Calibration Scorecard

This chapter shows how closely the baseline matches South African data: tax collections and debt in 2025, the National Treasury’s fiscal plan, the long-run ratios, and productivity growth. Every figure and generated table comes from the model’s standard example run, examples/run_og_zaf.py, with the packaged default parameters. examples/calibration_scorecard.py rebuilds them from that run’s output; the text around them is written by hand.

2025: the starting year

Tax collections, debt and capital in the model’s first year, compared with published 2025 figures. Tax rates are set to reproduce collections. Debt, its interest rate and public investment are set from the data. The primary balance is a result of the model. Differences are in percentage points of GDP, except for private capital.

Measure	Model	Published	Difference	Source
Personal income tax / GDP	10.16%	10.15%	+0.01	PIT R790.6bn, MTBPS 2026 Table 3.1
Company tax (incl. dividends tax) / GDP	5.02%	5.02%	+0.00	CIT R345.4bn + dividends tax R45.9bn, MTBPS 2026 Table 3.1
Payroll taxes (SDL and UIF) / GDP	0.67%	0.67%	+0.00	Skills Development Levy R26.0bn (MTBPS 2026) + UIF contributions R26.7bn (UIF Annual Report 2025/26)
Consumption taxes, net of SACU / GDP	8.92%	8.92%	+0.00	VAT + fuel levy + excise + customs + other indirect taxes (9.87%) less SACU payments (0.95%), MTBPS 2026 / BR 2026
Property and transfer taxes / GDP	1.61%	1.61%	+0.00	municipal property rates R101.5bn (Stats SA P9114, 2024/25) + transfer duty and securities transfer tax (BR 2026)
Estate and donations tax / GDP	0.08%	0.08%	+0.00	estate duty R4.8bn + donations tax R1.2bn, BR 2026 annexure
Total tax revenue / GDP	26.46%	26.57%	-0.11	MTBPS 2026 gross tax, less SACU payments, plus UIF and municipal rates; includes 0.12 points of minor taxes the model does not carry
Private capital / GDP	2.80	2.80	+0.00	private capital: PWT 11.0 cn/cgdpo 3.82 (2023) less the general government share 26.7% (SARB QB E-5)
Public investment / GDP	2.42%	2.42%	+0.00	general government GFCF 2025, SARB QB S-127
Real interest rate on government debt	3.69%	3.70%	-0.01	debt service / average gross loan debt 7.08% nominal (MTBPS 2026) less ~3.4% expected inflation
Primary balance / GDP	1.34%	1.30%	+0.04	MTBPS 2026, 2025/26
Gross loan debt / GDP	77.00%	77.00%	+0.00	Gross loan debt at the end of 2024/25, MTBPS 2026

Public finances to 2036

The model’s fiscal path against the National Treasury’s 2026 Medium-Term Budget Policy Statement (MTBPS). Debt holds roughly flat, while the Treasury plans a small decline by 2030: the model keeps today’s primary surplus, and the Treasury plans to raise it every year to 2029. The baseline does not assume that consolidation. Government consumption steps up in 2030, when spending starts to follow the long-run budget rule. The rate on government debt stays at the Treasury’s effective real rate.

Fiscal path: model and Treasury plan, 2025–2036

Model debt is the stock at the start of each year, compared with the Treasury’s figure for the end of the previous fiscal year. Values are % of GDP; the real rate on debt is in %.

Year	Debt, model	Debt, Treasury	Primary balance, model	Primary balance, Treasury	Tax revenue	Government consumption	Social transfers	Public investment	Real rate on debt
2025	77.0	77.0	1.3	1.3	26.5	19.0	3.7	2.4	3.69
2026	77.3	78.4	1.7	1.6	26.5	18.7	3.7	2.4	3.68
2027	77.2	78.2	1.4	2.0	26.5	19.0	3.7	2.4	3.68
2028	77.4	77.9	1.4	2.5	26.6	19.0	3.7	2.4	3.67
2029	77.4	77.5	1.4	3.0	26.6	19.0	3.7	2.4	3.68
2030	77.2	76.6	0.8		26.6	19.6	3.7	2.4	3.68
2031	77.6		0.9		26.5	19.5	3.7	2.4	3.69
2032	77.9		1.0		26.5	19.4	3.7	2.4	3.70
2033	78.0		1.1		26.5	19.3	3.7	2.4	3.71
2034	78.1		1.1		26.5	19.2	3.7	2.4	3.71
2035	78.2		1.2		26.5	19.2	3.7	2.4	3.72
2036	78.3		1.2		26.5	19.1	3.7	2.4	3.72

The long run

Ratios in the steady state, which the economy reaches after the transition, compared with South African targets. Eleven of the fourteen targets are within 5%. Three are further away. The real return on private capital is 11.0% against 12.0% in the Penn World Table; a higher return would push private capital below its target of 2.80 times GDP. Property and transfer taxes are 1.51% of GDP against 1.61%, and estate and donations tax is 0.10% against 0.08%: both were set to match 2025 collections, and their bases change relative to GDP over the transition.

Steady state: model relative to target

Ratios without a target are shown for reference.

Measure	Model	Target	Source
Personal income tax / GDP	10.30%	10.15%	PIT R790.6bn, MTBPS 2026 Table 3.1
Company tax (incl. dividends tax) / GDP	5.03%	5.02%	CIT R345.4bn + dividends tax R45.9bn, MTBPS 2026 Table 3.1
Payroll taxes (SDL and UIF) / GDP	0.67%	0.67%	Skills Development Levy R26.0bn (MTBPS 2026) + UIF contributions R26.7bn (UIF Annual Report 2025/26)
Consumption taxes, net of SACU / GDP	9.19%	8.92%	VAT + fuel levy + excise + customs + other indirect taxes (9.87%) less SACU payments (0.95%), MTBPS 2026 / BR 2026
Property and transfer taxes / GDP	1.51%	1.61%	municipal property rates R101.5bn (Stats SA P9114, 2024/25) + transfer duty and securities transfer tax (BR 2026)
Estate and donations tax / GDP	0.10%	0.08%	estate duty R4.8bn + donations tax R1.2bn, BR 2026 annexure
Real interest rate on government debt	3.71%	3.70%	debt service / average gross loan debt 7.08% nominal (MTBPS 2026) less ~3.4% expected inflation
Gross loan debt / GDP	76.50%	76.50%	MTBPS 2026 gross loan debt 2029/30
Foreign-held share of government debt	25.00%	25.00%	non-resident share of domestic bonds, BR 2026
Social transfers / GDP	3.74%	3.60%	social grants incl. SRD, BR 2026 Table 5.7
Public investment / GDP	2.42%	2.42%	general government GFCF 2025, SARB QB S-127
Foreign-owned share of private capital	21.24%	20.50%	upper bound: inward DI equity 22.7% + portfolio equity 34.8% of GDP (SARB IIP end-2025) / private K/Y; DI equity alone 0.08
Private capital / GDP	2.79	2.80	private capital: PWT 11.0 cn/cgdpo 3.82 (2023) less the general government share 26.7% (SARB QB E-5)
Real return on private capital	11.05%	12.00%	PWT 11.0 real internal rate of return, 2023
Government consumption / GDP	18.86%		not targeted
Net exports / GDP	5.66%		not targeted
Total investment / GDP	17.58%		not targeted
Household consumption / GDP	67.09%		not targeted
Household wealth / GDP	2.77		not targeted
Total tax revenue / GDP	26.79%		not targeted

Productivity and growth

South Africa’s productivity growth has been weak since 2008. In the baseline it recovers gradually: growth in output per hour follows the average of two estimates, a rule based on recent growth and the OECD’s long-term baseline (the firms chapter sets out both). The long-run rate, 1.4% a year, is South Africa’s growth in output per worker over 1994–2019 in the Penn World Table (see the macro chapter).

Growth in output per hour: model, baseline path and the two estimates

GDP grows faster than productivity because hours worked rise with the working-age population, at today’s employment rates by age. Total factor productivity starts above its long-run trend and declines to it, which holds productivity growth below the long-run rate.

GDP growth by source, and total factor productivity relative to trend

Average growth by period, % a year:

Period	GDP	Output per hour, model	Baseline path	Estimate 1: recent growth	Estimate 2: OECD	Hours worked
2026–2030	1.75%	0.53%	0.54%	0.28%	0.80%	1.21%
2031–2040	2.16%	1.10%	1.10%	0.80%	1.40%	1.04%
2041–2050	2.09%	1.18%	1.18%	0.93%	1.44%	0.90%

Why the baseline does not use the IMF forecast

Since the 2009 recession, the IMF’s medium-term growth forecasts for South Africa have been too high every year. Leaving out the Covid years 2020 and 2021, every forecast made five years ahead for 2009–2024 was above the outturn; for 2013–2024 the gap averaged 2.1 percentage points a year. Before 2009 the errors went both ways. The IMF, SARB and Treasury paths assume that energy and logistics reforms succeed, so we treat them as reform scenarios rather than the baseline. The data are in data/imf_weo_growth_forecasts.csv, from the WEO historical forecasts database and the current WEO.

IMF five-year-ahead growth forecasts for South Africa and actual growth

The path to the long run

The main ratios move smoothly from today’s values to the steady state, with no instability.

Main ratios on the baseline path, 2025–2084

Accuracy of the solution

How closely the solution satisfies the model’s equations in each year. Output not accounted for by spending is below 0.0000006 of output in every year to 2100, and household decisions are accurate to about 0.00000000001. The one spike, 0.0001 of output in 2145, is listed under known gaps.

Errors in the solution by year, log scale

Example policy: a higher company tax rate

The standard example raises the company tax rate to 30%. Capital, wages and GDP fall, as expected; hours worked rise slightly as households make up lost income. Changes are % from the baseline.

Higher company tax rate: change from the baseline

Variable	2025	2030	Average 2025–2034	Long run
GDP	-0.36	-0.14	-0.24	-0.62
Consumption	-0.18	-0.37	-0.35	-0.94
Capital	-0.75	-0.45	-0.63	-1.57
Labour	-0.01	+0.14	+0.10	+0.23
Real interest rate	-1.30	-1.41	-1.31	-0.57
Wage rate	-0.35	-0.28	-0.34	-0.85

What this calibration changed

Setting	Area	Before	After	Reason
Personal income tax schedule	Taxes	A flat 22% rate	A progressive schedule; brackets rise with average income over time	Collects 10.15% of GDP in 2025, as SARS did. The two lowest-income groups pay no income tax.
Consumption tax rate	Taxes	15%	15.9%	VAT, fuel levy, excise and customs, less payments to SACU partners: 8.92% of GDP
Company tax collections	Taxes	30% of the statutory rate collected	74% of the statutory rate collected	Company tax plus dividends tax: 5.02% of GDP
Payroll taxes	Taxes	None	1.27% of wages	Skills Development Levy and UIF contributions: 0.67% of GDP
Property and transfer taxes	Taxes	None	Up to 0.73% of household wealth a year	Municipal rates, transfer duty and securities transfer tax: 1.61% of GDP
Estate and donations tax	Taxes	20% of bequests	1.0% of bequests	Estate duty and donations tax: 0.08% of GDP
Depreciation rate	Production	5% a year	4.05% a year	Depreciation of private capital divided by the private capital stock
Household patience (discount factor)	Households	0.96	0.871	Sets private capital at 2.80 times GDP, the Penn World Table figure net of government capital
Long-run productivity growth	Growth	0.59% a year	1.4% a year	Growth in output per worker, 1994–2019 (Penn World Table 11.0: 1.44%)
Productivity path to the long run	Growth	At its long-run level from the start	Starts 8.3% above trend in 2025 and falls to it	Output per hour grows 0.4% in 2026 and about 1.1% in the 2030s, the average of two estimates
World interest rate	Open economy	4%	6.3%	4% plus South Africa’s sovereign premium of 2.26 points (Treasury)
Share of extra capital demand met from abroad	Open economy	90%	16%	Foreign-owned share of capital matches the balance of payments position (21%)
Foreign share of government debt	Open economy	16.2%	25%	Non-resident holdings of domestic government bonds
Government debt at the start	Debt	84.2% of GDP	77.0% of GDP	Gross loan debt at the end of 2024/25 (MTBPS 2026)
Long-run government debt	Debt	120% of GDP	76.5% of GDP	The Treasury’s projection for 2029/30
Real interest rate on government debt	Debt	Cross-country formula	3.7%, rising when debt is above its long-run level	The Treasury’s effective real rate on its debt
Public investment	Spending	0.03% of GDP	2.42% of GDP	General government fixed investment (SARB)
Government consumption	Spending	23.3% of GDP	19.0%, 18.7% then 19.0% of GDP	The first two years follow the Treasury’s primary balance plan; then the level that keeps debt stable
Start of the long-run budget rule	Spending	2045	2030	After the Treasury’s medium-term plan ends
Household wealth at the start	Households	Not set	Set from the capital stock	So that private capital is 2.80 times GDP in 2025
Population data	Population	UN data to the latest year	UN projections to 2098; death and birth rates differ by income group	Long-run population growth falls from 0.42% to −0.03% a year

The parameter names and exact values are in CHANGELOG.md.

Known gaps and choices

Item	Status	Detail
Primary surplus 2027–2029	By design	The model keeps a surplus of about 1.4% of GDP. The Treasury plans 2.0–3.0%. The baseline does not assume the planned consolidation, so debt stays near 77% of GDP instead of falling to 76.6%.
Spending step in 2030	Known	When the long-run budget rule starts, government consumption rises by about 0.6% of GDP in one year and the primary surplus falls to 0.8%.
Return on capital	Close	11.0% in the long run, against 12.0% in the Penn World Table (11–13% over 2015–2023). A higher return would push private capital below its 2.80 times GDP target.
Consumption taxes in the long run	Close	Set to match 2025 (8.92% of GDP); rise to 9.19% in the long run as consumption’s share of GDP grows.
Social transfers	Close	3.74% of GDP in the long run against 3.6% in the budget.
Long-run growth rate	Close	The model’s long-run productivity growth is 1.4% a year (1.41% compounded). The OECD estimate is capped at the measured 1.44%. The difference moves the baseline path by at most 0.02 points a year.
Error in 2145	Open	The accounting error is 0.0001 of output in 2145 alone, and below 0.0000006 in every year to 2100. 2145 is the last year of the listed income tax schedule; listing it to the end of the horizon should remove the error. No reported year is affected.
Growth above the IMF	By design	GDP grows about 2.1% a year in the 2030s, above the IMF’s 1.8%, because hours worked rise with the working-age population. Productivity growth is below the IMF, SARB and Treasury scenarios.
Unemployment	Covered	Built into earnings by age and income group: the earnings data are per person, the jobless included (see the earnings chapter). Labour supply settings are unchanged.
Informal sector	Covered	34% of jobs (ILO). Reflected in tax rates, which are set from actual collections rather than statutory rates (see the taxes chapter).
Earnings distribution	Not re-checked	This calibration does not change earnings by age and income group or the labour supply settings; their fit was not re-checked here.