Savings income tax base too low to match OBR costings

[MaxGhenis]

Summary

PolicyEngine's savings income tax yield from a 2pp rate increase is ~30x lower than OBR estimates:

• PolicyEngine: ~£0.02bn from 2pp increase
• OBR (Nov 2025 EFO): £0.5bn from 2pp increase (£0.6bn static minus £0.1bn behavioural)

Root Cause

The taxable savings income base in PolicyEngine is much smaller than what OBR uses:

Metric	PolicyEngine (2025)	OBR Implied
Total savings_interest_income	£4.3bn	-
Taxable savings_interest_income	£2.9bn	-
savings_income_tax (at 20%)	£0.33bn	-
Effective tax base	£1.66bn	~£25bn

The OBR's £0.5bn yield from 2pp implies a tax base of ~£25bn (0.5/0.02), while PolicyEngine has only £1.66bn effective tax base.

Note: This is different from #218

Issue #218 addresses projections being 2.5x higher than SPI targets. This issue is about the SPI targets themselves being insufficient to match OBR savings tax revenue estimates.

The HMRC statistics documentation notes:

"The coverage of savings income for the sample drawn from NPS prior to 2018 to 2019 was incomplete. This is because most Income Tax payers with savings income do not report it to HMRC."

Potential Solutions

1. Find better savings income calibration source - OBR or Bank of England may have better estimates of total UK savings income subject to tax
2. Calibrate to savings income tax revenue - Use HMRC self-assessment data on actual savings tax paid to back out the implied tax base
3. Use OBR costing as calibration target - Scale savings income so that a 2pp increase yields ~£0.5bn

References

• OBR November 2025 EFO, Table 3.5: "From April 2027, a 2 percentage point increase to the basic, higher and additional rates of saving income tax... This is estimated to yield £0.5 billion a year on average from 2028-29"
• HMRC Income Tax Liabilities Statistics: https://www.gov.uk/government/statistics/income-tax-liabilities-statistics-tax-year-2022-to-2023-to-tax-year-2025-to-2026

Impact

This affects the UK Autumn Budget 2025 dashboard estimates for the "Savings income tax increase (+2pp)" reform.

[MaxGhenis]

Additional Research: ONS Household Interest Income Data

Found authoritative data on total UK household interest received from ONS D.41g series:

Year	Household Interest Received (D.41g)
2022	£12.7bn
2023	£38.9bn
2024	£54.5bn

The dramatic increase reflects rising interest rates.

Comparison with Current Calibration

• Current SPI target: £2.52bn (incomes.csv)
• ONS total household interest: £54.5bn (2024)
• Gap: 22x lower than ONS total

This confirms the issue - SPI only captures savings income reported on tax returns, which excludes:

1. Interest sheltered by Personal Savings Allowance (£1,000 basic/£500 higher rate)
2. ISA interest (tax-free)
3. Interest below personal allowance

HMRC Savings Tax Receipts

Per GB News citing HMRC data, HMRC is projecting £6bn in savings income tax receipts for the current tax year (3x higher than 2022-23's £2bn).

If we assume ~20% average tax rate, this implies a taxable savings base of £30bn - matching the OBR-implied figure in the original issue.

Proposed Fix Approach

Rather than scaling SPI targets directly (which would distort the income distribution), we should:

1. Add a savings income calibration uplift based on ONS D.41g data
2. Apply the uplift proportionally based on existing savings income
3. The uplift factor would be approximately 10-15x to get from £4bn to £45-50bn gross (accounting for PSA/ISA deductions)

This would let PolicyEngine match both:

• The distribution shape from SPI (who has savings income)
• The aggregate level from ONS (total savings income in the economy)

[MaxGhenis]

Investigation Update

Confirmed that savings_interest_income IS being calibrated in loss.py (line 294) as part of INCOME_VARIABLES. The calibration targets come from incomes_projection.csv.

The problem is that the calibration targets themselves are too low:

Year	incomes_projection.csv target	ONS D.41g (household interest)
2022	£2.97bn	£12.7bn
2023	£3.21bn	£40.5bn
2024	£3.21bn	£54.5bn

The targets are based on SPI/HMRC taxable savings, which dramatically undercounts actual savings income because:

1. Most savers fall within the Personal Savings Allowance (£1,000 basic rate, £500 higher rate)
2. SPI only captures reportable/taxable savings

To match OBR costings, we need to either:

1. Update incomes_projection.csv with ONS D.41g-based savings income targets
2. Or scale the existing calibrated values via uprating factors (PR Fix savings income calibration to match ONS D.41g #229, but less accurate)

🤖 Generated with Claude Code