AgentSkillsCN

evidence-based-giving

深耕慈善评估方法论、资助最佳实践、成本效益分析与战略性捐赠领域。当从事慈善评分、评估框架设计、捐赠者指导或影响力评估等工作时,可激活此技能。

SKILL.md
--- frontmatter
name: evidence-based-giving
description: Philanthropic expertise on charity evaluation methodologies, grantmaking best practices, cost-effectiveness analysis, and strategic giving. Activates when working on charity scoring, evaluation frameworks, donor guidance, or impact assessment.

Evidence-Based Giving Expert

Expert guidance on impact-focused charity evaluation, anchored on Giving What We Can and Longview Philanthropy methodologies.

When This Skill Activates

  • Designing or modifying charity evaluation criteria
  • Writing donor guidance or recommendations
  • Building scoring algorithms for charity assessment
  • Analyzing cost-effectiveness or impact metrics
  • Comparing charities across different cause areas

Core Principles (GWWC + Longview)

The Central Insight

Not all charities are equal. The best charities are 30x more effective than average charities pursuing identical goals. Rigorous evaluation identifies these differences.

Impact Over Overhead

Traditional evaluators focus on overhead ratios. This is wrong:

"Even if 100% of your donation goes directly to a charity's program, if that program isn't accomplishing a lot compared to others, then your money won't be either." — GWWC

The Cash Benchmark

GiveDirectly (unconditional cash transfers) is the floor:

  • GWWC threshold: ≥2x cash to recommend
  • GiveWell threshold: ≥10x cash for "top charity"

Key Cost-Effectiveness Data

MetricValue
Cost to save one life (top charities)~$5,000
Cost to save one life (government programs)~$10,000,000
Multiplier2,000x

Primary Frameworks

Giving What We Can (GWWC)

See resources/giving-what-we-can.md for complete methodology.

Two-Step Framework:

  1. Identify high-impact causes using ITN (Importance, Tractability, Neglectedness)
  2. Select effective organizations based on evidence, cost-effectiveness, transparency

Key Contributions:

  • 30x variance claim between best and average charities
  • Cash transfer benchmark for cost-effectiveness
  • Impact-focused evaluator network (GiveWell, ACE, Giving Green)
  • Pledge community (10,500+ members, $522M+ moved)

Longview Philanthropy

See resources/longview-methodology.md for complete methodology.

Five Principles:

  1. Radical Impartiality — future beneficiaries count equally
  2. Intellectual Honesty — disclose uncertainties; include "case against"
  3. Hits-Based Giving — accept failures for occasional big wins
  4. Win-Win Scenarios — favor sustainable interventions
  5. Scientific Mindset — quantify impact; acknowledge limits

Five-Stage Grant Process:

  1. Map promising areas (ITN prioritization)
  2. Initial investigation (counterfactual, track record, risks)
  3. Deep dive (resolve uncertainties, predictions)
  4. Recommendation report (impact pathway + "case against")
  5. Follow-up (actual vs predicted; learn)

Critical Elements Often Missing:

ElementWhat It Means
"Case Against"Document strongest objections to funding
CounterfactualWhat happens if charity doesn't exist?
Uncertainty DisclosureConfidence levels for all claims
Comparative ContextMetrics vs peer benchmarks

The ITN Framework

For cause-level prioritization:

code
Expected Impact = Importance × Tractability × Neglectedness
FactorQuestion
ImportanceHow many affected? How severely?
TractabilityAre there proven solutions?
NeglectednessIs it underfunded relative to scale?

See resources/giving-what-we-can.md for application details.

Evidence Quality Hierarchy

TierEvidence TypeConfidence
1Randomized Controlled TrialsHighest
2Quasi-experimental studiesHigh
3Observational with controlsMedium
4Pre/post comparisonsLow-Medium
5Expert opinion/track recordLow
6Anecdotes/testimonialsLowest

Traditional Evaluators (Data Sources Only)

These provide useful data but are not impact-focused:

EvaluatorUseful ForLimitation
Charity NavigatorFinancial ratios, governance flagsOverhead ≠ effectiveness
CharityWatchDeep financial analysisProcess over outcomes
BBB Wise GivingGovernance standardsBinary, no impact measure
Candid/GuideStarTransparency seals, 990 dataDisclosure ≠ effectiveness

Use these for data collection, not impact assessment.

Impact-Focused Evaluators

Rely on these for effectiveness judgments:

EvaluatorFocus Area
GiveWellGlobal health/poverty (gold standard)
Animal Charity EvaluatorsAnimal welfare
Giving GreenClimate change
Founders PledgeMultiple areas

Charity Red Flags

Quick screening (not impact assessment):

SignalConcern
Program expense ratio < 65%Financial health
No audited financials (>$500k revenue)Governance
Board < 5 members, family-dominatedIndependence
Only testimonials as evidenceImpact claims
High executive turnoverStability

Writing Donor Guidance

Principles

  1. Impact over overhead — don't penalize moderate admin costs
  2. Acknowledge uncertainty — state confidence levels
  3. Comparative context — "37% more efficient than sector median"
  4. Include "case against" — intellectual honesty
  5. Respect donor values — informative, not prescriptive

Zakat Adaptation

GWWC/Longview ApproachZakat Adaptation
Cause-neutral optimizationFixed 8 asnaf categories
Global optimizationMay prioritize Muslim communities
Long-term framingImmediate need also valid

Resources

ResourceContent
resources/giving-what-we-can.mdGWWC methodology, ITN, cash benchmark
resources/longview-methodology.md5-stage process, "case against", counterfactual
resources/evaluation-frameworks.mdTraditional evaluator details (reference only)

External Sources

Primary (Impact-Focused)

Secondary (Data Sources)