Loading chart...
Conditional probabilities for the supplier with the largest observed import-share change within a selected candidate set, plus the inverse-Simpson concentration of the 2022 application probabilities.
Surface: /mercator/methodology
BACI HS92; training baseline years 2015–2020, backtest 2021–2022, application 2022
Mercator ranks the retained alternative supplier with the largest year-over-year import-share change after an incumbent loses share. Training and evaluation require suppliers to be observed in both years, excluding complete incumbent exits, new entrants and disappearing alternatives.
The conditional logit uses standardized world-export capacity, prior supplier share, distance, relationship headroom, governance, FTA access and RCA. Training baseline years are 2015–2020, with outcomes in 2016–2021.
The 2022 application uses world-export-capacity candidates, differing from the retained bilateral suppliers used for estimation. Its inverse-Simpson index summarizes candidate-probability concentration, not verified replacement capacity.
Coefficient, convergence and backtest values are served from the committed model artifacts. These descriptions do not imply recalibration or reconstruction of unavailable training inputs.
p(a|g) = exp(x_a β) / Σ_b exp(x_b β)N_eff = 1 / Σ_a p(a|g)²| Path | Method | Returns | Example |
|---|---|---|---|
| /api/mercator/model | GET | Committed coefficients, standard errors, convergence fields and selected-sample backtest metrics for the model, incumbency and capacity baselines. | /api/mercator/model |