Bilateral Migration Flow Imputation

They are complementary. Stock-based estimation derives the totals — how many people left each origin and arrived at each destination — from the change in migrant stocks. Imputation takes those totals as margins and fills in the full interior of the country-pair matrix, harmonizing any reported flows and smoothing sparse corridors with a prior. In the Abel and Cohen workflow the stock method supplies the scaffolding and the imputation completes the structure, so the two are usually run together rather than as alternatives.

Iterative proportional fitting makes the matrix match its margins but does nothing to tame the unreliability of corridors with only a handful of reported migrants, where raw counts are noisy and observed zeros may be false. The pseudo-Bayesian step shrinks each cell toward a structured prior, with a tunable prior weight, so thin corridors borrow strength from the overall pattern and do not swing on sampling noise. Combining shrinkage with margin fitting gives flows that are simultaneously consistent with the totals and statistically stable across large and small corridors alike.

Yes. Once the total bilateral flows are fixed, each corridor can be partitioned into sex- and age-specific cells using proportions fitted to whatever sex and age margins are observed, so the disaggregated table both sums back to the corridor totals and matches the compositional totals. This reuses the same margin-fitting logic at a finer level. The Abel and Cohen release provides such demographic disaggregation, which is valuable for studies of female migration, youth mobility, and the age structure of flows.