Barometer: Methodology & Validation
How the Financial Health Barometer is built, how it holds up against independent ground-truth data, and the open research question we're pursuing in the open.
What the Barometer is
The Financial Health Barometer is a real-time, readable summary of household economic stress across all 50 states and DC. It combines official U.S. government statistics — unemployment (BLS), poverty (Census SAIPE), rent burden (Census ACS), fair-market rent (HUD), house prices (FRED), and Harvard JCHS housing data — with search-interest signals, and expresses them on a single 0–200 stress scale across four indicators: Financial Anxiety, Food Insecurity, Housing Stress, and Affordability.
It is designed as a directional instrument for relative comparison between states and over time — not as a substitute for the underlying official measures. We say so on the tool itself, and this page makes the evidence behind that statement explicit.
How it validates
We tested the Barometer two ways, and publish both results openly — including the uncomfortable parts. The analysis is fully reproducible (links below).
1. It faithfully reflects the official data it is built from
Each index can be reconstructed from its published formula and the underlying official metrics almost perfectly — confirming the tool does exactly what it documents:
| Indicator | Reconstruction r |
|---|---|
| Financial Anxiety | 0.995 |
| Food Insecurity | 0.984 |
| Housing Stress | 0.997 |
| Affordability | 0.991 |
2. It tracks independent ground truth — as well as the official data does
Compared against the USDA's independent food-security survey (CPS-FSS, 2022–24 state averages), the Barometer's Food Insecurity index is a valid proxy — and it tracks that survey about as closely as the raw Census poverty rate does:
| Comparison | Correlation (r) |
|---|---|
| Barometer Food Insecurity vs USDA survey | 0.685 |
| Census poverty rate vs USDA survey | 0.685 |
| Barometer, controlling for poverty (partial) | 0.14 |
n = 51 states + DC · robust across data vintages (2019–21: 0.75 / 0.18)
What this means, plainly. Because the Barometer is built from official statistics, it faithfully reflects them — its readings are a clean, real-time summary of the government's own data. The flip side is that, by construction, it inherits the release lag of those sources. So the honest question is not "is it accurate?" (it tracks ground truth) but "can it warn us earlier than the official releases allow?"
The open research frontier
That question defines our current research: can real-time internet-search signals provide early warning of household financial distress beyond what lagged official statistics already tell us? Specifically, we are testing whether absolute-probability signals from the Google Health Trends API improve out-of-sample nowcasts of state-level distress — validated against the Census Household Pulse Survey — over and above the official inputs, and whether those absolute signals beat the freely available relative Google Trends series.
We're running this as a preregistered study and will publish the result either way — including if the answer is "search adds nothing." Transparency about what a tool can and cannot do is the point.
Data & reproducibility
The Barometer is open (MIT-licensed). Downloadable artifacts behind the analysis above:
Reproducible analysis scripts and the working paper live in the project repository under /docs.