Golden Rice Delay Dashboard

As of 2026, no country on Earth is growing Golden Rice commercially.

This is my estimate of what the Greenpeace-led campaign to block Golden Rice in every country from 2006–2024 has cost:

~106,000
children dead
range ~71,000–141,000
~210,000–425,000
children blinded
~7–12 million
years of healthy life lost

Cumulative, 2006–2024, across eleven target countries.[1][2]

† The range depends on estimates of how much vitamin A reduces child mortality (GiveWell's reading of the trials: 12–24%).[1]
Beta-carotene content is set at 6.3 µg/g — the average of the only two field studies that measured actual beta-carotene per gram of GR2E grain, in the Philippines (Swamy 2019) and Bangladesh (Biswas 2021).[3][4]

See the back-of-the-envelope estimate and calculations below — you can run them yourself. I'm actively seeking feedback to make them better; contact me at abio.substack.com.

Current status

Every major food-safety regulator that has reviewed Golden Rice has approved it: Australia and New Zealand (2017),[5] Canada and the US FDA (2018),[6][7] and the Philippines (2019).[8]

The Philippines approved commercial planting in July 2021—the first country to do so—and farmers began growing it. In April 2024 its Court of Appeals revoked the permit and ordered all cultivation to stop, citing the precautionary principle; the Supreme Court case is still pending.[9] Bangladesh's application has sat undecided since 2017.[10]

How I got these numbers

See the full calculation in Python. Below is a detailed breakdown of the countries included plus the cumulative total from 2006–2024.

Vitamin A supplements already reach about 75% of targeted children worldwide.[11] My estimate accounts for this and focuses on the children who haven't been reached by them. India is the largest contributor here because of its population size and rice-heavy diet. Only about 60% of children are reached there.[12]

Note: China is excluded because vitamin A deficiency prevalence is now very low.

The country breakdown

This table shows what would happen today if every grain of domestically produced rice in each country were Golden Rice.

How we calculated "vitamin A delivered per day"

  • Beta-carotene in GR2E grain: 6.3 µg/g (midpoint of the two field measurements of the deployed event; see note)[3][4]
  • Storage retention: 65% survives 3–6 months of tropical storage[13]
  • Cooking retention: 60% survives cooking[14]
  • Bioconversion: 3.8:1 ratio (Golden Rice–specific; see note below)[15]
  • Child rice intake: 30% of per-capita adult consumption[16]
  • Child RDA: 400 µg retinol activity equivalents (RAE) per day[17]

Where the 6.3 µg/g comes from: Only two studies have measured beta-carotene in milled grain of the deployed GR2E event. Swamy 2019 measured it directly in Philippine lines: 3.57 µg/g.[3] Biswas 2021 measured Bangladesh (BRRI dhan29) lines at 10.6 µg/g total carotenoids, of which the paper says 80–90% is beta-carotene — about 9 µg/g.[4] The midpoint is 6.3. Higher figures you may see (20–37 µg/g) are the original lab donor line before it was bred into farmers' varieties — not the rice that would actually be grown. Storage and cooking losses are applied separately, below.

Why 3.8:1 bioconversion instead of the WHO 12:1 default: The WHO uses 12:1 as a generic ratio for all plant-source beta-carotene—spinach, carrots, sweet potatoes, etc. But those foods store beta-carotene in green tissue, behind plant cell walls and chloroplasts that the gut has trouble breaking down. Golden Rice stores beta-carotene in the starchy endosperm, not green tissue. The only direct measurement of Golden Rice bioconversion (Tang et al. 2009) found 3.8:1 in healthy adults.[15] A second study in children found an even better 2.3:1, but it was retracted after uproar because researchers didn't disclose that the rice was genetically modified. It is unclear whether this disclosure was necessary. The scientific findings themselves were never challenged—only the lack of disclosure.[18] If someone prefers the WHO generic 12:1 default, all estimates below would be roughly halved.

Country VAD in children <5 Rice as % of diet Rice self-sufficiency Vitamin A / day
per child, from GR
% of daily need
= vit A ÷ 400µg RDA
Deaths prevented / year
India Rice-belt states only (55% of VAD children) VAS reaches ~60%, not the 83% S. Asia average (NFHS-4/5)[12] 57% NFHS-2, 2000[19] 31% national; 55%+ in rice belt[20] 100% (net exporter) 77 µg RAE 19% ~10,400
Indonesia 50% est., WHO SE Asia regional, 2000[1] 59%[20] 95% 72 µg RAE 18% ~1,200
Nigeria Rice is secondary staple (35% of VAD children eat rice) 30% est.[21] ~8% (yam, cassava dominate) 55% (~45% imported) 21 µg RAE 5% ~940
Myanmar 36% est., WHO SE Asia regional, 2000[1] 76%[20] 100% (net exporter) 104 µg RAE 26% ~510
Bangladesh 21% Nat'l survey, 2000[22] 73%[20] 97% 93 µg RAE 23% ~270
Vietnam 45% est., WHO SE Asia regional, 2000[1] 68%[20] 100% (net exporter) 88 µg RAE 22% ~240
Cambodia 45% est., WHO SE Asia regional, 2000[1] 80%[20] 100% (net exporter) 106 µg RAE 27% ~140
Philippines 38% FNRI, 2003[23] 40%[20] ~79%[24] 61 µg RAE 15% ~130
Laos 42% est., WHO SE Asia regional, 2000[1] 65%+[20] 100% 101 µg RAE 25% ~93
Tanzania Maize is primary staple (28% of VAD children eat rice) 33% DHS, 2004[25] Low (maize dominant) ~60% 17 µg RAE 4% ~88
Nepal Terai rice belt (65% of VAD children) 31% DHS, 2001[26] 44%[20] 90% 64 µg RAE 16% ~52
TOTAL across 11 countries ~14,000

How "deaths prevented" was calculated: For each country, we took the number of under-5 deaths attributable to VAD (from UN mortality data[2] and VAD prevalence rates), subtracted the children already protected by existing vitamin A supplement programs, then multiplied by the fraction of daily vitamin A need that Golden Rice would fill. The result is how many of those remaining deaths Golden Rice could avert. Full formula and every input are in "The calculation" below.

Deaths prevented per year if all domestic rice were Golden Rice

India
~10,400
Indonesia
~1,200
Nigeria
~940
Myanmar
~510
Bangladesh
~270
Vietnam
~240
Cambodia
~140
Philippines
~130
Laos
~93
Tanzania
~88
Nepal
~52

Multiplied by years

Golden Rice's current variety (GR2E) was ready by 2005.[27] We used a realistic adoption curve: slow at first, then accelerating, then leveling off at a 70% ceiling. We assumed deployment would have started country by country between 2006 and 2013, depending on regulatory capacity and IRRI partnerships.[28]

Under these assumptions, this model estimates approximately 106,000 children's lives lost (range 71,000–141,000) and 210,000–425,000 children blinded across these 11 countries from 2006 to 2024. The blindness estimate uses the WHO ratio: globally, 2–4x as many children go blind from vitamin A deficiency as die from it.[29][30]

'06'07'08'09'10 '11'12'13'14'15 '16'17'18'19'20 '21'22'23'24

Estimated annual lives saved per year if Golden Rice had been deployed starting 2006–2013 (varies by country).
Peak: ~10,300 in 2021. Hover/tap bars for exact values.

The decline after 2021 reflects falling VAD prevalence and child mortality—the underlying problem is slowly getting better even without Golden Rice, due to economic growth and vitamin A supplement programs.

Per-country cumulative lives lost, 2006–2024

India
79,000
Indonesia
10,500
Nigeria
3,600
Myanmar
3,530
Bangladesh
3,180
Vietnam
2,550
Philippines
1,790
Cambodia
940
Laos
495
Nepal
417
Tanzania
293

The calculation

Lives saved (per country, per year) = Under-5 deaths that year × VAD attributable fraction × (1 − effective VAS coverage) × adoption rate × (GR efficacy fraction) ^ 0.60 Where: VAD attributable fraction = P(RR−1) / (1 + P(RR−1)) P = VAD prevalence that year RR = 1.75 (relative risk of death for VAD children) Effective VAS coverage = VAS coverage × 0.70 (VAS only protects ~70% of nominally covered children) GR efficacy = min(RAE delivered / 400, 1.0) RAE delivered = child rice (g) × 6.3 × 0.65 × 0.60 / 3.8 Adoption rate = logistic S-curve ceiling = 70% × domestic rice fraction × rice-eating VAD fraction midpoint = 8 years after deployment steepness = 0.45 Years of healthy life lost: Deaths: 106,000 × 28 years lost per child death = ~3.0 million Blindness: 210,000–425,000 × 21 QALYs per blinded child = ~4.4–8.9 million Combined: ~7–12 million years of healthy life 28 years = WHO avg life expectancy lost per under-5 death 21 QALYs = ~35 remaining years × 0.6 disability weight (WHO blindness) QALY = quality-adjusted life year (a standard health economics measure)

Where each number comes from

Parameter Value Source
Under-5 deaths by country and year Varies (e.g. India: 2.3M in 2000, 560K in 2024) UN IGME / UNICEF[2]
VAD prevalence (baseline) 21–57% depending on country WHO VMNIS, national DHS/MICS, Stevens et al. 2015[1]
VAD annual decline rate 1.5–4% per year (country-specific) Stevens et al. 2015[1]
Relative risk of death (VAD vs. non-VAD) 1.75 Sommer et al. 1983, West et al. 1991, Fawzi et al. 1993[30]
VAS coverage by country and year 40–93% UNICEF / WHO joint estimates[11]
VAS effectiveness multiplier 0.70 Imdad et al. Cochrane 2022[31]
Beta-carotene in GR2E grain (milled) 6.3 µg/g (field range 3.6–9) Swamy 2019[3], Biswas 2021[4]
Storage retention (tropical, 3–6 mo) 65% (model); ~13% at 10 weeks (Schaub 2017) Schaub et al. 2017[13]
Cooking retention 60% IRRI standard[14]
Bioconversion (beta-carotene → retinol) 3.8:1 (Golden Rice–specific measurement) Tang et al. 2009[15]; WHO generic is 12:1[17]
Child RDA (vitamin A) 400 µg RAE/day WHO[17]
Adoption ceiling 70% Based on IR8 and HarvestPlus data[28]
Adoption midpoint 8 years IR8 historical data[28]
Dose-response concavity 0.60 Estimated; reflects that partial supplementation has diminishing returns[30]
Rice consumption (per capita/year) 32–200 kg (country-specific) FAO FAOSTAT[20]

The full script

The Python script uses no special libraries. Save it as check_calculation.py and run python3 check_calculation.py. It prints four tables that match the four sets of numbers on this page.

#!/usr/bin/env python3
"""
Golden Rice "cost of delay" — full calculation.

Run: python3 check_calculation.py

  PART A — vitamin A delivered per child per day, and % of daily need   (table)
  PART B — deaths prevented per year if ALL domestic rice were Golden    (table)
  PART C — cumulative children's lives lost 2006–2024                    (headline)
  PART D — blindness and years-of-healthy-life-lost                      (headline)
"""

import math

# ============================================================================
# 1. PARAMETERS  (sources cited on the web page)
# ============================================================================
BETA_CAROTENE_UG_PER_G   = 6.3     # µg beta-carotene per gram of GR2E grain (field-realistic)
STORAGE_RETENTION        = 0.65    # fraction surviving 3–6 months tropical storage
COOKING_RETENTION        = 0.60    # fraction surviving cooking
BIOCONVERSION            = 3.8     # µg beta-carotene -> 1 µg retinol (Tang 2009, GR-specific)
CHILD_RDA_UG             = 400.0   # µg RAE/day recommended for a young child (WHO)
CHILD_RICE_FRACTION      = 0.30    # a child under 5 eats ~30% of the per-capita adult portion

RELATIVE_RISK_VAD        = 1.75    # a VAD child's risk of death vs. a vitamin-A-replete child
EFFECTIVE_VAS_MULTIPLIER = 0.70    # share of "supplement-covered" kids actually protected
DOSE_RESPONSE_CONCAVITY  = 0.60    # partial vitamin A -> less-than-proportional benefit

ADOPTION_CEILING         = 0.70    # max adoption in the realistic S-curve scenario
ADOPTION_MIDPOINT_YEARS  = 8.0     # years after launch to reach half the ceiling
ADOPTION_STEEPNESS       = 0.45    # S-curve slope

MODEL_START_YEAR         = 2000
MODEL_END_YEAR           = 2024    # last year with real, filled-in data

# Years-of-life constants (WHO; used only for PART D)
YEARS_LOST_PER_DEATH     = 28.0    # discounted life-years lost per under-5 death (raw ~55-60)
QALYS_PER_BLIND_CHILD    = 21.0    # ~35 remaining years × 0.6 disability weight

# Blindness multiplier: WHO says 2–4x as many children go blind from VAD as die from it
BLINDNESS_LOW_MULT       = 2.0
BLINDNESS_HIGH_MULT      = 4.0

# How strongly does vitamin A actually cut child deaths? (the GiveWell / Cochrane question)
# The same Cochrane meta-analysis gives two numbers: a 24% reduction (random-effects model)
# and a 12% reduction (fixed-effects model, which gives full weight to DEVTA, the single
# largest trial). The model's raw output corresponds to the 24% (high) end; we center the
# headline on the MIDPOINT (~18%) and show 12–24% as the range. GiveWell uses this same band.
MORTALITY_EFFECT_HIGH    = 1.00   # 24%, random-effects (the model's raw output)
MORTALITY_EFFECT_CENTRAL = 0.75   # ~18%, midpoint of the 12–24% range
MORTALITY_EFFECT_FLOOR   = 0.50   # 12%, fixed-effects (full weight to DEVTA)

# ============================================================================
# 2. COUNTRY DATA  (inputs; sources cited on the page)
# ============================================================================
# under5_deaths: anchor years, linearly interpolated between
# vad_baseline: (year, prevalence fraction), declines at vad_decline per year
# vas: anchor years of vitamin-A-supplement coverage, interpolated
# rice_kg: per-capita milled rice, kg/year
# domestic: fraction of consumed rice grown domestically (reachable by seed system)
# deploy: counterfactual Golden Rice launch year in this country
# rice_eating_vad: fraction of VAD-affected children who actually live on rice
COUNTRIES = {
    "India": dict(
        under5={2000: 2_300_000, 2005: 1_900_000, 2010: 1_450_000, 2015: 1_000_000, 2020: 700_000, 2024: 560_000},
        vad_year=2000, vad_base=0.57, vad_decline=0.020,
        vas={2000: 0.44, 2005: 0.50, 2010: 0.55, 2015: 0.60, 2020: 0.60, 2024: 0.60},  # NFHS-4: 60.5%
        rice_kg=145.0, domestic=1.00, deploy=2009, rice_eating_vad=0.55),
    "Indonesia": dict(
        under5={2000: 290_000, 2005: 225_000, 2010: 160_000, 2015: 112_000, 2020: 80_000, 2024: 65_000},
        vad_year=2000, vad_base=0.50, vad_decline=0.030,
        vas={2000: 0.70, 2005: 0.78, 2010: 0.82, 2015: 0.80, 2020: 0.77, 2024: 0.75},
        rice_kg=135.0, domestic=0.95, deploy=2008, rice_eating_vad=1.00),
    "Nigeria": dict(
        under5={2000: 850_000, 2005: 780_000, 2010: 700_000, 2015: 610_000, 2020: 490_000, 2024: 420_000},
        vad_year=2000, vad_base=0.30, vad_decline=0.015,
        vas={2000: 0.40, 2005: 0.50, 2010: 0.55, 2015: 0.52, 2020: 0.50, 2024: 0.48},
        rice_kg=40.0, domestic=0.55, deploy=2012, rice_eating_vad=0.35),
    "Myanmar": dict(
        under5={2000: 120_000, 2005: 90_000, 2010: 59_000, 2015: 37_000, 2020: 25_000, 2024: 20_000},
        vad_year=2000, vad_base=0.36, vad_decline=0.025,
        vas={2000: 0.55, 2005: 0.65, 2010: 0.73, 2015: 0.74, 2020: 0.70, 2024: 0.58},
        rice_kg=195.0, domestic=1.00, deploy=2009, rice_eating_vad=1.00),
    "Bangladesh": dict(
        under5={2000: 250_000, 2005: 180_000, 2010: 115_000, 2015: 70_000, 2020: 48_000, 2024: 36_000},
        vad_year=2000, vad_base=0.21, vad_decline=0.035,
        vas={2000: 0.75, 2005: 0.87, 2010: 0.93, 2015: 0.92, 2020: 0.89, 2024: 0.87},
        rice_kg=175.0, domestic=0.97, deploy=2007, rice_eating_vad=1.00),
    "Vietnam": dict(
        under5={2000: 85_000, 2005: 61_000, 2010: 42_000, 2015: 27_000, 2020: 19_000, 2024: 16_000},
        vad_year=2000, vad_base=0.45, vad_decline=0.040,
        vas={2000: 0.76, 2005: 0.84, 2010: 0.87, 2015: 0.85, 2020: 0.83, 2024: 0.80},
        rice_kg=165.0, domestic=1.00, deploy=2007, rice_eating_vad=1.00),
    "Cambodia": dict(
        under5={2000: 50_000, 2005: 35_000, 2010: 21_000, 2015: 13_000, 2020: 8_500, 2024: 6_500},
        vad_year=2000, vad_base=0.45, vad_decline=0.030,
        vas={2000: 0.65, 2005: 0.78, 2010: 0.82, 2015: 0.83, 2020: 0.81, 2024: 0.79},
        rice_kg=200.0, domestic=1.00, deploy=2010, rice_eating_vad=1.00),
    "Philippines": dict(
        under5={2000: 65_000, 2005: 51_000, 2010: 37_000, 2015: 25_000, 2020: 17_000, 2024: 13_500},
        vad_year=2003, vad_base=0.38, vad_decline=0.040,
        vas={2000: 0.70, 2005: 0.80, 2010: 0.85, 2015: 0.83, 2020: 0.82, 2024: 0.80},
        rice_kg=115.0, domestic=0.85, deploy=2006, rice_eating_vad=1.00),
    "Laos": dict(
        under5={2000: 22_000, 2005: 16_000, 2010: 11_000, 2015: 7_000, 2020: 4_800, 2024: 3_800},
        vad_year=2000, vad_base=0.42, vad_decline=0.025,
        vas={2000: 0.55, 2005: 0.68, 2010: 0.74, 2015: 0.75, 2020: 0.72, 2024: 0.70},
        rice_kg=190.0, domestic=1.00, deploy=2011, rice_eating_vad=1.00),
    "Tanzania": dict(
        under5={2000: 225_000, 2005: 200_000, 2010: 160_000, 2015: 115_000, 2020: 78_000, 2024: 60_000},
        vad_year=2000, vad_base=0.33, vad_decline=0.018,
        vas={2000: 0.50, 2005: 0.62, 2010: 0.70, 2015: 0.70, 2020: 0.67, 2024: 0.63},
        rice_kg=32.0, domestic=0.60, deploy=2013, rice_eating_vad=0.28),
    "Nepal": dict(
        under5={2000: 55_000, 2005: 40_000, 2010: 26_000, 2015: 16_000, 2020: 10_500, 2024: 8_000},
        vad_year=2000, vad_base=0.31, vad_decline=0.030,
        vas={2000: 0.70, 2005: 0.85, 2010: 0.89, 2015: 0.87, 2020: 0.83, 2024: 0.80},
        rice_kg=120.0, domestic=0.90, deploy=2009, rice_eating_vad=0.65),
}

# ============================================================================
# 3. HELPER FUNCTIONS
# ============================================================================
def value_for_year(anchor_points, year):
    """Linear interpolation between anchor years; flat outside the range.
    e.g. if we know deaths in 2020 and 2024, estimate 2022 as the midpoint."""
    years = sorted(anchor_points)
    if year <= years[0]:  return float(anchor_points[years[0]])
    if year >= years[-1]: return float(anchor_points[years[-1]])
    for earlier_year, later_year in zip(years, years[1:]):
        if earlier_year <= year <= later_year:
            how_far_between = (year - earlier_year) / (later_year - earlier_year)
            return anchor_points[earlier_year] + how_far_between * (anchor_points[later_year] - anchor_points[earlier_year])
    return float(anchor_points[years[-1]])

def vitamin_a_delivered_per_day(rice_kg_per_year):
    """µg of vitamin A (RAE) a young child gets per day from Golden Rice."""
    rice_grams_per_person_per_day = rice_kg_per_year * 1000.0 / 365.0   # kg/yr -> g/day (×1000 is kg->g)
    rice_grams_a_child_eats       = rice_grams_per_person_per_day * CHILD_RICE_FRACTION
    beta_carotene_micrograms      = (rice_grams_a_child_eats
                                     * BETA_CAROTENE_UG_PER_G
                                     * STORAGE_RETENTION
                                     * COOKING_RETENTION)
    return beta_carotene_micrograms / BIOCONVERSION

def fraction_of_daily_need_met(rice_kg_per_year):
    """How much of a child's daily vitamin A requirement Golden Rice fills (capped at 100%)."""
    return min(vitamin_a_delivered_per_day(rice_kg_per_year) / CHILD_RDA_UG, 1.0)

def share_of_deaths_caused_by_vad(vad_rate):
    """Of all under-5 deaths, the fraction attributable to vitamin A deficiency.
    Standard epidemiology formula (the 'population attributable fraction')."""
    extra_risk = vad_rate * (RELATIVE_RISK_VAD - 1.0)
    return extra_risk / (1.0 + extra_risk)

def vad_rate_for_year(country, year):
    """Share of children who are vitamin-A-deficient in a given year (declines over time)."""
    rate = country["vad_base"] * ((1.0 - country["vad_decline"]) ** (year - country["vad_year"]))
    return max(rate, 0.02)   # floor: 2% residual in hard-to-reach populations

def adoption_s_curve(years_since_launch):
    """Share of the reachable crop that has switched to Golden Rice, 0..1.
    Slow at first, fast in the middle, leveling off — how new seeds really spread."""
    if years_since_launch <= 0:
        return 0.0
    return 1.0 / (1.0 + math.exp(-ADOPTION_STEEPNESS * (years_since_launch - ADOPTION_MIDPOINT_YEARS)))

# ============================================================================
# PART A — vitamin A per day & % of daily need  (the table's middle columns)
# ============================================================================
print("=" * 70)
print("PART A — Vitamin A delivered per child per day")
print("=" * 70)
print(f"{'Country':12} {'rice kg/yr':>10} {'µg RAE/day':>11} {'% of 400µg':>11}")
for country_name, country in COUNTRIES.items():
    vitamin_a            = vitamin_a_delivered_per_day(country["rice_kg"])
    percent_of_daily_need = vitamin_a / CHILD_RDA_UG * 100
    print(f"{country_name:12} {country['rice_kg']:>10.0f} {vitamin_a:>11.0f} {percent_of_daily_need:>10.0f}%")

# ============================================================================
# PART B — deaths prevented per year IF ALL DOMESTIC RICE WERE GOLDEN RICE
# ============================================================================
# This is the table's right-hand column. It is a "today" steady-state number:
#   - use the most recent year (2024) deaths and prevalence
#   - assume FULL reach: every domestically grown grain is Golden Rice, so the
#     adoption term = domestic_rice_fraction × rice_eating_vad_fraction
#     (NOT the 0.70 S-curve ceiling — this is the theoretical maximum)
print()
print("=" * 70)
print("PART B — Deaths prevented / year at 100% domestic adoption (2024)")
print("=" * 70)
YEAR = 2024
print(f"{'Country':12} {'deaths/yr':>10}")
total_deaths_prevented_per_year = 0.0
for country_name, country in COUNTRIES.items():
    under5_deaths_this_year         = value_for_year(country["under5"], YEAR)
    vad_rate_this_year              = vad_rate_for_year(country, YEAR)
    deaths_caused_by_vad            = under5_deaths_this_year * share_of_deaths_caused_by_vad(vad_rate_this_year)
    share_protected_by_supplements  = value_for_year(country["vas"], YEAR) * EFFECTIVE_VAS_MULTIPLIER
    deaths_not_prevented_by_supps   = deaths_caused_by_vad * (1.0 - share_protected_by_supplements)
    share_reachable_by_golden_rice  = country["domestic"] * country["rice_eating_vad"]   # full reach, no 0.70 ceiling
    golden_rice_effectiveness       = fraction_of_daily_need_met(country["rice_kg"]) ** DOSE_RESPONSE_CONCAVITY
    lives_saved_per_year            = (deaths_not_prevented_by_supps
                                       * share_reachable_by_golden_rice
                                       * golden_rice_effectiveness
                                       * MORTALITY_EFFECT_CENTRAL)   # 18% mortality midpoint
    total_deaths_prevented_per_year += lives_saved_per_year
    print(f"{country_name:12} {lives_saved_per_year:>10,.0f}")
print(f"{'TOTAL':12} {total_deaths_prevented_per_year:>10,.0f}")

# ============================================================================
# PART C — cumulative lives lost 2006–2024  (the realistic S-curve scenario)
# ============================================================================
# Same per-year logic as PART B, but now adoption ramps up along the S-curve
# starting from each country's launch year (capped at 0.70 × domestic × rice_eating),
# and we add up every single year from 2000 to 2024.
print()
print("=" * 70)
print("PART C — Cumulative children's lives lost, realistic adoption")
print("=" * 70)
print(f"{'Country':12} {'cumulative':>12}")
total_lives_lost = 0.0
for country_name, country in COUNTRIES.items():
    max_adoption              = ADOPTION_CEILING * country["domestic"] * country["rice_eating_vad"]
    golden_rice_effectiveness = fraction_of_daily_need_met(country["rice_kg"]) ** DOSE_RESPONSE_CONCAVITY
    country_total = 0.0
    for year in range(MODEL_START_YEAR, MODEL_END_YEAR + 1):
        under5_deaths_this_year        = value_for_year(country["under5"], year)
        vad_rate_this_year             = vad_rate_for_year(country, year)
        deaths_caused_by_vad           = under5_deaths_this_year * share_of_deaths_caused_by_vad(vad_rate_this_year)
        share_protected_by_supplements = value_for_year(country["vas"], year) * EFFECTIVE_VAS_MULTIPLIER
        deaths_not_prevented_by_supps  = deaths_caused_by_vad * (1.0 - share_protected_by_supplements)
        adoption_this_year             = adoption_s_curve(year - country["deploy"]) * max_adoption
        country_total += deaths_not_prevented_by_supps * adoption_this_year * golden_rice_effectiveness
    total_lives_lost += country_total
    print(f"{country_name:12} {country_total * MORTALITY_EFFECT_CENTRAL:>12,.0f}")   # 18% midpoint
print(f"{'TOTAL':12} {total_lives_lost * MORTALITY_EFFECT_CENTRAL:>12,.0f}")

# ============================================================================
# PART D — blindness and years of healthy life lost
# ============================================================================
print()
print("=" * 70)
print("PART D — Death range, blindness, and healthy-life-years lost")
print("=" * 70)
children_dead_central = total_lives_lost * MORTALITY_EFFECT_CENTRAL  # 18% midpoint (headline)
children_dead_floor   = total_lives_lost * MORTALITY_EFFECT_FLOOR    # 12% (low end of range)
children_dead_high    = total_lives_lost * MORTALITY_EFFECT_HIGH     # 24% (high end of range)
children_blinded_low   = children_dead_central * BLINDNESS_LOW_MULT
children_blinded_high  = children_dead_central * BLINDNESS_HIGH_MULT
healthy_years_lost_low  = children_dead_central * YEARS_LOST_PER_DEATH + children_blinded_low  * QALYS_PER_BLIND_CHILD
healthy_years_lost_high = children_dead_central * YEARS_LOST_PER_DEATH + children_blinded_high * QALYS_PER_BLIND_CHILD
print(f"Children dead (central, 18% effect): {children_dead_central:>13,.0f}")
print(f"Children dead range (12% – 24%):     {children_dead_floor:>13,.0f}  –  {children_dead_high:,.0f}")
print(f"Children blinded (2–4x of central):  {children_blinded_low:>13,.0f}  –  {children_blinded_high:,.0f}")
print(f"Healthy-life-years lost:             {healthy_years_lost_low:>13,.0f}  –  {healthy_years_lost_high:,.0f}")
print()
print("The range spans GiveWell's reading of the trials (12% gives full weight to DEVTA,")
print("the largest trial; 24% is the random-effects figure). Blindness uses the central.")

Biggest uncertainties

How much beta-carotene is in the rice?

Field tests show 3.57 µg/g in Philippine rice, roughly 9 in Bangladesh.[3][4] This model uses the midpoint, 6.3. Higher figures sometimes quoted (20–37) are the lab donor line, not the rice farmers would grow.

How much survives storage?

Beta-carotene degrades: Schaub 2017 found ~60% left after 3 weeks, ~13% after 10.[13] This model assumes 65% retention — i.e. shorter storage. IRRI is breeding lower-degradation lines.

Does vitamin A actually save lives?

This drives the death range. The old consensus was a 24% cut in child mortality; then DEVTA, the largest trial, found almost none.[31] Pooling the trials gives 12–24% (GiveWell's range). The headline (106,000) is the midpoint; the range 71,000–141,000 spans the two readings. I didn't use DEVTA's 4% alone — no single trial should be.

References

  1. Stevens GA, Bennett JE, Hennocq Q, et al. "Trends and mortality effects of vitamin A deficiency in children in 138 low-income and middle-income countries between 1991 and 2013." Lancet Global Health 3: e528–36 (2015). thelancet.com
  2. UN Inter-agency Group for Child Mortality Estimation (IGME). UNICEF CME data. data.unicef.org
  3. Swamy BPM, Samia M, Boncodin R, et al. "Compositional Analysis of Genetically Engineered GR2E 'Golden Rice' in Comparison to That of Conventional Rice." J. Agricultural and Food Chemistry 67(28): 7986–7994 (2019). Mean all-trans-beta-carotene in milled GR2E: 3.57 µg/g (range 1.96–7.31).
  4. Biswas S, Swamy BPM, et al. "Development and Field Evaluation of Near-Isogenic Lines of GR2-E BRRI dhan29 Golden Rice." Frontiers in Plant Science (2021). BRRI dhan29 background: 8.5–12.5 µg/g total carotenoids (~80–90% beta-carotene). ncbi.nlm.nih.gov
  5. Food Standards Australia New Zealand (FSANZ). Golden Rice approval. December 2017 / February 2018. See also: IRRI, "Golden Rice meets food safety standards in three global leading regulatory agencies." irri.org
  6. Health Canada. Golden Rice approval. March 16, 2018. ISAAA, "Golden Rice Gets Approval from Health Canada." isaaa.org
  7. US FDA. Golden Rice consultation completed. May 2018. ISAAA, "Golden Rice Gets Approval from US FDA." isaaa.org
  8. Philippines approved Golden Rice for food/feed (Dec 2019) and commercial propagation (July 2021), becoming the first country to do so. Farmers grew it in 2022. See: "For the first time, farmers in the Philippines cultivated Golden Rice on a larger scale and harvested almost 70 tons." phys.org
  9. De Guzman C. "What a Philippine court ruling means for transgenic Golden Rice." Science (2024). science.org. See also: ELAW, "Court Blocks Commercial Propagation of GMO Rice and Eggplant." elaw.org
  10. European Scientist. "Golden Rice finally on track for approval in Bangladesh." europeanscientist.com. See also: Science, "Bangladesh could be the first to cultivate Golden Rice." science.org
  11. UNICEF/WHO joint estimates of vitamin A supplementation coverage. data.unicef.org
  12. India vitamin A supplementation coverage: 60.5% of children aged 9–59 months (NFHS-4, 2015–16) — about 2 in 5 missed. (Coverage of the younger 9–35-month band rose to 71.2% by NFHS-5, 2019–21.) The same analysis found the supplementation–mortality link disappeared after adjusting for confounders. "Coverage of vitamin A supplementation among under-five children in India," British Journal of Nutrition, cambridge.org; and BMJ Global Health 7(6):e007972 (2022), ncbi.nlm.nih.gov.
  13. Schaub P, Wüst F, Koschmieder J, et al. "Nonenzymatic β-Carotene Degradation in Provitamin A-Biofortified Crop Plants." Journal of Agricultural and Food Chemistry 65(31): 6588–6598 (2017). Golden Rice retained ~60% of beta-carotene at 3 weeks and ~13% at 10 weeks; half-life ~25 days, then a plateau. pubmed
  14. IRRI. Standard cooking retention estimates for provitamin A in rice. ~20–25% loss during boiling/steaming.
  15. Tang G, Qin J, Dolnikowski GG, Russell RM, Grusak MA. "Golden Rice is an effective source of vitamin A." American Journal of Clinical Nutrition 89(6): 1776–1783 (2009). Bioconversion ratio: 3.8 ± 1.7:1 in healthy adults.
  16. Estimated. Children under 5 consume approximately 25–35% of adult rice portions; 30% used as central value. Consistent with NSSO household consumption data and FAO food balance approaches.
  17. WHO/FAO. "Vitamin and Mineral Requirements in Human Nutrition" (2002). Recommended dietary allowance for children 4–8 years: 400 µg RAE/day. Bioconversion standard: 12 µg dietary beta-carotene = 1 µg RAE.
  18. Tang G, Hu Y, Yin S, et al. "Beta-carotene in Golden Rice is as good as beta-carotene in oil at providing vitamin A to children." Am. J. Clin. Nutr. 96(3): 658–664 (2012). Retracted (July 2015) due to informed consent violations; scientific findings were not challenged in the retraction notice.
  19. India NFHS-2 (National Family Health Survey, 1998–99). VAD prevalence ~57% in children. See also: CNNS 2016–2018 finding 17.54% with updated methodology.
  20. FAO. "Rice in Human Nutrition." Food and Agriculture Organization, Table 8. fao.org
  21. Maziya-Dixon B, et al. "Vitamin A Deficiency Is Prevalent in Children Less Than 5 y of Age in Nigeria." Journal of Nutrition 136(8): 2255–2261 (2006). VAD prevalence: 29.5% nationally. jn.nutrition.org
  22. Bangladesh National Micronutrient Survey 2011–2012. Subclinical VAD prevalence: 20.5%. See also: Cambridge Core, "Vitamin A deficiency and determinants in Bangladeshi children." cambridge.org
  23. Philippines Food and Nutrition Research Institute (FNRI). National Nutrition Survey 2003. Subclinical VAD: ~38%.
  24. Philstar. "Rice self-sufficiency ratio recovers to 78.5% in 2023." (Dec 2024). philstar.com
  25. Tanzania Demographic and Health Survey (TDHS), 2004. VAD prevalence ~33%.
  26. Nepal Demographic and Health Survey, 2001. VAD prevalence ~31%.
  27. Paine JA, Shipton CA, Chaggar S, et al. "Improving the nutritional value of Golden Rice through increased pro-vitamin A content." Nature Biotechnology 23: 482–487 (2005). nature.com
  28. Laborte AG, et al. "Farmers' Preference for Rice Traits: Insights from Farm Surveys in Central Luzon, Philippines, 1966–2012." PLOS ONE (2015). plosone.org. IR8 adoption: 3% (1966) to 75% (1980) in Philippines. See also: FAO, "From the Green Revolution to the Gene Revolution." fao.org
  29. World Health Organization. "Vitamin A Deficiency." WHO Global Database on Vitamin A Deficiency. who.int
  30. Sommer A, Tarwotjo I, Hussaini G, Susanto D. "Increased mortality in children with mild vitamin A deficiency." Lancet (1983). West KP et al. "Efficacy of vitamin A in reducing preschool child mortality in Nepal." Lancet (1991). Fawzi WW et al. (1993). Relative risk range: 1.4–2.2; central: 1.75.
  31. Imdad A, et al. "Vitamin A supplementation for preventing morbidity and mortality in children from six months to five years of age." Cochrane Database of Systematic Reviews (2022). VAS reduces all-cause mortality by 24%.