Budget 2019 to 2020: feasibility of distributional analysis - study

5. Conclusions

The purpose of this project was to investigate the feasibility of undertaking a distributional analysis of the Scottish Government Budget. Data from the 2019/20 Scottish Budget was used to test the feasibility.

5.1 What is feasible

Our results suggests that distributional analysis is feasible by:

• Household net income decile; and
• Household type

For person level analyses, we have needed to first make an assumption around how certain taxes and benefits - paid to households - are shared amongst household members. By assuming these are shared equally amongst all adults aged 16 or over in the household, we are also able to produce distributional analysis by

• Age and Sex; and
• Age, Sex and Disability

5.2 What might not be feasible

It is technically possible to produce the following charts, however, due to small sample sizes and differences in service usage by groups, there is a risk that they may produce results that are unreliable or could be misinterpreted:

• Religion
• Ethnicity

There would be potential for misinterpretation of these charts without further breakdown, because the age profiles of these groups is very different and age (and life stage) is a key underlying driver of differences in spend.

For

• Age, Sex and Religion
• Age, Sex and Ethnicity

these analyses would need to be treated with care due to very small sample sizes.

5.3 Key caveats and limitations

The project has demonstrated that even when analysis is technically feasible, there are other issues that must be considered in deploying the analysis. These are outlined below.

Comparability

Comparability of the results with other outputs, such as that of HM Treasury, will be limited as different data sources and methods will have been used in aspects of the model development. For example, we have corrected for under-reporting of benefit income in our model and therefore the analysis is not directly comparable with distributional analyses that have not done this. The choices about which benefits and public services to include in our model will also be different from other models. The choices and assumptions made in this study were driven primarily by the purpose of this analysis rather than by comparability with other models.

Uprating

As the modelling is based on historical data is has been necessary to uprate incomes. Assumptions regarding uprating may be sensible for the population as a whole, but may be unrealistic for sub-groups of the population.

Complex relationships between characteristics

Life stage has been shown to be a critical consideration in analysis and interpretation, with the pattern of spending varying across the life course. For example, young people receive more higher education spend; in middle age we see higher tax paid due to higher incomes; and in old age there is higher receipt of benefits from pensions and higher health and care expenditure. This means that when considering distributional analysis by ethnicity, for example, results would ideally need to be age standardised to be able to compare ethnic differences within similar life stages. However samples are so small that the data at this level becomes unstable and unreliable. Gender will also have a strong intersecting impact but it is difficult to fully understand this because of the lack of understanding of how income from earnings and benefits are distributed between different genders within the household.^[41]

Missing groups

The survey upon which the modelling is based is a survey of private households. As a result, many student households and care home residents will be missing from the survey. As a consequence, either the survey population must be modified to include these people, through adjusting the weighting, or a decision needs to be made to exclude the public services which relate to these people. As expenditure on students and care homes is relatively sizeable, we have opted to adjust the underlying survey population rather than omit these public services.

Assumptions made

We have presented our assumptions and decisions about the taxes, benefits and services included in the different annexes. For any distributional analysis the assumptions used will determine the outputs that follow. This includes technical ways to address issues about income or which people to include in the analysis, as well as the overall approach and decisions about inclusion/exclusion. For example in health we decided to take an insurance based approach rather than an individual usage approach – and used population level usage information. There may be value in undertaking analysis using different assumptions to see what different approaches and resulting outputs tell us.

There will always be some debate about whether the assumptions used are the best ones for the analysis. However, whilst agreement on assumptions may be difficult if not impossible to achieve, it is important that if changes over time are to be understood, then the same assumptions must be used over time wherever possible. Similarly if comparison with other distributional analyses is required – for example 'what-if' analyses for specific policy areas - then this will require the same assumptions to be used.

Further development of the model

If this work is to be repeated for the 2021/22 Scottish Budget or future budgets, the model will need to be further developed to take into account the structural changes in the Scottish economy resulting from the Covid-19 pandemic (for example increases in benefits uptake due to increased unemployment). Modelling of mitigation activity in response to Covid-19 undertaken by the UK and Scottish Governments would also be needed. This is complicated by uncertainty regarding how long these measures will need to be in place and for which sectors. This modelling would need to be informed by Scottish Fiscal Commission forecast assumptions released for the 2021/22 budget documents.

5.4 Distributional findings

In testing the feasibility of our approach, data from the 2019/20 Scottish Budget was used. Therefore the project was also able to produce findings in relation to the distributional impact at the time of that budget. The key findings are:

• The progressive nature of tax by income decile is shown in the analysis.
• Spend and benefits is also progressive in that it is highest for income deciles 2, 3 and 4 and then reduces. However, those in the lowest income decile do not appear to receive as much in benefits and public spend as deciles 2, 3 and 4. This may relate to the composition of the group, which is complex containing people aged 55-64 in the pre-retirement phase who may have low income but a reasonable standard of living from savings or other wealth. The group also includes students, unemployed people, women with children and people who are temporarily sick.
• Life stage plays a critical role in understanding distributional analysis – and this is the case in looking at the Scottish offer. Our household analysis shows that households containing older people tend to receive more in benefits (from the state pension) and see a higher spend on health and social care. Households containing middle aged people tend to be paying the most tax (due to those age-groups tending to earn the most) and households with children and young people see the highest spend on schooling, Early Learning and Childcare and higher and further education.
• Analysis by ethnicity and religion shows that households with a head of household who identifies as non-British or a head of household who has a religion other than Christian, receive lower benefits and lower health and social care spend but higher further and higher education spend. This is primarily driven by the fact that the age profile is very different for these groups. The age structure is younger and we have seen that benefits and health and social care spend increases with age. There are also other differences such as the fact that a higher proportion of non-White British young people attend higher or further education than white British young people.
• Analysis at the individual level shows the impact of the gender pay gap and unequal childcare responsibilities in the differential amount of tax paid by men and women – being higher for men due to higher earnings. Conversely, women tend to receive more benefits, but some of this is due to child benefit – all of which is allocated to women in our model. Men aged over 65 had higher health costs while women aged over 65 tended to have higher social care costs, due to having longer life expectancy.
• The same structural issues can be identified amongst disabled men and women, with disabled men paying more tax than disabled women while benefit receipt looks relatively even between disabled men and women. Health care costs look similar across disabled and non-disabled groups, but that may be linked to the methodology used which allocated health care expenditure by age, sex and deprivation only. Social care spend is much higher for disabled men and women when compared with non-disabled men and women.

5.5 Potential role and value of analyses that are feasible

The project has demonstrated that it is possible to produce reliable outputs of distributional analysis for households by income level and for some if not all equality characteristics. It is also possible, however, for the outputs to be misinterpreted. Therefore a key finding of the project is that distributional analysis should be accompanied by robust explanation. This interpretation should explain apparent anomalies or unexpected results and provide details of what might be the underlying reasons for these. In some instances additional data and information will be helpful in explaining findings.

The analysis presented here provides a baseline understanding of budget impacts. It shows how big blocks of spend are distributed across households depending on income, age, gender, ethnicity and religion. It also shows the progressive nature of tax and spend. It shows which groups might see positive or negative impacts if changes were made to particular budget lines. However, it cannot look at the impact of any such changes in detail, since even relatively large changes in spend on individual budget lines would be very difficult to see when looking at all expenditure. There are benefits to the analysis here which aids understanding of how tax, benefits and spend add up for individuals, but this also needs to be supplemented by detailed intersectional analysis of particular items of spend across groups.

The project has shown that cumulative distributional analysis is feasible and can be a valuable tool for understanding the impacts of the Scottish Budget for different households and individuals. It is most useful in providing a baseline – which could be updated every few years (perhaps at the start and end of a Parliamentary term) – and which can help identify areas where more detailed analysis at individual policy level might be useful.