Employer NI annualises the secondary threshold as 52 × £96 = £4,992, not the statutory £5,000 #1967

Problem

ni_class_1_employer annualises the weekly secondary threshold (ST) as 52 weeks:

secondary_threshold = class_1.thresholds.secondary_threshold * WEEKS_IN_YEAR  # WEEKS_IN_YEAR = 52

secondary_threshold.yaml stores the statutory weekly £96 from 2025-26, so the model's annual ST is 52 × £96 = £4,992.

The law gives £5,000 a year. SI 2001/1004 reg 11(3A) says:

• (b): "where the earnings period is a year, £5,000".
• (c): for a period that is a multiple of a week, the annual figure / 52 × the multiple (rounded up to a whole pound under reg 11(4)).

The primary threshold and upper earnings limit already follow reg 11(3)(c) and 11(2A)(c): they store annual / 52 (12,570 / 52 and 50,270 / 52). The ST is the only Class 1 threshold stored as the reg 10 weekly figure. Before 2025-26 the two conventions agreed (52 × £175 = £9,100).

Effect: from 2025-26, employer NI is overstated by up to 15% × £8 = £1.20 per employee a year above the ST. For example, the model gives £5,251.20 on £40,000 in 2026-27; £5,250.00 is correct.

Fix

• Store the ST from 2025-26 as 5,000 / 52, with a reg 11(3A) reference.
• Update the tests that pin £4,992. These include ni_class_1_employer.yaml, the salary sacrifice YAML tests, and ni_threshold_freeze.yaml / test_ni_threshold_freeze.py from Hold NICs thresholds at 2026-27 levels through 2030-31 #1959, which has a strict xfail for this.
• Run an Enhanced FRS impact.

Found while reviewing #1959.