Policy findings, with the data and code behind each
This is the findings layer of Too Early To Say. Each row is a policy result we have published, paired with the one number that carries it, the data it rests on, the method that produced it, and a link to the full analysis with open code. We use an AI agent to pull data, draft pipelines, and run checks. The question, the identification, and the reading of each result remain ours.
This is for policy analysts, funders, and reporters who want the result and the proof in the same place. The problem it solves is simple. A striking number spreads without the data and method behind it, so a reader cannot tell a measured finding from a talking point.
One rule holds every row on this page. A finding earns its place only when someone else can retrace the steps to the same number. Each row names the study, the data, and the method, and every number below appears in the analysis it links to.
The findings
Proximity is not access
In Santa Clara County, SNAP participation varies 4.7x across neighborhoods with equal store access. A retail-density map reports no food deserts and still misses where the access gap is.
How we got here: The Food Desert MythThe gap county totals hid
Countywide SNAP participation stayed flat through the pandemic, yet the food-access gap between the least- and most-vulnerable tracts widened 49%. One number for the county hid a split between its neighborhoods.
How we got here: How COVID Widened InequalityWork above the food-security line
In 57 Santa Clara County tracts, more than 60% of working-age adults hold jobs while poverty runs above 10%. A paycheck in these neighborhoods does not clear the food-security line.
How we got here: When Work Isn't EnoughThe label that is not fraud
On the federal exclusion list, 40% of entries are license revocations rather than fraud. A model trained to flag fraud from that label learns the contamination instead of the crime.
How we got here: The Label That Isn'tThe cliff the relief postponed
When federal stimulus ended in 2011, California K-12 funding fell 8.7% in a single year, a sharper drop than the recession itself. The relief postponed the cliff rather than removing it.
How we got here: The Fiscal Cliff Schools FacedBehind each number
Every finding here is one instance of the same practice: a policy question becomes a defensible number through data we check, a method we can defend, and code a stranger can rerun. Each analysis above links to the full write-up with its open code, and the wider methodology section shows how the numbers are built. The list grows as we publish.