Getting Started with the ECB Data Portal API

Source on EconIndx: European Central Bank — free, no registration, 139,000+ time series, SDMX 2.1 API.

Access & Pricing

Free with no registration required. The ECB Data Portal API is a public institutional resource — no API key, no rate limit published, no contracts. The base URL is https://data-api.ecb.europa.eu/service. CSV format is the easiest response to parse; SDMX-JSON and XML are also available.

Your First Data Pull

The ECB API uses SDMX 2.1 conventions. A key is built as frequency.currency.denomination.type.variant. The most useful starting point is the exchange rate (EXR) dataflow:

📌 Note: The ECB API uses SDMX 2.1 — a standard for statistical data exchange. The key structure is dataflow/key?parameters. The key is a dot-separated combination of dimension values. A . in any position means ‘all values’ for that dimension. For example, EXR/D.USD+GBP.EUR.SP00.A fetches daily USD and GBP rates against EUR.

import requests
import pandas as pd
import io

ECB_BASE = "https://data-api.ecb.europa.eu/service"

def fetch_ecb(dataflow: str, key: str, start: str = "2010-01-01") -> pd.DataFrame:
    """Fetch ECB data in CSV format."""
    url = f"{ECB_BASE}/data/{dataflow}/{key}"
    r = requests.get(url, params={"startPeriod": start, "format": "csvdata"})
    r.raise_for_status()
    df = pd.read_csv(io.StringIO(r.text))
    return df

# Daily EUR/USD exchange rate
eurusd = fetch_ecb("EXR", "D.USD.EUR.SP00.A")
print(f"EUR/USD rows: {len(eurusd)}")
print(f"Columns: {list(eurusd.columns)}")
print(eurusd[["TIME_PERIOD", "OBS_VALUE"]].tail(5))

First Pull: What to Expect

EUR/USD daily from 1999 (euro introduction) — about 6,500 business days. The CSV response columns always include TIME_PERIOD and OBS_VALUE; other columns are dimension labels.

No data on weekends/holidays: ECB rates are published on TARGET2 business days only. Expect gaps in daily series — handle with forward-fill or explicit null rows, not as missing data.

Key Dataflows to Start With

ECB dataflows are identified by short codes. The most useful for economic pipelines:

Exchange rates (EXR):

D.{CURRENCY}.EUR.SP00.A  — daily reference rate
M.{CURRENCY}.EUR.SP00.A  — monthly average

Currency codes: USD, GBP, JPY, CHF, CNY, AUD, CAD

Interest rates (IRS):

• Deposit facility rate, ECB key rate: M.U2.EUR.RT0.IB.L40.YY._X.N
• EURIBOR 3M: M.U2.EUR.RT0.IB.L40.YY._X.N

HICP inflation (ICP):

• Euro area aggregate: M.U2.N.000000.4.INX
• By country: replace U2 with country code (DE, FR, IT, ES, etc.)

Monetary aggregates (BSI):

• M3, euro area: M.U2.N.A.A20.A.1.U2.2240.Z01.E
• M1: M.U2.N.A.A10.A.1.U2.2240.Z01.E

# Pull multiple currencies at once using + separator
multi_fx = fetch_ecb("EXR", "D.USD+GBP+JPY+CHF.EUR.SP00.A", start="2020-01-01")
print(f"Multi-currency rows: {len(multi_fx)}")
print(multi_fx.groupby("CURRENCY")["OBS_VALUE"].describe())

Data Tolerance & Validation

What’s normal:

• EUR reference rates are mid-market (not bid/ask). They differ from interbank transaction rates — don’t use them for trade pricing, use them for accounting and analytics.
• Gaps on TARGET2 holidays (not just weekends — also ECB-specific holidays). A gap of 1–4 days is normal; longer gaps indicate a data problem.
• M3 and BSI data revised monthly for the prior 2–3 months. HICP revised occasionally with annual benchmark.
• The OBS_STATUS column carries quality flags: A = normal, E = estimated, P = provisional. Filter or store accordingly.

⚠️ Weekend gap: The ECB’s exchange rate series only publishes on business days. Gaps for weekends and ECB TARGET closing days are expected — do not treat them as missing data errors. Use forward-fill sparingly and always flag filled values in your schema.

Validation checks:

def validate_ecb_pull(df: pd.DataFrame, series_name: str) -> dict:
    df["TIME_PERIOD"] = pd.to_datetime(df["TIME_PERIOD"], errors="coerce")
    latest = df["TIME_PERIOD"].max()
    days_stale = (pd.Timestamp.today() - latest).days

    # Check for suspiciously large moves (>5% daily for FX is extremely rare)
    df_sorted = df.sort_values("TIME_PERIOD")
    df_sorted["pct_chg"] = df_sorted["OBS_VALUE"].pct_change().abs()
    outliers = (df_sorted["pct_chg"] > 0.05).sum()

    return {
        "series": series_name,
        "row_count": len(df),
        "latest_date": str(latest.date()),
        "days_stale": days_stale,
        "stale_alert": days_stale > 5,  # daily series
        "outlier_moves": int(outliers),  # large daily moves worth investigating
    }

report = validate_ecb_pull(eurusd, "EUR/USD")
print(report)
# Expected: days_stale=0-3 (weekend gap), outlier_moves < 5 (crisis periods only)

Alert thresholds:

• Daily FX series: alert if more than 5 business days stale
• Monthly monetary data: alert if more than 45 days stale
• Any OBS_VALUE that is more than 15% different from prior period for an FX rate: investigate

Spot-Check Against Known Values

A useful sanity check — ECB publishes its reference rates on the website. Your pulled EUR/USD for a known business day should match to 4 decimal places:

# Spot-check: ECB published EUR/USD on 2024-01-02 = 1.0963
known_date = "2024-01-02"
pulled_value = eurusd[eurusd["TIME_PERIOD"] == known_date]["OBS_VALUE"].values

if len(pulled_value):
    diff = abs(pulled_value[0] - 1.0963)
    print(f"EUR/USD {known_date}: pulled={pulled_value[0]}, expected≈1.0963, diff={diff:.6f}")
    assert diff < 0.0001, "Reference rate mismatch — check API response"

Schema Stability

ECB SDMX dataflow codes are very stable. The CSV column names (FREQ, CURRENCY, TIME_PERIOD, OBS_VALUE, etc.) have not changed. The main risk is dimension value changes when new ECB member countries join — aggregates like U2 (euro area) are extended, which can cause a small break in aggregate-level series. Monitor the series metadata for composition changes.

Related Sources on EconIndx

• European Central Bank
• Eurostat
• IMF Data
• Bank for International Settlements

Related Guides

• Getting Started with the Eurostat API
• Getting Started with the World Bank API
• API Integration for Economic Data Sources
• API Versioning Strategies for Economic Data Sources