If you want to make sure the right Land Transaction tax is paid, on the right property, by the right person, it goes without saying that the location of the property needs to be accurate. The Welsh Revenue Authority wanted to understand why – in some cases –address data wasn’t accurate for some returns. There are two other drivers to understand why location data isn’t always right; the first is the creation of a project team in April 2022 to explore a Land and Property data platform delivering high quality geographic datasets across Wales. Any platform would need good quality location data for it to work well. The second, the announcement of a new local rate of Land Transaction Tax on 4th July 2022 has intensified people’s need for better quality geographic property data.
In June 2022 we did research with a handful of solicitors and conveyancers to understand what happens when they enter property address into the Land Transaction Tax return.
Why a postcode lookup can be a problem
On the face of it - it’s hard to see how address could go wrong. We watched users enter in postcodes for properties and choose addresses from a list. A selection of addresses are returned for the post code the person enters. We hypothesized about whether there was some problem with the address lookup or if people were not using it. In reality, all our participants used the address lookup - most without any issues.
However, for one of our participants, and maybe this behaviour offers a clue to what is happening some of the time, this is what we saw and heard:
Participant enters postcode.
A list of addresses is returned.
Participant says “The address I am looking for isn’t on this list”
Looks again at the list of addresses.
“Oh yes it is”
I ask what they have selected.
“So I have picked out - “former post office”’
Participant says “This isn’t quite the same as on my list next to me and I know from checking the office copies that this property is now known as “The Old Post Office” not “Former Post Office” - I don’t know if [this address here] is one and the same but I am informing the WRA that this is the property that I want to pay stamp duty on” [P4]
Participant manually edits OS places address from “Former post office” to “The Post Office”
Small edits like this, align or “correct” the data to their own data set which in this case is the title deed or transfer document. The agent needs the data to match the title deed and transfer document. The underlying dataset being used doesn’t align to what they need it to say. So the agent needs to edit it. However once you have manual entering of address data there is a discrepancy between the address data set and the users’ data.
How errors creep into address through manual processes
A second thing we learned was around the process of completion of the return. The manual aspects of this had a significant impact on the accuracy of the completion.
This participant is explaining to me how they get the details needed to input into the online return form:
“This is a scanned copy so I can’t copy and paste it - so that’s why I write it down”
The participant has a physical piece of paper in front of them.
They manually write down the details they need to complete the return on the paper, often from PDFs or from their own bespoke internal systems.
Then they use what they have on their piece of paper - rekeying it all into the return form.
Another participant comments: “a lease might be 50 pages long - the info you need might be in the definitions but it might not (so marking it out on a notepad) saves having to flick through the whole lease” P2
Clearly the cognitive effort of finding the right information in lengthy documents, manually transcribing it to a piece of paper and then rekeying the information into the online return is a process that can easily go wrong.
Although no one is going to volunteer that things go wrong, or the frequency with which this happens, participants did comment that one of things they really appreciated about Land Transaction Tax was the ease with which they could make and submit amendment requests(!).
Why only one address is captured when three are asked for
Often only one address is entered; the property address - even though there are two other address captures - the seller’s address and the buyer’s address.
It was explained to me like this [paraphrased]:
The buyer is going to move to the property address - so we put the property address as their address. The seller - we don’t know where they are moving to - so we put their address down as the property address too.
So there is one address for the buyer and the seller. This undermines the accuracy of the data and begs the question - is there a value in asking for three address captures?
We also explored scenarios where there is no address.
Lack of existence of an address was rarely a problem for residential
We discovered that for residential property having no address was not an issue. This was only an issue for commercial property, particularly if it was a land only transaction. We heard how a plan sometimes needed to be emailed to WRA from the agent (there is no upload facility). And we discovered that there is no agreed standard for explaining the location other than reflecting the words written in the title deed.
There is a mismatch between when the LTT return is completed and the agent’s own process
And finally we discovered that because the return is completed once the transaction has happened - there is little engagement from the agent or their client with the details required to be submitted. The agent’s incentive is to ensure that all the details on the return match what they have in their systems. Once the transaction has happened the agent’s client (the property owner) is disengaged (and we heard) often not responsive.
Additional information (not asked as part of a transaction) like the hectarage caused delays in submitting the form because it is not something the agents need to know and so they are not invested in getting that information once the transaction is completed.
We are using the research findings to think about the future
Our research findings about challenges with Land Transaction Tax returns are helping us to focus on what things might also be true for an emerging local Land Transaction Tax rate.
One agent said to us:
Provided we collect the correct data we can generally submit our returns [To HMRC] with a one-click option from our system whereas with LTT we have to log in and enter a reasonable amount of information into [the LTT system] and its double entering because we have already entered it into our portal
We now know that inaccuracies in location data seem to be primarily happening because of manual rekeying into the Land Transaction Tax return. We are going to focus our energies on looking at what aspects of the Land Transaction Tax process could be changed for the better to improve the delivery of Land Transaction tax, regional Land Transaction Tax and also improve the accuracy of location data for everyone.
Each week we share a short update on what we’re learning, along with links to things we think are interesting or relevant.
- The technical architecture
- Policy challenges
- Every day's a school day – lessons learned from the land and property platform
- Why are inaccuracies creeping into addresses on the Land Transaction Tax return?
- How a map can paint a bigger picture
- Summarising the Proof of Concept
- What we learned about users
- How we used weeknotes
- What we've learned about working in the open
- What do we mean by a data platform?
We built a number of prototypes to test our assumptions, platform capability and ideas.