`total_income` leaves out `other_investment_income`, which `income_tax` charges

[juaristi22]

Summary

total_income (variables/gov/hmrc/income_tax/total_income.py) cites ITA 2007 s. 23. Step 1 of s. 23 sums every amount of income on which the person is charged to income tax. The variable adds eight inputs:

• employment_income;
• private_pension_income;
• social_security_income;
• self_employment_income;
• property_income;
• savings_interest_income;
• dividend_income;
• miscellaneous_income.

It does not add other_investment_income, which the input's documentation describes as National Savings interest, securities interest and interest from trusts or settlements. Yet the model charges it:

• taxable_savings_interest_income = max(0, savings_interest_income + other_investment_income − tax_free_savings_income);
• so savings_income_tax, income_tax and adjusted_net_income all include it.

For anyone with other investment income, total_income is lower than the income the model taxes. Someone whose income is mostly other investment income can show a total_income in the lowest taxpaying band while owing tax on a much larger income.

Where it matters

Five baseline formulas in 2.122.2 read total_income, and each understates the income of anyone with other investment income. Main (2.123.1) still leaves it out.

• post_tax_income adds total_income and subtracts income_tax, so it subtracts tax on income it never added.
• winter_fuel_allowance uses it as the taxable-income test.
• household_benefits_individual_non_dep_deduction and council_tax_reduction_individual_non_dep_deduction band non-dependants' weekly income on it.
• housing_benefit_net_earnings uses it, through its adds, for the earnings share of income.

Two contributed basic-income reforms, bi_individual_phaseout and bi_household_phaseout, read it too.

Datasets that band people by total income inherit it too. Microcosm UK binds 47 HMRC targets on total_income: the SPI regional cells (Table 3.11), the national SPI band tables and the ITL rows by total income band. On PolicyEngine/microcosm#1121's build:

• The West Midlands £12,570–15,000 income-tax cell started at 6.1 times its target. Every other region's cell in that band started at 0.6–1.1 times.
• The cause is fewer than 10 records whose income is mostly other investment income. They fell into that band and carried all their tax into it.
• Banded on total_income + other_investment_income, the cell starts at 0.92 times its target, and the other 119 regional income-tax cells move by 8% at most.
• UK: port the remaining uk-data fixes tracked in #1095, engine items included microcosm#1121 defers that cell until this is fixed.

Proposed fix

• Add other_investment_income to total_income.adds.
• Add a test: a person whose only income is other investment income has total_income equal to it, and post_tax_income equal to it net of the tax on it.
• housing_benefit_net_earnings iterates total_income.adds to undo losses, so it picks up the new component without a change. Its result should be checked against its existing tests.

Related: statutory pay

housing_benefit_net_earnings already notes the same kind of gap for statutory pay. Its comment says statutory sick and maternity pay are taxed as employment income (employment_benefits, through taxable_employment_income) but left out of total_income.

• Adding employment_benefits to total_income.adds would close that gap as well.
• But housing_benefit_net_earnings currently adds statutory pay separately on top of total_income, so it would need adjusting in the same change.
• Whether to do both here or separately is the maintainers' call.

An alternative is to define total_income as the sum of the taxable components that adjusted_net_income already sums. That would also change how losses and the property and trading allowances enter total_income, because those components floor at zero after the allowances. That's a larger change for every reader above.