Credit risk & IFRS 9 macro scenarios

IFRS 9 expected-credit-loss models need forward-looking macroeconomic variables, and a clean history to fit them on. PakDataHub gives Pakistani lenders, microfinance banks and auditors those variables from official sources, in one consistent format.

#Typical macro variable set

#Build a quarterly model frame

import pandas as pd
KEY = "pk_live_xxx"
def s(sid, **p):
    q = "&".join(f"{k}={v}" for k, v in {"format": "csv", "sort": "asc", "limit": 10000, **p}.items())
    x = pd.read_csv(f"https://api.pakdatahub.com/v1/series/{sid}?{q}&api_key={KEY}", parse_dates=["date"])
    return x.set_index("date")["value"].rename(sid)

frame = pd.concat([
    s("inflation.cpi.national.yoy"),
    s("fx.rate.avg.usd", transform="yoy"),
    s("banking.six_months_karachi_interbank_offer"),
    s("industry.lsm.qim", transform="yoy"),
], axis=1).resample("QE").mean()

#Audit trail

• Sourcing: every series carries its official source (meta.source, source_url in the catalog), so auditors can trace each input.
• Vintages: with revision vintages (Pro) you can rebuild the exact dataset used at a past reporting date, even after SBP or PBS revise the numbers.
• Stable ids: series ids don't change, so model code keeps working between reporting cycles.

See also Banks, treasury & ALM and For macro researchers.