Skip to main content

How Welsh Revenue Authority (WRA) statistics meet the Code of Practice for Statistics and the dimensions of value, trustworthiness and quality.

First published:
27 June 2019
Last updated:
View update history


Introduction of LTT

We introduced these statistics to meet the immediate user requirement for data on Land Transaction Tax (LTT), following the WRA being established. See further information on Land Transaction Tax

Statistics published in other UK countries

HMRC and Revenue Scotland publish statistics for the equivalent taxes in England and Scotland, respectively.

Users may wish to be aware that in the Monthly Property Transactions publication, the number of Welsh transactions in the latest month may differ slightly to LTT statistics. This is because in the UK publication, initial estimates for the month have a grossing factor applied in order to estimate the eventual number of transactions for that month.

Comparing LTT statistics with property sales data used in the UK House Price Index (HPI)

The Office for National Statistics (ONS) publish monthly statistics on house prices in the UK. The ONS use a variety of data sources to produce the UK House Price Index (HPI). This includes data from HM Land Registry on land registrations for England and Wales. Alongside HPI statistics, the ONS publish the monthly number of residential property sales.

We compared the numbers of property sales from this source with our LTT statistics. While the trends are broadly similar in the two sources, the number of residential LTT transactions is generally higher than in the sales data from the HPI. We believe the main reason for the difference is that the ONS excludes particular transactions when producing the HPI statistics. We believe the two main factors are

  1. Commercial transactions of residential properties (namely, where the buyer or seller is a corporate body, company or business). We can broadly account for these in LTT data, and these transactions can be seen to account for some of the difference.
  2. Sales that were not for full market value. Although we can’t easily account for these in the LTT data, the average property value in LTT transactions is generally below that within the HPI statistics (despite the counts of transactions being higher). This suggests that including what are lower value transactions in the LTT data is having an impact on the comparison.

The above transactions are included in LTT statistics, but not within HPI statistics.

The ONS also exclude certain other transactions from the HPI, which we believe have less impact on the comparison. Further information on these exclusions is available in ‘About the UK House Price Index’ reports.

Dates used in LTT statistics

Data presented in our statistical releases is based on the effective date of the transaction. The effective date is when the tax becomes liable to be paid, usually when a transaction is completed on a property. Whilst using the effective date in analysis can lead to greater volatility in the data (for example, due to a change in taxation rates), and revisions in successive releases and data reports, this date relates to the point at which the transaction took place and not a notional future date when the tax return was received. This also means that the series created from our analysis will reflect changes in tax rates and policy at the time that any changes take place.

We are aware that some other publications in the UK base their analysis on the date that the tax return is submitted. We have therefore produced some comparable figures to other UK countries (using date submitted) in our annual LTT statistics.

Revisions to and timing of LTT statistics

Land Transaction Tax statistics are regularly revised. The effect of revisions is analysed in monthly and quarterly LTT statistics, with commentary provided in quarterly LTT statistics.

We explain the timing of LTT statistics in our policy on statistical outputs.


We have produced these statistics in line with the Code of Practice for Statistics. 

As a new producer of official statistics, we are developing our statistical publication processes and have previously published our policy on statistical outputs. This includes:

  • the professional standards which were adhered to as part of the creation of these statistics
  • how the content and timing of outputs is independently managed by the WRA Lead Official for Statistics
  • how we notify users of upcoming outputs
  • how data is collected, stored and managed, and
  • that staff involved in producing statistics will undertake continuous professional development in line with the Civil Service competency framework and the Government Statistical Service (GSS) competency framework


We have assessed the interest and potential quality concerns in line with the Government Statistical Service guidance on the use of administrative data. The table below presents our current assessment.

Date source Public interest profile Level of risk of quality concerns Level of assurance information to be developed
LTT Registration data Low Low A1 – basic assurance
LTT Tax returns Medium Low A2 – enhanced assurance

While preparing each statistical release, we have continued to work closely with WRA operational staff to identify further issues during data collection and processing of individual returns. We continue to apply quality controls which provide immediate analysis of tax returns which flag up potential areas of concern. Where a tax return is amended by the WRA operational team, the organisation filing the return is contacted to confirm the amendment.

We are also continuously working with internal colleagues to identify options to mitigate these issues at the point of collection for future returns, which will indirectly lead to improved data quality. 

The WRA operations team are also considering risks to tax collected. Addressing risks could indirectly improve data quality in some areas.

It is also worth noting that most LTT data are supplied by organisations working on behalf of the taxpayers, some of whom submit transactions on a regular basis (around 10 per cent of organisations supply around four-fifths of the tax returns). This has led to data quality improvements as the WRA has developed its relationships with many of these key organisations.

Around 3,300 organisations are registered for online LTT submissions, with a total of around 9,300 registered online users. Most organisations have submitted at least one return. Examples of quality issues we have previously discovered and our response to date are:

Example 1

As part of the quality assurance for this release we studied non-residential transactions due to the volatility of the series, particularly when considering smaller subsets of the data than are presented in the release. We reviewed the reasons for the volatility which generally relate to small numbers of large value transactions which can occur at any time but may not be present in each month. Because non-residential transactions are usually relatively low in value, the impact of these large value transactions has a large effect on the volatility. 

We have concluded that there are no immediate concerns with the data we are currently publishing at the Wales level as a result of this exercise. However, this issue combined with some missing or invalid postcode information has led to some concerns when presenting further disaggregation of the data (see the Quality of geographic data section) as part of our annual release and our StatsWales tables.

Example 2

By assessing the tax due against the data that has been supplied for each transaction, we have been able to identify some errors in data. This has led to some corrections to the transactions, in conjunction with the agents that have submitted them.

One example is where we have checked whether the correct option for the type of transaction was selected when filing. 

A transaction can be either residential or non-residential (which includes cases where a property is not wholly residential). In addition to this, a higher rate can also be applied to residential properties depending on a few factors. See our technical guidance for further information.

By carrying out analysis on the tax due, and supplementary data, we have been able to identify cases where the incorrect type of transaction was selected. We have then subsequently corrected the classification for the purpose of the statistics presented here. The data is presented by transaction type in Table 2 in monthly and quarterly statistics.

Why have we done this?

If within three years of completing a higher rate LTT transaction, the buyer sells their previous main residence, they may be eligible for a refund of the additional higher rate of LTT. It is therefore important to be able to estimate this figure as accurately as possible. 

Example 3

We have made several changes to our data tables in this release, based on user feedback and to correct for quality issues with data previously presented.

On 22 February 2019, we made a small downwards revision to ‘additional revenue from higher rates’ tax due in Table 1 in our monthly LTT statistics. This revision was made in our data-only release for January 2019 LTT statistics, published on the StatsWales website. Previously, we had assumed that the additional revenue from higher rates was 3 per cent of the consideration (for all higher rate transactions). However, this is not always true, for example if there are other reliefs applied to the transaction which reduce the tax due. Our small downward revision to the statistics takes this into account for these cases.

Non-residential transactions where a new lease is granted have either or both:

  • a non-rental value (or premium); and 
  • a rental value (which relates to the length and terms of the lease). 

These two elements contribute differently to the tax due on a lease transaction. Previously, Table 4 (non-residential transactions by value) in our monthly and quarterly statistics classified transactions according only to the non-rental value.

We received user feedback that Table 4 should include the rental value of non-residential transactions (where applicable) in classifying the transactions. Therefore, we have developed our methods to allow us to add this breakdown. This has involved splitting the tax due on non-residential transactions into the two elements pro rata to the rental and non-rental values and presenting these in separate columns. This introduces some double counting in the columns shown in the table, and care should be taken in adding up the columns (as explained in the table footnotes). We have also added the rental value of non-residential properties into Table 1.

Example 4

In our analysis of reliefs, we have found a number of transactions for which the reliefs seemingly had no effect on the tax due. Over the course of the last year, we have been looking at why this is the case and sought to reduce the cases where the relief was claimed unnecessarily. In some cases, we know the relief was incorrectly claimed and our operations team have worked with the agents supplying the data to correct the underlying transactions, with commensurate improvement to our data.

In other cases, however, we know the relief was claimed but other aspects of the transaction were incorrect, which led to the assumption that the relief had no effect. Because the underlying transactions are awaiting correction, we have made some adjustments to reflect the corrections which we believe should be made as part of the process of deriving our statistics.

Both these sets of adjustments have resulted in a decrease in the number of transactions for which the reliefs seemingly had no effect on the tax due, and in many cases has added to the amount of tax shown as relieved in section 5 of the annual statistical release.

We are continuing to look at this category of reliefs and it is likely that further adjustments will be made over the coming year.

Example 5

In June 2019, we discovered nine transactions linked to each other where the consideration entered for each was incorrect. Each of the transactions reported the same overall consideration that applied across all the nine transactions, rather than the individual considerations applied to each transaction which should have added up to that total. As such, we counted the overall consideration for these linked transactions nine times (an eight-fold over count). Once we discovered this error, we contacted the agent who amended each transaction to include the individual considerations.

Example 6

We have discovered a number of tax returns where the consideration is 100 times higher than what it should have been. This is possibly due to agents using third-party software to complete and submit the tax returns. In these cases we have identified, we have contacted the agent to request that they check this and submit an amended return if required. This issue has affected some of our statistics on property value and also our reliefs statistics (where a relief has been claimed on these transactions).

In addition to these examples, we have also considered the restricted reporting of certain larger transactions, which may be at risk of disclosure. With any transaction of sufficient size, there is a risk of revealing something about them when publishing aggregate statistics.

Our approach is to balance that risk against the need to publish a meaningful figure for the total tax raised under LTT.

To do this, we have created a bucket category for each year, in which only the total tax due of all these transactions is reported, rounded with less precision (to the nearest million pounds). This bucket will reveal nothing other than the year in which the transaction occurred.

The number of transactions going into the bucket, the total value of the land/property involved in those transactions, and any sense of the distribution of the transactions (e.g. between residential and non-residential) is not presented.

An alternative approach of simply leaving out these bucket transactions was considered. However, this would provide an incomplete picture of LTT revenue. Furthermore, any payments of tax associated with these transactions will legally need to appear in the WRA’s annual accounts. Any difference between those accounts and our published statistics (if we excluded the tax due) would therefore reveal the bucket value with limited accuracy anyway. The approach we have taken to include the bucket is therefore considered to be a proportionate and balanced one.

In our published analysis of transactions by narrow value band, we have stress tested the use of nearest £100k rounding on some of the forecasting models and found there to be little impact when this was reduced to £10k.

Quality of geographic data

Our annual statistics provide a breakdown by local authority and Welsh Index of Multiple Deprivation (WIMD) area. This data is also published on the StatsWales website alongside data for National Assembly constituencies.

The local authority field is mandatory for completion on all tax returns, while the postcode is an optional field on the tax return. We can only use the postcode on the tax return to establish the National Assembly constituency or Welsh Index of Multiple Deprivation areas in which a transaction took place. The postcode is either missing or invalid on nearly 5% of tax returns.

In our analysis of local authority data, we derive the local authority from the postcode by looking this up on the ONS’s National Statistics Postcode Database (NSPD). We then compare this information to the local authority selected on the tax return.

  • In 95% of cases, the local authority derived from the postcode matches the local authority provided on the tax return.
  • Where this is not the case, we look at whether the two local authorities are neighbouring (for example, Cardiff and Newport are neighbouring):
    • if yes (neighbouring), we use the local authority provided on the tax return for our statistics. This is the case for around 0.2% of tax returns
    • if no (not neighbouring), we use the local authority derived from the postcode for our statistics. This is the case for less than 0.1% of tax returns. In this circumstance, we believe the postcode is more likely to be correct and that the local authority provided on the tax return is more likely to be incorrectly chosen.
  • In the other 5% of cases where the postcode is missing or invalid, we take the local authority as supplied on the tax return.

Most transactions have one land item associated with them, although some transactions have more than one piece of land associated. We must use details of each land item to determine the local authority. To avoid duplicating transactions with more than one piece of land, a fraction (equal to one divided by the number of land items) has been applied to the transaction and tax due for each. This ensures the data are apportioned between the relevant local authorities whilst maintaining consistency with the overall count of transactions and value of tax due at Wales level.

In our annual statistical release, local authority data is presented for residential and non-residential transactions and tax due, whilst local authority data on the value of properties taxed (known as the consideration) is presented for residential transactions only. This is because there are some non-residential transactions with a particularly large consideration and a possible risk of identifying a taxpayer if we were to publish annual local authority data on these. In future, we will investigate the viability of combining several years of non-residential transactions to support safe publication of consideration data.

In addition, we have published analysis for National Assembly constituencies and Welsh Index of Multiple Deprivation (WIMD) for residential transactions only. As the postcode on the tax return is used to derive these geographies, we have discovered that where the postcode is not supplied, there is a clear bias towards larger non-residential transactions. As these transactions cannot be allocated to a National Assembly constituency or WIMD area, the resulting statistics are not reliable. Therefore, it is not currently appropriate to produce statistics on non-residential transactions for National Assembly constituency or WIMD areas.

Properties sold more than once in a year

All our LTT statistics count transactions during the time period stated, in which the same property or piece of land may have been sold more than once. Our best estimate is that between 3% and 4% of transactions involved a piece of land which has been sold more than once in April 2021 to March 2022.

When do purchasers pay higher rates?

A number of factors can mean a residential transaction is subject to higher rates. These include:

  • purchasing buy-to-let properties
  • buying a second home or holiday home
  • buying a new property while trying to sell an existing one
  • companies like social housing providers buying properties

The LTT statistics only include properties sold in the past year. They don’t represent the full stock of properties in any local authority.

As a useful source of information on people with second addresses, the Office for National Statistics have published a statistical bulletin and dataset on this topic using data from the 2011 Census.