Hot Potato

Most analyses of household debt split it by loan type. This one splits it by who holds it, because whoever holds a loan takes the loss if it goes bad: the federal government, the GSEs and agency pools, banks, private securitization and others. In 1952 the public sector held 3.2% of American household mortgages and consumer credit; by 2026 it held 56.4%. The project started from the thesis that each crisis ratchets that share up, and the data rejects it as a general pattern: the shift after a crisis looks like any other period’s, and the long climb is a slow structural build-out of the GSEs. One crisis is the exception. From 2007 to 2011 private-label securitization lost almost two-thirds of its book, the public share jumped 13 points, and it never went back. The story is told as an animated dot flow in which each dot is $50 billion of debt in 2026 dollars, moving between holders as the reader scrolls.

Question

Have economic crises since the 1950s moved the credit risk on American household debt onto public balance sheets?

My role

Individual. The question, the data pipeline and warehouse, the analysis and the dot-flow site.

Method

Python · PostgreSQL · Next.js · Canvas animation

Tools

Python (pandas, psycopg) · Supabase Postgres · Next.js and HTML Canvas on Vercel · Vitest

Evidence

Federal Reserve Z.1 Financial Accounts, 1952 to 2026Q2: holders of 1–4 family mortgages and of consumer credit by type; FRED price and labour series; the New York Fed Household Debt and Credit report from 2003.

Result

56.4%Held by the public sector, 2026

3.2%

Public share, 1952

+13.1 pp

Public share, 2007Q4 to 2011Q4

$50B

Of debt per dot, 2026 dollars

Findings

• The public share of household mortgages and consumer credit rose from 3.2% in 1952 to 56.4% in 2026
• Crises do not reliably ratchet it up: across eleven recessions the shift after a crisis looks like any other period’s, and the 1980s alone added about 19 points with no crisis at all
• One crisis did: from 2007Q4 to 2011Q4 the share rose from 36.1% to 49.2%, larger than 98% of all four-year windows, and it never reversed

Output

Live interactive dot-flow story on Vercel.

BoundaryDescriptive: the share is measured by holder from the Fed’s Financial Accounts, and the crisis test compares crisis windows with ordinary ones rather than identifying what caused any shift. Holder is a proxy for who bears the credit risk: federally insured loans held by banks count as private, and risk the GSEs pass on to private investors still counts as public.