will say no to 75% of talks that I am offered. I'm only allowed to accept a certain number of invitations to speak overseas a year in order to limit my own sort of chance to be like, yes, sure, of course I can absolutely do everything. Why wouldn't I? So, you know, important life lesson for everyone. Don't have a child, get a new job. And write a book in the same year. Free Advice. Past me was a flow decision maker. Present me thanks to this behavioural insight is a perfect decision maker and I'll never make that mistake again. Honest.

Elaine Ung:

Great tip. Thank you. And throughout the last six and a half years that BIT, could you share some reflections on the evolution of behavioural economics and behavioural insights, especially in public policy and also where do you see BE going in the field?

Michael Sanders:

So I'm very fortunate that when I joined a lot of my colleagues, so Sam Hain, Simon Ruda, Felicity Holgate, Michael Hallsworth, Owain Service, and David Halpern had done a lot of the really hard graft of convincing people. So Michael Holdsworth in particular ran this text letter study. Now that I have people that take some time. And that was really the moment where behavioural insights came alive in the minds of a lot of people in government. So I sort of arrived, it was still a little bit of trench warfare, but like the back had been broken of many people in terms of being like as a tool for good. I think in terms of the evolution over that time, we have seen a gradual rolling out within the domestic policy sphere in the UK and internationally of convincing more and more people. And that sort of is generally a not here kind of argument where, Oh sure that works in the UK, but it's not going to work here in Guatemala or Moldova.

Or, oh yeah. It's fine in Britain where you have all of these like ingrained social problems. But here in Australia, one of the most diverse and developed countries in the world that certainly isn't going to work. Or in Singapore, which has this sort of hole is a very different sort of environment in many ways. Or in the domestic policy sphere, it's like, oh, it works in tax, sure, but it's not going work in education. Oh, it works in education but it's not going to work in children's social care, or oh, it's not going to work in development policy. I think that we've gradually been rolling out and convincing people piece by piece through ... The best convincing tool that we've got and the engine of this revolution really is the bar chart. So somebody comes in and says, we did this thing, control group did x, treatment group did x plus 5% and up as good in this case.

And people just start to say, oh, that's that many kids or that that many dollars or that many pounds or that many antibiotics as prescribed. That is the way this is [inaudible 00:15:24]. It's not actually very sexy. It's like, here's a bar chart and this is the engine of our revolution. But it has been, I think, just phenomenally successful so far. And so it was heartening to come to be X every year and see, oh, and now that's been done, this 10, 15 other countries. Or oh, this is how someone's taking this idea and really run with it. So I think that's about the evolution so far. I think what I want to see in terms of the future is I guess a couple of different things. The first is there are big important policy areas where we haven't done enough yet. So in the UK that feels like children's social care.

It feels like a social connectedness across a range of dimensions. I understand that in Australia there's a lot of work being done by our colleagues on domestic violence, which is a huge and so important issue. And we haven't really grappled with these big, very, very important items yet. I'm really gratified to see us doing that, but I just want to see more and more and more of it. The other thing is around machine learning in the data science, which is very cool at the moment. And I think a lot of us are running off and be like, okay, let's use a machine learning algorithm will predict X, Y, and Z, and that's really half of the discussion at best. The other half of the discussion has to be what next? How do we put this into the hands of practitioners? So in general and social care, for example, we created an algorithm that tells you that this particular kid has a very high risk of that case escalating.

How do we give that to social workers in the way that they're actually going to listen to it where it's appropriate but not to too much. And then how do we tell them what to do next? Right? So machine learning algorithms by definition, almost produced black boxes. How do we open up that black box in a way that doesn't require you to have advanced degrees in statistics in order to be able to tell what's going on? So the social worker practitioners who are incredibly busy on the front lines are able to actually use it in their everyday practise.

And that isn't a statistical issue. That's a human issue. That's a behavioural insights or behavioural economics issue. So I think that's the real exciting frontier for me. So the thing that I've been saying probably far too much over the last few weeks is that behavioural economics has been about putting humanity back into economics, behavioural insights, and about putting humanity back into public policy and what the next frontier for us is putting humanity back into robots and machine learning. So that's sort of where I hope we're going to go next.

Elaine Ung:

It's absolutely an exciting time to be working in the economics and behavioural insights fields. And I think coming together at BX, it's really exciting to hear what other people are doing and the different challenges that everyone's facing, which are actually all very similar. So it's good to have that learning across the board.

Michael Sanders:

They are all very similar. There are also substantial differences, right? So one of the ... every year in my class, which I teach, we talk about should we nudge people who vote, and in the UK it's the public policy. We don't nudge people to vote, we might nudge them towards voter registration, because you don't feel comfortable with government getting involved in that kind of business. If you talk to Americans, sort of people like Todd Rogers and Don Green and others have done a lot of work in nudging people to vote and get out the vote, and typically on the democratic side of that debate.

And so they think obviously that's fine to nudge. You're not going too far at all. And then I talk to Australians, I hear that, well why would you nudge someone to vote? Why don't you just mandate it? And that seems to work absolutely fine. And so there are all these interesting differences where we consider different problems. We consider different nudges and whether or not we reach into the basket of traditional government tools, legislation, taxation and information, or whether we reached for a nudge. There is a lot between contacts. I think that is a really interesting thing to try and bottom out at this kind of conference when we're all here together.

Elaine Ung:

Yes, absolutely. And one final question. What's next on the agenda for you?

Michael Sanders:

So finishing this book in the next two months is the really big thing. The other thing is the scaling up of our work in general and social care. This is so, so important. We have such a large population of vulnerable children who we don't do enough for and who I personally feel like the behavioural insights field hasn't done enough for yet. It's going to be so important to get it right and to go carefully. But you know, we're scaling up from one pilot local authority last year to something like 10 pilot local authorities this year. We're looking to keep expanding the out building digital tools that can be put into the hands of social workers and their managers to try and really improve outcomes in a meaningful way. And then of course to evaluate those digital tools using randomised controlled trials. So for me, that's the frontier I'm most excited at pushing them.

Elaine Ung:

Well, thank you so much for sharing, and all the best with all the work you're doing. Can't wait to see what comes out of it. So thanks again.

Michael Sanders:

Thank you.

Elaine Ung:

Thank you Michael. I hope you enjoy the rest of BX.

Michael Sanders:

Thank you.

Elaine Ung:

Hi, again. Thanks for tuning in. If you want to learn more about Michael's work, you can read about the work and research he's done on the Behavioural Insights Team's website.

Thanks for tuning in. If you haven't heard our previous episodes, listen to them at www.pmc.gov.edu/beta. Until next time.