Impute student loan balance from WAS to FRS

[MaxGhenis]

Summary

Add student loan (SLC) balance imputation from the Wealth and Assets Survey (WAS) to the FRS, following the same pattern as existing wealth imputations in wealth.py.

WAS Data Available

Student loan balance can be derived from WAS Round 7 (April 2018 - March 2020):

• Tot_LosR7_aggr (total loans) minus Tot_los_exc_SLCR7_aggr (total loans excluding SLC) = SLC debt

WAS Statistics

Metric	Value
Unweighted households with SLC debt	701
Weighted households with SLC debt	1.66m
Mean balance (those with debt)	£20,028
Median balance	£14,000
Max	£210,000
Total weighted	£33.4bn

Comparison to Admin Statistics (~2019-20)

Metric	WAS	Admin (SLC)
Borrowers/HH with debt	1.66m HH	~4.5m individuals
Total outstanding	£33.4bn	~£140bn UK
Coverage	24%	100%

WAS undercounts due to young graduate underrepresentation and household vs individual unit. However, it captures the distribution shape which is valuable for imputation.

Implementation

1. Add to wealth.py RENAMES dict:

"Tot_LosR7_aggr": "total_loans",
"Tot_los_exc_SLCR7_aggr": "total_loans_exc_slc",

2. Derive SLC debt in generate_was_table():

was["student_loan_balance"] = was["total_loans"] - was["total_loans_exc_slc"]

3. Add "student_loan_balance" to IMPUTE_VARIABLES

4. Consider scaling imputed values to match SLC admin totals (currently £267bn for England, March 2025)

Why Needed

• FRS student_loans (tuborr) is amount borrowed THIS YEAR by current students, not outstanding balance
• 98.7% of FRS repayers have student_loans = 0
• Without balance data, we can't cap repayments at outstanding balance (high earners get unrealistic modelled repayments)
• Needed for interest accrual calculations

Related Issues

• Add student loan repayment calibration targets #237 - Add student loan repayment calibration targets
• policyengine-uk will need a student_loan_balance variable to use the imputed data

Sources

• SLC Statistics
• Student Loans in England 2024-25

[nwoodruff-co]

@PolicyEngine could you try this

[policyengine]

I've implemented student loan balance imputation from WAS to FRS by making three changes to wealth.py:

1. Added total_loans and total_loans_exc_slc to the RENAMES dictionary to pull these columns from WAS Round 7
2. Derived student_loan_balance in generate_was_table() as the difference between total loans and loans excluding SLC
3. Added student_loan_balance to IMPUTE_VARIABLES so it gets imputed to FRS households

This follows the exact same pattern as the existing wealth imputations. The model will now impute student loan balances based on WAS data (1.66m households with £33.4bn total SLC debt) to provide the balance data needed for accurate student loan repayment calculations.

Created PR #252 with the changes. The imputed values can be scaled to match current SLC admin totals (£267bn as of March 2025) in downstream usage if needed.