A data-science study of how visible women's economic activity is: measurement model, uncertainty, sensitivity, scenarios and value of information - Saudi Arabia

This paper studies how visible women’s economic activity is in Saudi Arabia, one of the 19 sovereign G20  members scored in the series, using the Gender Economic Visibility Index (GEVI). The index converts 13  indicators into parity scores, averages them into five dimensions (labour participation, job quality and  seniority, enterprise and legal rights, institutional visibility, and statistical and political voice) and combines  the dimensions with a weighted geometric mean, so that a strong dimension cannot fully compensate for a  weak one. Every input carries an evidence grade that sets its uncertainty. Saudi Arabia scores 51.8 out of  100, in the Partial band, with a 90% uncertainty interval of 48.1 to 56.1 from 10,000 joint Monte Carlo draws  (median 51.8) and a point rank of 18 of 19 (rank median 18; 5th to 95th percentile range 18 to 18). The  strongest dimension is D2 Job quality and seniority (0.900); the weakest is D4 Institutional visibility (0.360).  A Shapley attribution of the 48.2-point gap to full parity assigns the largest share, 14.5 points (30%), to D1  Labour participation. Of the 13 inputs, 12 are grade A and one (senior and middle management) is grade C,  imputed from the series median, giving a blind-spot ratio of 0.08; weight uncertainty explains 73.5% of  simulated output variance and the imputed input 3.3%. Rank stays at 18 across the seven weighting  schemes and the ten alternative specifications (benchmark, parity cap and imputation rule), with scores  moving by at most 3.5 points. The largest single-lever scenario is halving the gap to parity in labour access  (S4, +6.6 points), and a greedy pathway of +0.10 steps reaches the 65-point threshold after fifteen steps.  These pathways order priorities; they are not forecasts. The index contains no direct measure of women’s  visibility in taxpayer records, so the paper proposes adding a sex field to wage-tax and taxpayer records and  publishing an annual gender tax profile; this would extend the model and is not part of the current score.  The priority data actions are to publish an annual appointments series by sex (I11), record owner sex on  business registrations (I7) and diagnose, from survey microdata, whether the youth unemployment gap is  demand-side or a matching problem (I3). The weights and uncertainty settings are judgemental  assumptions, the study is cross-sectional, and the scenarios describe index arithmetic rather than policy  effects. All results are reproducible from the accompanying Python scripts with a fixed random seed of 2026;  scores and intervals are research estimates, not official statistics.