Benchmark Results

Overview

The GRI has been benchmarked against five major survey programs with different design philosophies, geographic scopes, and sample sizes:

Global Dialogues — A longitudinal survey series capturing public perceptions of AI, deliberately designed for global demographic representation (~1,000 participants per wave from 50+ countries).
World Values Survey — The largest global social survey program, running since 1981, with ~58,000 participants per wave across 60–100 countries.
Pew Global Attitudes — Pew Research Center’s annual international survey, with ~1,000 respondents per country across 35 countries in Spring 2024.
Afrobarometer — A pan-African survey network covering 39 countries, with ~53,000 respondents in Round 9 (2021–2023).
Latinobarómetro — A Latin American public opinion survey covering 17 countries, with ~19,000 respondents per wave.

Global vs. regional benchmarking

Global Dialogues and WVS claim to represent the global population, so their GRI is calculated against global demographic benchmarks. Regional surveys—Afrobarometer, Latinobarómetro, and Pew—are scored against the populations of the countries they actually survey. A GRI of 0.80 means “80% representative of the population you claim to cover,” making scores comparable across programs regardless of geographic scope.

Cross-Survey Comparison

Five survey programs compared on representativeness and statistical precision.

GRI Scores (Key Dimensions)

GD and WVS are averages across waves. Latinobarómetro is averaged across 2023–2024.
Dimension	GD (global)	WVS (global)	Pew (35 countries)	Afrobarometer (39 African)	Latinobarómetro (17 LatAm)
Country x Gender x Age	0.34	0.20	0.48	0.53	0.48
Country x Religion	0.50	0.30	0.53	0.62	0.52
Country x Environment	0.42	0.32	0.55	0.63	0.51
Country	0.60	0.31	0.56	0.63	0.53
Religion	0.82	0.75	0.72	0.91	0.87
Environment	0.63	0.85	0.78	0.94	0.92
Age Group	0.71	0.83	0.93	0.91	0.97
Gender	0.99	0.98	0.99	0.99	0.99
Overall (13 dim. avg)	0.64	0.55	0.69	0.81	0.78

Precision Retained (Key Dimensions)

Fraction of nominal sample size that contributes to inferential precision after reweighting.
Dimension	GD (global)	WVS (global)	Pew (35 countries)	Afrobarometer (39 African)	Latinobarómetro (17 LatAm)
Country x Gender x Age	34%	43%	52%	50%	40%
Country x Religion	30%	10%	42%	40%	31%
Country x Environment	10%	22%	43%	45%	23%
Country	47%	25%	40%	42%	33%
Religion	74%	49%	52%	95%	91%
Gender	100%	100%	100%	100%	100%

Key Takeaways

Deliberate design outperforms convenience sampling. GD achieves 40–70% higher GRI on intersectional dimensions with ~1,000 participants vs. WVS’s ~58,000.
Regional surveys excel within their scope. Afrobarometer and Latinobarómetro achieve GRI > 0.78 against their claimed populations, demonstrating strong representativeness of the regions they target.
Equal-per-country designs sacrifice country-level representativeness. Pew and WVS both allocate ~1,000 respondents per country regardless of population, creating large design effects at the country level.
Intersectional precision is universally expensive. No survey retains more than 52% of its nominal precision on Country x Gender x Age—the combinatorial challenge of matching multiple distributions simultaneously is inherent.
Afrobarometer achieves the highest overall GRI (0.81) partly because its 39-country scope represents a more homogeneous religious landscape than the global population, making religion matching easier.

Global Dialogues (GD1–GD8)

Wave	Participants	Period
GD1	1,280	Wave 1
GD2	1,105	Wave 2
GD3	971	Wave 3
GD4	1,050	Wave 4
GD5	1,057	Wave 5
GD6	1,037	Wave 6
GD8	904	Wave 8

Primary Dimensions

Dimension	GD1	GD2	GD3	GD4	GD5	GD6	GD8
Country x Gender x Age	0.337	0.329	0.348	0.361	0.347	0.341	0.296
Country x Religion	0.482	0.488	0.514	0.535	0.499	0.495	0.496
Country x Environment	0.424	0.394	0.435	0.453	0.414	0.406	0.402

Regional Dimensions

Dimension	GD1	GD2	GD3	GD4	GD5	GD6	GD8
Region x Gender x Age	0.544	0.543	0.570	0.577	0.563	0.559	0.529
Region x Religion	0.597	0.587	0.632	0.647	0.609	0.621	0.625
Region x Environment	0.538	0.507	0.562	0.576	0.520	0.518	0.522
Region	0.745	0.739	0.791	0.799	0.738	0.734	0.769

Single-Axis Dimensions

Dimension	GD1	GD2	GD3	GD4	GD5	GD6	GD8
Country	0.590	0.573	0.618	0.639	0.592	0.587	0.581
Continent	0.832	0.830	0.886	0.883	0.773	0.802	0.856
Religion	0.817	0.819	0.833	0.826	0.813	0.806	0.840
Environment	0.629	0.623	0.642	0.628	0.635	0.620	0.636
Age Group	0.656	0.684	0.706	0.723	0.746	0.756	0.699
Gender	0.989	0.990	0.996	0.979	0.986	0.995	0.998

Design Effect and Precision Cost (GD Averages)

Averages across GD1–GD8 (N ≈ 1,000 per wave).
Dimension	GRI	Design Effect	Effective N	Precision Retained
Country x Gender x Age	0.34	3.0	189	34%
Country x Religion	0.50	4.1	241	29%
Country x Environment	0.42	10.9	102	10%
Continent	0.83	1.1	965	89%
Religion	0.82	75	–	–
Gender	0.99	1.0	1,083	100%

World Values Survey (W1–W7)

Wave	Participants	Period
W1	9,144	1981–1984
W2	24,553	1990–1994
W3	77,761	1995–1998
W4	60,012	1999–2004
W5	83,879	2005–2009
W6	85,219	2010–2014
W7	62,927	2017–2022

Design Effect and Precision Cost (WVS Averages)

Averages across W1–W7 (N ≈ 58,000 per wave).
Dimension	GRI	Design Effect	Effective N	Precision Retained
Country x Gender x Age	0.20	2.7	5,695	43%
Country x Religion	0.30	29.4	3,634	10%
Country x Environment	0.32	45.4	5,832	22%
Continent	0.67	1.6	40,371	66%
Religion	0.75	–	–	–
Gender	0.98	1.0	57,550	100%

Pew Global Attitudes (Spring 2024)

Multi-country survey (35 countries)

Pew Spring 2024 covers 35 countries across all continents. GRI is calculated against the combined population of these 35 countries, not the full global population. Urbanicity data is only available for face-to-face interview countries (20 of 35).

Participants	41,483
Countries	35 (phone, face-to-face, and online modes)
Period	January–May 2024
Benchmark	35-country combined population

GRI Scores by Dimension

Benchmarked against 35-country population.
Dimension	GRI	Diversity	Design Effect	Precision Retained
Country x Gender x Age	0.476	0.996	1.9	52%
Country x Religion	0.534	0.806	2.4	42%
Country x Environment	0.547	0.524	2.3	43%
Country	0.560	1.000	2.5	40%
Region x Gender x Age	0.686	0.994	2.4	41%
Region x Religion	0.611	0.714	2.0	49%
Region x Environment	0.693	0.625	1.3	76%
Region	0.683	1.000	2.0	50%
Continent	0.745	1.000	1.3	77%
Religion	0.715	0.857	1.9	52%
Environment	0.776	1.000	1.0	100%
Age Group	0.931	1.000	1.0	97%
Gender	0.992	1.000	1.0	100%
Overall	0.688	0.886

Notes:

Country x Environment diversity is low (0.524) because urbanicity is only recorded for face-to-face interview countries. 15 phone/online countries have no urban/rural data, creating 29% uncovered population in that dimension.
Equal-per-country sampling (~1,000 per country) means India (1.4B) and Singapore (5.8M) receive equal representation, producing a Country design effect of 2.5.
Religion coverage: 94% of respondents have valid religion data. 6 of 7 major GRI categories are represented (no Folk Religions category).

Afrobarometer Round 9

Regional survey (39 African countries)

Afrobarometer claims to represent the population of 39 African countries. GRI is calculated against that 39-country population, not the global population. Scoring against global benchmarks would yield an Overall GRI of only ~0.35 (since Africa is ~17% of the world population), which would be misleading.

Participants	53,444
Countries	39 African countries
Period	2021–2023
Benchmark	39-country combined population

GRI Scores by Dimension

Benchmarked against 39-country African population.
Dimension	GRI	Diversity	Design Effect	Precision Retained
Country x Gender x Age	0.532	0.995	2.0	50%
Country x Religion	0.623	0.974	2.5	40%
Country x Environment	0.632	0.985	2.2	45%
Country	0.634	1.000	2.4	42%
Region x Gender x Age	0.800	1.000	1.4	71%
Region x Religion	0.835	0.771	1.5	65%
Region x Environment	0.817	1.000	1.3	77%
Region	0.850	1.000	1.3	76%
Continent	1.000	1.000	1.0	100%
Religion	0.912	0.857	1.1	95%
Environment	0.942	1.000	1.0	99%
Age Group	0.912	1.000	1.0	96%
Gender	0.995	1.000	1.0	100%
Overall	0.807	0.968

Notes:

Continent GRI = 1.000 because all surveyed countries are in Africa—the survey perfectly represents its claimed continental scope.
Religion GRI = 0.912 reflects good representation of Africa’s predominantly Christian and Muslim population distribution.
Country-level design effect = 2.4 indicates roughly equal allocation per country despite large population differences (Nigeria 220M vs. Cabo Verde 0.6M).

Latinobarómetro (2023–2024)

Regional survey (17 Latin American countries)

Latinobarómetro claims to represent 17 Latin American countries. GRI is calculated against that 17-country population. Urban/rural is derived from city size (< 10,000 = Rural).

Participants	~19,200 per wave
Countries	17 Latin American countries
Period	2023 and 2024
Benchmark	17-country combined population

GRI Scores by Dimension (Averaged Across 2023–2024)

Benchmarked against 17-country Latin American population. Averaged across 2023 and 2024 waves.
Dimension	GRI	Diversity	Design Effect	Precision Retained
Country x Gender x Age	0.481	0.993	2.5	40%
Country x Religion	0.516	0.969	3.2	31%
Country x Environment	0.506	0.971	4.5	23%
Country	0.529	1.000	3.0	33%
Region x Gender x Age	0.932	1.000	1.0	96%
Region x Religion	0.861	0.525	1.2	83%
Region x Environment	0.894	1.000	1.7	60%
Region	0.942	1.000	1.0	98%
Continent	0.787	1.000	1.2	84%
Religion	0.873	0.571	1.1	91%
Environment	0.920	1.000	1.0	93%
Age Group	0.965	1.000	1.0	99%
Gender	0.992	1.000	1.0	100%
Overall	0.784	0.925

Notes:

Region GRI = 0.942 shows excellent representation at the sub-continental level—Latin American countries cluster tightly within one UN region.
Religion diversity = 0.571 is low because Latin America is predominantly Christian (~90%)—only 4 of 7 major religious categories are represented, which is an accurate reflection of the region.
Country x Environment design effect = 4.5 suggests uneven urban/rural sampling across countries, likely due to city-size-based classification.

Sampling Efficiency

Raw GRI scores can be misleading without context. A score of 0.35 on Country x Gender x Age sounds low, but the maximum achievable score at N = 1,000 is only 0.79. Efficiency ratios normalize scores against these theoretical ceilings:

Dimension	Max GRI (N=1000)	GD Avg. Actual	Efficiency
Country x Gender x Age	0.792	0.344	43%
Country x Religion	0.938	0.502	54%
Country x Environment	0.950	0.421	44%

GD achieves 43–54% of the theoretical maximum across primary dimensions, indicating room for improvement through targeted recruitment while capturing a meaningful share of achievable representativeness.

Compare Your Survey

The GRI framework can be applied to any survey with demographic data. See the Python library documentation to calculate GRI scores, design effects, and effective sample sizes for your own data.