How to get imputed multilateral ODA by sector

one.crs.imputed_multilateral_oda spreads each provider's core contributions to multilateral organisations across recipients and sectors, in the proportions each organisation spends its own money. Ask for it like any other indicator and split it with by=.

Check how the money was spread

Every core dollar lands in one row, and allocation_status says how: imputed from the organisation's own spending, stale_share from an older window of it, proxy from a related fund's spending or one fixed purpose, or unallocated.

  allocation_status   value
0           imputed  6060.3
1             proxy   451.3
2       unallocated   219.6

import oda_data as oda

df = oda.get(
    "one.crs.imputed_multilateral_oda",
    years=2023,
    providers="GBR",
    by=["allocation_status"],
)
print(df[["allocation_status", "value"]].round(1))

The three rows add up to USD 6,731 million, the United Kingdom's core contributions in 2023 (oecd.multisystem.core_contributions). An allocation that loses or adds money raises ImputationConservationError.

Add imputed to bilateral ODA by sector

Bilateral CRS ODA and imputed multilateral ODA share the gross disbursement basis, so the two add up sector by sector. Some sectors hold only one of the two, so fill both sides with 0, name the columns by indicator id, and check the table against both totals.

indicator                       bilateral  imputed   total
sector
Refugees in Donor Countries        5311.0      0.0  5311.0
Health                              935.0   1970.0  2905.0
Humanitarian                        968.0    465.0  1433.0
Administrative Costs of Donors      881.0    330.0  1210.0
Multi-sector                        636.0    306.0   942.0

import pandas as pd

import oda_data as oda

bilateral = oda.get("oecd.crs.bilateral_oda", years=2023, providers="GBR", by=["broad_sector"])
imputed = oda.get("one.crs.imputed_multilateral_oda", years=2023, providers="GBR", by=["broad_sector"])

both = pd.concat([bilateral, imputed])
both["sector"] = both.broad_sector_name.astype("string").fillna("Unallocated")
wide = both.pivot_table(index="sector", columns="indicator", values="value", aggfunc="sum")
wide = wide.rename(
    columns={
        "oecd.crs.bilateral_oda": "bilateral",
        "one.crs.imputed_multilateral_oda": "imputed",
    }
).fillna(0)
wide["total"] = wide.bilateral + wide.imputed

assert abs(wide.total.sum() - (bilateral.value.sum() + imputed.value.sum())) < 1e-6
print(wide.sort_values("total", ascending=False).head(5).round(0))

Two-thirds of the United Kingdom's health ODA in 2023 went through multilateral organisations' core budgets, so a sector chart from bilateral CRS alone understates it threefold.

The assertion holds because both kinds of one-sided row survive. Action Relating to Debt has USD 5.5 million imputed and no bilateral rows, so a total that fills only imputed leaves it null and drops the money. The USD 219.6 million with allocation_status unallocated has no sector name, and pivot_table drops a row whose index is missing, so it gets a label first. The table adds up to USD 18,858.3 million, bilateral USD 12,127.1 million plus imputed USD 6,731.2 million.

Filling with 0 suits a complete year such as 2023. In a partial last year a missing component has not been reported yet, so read the OdaDataWarnings of both calls before you fill.

See which channel lent its shares

oda.records() returns one row per provider, channel, recipient, and purpose, with share_channel and share_years naming the spending behind each share.

channel  share_channel
44007    44002            409.6
Name: value, dtype: float64

import oda_data as oda

rows = oda.records("one.crs.imputed_multilateral_oda", years=2023, providers="GBR")
proxy = rows[rows.allocation_status == "proxy"]
print(proxy.groupby(["channel", "share_channel"], observed=True).value.sum().round(1))

IDA's Multilateral Debt Relief Initiative (channel 44007) reports no spending of its own, so its USD 410 million takes the shares of IDA's ordinary lending (44002). The other USD 42 million of proxy money goes to one purpose with no recipient, so its share_channel is null.

/// warning | Heads up Unallocated amounts and fixed-purpose proxies have no recipient. A call with recipients= leaves them out, so its total is smaller than the provider's core contributions. ///

Sum across providers

EU members count what they pay EU Institutions (918) as their own core contributions, so 918's contributions on top would count that money twice. A call that sums providers, with provider outside by, leaves 918 out of the default providers. A call split by provider keeps 918's row.

   year    value
0  2023  62500.0
   provider  value
48      918   88.1

import oda_data as oda

summed = oda.get("one.crs.imputed_multilateral_oda", years=2023, by=["year"])
split = oda.get("one.crs.imputed_multilateral_oda", years=2023, by=["provider", "year"])
print(summed[["year", "value"]].round(1))
print(split[split.provider == "918"][["provider", "value"]].round(1))

/// note | Good to know The provider rows add up to USD 62,588 million in 2023, which is 918's USD 88 million more than the sum. explain() says whether a call left 918 out. Naming 918 with an EU member and summing them raises UnsupportedQueryError. ///

Troubleshooting

UnmappedChannelError means a multilateral provider and agency pair in the CRS has no channel in the bundled crosswalk, usually a new OECD agency. The message names the pair and the money at stake. Report it on the issue tracker, since the crosswalk is reviewed by hand.
ImputationConservationError means the rows of a provider, channel, and year did not sum to the core contribution. It signals a defect in oda-data. Report it with the call that raised it.