Civic Forecast

Public data, handled with care.

Civic Forecast is a research and civic-education project from API Automations. We normalize public, licensed, and government data into source-attributed, aggregate civic signals, and we hold every source to the same governance, attribution, and privacy commitments.

What the pipeline actually carries

Governance is a claim about how something is run, and it is only meaningful once you know how much there is. These four figures are read from the platform itself rather than written here, so they cannot fall out of step with it.

A governed data pipeline

Every source moves through an auditable pipeline: a repeatable pattern of dataset → normalized signal → public attribution → civic context and alerts. Each layer is governed, and no layer re-exposes source records.

Civic Forecast Findings

The research output the pipeline exists to produce: figures Civic Forecast computes itself, from two or more datasets it already holds, that neither source publishes on its own. BLS knows what an hour of work pays; HUD knows what a two-bedroom costs. Neither publishes what share of a paycheck the rent takes, because neither holds both halves. Civic Forecast does, so it can.

The discipline behind a Finding

A derived figure that names two trusted publishers and is quietly wrong is worse than no figure at all. So every Finding names its inputs, aligns them to the same period before combining them, withholds rather than defaulting when an input is missing, and is recomputed as publishers revise.

See the formula behind each Finding

Findings are our data posture in practice: facts are not copyrightable, so where we can compute our own statistic from the raw data we do: better journalism and a smaller licensing footprint than republishing someone else's. It is also why a partner's dataset is often most valuable in combination: most data is inert on its own and says something the moment it meets another source.

The complete set

Those four are illustrations, not the inventory. Every Finding the platform publishes is listed below, generated from the platform itself rather than maintained on this page, so it cannot quietly fall behind what the app is showing.

43 Findings publish today: 29 measures, 14 of which are also mapped state by state.

Housing and rent (13 Findings)

Prices, rates and debt (8 Findings)

Democracy and influence (11 Findings)

Work and pay (5 Findings)

Health and poverty (4 Findings)

Environment (2 Findings)

Publishing the work

Findings are computed to be used in the app. Some of them are also worth putting in front of peer review, where a method either survives being read by people paid to find its faults or it does not. That is a program with gates, not an intention.

What can never be published

Civic Forecast holds statewide voter files for a handful of states, acquired lawfully for postal outreach. They are kept on an operator workstation, never reach the app or the API, and cannot feed a paper: not in aggregate, not with names removed, not at all. Being aggregate does not convert a restricted record into a publishable one. A manuscript naming such an input stops and goes to counsel for a written determination and a separate, named purpose. There is no expedited path.

Everything published is computed from open federal data and state-published aggregates: HUD, BLS, Census, CDC, FEC, Congress.gov and EPA ECHO, plus North Carolina turnout and Florida registration, which the platform already ingests.

Five gates, in order

What a university partner changes

The cleanest route through the ethics gate is an academic collaborator whose institution acts as review board of record. That is a real dependency rather than a courtesy, and we would rather say so: it turns the conversation from asking whether we can help into establishing that we need each other. Partners get a documented method, a governed pipeline and a corpus that is already computed, not a data dump to clean.

Talk about a research collaboration

Data governance & privacy