economicpolicyuncertainty

A Python client for policyuncertainty.com. Fetch any EPU index as a pandas DataFrame, cached locally in SQLite so repeat calls don't re-hit the source.

Installation

pip install economicpolicyuncertainty

With the optional REST server:

pip install economicpolicyuncertainty[server]

Quick start

import economicpolicyuncertainty as pu
import datetime as dt

data = pu.download("UK_DAILY", refresh=True ,start="2021-01-01", end=dt.datetime.now())              

print(data)

start

import economicpolicyuncertainty as pu

pu.available()                                        # DataFrame of every series
pu.download("UK_DAILY", refresh=True)                 # fetch + return DataFrame indexed by date
pu.download("ALL_COUNTRIES_MONTHLY", column="UK")     # one country from the panel file

s = pu.Series("UK_DAILY")
s.info                                                # registry metadata dict
s.history(start="2020-01-01", end="2020-12-31")       # filtered DataFrame
s.columns()                                           # for wide series, list available columns
s.latest()                                            # most recent record as a dict
s.refresh()                                           # force re-fetch from source

pu.refresh_all()                                      # refresh every series; failures are isolated

Indicators

Multi-column series

ALL_COUNTRIES_MONTHLY, US_CATEGORICAL, and US_STATE_EPU each contain many columns in a single file. Use column= to filter to one, or .columns() to see what's available:

s = pu.Series("ALL_COUNTRIES_MONTHLY")
s.columns()
# ['australia', 'brazil', 'canada', 'chile', 'china', 'france',
#  'gepu_current', 'gepu_ppp', 'germany', 'greece', 'india', 'ireland',
#  'italy', 'japan', 'korea', 'mainland_china', 'mexico', 'pakistan',
#  'russia', 'scmp_china', 'singapore', 'spain', 'uk', 'us']

pu.download("ALL_COUNTRIES_MONTHLY", column="uk")

s = pu.Series("US_CATEGORICAL")
s.columns()
# ['1_economic_policy_uncertainty', '2_monetary_policy', '3_taxes',
#  '4_government_spending', '5_health_care', '6_national_security',
#  '7_entitlement_programs', '8_regulation', '9_trade_policy',
#  '10_sovereign_debt_currency_crises', 'financial_regulation',
#  'fiscal_policy_taxes_or_spending']

Column names are lowercased and have spaces replaced with underscores. Pass either the raw name ("United Kingdom") or the sanitized form ("uk") — both work.

API reference

pu.available() → DataFrame

Returns a DataFrame listing every series and its metadata (key, frequency, shape, url, notes).

pu.download(key, start=None, end=None, column=None, refresh=False) → DataFrame

Returns the series as a DataFrame indexed by date. Reads from the local cache by default; pass refresh=True to fetch fresh data first.

pu.Series(key)

Object-oriented interface to a single series.

pu.refresh_all() → list[dict]

Refreshes every series. Each failure is caught individually so one bad URL doesn't block the rest.

Local REST API (optional)

Useful for accessing data from non-Python environments.

python server.py          # runs at http://127.0.0.1:5000

Refreshes all series on startup, then daily at 06:00 local time (REFRESH_HOUR / REFRESH_MINUTE env vars to change it).

Handling the source's bot check

The site occasionally serves a JavaScript challenge page to automated requests. Every fetch tries a browser-header requests session first, then falls back to cloudscraper. If both fail, download the file manually and feed it in:

from economicpolicyuncertainty import fetch, db, registry

entry = registry.get("UK_DAILY")
with open("UK_Daily_Policy_Data.csv", "rb") as f:
    rows = fetch.parse(f.read(), entry)
db.upsert_rows(entry["key"], rows, shape=entry["shape"])

Verifying sources

If a series stops returning data, run:

python check_sources.py

It reports OK <rows> or FAILED <error> for every registry entry. A broken URL is a one-line fix in policyuncertainty/registry.py.