Nudging Towards The Good: In Conversation with Gus O'Donnell

The former Head of the Home Civil Service in the UK discusses behavioural approaches to public engagement, policymaking and sound values for the civil service.

Date Posted

15 Jun 2014


Issue 13, 14 May 2014

On Social Media, Sense-making and Getting a Feel for Ground Sentiment

There are various people now starting to monitor public sentiment. The private sector was the first to do a semantic analysis of social media, for example, linking certain trends in Twitter sentiment to movements in the stock exchange. The Bank of England tried to analyse all of their reports running up to the financial crisis to see if they could pick up any clues from the tone: interestingly there was an increase in worry but nothing was translated into action. Unless someone is watching these trends and doing something about them, it makes no difference. Nowadays of course, once you put something out there, everyone starts looking at it, which distorts things and can be manipulated, so you have to be quite careful. So far this has been pretty limited for government use. Some political parties, for instance, may use social data analysis, taking a cue from the Obama elections and what we can learn based on behavioural sciences, in order to target communications to certain groups.

I think it could be useful as another way to pick up indications about some of the concerns that citizens have. You may have to wonder about how representative social media is, but it seems like it could be much more representative in Singapore than in the UK. The question is how to get the views of those who are not as well represented.

If you are trying to get people’s honest feedback, you should make sure that the one asking for the feedback is perceived as completely neutral.

The hardest thing, I found, was for the government to get money to carry out proper evaluations for implemented policies — you’re doing well if you get to pilot something. So how do you found out whether your policy is working? Quite often there is going to be an enormous lag between a policy implementation, say a change to the benefits system, and before you get any useable data. So how do you get a more immediate sense?

In the UK, there is a Citizens Advice Bureau: basically a free walk-in agency where someone can come in for advice on legal, financial and other issues. It is an independent organisation, but they work with government. The Bureau sees more people who have difficulties or are upset, and is able to compile real data on policies and people’s reactions to them across the country, and identify things that are not working well. So you get a kind of live measure of what the issues are, for those people affected by these policies.

If you are trying to get people’s honest feedback, you should make sure that the one asking for the feedback is perceived as completely neutral. Behaviourally, we find that people don’t tend to give honest feedback to someone who is in a position of authority relative to themselves. We find that individuals, who are quite good at giving honest feedback to people working for them, are terrible at giving honest feedback to their boss. They feel that if they criticised their boss, this would be held against them and you can understand why. So you have to find alternative channels and incentives. You could ask whether certain things should continue or not — something most people would give feedback on. You explore different ways to find out what people are generally prepared to give you an answer about, and then you work with that information.

The number one behavioural principle in all of this is to keep it simple. If you build feedback mechanisms into a service system by default, as a prompted choice, you are more likely to get accurate, immediate and comprehensive results rather than ones that are mainly from those who have strong views one way or another.

On Using Big Data and Behavioural Methods as Policy Levers

The idea behind behavioural levers is that you nudge people in the right direction without having to put in place regulations to do so.

For example, giving people immediate contextual information — a little help at the point at which they make decisions — turns out to be more useful than offering abstract information, such as from a course on financial planning. A new book, Behavioural Public Policy,1 examines this sort of phenomena, such as how people can become psychologically numb to large numbers.

We can crunch big data sets to look at the needs of individuals across different government services and offer it to them at the relevant time. If the data is in the aggregate, it is to everyone’s advantage and there is no privacy problem. But for personalised information, privacy remains an issue. The UK has many of the same privacy laws that prevent the sharing of data. However, there is an idea in the UK about making any machine data on individuals, but held by companies, available to the former on request. Now, that’s not much use to them individually, but what we are hoping is that this will enable a market.

So for instance you go to a supermarket and your trolley has an app that allows you to put in your own private information, such as your preferences, your budget, your diet, your health data, your allergies, whether you want to reduce your calories or avoid certain products and so on. And then as you zip around, the app on the trolley can keep track of your shopping and give you reminders or advice based on your needs — whether you really want the non-vegetarian or high cholesterol item and so on.

Then looking at the public sector, you can imagine data sets from different areas being put together to tell you how your financial planning is going, and what recommendations can be made based on your needs. For example, the one thing that seems to work is when you can help people make comparisons between similar products and services, such as insurance or energy companies, when they are making purchase decisions.

We could also use this data to nudge citizens towards behaviours that represent better outcomes. For instance, we could encourage car owners to switch to a more fuel-efficient, lower emissions car than what they’ve got, and save money in the process. This sort of cross-selling is very underdeveloped in the public sector.

The flip in paradigm is that instead of being told what to do, you decide on your long-term goals and how your own data is going to be used to help you achieve your own goals — lose weight or save money and so on. You give citizens their own data back in ways that they can use to make informed decisions. This also gets around the privacy issues surrounding personal data, which people can be very suspicious about. Once people see the advantage in doing this, hopefully it can become a norm. The public sector can help overcome the obstacles that prevent these tools and services from being developed, or to lower switching costs between providers.

Private sector financial services in particular are starting to use behavioural insights on customers. In the UK, the Financial Conduct Authority is looking into whether they are exploiting these to the detriment of customers, particularly those who are less financially literate. So if they are making very clever behavioural nudges to get customers to buy what they don’t need or cannot afford, we need to nudge them back. However, you have to be incredibly careful not to starve innovation that can leave people better off. Instead, the government should look out for areas where there is potential for abuse and address those.

On the Need for Civil Services to be More Politically Savvy and Sensitive to the Public

Civil servants have to be very good at explaining and giving their political masters the raw material they need. So for most of the policies we make, we need to have the distributional numbers. Economists are very bad at this because they do cost-benefit analysis with the tendency to assume that the benefits can all be added up irrespective of who they go to, and that cost can all be added up irrespective of who bears it. But diminishing marginal utility means things do not add up that way.

What good public servants should do is to work out who the winners and the losers are. We can expect some groups to be more vocal than others and so on. Can we do something in advance? What can we say to those who will be worse off after the policy?

I think it goes back to your communication style. People mostly get upset when their expectations have been confounded. So if you warn people in advance that they may be adversely affected, together with the reasons why you need to do this and what you are trying to do, you give people a better chance to think. You probably need to explain in terms of stories what is going to happen to them and why, or have the families themselves — rather than the government — tell the stories, which would be much more powerful.

There are times when politics trumps good economics. In the UK, there was a policy on winter fuel payments where everyone above a certain age got money from the government for winter fuel. It wasn’t related to the amount of fuel they used or to their income. The policy was expensive and indiscriminate but politically the government thought it was doing the right thing in helping the aged. Although it would have only taken a fraction of the cost to help the poorest who needed it, the politicians were committed to keeping the policy come election time. I call this history as policy: once you introduce a poor policy, it’s hard to get rid of it later.

So what could the civil servants have done? I think we should have come up with a better package which would have targeted the needy old. Instead of just saying that it’s a terrible policy and we shouldn’t do it, we should have persuaded the politicians that we understand they want to help the aged but there are better solutions. In this case, we didn’t have enough time to come up with a good alternative. Part of the problem is that the civil service mindset is often not proactive enough on policy: this wasn’t a public issue we had thought about.

You have to be incredibly careful not to starve innovation that can leave people better off.

On Nurturing Healthy Values Within the Civil Service

For the top level of the civil service, say among the top 200 employees, there is a lot of reflection on values, particularly when it is coming up to the elections. For the UK, a change in government is a real option with every election, and we need to think about how we might cope with changes in administration, particularly for those who have not seen a transition for a long time.

We also sort for values at recruitment. One aspect of the UK Civil Service Fast Stream is what is called an e-Tray exercise: a timed set of scenario-based questions. It might be a flu epidemic and you have a limited supply of flu medicine and so on. A question is posed on how you might approach the problem, with four possible answers. One of the answers might be that the optimal way to do this is to sell off the flu vaccine, because the people who want them the most will pay the most. Another might be that you distribute it according to those groups who benefit most from the vaccine. Another answer might be that you can’t make any of these judgements because the underlying data is not good enough so you have to give it out first come, first served. Values are implicit in each of these options, and you have to pick the best and worst options out of the four.

The way we do things today evolved in a different age, and are geared to an older model based on theoretical economics, not human beings.

The answers get marked based on our assessment of the options and how we think we would want our civil servants to answer such questions. Interestingly, the e-Tray exercise is probably the biggest differentiator for recruitment: it’s not grades, because they all have good grades. So values are at the centre of identifying the civil service leadership of the future. We are saying: here’s what you do, you lead. Here’s how you do it: by using our values.

What would I say to bright young civil servants coming in today? I would want them to challenge. There are so many aspects of what we do that have just developed historically. The way we do things today evolved in a different age, geared to an older model based on theoretical economics, not human beings. I want them to understand the power of behavioural insights and understand that it is possible to change things. The world is changing so let’s use all this new data and new tools, and try new ideas. We need to be more innovative and take some risks, and try to build a system that will allow for more failure. If I were a young civil servant coming in, I would want to think that even if I try something new and fail, it is not going to be the end of me.

In his second visit as CSC Senior Visiting Fellow in November 2013, Lord Gus shared these views in conversation with ETHOS Editor-in-Chief Alvin Pang, CSC Senior Researcher Vernie Oliveiro, futurist and Deputy Director in the Futures & Strategy Division Lee Chor Pharn from the Ministry of Trade and Industry and Jonathan Ng, Senior Assistant Director from the Strategic Policy Office, Public Service Division.


Lord Augustine O’Donnell has been non-Executive Chairman of Frontier (Europe) since October 2013. After joining the Treasury in 1979, Lord Gus held various positions at the British Embassy in Washington, the International Monetary Fund and the World Bank. From 2002 to 2005, he was Permanent Secretary at the Treasury and in 2005 became Cabinet Secretary. He held this position until 2011, serving three Prime Ministers. Lord Gus studied economics at the University of Warwick and holds an MPhil in Economics from Nuffield College, Oxford. He lectured in economics at the University of Glasgow.


  1. Behavioural Public Policy, edited by Adam Oliver (UK: Cambridge University Press, 2013),