Local Government, Housing and Planning Committee

Yes. Good morning, all. Thanks ever so much for joining us today. We're undertaking this important bit of evidence around elections. I want to look first at evaluation of the pilots. We'd like to understand what the challenges were in delivering for you. How was it? What's your view of what the pilots were telling you? There are going to be lots of connected questions, especially as we're moving a lot closer to the next elections. Can you give us a flavour of what you found and what the challenges might be? Or was it good? You may not all want to answer the same question, I accept that.

Everyone's looking at me, so I guess I'm going to begin. As a pilot authority, ours is quite a lived experience. They were successful in Powys. It was possible to use local data sets to identify individuals that weren't registered and send them a notice of registration. Again, those notices of registration were generally quite well received. We had a very low number of opt-outs. Obviously, where people did feel strongly and did not wish to be automatically registered, they felt very strongly, as I'm sure you've all experienced on the doorstep sometimes. Overall, as a process, it was successful. We found just over 5,000 individuals, which worked out at approximately 5 per cent of our register. I think there was overall confidence that the people we found were accurate. They had a link to the property. Having the assurance of finding an individual on two or three or even four data sets gave us that assurance that there was quality and accuracy in that result.

I'm sure, from the elector's perspective, it felt very automatic and very seamless. From the administrator's experience in the back office, it was probably anything but. There was a lot of manual intervention and work, which I think the Electoral Commission report does highlight. Data matching relies on good-quality and standardised data, none of which is what we have in local authorities. A lot of the data sets have been created for other purposes. They're maintained not because that person is resident at the property, but maybe for another reason, et cetera. So, it is difficult to evaluate certain data sets and their usefulness until we actually receive them. In fact, there were two or three data sets that we received that didn't have addresses on, or the quality and completeness of their addresses weren't good enough for us to even use. So, it was balancing that expectation of what we requested and what we thought we were going to get with what actually arrived.

Data cleansing the data sets that were of use was quite time consuming, stripping out the incomplete records, ensuring that they were all of a standard good enough to take forward. And even within those data sets, obviously you have inconsistencies that an individual might use with regard to their name or even an address, and I'll give an example of that. On one data set, somebody might be known as David John Edwards. That individual might also be known as Dai John Edwards. On another data set, it might be D.J. Edwards, or just good old Dave. Trying to match all of those up and make the assumption that that's one and the same can be quite difficult. And it's obviously not 100 per cent then unless you've got some other information such as a date of birth to link them. To add to that complexity, then we add children into the mix. Something I always say is that parents can sometimes lack imagination when they name their children and we might have another D.J. Edwards at the property. If we're lucky, it might be a David Jasper, for example—same initial, but at least there's a distinction of a middle name. If we're very unlucky, the worst-case scenario is we have a David John Jr. Again, there are quite a few data sets that don't have 'Sr.' or 'Jr.' So unless we have a date of birth to identify the difference—. All I can say is opening the post in those addresses must be quite fun. And this happens more often than we realise.

Those sorts of issues meant that there was a lot of manual checking and, even after the data match exercise, human eye-balling the records that came out before we then actually wrote out. And it's not just elector information; property addresses come in all sorts of incarnations. There is a standard for an address. There's a unique property reference number, which some of you may be aware of. It's called a UPRN. However, again, in Powys, our council tax records don't have a fully complete UPRN on every address. I know that some local authorities across Wales do have that, because when they data match during the canvas, that makes their life a lot easier. I think there's about 50-odd per cent of our council tax that have UPRNs. So, that was the level of accuracy: we know that, if that number and that number matches, it's exact. However, for the other proportion, that's not so possible. And there are a lot of the data sets that don't have UPRNs, so any discrepancies in the way addresses have been typed up—. For example, you could have '1 High Street', you could have 'Flat 1 High Street', 'Flat 1' without a property name and 'High Street'—are all of those incarnations the same? If you didn't have a UPRN number or maybe a surname that linked them, that sometimes, as well, could be quite difficult, and we did find incomplete addresses being a problem.

Those are very clear lived challenges that you managed to get around. Do you believe that, if that was rolled out all over, there are answers to those challenges, which can make this truly deliverable? Or would we have a half-cocked process that couldn't manage?

It requires a lot of manual checks. After the matching and the data mining, there needs to be a lot of quality checking. We spent probably another further week just going through the 5,000 records, just checking, doing some spot checks, ensuring that there weren't any duplicates. We had a couple of issues with the software, just a bug that didn't link people correctly, so even though they had been identified, they weren't linked. So, it just requires checking, whereas now we operate on the fact that an elector or an individual has been proactive and sent us an application that we then verify. We don't want to write out to households to put somebody in their property and register them if they've got no link, because that will undermine the trust in the register. And that wasn't the case with those 5,000 individuals. The records and the checks that we completed obviously did work, because we had very few phone calls—I think only a couple—where somebody rang us and said, 'Actually, these individuals don't live here.' And that was only with the education records where we had relied on that one data set; on all of the others, they were complete enough.

The other problem with the process was that we obviously identified another 12,000-odd people with one data set. So, whilst they couldn't be automatically registered, we then had to follow them up to make sure that we weren't missing—. We obviously couldn't do anything with them with regard to the AVR process, but we have a duty as an ERO to follow those up and encourage them, if they are there and they are genuine, to register to vote.

Yes. Are there any other—? Does anybody want to add anything else on additional challenges?

I think in terms of what Sandra's outlined, obviously it's Powys's experience, but the implementation of the pilot was over quite a long period of time, and I think our concerns on the manual process—and, as we said, there's nothing automatic, from an administrator's point of view, about automatic registration—are the amount of time it takes to manually cleanse the data before you do the data match. Then there's the manual quality assurance process—the time that takes and the window that we're looking at in doing that, and the fact that administrators in other authorities, who haven't had the experience of the pilots, were in a very condensed time frame. So, while some risks can be mitigated in terms of that, I think the biggest risk we've got is this short time frame we've got to be able to do that in the other 19 authorities that haven't had the experience that Sandra and Powys have had in that respect.

I think, in terms of the reliance on manual systems, when a human is responsible for doing those data checks, there's always that element of risk and error and things as well. And I think our biggest concern is that, if this process appears to be rushed in any way, shape or form, or sufficient time is not given to allow for people to do the work correctly, there's the risk of adding people incorrectly to the register, because those checks haven't been done fully or properly. It's the impact that that has long term on the integrity of the process and on the delivery of elections going forward as well.

And I think, again, it's also those concerns around the access to information as well. We're very much aware that a lot of authorities in Wales are currently struggling to access the local data sources. They need to do this exercise. There are regulations, but the regulations that are relied upon that allow electoral registration officers access to things have always been a barrier. UK Government have recognised this and they're currently, in the Representation of the People Bill, trying to address that to make sure those barriers don't exist. And it's a shame that Welsh Government haven't carried out those things first to allow greater access to those data sources to take away some of those barriers that currently exist as well. So, again, it's about having that time period to make sure that the legislative framework and the digital framework are in place to successfully allow for what could be a really successful initiative to get more people on the register and to exercise their democratic rights.

Thank you.

Just maybe a couple of comments. Obviously, you've got some brilliant lived experience from Sandra in Powys, who's obviously undertaking that role on a daily basis, and Clare representing electoral administrators. I guess, as an electoral management board, we both have a statutory role in terms of elections, but also a co-ordinating role. So, I suppose I'm going to give more of an overview, and I think one of the things Sandra's pointed out is that that was successful in Powys. Not all electoral teams are the same in terms of their size and capacity. I think it's fair to say that they're an extremely skilled workforce, so they have the skills to do it. I think, as Clare's described, it's the speed and the timing that's the issue that needs careful management.

As an electoral management board, we've co-ordinated an impact assessment to try and make sure those impacts are really, really clear, and there is a Welsh Government AVR working group in terms of making sure those impacts are mitigated. I think that's just really critical that there's momentum on that—so, those risks around software, timescale and also funding. And we're grateful that funding has been given to support the additional resources that are required, but it remains a live risk in terms of delivery for the 2027 period, and I think just being tight around managing that live risk is really critical and important.

Thanks for that. There'll be several other areas drawn out later on that refer to many of the points you've raised.

If I could go on—. Should I go on, Chair, to notification periods, or does somebody else want to come in?

No, it's just you, Peter.

So, if I could explore briefly a couple of points around the notification period to opt out, because you'll know that, originally, the pilots were moving forward with a 60-day opt-out. There have been various recommendations to go to a 28-day opt-out, and then there is a set of concerns from similar people saying, 'Well, yes it's a good idea, but what about the safeguarding side of that?' And I just wondered if you could give us a reflection on the benefits and potential drawbacks of that shorter 28-day period as you see it. That would be really handy.

I think that, in terms of the benefits, if this is to be rolled out before the local government elections next year, you have to have a 28-day period. You've run out of time to have a 60-day period without it impacting detrimentally on the delivery of the local government elections next year. With a 60-day window, the earliest we're looking at notification letters going out—. Given that we haven't got legislation in place yet, the EMS suppliers haven't got what they need, the Electoral Commission have had to produce guidance before the legislation has come out, again, we're in a position where January, early January is the absolute earliest we can look at notification letters going out. A 60-day window would then push you into those people being added to the register during the notice of election period. So, you wouldn't have poll cards for those people until later, which, again, if we're introducing people to a democratic process and then putting barriers in their way or sort of not giving them a good first experience is detrimental there. So, having a 28-day window in this case to allow you to have those people on the register by the 1 March notice of alteration is crucial to make sure that they're included in the data that electoral administrators will send to printers to administer that election.

There are obviously drawbacks, I guess, in that, if you've got anonymous electors or if you've got ineligible electors who've unintentionally been sent correspondence—. A longer period obviously gives people more time to respond to that. But I think that, in terms of what the pilots showed, most people tended to respond within a two to three-week period in that window, so that 28-day period would be sufficient. I think Sandra's experience, which I'm sure she can elaborate on as well, is that when they sent reminders during that 60-day window, because it allowed for that to happen, you sometimes still had people responding to it, but generally speaking, I think the pilots have shown that those risks were relatively minimal and a 28-day window is sufficient.

I'm conscious that a 28-day period might need to be put in place so that we can get ahead with this election, but if that wasn't in the way, is a 28-day period still right?

I think 28 days allow electors a sufficient period of time in which they can raise concerns. I think the key thing is around the messaging that sits behind it, so the communications campaign around this taking place, the messaging on the outer envelopes that are sent to make people aware what the content is and the need to engage with it, and obviously the communication, then, that's within that correspondence as well to make people aware, if they are a vulnerable elector and need to pursue anonymous registration, what their options are in terms of that to mitigate some of those risks and concerns. The Welsh Government has recently passed legislation that removes the open register for the local government register, so, again, that takes away some of those risks about where their information would appear as well.

Some of the points—. I'll raise this question, and you may all want to come in on it and I'll be done then. Some of the issues that have been raised—. As I touched on earlier, there were some safeguarding issues. Certainly, One Voice Wales could see the benefit of 28 days, but they also had great concerns about, perhaps, people who might be stuck in domestic violence situations or who might be disabled and might not have good access to communications or who might face challenges in using communications. Have we got the balance right? Would the balance be right between 28 days and enabling, or making sure that we have those safeguards in place?

Going on from what Clare's just mentioned, from our experience, I think 28 days would be enough. I think that we were the only local authority within those 60 days to send a reminder out. So, we had three communications: the initial notification of registration; then about 25 days later we sent a reminder, a prompt—'You still have until the end of August to opt out of the automatic registration, but obviously the chance to register fully'; and then the final notification. With regard to the phone calls and the opt-outs, the early opt-outs with regard to concern about safety and safeguarding were all in the first tranche of phone calls. Anyone who wanted to explore or needed to consider anonymous registration all rang us within the first two weeks of receiving that communication. Any parents who felt strongly that their child should not be on the register at that age rang and wanted to opt them out at that point.

During the reminder stage, or the prompt stage, as I call it, we did have opt-outs. We did have a handful, a very small number; we only had 28-odd for the whole of the mailout. But we had a few, and I took some of those phone calls myself—I happened to be in the office—and when I questioned, 'Why did you leave it until this point when you had the time?', it was, 'I just put it on the mantelpiece' or 'I put it in the recycling'. So, they weren't concerned about their safety, they just wanted to opt out. They'd read it fully by then, considered it and obviously wanted to opt out at that stage. When we sent the final notification, again, we had, I think, just one person querying something, but nothing with regard to safeguarding. All those phone calls with concerns about their name appearing on a public register, or because of their safety or circumstance, all happened within the first tranche, and our team were able to advise them accordingly.

So, based on that experience, I think 28 days would give sufficient time. I do think—and Clare made the point, and it was something that we experienced—that we need to have clear messaging on the envelope. It's amazing how many people receive an envelope and don't do anything with it. Certainly, that would also aid with the accuracy, where, if somebody has been matched and identified against the wrong property, what tends to happen is the householder just thinks, 'Oh, they might be the past occupier' or that it's a bureaucracy or admin error, and they just chuck it in the recycling. We need them to phone us. 'If this person does not live at your property, we need you to contact us immediately.' And, again, with regard to safeguarding, there needs to be a clear message that if you do not wish to be automatically registered, you need to take action by a certain date, or immediately. So, I think there are mitigations around that, with clear communications and flagging that.

And I think—. We had a by-election. We had several by-elections, actually, during that 60-day period—never a quiet time—and, obviously, that 60-day period is quite difficult. You're telling people they're going to be on a register at a certain point, but, actually, there's an election in the meantime. It's about the messaging to the public. 'Oh, does that mean I can vote for this one?' 'No, it doesn't, you need to do something else.' And 60 days was quite a long period for that, so I think 28 days probably sits better with an election timetable, because we don't just have elections every four years, or even once a year; they happen throughout the year, so it's something that, for the electorate, needs to be a bit more immediate.

Two summary points, which I think both Clare and Sandra picked up. It feels an appropriate balance, but the communication is absolutely critical. And I think what electoral teams have got really good at is engaging in ways that reach people who haven't been reached in the past. So, I think using those networks to communicate, as well as campaigns, is just really critical.

Thank you, Chair.

Thank you, Peter. Mike, do you want to come in on the registration application exercise?

The first one is about areas with high turnover. These are areas with students, but also areas with lots of houses in multiple occupation. And Cathays in Cardiff, Treforest in Pontypridd, Uplands in Swansea meet those criteria. How are you going to make sure you get those matched as people move?

Me again? [Laughter.] I think the frequency—. The flexibility in the regs does allow EROs some idea of timing an AVR exercise that might suit their circumstances and their type of local authority area better. I would say that AVR isn't the only mechanism that the electoral registration officers have to get people on the register. We already do quite a lot of work, and I think this is pretty much commonplace across local authority areas. We regularly receive updates from our council tax team with regard to movers coming into the authority, account changes, and most electoral registration officers and teams would be fairly proactive in reaching out to those individuals and encouraging them to register because they've moved address and, therefore, they will have, potentially, lost their right to vote at their last address. So, I think electoral registration teams have got quite good at becoming proactive and chasing up changes to addresses. So, I would say that AVR isn't the only mechanism we have to get these people registered. Obviously, it does require a little bit more input from those individuals themselves.

You will have a lot of students joining in the next couple of weeks. So, the matching system and the work with universities will, obviously, help you with that and you'd hope to get the halls of residence and other university accommodation picked up. So, this is the time of year to do that, isn't it, for universities? But, for people in HMOs, people in other types of accommodation who move very regularly, people in bedsits who move very regularly, people who are in supported accommodation who move fairly regularly, how are you going to make sure you keep on picking them up?

[Inaudible.]—as you said, with those particular demographics, it is having the local data available and that that local data is as up-to-date as possible to allow you to automatically register them. And I think the actual register, in an ironic sense, is probably the most up-to-date record that we've got in a local authority, in terms of the fact that you carry out a canvass every single year and the information is verified. And, as Sandra said, looking at alternative methods in terms of those, because having access to data that accurately reflects those high churn types of properties and things, is a challenge. There's no two ways about that.

I think the point around university students is quite an interesting point as well, as you said. It was interesting, during the pilot, when you had the data matching taking place in the May period, the risks that that sometimes can introduce as well. You would have automatically registered a higher volume of university students in one of the pilot areas, but, due to the nature, as you said, of student registrations, as soon as you'd put them on the register, they'd effectively moved by the time that you published the register. So, again, there are concerns around how different demographics interact, the availability of data that you've got that allows you to do that, and also the mechanisms that you have then to review the register. For types of properties like that, there's a direct registration process that the Welsh Government are proposing, but there's not a way of automatically removing those people that you can see where there are those high churn areas as well. And our concern is that you can end up then with register bloat, potentially, where you're adding lots of people and so you're getting a more complete register, but whether that register is more accurate is questionable, unless you've got the mechanisms to easily remove people and identify them to take them off the register. Again, the UK Government's Representation of the People Bill is looking at introducing that side of it as well. So, we've got concerns, which I think is what is being stated here, about how you get these high churns on. It's also how you get those high churns off to make sure that register accuracy is maintained across the board, and also then trying to find those data sources that more accurately help you identify—. For universities, it's clear where that information might come from. But, for things like HMOs and things, it's having access to information that local authorities have access to, to be able to accurately and quickly be able to do that as well.

I think the point that it brings out, which has come through the other questions as well, is those individual issues are—. The registration teams are skilled and capable of managing those and, spending that time having overseen a registration team with a high university population, they were very good at making sure there are additional things in place to make sure we didn't miss those.

It's the cumulative effect of doing absent voters, overseas voters, of doing people who've got high churn, in a quick timescale, I think, which is probably the biggest area of risk to manage in this particular time frame. Over a longer time frame, in terms of future elections, I think that all becomes more manageable. So, I think the large risk rise is in the current situation, with the implementation for 2027, with all of those factors that are sitting with electoral registration teams. They are manageable, but they need really tight management, and reviewing how that's gone is probably critical because of the fact that it's the first time and, also, those delivery risks are not insignificant.

My final question is about supported housing. Are you getting data on supported housing so you can contact the people who may be in the Wallich for three months and then move on to another organisation for six months, and then move into a HMO?

I think our experience is, from what we've heard from electoral administrators, is that a lot of them are struggling to get even access to council tax data in their authorities at the moment. So, while there are provisions there, as we said earlier, that allow them to have access to that, there are definitely difficulties in getting the relevant other teams in the council to provide the information and to provide it on the frequency of the basis, particularly in those sorts of circumstances, for them to actually have that. I don't know whether, Sandra, your experiences are—

It's the same. Whilst we are very lucky in Powys, as we have good relationships with our department—I have an excellent data protection officer who backs up the EROs' power to request data—having regular data sets updated is a problem. That will be the case across lots of local authority areas, just because the teams themselves that maintain those records, they have to supply them, and that relationship needs to be built up over time, not just ad hoc.

Okay. David.

How confident are you that there's no duplication? Because, obviously, when a person's registered, say, in Powys, and they go off to university in Pontypridd, they register there and they register in Powys. How can you tell if they register in two places so that they're able to vote in two places?

I think, in that particular example, in terms of university students, university students are eligible to register in two places. I think what we're concerned about is that if you are using university data to automatically register students, is the unintended risk—well, they can be registered twice; they can't vote at the same election twice—of them unintentionally being in a position where they could do that and not understand the consequences through automatic registration. While it's correct that they can be registered in both places and that could possibly be the case, I think our concern would be more what Sandra alluded to earlier, that because of the nature of the way that names are on different sorts of documentation and things, and different council records, there's the ability to almost register the same person twice because of differentiations within their name. There's a risk of that because you haven't necessarily got the additional information that can uniquely identify somebody, such as a national insurance number or a date of birth, within local government records. So, using local data, there is a risk of duplication.

I think we referenced in our written evidence that the Democracy Club, who administer polling station staff, and Financial Data Intelligence Framework, who have huge expertise in data—far better than I definitely do—have explained some of these things. They've got huge concerns over the nature of fuzzy matches, which then naturally lead to potential for duplications. You've got the manual checks that kind of mitigate some of that, but, again, without additional information that some national data sets, for instance, would potentially have, for you to be certain that it is a unique, different person, or that it's the same person—. That's more our concern, I think, in terms of duplications.

Thank you very much.

Thank you, David. Elyn.

I guess this would be directed at Sandra. In terms of the eight-week period that is recommended, do you think that's adequate to do the data matching that you need? And is there any work being carried out in Powys as a result of the pilot to ensure that you're retaining benefits from that first load of data matching so that the work reduces over subsequent years?

So, I'm hoping, having done the pilots, that, moving forward, we won't be having the whole data set again, just the updates from the last 12 months. So, that's going to make our job quite a lot easier, and for the other pilot areas as well.

The difficulty of the timing will be—. I don't know the number of records that some of these local authority data sets hold. We had, sort of, 60,000 records in some data sets, others had 80, some may only have had 8,000. So, it really will depend on the decision of the EROs as to what data sets they request. We, obviously, started the process quite early, in that we started looking at data sets—some were discarded because they weren't of good enough quality or didn't have the sufficient information we needed. But I think eight weeks should be enough for most of the teams. Even those large local authorities, who will have much bulkier data sets, will have, probably, a larger team to help cope with that and the funding to try and support that. Also, there's some work being done now in the next week or two with the AEA to share some knowledge and good practice from local authorities that currently do a lot of data matching and data cleaning. So, hopefully, sharing those skills to be able to roll that out will be helpful.

Thank you. In terms of going forward, do you think it's possible to establish a common data point or a common data standard? So, you mentioned earlier UPRN, which probably wouldn't work in Powys, but might in other local authorities, so that, again, the amount of work needed to be done to do that matching reduces in future.

Again, in some local authorities, they already are are really good. They've all got UPRN numbers against their council tax database, for example. Powys is working on it and, actually, as a direct result of our experience last year with the pilots—it was only a few weeks ago at our corporate services event—we were discussing trying to look at a corporate standard for data sets. So, that is something, probably, all local authorities will find as part of this process. So, hopefully, it will benefit the wider community within local authorities.

Raising awareness of the UPRN number, raising awareness of, perhaps, having a certain standard when you enter somebody's name and certain fields that should be a must, I think, will develop over time. Again, one of the reasons we wanted to pilot is to see and to test data sets. We often hear from electors, 'Well, I've told council tax, I've told one of your other departments'—I'm sure you know, you often hear it—and yet we don't have this joined-up approach. So, moving forward, there is definitely an opportunity where somebody has contacted the local authority, as they do for registrars with Tell Us Once, that works seamlessly and hopefully alleviates the problem, where somebody has passed away, that, actually, they don't need to contact 10 other departments where that person, an individual, might show up on. So, I think it's longer term, but it's not going to happen overnight. You've probably been aware with local authorities that these things do take time to bed in, and encouragement, even, from you to local authorities to work on this, moving forward, would help.

I think it's bigger also than just elections as well, that input. This is being done and, as Sandra said, it's influenced, maybe, what Powys have done through their lessons learnt. But I think it's more around how you create an infrastructure within local authorities and even with national data sets. Through conversations with the Democracy Club, they've talked about the quality of data, of addressing that across the board, to allow you to have the data sets that take away properly the risks that sit around those things, such as assuming everyone has a UPRN, on what data, or how the context or standardisation is, as you said. But that's a wider, bigger problem across the board.

That's where the roles of returning officers, chief executives, sitting with that dual hat and viewing elections as an important role to be done across a range of areas of the council, are really critical. Obviously, that's why the complementary work of the AEA and the electoral management board, bringing returning officers together with electoral administrators so that that co-ordination happens effectively and in a high-quality way is really, really critical. As an EMB, we're only 15 months in, but I think what you'll see is that best practice rolled out as a result of that effective co-ordination. I think that's some of the findings from the Electoral Commission's report on the Senedd elections, that you're starting to see stronger and more effective co-ordination.

And then, if I can, Chair, lastly from me. What's really struck me today is the barriers to accessing data sets, even within one's own local authority. So, how do you best see that being tackled?

Me again. It's something, actually, we raised at the AVR working group yesterday. I know colleagues have reached out to departments within their councils and have been denied access to that. I think, ahead of the Senedd election and, actually, with the EMB, there was good work done with data protection officers co-ordinating across the whole of the RO areas, in order to share information for, in that regard, the candidate platform. I think a similar approach we've raised with Welsh Government officials to try and co-ordinate. If you get the backing of your data protection officer, to where a head of service might have a concern about releasing that data set, a DPO officer can help you with that. As I said, we've got an excellent one in Powys, who works to enable us to get that, where there, perhaps, are some concerns about why we're using it, et cetera. So, we've suggested that perhaps trying to get a co-ordinated approach, through some of the DPOs. I think EROs need to be aware, and, again, we've suggested to some of my colleagues who have been having problems, 'You need to go to the top.' As Ian mentioned earlier, EROs need to have oversight. Again, it's something I've always been very lucky with, that our ERO and management team, we raise awareness that, 'This request is going out in the next month, six weeks; your teams need to be aware of it, you need to support that.' As I said, this is where, perhaps, local authorities' experience will differ, because they perhaps haven't used data sets in the same way as we have over the years, and so they need to build that relationship, that trust, that this data is required and how it's going to be used.

As a result of that being raised yesterday, we'll be writing out in the next few days to returning officers, chief executives, to say, 'Please can you place that emphasis?' Often, electoral teams sit within the legal area of a council, the monitoring officer area of a council, which is often where data protection sits as well, so unlocking that co-ordination shouldn't be that difficult; it just needs that emphasis and priority from the top of the organisation. But, again, I think that also brings home that this is a live issue, the importance of that working group co-ordinating and making sure it addresses the risks that you're picking out today and mitigates those in time for May 2027.

[Inaudible.]—but I think also I mentioned earlier about looking at the legislation in terms of the regulations, in the Representation of the People (England and Wales) Regulations 2001, which are used as the basis for accessing information, and reflecting what UK Government has done to help remove those barriers, and also looking at the nature of some data sets as they are formed from national data sets—so, housing benefit would use DWP data—and then that's acted as a barrier as well to obtaining a really good data source potentially that could be used for that. So, again, just looking at the framework that sits around the legislation to allow greater access as well.

Thank you, Elyn. Rebeca, I believe you have some questions.

Yes. Thank you all for coming in this morning. So, I think most of us in this room will know that local authorities are already stretched when it comes to preparations for the elections next year. How possible is this actually going to be, that this roll-out is actually going to happen? And also, lessons learned from the pilot, are they going to be shared in terms of the lived experience, which is, obviously, really important as well?

Maybe if I start with this one in terms of the 'possible' question. It's possible. I think we've all identified risks that have to be mitigated if that's going to work effectively, and I think we would stress that they are live risks that are being managed through that co-ordination. But I think we've all flagged up as organisations that the pressure those small electoral registration teams are under is not insignificant, and, actually, the momentum and the resource that needs to be put in to make sure any issues they faced are removed and addressed, whether that's from software providers being able to deliver on time, whether it's in relation to the funding that's available, I think it's that responsiveness that's really, really critical if we're to see this be successful for 2027.

I think we've been clear in our evidence that the accelerated timetable that we're facing, amongst all the different responsibilities that electoral services teams are administering at the moment, from the annual canvas process, with the registers due to be published on 1 December, they've got Senedd accounts to submit by November, they worked through the overseas renewals process for the very first time, they're going through the online absent vote conversions process, which is a consequence of Welsh Government not agreeing to the changes back in 2023, which has then led to an administrative burden and built a confusion in relation to that to try and fix that solution, as well as the numerous community and electoral review implementation of changing boundaries, changing electoral arrangements, it's—. I think—some of you, I know, are previous councillors—local government elections are the most difficult set of elections for any electoral services team to run. They may not be the most high profile, like a UK parliamentary or a Senedd election, but the scale of all the different modes that have to be—Powys being a case example—the number of community councils that some areas have—I know Sandra's definitely winning the league table in Wales in terms of that—.

But to administer those polls and to have this placed on top, while there can be mitigations and everything, our biggest concern is that we've already run out of time to a certain degree. So, while Welsh Government have now delivered what funding is available—and that's been quite generous, I think, in terms of what's been made available—having notification of that on 3 September for a process that you're meant to roll out from November onwards does not give sufficient time necessary to then recruit. We all know that recruitment processes for local authorities cannot happen just overnight. To actually implement those changes takes time. We're still waiting for the electoral management software, which we're so dependent on for delivering this. The final specifications are not due to be ready with those suppliers until next week, so the pressure that's putting on them to deliver in time.

Again, there are ways and means around it, but there's a lot of things. The Electoral Commission have done a great job in producing guidance, so that administrators can actually start to get to grips with what they need to do. But, again, it's very rare for the commission to produce guidance before the legislation is in place, for obvious reasons. The fact that legislation is not going to be in place until probably early November now at the earliest as well, and then to expect administrators to literally go the next day and administer this process is really an unusual situation to be in. It doesn't comply with the Gould principle, where there should be at least a six-month period for people to understand what is required of them.

So, while there are mitigations and while the Welsh Government are doing everything I think they can to help that, I think it's just our concerns from the working group meeting that we went to in January, where we discussed that the funding needs to be in place now, the clarity around this needs to be in place now. There is definitely a risk to delivery when people are still unsure of what that process is, have not got clarity on the legislation, and are required to implement something that could then have knock-on consequences on the delivery of a local government election, which is what we're concerned about.

Is there a cut-off point, as in—?

I think, as we said, in terms of delivering this, to not have a knock-on consequence in terms of local government elections, you need to have anybody who needs to be added to the register guaranteed to be on the notice of alteration that is published in March, to allow people to know how many electors are in each ward and the ballot paper requirements, and not introduce the risk that exists there. So, it's knowing that everything that falls back from that, that you can send notification letters out by this point and that the software is ready to do the data matching that allows you to do that.

I think we're very close even now of knowing that everything is in place to be able to deliver that with confidence. At the moment, there is no ability for anything to slip, and that's worrying. Experience from the pilots is that naturally things slip slightly, but while we're still doing this, to expect someone to implement it before the local government elections next year is definitely a scary prospect, I think, that a lot of administrators are finding themselves facing.

I would concur, in the sense that I feel quite lucky and probably quite smug because we were a pilot area, and therefore I have a greater knowledge and understanding about the process. I would say it is possible. Electoral services teams, if any, will pull it off because we do—we get a snap general election, it gets delivered. However, as I said, it would be remiss of me if I don't relay the concerns of my colleagues who feel generally quite overwhelmed at the moment, and probably, I think it's fair to say, perhaps unprepared for this with everything else, as Clare's mentioned.

In the next few weeks, there'll be a lot of training, a lot of raising awareness, and perhaps they haven't engaged as they should have because they've been so busy with the Senedd elections, by-elections, the annual electoral registration canvas, that it's not been something that has been on their horizon or 'We didn't know if it was going to happen or not.' So, it is possible. As all things, if the right resources are there, we will turn it around, but it's obviously now at risk.

Thank you very much. Diolch.

Okay, Rebeca?

Yes, thank you.

Thank you. Peter, you've got a question.

I suppose I have some anxieties that there'll be some authorities that—. I'm thinking of my old authority, Monmouthshire: a very small election team, and you couldn't pick a worse set of elections to get it prepared for. If it was preparing for a Senedd election or a general election, it would be so much easier, wouldn't it? But you've got so many candidates to sort out, so many people to get on board, there's a lot of pressure on those teams as well. I'm a bit anxious, for a start, about where the additional staff are coming from. What skills do they need? How long will it take for them to be trained? That's probably quite quick, I would imagine, but I don't know. But all of those things that we talked about in the challenges that you had in Powys, can we really gear up to do it? Or is there even an opportunity where you can partially do it? So, you're doing some of it, and you've just got to accept you're only going to get so far on it. So, you bring so many people on, but that's better than none, and then you sort of build on it. I mean, is there a hybrid? Does it have to be that you do it or you don't do it, or can it be organic?

I think, from the experience of the pilots, there are different approaches. I know in Newport, for example, it was another team in the council that completed the data cleansing and the data matching. Something that we've raised at the working groups is that perhaps the ROs need to reach out within their teams, as Ian has mentioned, for wider support from the local authority, so that if those skill sets already exist, as they do probably in the business intelligence and performance teams, perhaps that could be drafted in for that work. I know the Newport experience was probably a lot cleaner in the sense they identified—. I can't remember the exact numbers, but around 1,500 new individuals had been matched with two or more data sets. They discarded all other matches as being inaccurate, and so just moved forward with those, and it seemed a more streamlined process.

I think with regard to capacity and timing as well, it will also depend on how many data sets the decision is to use and, as I mentioned, the actual volume in those data sets. So I think it can be managed, and certainly something the Welsh Government was looking at is what other software could be used and brought in. I know there has been some discussion about AI solutions. I tend to be a little bit wary of those, but again, with the right support and the right knowledge, that is also something else that could be considered.

I think there are workarounds and other possibilities, even maybe for another local authority that does have that capacity and that skill set to do the work on behalf of a neighbour, for example. We already know that there are confederations of local authorities, maybe in the south, for example. So, it isn't impossible. The recruitment, however, and onboarding of staff—. Local authorities can be difficult, for example, in terms of getting them access to certain software, as we've all experienced, trying to get a member of staff started. People need to start hitting the ground running. That's what I would say needs to happen if all of this is to work.

We're talking about 22 local authorities, aren't we, who are all, as you are fully aware, uniquely different, and in essence the ability to manage those risks both in a planned way but also to respond to those risks that the smaller teams bring out. The support that's provided is really critical. It will only be achieved if a whole-council approach is taken in those 22 authorities and particularly in those smaller authorities, which is why that emphasis is needed at the top level of the council.

I agree totally with what Sandra said. There are lots of individuals within local authorities who've already got the skills for doing the data matching, data mining. Electoral administrators like to think we're jacks of all trades, but in terms of actually providing that expertise and capacity to do the work, again, not to hark back, but it's having enough notice to be able to reconfigure your resources to allow you to do that. Again, we're in a period where I don't know whether you can suddenly change workload as quickly as that with a month's notice. I think there are concerns around that aspect as well in terms of that.

David, do you want to come in?

Having come from Merthyr, which is the smallest authority involved, I see they're already on a staff of bare bones. They can't get people in because they can't afford to increase their—. In fact, they’re actually getting rid of people. So to bring people into you, and to use other teams to do the data—it's not possible. I can see in Merthyr it wouldn't be possible, to be honest with you. So, I feel—and thank you; you have painted a picture for me—that 2027 is a bit too far, to be honest with you. I think we're going a bit fast. I think that we put a lot on councils who are already stretched and we're expecting them to take another challenge yet again when they’re already too stretched to do the work that they're there to do now. So, that's my point there.

Thank you, David. Marc, do you want to come in?

I'll put the exact opposite point, I'm afraid. I have also come from a relatively small council, where Ian was the returning officer until earlier this year, and I think what I've seen with returning officers, or electoral registration offices, is they're incredibly resilient. They meet every challenge. I think in ours there were three members of staff, and despite that, they're able to process huge amounts of information and change. The question I have is this: in terms of data sets, are there, for example, two data sets that other councils should be using as a result, rather than scrambling to look at half a dozen?

I think that will vary between local authorities. Council tax is the obvious one. We use housing tenants records, but, for some local authorities, that won't be possible because they're outsourced. So, there are going to be some difficulties. We used adult social care data, which was very good in Powys, because we're a large rural area and take up quite a large portion of Wales. Employee data was very good, because quite a lot of our staff work within the county. That won't be as useful for others in the small urban areas. Sometimes, I think they're going to need to start those discussions and find out what—. I'm not sure if I can recommend, because I don't know the quality of some data sets in other local authorities. But looking at the standard that came across from the pilots, we all used two or three core sets, as I said, like council tax, education records, and it's then looking at what else might be strong within your council, because the EROs have to have faith that the data that they receive is of good enough quality.

One thing that we have concerns with as well is—. We've been speaking to some electoral services teams in England—one in particular—who have been extensive in data matching and mining for a number of years, which they've done work with. In terms of their evaluation of their data sets within their authorities, they use council tax and they use education. We've talked about physically getting hold of information. Sandra has alluded to those being the two key sets.

While you can use your education records and obviously add someone to the register automatically in that sense, without that second data source—. It's actually having the ability to have information that allows you to add those people to the register. I think that's the concern there: the quality of it and the access to it. They can get council tax, but finding that other second data source is going to be, I think, a struggle for a lot of people, based on Sandra's experiences, from their evaluation of their data and from what we know from English local authorities who do data matching.

The other thing that they flagged up as well is the checks. With this one local authority in particular, a lot of their other local data sources are verified against the electoral register, and they've said they can't use data that has been verified. Again, as I said, the electoral register is considered one of the most comprehensive and complete data sources, so it's that challenge of making sure that you're not using data that has been verified against your own records.

I think as well, if I could add, that should be a key part of any subsequent review as well. Obviously, looking back, once all 22 authorities have been through the process, it gives you an opportunity to look at what worked and what didn't. And again, that's where the EMB could come in with sharing good practice after the fact.

Thank you. You'll be glad to know there's just one more question left and it's from me. Thank you very much for your time today. I think it touches on something that Clare mentioned about funding. Peter discussed potentially flexing up with staff in terms of getting this over the line. The WLGA has told the committee that the cost and delivery implications for local authorities are not yet fully understood. So, I guess this is a question to all of you: what are your views on the potential cost implications of implementing AVR and how might this vary across the different authorities?

I'll start. I'm going to answer in a broad way, as you'd expect, probably, with that strategic overview. I think the real importance with funding is it's not one-off funding and that that funding is put into budgets and ring-fenced for this activity on an ongoing basis. I think what we've seen so far is a response from the Welsh Government to support one-off related costs up to the 2027 election, but I think the real risk is that that isn't ongoing. So, I think that's the big strategic point we would want to get across.

I think that most authorities who received their grant funding letters for the initial process for it on 3 September have probably got sufficient funding for this year. But, as Ian said, it's knowing what challenges may arise out of things and then the ongoing funding. As I said, it's not just delivery here, it's what software development needs to continue to develop—we feel like what's being delivered this time is probably a minimum viable product. But it's looking at how you advance the way that you move from manual verification processes of data, how to continue to expand on those to bring capacity to electoral services teams.

I think as well it's the unintended consequences in terms of costs that automatic voter registration brings. As Sandra said, their pilot identified 12,000 electors that weren't on the register, but they didn't have enough data to automatically register them. The system that exists within the UK means that, once you've identified somebody, you need to send an invitation to register within 28 days, you then need to send two follow-ups and potentially door-knock those properties. It's those costs as well. While, obviously, it's the electoral registration officer's duty to look for people and get them on the register, it's the scale at which automatic voter registration builds that picture, and that's going to be an ongoing cost throughout time as well.

I think it's also the knock-on consequences for the annual canvass proces. Because this is being done just on the local government register and does not affect the parliamentary register, it prevents you using the more cost-effective routes that the canvass offers you. Most properties now go down route 1, which doesn't require a response to that sort of communication. When you've got an automatically registered person and you don't know whether or not they can be registered for the parliamentary register if they haven't actively gone and registered themselves for that, it then leads to them having to go down route 2, which requires a response, which requires a chasing cycle. So again, it's the knock-on consequences of funding and different electoral registration processes throughout the year, which also requires funding, as well as the ongoing support and maintenance for delivering automatic voter registration.

If it's okay just to come back, I think I would also pick up, in light of that, that obviously it's a policy decision for the Senedd and the UK Government whether they take different approaches to AVR, and obviously the UK Government are going through that debate at the moment. That's another area where, if there is a divergence, clearly, cost and time increases. So it's just important to be aware of that.

I think that's one of the—not concerns, but one of the thoughts I had in my head. We're discussing funding, we're asking all these local authorities to meet that challenge that we're going to place in front of them, and I'm sure there will be evolution of this system. Has that potential additional expenditure been considered by anybody?

I don't know the answer to that question, but knowing that the UK Government have got direct registration and direct alteration principles and are looking to pilot in the very near future as well, the system will look different to what is being proposed by the Welsh Government currently. And as Ian said, there will then be this sort of friction, potentially, that that system introduces national data sets into the equation, which then changes fundamentally what that means. If you're looking then to pilot the UK system within Wales, then you've got this dual process that butts heads a little bit in terms of that. I take Ian's point completely that each Government has its own legislative capacity to decide what is best, but given that UK Government are also looking at this, there are definitely benefits to looking at how the two systems integrate better together before proceeding with implementation.

I'd also like to add support to Ian's point, because it is a concern that's been raised by my colleagues: that we will recruit new staff to do this process, but if the funding dries up after March or that our local councils don't prioritise it because it's not ring-fenced to us, those staff will leave and so will those skills, that knowledge, and this will be a recurring process where we will lose that skill set. There's a huge resource taken to train people up, to onboard them. So, I would support Ian's point that this is one of the concerns: it will take an awful lot of time to get them recruited, to get them trained up and skilled and knowledgeable about the process, and hopefully support us into the local elections next year, and we'll potentially lose them if we can't get that funding based in the teams. If there is divergence, or whatever decision is made in the future, we want to have those skills in the team to be able to react to whatever decision each Government makes.

Are there any more questions from anyone? No. On behalf of the committee, I'd like to thank you all for coming here today. It's been challenging, giving evidence and taking questions from the floor. We will send out to you the record of today's meeting so that you can ensure that we've captured that discussion accurately. Again, thank you very much.