uninsured

WarningStaging

Built from the latest training run on 2026-09-30 17:45 UTC, not from a released model set. The released documentation is at https://matter.indigo.engineering/docs/.

Overview

This model predicts whether a voter is currently uninsured — the policy-relevant minority in need of enrollment assistance and coverage-access outreach.

Trained on survey data, it uses age, employment, income, and area-level coverage indicators to identify voters without any current health insurance.

Organizations can use these scores to target enrollment-assistance campaigns, Medicaid and ACA marketplace outreach, and coverage-gap advocacy.

  • approach: boosted tree classifier
  • training rows: 17,117
  • last updated: 2026-09-06 19:39:03

Validation

metric value
area under ROC curve 0.76
area under precision-recall curve 0.15
accuracy 0.55
precision 0.11
recall 0.88
f1 score 0.20
mean log loss 0.91
gini coefficient 0.53
brier skill score 0.03

Lift Chart

Each bar shows how much more common the outcome is within a score decile than in the population overall. Bars taller than the dashed line (lift of 1) mean the model’s top-ranked groups concentrate the outcome; a steady decline from decile 1 to 10 indicates the model is ranking effectively.

Calibration Plot

Bars of similar height within each decile indicate well-calibrated predictions. Systematic differences reveal where the model over- or under-predicts.

Cumulative Gains

The further the model curve bows above the diagonal, the better the model concentrates the outcome in its top-ranked predictions.

Decile Analysis Table

decile count mean actual mean predicted mae lift
10 432 0.18 0.20 0.03 2.84
9 431 0.14 0.10 0.04 2.21
8 431 0.09 0.07 0.02 1.48
7 431 0.08 0.06 0.02 1.22
6 431 0.06 0.05 0.01 1.03
5 431 0.03 0.04 0.005 0.52
4 431 0.009 0.03 0.02 0.15
3 431 0.01 0.03 0.01 0.22
2 431 0.02 0.02 0.007 0.26
1 431 0.005 0.02 0.02 0.07

Architecture

This model is a boosted-tree classifier trained with BigQuery ML.

Hyperparameter configuration

  • max iterations: 20
  • learn rate: 0.2282
  • l1 regularization: 0
  • l2 regularization: 4.904e-11
  • min relative progress: 0.01
  • early stop: TRUE
  • input label columns: health_uninsured
  • data split method: AUTO_SPLIT
  • max tree depth: 4
  • subsample: 0.6395
  • min split loss: 0
  • num trials: 12
  • max parallel trials: 4
  • hparam tuning objectives: ROC_AUC
  • enable global explain: TRUE
  • category encoding method: LABEL_ENCODING
  • booster type: GBTREE
  • num parallel tree: 1
  • tree method: AUTO
  • min tree child weight: 1
  • instance weight column: sample_weight
  • xgboost version: 0.9
  • auto class weights: FALSE

Features

Importance

Interaction terms are split evenly between the features they combine.

Sourcing

Agency for Healthcare Research and Quality

  • tract % direct-purchase insurance only
  • tract unemployment rate

CDC PLACES

  • tract % adults 18-64 uninsured

Federal Reserve

  • county debt-to-income decile

Georgetown University Center for Children and Families

  • county medicaid / chip coverage rate

Revelio Labs

  • avg salary differential
  • current position salary
  • current position tenure (days)
  • min salary
  • number of jobs held

State voter files (via TargetSmart)

  • age
  • years registered to vote

TargetSmart

  • campaign finance reform score
  • catholic score
  • high school only score
  • marriage score
  • midterm general turnout score
  • non-presidential primary turnout score
  • off-year general turnout score
  • path to citizenship support score
  • race: white

U.S. Census American Community Survey

  • block group % age 50–64
  • block group % high school diploma

Verisk (via TargetSmart)

  • clerical occupation
  • newspaper reader

Implementation

Prediction distribution

statistic value
mean 0.09
std dev 0.07
min 0.01
25th 0.04
median 0.07
75th 0.11
max 0.80

Segments

segment lowest score highest score share of records
Most likely 0.18 0.80 9.3%
Somewhat likely 0.13 0.17 9.1%
Least likely 0.01 0.12 81.5%

External validity

Our model indicates that 9% of people with scored records are currently uninsured. As of 2024, ~23.5M U.S. adults were uninsured — roughly 9% of the U.S. adult population (2024 NCHS National Health Interview Survey).