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:

  1. Get the list of available dataflows
  2. 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
(225, 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 PI_WCA_2026_MAY_VINTAGE Production Indexes, World and Country Group Ag... IMF.STA 2026-05-27T14:17:19.149352Z
1 FFS Fossil Fuel Subsidies IMF.STA 2026-04-06T16:34:21.566778Z
2 COFER Currency Composition of Official Foreign Excha... IMF.STA 2025-12-19T16:09:51.115672Z
3 IL_2026_FEB_VINTAGE International Liquidity (IL) 2026 February IMF.STA 2026-02-25T20:31:17.183284Z
4 MFS_DC_2026_JAN_VINTAGE Monetary and Financial Statistics (MFS), Depos... IMF.STA 2026-02-09T20:07:05.015110Z

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
12 CTOT Commodity Terms of Trade (CTOT)
32 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
2 COFER Currency Composition of Official Foreign Excha...
28 BOP_2026_APR_VINTAGE Balance of Payments (BOP) 2026 April
81 ITS International Trade in Services (ITS)
102 BOP_AGG Balance of Payments and International Investme...
118 BOP_2026_MAY_VINTAGE Balance of Payments (BOP) 2026 May

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
155 FSI_COUNTRY_METADATA_TABLE_2 Financial Soundness Indicators (FSI), Country ... 2026-10-09 22:39:46.289903+00:00
133 FAS Financial Access Survey (FAS) 2026-10-09 19:09:56.991362+00:00
72 AFRREO_2026_APR_VINTAGE Sub-Saharan Africa Regional Economic Outlook (... 2026-10-07 19:23:15.996650+00:00
222 WEO_2026_APR_VINTAGE World Economic Outlook (WEO) 2026 April 2026-10-05 21:05:13.368993+00:00
79 CSR Comprehensive Surveillance Review (CSR) 2026-10-02 14:31:23.878915+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
32 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.