EconomicIQ · County & city economic intelligence
See your economy as a living system.
Ask your county a question and get an answer built from 30 public datasets, loaded for that county — reconciled against each other and traced back to the agency that published every number.
Explore how jobs, housing, income, business, climate and public records connect. Every relationship leads back to evidence.
One question
Every number arrives with its margin
The assistant does not summarize a PDF. It reads the county's loaded data, and each figure it returns carries the address of the cell it came from — so the question "where did that come from?" has an answer before anyone asks it.
“Is our housing stock shrinking?”
ACS.2019_2023.37121
.housing_units
No. Within one consistent control base the stock rose by 34 units, and the margin of error on that is ±61 — wide enough that the direction itself is not settled.
REGISTER
.housing-stock-decline-134
The 134-unit decline you may have seen quoted is a census control-base switch between vintages, not a real loss of housing. It is on the register of claims this system will not make.
Wording is illustrative. The cell addresses, the register entry and the ±61 margin are the real ones.
Inside the portal
Six ways into the same evidence
One county, one set of loaded sources, six views over it. Nothing here is a separate export that can drift from the others — the report, the graph and the assistant all read the same cells.
-
Report
The full written edition
A governed county report. Every figure in it resolves to a source cell, and every claim it makes is one the underlying data supports.
-
Analyst
Ask the county its own questions
Ask in plain language. The assistant answers over this county’s data, cites what it used, and declines what the data cannot carry.
-
Relationships
Sources to cells to claims
The graph behind the report. Follow any number back through the cell it came from to the source that published it.
-
Indicators
Every indicator with its evidence trail
The measures themselves — each one with the vintage, the geography, and the margin of error it was published with.
-
Compare
Structural peers, side by side
Counties matched on structure rather than proximity, so the comparison is to a place with the same economy shape.
-
Sources
What is held, and what it covers
The register of what is loaded for this county, what each source covers, and where coverage stops.
The register
The useful part is what it will not say
A confident wrong number is worse than no number, because it survives into a grant application, a council packet, or a site-selection memo. So the claims this system must never make are written down as data — 7 named false claims and 8 data traps that produce a wrong number even from real source data. They are enforced when a tool answers, and screened again on the way out.
-
False claim
Housing stock shrank by 134 units.
housing-stock-decline-134
A census control-base switch between ACS vintages, not a decline. Within one consistent base the stock rose 34 units — against a margin of error of 61, which does not even settle the sign.
-
Data trap
This county’s fair-market rent rose to the state floor.
hud-fmr-statewide-floor
The published figure is a statewide floor applied to the county, not a county measurement. Reading it as local rent movement invents a trend the source never reported.
-
Data trap
Each mining job supports 3.4 other jobs.
no-economic-multiplier-ever
No employment multiplier exists in anything loaded here. A plausible number with no source behind it is the most expensive kind of wrong.
The register is data, not a code branch: a newly found trap is an entry plus a test. When a question lands on one, the answer says so and stops, rather than reaching for the nearest plausible figure.
Coverage
30 datasets, and an honest edge
30 datasets are loaded per county, and every one of them is backed by real rows rather than a listing. That is a smaller number than the full Agentic Data Mining catalog on purpose: a dataset keyed on a stock number, a patent assignee or an SEC registrant has no county in it, and assigning one would be the first false claim on the register.
Every source clears licence review and a QA gate before it is loaded, and new ones join as they clear. Coverage still stops somewhere for every county, and the Sources view says where — what is held, what vintage it is, which of it this edition actually cites, and what it does not reach.