Oral evidence: Fraud and error stocktake, HC 394

Oral evidence, 14 September 2015.

Witnesses

Robert Devereux, Permanent Secretary, Department for Work and Pensions, Jon Fundrey, Senior Responsible Officer, Fraud, Error and Debt Programme, DWP, Lin Homer, Chief Executive and Permanent Secretary, HM Revenue and Customs, and Nick Lodge, Director General, Benefits and Credits, HMRC, gave evidence.

Chair

Welcome, everyone, to the Public Accounts Committee. We are discussing today the NAO Report on fraud and error with HMRC and the Department for Work and Pensions. I am pleased to welcome Robert Devereux, permanent secretary at DWP, and Jon Fundrey, the senior responsible officer for the fraud, error and debt programme at that Department. I also welcome back, within a week, Lin Homer, the chief executive and permanent secretary of HM Revenue and Customs, along with Nick Lodge, who is her director general for benefits and credits and the SRO for fraud and error at HMRC. We have a lot to get through this afternoon and we hope to finish before 6 pm, so if witnesses can be brief in their answers, we will endeavour to be brief in our questions. I am going to hand straight over to Nigel Mills.

Nigel Mills

We will start with you, Ms Homer, because you probably have the better story to tell of the two Departments—your fraud and error rate is down a reasonable amount in the past few years. Can you talk us through how you have achieved that? It is not entirely clear exactly from the NAO Report.

Lin Homer

It is probably fair to say that we had slightly more to go at than DWP, which I am sure has something to do with the difference in progress. As I mentioned in the sitting last week, there was quite a good chapter on fraud and error in the annual report, which we did not deal with last week because this hearing was coming up. In some ways, that chapter gives quite a good summary of what we have been doing and identifies that we had a fairly clear strategy about targeting particularly high-risk areas, focusing our interventions on those areas, and being prepared to use design of both policy and process to improve what we were doing, while also trying to evaluate what we were doing in order to work out what was working and what was not. Nick could give you more detail on many of those points, but I would say that we are very clear that there are still some fairly big challenges ahead for us, one of which is the significant time lapse between when we do something and when we know whether it has worked. That can mean that it takes us quite a long time to finesse things. The other thing is that we are still struggling—as are, I know, Robert and DWP—on undeclared partners. That remains a big risk for us, although we do think we have made some progress on that over the time period. So, we have been focusing on strategy, we have used targeted interventions and we have been prepared to do policy process and productivity, but we recognise that there is still some way to go. The final thing is that we have not focused absolutely on a complete redesign of the policy because tax credit has only a limited life. That was our strategic response to the problems with tax credit, which is one of the most complex and difficult-to-operate processes that I have been involved with. The longer-term strategic answer is for it to be absorbed into universal credit.

Nigel Mills

You mentioned undisclosed partners. A decent chunk of the reduction was down to you restating your estimate on that. Are you confident that that restatement was valid and will not be reversed back again at some point?

Lin Homer

We are trying to understand it fully, but our initial look at it does focus on the fact that we have done quite a lot more work on undisclosed partners, so we do feel that we have been bearing down on that rather more. Again, if you want some more detail, perhaps this would be a good point for me to hand over to Nick. We have significantly increased the number of interventions generally, but that has included interventions in this area. We are still looking at that and trying to understand it. Our rate of progress largely trends downwards but occasionally turns upwards across all six areas of risk, but we are confident that there is an underlying downward trend. I suspect there is more for us to learn. Have I missed any salient points, Nick?

Nick Lodge

No, but perhaps I can just elaborate a bit. We have done a fair bit of work and will be taking the NAO through all of the detail in the coming weeks, just to make sure that we are very satisfied that the reduction in the figure from 7% to 5.3% is right, and we believe that it is. We understand some of the underlying reasons for that. There is a bit more analysis to really understand some of the detail. For example, we know that there were some appeals outstanding at the point that we first did the calculation. We know that there were some unfinished cases. We know that we had to take into account a bit more accurately than we did something called notional entitlement—I am very happy to expand on that if you would like me to. There are a variety of reasons, but we are satisfied that the 5.3% figure is robust.

Nigel Mills

So what is your target, going forward? You have got down to 4.4%, so you are ahead of the aim you set yourselves as a Department. Where do you want to get in this Parliament? There will be a lot of tax credits still around by the end of this Parliament, I suspect. It is still a big number.

Lin Homer

At the moment, all Departments are working towards the spending review in the autumn. We are doing a lot of work with our colleagues in Treasury, and obviously with Ministers, about what our spending review will look like. I think we will be better able to look at what projections we think we should be achieving thereafter, but in overall terms we are very determined to try to ensure that we continue to work with colleagues in DWP so that we are handing over tax credits in the best shape we can as it moves into universal credit. Although the final transition out of tax credits and into wholly UC is a few years away, it is ramping up. It is really important that that collaborative work between us at least maintains if not improves the progress we have made, so that we are vesting the legacy of tax credits in as good a shape as we can into DWP.

Nigel Mills

I am not totally sure why the spending review means that you cannot have a view on how low you think you can get fraud and error. It is down to 4.4%. Presumably zero is a long way too optimistic, but is 1% or 2% achievable?

Lin Homer

I think it’s early days. We have had some significant policy changes, as well as resourcing changes, and in some ways what we might all anticipate is that the transitional period will give us some additional challenges. We are very determined to keep making progress, but it is just the wrong point in the cycle for me to be putting a wet finger in the air. It needs to be subject to that debate with Government later in the year.

Nigel Mills

So when do you think you will be able to set a target?

Lin Homer

I am hoping the spending review will be settled quickly and cleanly in the autumn, and I am sure I will have a revised remit letter from the Chancellor as a consequence of it.

Nigel Mills

It looks pretty clear from the evidence that having a target in place focused the Department’s mind on making progress on this. I am not quite sure I understand why, given the work you have done so far, you don’t think, “Great. We’ve got to 4.4%, but we can get to 3% in the next three years with these initiatives.” It just seems a bit of a drift, doesn’t it?

Lin Homer

As I say, one of the challenges for us is that our learning is all lagged 14 months. We have had some fairly complex policy changes to implement in the summer Budget. Some of the disregards have changed. Some of the thresholds have changed. We think it is wise for us to try to have a look at our entire agenda before we make projections about the future, but you have my assurance: we do intend to keep focusing on this. We are very clear, even with the progress we have made, that there is a significant amount of spend covered by the fraud and error figures that could be spent on other things if it was reduced further.

Nigel Mills

Of course, your level of fraud and error is still higher than the working-age average of the DWP.

Lin Homer

As I said, we had a bit more to go at. It is getting closer, but there is clearly progress to go before we are at that level. Indeed, there are gains from universal credit itself that will help in relation to some of the in-built problems with tax credits.

Nigel Mills

I assume you would quite like not to have your accounts qualified for this issue if you could achieve it, but it does not sound like you have a target or a plan to achieve a target, so it does not strike me that you will be avoiding it any time soon. Is that qualification something that keeps you awake at night? Is it a priority to get rid of it?

Lin Homer

It is something we discuss regularly with the NAO. The accounts have been qualified every year since tax credits were in existence, initially in the trust account and now in the resource account. We have slightly asked that question of our colleagues in the NAO: given the complexity of the current system—which we are clearly not going to completely overhaul before universal credit—what level of improvement in fraud and error might it take to reach that halcyon point? We are on the verge of our Northern Ireland NIF accounts not being qualified, so we are very focused on that.

Chair

Sorry, could you just explain what you mean by Northern Ireland NIF accounts?

Lin Homer

Sorry: Northern Ireland national insurance fund. That is also qualified, or has been, because of the level of fraud and error—it was not last year and we are hoping that it will not be this year. It is slightly the reverse. Our colleagues in the NAO have given us a slight advance warning that if and when tax credits is not on the books, they will look more closely at the level of error in child benefit, which is £175 million and about 1.5%. The level of materiality goes quite low if £175 million at 1.5% is beginning to attract interest. I suspect that we could double our performance again and probably—Amyas would say, rightly—they would still qualify us.

Sir Amyas Morse

Who knows whether I’ll say that? Forgive me for a second, but may I just check something? Do I understand that the level of further reduction might be one of the negotiating factors in the spending review?

Lin Homer

No, not specifically, but we will end up with a balanced agenda. Most of the fiscal events have a variety of issues for us. We make some decisions about which of our obligations we need to balance in which ways. We have tended to try to make that a connected set of objectives, not disaggregated, because it is the nature of our business that the parts of our business play into each other. It is not a specific, targeted intervention but the point at which we will adjust the business plan and effectively republish is when we put the whole thing together.

Nigel Mills

Mr Devereux, it is not quite such a rosy picture, is it? To me, you do not look to be anywhere greatly near meeting the aspiration you set five years ago of being down to 1.7%. What has gone wrong?

Robert Devereux

I guess I want to make two points at the start in answer to that question—two things that are slightly different from HMRC. The figures that have been quoted in this Report are perfectly taken from our own statistics and they are measuring the extent to which there is fraud and error in the system at any one moment. They are not seeking to estimate how much of that we then subsequently recover. In my mind’s eye, I am thinking of two concepts: the gross loss, which is what is wrong today; and the net loss to the taxpayer, which is what happens when I have collected something back. That is point one. The second thing that the NAO has drawn out in its Report is that, for my system, because I have multiple benefits, the headline number is as affected by the relative size of the spend on the different benefits as it is by the level of incorrectness. If I end up with a lot more money being spent on the state pension, which is really low fraud, my headline number goes down. I am slightly with the NAO in thinking that the right answer to your questions is, “Well, can I just get underneath these headlines a bit?” and I draw out three things. First, we have a very stable and low rate of fraud and error on half of my spend, which is state pension. Secondly, we have had a rise in both the gross and the net loss on housing benefit, which is what I was before the Committee for 12 months ago, when we had the long conversation about what we were doing. I’ll come back to that. On all the other benefits—and I mean all the other benefits—over the course of the Parliament, we have reduced the gross loss by 20% and the net loss by 30%. You cannot see that from the NAO Report because they have chosen not to put the net together and because they have not gone below the headline numbers, despite the recommendation saying that we should write strategies that way, with which I agree. I need to put that on the table. So I have three different stories: a lot of money on which there is very little fraud; a substantial amount of money on which things have been getting worse, in part, as I explained last time, because housing benefit is a benefit paid in work and out of work, so when the composition changes and more people are in work on housing benefit, which has happened during the recession, I have a compositional problem; but then on income support, JSA, pension credits—you name it; all the rest of them—there have been very substantial reductions in the gross and the net amount. That is partly because we’ve done good things on fraud and error, and partly because we are doing much better things on overpayment and recovery.

Nigel Mills

So are you saying you would rather have a separate target and a separate assessment for each benefit? Is that something that you would like?

Robert Devereux

I think it’s a good question for the Committee to think about, because if you hold me to account for one number, there are so many countervailing factors that I think we’re always going to have a long conversation. The advantage that we’re moving towards—we might think about doing this, since we are setting, as Lin said, spending targets right out to the end of the Parliament—is that we’re getting to a world that is going to be universal credit, pensioner benefits and the state pension. And at that point, these are three moderately coherent things, so you could ask us what we are doing on each. I don’t think I want to get to the point where I will have a target for everything, because of winter fuel benefit and some tiny ones knocking around, but I think it would be perfectly sensible for the Government to say, “Well, here are some big pieces”—tens of billions of pounds worth of spending—and I could put them into sensible chunks and tell you a story about each one, which is essentially the principal recommendation of the report.

Chair

Can I just be clear that you don’t think you can do that now with the current benefits? Because this figure that is highlighted masks a range of—

Robert Devereux

I could do that, yes. The target that was set in the last Parliament, just before I arrived, was an aggregate target—to try to get some sense of, “We’re doing something about the whole of fraud and error”—but I think this compositional effect is real.

Chair

So you say you could do it. Could you do it quickly and easily, or are you saying you could theoretically do it?

Robert Devereux

So, what you’ve got in the Report is actually an estimate of the total amount of spending—this is figure 8—and it is a projection, as best we can, of both the compositional effect of the benefit system, including taking over the whole of tax credits and bringing them into universal credit, as well as projected levels in the improvement in the rate of incorrectness. What that red line shows you in figure 8, on a like-for-like, apple-and-apple basis, is a reduction from 4% in the current year to 2.5% at the end of the Parliament. That is a set of projections that we—

Chair

But this 1.7% figure masks—as you know, more than half of that is state pension, and then there is the other—

Robert Devereux

Almost exactly half.

Chair

So how easy would it be for you to sit here now and tell us what you’re doing bit by bit? Is it just a question of the metrics being pulled out from the Department, or is it more complicated than that?

Robert Devereux

I did it in preparation for coming here. If I just stick with the net loss position, on the state pension, it is £86.5 billion. That is 0.2%—tiny. On housing benefit, it is £24 billion—3.5% net loss. And on everything else, it is £42 billion—2.5% net loss.

Mr Bacon

Could you just repeat the housing benefit one?

Robert Devereux

3.5% net loss; 5.7% gross loss.

Nigel Mills

There is a risk to using net loss, in that you cause a lot of pain and misery to people who get given too much at the start and then you have to get it back from them. There is a risk that you won’t get it back and there is misery for people who have got stuff they shouldn’t have.

Robert Devereux

Sure. So, I’m absolutely of the view that we should try to get it right to start with, but quite often, if I’m honest, the debate about fraud and error characterises it all as loss to the taxpayer. For my system, an awful lot of what gets measured as fraud and error is actually—we know for a fact—claimants reporting one week, one month, two months late. So you can see from the analysis of our overpayment data that quite a lot of people tell us eventually what they’re doing, but they don’t tell us at the time. So when the statisticians go out and do samples, they find it to be incorrect, and that’s a perfectly good statistical result. In practice, the population tends to correct it a bit behind the time. I don’t think I’m making an unfair proposition in terms of what is the loss to the taxpayer. However, would it be better to get it right? Absolutely. Let us take universal credit for one. The very fact that I’m only paying those people in employment when I know what their employer has declared their earnings to be, rather than paying first and asking questions afterwards, precisely goes after—

David Mowat

If I could very quickly interrupt, I am a little bit puzzled by your answer. The NAO Report quite clearly says that HMRC is on target with this and it says that DWP is not on target with this—that is in conclusion No. 13. Now, what you appear to be doing is defining slightly different indicators, one of them being the net one. Do you agree with the NAO conclusion that you are not on target, or is the issue that the NAO has just used different indicators to you?

Robert Devereux

No. The Government set a target, which included an aggregate percentage to be achieved by my Department, which we have not achieved. I have tried to explain, in answer to the question, why that would be the case when—

David Mowat

I understand; and when the Government set that target, was it a fair target? I mean, did you accept it yourself?

Robert Devereux

Yes.

David Mowat

So you were working to that target?

Robert Devereux

We were working to that target, yes, but two things have changed in the process. One is that some of the things we hoped to do have not paid off as well as we hoped. Secondly, the compositional change of benefits has changed as well, with both pluses and minuses. If you want to understand the position, I’ve got a strong story on most benefits, a getting-worse story on housing benefit and something which isn’t really a problem on the state pension. So in answer to the question, “How do I understand it?”, that is how I understand it.

Nigel Mills

The reason why I was asking about gross versus net is that surely you should be targeting is not making the mistake in the first place, because that has less risk for you and the least misery for other people.

Robert Devereux

I agree, but if Parliament wrote rules that say, “These are means-tested benefits. I don’t expect you to pay if somebody’s earnings change,” I would have two choices. Either I have a fairly elaborate system, which I have at the moment, to try to get people to tell me when their earnings change, or—good news: this is what we have now done, based on HMRC’s system—I know when their earnings have changed. The development in the technology in the tax-collecting system has enabled me to do something better. I could achieve a better result by just having higher disregards and forgetting a bit of change in the system. My data would be better, but—

Nigel Mills

If you could get your way on that. But you are still expecting, based on your Department’s forecasts, a very large level of fraud and error in 2020—I think £5.8 billion.

Robert Devereux

Yes. So in 2020 the position is, as it shows in figure 8, 2.5%—that is gross, so take another discount off, which I haven’t calculated, for the net positon. At that point it is 2.5% of an even bigger number. When tax credits are in together, we will have a total departmental spend of the best part of £225 billion. A tiny percentage of £225 billion is a big number.

Nigel Mills

So you are estimating 5.8%. What do you think you could get that down to?

Robert Devereux

I didn’t say 5.8%; I said 2.5%.

Nigel Mills

Sorry, £5.8 billion. What do you think you could get that down to by 2020 with a full effort at tackling fraud and error? What do you think your target should be for this Parliament?

Robert Devereux

The table that the NAO produced for you is, as at autumn statement ’14, our best assessment of what was possible. You will see that figure 8 shows a like-for-like improvement—no doubt we will have to explain overpayments in the tax credit system to you shortly—from 4% to 2.5%. By the standards of the sorts of movements you have seen in both Departments in recent years, that is still quite a big change. It is possible that we have to do better than that. On some of the things Amyas is asking about, such as how this will work in the spending round, there are still choices for Ministers to make about the roll-out of things, what amount of expenditure they are prepared to put into the Department, and how much of the effort we think we can make the OBR will score as an AME saving, even if I believe it to be true. That is part of the reason the spending round is a relevant concept, but this was a best estimate, as at autumn statement ’14.

Chair

But did you make different progress on different benefits within the half that is more controllable?

Robert Devereux

Yes.

Chair

What have you learned from the best to influence where you are not doing so well?

Robert Devereux

We went through two things. To slightly paraphrase the housing benefit hearing, there is a conversation about the extent to which the claimant make-up has shifted to make it more difficult to manage, because more people are in work and have to declare their earnings—I get that problem. Over and above that—leaving that aside—there is an increase in the rate of fraud and error. The Committee challenged us and asked whether we were incentivising local authorities to do the right thing. On the back of that—we told you we were doing it anyway—we changed the incentive system. We put more money in local authorities’ hands, and they are out doing things now that they weren’t doing the last time we met. It is a complicated part of the system. There are many actors trying to do something on a particular benefit. One of the things I deduced from that is that it would be quite a good idea to make sure that you have a rather more streamlined system. I am still running a system in which I have six principal working-age benefits knocking around. Come the day when there is only one, it becomes radically simpler for the claimant who currently has to remember what the rules are, whether they are different, who they have to tell and how it all works. That gives me an excuse to talk about two things that have gone on in this Parliament. One is action on the actual legacy system, which is what this Report is reporting. Part 3 of the Report says that the other reason why the Department has been busy is that it has been landing all these reforms, designing them and making sure we have got some sensible rules in there. A lot of stuff that you as MPs passed in various welfare reform Acts has bit by bit constructed a system that is better designed and simpler than the system that we inherited. If the CAG insists on regularity being measured against your own rules, you have simplified the rules, which will make it easier for me to keep within them.

Nigel Mills

You are also not particularly keen to set a target for where we might be in 2020 in terms of how far you can get fraud and error down. Presumably when you are making a bid to the Treasury for investment, you can say, “If you were to give us £1 billion to spend on fraud and error, we could get the rate down.”

Robert Devereux

I could, up to a point, with one slight subtlety that I hesitate to say, but that is none the less true. All the Treasury figures will have to be approved by the Office for Budget Responsibility. They are cautious people, and when I say, “If I do this, this and this, I can confidently expect to save x, y and z,” there is a judgment call on their part about whether they actually buy all that. Consistently, they are cautious about what they score. At the margin, our preparedness to say to the Chancellor, “Give me a pound and I will make it five for you,” depends a little bit on him being confident that the OBR will score the fund, so my case has to be good. As we go through the spending round, I do not know quite where some of those options will come, nor where Ministers will want to be in terms of the choices they make about the overall cost of the Department. We will seek to run it as rationally as possible.

Nigel Mills

Okay, that was a long way round of not answering, but I want to work out something. It is very hard for us to assess how well you do this without knowing exactly where you think you can get over the next few years. Have you any assessment—

Robert Devereux

Sorry, figure 8 shows you, as in the autumn statement, on a consistent basis—

Chair

Perhaps you could let Mr Mills finish and answer his question.

Nigel Mills

I am trying to work out what you think you could get this down to with the right investment. It is a huge amount of money. We are talking about more being lost in fraud and error than the tax credit savings we have to vote on tomorrow. That is why we would like to see a bit of ambition as to what you think you could get this down to if you made use of all the resources that you could have.

Robert Devereux

I have a view that it is quite difficult to land all the welfare reforms, but figure 7 tells you that the consequence of doing what we are planning is to take £2 billion of taxpayers’ money that would otherwise disappear and put it to good use. It is that £2 billion that is driving the red line down in figure 8. If you ask me, “Could you do more? Could you take fraud below 2.5%?”, the answer is, “Maybe, but I have to land all this first,” so forgive me if I am recognising that if I come back and say, “Don’t worry, it will all be fine tomorrow,” you would tell me, “Well, what about your track record?” So, 4% to 2.5% strikes me as not being a bad input for a Parliament when that was not what we achieved in the present one. I have some serious plans around each of these big reforms. Whether it is personal independence, state pension or the pension credit, each of these things is giving you a much better platform on which to operate.

Nigel Mills

To be fair, some of the years in that chart are in the previous Parliament. It is not like 4% to 2.5% is all in one Parliament, is it? I suppose we can overlook a technicality. You set a lot of stall around the changes that universal credit and the new state pension will bring. Have you done a full assessment of the risks of what goes wrong if you mis-implement those? There is obviously a lot of complexity in universal credit. There must be a fair risk that people will end up on the wrong state pension when they start their new one. Have you got an assessment of the risk of something going wrong and how much that might cost?

Robert Devereux

Yes. To start with the universal credit figures, the figures that you are seeing in the Report are the ones that we had in the autumn, and they represent all the good things we hope to have, but also take account of the fact that under universal credit there will be much higher rates of take-up, so we have all the pluses and minuses. Then there is the stuff that we have not yet fully costed, such as exactly what is the most likely level at which some of the new risks consequent from universal credit will come out. We are trying, each time we have some data as the roll-out progresses, to establish the possible consequence of operating in a different fashion. We are trying to work that through. The figures that you have in front of you are the best ones—

Nigel Mills

So are you prepared to set targets for those two major new benefits on what you think you can get the fraud and error rate down to or what it will be?

Robert Devereux

I think that would be a very sensible thing to do. I cannot see the Government not wanting to target fraud and error. I personally do not think that an aggregate welfare target makes a lot of sense for the reasons I have explained. I think disaggregating it into the principal components would be sensible. It seems to me that over this Parliament you will end up with a state pension, universal credit legacy benefit and the personal independence payment.

Nigel Mills

We are sat here six months from the roll-out of the new state pension and we are rolling out universal credit. Don’t you think you could have that assessment and target in place by now?

Robert Devereux

I have shown you what my best estimate currently is, and we have both explained that the spending round is important because it has a bearing on the resources both of us are given to attack the problem.

Nigel Mills

Okay, so what do you think the errors in the first year of the new state pension will be? How much do you think you might lose to miscalculation that has to be sorted out afterwards? Is that an assessment you have done?

Robert Devereux

I do not have that to hand, no.

Nigel Mills

But you could write to us, presumably.

Robert Devereux

I could do. With the state pension, as you can see from the sorts of numbers we are talking about—

Nigel Mills

It is huge.

Robert Devereux

It is a huge number, but a tiny—basically it is based on historical information about your circumstances, so it is not a surprise that it is relatively easy to get it straight. With the new state pension, if I know that you have made 35 years of national insurance contributions, ping; I can pay the—

Nigel Mills

I get that. I guess the question is whether you know—as long as the records are complete for people and you do not suddenly find out you have got stuff wrong.

Robert Devereux

Yes, and there is a problem, but just be careful. The state pension historical problem is one I have at the moment. When you and I both retire, we hope our records will be up to date, but that problem is going to occur with or without the new state pension. As soon as we get to 16 April, your contribution and mine will be a lot easier to find, so our retirement pension will be a problem that I already have, whether or not we reform it, and something for all years, after 16 April, that is easier to do, so I ought to be better off.

Nigel Mills

For the Committee, when we come back and look at the implementation of the new state pension—whenever we do it— I think it might be useful if we know that the Department thought that, on the roll-out, there would be an error rate of this much, and then either that it had turned out much lower and everyone was happy, or that something went horribly wrong—a whole load of calculations go wrong—and you started overpaying people for whatever reason. We need something to hold you to account on. You say that that kind of assessment and target exists.

Robert Devereux

I think we should not be making the rate of error on the state pension worse because of the new state pension. If you want to stick with the number we first thought of—in the current year, state pension was running at 0.2%—you ought not to accept it getting worse consequent on the introduction of reforms that, in my view, make it easier.

Sir Amyas Morse

May I ask something? I am genuinely asking for information. Is there enough risk in transitioning into the new system that you might have some one-off transitional problems? I am asking only because I think that might be what you are driving at. Rather than saying what the sustainable level of error might be—I fully understand that it should not be more than it was before—I am curious to know whether, when you are thinking about the transition, there are risks. It is not evident to me what they are, so I am genuinely interested to know.

Robert Devereux

My own view is that I do not think I do have a consistent problem, because the transition up to 31 March ’16 is exactly the position I am trying to manage at the moment. After that, I have a simpler system. You are right that the process that we are endeavouring to do on 31 March ’16 is to fossilise, as it were, all our entitlements to the old system, on to which I just have to add easy things. The most important thing is whether, by the time you actually get to retirement—as opposed to the day after 1 April ’16—it is right then. My expectation is that the work we are doing—by the way, a lot of it is to do with tidying the national insurance record—should mean that as each of us gets to retirement, we should be confident that that sum has been done, but that sum is not more difficult post-reform than it is pre-reform.

David Mowat

Just a couple of observations about points you made earlier. Mr Devereux, I think you used the phrases “if the rules were simpler, this would be easier” and “elaborate”. Ms Homer, you used the words “complex policy changes”. I think that that is right. I agree that if this were simpler, it would be easier to do it without so many errors. The question that arises, though, is whether, when policy is being developed, you are active in informing people as they make policy that some of this stuff is going to be hard for them to build into their systems and likely to result in high levels of error. Are Ministers aware of that? Is there a kind of feedback, or are you just passive recipients?

Lin Homer

No, and tax credits is a really good case study. We have been aware since early on in the system that it was very complex. A significant amount of the fraud and error we face is because of the fact that it is effectively retrospective, so about 70% of the fraud and error enters the system through, as Robert has indicated, late changes of circumstance. We have made a number of proposals that have led to policy change in the tax credit system, but it is one of the reasons why, if I’m honest, we are so supportive of the morphing into a universal credit system that has a lot of real-time self-rectifying built into it, because it is the biggest problem in our current system.

David Mowat

Right, but you have both said that one of the issues is complexity, which implies that you think there may be more complexity than we need, which I’m sure is true.

Lin Homer

Well, there definitely was in tax credits. It was retrospective complexity, but it was what Parliament wanted to achieve, because Parliament was trying to give families support at the beginning of the year and reconcile at the end, so it built that in—I think the NAO says it in the Report.

David Mowat

No, to be clear, I think that some complexity is an inevitable cost of policy, and it is right that it’s done, as long as everyone understood that that was what was happening. What bothers me more—I don’t know this for sure, but this is my guess—is that quite often things are done and people just do not realise the complexity that that is going to lead to in your organisations, and you do not tell them clearly enough that that is the case. It just happens, and then you turn up here saying, “We are losing £200 per family per year.”

Robert Devereux

I understand why you say that, but that is not what it feels like to be in the Department. It is not only that this complexity ends up with a cost in fraud and error; it is a cost in actually running the Department. You have asked me to be 25% more efficient in the previous Parliament, which we were. That has been achieved by stripping out wasteful process. To come back to the example of the new state pension, its effect is essentially to float a whole load of people off pension credit, a means-tested benefit that, being a means-tested benefit, is a difficult benefit—it currently has a rate of error pretty much the same as that for housing benefit. I have been thinking through with Ministers, “What happens if I do this?” The total amount of money that people are going to get at the bottom end is actually going to be very similar, but making it up by way of a new state pension and very little pension credit gets you—

David Mowat

To summarise, you are both happy that you get sufficient input into the policy development such that unnecessary complexity is not built in, so it is all working pretty well.

Lin Homer

I think we are both satisfied—well, I am satisfied—that we have an influence. As you said, there are times when the Government of the day will accept complexity in order to achieve their policy objectives.

Mr Bacon

I am extremely interested in Mr Mowat’s line of questioning because I have been asking these questions of permanent secretaries—certainly Sir Richard Mottram and Rachel Lomax—for years. The question I always put was: if it is sufficiently complex that it causes problems in and of itself, and it might be, Ms Homer, that the Government of the day want to do x, or have decided that it is a price worth paying, the issue is whether you, as accounting officers, have decided that it is worth paying and, if you haven’t, what you have done about it. I look in vain for requests for letters of direction about this saying, “Minister, what you’re asking me to do will mean that it is complex to such a high degree that it will be impossible for me to fulfil my remit to Parliament to be as effective, economic and efficient as I want to be so, if you want me to do this, you’ll have to direct me to do it.” That conversation does not happen. Coming back to Mr Mowat’s question, you said, Ms Homer, that you are comfortable; are you, Mr Devereux, comfortable that you have enough input into the policy—not its remit, but its consequences in terms of effectiveness, efficiency and economy? Are you comfortable that you have enough input into the shaping of the policy so that, if it does create added complexity that makes it either more difficult or impossible for you to do your job as an accounting officer, you can and do push back?

Robert Devereux

That is a very fair question. In trying to reach that judgment, I guess I am also conscious of what the Government are trying to achieve overall. For example, if they are seeking to save money and change how some of the benefits operate, you can see a trade-off between the generosity of disregards, the size of taper and the incentives for people to play the system. Those are actually quite difficult to get a really good grip on ex ante, but you can see for sure that it will end up being a lower-cost system. I am not making that point for our current Government, but just generally. These rules are almost all the consequence of Parliament and the Government seeking to constrain the total amount of money that is being spent on welfare.

Stephen Phillips

The trouble with that, Mr Devereux, is that that’s not your job, is it? You are the accounting officer within the Department. It is not your job to look across the whole of Government and to support the Government’s agenda to reduce the size of public spending; your job is to deliver value for money within the Department of which you are the permanent secretary. If something is so complex that in fact it is going to be difficult to implement and is going to result in a greater level of fraud and error, that is something to which you ought to draw Ministers’ attention and, if necessary, seek a letter of direction about, is it not?

Robert Devereux

It is, so what a good position I find myself in, to be here at a point at which we are simplifying benefits very substantially. In the period in which I have been here, we have been doing reforms that for the most part have all worked in the direction of improving fraud and error, which is why the totals are going down.

David Mowat

Just to leave this line, I think we all agree that what we are trying to achieve is sensible. All I would say is that there are two ways in which you can look at the process of specification. You can say it is done by Government and you are passive recipients of it, or you can say that in certain instances there is push-back and it is part of your role that you prevent what I would call unintended consequences. I spent quite a large part of my life developing computer systems, and you quite often see a lot of cost in areas that the people who designed the thing didn’t know and didn’t care about while they were doing it. They hadn’t been told and that loop hadn’t been closed. I think we should agree that that is your role.

Robert Devereux

I agree. I will give you an example. Somewhere in the bowels of the universal credit system there are rules we are having to construct about what we will do about self-employed earnings.

David Mowat

I don’t want to spend too long on this. Is the position of both of you on this that over-complexity and over-elaboration are a bad thing, but that you are satisfied that, within your Departments, when you need to give that steer to Ministers it is given?

Lin Homer

Yes.

Robert Devereux

Yes.

David Mowat

Right. One final question on detection: in figure 5—this is a DWP thing—the line on prevention shows a massive underspend in terms of the programmes. You spent £27 million as opposed to £192 million, although for every pound you spent you got even more back than you expected. That raises the question of whether, if you’d spent another £27 million, you could have saved even more. I don’t know. What changed between the original programme and what you ended up doing?

Robert Devereux

What changed is covered in footnote 1, slightly, and in paragraph 2.21. In the last Parliament, we embarked with the intention of constructing a system called IRIS, which is around risk assessment and testing the authentication of whether you are who you say you are when you are in the middle of an online session. The conclusion we reached in the middle of 2013, along with lots of other things on universal credit, is that, desirable though such a system was, its cost was far too great, so we did not spend the money that we thought we would spend on that system. Instead, we have started to build a system differently and at a much lower cost to achieve principally the same aims. We are currently doing two things. One is that we have constructed and are now using in universal credit—and will continue to build along with universal credit—an analytical hub. That basically takes the data, mashes it together and makes some sensible suggestions about whether Mr Devereux’s claims are likely or not to be incorrect. Separately—and again, we have built this and it is operating for the universal credit service—is something that says, “What do I know about the current session we are in? is it likely that we are currently being spoofed?”

David Mowat

Okay, but you were hoping for savings that you didn’t get, of £860 million. You didn’t get those, so are you saying that they weren’t there to be got or that it would have cost you even more than that to get them?

Robert Devereux

No. What I said was that the potential solution that would have caused us to spend that money, and, had it worked, would have generated those sort of savings, was, in our judgment, too expensive for the thing that we were getting.

David Mowat

But your original programme estimated that the cost of that solution was about £150 million at the very most, on these numbers—

Robert Devereux

It was £180 million.

David Mowat

No, £190 million, and then you spent £27 million. Okay, £180 million. Are you saying that was wrong?

Robert Devereux

Yes.

David Mowat

So it wouldn’t have been £180 million in the end—it would have to have been a much bigger number?

Robert Devereux

No, sorry. When we constructed the programme, the prospective future spending was estimated. In that original column, that table is talking about what we thought was going to be the case. We hadn’t got very far with it before we concluded that that was way too much to be spending for that capability and consequently did not spend it.

David Mowat

How could £180 million have been way too much to have been spending if it would have saved £860 million?

Robert Devereux

That is a good question.

Chair

We are hanging on your answer.

Robert Devereux

So—what is a good answer to that?

Mr Bacon

Good job this one is not televised—it probably is.

Robert Devereux

Yes, I am sure it is. We could have spent that money and potentially made those savings, I agree. I think people were just thinking that this was going to lock us into a way of doing big IT things that was outsourced to a provider doing something and that was just the wrong thing to do. The sort of sums we are spending now are a fraction of this. You are right: on a cost-benefit analysis, actually this might have been a good thing to do. It would have given us a capacity which I am fairly sure I would have been back in front of you, with you saying, “How can you possibly be spending all that money on it, because it could have done it a lot cheaper than that?” It is that that we are doing.

Chair

Mr Fundrey may have an answer.

Jon Fundrey

Can I help a little bit? The reduction in cost is primarily because we cancelled IRIS, which was unaffordable. The reduction in benefit—

Chair

How much had you spent on IRIS before you cancelled it? Remind us—it is somewhere in there.

Jon Fundrey

To be honest, I cannot recall.

Chair

Okay. Perhaps Max can help us there. Carry on.

Jon Fundrey

Whereas the reduction in savings is a result of loss of benefits to IRIS, but also a reduction in what we thought we were going to save from the ATLAS system, which is a system we built to share information with local authorities. So you are not comparing like with like. Roughly, IRIS was around £170 million and we have lost £370 million of benefits, so it is two to one.

David Mowat

When you say I am not comparing like to like, all I am doing is looking at the table that says that you spent a lot less than you said, but you got massively less benefits than you had hoped.

Jon Fundrey

Yes.

David Mowat

Okay, it is a combination of two systems, but, nevertheless, even on the numbers you just said then, £180 million to £380 million—I think you used to work for BOC, and had you gone to your CEO at BOC and said, “I can spend £180 million and give you £380 million benefit,” he would have promoted you, wouldn’t he? So why did you not do it here?

Chair

Who is going to answer that? Mr Devereux, you are the accounting officer.

Robert Devereux

I think the confidence the Government had in whether the system would do what it said and the obvious expense of the system caused us to pause—

David Mowat

So when you looked harder again at the numbers, you thought, “Actually, these benefit figures could be flaky.”

Robert Devereux

You are making a very good point. If all cash is equal, we could have spent this money and potentially made some of these savings with a system which, I am sure by now, Mr Bacon would be asking me questions about why on earth we were spending a number like that, because he has got a friend who could do it for a tenth of that.

Chair

I am going to bring in the Comptroller and Auditor General.

Sir Amyas Morse

Wouldn’t it be fair to say that there was also much higher uncertainty about whether the system could have been delivered in a functional way? When you are talking about big system projects, it is not just a question of comparing costs; it is a question of understanding that, if you have had a lot of difficulty with a system, it may be that it will take a long time and you do not know when or whether it will deliver the functionality you were hoping for. So, in some ways, one of them is not only a lower cost, but a lower risk.

David Mowat

Yes, I understand that. What you said just then is that you are dealing with probabilities of outcome. I am, too, but maybe that complexity was too complex to put on to table 5. All I am saying is that—

Chair

We are in danger of going off on a tangent here.

Robert Devereux

The way accountability works is whatever we thought in 2010 is in this table. Some of those things we knew very well what was going on, and some of them were less well cooked, because you are talking about 2010 and the Government had only just arrived and decided that they wanted universal credit. So I do not think that it is unreasonable, as some of them were at different levels of maturity. But your basic point is right. If we were absolutely confident about both sides of the thing, then you could have spent even a lot more money, but, as we worked it through, we were able to ask, “Is this really the best answer to the question?” and our conclusion was, “No, it isn’t.”

David Mowat

Okay.

Chair

I am going to bring in Stephen Phillips in a minute, but can I pick up on lessons learnt? Lin Homer, admittedly you had a lot more to bite—or whatever phrase you used—but you had some success on tax credits. What lessons have you learnt for future reductions in fraud and error? We are hearing where assumptions were made that it would not be brought in, but you have got the real proof of the pudding.

Lin Homer

I was going to, at risk of distracting you from Robert, make the point that, at any one time, both of us have got a significant number of initiatives running. I think Nick has got around 60 initiatives in personal tax credits. And at some points you also have to make a judgment about which you are going to give management time to as well. I think what we have learnt—yes, figure 6 sets that out. You fairly criticise us to a degree for spending less attention on prevent than detect. Part of the answer for us is: absolutely, redesigning our system to get to the heart of prevent is not sensible when it is not going to last very long. What we have also learnt, significantly—Nick could probably give you a better example than me—is that focusing on high-value fraud and error has got us a significant return. We have increased the number of interventions overall from around 100,000 to more than 2 million. We have altered the balance of risk for our claimants as to whether we will lift the stone and look; that is something that we have learnt.

Chair

Sorry—so you were looking at 100,000 cases. Is that individual claims?

Lin Homer

Yes.

Chair

And you are now looking at over 2 million, just to be really clear.

Lin Homer

Yes. So that’s something that we have learnt. You have to change the balance of risk for the people in the system. We have also looked at high-risk claimants particularly deeply and got two and a half times as much return from them as we were doing previously. We have learnt to target. As a consequence of giving our staff more information about things so that as they are interacting in real time with a claimant they can work out the level of risk, we have massively increased the productivity of each of those interventions. I do not absolutely have the figure off the top of my head, but, for not very many more staff, there has been a massively bigger return.

Chair

It would be good to get that figure, if you could send it to us at some point or find it while we are here.

Lin Homer

I’m sure that I’ve got it somewhere. I may be able to send it while we go. Level of intervention, targeting on high risk and giving our staff much better prompt cues—“If you see this, ask about this. If that happens, ask about that.”—are all much better supported by predictive analytics and data analytics than the first time that I appeared before you.

Chair

So in summary, you are saying that a mix of automated risk assessment and the human factor is important.

Lin Homer

Using the higher skill of our people because we are feeding them better information.

Robert Devereux

We have a similar story. In that same table that you have just been referring to, the intervention of staff is paying us £10 for every £1 that we spend. It is a huge return.

Chair

Right, so staff well used—

Stephen Phillips

I will come back to you, Ms Homer, because it is really this issue for both of you. As the preamble says to figures 5 and 6, you are both focused on detecting fraud and error, rather than on preventing it. Go back in the Report to the key findings on page 7, and to key finding 14. Again, the NAO says that you are “focused primarily on correcting claims rather than preventing errors.” Mr Mills put a very interesting point to you, Mr Devereux—I appreciate that it does not concern you so much because we are the politicians and you are the people who implement the policy—that these are real people. They are our constituents. They get paid too much and then either DWP or HMRC comes back to them afterwards and says, “We want our money back. You’ve been overpaid.” We have to deal with that and, what is more important, our constituents have to deal with that. These are people who do not have a great deal of money. Why are you focusing so hard on correction and detection of overpayments in the past and not focusing more on prevention, which is what both of you said you were going to do when we got to the spending round in 2010?

Robert Devereux

There is part of your premise which has simply ignored all the work that the Government have done on redesigning the benefits system. There are two chapters to this story. There is what we have been doing in the current five years, but the principal reason that things will get better is because the product set that we will be operating will be better because of all the hard work we have been doing in the past five years. That is an important point to make. I am not going to accept that the Government have somehow only used that in detection. The entire point of reconstructing the benefits system is to give us the chance not to be doing that. That is the first thing. The second is about paying this out and paying us back. Most of this problem is people who have not told us things that they should tell us, so I am consequently trying to find ways to find that out promptly and effectively. A lot of the productivity gains that we have had are from getting our staff, whenever they are on the phone, to say, “By the way, can I just check that x, y and z still remains the case? Is there anything else you want to tell me?”, which stops us having to have after-the-fact, customer-led correction.

Stephen Phillips

On the second of those, you were the ones who expected to spend £127 million, I think it was—the figure is there under figure 7 somewhere—on detection, which would save our constituents from being paid too much money. We have just heard that you spent a lot less than that. That has a cost associated with it in terms of people having to deal with overpayments. There does not seem to be any appreciation on your part or that of the Department that that is the case.

Lin Homer

So—

Stephen Phillips

No, Ms Homer. I’m just dealing with DWP at the moment.

Robert Devereux

The detection line—

Stephen Phillips

Paragraph 2.21 says that you expected to save £1.12 billion. You were going to spend £192 million on projects for preventing high-risk payments. In fact, it says in the last sentence that you now project you are only going to spend £27 million.

David Mowat

That is the top line of figure 5.

Stephen Phillips

If you go back to finding 14, you have undershot what you told the Treasury you were going to do in terms of preventing errors by £100 million.

Robert Devereux

In the long answer I gave your colleague, I went through that top line and explained that we had one particularly large intervention.

Stephen Phillips

Hold on; as I understood your answer, you are blaming individuals for not telling you about changes in their circumstances, which is resulting in significant overpayments—fraud and error. Why do you not do more to tell people what they ought to be doing?

Robert Devereux

With the greatest of respect, that is precisely what we have been doing in redesigning the benefits system. Most of these claimant errors are to do with earnings—undisclosed earnings.

Stephen Phillips

That was your first point. I was actually on your second point. You are now saying that it is all going to change under universal credit. Is that right?

Robert Devereux

I am slightly losing track of which question you are asking. You were asking me about claimants and whether I was doing enough to help them get it right. My best answer to that is: I am redesigning the system to make it so much easier for them to get it right.

Stephen Phillips

You gave me two reasons. The first of your reasons was, “It’s going to get better in terms of prevention because we’re redesigning the system.” Is that a reference to universal credit?

Robert Devereux

It’s a reference to universal credit. It’s a reference to how the new state pension will take people off the pension credit. It’s a reference to the changes we made to personal independence payments.

Stephen Phillips

How it is going to get better from universal credit is figure 7, as I understand it, on page 25 of the Report.

Robert Devereux

Correct.

Stephen Phillips

So the headline here, if you like, is that you expect universal credit to reduce fraud and error by £500 million each year once fully rolled out. That is what the NAO say. Is that what you told them you expect to happen?

Robert Devereux

I alluded to this earlier on. There are two—

Stephen Phillips

Sorry, Mr Devereux. Is that what you told them you expected to happen, in these specific terms, as a result of the roll-out of universal credit—yes or no?

Robert Devereux

I helped them to construct a table that shows there is a £2 billion saving. The point about it is that some of it gets scored technically as fraud and error and some of it was never scored previously, but it is all wrong.

Stephen Phillips

Right. Now, in 2010, as we know from key finding 14, you thought you were going to save £100 million more in terms of prevention than you have actually achieved. Why should we have any more confidence in what you say will happen as a result of the roll-out of universal credit, given that you were wrong last time round?

Robert Devereux

A fair observation. That was the strategy written within a few months of Ministers arriving and launching on a thing called universal credit, which I think had not even got to its White Paper at that stage. I am now four or five years further down the line, and I know a lot more about the subject. I have already built some of the things that are in operation, so it is the difference between a plan to the best of the Department’s ability at the time and—

Stephen Phillips

Let’s get this right. It was a rubbish guess in 2010. Now it is based upon four or five years of experience, so you are confident you can get the £2.1 billion.

Robert Devereux

With the greatest of respect, some things in life are hard to estimate in advance.

Stephen Phillips

Ms Homer, the same question to you. Figure 6 shows your fraud and error investigations. As I understand it, HMRC does not take responsibility for any fraud and error at all. According to you, it is all customers not telling you what is going on. Is that right?

Lin Homer

Sorry?

Stephen Phillips

HMRC does not accept any fraud and error in its own systems; it is all down to customers not telling you about changes in circumstances.

Lin Homer

Yes, but this is the design of the system. We have said in front of this Committee on numerous occasions that the complexity of the system is largely what causes that. We are not just saying it is sloppiness. It is a very complex system. It makes it very difficult for people to get things right. The comment I was going to make—

Stephen Phillips

Just pause there, if you would. The problem with that is that if you do not accept any responsibility at all for errors that people make in not telling you about changes in circumstances, you do not have to focus on prevention at all as an agency, for the simple reason that none of it is your fault.

Lin Homer

No, that’s not the line we take at all. The point I was going to make is that, although prevention has been a smaller part of what we do—indeed, short of revising the whole system, it would be difficult to make it bigger—I strongly believe that the work we do on correction, which is big for us and for Robert, is us absolutely accepting our responsibility. That, too, goes to trying to ensure claimants are helped to avoid the consequences that, as Mr Mowat and you rightly pointed out, can be dire for them.

Stephen Phillips

Well, 80% of the losses you are saving are correction and—

Lin Homer

Yes, but can I give you an example? If we can prompt a claimant really quickly, either by using our data analytics or by using real-time information to say “We think you might have got this wrong”, we might be correcting something that could lead to a 15-month overpayment in a matter of weeks or short months. That isn’t as good as getting it absolutely right first time, but it’s an awful lot better than waiting 15 months and then saying, “That’s wrong.” So we do accept responsibility.

Stephen Phillips

I understand that, because of the time lag, it is difficult for the agency. I understand that for DWP and in relation to housing benefit, too. Why aren’t you focusing more effort on educating people to prevent such errors from occurring in the first place?

Lin Homer

We are, and in the annual Report—not this Report—the NAO accepts that one of the things we have done well is educate claimants. I realise that you have not got it in front of you, but we did talk briefly last week. In recognising some of the things we have done, the NAO made the point that by giving our telephony staff better information we were able to educate and support the claimants and reduce the overpayment risk—perhaps not to nothing via prevention, but to a much smaller amount. I think that is us accepting our responsibility.

Stephen Phillips

For both of you, the Committee is going to have to form a view when we report this matter to the House. Do you think that both of you could be focused a bit more on prevention than you currently are?

Lin Homer

I think it would be difficult for us. We have looked very seriously at major legislation in relation to tax credits, as opposed to marginal, but in the timescale involved, we think the complexity you would cause the claimants for what will be a transitional period anyway is probably not sensible. We think our major investment in digital—allowing tax credit claimants to renew online and prompting them to tell us about changes of circumstances online—is a good bridge between a complete redesign of the system, which is not justified before UC, and doing nothing. With high-risk claimants, we could get to the position where we are prompting them much more regularly. With children in full-time education—

Chair

Could you just define a high-risk claimant?

Lin Homer

Someone who has made significant mistakes that have led to overpayments lots of times. It may be because of the volatility of their earnings or their family circumstance. We can say, “Having plotted and seen your pattern of behaviour, if we prompted you to update us on your little one’s childcare costs more regularly, would that help you stay in good shape?” We think that is a legitimate simplicity, or a smoothing of the complexity, in a programme that is already in transition.

Stephen Phillips

Do either of you have an app, so people can notify you on a smartphone that their circumstances have changed and then you can call them?

Lin Homer

Yes. Last year we introduced a digital renewal for tax credits, and in the final week of tax credits this year 48% of people renewed using the digital line, many of them using smartphones. They were tweeting each other to say, “This is really easy. Use it.”

Stephen Phillips

That is very helpful. Mr Devereux, the same question to you. This Committee is going to have to form a view about it. Do you think you could and should be doing more on prevention? I know you are doing work on correction and detection, which seems to be resulting in the recovery of significant amounts of taxpayers’ money, but we are talking about real people who are overpaid, spend the money and haven’t got it to pay you back. Could you do more on prevention?

Robert Devereux

Potentially.

Stephen Phillips

Thank you. That is very fair.

Chair

We may come to this in a bit. I want to touch on the issue of the £1.6 billion a year in underpayments. Little emphasis is placed on reducing underpayments. Could you set a target for that, Mr Devereux?

Robert Devereux

I was going to elaborate until you decided to go for my one-word answer.

Chair

We like one-word answers sometimes.

Robert Devereux

I know you do. One of the reasons why we are getting better at doing the sort of work you are asking us to do is that we are getting better at engaging with staff on a holistic basis. As you probably know, I used to have a contact centre arrangement and a set of people processing benefits. I brought those two together, so when people ring up they get the entire service in one go. The advantage of that is that the person who knows all the rules and knows how to deal with the person is dealing with you straight away. Let me pick up one of the points made by the National Audit Office, which was: “Yes, but you’re chasing people to process quickly, because the other thing you want is your claimants paid promptly. Is there a risk that the speed is working against the accuracy?” We put quite a lot of effort into the quality side of the house, which has changed the NAO’s view of—

Chair

Can you set a target to reduce this? That is the question.

Robert Devereux

We have not historically set targets and no Government have previously.

Chair

Would you be able to do that? We are tempted on that score, because we have to make recommendations to the House, as Mr Phillips has reminded you.

Robert Devereux

Yes, well, I shall look forward to reading them.

Chair

But do you think it is something that you could technically set a target for, given what you know—all the metrics you’ve got, and, as you say, bringing together two different teams?

Robert Devereux

One of the things that is going on in universal credit is that a significantly greater take-up is going to take place. Underpayments—“You were entitled to it but you weren’t getting it”—are going to be significantly helped by the fact that actually, you haven’t got to make the six claims, or dodge around the system and find three authorities; the local authority, me and—

Chair

We are hearing a lot about universal credit, but let’s move on to Lin Homer on this same point about underpayments. Do you think it is a target that you could technically set and achieve?

Lin Homer

I think we would clearly want, for the rest of the period in which we are running tax credits, to try and minimise the risk of inaccuracy. What we have traditionally done is send quite a complex claimant settlement letter, which is very hard for people to understand.

Chair

Absolutely. I get them at my surgeries.

Lin Homer

We say, “Please check this because this is your formal record”, so one of the things we are looking at is whether we can help—again, using the digital account, but also clearer information—to allow people to check very carefully. But with the complexity of the system, we are always going to have some underpayments as well as overpayments in the current system. They are quite a small part but we try to prioritise them when they happen, don’t we Nick?

Nick Lodge

The renewals part of the process is really important for tax credits, so it is at that stage, really, that we are reconciling what has happened through the year with claimants. Obviously, we do everything we can, including having calculators online and that sort of thing, so that people can make sure they get the information to us correctly.

Chair

You talk about transparency. We see those letters at our surgeries, both from the DWP and HMRC. They are mostly impenetrable, frankly. When you are at a busy surgery with 40 people queueing and someone presents you with a letter with a load of figures, you think “No wonder they can’t understand it”, and it takes quite a lot of work to get into it. I think there is an issue of cost shunting, which, as you know, the Committee is quite concerned about. When something is very complicated they have to go to some other agency to have them explain a letter that should be transparent in the first place. On underpayments, we know—it has come out from what you said, but also from the Report and other places—that most of the challenges come with renewal, because of the change of circumstances, and that is when error very often creeps in. Have you got really good plans for making that simpler for people? You have talked a little bit about online and about whether people are asking the right questions, but what else will really bite with people, because it is very complicated when you are the claimant? How are you going to do that?

Lin Homer

I might ask Nick to talk you through some of the things we are doing. I am not 100% optimistic that we can make the tax credit system easily understood by MPs or claimants in the time we have left, but we are doing some things to try and ensure the right prompts are ours at the point when people telephone.

Nick Lodge

For example, we have used real-time information—the RTI system—to prepopulate people’s renewals forms for the first time this year. We use that same information when they respond to us to try to get their award right on the basis of the correct income. As Lin mentioned, we have developed an online service and an app that people can use through their smartphones, and we are going to deploy that through the whole year, so that we can build something that people find easy to use and can tell us about their changes on easily and quickly, 24 hours a day, at their own convenience. We have found that, actually, tax credit claimants have a need for easy contact and actually like using the online service and did so—

Chair

That’s great and I’m a big fan of RTI, but I do have constituents—I know we all will—who get wrong information from their employer on their payslip. Perhaps the hours are miscalculated or overtime is calculated in different ways. That can cause a myriad problems. What sanctions are there for employers or how do you encourage employers—whichever way you choose; I don’t know whether you prefer sanctions or encouragement—because that has a devastating impact on the individual, and once RTI gets in, that becomes the golden truth, whether it’s right or wrong?

Lin Homer

We did an awful lot of work with employers to clean up their data to go into RTI. I’m not suggesting there aren’t cases where double-reporting or failure to report is happening, but I think the introduction of RTI has seriously cleaned up employee data in a way that really helps us all.

Chair

Do you know what the percentage of error is in RTI?

Lin Homer

The percentage of accuracy is in the high 90s, so it is very tiny sums. What we are continuing to try to do with RTI is spot indicators that something is going wrong, so if we see an employer give us 11 months of info and then—I know it sounds silly—not deliver month 12, as opposed to assuming they had no staff that month, we’ll go back and say, “Are you sure you haven’t missed something?” So we are doing what we can, all through the system, to build in-built triggers. Generally, HMRC is now building some of our triggers into the software that is sold to big and small employers, so that the software providers can build in something that says, “HMRC will ask you a question about that. Do you want to check it before you submit it?” We are right at the foothills of this much cleverer use of data and technology, but we believe it’s very powerful in nudging people into checking and getting things right much earlier in the system. We are just beginning to play with the value of doing that for individuals as well as, as Nick says, with tax credits, but I do think we’ve got to learn to run before we can walk. I am confident we are now basing income data of a wholly more up-to-date PAYE system than we were when I first sat in front of you three or four years ago.

Chair

I think we would acknowledge that there has been a massive change and improvement, but it is interesting to know what the accuracy rate is. Mr Devereux, from your side of things, how are you going to tackle this?

Robert Devereux

Using the same technology, effectively. The way in which we are constructing the universal credit declaration, on a monthly basis, is to give people the opportunity, including on their smartphone, to say, “Actually, these are the numbers for me this month.” Instead of it being a system where at worst, under tax credits, you are asked at the end of the period, this is something which time after time they will be declaring: “This is my position. Do your sums on this basis”—with, by the way, the exception of the earnings, which we are getting from a third party to stop them making the errors. I genuinely think that the combination of the work that Lin is doing with employers and the design of the system that puts in front of them on a regular basis—remember: this is working two ways. It’s not just about getting the benefit correct; for a lot of these people, we are also trying to encourage them into work or to change their hours, so the dialogue between us is a continuous thing. In that context, even when they are out of work, they are getting used to engaging with us much more frequently and remotely than they are used to at the moment.

Chair

Well, I hear lots of things, but this is not the place to discuss the ins and outs of things like Universal Jobmatch—the interface with people. But let’s face it: they do not all run smoothly. We’ll come back to that at another hearing. The fraud and error rate that you predict when universal credit is rolled out is 2.5%. Is that right?

Robert Devereux

No, 2.5% is the figure in that table for aggregate welfare benefits over £225 billion—

Chair

Okay. Once it’s down to that level, though, can we not be working to prioritise prevention?

Robert Devereux

The question then is: what do we think will be the principal issues in that projection? We know that at the moment by far our biggest problem is undeclared earnings and we know that the combination of RTI and the design of UC will help that. The next one up is going to be household formation.

Chair

Which is a huge challenge.

Robert Devereux

It is a huge challenge because it’s technically difficult to do. I could make it go away immediately if we just decided that we would pay a couple twice—two singles—but we don’t do that, because we think, perfectly properly, that they only deserve 1.7 or something, thus there is a very strong in-built incentive for you to misdeclare. So what we’re trying to do there is to say, “Well, what are the cute ways of doing that?” I do think that the Revenue have found some ways, through the way they’ve been thinking about life, which we would hope to borrow as they develop their systems. But that is an intrinsically difficult thing—what is going on in household formation. For some of the rules that we apply, when you come to modern relationships, it is quite difficult to be really clear about what it is, explain it and police it from a distance.

Chair

And codifying it—we recognise that there are challenges there particularly. One other thing is that we have talked a lot about is changing the design of the schemes, whether that is the rules or the interfacing, but that requires, as you highlighted, more frequent reporting requirements. That will have an effect on claimants. Some of my constituents will find that online interaction difficult and others have literacy or language issues. Have you modelled the impact on those claimants? Have you looked right down to that granular level? You are both dealing with huge programmes but, at the bottom level, they impact heavily on individuals. What have you done to mitigate against the impact on them?

Robert Devereux

If I go back to the spending round conversation that we have had, in building up what the cost of my Department might be in 2020 with universal credit, we have gone through in really quite a lot of detail. We say: at the point of initial claim, what assumption do we make about the people who are happy to do it online, are capable of doing it and get all the way through with no intervention at all? Answer: x%. How many of those people online might actually want to ring a friend in a contact centre as we go along? How many of those people would be happy to do it online in one of my offices with a bit of help? We have gone through it item by item. We have not simply said, “Good news. Online. 90% pots of savings.”

Chair

No, we are not suggesting that. Even if you are not doing it online remotely from your home, if you are lucky enough to have broadband and internet access—which, in Norfolk, you don’t—it is still a challenge. If you talk to anybody about universal jobs match, the people who are going through that find it quite clunky at times. There is an impact on them, even when you have a system that you have modelled and checked all this on.

Robert Devereux

All I am saying is that we are not putting a bid into the Chief Secretary to say, “This will all be online, you don’t have to worry about it.” I am actually pricing in help, whether it is my own staff or whatever, and that says, “You cannot assume that this will all be straightforward.” The size of my Department in 2020 will reflect my best assessment at the minute of how much effort people will need in the way of hand-holding to run a system, even when it has been well designed. For the great majority of people it will work well but, for a lot of other people, we will need assistance.

Chair

You have done some modelling. You talk about the percentages who would be able to do it easily and so on. Have you tried it with real people?

Robert Devereux

Yes, we have.

Chair

I know you have with the early bit of UC but, more properly, with the changes that will be coming down the line. Do you try that with real people?

Robert Devereux

Yes, we do. As you know, we have been doing this in Sutton and we are going to Croydon. At the moment, the percentages of people getting through without any trouble seem to be a lot better than some of our more conservative assumptions but—

Chair

But you are dealing with simple cases at the moment, aren’t you?

Robert Devereux

Not in the service that is running in Sutton, which is wholly built on a digital system. It is all-comers. That was done that way very deliberately. We chose Sutton—I hope I don’t offend anybody—because it is a very average place. As far as our statisticians are concerned, it is as average as it comes. We are trying to get some sense about what the spread will be because it is critical to projecting an entire Department. If I had many of these people, they would be in different categories.

Chair

Can I just ask one more about the impact of the changes on more vulnerable claimants—this is for things like pension credit and PIP? I don’t need to tell you this, but we have had a Minister standing up in the House and apologising about PIP before it even gets any further, so there have been problems there. Have you modelled the changes—

Robert Devereux

The principal change with pension credit is that you will not be claiming it in anything like the numbers that people are claiming it at the minute. People simply will have no entitlement to it because their state pension will have floated them off any entitlement to means-tested support. As far as pensions are concerned, that design change is a huge difference to their ability to interact with the system. They just get a pension—done.

Mr Bacon

I have one specific question about something that you were discussing with the Chair a moment ago. Mr Phillips has just been berating me for having this old-fashioned BlackBerry rather than a posh, modern smartphone, but I like simple technology. I was in Argos the other day buying something. They used to have a 1400-page catalogue. It doesn’t exist anymore; it is just a large tablet fixed on a stick. It is very simple to use and almost impossible to get wrong, even for someone like me. When you are designing your systems for use in centres where people actually turn up, to what extent are you looking at what is done in the private sector—for examples, in places such as Argos or others—and copying the good ideas that you see? Are you doing that?

Lin Homer

Yes, and more than that we are saying to the designers of APIs[1] and—

Mr Bacon

APIs? We try to avoid TLAs in this Committee.

Lin Homer

These are the systems that work on other people’s products—mini-programmes that are available on your smartphone.

Chair

Apps.

Lin Homer

Yes; well, the program that sits behind the app. We are trying to design ours so that they are tested with real folk. If you remember RTI, we introduced it quite slowly with groups of people, and we iterated it as we went. What we are also saying to others is, “If you can design a gateway into our system that is simple to use and that people want to use, and if the security docks in the right way, we will let them come in via your system.” I mentioned software. Most small businesses—there are 5 million of those—will use some form of software to run their business. If people design that software so that it completes our needs as well, the business just uses the one piece of activity and we take our feeds off that. We do not even think we copy them; we let them do it for us, if we can make the interfaces work.

Mr Bacon

So there could be an almost infinite number of different front ends from the consumer’s point of view, but you are saying it doesn’t matter so long as—

Lin Homer

Provided that the interfaces work. If you like to use old-fashioned parlance, for those of us who still hanker somewhat for the old days, it is a bit like a kitemark. Mark Dearnley will invite developers in, and they will joint-develop with us. We will check that we are satisfied with both the cyber-security and the ease of use, and then that product will be allowed to join with us. We think that that is the way you keep the best, most modern design, because we doubt we will always be at the leading edge.

Chair

I want to ask Mr Devereux about sanctions. Again, speaking as a constituency MP, that comes up for me all the time. What level of sanctions have you applied across the board on benefits in the last year?

Robert Devereux

I am guessing you mean penalties for non-compliance, not sanctions for not—

Chair

Yes, penalties to claimants, particularly of JSA.

Robert Devereux

At one point in the preparation I had some numbers. Do I have the volumes here? This may be too long a period, but since 2012, we and local authorities have imposed nearly 70,000 penalties for fraud and more than 150,000 civil penalties for claimant error.

Chair

For claimant error. Does that include fraud?

Robert Devereux

No. I gave you one number for fraud and another number for civil penalties. With civil penalties, by definition, there is no admission of guilt.

Chair

Sorry, I cannot hear you. Can you speak up?

Robert Devereux

I am sorry. The paragraph says “Since 2012, DWP and local authorities have imposed nearly 70,000 penalties for fraud and more than 150,000” civil penalties for claimant error.

Chair

What evidence do you have about the impact of punishing people on changing behaviour at a later stage? Does it have one? We often have this discussion vis-à-vis tax about the deterrent impact of punishment and whether it makes people stop doing it in future, or sends out the message more widely to other people that they will get caught, versus just happening once when a person gets a sanction and the money comes in, if you have put a financial sanction on people.

Robert Devereux

Sorry, are you asking me whether I know that people who have had a sanction—

Chair

Are more or less likely to do it again.

Stephen Phillips

There are two questions there. There is the effect on their behaviour, and the effect on the behaviour of claimants generally as a result of publicity surrounding the fact that you have imposed sanctions.

Robert Devereux

I don’t know that I have information on the first question. I will go away and check. Because the sanction regime has been changed significantly in the last couple of years, it is quite hard to know some of that answer on a consistent basis, but let me go and find out. In general, ever since I was first doing this in 1998, all Governments have sought very consciously to keep in the public eye the fact that this is not what the welfare system is about. We deliberately take cases to court to make a point of a deterrent nature, because these cases cost us money to prosecute.

Chair

It is difficult to assess the impact completely accurately, I recognise, but you are giving quite a vague answer, if I may say so.

Robert Devereux

I am giving you a vague answer because I am not confident that if I went back to the ranch and said, “Prove beyond peradventure what the deterrent effect of prosecutions is,” it would be a knowable answer.

Q108Chair

Do you know about people who are repeat offenders, for instance? Do you have figures like that?

Robert Devereux

Yes, I get the line. Can I write you a note because I just do not have the data with me?

Chair

Please do, because it is quite important that we get a handle on it.

Robert Devereux

If it turns out that I don’t know, I will let you know, too.

Stephen Phillips

When did you last ask the Department’s economists to look at the financial benefit to the Department of the prosecutorial line that you take in terms of the number of people you prosecute?

Robert Devereux

Clearly, over the course of the past few years, we have significantly changed these to civil penalties to get out of the very large costs. The answer to the question of whether I have asked the economists to do that is that I haven’t, so let me go and check.

Stephen Phillips

Do you think it might be a good idea to get them to look at the enforcement regime—criminal and civil—asking, “How much this is benefiting the Department? Could we do it in a different way that might benefit us even more and let’s model that?” Isn’t that a piece of work that ought to be done?

Robert Devereux

I am making a distinction between whether I have personally asked them—I have not—and whether the question is none the less being asked by officials, which it may well be.

Stephen Phillips

I appreciate that you cannot know everything that is going on. Perhaps you would let us know, Mr Devereux?

Robert Devereux

I will.

Chair

I have a few quick points to raise—some of them, forgive me, are slightly off the Report, but I will raise them as you are here—and Teresa Pearce is going to come in on something to do with tax credits. On credit reference agencies, which are referred to in the report, HMRC is using them very effectively, but we are told that DWP does not use credit reference agencies in the same way. Why not, and are you going to try to learn from HMRC?

Robert Devereux

The answer to the second question is yes. The answer to the first question is that we do have access to them in cases of individual investigation. So, if someone says to me, “Mr Devereux, here’s a bad thing,” I send a fraud investigator off and they can go and find the credit reference agency data. The thing that we tried to do was to see whether credit reference agencies could help us to identify ab initio potential cases of risk—looking at your data, can you spot people who are living together or whatever it is? We got really poor results from that question, partly because they were finding people who were living together who turned out to be sisters, brothers or uncles. They were all at the same address and all had different accounts, but they had not established any economic relationship between them. If I have understood it correctly, what the Revenue has done is that because it has been building a system progressively over time primarily to chase after tax and to find out things, it has built a system that has now got slightly better information in it about some of the things we are interested in. Both of us know whether we are running more than one bank account into a particular organisation. You have a bigger incentive to run more than one bank account if there is a bit of tax avoidance. I think that Lin is ahead of me. Our view is that where the Revenue has used credit reference agencies on top of risk assessment done in house, as it were, it has got a good result. We were using them without the benefit of having that, partly, coming back to the answer I gave you earlier, because we had not constructed the IRIS system, but in principle at least the data that we now have access to—I have the same access to historical records as Lin—should mean that in the fullness of time I will get the same benefit from it, which will be about trying to make sure that we have gone beyond a shot in the dark. I am afraid that the credit reference agencies are not really more capable of doing that than we are.

Chair

Okay, but are you going to be using them more before UC comes in? Universal credit, I should say, in case anyone is listening.

Robert Devereux

The short answer is that if we can demonstrate that we have a better a priori view that the following 20 people might be worth the credit agencies washing them through their system, I think we will do that, but we have to establish that we can get to a better a priori view rather than giving them 100 cases and their coming back with rubbish, because it consumes lots of resource checking up only to find that it is two sisters.

Chair

It is useful to know that. The single fraud investigation service—how is that going?

Robert Devereux

Fine.

Chair

A one-word answer. Perhaps we could have a little more, Mr Deveraux, on this occasion. We like short answers, but a little bit more than that would be good.

Robert Devereux

Jon, why don’t you have a go?

Chair

Mr Fundrey, you have been sitting here for a long time, so perhaps we can hear from you.

Jon Fundrey

Thank you. Very well.

Chair

Surprise surprise.

Jon Fundrey

I would say that, wouldn’t I? We have 244 local authorities signed up and 136 to go, and 1,000 staff transferred. The transfer is going very well. I think it is one of the biggest TUPE transfers, or COSOP transfers, in Government. It is running very well and we are finding that the benefit we hoped for—more pan-benefit investigations—is being demonstrated.

Chair

Are you confident that that will continue in the future? One of the concerns raised about this was that we would lose that local granular knowledge. I think, Mr Devereux, that from memory you said that that will not happen because they will be in local teams. They are being TUPE-ed over, and that is fine now, but what about in three years’ time? Are you predicting that we will keep that local granularity that you promise?

Robert Devereux

Yes. It is not with SFIS for this. What we have found in the work we have done on the universal credit to date is that because it incorporates universal credit, you need people locally for them to understand what the housing market looks like. Whether it is investigations or the administrative side, there is something intrinsically local about housing, so we are trying to make sure that we have the right staffing, including local authority and ex-local authority staff, to make sure that we have that local knowledge. We have a very strong interest in keeping that refreshed.

Chair

Forgive me; I should have warned you that we would be going into this a bit. Will you be benchmarking from what you inherited to where it is going to be in a few years’ time when everyone has transferred over? I have had concerns raised with me that there will be a lot of loss in detection because of the change, and we need to follow that through as a Committee.

Jon Fundrey

Yes, we will. To be honest, when we modelled it, we also assumed that there would be initial drops in productivity in the early days. We are not actually seeing that. It is early days, and we will be tracking it.

Chair

Okay. We will be coming back to that, I am sure. Ms Homer, this is slightly off the subject, but while you are here, I am interested in how the Customs side of your operation is going, particularly on fuel fraud in Northern Ireland. Do you have anything you could tell us? I am not going to hold you to every figure, because I haven’t given you prior warning.

Lin Homer

I am not sure I will be able to give you absolutely every figure, but as I am sure a number of you know, the laundering of red diesel is a big issue, not just in Northern Ireland; with the boundary with southern Ireland, there is a particular angle. We have recently introduced a new marker, and we have recently improved our ability to do roadside testing. We have had some initial bedding-in issues, and we discovered we had to finesse one of our protocols around testing, to make sure that the tests were valid. But we are confident that we are now making it much more difficult for people to pass off red diesel as fuel. All I would say is that I think the XST went on a visit himself recently, and not everyone is bothered about whether they are hiding red diesel. On one of the raids that he was involved in, it was unadulterated red diesel, so we should not think that markers will be the sole solution. We need to continue traditional enforcement as well as markers to make sure that the incentives for avoiding that duty are minimised.

Chair

One of the concerns that has been raised by the Northern Ireland Affairs Committee and colleagues representing Northern Ireland in the House—10 years ago, I was looking at this when I was on the Northern Ireland Affairs Committee, so I know that the fuel problems are an issue—is the Dow fuel marker. That is the new fuel marker that you have brought in, which one of your staff, Pat Curtis, who is the specialist investigator into fuel laundering in HMRC, described as “ground-breaking technology. There is no other trial of these...in vehicles like this anywhere else in the world.” That was in June this year. I have been told that there are serious problems with this marker and that, when distilled, it disappears. Is that true?

Lin Homer

No. We have been seeking to test the Dow marker under laboratory conditions, and so far we have not found it possible to easily and readily launder the marker out. However, it is our experience over a number of years that the criminal profits to be made in these areas—a bit like the importation of tobacco or alcohol—are such that it brings all sorts of ingenuity to the fore. We would intend to—

Chair

So they are getting ahead of you?

Lin Homer

Well, they catch up quicker than we would like, rather than getting ahead. As I say, so far this marker has proven to be very difficult to launder, even in laboratory settings, whereas earlier markers could be relatively easily laundered out in still substantial—you could spot the factories with heat-seeking cameras from above, but when you got there, they were fairly crude. This needs much more sophisticated machinery, and we think still leaves traces. We think it is robust. Probably, it would be wrong of us ever to declare success for all time, and I think we will continue to look at the market to ensure we know both what the criminals are up to and what the other developments are in the market. Both for the criminals to cheat us of our marker and for other people to bring products to market—both are sensible markets for us to keep our eye on.

Chair

Okay. Just to flag up to you, if you are not already aware, that that is a very big concern to Northern Ireland colleagues, who were particularly keen that I raise it.

Lin Homer

Yes, I am very aware of that, and the Minister is very keen that we stay in close discussion with Northern Ireland[2].

Chair

Great. I am going to ask Teresa to come in on tax credits and enforcement.

Teresa Pearce

Ms Homer, I believe your Department has entered into a three-year contract with Concentrix to help you bring down the fraud and error bill. It is reported that it is a £75 million contract. Is that correct?

Lin Homer

It is a three-year contract, which we think will be between £50 million and £70 million if all goes according to plan.

Teresa Pearce

Is it a payment-by-results contract?

Lin Homer

It is payment by results, which again is perhaps an indication of something we talked about earlier. Our initial estimate was that this might bring in as much as £1 billion. Partly because circumstances have changed and the level of fraud in the system is now lower, our current view is that there is about £420 million of AME savings to be made. If you look at that in broad terms, there is getting on to the 10:1 or 11:1 return that we generally make. That is, in a sense, an extension of resource in the private sector to do what Nick is already doing in-house in a very similar risk range.

Teresa Pearce

Do they have set targets within their contract?

Nick Lodge

The contract is published in accordance with the general rules, but there are various elements of the contract around performance, not just savings but other elements of performance that you would expect such as serving customers and that sort of thing.

Teresa Pearce

When they take up a case and write to somebody, who has indicated that that is the person who should be looked at? Is it the Department, or is it the company?

Nick Lodge

As Lin said, this is an extension of what we already do. We have around 1,500 people inside HMRC carrying out checks throughout the year, as we described earlier, to help to get awards on the right footing.

Chair

Sorry. Could you speak up a bit, Mr Lodge? There are bad acoustics in this room.