Fetching primary parquet sources and computing exhibits.
method research · bilateral trade flows
Bilateral trade flow imbalances
These charts describe differences between goods exports in opposite directions. In 2023, 3,021 country pairs have at least $50M in each direction. Their median absolute transformed imbalance is 10.0%, with a 90th percentile of 27.8%. These are descriptive statistics, not tariff equivalents or estimates of one-way trade costs.
What the transformation measures
Let A and B follow alphabetical ISO3 order and X_AB denote goods exports from A to B. We report T = 100 × [(X_BA / X_AB)1/7 − 1], using a fixed power of 1/7 to compress the flow ratio. A positive T means B exports more to A than A exports to B. Most charts show |T|. Reversing the labels changes its magnitude, and changing the power can change rankings by absolute T. It does not isolate policy, logistics or regulatory costs.
Novy (2013), equations 4–5, identifies the geometric mean of bilateral trade costs relative to domestic trade costs: [(X_AA × X_BB) / (X_AB × X_BA)]1/[2(σ−1)] − 1. This product relationship does not separately identify the two directions from the observed bilateral-flow ratio. We do not estimate that geometric mean here: domestic gross trade is required, and GDP minus gross exports is not an interchangeable measure.
Pairs with large flow ratios
Figure 1
Top 20 pairs by absolute log flow ratio, 2023
Scroll horizontally to view the full chart.
China exports $5.8B to Liberia and receives $85M in reverse. Bar lengths show |T|; the ordering uses |ln(X_BA/X_AB)|, so bar lengths need not descend.
Figures 1 and 5 retain the existing sample filter requiring positive GDP-minus-goods-exports residuals for both partners. That residual is used only for eligibility, not as domestic trade or a trade-cost estimate. Other figures do not use this additional filter.
CEPII BACI 2023, current USD, FOB in both directions. Both flows ≥ $50M.
Association with distance
Figure 2
Largest deviations from a descriptive distance fit, 2023
Scroll horizontally to view the full chart.
Fixed regional comparison sets
Figure 3
Median absolute transformed imbalance within comparison sets, 2023
Scroll horizontally to view the full chart.
Sets are EU27, ASEAN-10, and Argentina, Brazil, Paraguay, Uruguay and Bolivia. Membership is held fixed; this is not an estimate of an agreement's effect or historical legal coverage.
CEPII BACI 2023. Unweighted median |T| among qualifying pairs within each set.
Annual distribution
Figure 4
Absolute transformed flow imbalance, 1996–2024
Scroll horizontally to view the full chart.
Relative imbalances and dollar gaps
Figure 5
Transformed ratios and absolute dollar gaps, 2023
Scroll horizontally to view the full chart.
One-way trade costs: estimates unavailable
Figure 6 is withheld. A bilateral-flow ratio alone cannot separate forward and reverse trade costs. The previous directional estimates combined an unsupported ratio identity with a geometric-mean expression and took an additional square root. Neither those levels nor their distance-bin comparisons are supported by the cited method.
Fixed GDP-per-capita bands
Figure 7
Median absolute transformed imbalance by GDP-per-capita band, 2023
Scroll horizontally to view the full chart.
Dyadic characteristics
Figure 8
Median absolute transformed imbalance by pair characteristics, 2023
Scroll horizontally to view the full chart.
OLS relates ln(|T|/100) to ln(distance) for 2,517 pairs with positive distance and |T| greater than 0.1%. The fitted slope is 0.14. Bars show 100 × [exp(residual) − 1], the relative deviation from the fitted magnitude. This is an association, not evidence that residuals identify unobserved trade barriers.
CEPII BACI 2023; distance from Legacy gravity_bilateral table (exact release provenance unverified), 2021 records.
The median moves from 7.5% in 1996 to 10.0% in 2024, a change of 2.5 percentage points. The qualifying sample can change each year. This is not a trend in identified trade costs.
CEPII BACI 1996–2024. Annual continuous 50th and 75th percentiles of |T|; both directions ≥ $50M in each year.
Bar lengths show |T|; labels give |X_AB − X_BA| in USD. A ratio and a dollar gap describe different aspects of the same two flows, not separate economic effects. These are the first 12 pairs from Figure 1.
CEPII BACI 2023. Same pair order and eligibility as Figure 1.
Bands use 2021 current-USD GDP per capita: B1 ≤ $1,136; B2 > $1,136 and ≤ $4,466; B3 > $4,466 and ≤ $13,845; B4 > $13,845. These fixed descriptive bands are not World Bank Atlas-GNI income classifications. Higher bands are listed first in each pair.
CEPII BACI 2023; GDP per capita from Legacy gravity_bilateral table (exact release provenance unverified), 2021 records. Unweighted median |T|.
Pairs are grouped by recorded border, official-language and common-colonizer flags. Unflagged groups include missing flags treated as zero in the existing extraction. The comparisons do not identify causal effects or directional trade barriers.
CEPII BACI 2023; dyadic flags from Legacy gravity_bilateral table (exact release provenance unverified), 2021 records. Unweighted median |T|.