Methodology · Defensible Rebuild

How the account is built — and what it leaves out

The rebuild replaces a single R155.1bn headline with three transparent registers. Defensible activity is counted; real-but-unmeasured activity is tracked; and ineligible items are removed in plain sight.

Reconciliation: R155.1bn → R133.6bn

Every adjustment is visible and reversible if evidence arrives.

Original NSSA headline
R155.1bn
Watch Register (7 lines demoted)

Real activity, indefensible evidence — tracked annually

R8.3bn
Eliminated Register (6 lines removed)

Not sport per Vilnius 3.0 or no market exists

R5.4bn
Share corrections (2 lines halved)

C10 bicycle 70%→35%; C22 betting 60%→30%

R7.8bn
Core defensible GVA
R133.6bn

Share corrections

C1070% 35%

Bicycle retail

Most bicycle-kilometres in SA are commuting, not sport.

R1.2bn GVA

C2260% 30%

Sports betting

The 60% conflated gambling market share with value-added share. Casino, LPMs, bingo and lottery dominate gambling GVA.

R6.6bn GVA

Confidence-weighted view

Each line is weighted by evidence quality. The R133.6 billion core becomes R74.5 billion once weighted — a 44% quality discount.

TierWeightLinesSport GVAWeighted GVA
Medium×0.78R70.3bnR49.2bn
Low×0.417R63.3bnR25.3bn

Watch Register

Real activity that fails the evidence test today. Tracked, not counted.

R8.3bn tracked

Eliminated Register

Items removed because they fail the Vilnius 3.0 eligibility rules.

R5.4bn removed

Eligibility rules

Rule 1

Vilnius eligibility

The product must be a sport product as defined by the Vilnius 3.0 list.

Rule 2

Multipurpose exclusion

Multipurpose durable goods are excluded unless they are dedicated sport equipment.

Rule 3

No financial services

Insurance, fund management and pensions are outside the satellite account perimeter.

Rule 4

No phantom GVA

No GVA may be asserted where no measurable market exists.

Rule 6

Prudence on shares

Where a sport share is contested, the prudent (lower) share is used until evidenced.

Multipliers and indirect/induced effects are intentionally excluded. They require multi-region input-output (MR-IOT) modelling and are reported separately, never assumed as a flat constant.