Add salary sacrifice NI relief as calibration target

Target the actual NI relief from salary sacrifice (£4.1bn from SPP/HMRC data)
rather than gross contributions. This uses a counterfactual simulation approach:
calculate what NI would be paid if salary sacrifice were converted to employment
income, then take the difference.

Current FRS data shows £0.63bn relief vs £4.1bn target, so calibration will
scale up weights approximately 6.5x to match the external benchmark.

Changes:
- tax_benefit.csv: Add salary_sacrifice_ni_relief target (£4.1bn for 2023)
- loss.py: Calculate NI relief via counterfactual simulation

Closes #215

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: MaxGhenis

changelog_entry.yaml

@@ -0,0 +1,4 @@
+- bump: minor
+  changes:
+    added:
+      - Add salary sacrifice NI relief as calibration target (£4.1bn from SPP/HMRC data)

policyengine_uk_data/storage/tax_benefit.csv

@@ -41,4 +41,5 @@ universal_credit_jobseekers,gbp-bn,obr_march_2024_efo,2.0,3.0,16.6,13.0,9.8,11.1
 universal_credit_non_jobseekers,gbp-bn,obr_march_2024_efo,6.2,15.3,21.7,27.6,32.1,41.0,52.2,60.6,67.1,68.6,70.9,73.4
 vat,gbp-bn,obr_march_2024_efo,,,,,159.7,168.9,170.3,178.8,186.6,193.9,201.7,209.8
 winter_fuel_allowance,gbp-bn,obr_march_2024_efo,2.0,2.0,2.0,2.0,2.0,2.0,0.3,0.3,0.3,0.3,0.3,0.3
-private_school_students,person-k,obr_march_2024_efo,557,557,557,557,557,557,557,557,557,557,557,
+private_school_students,person-k,obr_march_2024_efo,557,557,557,557,557,557,557,557,557,557,557,
+salary_sacrifice_ni_relief,gbp-bn,spp_ni_relief_2023,,,,,,,4.1,4.2,4.3,4.5,4.6,4.7

policyengine_uk_data/utils/loss.py

@@ -158,6 +158,37 @@ def pe_count(*variables):
     # Not strictly from the OBR but from the 2024 Independent Schools Council census. OBR will be using that.
     df["obr/private_school_students"] = pe("attends_private_school")
 
+    # Salary sacrifice NI relief - SPP estimates £4.1bn total (£1.2bn employee + £2.9bn employer)
+    # Calculate relief via counterfactual: what additional NI would be paid if SS became income
+    ss_contributions = sim.calculate("pension_contributions_via_salary_sacrifice")
+    employment_income = sim.calculate("employment_income")
+
+    # Run counterfactual simulation with SS converted to employment income
+    counterfactual_sim = Microsimulation(dataset=dataset, reform=reform)
+    counterfactual_sim.set_input(
+        "pension_contributions_via_salary_sacrifice",
+        time_period,
+        np.zeros_like(ss_contributions),
+    )
+    counterfactual_sim.set_input(
+        "employment_income",
+        time_period,
+        employment_income + ss_contributions,
+    )
+
+    # NI relief = counterfactual NI - baseline NI
+    ni_employee_baseline = sim.calculate("ni_employee")
+    ni_employer_baseline = sim.calculate("ni_employer")
+    ni_employee_cf = counterfactual_sim.calculate("ni_employee", time_period)
+    ni_employer_cf = counterfactual_sim.calculate("ni_employer", time_period)
+
+    employee_ni_relief = ni_employee_cf - ni_employee_baseline
+    employer_ni_relief = ni_employer_cf - ni_employer_baseline
+
+    df["obr/salary_sacrifice_ni_relief"] = household_from_person(
+        employee_ni_relief + employer_ni_relief
+    )
+
     # Population statistics from the ONS.
 
     region = sim.calculate("region", map_to="person")