Remittances

Remittances methodology

The figures on /remittances and on the 185 country pages are computed at build time by the queries in src/lib/remittances.ts. These numbers there are choices rather than query results: the 10-year receipt minimum that decides which countries get a page, the 5% of GDP cut used to count remittance-heavy receivers, the 2000 start of the balanced panel, the 1%, 5% and 20% bands used to grade how far the modelled corridor matrix sits from reported receipts, and the five-largest cut used to measure corridor concentration. This page states what the queries read and what they cannot see.

Sources

• Flows. World Bank World Development Indicators, BX.TRF.PWKR.CD.DT (personal remittances received) and BM.TRF.PWKR.CD.DT (personal remittances paid), current US dollars, CC BY-4.0. WDI defines personal remittances as personal transfers plus compensation of employees, so the series includes the earnings of short-term and border workers, not only transfers by long-term migrants. Coverage in the parquet: 201 economies, 1966 to 2024, 14,086 country-year-direction rows.
• GDP denominator. WDI NY.GDP.MKTP.CD, GDP in current US dollars, same vintage. Numerator and denominator therefore come from the same file, and neither is PPP adjusted.
• Current account. IMF Balance of Payments, current-account balance in US dollars, used only for the subtraction test on the index page.
• Bilateral corridors. World Bank / KNOMAD Bilateral Remittance Matrix, 2021 vintage: 10,229 corridors, 212 senders, 175 receivers, $781.08B of modelled value. The source publishes values in US$ millions; this module converts to dollars before any comparison with the flows table, which is in raw dollars.

The corridor matrix is a model

The World Bank does not observe bilateral remittance payments. It takes each country’s recorded total receipts and allocates them across origin countries using bilateral migrant-stock estimates and origin-destination incomes. Every corridor value in the matrix is the output of that allocation rule, not an observed payment.

Summing the matrix by receiver and comparing with the 2021 balance-of-payments receipts of the same country: 168 receivers can be compared, 71 agree within 1%, 66 differ by more than 5%, and 33 by more than 20%. The largest gap in dollars is China: $53.00B modelled against $17.12B reported. A further 7 receivers in the matrix have no 2021 reported figure at all.

Construction rules

• The latest year is derived, never typed. It is the maximum year with a receipt observation, currently 2024. GDP is joined on the flow year, never on the GDP table’s own maximum year, which is 2025: joining on that would match no flow row at all and blank the share column rather than break the build.
• Time series run on a balanced panel. The trend chart uses the 128 countries that reported receipts in every year from 2000 to 2024. An unbalanced total would move with the set of reporters as well as with the money.
• Cross-sections state their N. 160 countries report receipts for 2024, and all of them also have a 2024 GDP figure. 162 report payments. 158 have both flows and an IMF current-account balance.
• The current-account test is a subtraction, not a forecast. Remittances are recorded inside the current account: personal transfers in secondary income, compensation of employees in primary income. Removing net remittances from the recorded balance leaves the rest of the account exactly as reported. It does not model what a country would do without the inflow, and the page says so.
• The build asserts its own assumptions. The build throws if the direction column stops being exactly "inflow" / "outflow", if the bilateral table stops being a single year, if it acquires self-corridors or negative values, or if the balanced-panel start year loses its observations. A refresh that breaks a stated assumption fails the build instead of blanking a section.

Which countries get a page

A country gets a page when it has at least 10 annual receipt observations. 185 of the 201 economies in the table qualify. The other 16 are listed on the index with their observation counts: United Arab Emirates, Bahrain, Cayman Islands, Bahamas, Andorra, Brunei, Eritrea, Micronesia, Equatorial Guinea, Libya, San Marino, Somalia, Turks and Caicos, Chad, Turkmenistan, Vietnam.

A page does not imply a current country. 185 countries qualify on history, but only 160 report for 2024, so a page can carry a last observation years old. Each country page states its own last year in the headline.

Reporting coverage by year

The best-covered years are 2011 and 2014, at 181 reporting countries; 2024, the last year in the table, has 160. Every time-series total in this module is therefore summed over the balanced panel, and every cross-section states its own N.

What the data cannot tell you

• The WDI figures are compiled from balance-of-payments reporting. The parquet carries four columns (iso3, year, flow_type, value_usd) and no field that separates formal from informal transfer, so this module cannot measure or bound hawala, hundi, or cash carried home by hand. Whether a given country's reported figure embeds an estimate of those flows is not something this data can answer.
• The receipts side and the payments side do not reconcile: 160 countries record $856.61B received in 2024 while 162 record $619.28B paid. The $237.32B difference is a gap between two incomplete sets of reporters.
• WDICSV.csv, the bulk file this module is built from, is a country-by-year value matrix whose only columns are country, indicator and year. It carries no per-observation estimate flag, so this module cannot separate national submissions from World Bank staff estimates.
• No comparison of remittances against FDI or official aid is made anywhere in this module. Such a comparison would pair this balance-of-payments flow with a series from a different collection on a different universe, and it is left absent here rather than approximated with the wrong series.
• The corridor matrix exists for 2021 only, so it supports no corridor trend, no pre-pandemic or post-pandemic comparison, and no statement about a corridor opening or closing.