← Back to ResearchGEVI Research Series · G20 Countries · Paper 5

ChinaEconomic Visibility

Shape. The strongest dimension is D2 Job quality and seniority (0.915); the weakest is D4 Institutional visibility (0.428). A Shapley attribution of the 26.4-point gap to full parity assigns the largest share, 10.9 points (41%), to D4 Institutional visibility.

73.6GEVI / 100 · Emerging band

Evidence. Of 13 inputs, 11 are grade A, 1 is grade B and 1 is grade C (imputed), a blind-spot ratio of 0.08. The imputed inputs are placeholders and widen the interval.

AuthorSujoy Maitra

0009-0009-1319-3706

PaperMCWP-2026-57

Publication Date29 Sept 2026

INSTITUTIONAL VISIBILITYCHINA

D4 ACCOUNTS FOR 41.4% OF THE PARITY GAP

GEVI score73.6

Rank12 / 19

Blind-spot ratio0.08

Reported research measures. Animation reveals the graphic; it does not represent change over time.

Abstract

Result. China scores 73.6 out of 100 on the Gender Economic Visibility Index, in the Emerging band. The 90% uncertainty interval from 10,000 joint Monte Carlo draws is 68.7 to 77.5 (median 73.3). Its point rank is 12 of 19, and across the same draws the rank has a median of 12 with a 5th to 95th percentile range of 10 to 13.

Evidence. Of 13 inputs, 11 are grade A, 1 is grade B and 1 is grade C (imputed), a blind-spot ratio of 0.08. The imputed inputs are placeholders and widen the interval.

Download full paper PDF

Inputs and parity calculations

Inputs and parity scores

ID table
ID	Indicator	Raw	Score	Grade	SD
I1	Labour force participation	women 59.1, men 69.9 (2025)	0.8455	A	0.02
I2	Youth NEET	women 14.5, men 10.9 (2000)	0.7517	B	0.06
I3	Youth unemployment	women 14.2, men 17.1 (2025)	1.0000	A	0.02
I4	Non-vulnerable employment	women 32.7, men 34.0 (2025)	1.0000	A	0.02
I5	Senior and middle management	not available	0.7460	C	0.12
I6	Wage and salaried employment	women 64.6, men 60.4 (2025)	1.0000	A	0.02
I7	Firm ownership	44.0% (2024)	0.8800	A	0.02
I8	WBL legal frameworks	70.0 (2025)	0.7000	A	0.02
I9	WBL enforcement	63 (2026)	0.6300	A	0.02
I10	WBL supportive frameworks	69 (2026)	0.6900	A	0.02
I11	Women ministers	8.33% (2024)	0.1666	A	0.02
I12	Women MPs	26.5% (2026)	0.5300	A	0.02
I13	Data availability	11 of 12 rows	0.9167	A	0.02

The labour, management, enterprise and legal inputs (I1 to I11) come from the World Bank Gender Data Landscape for this economy (April 2026 edition, including Women, Business and the Law 2026). The parliamentary share (I12) comes from the Inter-Parliamentary Union monthly ranking as of 1 August 2026.

Step 1: Indicator calculations

GEVI has three layers: 13 indicators, 5 dimensions and one composite. Each indicator is converted to a parity score, each dimension is the simple mean of its indicators, and the composite is a weighted geometric mean of the dimensions. Three principles govern the design: parity scores avoid arbitrary goalposts; substitutability is allowed within a dimension but not across dimensions; and every input carries an evidence grade that sets its uncertainty.

ID table
ID	Calculation	Score
I1	min(59.1 / 69.9, 1) = 0.845	0.845
I2	min(10.9 / 14.5, 1) = 0.752	0.752
I3	min(17.1 / 14.2, 1) = 1.000	1.000
I4	min((100 - 32.7) / (100 - 34.0), 1) = min(67.3 / 66.0, 1) = 1.000	1.000
I5	imputed series median 0.746	0.746
I6	min(64.6 / 60.4, 1) = 1.000	1.000
I7	min(44.0 / 50, 1) = 0.880	0.880
I8	70.0 / 100 = 0.700	0.700
I9	63 / 100 = 0.630	0.630
I10	69 / 100 = 0.690	0.690
I11	min(8.33 / 50, 1) = 0.167	0.167
I12	min(26.5 / 50, 1) = 0.530	0.530
I13	11 / 12 = 0.917	0.917

Step 2: Dimension scores

Dim table
Dim	Calculation	Score
D1	0.845 + 0.752 + 1.000 = 2.597; / 3	0.8657
D2	1.000 + 0.746 + 1.000 = 2.746; / 3	0.9153
D3	0.880 + 0.700 + 0.630 = 2.210; / 3	0.7367
D4	0.690 + 0.167 = 0.857; / 2	0.4283
D5	0.530 + 0.917 = 1.447; / 2	0.7233

The base weights (0.25, 0.20, 0.20, 0.15, 0.20) reflect the emphasis on labour participation and on statistical voice. They are a prior, not an estimate, which is why the paper tests seven weighting schemes, including data-driven entropy and principal-component weights computed from the 15-economy dimension matrix.

Aggregation, uncertainty and sensitivity

Step 3: Aggregation trace

Term table
Term	Calculation	Value
D1	0.8657 ^ 0.25	0.96460
D2	0.9153 ^ 0.20	0.98246
D3	0.7367 ^ 0.20	0.94071
D4	0.4283 ^ 0.15	0.88057
D5	0.7233 ^ 0.20	0.93728
Product	0.9646 x 0.9825 x 0.9407 x 0.8806 x 0.9373	0.73578
GEVI	100 x product	73.58

Check in logs: sum of w ln D = -0.30683; exp = 0.73578; times 100 = 73.58.

Step 4: Uncertainty and sensitivity

For each of 10,000 draws the model samples all 13 indicator scores from truncated normal distributions and the five dimension weights from a Dirichlet distribution, then recomputes the index. The same seed (2026) is used for every economy, which means all economies share the same weight draws. That is intentional: weight uncertainty is common to the whole series, so rank comparisons are made under identical weights in each draw.

Each input has a grade with a standard deviation applied to its score: A = 0.02, B = 0.06, C = 0.12. The standard deviations are judgemental settings that express how much a reader should trust a value; they are not estimated from replications. Section 7.5 shows the effect of changing the settings that matter most.

Robustness. Across seven weighting schemes the rank runs from 10 to 14; across ten alternative specifications (benchmark, parity cap and imputation rule) it runs from 11 to 13. The score changes by at most 6.6 points across those specifications.

p1 table
p1	p5	p10	p25	p50	p75	p90	p95	p99	Mean	SD
66.63	68.75	69.78	71.51	73.31	75.07	76.66	77.49	78.96	73.24	2.67

Weight uncertainty explains 71.7% of the output variance; the grade C (imputed) inputs explain 5.6%. Weights and measured inputs, not imputation, drive the spread.

Gaps, scenarios and priorities

Step 5: Where the gap is

The gap to full parity is 26.4 points. Because the geometric mean is non-additive, the contribution of one dimension depends on the state of the others. Shapley values average each dimension’s marginal contribution over all 5! orderings of the dimensions, so they sum exactly to the gap. v(S) is the GEVI obtained when the dimensions in S are set to 1.

Dim table
Dim	Shapley points	Share of gap	Shortfall of score (x100)
D4	10.94	41.4%	57.2
D5	5.58	21.1%	27.7
D3	5.27	19.9%	26.3
D1	3.11	11.8%	13.4
D2	1.53	5.8%	8.5

Scenario simulation and pathways

Levers. The largest single-lever scenario among S1 to S6 is "S2 Institutions: supportive frameworks 50%, ministers 15%", worth +1.6 points. A greedy pathway of +0.10 steps on measured indicators reaches the next threshold, 80 points, after 7 steps; at an illustrative improvement of 0.02 a year on every measured indicator it takes 5 years.

Scenario table
Scenario	GEVI	Gain	p5	Median	p95
S1 Data coverage to 9 of 12 rows	73.6	+0.0	68.7	73.3	77.4
S2 Institutions: supportive frameworks 50%, ministers 15%	75.2	+1.6	71.1	74.8	78.4
S3 Legal: legal and enforcement scores +0.10	74.9	+1.3	69.9	74.6	78.8
S4 Labour access: halve the gap to parity	75.0	+1.4	69.8	74.7	79.1
S5 Politics: women 15% of parliament	73.6	+0.0	68.7	73.3	77.4
S6 Enterprise: management and firm ownership to 0.75	73.6	+0.0	68.7	73.3	77.4
S7 Combined S1 to S6	77.9	+4.4	73.4	77.6	81.5
S8 Series best-in-class on every measured indicator	95.1	+21.5	91.9	93.8	95.4

The pathway adds 0.10 to one measured indicator at a time, always choosing the step that raises GEVI most, until the composite reaches the next threshold above the current score (50, 65, 80, 90 or 95) or no measured indicator can rise further. It is an ordering of priorities, not a forecast.

Step table
Step	ID	Indicator	Score change	GEVI after
1	I10	WBL supportive frameworks	0.69 to 0.79	74.8
2	I10	WBL supportive frameworks	0.79 to 0.89	75.9
3	I10	WBL supportive frameworks	0.89 to 0.99	77.0
4	I12	Women MPs	0.53 to 0.63	78.0
5	I12	Women MPs	0.63 to 0.73	79.0
6	I11	Women ministers	0.17 to 0.27	80.0
7	I12	Women MPs	0.73 to 0.83	80.9

Priorities

I11 Women ministers. Publish an annual appointments series by sex for ministerial and board posts.

I12 Women MPs. Treat parliamentary representation as a policy question for parliament and parties; a revenue authority can supply the data but not the remedy.

I9 WBL enforcement. Publish complaint, prosecution and remedy data so that enforcement can be measured, not only perceived.

Restore a fiscal indicator. Add a sex field to wage-tax and taxpayer records and publish an annual gender tax profile, so that D4 can include a direct measure of visibility in the tax system.

Practical use and data strategy

How to use this data

Data product table
Data product	Primary user	Decision it informs	Cadence
Ministerial and board appointments by sex	Cabinet office, public service commission	A measured record of decision-making representation	Annual
Candidate and elected members by sex by party	Parliament, electoral commission	Baseline for any special-measures discussion	After each election
Enforcement statistics linked to labour and equality law	Justice ministry, labour inspectorate	Where the gap between law and practice arises	Annual
Sex-disaggregated taxpayer and wage-tax statistics	Revenue authority, treasury	Who is in the formal tax base and who receives refunds and credits	Annual

China’s use case is to turn parity gaps into named data products with named owners, so that each year’s re-run of the index shows which product moved which indicator.

Data strategy

Data product table
Data product	Owner	Indicators	Effect
Sex field on wage-tax returns and taxpayer registrations	Revenue authority	Adds a fiscal indicator (I14, proposed)	Restores the fiscal indicator the index was designed to have
Keep the core 12-row table current and add a fiscal (taxpayer) row	National statistics office and revenue authority	I13 and a proposed I14	Coverage is already 11 of 12; the gain is a new fiscal indicator, not more rows

Scope and limitations

Weights are a prior. The weights express a view, not an estimate. Chapter 7 shows their effect, but no data can identify the right weights.

Uncertainty settings are judgemental. The standard deviations 0.02, 0.06 and 0.12 are settings, not measurements. If they were larger the intervals would widen.

Source heterogeneity. Inputs come from different years and sources; older values carry a wider standard deviation, but a level shift caused by definitions is not modelled.

No causal claim. The scenarios show what the index would do if an indicator changed. They do not show that a policy would change the indicator.

Fiscal indicator absent. No sex-disaggregated taxpayer statistic was found, so the index does not yet measure visibility in the tax system directly.