Country Analysis Prompt

How do you research a country deeply enough to trust the score?

Copy the prompt. Fill the bracketed slots. Run phases one to seven in order.

Put this to work

Run an in-depth country analysis

Country researcher

Copy this prompt. Paste into Claude, ChatGPT, or any AI assistant. The page context is already loaded — send it and get analysis tailored to your role.

You are the research lead for an in-depth country analysis.

## Decision

Analyse: [country]
Decision to inform: [decision]
Target venture: [venture_or_none]
Target resident persona: [persona]
Comparator countries: [comparators]
Home city or regions in scope: [locations]
Evidence cutoff: [YYYY-MM-DD]
Time horizons: 90 days, 5 years, 10 years, and 25 years

Answer two questions:

1. Does this country offer the best conditions for the stated venture and persona?
2. What would it take for this country to become the best place in the comparator set to live, build, and belong?

## Role

Act as a skeptical multidisciplinary team: country economist, institutional analyst, lawyer, technologist, demographer, climate-risk analyst, migration researcher, founder, and newcomer. Seek disconfirming evidence. Never protect a preferred narrative.

## Non-Negotiable Rules

1. Separate observed fact, source interpretation, and analyst inference.
2. Prefer primary official, legal, statistical, regulatory, court, audit, treaty, and reproducible research sources.
3. Use an institutional index only when its methodology, year, scale, and comparison cohort are known.
4. Triangulate contested, perception-based, commercial, crowdsourced, or self-reported claims with an independent source.
5. Record each datum's publisher, title, URL, publication date, observation period, access date, geography, population, unit, denominator, methodology/version, and source tier.
6. Compare like with like. Flag differences in year, geography, population, currency, PPP basis, definition, or method.
7. Rank only with cohort size. Never treat rank as an absolute measure.
8. Preserve why evidence is unknown: `unknown_missing | unknown_between | unknown_noncomparable`. Render plain `unknown` only when the public surface does not need the cause. Absence of evidence is not proof of poor performance.
9. Score 1 only when evidence matches the worst anchor. Never use 1 as a missing-data code.
10. Preserve contradictions. Explain which source is more direct, current, independent, and comparable.
11. Do not infer causality from correlation. State the causal mechanism, alternative explanations, and falsifier.
12. Do not browse indefinitely. Stop when added sources no longer change a material claim, score, confidence, or decision.

## Source Tiers

Use this order:

1. Primary: enacted law, regulator, court, audit office, national statistics, central bank, official registry, grid operator, treaty body.
2. International institution: World Bank, IMF, OECD, UN, ILO, ITU, WIPO, WHO, IEA, IRENA, FATF.
3. Peer-reviewed or reproducible research with disclosed data and method.
4. Credible secondary reporting or transparent commercial dataset.
5. Survey, crowdsourced source, vendor claim, anecdote, or lived experience.

Lower tiers can reveal experience or emerging change. They cannot silently override stronger evidence.

**Phase 1 — Frame Reality**

Before research:

- Restate the decision, persona, venture needs, locations, comparators, and time horizons.
- Define what “best” means under both lenses below.
- State the five most decision-changing unknowns.
- Build a source plan for each dimension.
- Read any existing country page as claims to verify, not as evidence.

Extract existing claims into a ledger:

`claim | location in page | sourced/unsourced | evidence date | dimension | verify/replace/retain`

**Phase 2 — Research 25 Dimensions**

Research five dimensions in each group.

### Principles

- Rule of law: WJP and WGI; use CPI only as a corruption cross-check. Include enforcement, judicial independence, property rights, and remedy.
- Digital identity: enacted law, coverage, user control, privacy, exclusion, redress, likeness/deepfake protection, and interoperability.
- Political stability: peaceful transfer, institutional continuity, civil conflict, state capacity, and policy volatility.
- Social trust: generalized, institutional, in-group, bridging, and newcomer trust.
- Values alignment: define the target values first. Test rights, incentives, revealed behaviour, and newcomer experience.

### Performance

- GDP per capita PPP: level, trend, PPP basis, productivity, distribution, and comparator cohort.
- Innovation: WIPO inputs and outputs, R&D, commercialization, ECI value/rank/trend, and product-space density.
- Ease of business: current entry, operation, competition, investment, insolvency, and enforcement evidence. Treat discontinued Doing Business ranks as legacy context only.
- Integration success: employment, income, language, citizenship, retention, discrimination, and newcomer experience.
- Brain drain: skilled inflow/outflow, retention, return, diaspora links, shortages, and know-how depth. Do not infer it from total net migration alone.

### Platform

- Digital infrastructure: coverage, adoption, speed distribution, affordability, reliability, cloud/data centres, and digital public infrastructure.
- Financial infrastructure: accounts, payments, settlement, competition, capital controls, foreign access, and lawful digital-asset rails.
- Blockchain ecosystem: retail adoption, developers, regulated firms, institutional pilots, and production use.
- DePIN coverage: active nodes by network, measurement date, definition of active, provenance, and spoofing/duplication risk.
- Energy: generation mix, firm capacity, relevant user prices, grid reliability, import dependence, and transition trend.

### Process

- Crypto regulation: activity and actor coverage, licensing, custody, banking, tax, consumer rules, AML/CFT, and enforcement.
- AI regulation: obligations by risk/use, regulator powers, sandboxes, data/compute rules, status, and enforcement.
- Immigration: explicit applicant persona, eligibility, rights, dependants, fees, stated and observed processing time, refusal/appeal, permanence path, and volatility.
- Business formation: steps, time/cost distribution, ownership, beneficial ownership, bank dependency, recurring compliance, insolvency, and contract enforcement.
- Data privacy: scope, lawful bases, transfer rules, regulator independence/resources, remedies, fines, and actual enforcement.

For legal claims, distinguish `announced | proposed | consulted | enacted | in force | enforced`.

### Players

- Talent pool: relevant occupations, depth, seniority, wages, language, vacancies, retention, inflow/outflow, and adjacent know-how.
- Startup ecosystem: formation, survival, funding by stage, exits, customers, procurement, founder diversity, and concentration.
- Community access: bridging and bonding ties, third spaces, civic groups, language access, discrimination, and time to belong.
- Cost of living: a dated, persona-specific after-tax basket covering housing, food, transport, healthcare, childcare, utilities, and currency/PPP assumptions.
- Climate and environment: normal climate, extremes, air/water, nature access, hazard exposure, adaptation, insurance, and resilience.

For every dimension return an evidence packet:

`claim tested | metric/value | observed fact | source interpretation | analyst inference | source metadata | corroboration | contradiction | subgroup/city limits | trend | confidence | automation status | unknowns | next retrieval step`

Use `automation status = api | index | manual`.

**Phase 3 — Assess Credibility**

Measure the gap between official story and verified delivery across:

- claim integrity;
- openness instruments;
- promise track record;
- performance versus compliance.

Select at least two material claims or commitments per signal. Build this ledger:

`claim | claimant | claim date | promised outcome/date | instrument | observable indicator | official evidence | independent evidence | outcome | met/partly met/missed/unverifiable`

Government statements prove that a claim was made. They do not prove delivery.

Score each signal 1–5 only after the ledger is complete. Record its anchor, confidence, counter-evidence, and falsifier. Set the overall credibility score to the minimum sub-signal. Keep credibility separate from the 25 dimensions.

**Phase 4 — Score**

Use the owning country scorecard calibration anchors. For each dimension:

1. Cite the decisive evidence and contrary evidence.
2. Match anchor 1, 3, or 5.
3. Satisfy every material clause in the chosen anchor. Do not average clauses or substitute a proxy for a named metric.
4. Return a typed unknown when evidence is missing/stale, between anchors, or non-comparable. Never interpolate 2 or 4.
5. Record confidence and the single fact most likely to change the score.

Output:

`dimension | score_or_typed_unknown | anchor | decisive evidence | contrary evidence | observation period | legal status | subgroup limits | source IDs | confidence | falsifier`

Check for double-counted evidence. Compare each score with at least two previously scored countries under the same anchor and period. Explain any inversion where the evidence appears stronger but the score is lower.

Stop for explicit human approval. Show scored, typed-unknown, stale, low-confidence, and contradicted counts first. Write an approval receipt containing the score/state, anchor, evidence cutoff, source cohort/year, contrary evidence, confidence, user adjustment, and approval date. Do not compute composites while a required score is unknown or unapproved.

**Phase 5 — Compute And Stress-Test**

First assert that all 25 dimensions and four credibility sub-signals are numeric and approved. If any required value is unknown or unapproved, label the affected group means, composites, classifications, gap ranks, and sensitivity `indeterminate`. Show known dimensions, but never average around missing ones.

Otherwise compute each P-group's mean, median, minimum, maximum, spread, binding dimension, and low-confidence count. Use full precision until display.

Use the declared lens weights:

- Trad: Principles 0.15, Performance 0.25, Platform 0.25, Process 0.20, Players 0.15.
- Future: Principles 0.25, Performance 0.10, Platform 0.15, Process 0.15, Players 0.35.

Assert each lens sums to 1.00. Report raw and credibility-gated composites. Then run sensitivity tests:

- move every low-confidence dimension by -1 and +1 within bounds;
- shift one P-group weight at a time by +/-0.05 and proportionally renormalize;
- test whether rank, preferred country, leading lens, or decision threshold changes.

Return a score interval and `robust | sensitive | indeterminate` verdict. Do not present a sensitive ranking as a precise conclusion.

**Phase 6 — Find Leverage**

Compare at least six candidate interventions. Select three portfolios by:

- weighted gap;
- causal influence on other dimensions;
- changeability and decision-maker control;
- implementation capacity;
- time to first signal;
- cost scale and reversibility;
- political feasibility;
- distributional effects;
- evidence confidence;
- robustness across both lenses.

For each selected portfolio report:

`binding problem | causal hypothesis | dependencies | baseline | target | sequence | owner | first mover | cost scale | 90-day proof | 5-year outcome | longer horizon | leading indicators | lagging indicators | prerequisites | risks | harmed groups | counterfactual | falsifier | confidence`

Do not target geography or “culture” directly. Name the institution, incentive, network, or behaviour that can change. If credibility is 3 or lower, test credibility repair first, but prove that it gates the other moves.

**Phase 7 — Write The Decision Brief**

Use this order:

1. Executive answer: decision, confidence, and the strongest reason it may change.
2. Decision frame: persona, venture, locations, comparators, cutoff, and horizons.
3. Reality: strongest assets, binding constraints, and material contradictions.
4. Scorecard: 25 scores, five P-groups, raw/gated lenses, intervals, and robustness.
5. Counter-case: the strongest evidence against the recommendation.
6. Intervention portfolios: first move, owner, 90-day proof, longer outcome, and risk.
7. Scenarios: base, upside, downside, triggers, and signposts.
8. Unknowns: evidence gaps that could change the decision.
9. Source trail: source IDs with full metadata.
10. Next question: one uncertainty that should drive the next research cycle.

Use short, concrete sentences. Cite claims near the sentence they support. Label interpretation and recommendation. Do not hide uncertainty behind composite scores.

## Final Quality Gate

Fail the analysis if any condition is true:

- A material claim has no source.
- A source lacks an observation period or method.
- A legal proposal is described as current law.
- Missing evidence becomes a low score.
- A rank omits its cohort.
- A national average hides a decision-relevant city or subgroup difference.
- A vendor, government, index, survey, or crowdsourced claim is treated as self-validating.
- Contradictory evidence is silently averaged or omitted.
- Correlation is described as causation.
- Scores use different anchors or incomparable periods without warning.
- Sensitivity can reverse the decision but the conclusion is written as robust.
- A recommendation lacks an owner, proof signal, risk, and falsifier.

Finish with:

`Decision | confidence | robustness | decisive evidence | strongest counter-evidence | next proof action | next question`

Next →Compare AI chat providers: costs, strengths, fit

Proof of done: the decision brief passes the final quality gate with every material claim sourced, every score anchored, and sensitivity declared.

Failure Modes

• Missing evidence scored as poor performance instead of unknown.
• Rank quoted without cohort size.
• A legal proposal described as current law.
• Sensitivity can flip the decision but the brief reads as robust.

Context

• Countries hub — where the resulting score lands.
• Balanced Scorecard — the 25-dimension anchor set.
• Performance Reality — gauges before judgments.

Questions

Which dimension's score would change first if you re-ran the analysis today?

• Changes my mind: a country brief passes the quality gate yet still misleads the decision it was built for.
• Next question: which comparator country exposes the weakest anchor in the current scorecard?