Project 2 Data: Credit Default

The consumer-lending dataset for Project 2 (Predicting Credit Default): 29,881 customer accounts, one row each, with whether the account defaulted. Get it in whatever form suits you.

Start here: the project guide

The full Project 2 brief – the scenario, deliverables, rubric, workflow, and the Challenge & Verify log. Due Wednesday, July 15, 12:00 noon MT.

Read the Project 2 guide

Get the data

Download everything (.zip)

The CSV, the SQLite database, and the data dictionary in one file.

Download all (.zip)

CSV

For pandas or Excel. Opens on GitHub (right-click → Save, or use the zip above).

Get the CSV

SQLite database

The same data as one table, accounts, for any SQL tool.

Download .sqlite

Explore it in your browser (no install)

Run SQL against the data right here, with an AI helper alongside. Nothing to download or set up.

Open the data explorer

Load it in a notebook

In your Project 2 notebook, read the CSV straight from its URL, no download step:

import pandas as pd
url = "https://raw.githubusercontent.com/sean-mccaman/acctg5150-090/main/2026-summer/project-2/ar_default_data.csv"
df = pd.read_csv(url)
df.head()

What's in it

Column	Meaning	Role
`bad_ar`	Defaulted next month (1 = yes, 0 = paid)	Target
`credit_limit`	Normalized credit limit (higher = more credit)	Feature
`outstanding_balance`	Normalized current balance owed	Feature
`pay_this_month`	Payment status (0 = on time, 1-3 = months behind)	Feature
`no_activity_this_month`	No payment activity this month (0 = active, 1 = inactive)	Feature
`last_payment_portion`	Fraction of the balance paid last cycle	Feature
`num_recent_payments`	Count of payments in recent months (0-6)	Feature
`gender`	Customer gender	Demographic
`education`	Education level	Demographic
`marital_status`	Marital status	Demographic
`age_decile`	Age group (20, 30, 40, 50, 60, 70)	Demographic

About 22% of accounts default. The full reference is in the data dictionary inside the zip.

The four demographic columns are included so you can examine whether they should be used. Using them in a real credit decision raises fair-lending questions under the Equal Credit Opportunity Act (ECOA). See the project guide for how to treat this. The data contains no personally identifiable information.