How SA Policy Space Works
A transparent account of how policy ideas are sourced, assessed, categorised, and presented — including what this tool can and cannot tell you.
Data Sources
All policy ideas are sourced from committee proceedings documented by the Parliamentary Monitoring Group (PMG), an independent organisation that monitors the South African Parliament and maintains the most comprehensive public record of committee deliberations. PMG is the authoritative source — this project reads and synthesises their record; it does not reproduce it.
The database draws on proceedings from the following parliamentary portfolio and select committees:
Coverage spans 2015 to the present, with particular depth for the 7th Parliament (2024 onwards) following the Government of National Unity. Meeting records include committee minutes, briefing documents, and oral evidence sessions. Committees are listed under their current names: the 7th Parliament renamed several — Energy became Electricity and Energy, Mineral Resources became Mineral and Petroleum Resources — and dissolved Public Enterprises, whose state-owned-enterprise oversight moved to the relevant line committees. Older meeting records retain the committee name in use at the time.
Extraction Process
Parliamentary committee transcripts are rich in policy content but rarely organised around discrete, actionable ideas. The extraction process turns deliberative text into a structured database of reform proposals.
Thematic scanning
Each meeting record is read for recurring themes: explicit proposals by members or witnesses, references to draft legislation, performance gaps acknowledged by departments, and cross-committee patterns (the same constraint surfacing in multiple committees signals a binding bottleneck rather than a one-off complaint).
Idea synthesis
Each distinct policy proposal is written up as an original synthesis — not verbatim PMG text. We focus on proposals with measurable outcomes: ideas that could, in principle, be evaluated against a counterfactual. Vague aspirations (“improve coordination”) are excluded unless linked to a concrete mechanism.
Deduplication and linkage
Recurring proposals (the same idea discussed in 2019, 2022, and 2025) are merged into a single record with a meeting history. Frequency of appearance is one signal of political salience, but it does not directly enter the growth-impact or feasibility score.
Scoring Methodology
Each policy idea is rated on two dimensions, both on a 1–5 scale. These are researcher assessments, not algorithmic scores — they reflect judgement informed by the economic literature, international precedent, and South Africa–specific context.
Growth Impact (1–5)
Potential contribution to GDP growth or productivity over a 5–10 year horizon, anchored to the South African context. Draws on macro-econometric estimates (where available), analogous reforms in comparable economies, and the degree to which the reform addresses an identified binding constraint.
Feasibility (1–5)
Composite assessment of implementation likelihood, considering political feasibility (coalition alignment, veto players), institutional capacity (does the responsible department have the bandwidth?), budget requirements, and legislative complexity. Time horizon also matters: quick wins score higher even if long-run impact is moderate.
Stakeholder coalition score, and when it is withheld
The idea page's coalition score weights each mapped actor's position on that idea (support +2, conditional +0.5, neutral 0, oppose −2) by the actor's influence, then normalises to 0–100. Because it normalises by the influence of the actors that happen to be mapped, a single supportive actor would otherwise score a perfect 100.
So a number is shown only when the map is wide enough to mean something: at least 3 recorded stances, or mapped actors carrying at least 25% of all 38 actors' influence. Below that the page reads “N of 38 actors mapped — not scored” and the stakeholder cards stand on their own; the coalition axis of the matrix leaves those ideas unplotted. The tier labels say what the number is a reading of — mapped support: strong / mixed / contested / weak / outweighed — rather than passing a verdict on the reform.
Every stance row is drafted by a model and none have yet been editorially reviewed; the score card carries that count in its header. The Simulator's political axis is a different quantity — actor-level engagement posture across the 16–22 actors attached to each reform package — and is not gated by this floor.
Binding Constraints Framework
Policy ideas are organised around structural binding constraints on South African growth. The conceptual foundation is the Hausmann–Rodrik–Velasco (HRV) growth diagnostics framework, which holds that growth is best accelerated by removing the most binding constraint rather than attempting comprehensive reform across all fronts simultaneously. The approach asks: what is the shadow price of relaxing each constraint? Where is the return to reform highest given current conditions?
The ten constraints are drawn from the World Bank's South Africa Country Economic Memoranda, the National Planning Commission's diagnostic, and the academic literature on South African structural bottlenecks.
Energy & Electricity
Load-shedding and unreliable supply remain the most acute short-run constraint on output. Reforms cover generation, transmission, distribution, and the regulatory framework.
Logistics & Transport
Transnet's network deterioration and Durban port inefficiency raise the cost of traded goods and undermine export competitiveness, particularly in mining and agriculture.
Skills & Education
A skills mismatch — scarce technical and professional graduates, large numbers of youth neither in employment, education, nor training — constrains both the formal labour market and firm investment.
Regulation
Complex product market regulation, licensing burdens, and slow permitting processes raise the cost of doing business and deter investment, especially for smaller firms and new entrants.
Crime & Safety
High rates of violent crime and business disruption (including organised crime targeting infrastructure) impose direct costs on firms and households and deter tourism and foreign investment.
Labour Market
Rigid collective bargaining structures, high minimum wages relative to productivity in low-skill sectors, and search frictions contribute to structural unemployment above 30%.
Land & Property
Unresolved land tenure, a backlogged title registry, and slow urban land release constrain agricultural investment, housing supply, and household balance sheets.
Digital Infrastructure
Broadband penetration remains uneven, with high wholesale prices limiting connectivity for small firms and low-income households — a constraint on the service economy and e-government.
Government Capacity
Depleted technical capacity in key departments (SARS, DPWI, ESKOM oversight) and municipalities means that even well-designed reforms struggle to clear the implementation bottleneck.
Corruption & Governance
Corruption in procurement, licensing, and service delivery raises costs, diverts public investment, and undermines the credibility of institutions needed for all other reforms to succeed.
Reference: Hausmann, R., Rodrik, D., & Velasco, A. (2005). Growth Diagnostics. Harvard Kennedy School Working Paper. Applied to South Africa in NPC (2011), National Development Plan 2030.
Reform Packages
Individual policy ideas are grouped into five sequenced reform packages. Packaging reflects three considerations: (1) implementation dependencies — some reforms are preconditions for others; (2) political economy sequencing — early wins build credibility and coalition support for harder subsequent reforms; and (3) institutional ownership — ideas within a package tend to share a common lead department or coordinating body, reducing handoff costs.
Infrastructure Unblock
Energy, logistics, and digital infrastructure — the most acute short-run constraints. Package 1 reforms are largely within executive reach and would generate immediate investment signals.
SMME & Employment Acceleration
Regulatory, labour-market, and finance reforms that lower barriers for small business formation and formal employment — targeting structural unemployment.
Human Capital Pipeline
Basic education quality, TVET expansion, and graduate market alignment. Longer time horizon but essential for sustaining growth beyond the infrastructure cycle.
Trade & Industrial Competitiveness
Export facilitation, industrial policy rationalisation, and regional integration. Builds on infrastructure and skills gains from Packages 1–3.
State Capacity & Governance
Anti-corruption, professionalisation of the civil service, and municipal fiscal reform. These are cross-cutting reforms that raise the return on investment in all other packages.
The sequencing logic is described in detail on each Package page, including dependency maps and theory-of-change narratives.
International Comparisons
The database includes 59 case studies drawn from countries that have implemented reforms analogous to those in the SA policy space. Peer cases are selected on three criteria:
Countries represented include Chile, India, Vietnam, Brazil, South Korea, Rwanda, Kenya, Indonesia, Botswana, Morocco, and others. Explore the full dataset at International Comparisons.
Limitations & Caveats
Research tool, not a policy blueprint. SA Policy Space is designed to surface ideas and facilitate analysis. It should not be read as an endorsement of any particular reform path or a substitute for formal policy evaluation.
Scores reflect researcher judgement. Growth impact and feasibility ratings are not outputs of an econometric model. Reasonable analysts will disagree, especially on feasibility scores, which are sensitive to rapidly changing political conditions.
Coverage is not exhaustive. The database prioritises ideas with measurable growth linkages from key economic committees. Social policy committees, security and justice committees, and some NCOP proceedings are less fully covered.
Periodic updates. The database is updated periodically but does not reflect real-time parliamentary activity. Status updates (e.g., “implemented”, “stalled”) may lag actual developments by weeks or months.
Original synthesis, not verbatim PMG. Policy idea descriptions are the researcher's interpretive synthesis. They should not be attributed to committee members, departments, or PMG. For verbatim records, consult the PMG directly.
Not an official resource. This is an independent personal research project — not affiliated with the South African government, Parliament, PMG, or any employer of the author.
Reform Index: Weights & Sensitivity
The Reform Progress Index is the site's one composite number, so its construction is published in full here rather than described. Every weight below is read from the code that computes the score ( lib/reform-index.ts ), not transcribed, so the table cannot fall out of step with the index.
The formula
where np,s is the number of ideas in package p at status s, and np is the package's total idea count — recomputed from the live idea set on every read since the August 2026 re-derivation, not read from a stored literal.
Package weights
These are not equal and they are not estimated. They are a researcher judgement about strategic importance and cross-sector growth multipliers, informed by each package's theory of change and by the binding-constraint literature for South Africa. They sum to 1.00. Infrastructure carries the most because its constraints bind on every other package's reforms.
| Package | Weight | Ideas | Sub-index | Contribution |
|---|---|---|---|---|
| Infrastructure Unblock | 0.28 | 51 | 42 | 11.8 |
| SMME & Employment Acceleration | 0.16 | 26 | 26 | 4.2 |
| Human Capital Pipeline | 0.22 | 30 | 32 | 7.0 |
| Trade & Industrial Competitiveness | 0.14 | 24 | 27 | 3.8 |
| State Capacity & Governance | 0.20 | 40 | 33 | 6.6 |
| Index | 1.00 | 171 | 33 |
Status weights
The credit an idea earns for sitting at each status. Also not equal, and also a judgement: the jump from under review to partially implemented doubles the credit, on the view that partial implementation delivers real reform benefit while active attention delivers none yet.
| Status | Index credit |
|---|---|
| Implemented | 1.00 |
| Partially implemented | 0.60 |
| Under review | 0.30 |
| Drafted | 0.25 |
| Debated | 0.20 |
| Proposed | 0.10 |
| Stalled | 0.05 |
| Abandoned | 0.00 |
The aggregates the index reads distinguish only five of these eight rungs — implemented-or-partial and stalled-or-proposed arrive as combined counts, split 55/45 and 40/60 respectively, and everything else is treated as under review. Those splits are assumptions, and they are among the things the band below is testing.
Sensitivity: ±1 rung on every status
The usual robustness check for an index like this is to perturb the underlying scores by ±1 and see how far the headline moves. That check cannot be run on the feasibility ratings here, because — as noted above — feasibility is not an input. The one per-idea judgement the index does consume is which status rung an idea sits on, so that is what is perturbed: every idea is re-graded one rung up, then one rung down, and the index recomputed each time through the same function that produces the published score.
| Scenario | Index | Reading |
|---|---|---|
| Every status one rung lower | 21 | Systematically over-credited throughout |
| As published | 33 | Statuses as recorded |
| Every status one rung higher | 52 | Systematically under-credited throughout |
Under ±1 rung on every status, the index moves within 21–52 against a published 33 — a band 31 points wide, and asymmetric, because the ladder is not evenly spaced. That is wide, and it should be reported as wide. Two honest caveats in the other direction: both bounds assume the misclassification runs the same way for all 171 ideas simultaneously, which no real coding error does; and the index is built to be read as a trajectory rather than a level, so a band on the level matters less than whether the direction is robust. Quote the level with the band attached, or quote the direction.
Sensitivity: one rater, one notch
The growth-impact and feasibility ratings in section 3 are ordinal 1–5 scores assigned by a single rater, once. There is no inter-rater agreement to report and no standard error to attach, because nothing was sampled; they are judgements, and are labelled as such rather than dressed as measurements. Their leverage on the index is nil by construction — neither enters it — so the question worth asking is what one notch of the judgement the index does consume, the status rung, is worth. The arithmetic is short. The headline is the weighted sum of five sub-indices, so one point on a package sub-index moves it by that package's weight: 0.28 for Infrastructure Unblock at most, 0.14 for Trade & Industrial Competitiveness at least. One idea moving one rung moves its package sub-index by 100 × (credit step) ÷ (ideas in the package); with adjacent rungs 0.05 to 0.40 apart and packages of 24–51 ideas, a single reclassification is worth between 0.025 (State Capacity & Governance, smallest step) and 0.293 (Human Capital Pipeline, largest step) headline points — beneath the integer grain at which the score is published, so one disputed status usually leaves the printed number where it was and can never move it by more than one. What moves it materially is the band above: every one of 171 statuses wrong in the same direction.
Licence
Content and data on this site are published under Creative Commons Attribution 4.0 International (CC BY 4.0). Reuse them freely, including commercially, with attribution to “SA Policy Space / Laurence Wilse-Samson”. The application code is MIT. Parliamentary source material remains the property of the Parliamentary Monitoring Group and is linked, not reproduced. The licence above covers this site's original synthesis, assessments, and derived datasets.
How to Cite
For academic use, please cite as follows. Figures here are recomputed from live data, so the accessed date and the data version are the parts that make a citation reproducible — a reader who has both can tell whether the number you quoted is the number they are looking at.
BibTeX:
@misc{wilsesamson2026sapolicyspace,
author = {Wilse-Samson, Laurence},
title = {SA Policy Space: Mapping South Africa's Path to Reform},
year = {2026},
url = {https://sa-policy-space.vercel.app},
version = {2026-10-12.218dea82},
note = {Accessed 2026-10-12; data as of 2026-10-12}
}Filled at render from the live data, so the block above is already correct for this visit — copy it as it stands. The same pair appears in the footer of every page, in the meta block of every API response, and is explained on the changelog.
For questions or feedback: lw3387@nyu.edu