Admin-based income statistics, England and Wales: tax year ending 2016

These admin-based income statistics (ABIS) are Experimental Statistics¹. The small area income estimates are the official estimates of household income for small areas. The following statement must be included in all releases and accompanying metadata and these outputs must not be reproduced without it.

The admin-based income statistics (ABIS) bring together data from the Pay As You Earn (PAYE) and benefit systems to derive estimates of net and gross income. The ABIS are defined as experimental, because both the income measure and coverage are currently incomplete; therefore, these statistics have limited use for decision-making. Instead, the ABIS demonstrate the potential to produce small area income statistics from administrative data and allow some interim evaluation to be made (taking their partial coverage into account). They also allow us to start comparing administrative data-based estimates with other survey-based household income statistics.

Office for National Statistics (ONS) will seek National Statistics status once more components of income have been included and after further consultation with users.

As Experimental Statistics, we welcome feedback on how we might improve these statistics to meet user needs. Please see Section 8 for the ways you can do this.

Notes for: Disclaimer

1. For further information on Experimental Statistics, please see the ONS’s Guide to Experimental Statistics.

These admin-based income statistics (ABIS) are the continuation of the Income Research Outputs published last year to produce administrative data-based, small-area, multivariate income statistics. The overall aims of this research are detailed in Annex C: Background and previous research.

In response to user feedback (see Annex A: Feedback from 2017 publication), these ABIS now include:

• percentiles of individual and household income for Lower layer Super Output Areas (LSOAs) in England and Wales for tax year ending 2016

• estimates of both gross and net income

• imputation of additional income components

• a change in the equivalisation method (from one-person to two-person reference households)

• household income distributions where people were used as the analysis unit rather than households

• multivariate analysis showing household income percentiles by household composition

• a comparison to survey-based household income data

Due to the nature of the data sources used to construct the ABIS, please remember when interpreting the findings that:

• the term income refers to nominal¹ income from Pay As You Earn (PAYE) and benefits only – it does not include income processed through the Self Assessment system

• these estimates must not be compared with previous research as a time series due to changes to the methodology

• only individuals and households present on the population base (Statistical Population Dataset (SPD) V2.0 for 2015²) were included in these statistics – the quality of the population base will impact the quality of the outputs presented here, but we are continuing work to make improvements to the base we use

• when referring to household income, the ABIS are actually referring to income at an address – you can find more information on the differences between a traditional “household” and an “occupied address” in the Occupied Address Research Output

Notes for: Things you need to know about this release

1. Nominal income has not been adjusted for inflation.

2. Statistical Population Datasets (SPD) estimate the size of the population by linking records across multiple administrative data sources and applying a set of inclusion and distribution rules.

This section outlines the definitions constructed and used in this article. Annex D details the Canberra Handbook definitions used in survey estimates, which serve as a target for these statistics to try and replicate. However, some important administrative data sources are not yet included in this research (for example, Self Assessment data), so the admin-based income statistics (ABIS) do not match the Canberra Handbook definition. Therefore, it’s important to be clear on what is and is not included in gross and net income for this research.

Gross income

Components of income covered by the administrative data and included in these admin-based income statistics (ABIS) are¹:

• gross earnings (net of employee and employer pension contributions) from employment, including any benefits-in-kind paid through Pay As You Earn (PAYE)

• state support, which consists of most state benefits, including Tax Credits and Child Benefit

• income from occupational and personal pensions processed through the PAYE system

Following feedback from the Department for Work and Pensions (DWP), Winter Fuel Payment and Christmas Bonus payments have been imputed this year to increase the coverage of income components included in the measure.

Components of gross income included in the international definition of gross income (as defined in the Canberra Group Handbook), but not included in the ABIS income measure, are:

• income from self-employment or income from an employer not paid through PAYE

• investment income including income from property and interest from Individual Savings Accounts (ISAs) and other saving accounts, bonds, stocks and shares

• some state support – some benefits including Universal Credit and Personal Independence Payment²

• current transfers received such as parental contributions, child maintenance payments and educational grants

We hope to include some of these income components in the future as the availability of administrative data increases.

Student loans are not included in the ABIS gross income measure, as they are not included in the Canberra Group Handbook definition of gross income. If you have a requirement for student loans to be included in the income measure, please let us know by emailing Admin.Data.Census.Project@ons.gov.uk.

Net income

The net income measure is calculated by deducting Income Tax and Class 1 National Insurance contributions from the gross income measure described previously (which is already net of pension contributions). Other payments, such as Council Tax, child maintenance or parental contributions, were not deducted.

Income Tax and Class 1 National Insurance contributions were estimated using the standard tax and National Insurance rates, and the age-dependent personal allowance. Other allowances, such as Marriage Allowance and Blind Person’s Allowance, were not taken into consideration.

In the future, access to more detailed administrative data will allow for a more accurate calculation of Income Tax and National Insurance contributions. Access to additional data sources will also allow other taxes, such as Council Tax, to be deducted. These developments will help move the ABIS towards a measure of disposable income.

Notes for: Current income definitions

1. Further information on the income components included and excluded from the income measure are available in these publications: Income Research Outputs 2016 and Data source overview: income and benefits.

2. Universal Credit and Personal Independence Payments were introduced during the tax year ending 2014 and were initially rolled out in a limited number of areas. During the tax year ending 2016, the number of claimants on Universal Credit rose from approximately 50,000 to just over 200,000. Despite this increase, the staggered roll out means the impact on the ABIS for tax year ending 2016 is expected to be small. However, it may affect some geographic areas more than others and will have more of an effect on the ABIS in future reference years, if these components are still not included.

Please ensure that you have read Section 1 and Section 2 to understand the context of these research findings. Information on how these individual income distributions were constructed can be found in Section 7.

Where an individual or household had income information included in the admin-based income statistics (ABIS), their whole income may not have been captured. The ABIS often underestimate individual and household income due to the components of income that are missing from the income measure¹.

Throughout Section 4 and Section 5, we’ve also explored the percentage of individuals or households with no income information included in the ABIS. This does not reflect that these individuals or households had no income, rather their income was not captured by the administrative data included in this research. This could include, for example, individuals who received all their income from self-employment or investments.

There also appears to be a close relationship between the number of individuals or households with no income information and the value of the 10th income percentile. For example, age, region and household composition groups with a higher percentage of individuals or households with no income information, often also have lower 10th income percentiles. This could be an indicator that these lower percentiles are disproportionately affected by the missing income components and should be interpreted with extra care.

For a discussion of the findings for gross Pay As You Earn (PAYE) and benefits income, see the 2016 and 2017 Income Research Output publications.

Net individual PAYE and benefits income by age and sex

The ABIS include some income information for 88% of people aged 16 years and over on the population base (Statistical Population Dataset V2.0 for 2015).

Figure 1 shows how both coverage of the population and median net PAYE and benefits income varied by age and sex. For both sexes, coverage of the population base with some income information increased between ages 16 and 24 years, from below 20% at age 16 years to over 80% by age 24 years. Median net PAYE and benefits income also increased between these ages for both sexes from below £2,000 at age 16 years to over £12,000 by age 24 years.

Lower coverage and income for the younger ages was expected due to the high proportion of 16- and 17-year-olds in full-time education. The Labour Force Survey showed that between May and July 2015, 88% of individuals aged 16 to 17 years in the UK were in full-time education. The continued increase in both coverage and income between ages 17 and 24 years was also expected, as many individuals leave education and enter the labour market during this period. For example, the Labour Force Survey showed that, for the same period, the percentage of individuals aged 18 to 24 years in full-time education was lower, at 33%. The Labour Force Survey also showed that only 25% of 16- and 17-year-olds were in employment between May and July 2015, compared with 61% of 18- to 24-year-olds.

Between ages 25 and 60 years, coverage of women with income information continued to be higher than men. There are two main reasons for this. Firstly, a greater proportion of individuals in receipt of Child Benefit were women (89% of claimants in this output). All women claiming Child Benefit will have had some income information included in the ABIS. Secondly, the Labour Force Survey has shown that between June and August 2015, more than twice as many men were self-employed than women. If self-employment was an individual’s only source of income in the tax year, they will not have had any income information included in the ABIS.

Despite higher coverage for women aged between 25 and 60 years, men had a consistently higher median net income, which also peaked at an older age. Median net PAYE and benefits income for men peaked at approximately £19,300 at age 42 years, whilst the median income for women peaked at approximately £15,700 at age 36 years.

For women, coverage increased at State Pension age (age 62 years²) due to the inclusion of State Pension income in the ABIS. For men, coverage did not increase until age 63 years. The State Pension age for men was higher than for women in tax year ending 2016 (age 65 years for men compared with age 63 years for women), but coverage for men increased sharply at age 63 years, rather than age 65 years. This is due to the inclusion of Winter Fuel Payment (which was imputed in the ABIS for eligible individuals aged 63 years and over). Higher coverage remained throughout the older ages for both sexes, staying above 99% after age 63 years for women and after age 66 years for men.

At State Pension age for both sexes, there was also an increase in median PAYE and benefits income. This is likely to reflect the increased coverage of income components for these ages, and the potential inclusion of both State Pension income and employment income for individuals still in employment.

After State Pension age, median net PAYE and benefits income for men declined. For women, it declined up to age 72 years before starting to rise again. This might be because women have a longer life expectancy than men and they were inheriting their spouse’s pension.

Figure 1: Percentage of individuals with some Pay As You Earn (PAYE) and benefits income information, alongside median net PAYE and benefits income

By age and sex, men and women aged 16 years and over, England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

Figure 2 shows the percentage difference in net income for men and women expressed as a percentage of the income for women. At all ages, the 90th percentile was higher for men than for women and the 20 to 24 age group was the only age group where the median income for women was higher than the median income for men. There was more variation in the difference between the sexes at the 10th percentile. For the younger age groups (16 to 34), and age group 60 to 64, women had a higher 10th percentile than men. However, as discussed at the start of Section 4, the 10th percentile may be disproportionately affected by the income components not included in the ABIS income measure and should be interpreted with extra care.

There was a decrease in the difference between men and women at age group 60 to 64, which was most extreme at the 10th percentile. This decrease in the difference reflects the earlier State Pension age for women than for men (age 62 years for women and age 65 years for men6). Women within this age group would have started to receive the State Pension, whereas men would not. Apart from the 60 to 64 age group, the difference between the sexes generally increased throughout the working ages (between ages 20 to 24 and 50 to 54 years) and decreased with age from age 70 years and over.

Figure 2: Percentage difference between women and men in net individual PAYE and benefits income at the 10th percentile, median and the 90th percentile

By five-year age group, England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

Geographical analysis of net individual PAYE and benefits income

Over 70% of individuals aged 16 years and over in each local authority of England and Wales had some PAYE and benefits income information included in the ABIS.

Figure 3 shows that coverage varied more at Lower layer Super Output Area (LSOA) level. At LSOA level, the percentage of the aged 16 years and over population with no PAYE and benefits income information included in the ABIS varied from 3% to 85%.

Some of the main reasons a high percentage of individuals in an LSOA would have no income information included in the ABIS are:

• high levels of self-employment in the area (self-employment income is not currently included in the ABIS income measure)

• a high proportion of students in the area (student loans and maintenance grants are not currently included in the ABIS income measure)

• high concentrations of armed forces personnel in the area (due to the methodology used to construct Statistical Population Dataset V2.0, it was not possible to include income information in the ABIS for much of the armed forces)

• a higher proportion of benefit claimants on Universal Credit (the staggered roll out of Universal Credit will mean some geographic areas are affected more than others by the absence of Universal Credit from the ABIS income measure)

Figure 3: Percentage of individuals aged 16 years and over with no Pay As You Earn (PAYE) and benefits income information included in the admin-based income statistics

Lower layer Super Output Areas, England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

2. Data displayed in the map relate to the Lower layer Super Output Area and not the postcode.

Figure 4 shows the median PAYE and benefits income for individuals aged 16 years and over in England and Wales for the tax year ending 2016, by region. The 10th and 90th percentiles of PAYE and benefits income are also shown.

Median net individual PAYE and benefits income for England and Wales was £13,867. London had the highest median of all the regions at £15,262. It also had the highest 90th percentile (£34,683) and the lowest 10th percentile (£3,069). As discussed at the start of Section 4, the 10th percentile may be disproportionately affected by the income components not included in the ABIS income measure. The absence of self-employment income in the ABIS and the high levels of self-employment in London may be the reason London has the lowest 10th percentile of all the regions.

Figure 4: Net individual Pay As You Earn (PAYE) and benefits income

By region, England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

Figure 5 allows you to explore median net individual PAYE and benefits income for England and Wales at LSOA level. Results are largely as expected, with many LSOAs with higher median PAYE and benefit incomes located in London and the South East.

Figure 5: Median net individual Pay As You Earn (PAYE) and benefits income

Men and women aged 16 years and over, Lower layer Super Output Areas, England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

2. Data displayed in the map relate to the Lower layer Super Output Area and not the postcode.

Notes for: What do the admin-based income statistics show for individuals?

1. More information on components not included in the income measure can be found in Section 3.

2. State Pension ages correct for the tax year ending 2016.

Please ensure that you have read Section 1 and Section 2 as well as the start of Section 4 to understand the context of these research findings. Information on how these household income distributions were constructed can be found in Section 7.

Geographical analysis of net equivalised household PAYE and benefits income

Some Pay As You Earn (PAYE) and benefits income information was included in the admin-based income statistics (ABIS) for over 85% of households in each local authority. As for individuals, household coverage varied more at Lower layer Super Output Area (LSOA) level than at local authority level. Figure 6 allows you to explore the percentage of households in each LSOA with no PAYE and benefits income information.

Figure 6: Percentage of households with no Pay As You Earn (PAYE) and benefits income information

Lower layer Super Output Areas, England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

2. Data displayed in the map relate to the Lower layer Super Output Area and not the postcode.

Households with income information included in the ABIS, may not have had income information included for all adults in the household. Although 98% of households in England and Wales had some PAYE and benefits income information included, only 83% of households had income information included for all adults. In 15% of households, income information was only included for some of the adults in the household.

In London, 74% of households had some PAYE and benefits income information included for all adults in the household. In 21% of households, income information was only included for some of the adults. Income information was not included for any adults in 5% of London households. As discussed previously, these lower coverage levels are likely to reflect the high levels of self-employment in London and the absence of self-employment income in the ABIS’s income measure.

Figure 7 shows the equivalised net household PAYE and benefits income distribution for England and Wales combined and by region. At regional level, median net equivalised household income ranged from £22,595 in Yorkshire and The Humber to £25,736 in the South East.

Equivalised net household income was highest in London at the 90th percentile and lowest in London at the 10th percentile, suggesting London had the greatest spread of incomes. However, as discussed in Section 4, the 10th percentile may be disproportionately affected by components of income not included in the ABIS. This may affect London more than other regions due to the lower coverage of households in London with some PAYE and benefits income information.

Figure 7: Equivalised net household Pay As You Earn (PAYE) and benefits income

By region, England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

Figure 8 shows the median net equivalised household PAYE and benefits income at LSOA level. As for individual income, results are largely as expected, with many LSOAs with higher median PAYE and benefits incomes located in London and the South East.

Figure 8: Median net equivalised household Pay As You Earn (PAYE) and benefits income

Lower layer Super Output Areas, England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

2. Data displayed in the map relate to the Lower layer Super Output Area and not the postcode.

Net equivalised Pay As You Earn and benefits household income by household composition

Earlier in 2018, we published an update on developing household statistics for an Administrative Data Census. A household composition measure was developed, which has enabled the production of a breakdown of equivalised household PAYE and benefits income by household composition. In future releases, we hope to be able to break this analysis down further, by other household characteristics or to smaller geographic areas.

Table 1 shows the percentage of households with no PAYE and benefits income information by household composition. For household types where all individuals were aged 65 years and over, a low percentage of households had no income information because of the inclusion of State Pension data and Winter Fuel Payment in the income measure. The household type where the greatest percentage had no income information included were households with only full-time students.

Figure 9 shows net equivalised household PAYE and benefits income by household composition in England and Wales for the tax year ending 2016.

Couple households with no children had the highest median net equivalised household income at £30,369. Households with all full-time students had the lowest median net equivalised household income. This does not accurately reflect the living standards of students as the ABIS do not currently include student loans, educational grants or parental contributions in the income measure. The student income and expenditure survey for England and Wales showed that, in the academic year ending 2015, almost two-thirds of full-time students’ income came from student loans and grants.

Lone-parent households with dependent children had the second-lowest median net equivalised household income at £17,821.

Figure 9: Net equivalised household Pay As You Earn (PAYE) and benefits income

By household composition, England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

2. For more information on the household composition breakdown, see the update on developing household statistics for an Administrative Data Census.

Figure 10 compares the distribution of net equivalised household PAYE and benefits income for households with dependent children against all other household types. No PAYE and benefits income information was available for 1% of households with dependent children and 3% of all other households. Higher coverage was expected for households with dependent children due to the inclusion of Child Benefit income in the ABIS’s income measure.

At all percentiles, the income for households with dependent children was lower than the income for all other households. The median net equivalised household income for households with dependent children was £20,924, compared with £26,607 for all other households.

This may in part be due to the different levels of coverage. Most households with dependent children will have had some income information included due to Child Benefit. Therefore, a self-employed household with children will have been included in the analysis with only their Child Benefit income. In contrast, a self-employed household without children and no other income sources will have been excluded from the analysis.

Figure 10: Net equivalised household Pay As You Earn (PAYE) and benefits income for households with dependent children and all other households

England and Wales, tax year ending 2016

Notes:

1. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

Our aim is for the admin-based income statistics (ABIS) definition of income to harmonise as closely as possible with the definitions outlined in the Canberra Group Handbook. A Comparison of Canberra Handbook Income components included in ABIS, HBAI and ETB table is available, which outlines the components of income included and excluded from the ABIS’s income measure.

The table also presents a comparison of the components of income currently included within ABIS, against the components of income included in two measures of household income statistics published across the Government Statistical Service (GSS). These are the Department for Work and Pensions’ (DWP) Households below average income publication (HBAI) and ONS’s Effects of taxes and benefits on UK household income publication (ETB) – both of which hold National Statistics quality accreditation. The table shows that coverage of income components within HBAI and ETB is comprehensive – income components not currently captured by the surveys are small.

Within this section, comparisons are limited to household income statistics that are produced for larger geographies. Comparisons are not made with the small area income estimates.

ABIS comparison with effects of taxes and benefits publication

The ONS publication Effects of taxes and benefits on UK household income (ETB) is based predominantly on Living Costs and Food Survey (LCF) data. LCF is a voluntary survey of approximately 5,000 households in the UK¹.

For the purposes of comparison with the ABIS, the ETB figures have been adjusted to remove income from self-employment and investment, to align more closely with components of income currently included in the ABIS estimates². A three-year average is presented of ETB data from tax years ending 2015, 2016 and 2017, with annual incomes in 2015 to 2016 prices. A comparison is made of gross income figures only. The ETB analyses used here are not comparable with the published ETB National Statistics and cannot be used for independent ETB analysis. The ETB figures presented here are specific to this partial income concept, developed to correspond more closely with the ABIS.

Figure 11 presents the percentage differences, or relative differences, at regional level between ABIS’s and ETB’s 10th percentile and median gross equivalised household income figures (using household weighting³).

Figure 11 shows the ABIS’s 10th percentile was higher than the ETB’s in Wales, the South East, London and the West Midlands. ABIS include administrative data for most state benefits, including Tax Credits. These benefits are likely to have the most impact on lower-income households, which are present in the 10th percentile. We acknowledge that there is under-reporting present in the top and bottom of the ETB income distribution, as well as non-response error. This under-reporting for lower-income households in the survey data that ETB relies upon may explain why the ABIS’s 10th percentile is higher than the ETB’s in these regions.

We might have expected to observe this trend across all the regions. However, in the other regions, the ETB’s 10th percentile was higher than the ABIS’s. In Section 4, we discussed that the ABIS’s lower percentiles should be interpreted with extra care as there appears to be a close relationship between the percentage of households with no income information and the bottom 10th percentile. Although self-employment and investment income have been removed from the ETB measure, other components of income missing from the ABIS may still be affecting the bottom percentiles. Uncertainty around the bottom percentiles in the ABIS may therefore explain these unexpected findings.

Figure 11 also shows that median gross equivalised household income in the ETB estimates is higher than in the ABIS for all regions except the North West (where the medians were very similar) and the North East. The general trend of higher ETB medians was expected due to the components of income missing from the ABIS income measure, even after the removal of self-employment and investment income from ETB. Under-reporting in the survey underpinning ETB will also have less of an effect as you move up the income distribution, as welfare benefits form a smaller component of household income.

Figure 11: Percentage differences between ABIS’s and ETB’s 10th percentile and median gross equivalised household income figures

By region, England and Wales, tax year ending 2016

Notes:

1. ABIS refers to admin-based income statistics and ETB refers to effects of taxes and benefits estimates.

2. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

3. ETB figures are based on a three-year average of tax years ending 2015, 2016 and 2017.

4. The ETB analyses used here are not comparable with the published ETB National Statistics and cannot be used for independent ETB analysis.

5. The ABIS and ETB figures compared both use households as the analysis unit. Both figures are rounded to the nearest £100.

ABIS comparison with households below average income publication

The Department for Work and Pensions’ (DWP) households below average income publication (HBAI) is based predominantly on Family Resources Survey (FRS) data, although adjustments are made for top incomes using administrative data from HM Revenue and Customs. The FRS is a representative sample of over 19,000 UK households.

As with ETB, the HBAI figures have been adjusted to remove income from self-employment and income from investment, to align more closely with components of income currently included in the ABIS estimates⁴. A three-year average is presented of HBAI data from tax years ending 2015, 2016 and 2017, with annual incomes in 2015 to 2016 prices. A comparison is made of gross income figures only. The HBAI analyses used here are not comparable with the published HBAI National Statistics and cannot be used for independent HBAI analysis. The HBAI figures are specific to this partial income concept, developed to correspond more closely with the ABIS.

Figure 12 shows that except for Wales, unlike ETB, ABIS are not higher than HBAI at the 10th percentile. This might be due to more complex benefit editing undertaken by DWP on FRS data. DWP data may still be affected by an under-count of benefit income relative to the administrative data used by ABIS, but this under-counting is having less of an effect than the ABIS partial coverage of income components.

As more components of income are incorporated into the ABIS income measure, we will continue to investigate why these differences occur and why differences may vary geographically.

Figure 12: Percentage differences between ABIS’s and HBAI’s 10th percentile and median gross equivalised household income figures

By region, England and Wales, tax year ending 2016

Notes:

1. ABIS refers to admin-based income statistics and HBAI refers to households below average income.

2. Data source: Pay As You Earn (PAYE) employment and pension data and Tax Credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

3. HBAI figures are based on a three-year average of tax years ending 2015, 2016 and 2017.

4. The HBAI analyses used here are not comparable with the published HBAI National Statistics and cannot be used for independent HBAI analysis.

5. The ABIS and HBAI figures compared both use the individual as the unit of analysis. Both figures are rounded to the nearest £100.

Figure 13 compares gross equivalised household income in the ABIS, ETB and HBAI across the percentiles. There’s a general increase in the differences between ETB and HBAI with the ABIS towards the top end of the distribution. There are likely to be several factors contributing to this.

Firstly, there’s a difference in the populations that are being compared between ETB, HBAI and the ABIS. The LCF and FRS survey data that ETB and HBAI estimates are drawn from, respectively, are based on private households, whereas ABIS estimates are based on occupied addresses⁵.

Certain components of income also disproportionately affect the upper end of the distribution, such as benefits-in-kind and salary bonus payments. These are included in the ABIS estimates through Pay As You Earn (PAYE), but their inclusion may not align with the survey data used for ETB and HBAI. Within PAYE data, these payments are recorded in the month in which they are paid to the employee. In the collection of the survey data, respondents are asked about any benefits-in-kind or bonus payments received within a certain reference period. These time periods may not align and could contribute to the differences in the estimates.

Definitions of employment income within the survey data may also have an effect. Despite rigorous survey design, the possibility of respondent error remains. Respondents may incorrectly think of themselves as employees when they more accurately fit the definition of self-employed, and so the source of their income is erroneously attributed to employment.

A further difference is the weighting and imputation methodologies used in the LCF and FRS data processing. These methodologies are not used when calculating the ABIS figures.

ABIS, ETB and HBAI inequality rankings

ETB and HBAI sources are frequently used to report on inequality of income⁶. To allow comparison with the ABIS experimental inequality estimates of gross income, Table 2 presents the ratio of the 90th percentile to the 10th percentile. This is the ratio of the upper band of the 90th percentile with that of the 10th percentile, as a measure of inequality. This is presented by region and uses gross household income figures, with household weighting for ETB and person weighting for ABIS and HBAI⁷. It’s important to be clear that these are not regional inequality statistics, but are comparisons in the inequality rankings between ABIS, ETB and HBAI when examining a particular income concept corresponding to the current ABIS definition. Table 3 presents the rankings of these ratios for each income publication by region (where 10 is the greatest inequality and 1 is the least).

A comparison of the 90th percentile and 10th percentile income ratio calculated for all three income measures shows agreement throughout the bottom end of the rankings of the regions and also at the very top of the rankings. There are just a few regions that switch positions towards the upper end. However, the rankings remain similar. This suggests that despite some of the variation in the percentile figures between the three income measures, the trends of regional income inequality remain very similar. Where the surveys’ rankings agree and the ABIS disagree, the ratios tend to be very close. However, it’s likely that the components of income missing from ABIS and that we have not been able to be adjust for in the analysis, are contributing to the difference.

Notes for: Admin-based income statistics (ABIS) comparison with survey-based household income data

1. From the reference period tax year ending 2018 onwards, the core sample for household income that ETB estimates are based on is increasing to 19,000 households.

2. For an overview of what is currently included within the ABIS estimates, please refer to Section 3: Current income definitions.

3. Households were used as the analysis unit for the purposes of comparison with ETB. This is because ETB uses households rather than people as the unit of analysis. See Section 12, bullet point 4 of the ETB release for more details. A discussion of households versus individuals as the unit of analysis is included in Annex B.

4. For an overview of what is currently included within the ABIS estimates, please refer to Section 3: Current income definitions.

5. You can find more information on the differences between a traditional “household” and an “occupied address” in the Occupied Address Research Output.

6. For more information, see Unequal results: Improving and reconciling the UK's household income statistics.

7. ETB continues to use households rather than people as the unit of analysis. See Section 12, bullet point 4 of the ETB release for more details. A discussion of household analysis units is included in Annex B.

Data sources and linkage

Five administrative datasets on Pay As You Earn (PAYE) income and benefits (including tax credits) were linked to derive an annual PAYE and benefits income total for each individual. Linkage was performed using a unique identifier created by the Department for Work and Pensions (DWP) specifically for our purposes. These datasets were:

• HM Revenue and Customs (HMRC) tax credit data

• HMRC’s PAYE data

• DWP’s National Benefits Database (NBD)

• DWP’s Single Housing Benefit Extract (SHBE)

• HMRC’s Child Benefit data (new in the 2017 publication)

This linked income dataset was then linked to the Statistical Population Dataset (SPD) V2.0 for 2015¹. This is the population base for the statistics. Income information was only included for individuals present on SPD V2.0. Income information was linked to 88% of individuals on the SPD aged 16 years and over.

Calculating income

Demographic information on SPD V2.0 and benefit information on the NBD were used to calculate Winter Fuel Payment and Christmas Bonus amounts for eligible individuals. Income from the five administrative datasets, Winter Fuel Payment and Christmas Bonus amounts were totalled to derive gross annual estimated income amounts.

Income Tax and National Insurance amounts were estimated using standard thresholds and subtracted from gross annual income to estimate net annual income.

Individual income percentiles were produced for all individuals aged 16 years and over present on SPD V2.0. Individuals who appeared within the five PAYE and benefit administrative datasets with a zero-income amount were included when calculating the percentiles. Individuals who did not appear in the PAYE and benefits administrative data were excluded.

To calculate household income, the concept of an “occupied address” rather than a traditional “household” has been used. More information on the differences between a traditional “household” and an “occupied address” is available in the Occupied Address Research Outputs publication.

Income information for individuals at the same unique property reference number² was totalled to calculate an un-equivalised household income figure. Income of all household members (including people aged under 16 years) was included in the household income measure.

Household income was equivalised and percentiles of household income were produced using individuals as the unit of analysis. The following sections contain more information on why we equivalise household income, the equivalisation scale used and the choice of analysis unit. As for individual income, households with individuals who appeared within the PAYE and benefits administrative data but with a zero-income amount were included when calculating the percentiles. Households that did not contain any individuals who appeared within the PAYE and benefits administrative data, and therefore had no income information, were excluded.

Income information for much of the armed forces population was excluded from the individual and household income measures. This is because the armed forces populations were added to SPD V2.0 at aggregate level. Therefore, for most armed forces personnel, it was not possible to link their individual income information or allocate them to a unique property reference number.

For more detail on the methods used to calculate individual and household income, see the 2016 and 2017 Income Research Output publications.

Equivalisation of household income

Income information in the admin-based income statistics (ABIS) is collected at an individual level. However, when considering economic well-being, it’s common to consider measures of household income. This is because incomes are often shared between household members. Therefore, the ABIS aggregate individual income data to household level. This household total is then equivalised to account for different resource needs of households of different sizes and compositions.

When using measures of household income, we assume that the income is shared evenly between all household members. In reality, this assumption might not be true for all households.

We’ve equivalised our measures of gross and net income using the Organisation for Economic Co-operation and Development-modified (OECD) equivalence scale. Following feedback on the 2017 Income Research Output publication, and to ensure coherence with other income statistics, we’ve changed our methodology to equivalise with an adult couple household with no children as the reference group.

During equivalisation, each member of the household was allocated a standard weighting, as outlined in Figure 14. The weights of all household members were then added to calculate a total household weight. The weight for a household with an adult couple and no children was 1 (the reference).

The total household income was then divided by the household weight to give an equivalised household income.

Figure 14: How we have equivalised household income

We equivalise household income to account for the different resource needs of households of different sizes and compositions

Source: Office for National Statistics

Download this image Figure 14: How we have equivalised household income

.svg (18.3 kB)

Analysis unit for the distribution of household incomes

After equivalisation, individuals were ranked in order of their equivalised household income. The individual halfway through the distribution was selected and their household income was used as the median equivalised household income figure. Percentiles of household income were calculated in a similar way.

This approach has changed following feedback on the 2017 Income Research Output publication, where households were used as the analysis unit for the distribution of household incomes rather than individuals.

For more information on the difference between these two approaches, and a summary of our analysis comparing the two, please see Annex B.

Imputation of Winter Fuel Payment and Christmas Bonus

Winter Fuel Payment and Christmas Bonus were imputed into this year’s income measure. This increases coverage of income components and is a response to feedback received from the Department for Work and Pensions (DWP).

Demographic information on the population base and benefit information held within the National Benefits Database (NBD) were used to identify the eligible population and allocate them a benefit amount.

Christmas Bonus is a one-off £10 payment paid to eligible people claiming certain benefits during a qualifying week. Christmas Bonus was allocated to individuals in the ABIS when they were in receipt of eligible benefits during December 2015. Christmas Bonus payments could only be allocated to individuals in receipt of eligible benefit types included in the NBD³. It was not possible to allocate Christmas Bonus payments to individuals in receipt of some eligible benefits, for example, Personal Independence Payment. Using this method, ABIS allocated Christmas Bonus payments to approximately 13 million individuals in England and Wales compared with the 16 million individuals reported by DWPacross the UK.

Winter Fuel Payment (WFP) is an annual payment of between £100 and £300 paid to eligible individuals during the winter to help pay for their heating bills. For the tax year ending 2016, individuals born before 5 January 1953 qualified for WFP if they met the residency criteria. To impute WFP in the ABIS, age is taken at the reference date of the SPD (30 June 2015). Therefore, it was not possible to exactly match the eligibility criteria of the benefit. Individuals aged 63 years and over on the reference date were allocated a WFP. The amount was calculated based on their age and whether anyone in the household was claiming Pension Credit, income-based Jobseeker’s Allowance, income-related Employment and Support Allowance or Income Support.

Approximately 10.7 million individuals aged 63 years and over were allocated a WFP amount in the ABIS. This compares with the approximate 11.1 million individuals reported by DWP. Fewer individuals may have been allocated WFP in the ABIS due to the different residency criteria for the Statistical Population Dataset (SPD) V2.0 and for WFP, and the use of age 63 years in the ABIS when some individuals aged 62 years will have been eligible.

Notes for: Methodology

1. A Statistical Population Dataset (SPD) is a single, coherent dataset that forms the basis for estimating the population. It’s produced by linking records across multiple administrative data sources and applying a set of inclusion and distribution rules.

2. A unique property reference number (UPRN) is a unique alphanumeric identifier for every spatial address in Great Britain and can be found in Ordnance Survey’s address products.

3. Eligible benefit types included in the National Benefits Database include: Attendance Allowance, Carer’s Allowance, contribution-based Employment and Support Allowance, Disability Living Allowance, Incapacity Benefit, Pension Credit, State Pension, Severe Disablement Allowance, and widow’s benefits.

We’re keen to get your feedback on these admin-based income statistics (ABIS) and the income measure presented in this publication. Feedback may also include suggestions on how to improve the outputs, further uses of the data and any requirements for output area-level data.

Please send feedback to Admin.Based.Characteristics@ons.gov.uk and include “ABIS 2018” in your email subject.

We received feedback on the 2017 Income Research Output publication by email and at a series of engagement events held with representatives of various local authorities. This section provides a summary of the feedback we received and how we’re responding to this feedback in these admin-based income statistics (ABIS).

Previously, some users have presented a need for income data at output area level. This use of data was presented to the [National Statistician’s Data Ethics Advisory Committee][1]. They recommended further consultation with users to more fully understand the need for output area-level data. Once sufficient evidence of user need has been collected, the case can be resubmitted to the committee. If you do have a need for income outputs at output area level, please contact us to tell us more about your needs.

The distribution of household incomes is generally analysed after equivalisation. The two analytical approaches are as follows.

Household analysis (used in ONS’s effects of taxes and benefits (ETB) publication)

The household is used as the basic analysis unit. Every household, no matter its size, contributes the same to the distribution and has a weight of 1. Therefore, an individual in a four-person household would have one-quarter the representation in the distribution of an individual who is in a single-person household.

Calculating the median: To calculate median equivalised household income, the households are ranked and the income of the household halfway through the distribution is selected.

Individual analysis (used in DWP’s households below average income (HBAI) publication)

The individual is used as the basic analysis unit. An individual within a four-person household would have the same representation in the income distribution as an individual in a single-person household.

Calculating the median: To calculate equivalised household income, the individuals are ranked rather than households. The individual halfway through the distribution is selected and their household income is the median income figure.

Figure 15 illustrates these two approaches, and how the choice of analysis unit might impact median income. In this example, household analysis has resulted in a higher median income figure than when individual analysis is used, £15,000 and £10,000 respectively. This is because, in this scenario, the household with the lowest income is much larger than the other households. If the reverse occurred and households with higher incomes were larger than households with lower incomes, the household analysis would result in a lower median income than the individual-level analysis.

Figure 15: How we calculate the median equivalised household income

Source: Figure 15: How we calculate the median equivalised household income

Notes:

1. In this illustrative example, the household income is lower using individual analysis. However, it's important to recognise that the opposite can also occur and incomes can be higher when using individual analysis.

Download this image Figure 15: How we calculate the median equivalised household income

.svg (14.9 kB)

The choice of analysis can also impact measures of poverty. In Figure 15, if the absolute poverty threshold were set at £12,000, we could either conclude that the poverty rate is 33% (one in three households if using household-level analysis) or 64% (7 out of 11 individuals if using individual-level analysis).

When using household-level analysis, the poverty rate could go up or down if there was a change in the number of households. For example, if household 1 were to split into two, or if households 2 and 3 combined, then the poverty rate would change. This change would happen even if there had not been an actual change in living standards.

Analysis has been run on ABIS data using both levels of analysis. For England and Wales combined, individual-level analysis and household-level analysis result in very similar gross equivalised median income figures (£26,419 and £26,375, respectively).

At local authority level, there was greater variation between the two analysis levels (Figure 16). The difference in median gross household PAYE and benefits income ranged from £1,785 in Richmond upon Thames (where individual-level analysis produced a higher median than household-level analysis), to £3,338 in Tower Hamlets where the reverse occurred. Areas such as these, with larger differences, will have had a stronger relationship between household size and household income. For example, in Richmond upon Thames, households with higher incomes will have been larger on average than households with lower incomes. Whereas in Tower Hamlets, households with lower incomes will have been larger on average than households with higher incomes.

For 89% of local authorities, the difference between the two approaches was less than plus or minus £1,000.

Figure 16: Difference in median gross equivalised household Pay As You Earn (PAYE) and benefits income between household-level analysis and individual-level analysis

Local authorities, England and Wales, tax year ending 2016

Notes:

1. Data Source: Pay As You Earn (PAYE) employment and pension data and tax credits data from HM Revenue and Customs, and benefits data from the Department for Work and Pensions.

This year’s ABIS use individual-level analysis, ensuring that individuals have the same representation in the income distribution regardless of their household size¹. As discussed previously, this will be important if poverty thresholds based on these income statistics are developed in the future. Adopting individual-level analysis has also made ABIS more consistent with the advice given in the Canberra Handbook²:

“Equivalised disposable household income can be household weighted³, but since it can be viewed as a measure of the economic resources available to each individual in a household, income measures for equivalised estimates are generally based on numbers of people rather than numbers of households. This is referred to as individual weighting and ensures that people in large households are given as much weight in the distribution as people in small households.”

Notes for: Annex B: The difference between household- and individual-level analysis

1. In Section 6 of the ABIS, household-level analysis was used to enable comparisons with ONS’s effects of taxes and benefits publication.

2. For more information, see the UN Canberra Group Handbook 2011.

3. The terms “household weighting” and “individual weighting” used here refer to the household-level analysis and individual-level analysis described throughout Annex B.

The 2021 Census topic consultation identified user needs for small area multivariate income statistics. This user need cannot be met through the traditional census due to the negative impact on response rates and poor quality of resulting data. Therefore, in these admin-based income statistics we're exploring the potential to fill this gap in official statistics using administrative data. Two sets of Income Research Outputs have previously been published, in 2016 and 2017, demonstrating the progress of this work.

This administrative data-based approach could be used to produce enhanced income outputs for the 2021 Census, or income outputs with an administrative data population base (which constitutes the analysis presented in this publication).

Our current focus is on demonstrating the feasibility of producing small area income statistics and multivariate income outputs. While there are still parts of income missing from this year’s income measure (self-employment and investment income), this publication demonstrates how we’re developing our methodologies and income definitions to meet the user need. We plan to include self-employment and investment income in our measure in 2019, subject to data access.

The aim of this research is to meet user needs identified through the census topic consultation as well as the needs of ONS transformation to produce outputs on an “administrative data-first” approach. That is, to make efficient use of already collected administrative data and use surveys to fill in the gaps. We’re working with experts across the Government Statistical Service (GSS), particularly within the Department for Work and Pensions (DWP) and HM Revenue and Customs (HMRC), to develop a coherent approach to producing harmonised income measures.

The aim of this research and future research will be to ascertain the extent to which administrative data can meet definitions of gross and disposable income as outlined in the Canberra Group Handbook.

Components of gross income included in the target definition are:

• income from employment (which includes employee income and income from self-employment)

• property income (which includes investment income from both financial and non-financial assets)

• current transfers received (which includes social security schemes and pensions)

Components deducted from gross income to calculate disposable income in the target definition are:

• direct taxes (for example, Income Tax and Council Tax)

• social insurance contributions (for example, National Insurance)

• current transfers paid (including transfers between households such as child maintenance and to non-profit institutions such as regular charity donations)

• compulsory fees and fines (for example, county court judgement)

Student loans are not included in the current target income definition. However, feedback has indicated that some users would like to see student loans included to accurately reflect the living standards of students. If receipt of student loans is included in the income measure, student loan repayments would also need to be deducted from the measure.