Discovering Datasets

Dataflows

The IMF publishes hundreds of separate datasets through its API. In SDMX — the standard the API speaks — a dataset is called a dataflow, and every request has to name one. So before you can fetch any data, you need to:

Get the list of available dataflows
Find the ID of the one you want

That ID is the first argument to every other function in imfp.

Listing Dataflows

imf_get_dataflows returns the full catalogue as a tidy DataFrame:

import imfp

dataflows = imfp.imf_get_dataflows()
dataflows.shape

(222, 7)

Each row describes one dataset:

Column	Meaning
`id`	The dataflow ID, e.g. `PCPS`. This is what you pass to the other functions.
`name`	Human-readable title.
`description`	Longer prose description, where the IMF provides one.
`version`	Version of the dataflow definition.
`agency`	The IMF department that publishes it, e.g. `IMF.STA`.
`structure`	URN of the datastructure definition that backs it.
`last_updated`	When the dataset was last refreshed.

dataflows[["id", "name", "agency", "last_updated"]].head()

	id	name	agency	last_updated
0	IIPCC_2026_APR_VINTAGE	Currency Composition of the International Inve...	IMF.STA	2026-04-27T15:03:17.665208Z
1	ANEA	National Economic Accounts (NEA), Annual Data	IMF.STA	2025-03-28T08:00:46.650082Z
2	GFS_COFOG	GFS Government Expenditures by Function	IMF.STA	2025-06-06T01:34:26.584454Z
3	MFS_FMP_2026_MAY_VINTAGE	Monetary and Financial Statistics (MFS), Finan...	IMF.STA	2026-05-27T14:59:17.737322Z
4	LS_2026_FEB_VINTAGE	Labor Statistics (LS) 2026 February	IMF.STA	2026-02-25T20:34:44.315994Z

Finding the Dataset You Want

Because the result is an ordinary DataFrame, searching it is an ordinary pandas filter. Search the name column for a keyword:

dataflows[dataflows["name"].str.contains("Commodity", case=False, na=False)][
    ["id", "name"]
]

	id	name
85	CTOT	Commodity Terms of Trade (CTOT)
211	PCPS	Primary Commodity Price System (PCPS)

The description column often contains terms the title does not, so it is worth searching too:

matches = dataflows[
    dataflows["description"].str.contains("balance of payments", case=False, na=False)
]
matches[["id", "name"]].head()

	id	name
42	IL	International Liquidity (IL)
55	BOP_2026_FEB_VINTAGE	Balance of Payments (BOP) 2026 February
78	COFER	Currency Composition of Official Foreign Excha...
123	BOP	Balance of Payments (BOP)
143	BOP_AGG	Balance of Payments and International Investme...

Checking How Current a Dataset Is

last_updated tells you when each dataset was last refreshed, which is useful when you are deciding whether a series is current enough for your purposes:

import pandas as pd

recent = dataflows.dropna(subset=["last_updated"]).copy()
recent["last_updated"] = pd.to_datetime(recent["last_updated"], format="mixed")
recent.sort_values("last_updated", ascending=False)[["id", "name", "last_updated"]].head()

	id	name	last_updated
214	IRFCL	International Reserves and Foreign Currency Li...	2026-06-19 15:17:41.650334+00:00
11	ISORA_LATEST_DATA_PUB	ISORA Latest Data	2026-06-15 17:13:23.041621+00:00
99	GPT	IMF Global Policy Tracker: How Countries are R...	2026-06-12 16:15:44.521971+00:00
19	FM	Fiscal Monitor (FM)	2026-06-03 16:00:56.895956+00:00
111	MFS_OFC_2026_MAY_VINTAGE	Monetary and Financial Statistics (MFS), Other...	2026-05-27 15:19:21.834692+00:00

Reading the Dataflow ID

The id column is the value you pass everywhere else:

pcps = dataflows[dataflows["id"] == "PCPS"]
pcps[["id", "name", "agency", "version"]]

	id	name	agency	version
211	PCPS	Primary Commodity Price System (PCPS)	IMF.RES	9.0.0

Note the agency. Different IMF departments publish through the same API but do not all support the same features — in particular, only IMF.STA currently honors server-side time filtering. imf_get warns you when you ask for a time window that the publishing agency will ignore. See Fetching Data.

Next Step

With a dataflow ID in hand, the next step is finding out how that dataset can be filtered. See Dimensions and Codes.