enrolled in chip

Overview

This model predicts enrollment in the Children’s Health Insurance Program (CHIP), which provides coverage to children in families that earn too much to qualify for Medicaid but cannot afford private insurance.

Trained on survey data, it uses household composition signals, economic indicators, and neighborhood demographics to identify likely CHIP households.

Organizations can use these scores for outreach around children’s health coverage and family benefits access.

  • approach: boosted tree classifier
  • training rows: 16,076
  • last updated: 2026-09-02 16:37:30

Validation

metric value
area under ROC curve 0.76
area under precision-recall curve 0.12
accuracy 0.52
precision 0.05
recall 0.83
f1 score 0.10
mean log loss 0.95
gini coefficient 0.51
brier skill score 0.04

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
10 399 0.11 0.11 0.004 3.38
9 398 0.07 0.04 0.02 2.00
8 398 0.05 0.03 0.02 1.39
7 398 0.03 0.02 0.004 0.85
6 399 0.02 0.02 0.0008 0.69
5 398 0.02 0.02 0.003 0.54
4 398 0.01 0.02 0.007 0.38
3 398 0.02 0.02 0.004 0.46
2 398 0.002 0.02 0.02 0.08
1 399 0.008 0.02 0.009 0.23

Architecture

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

Hyperparameter configuration

  • max iterations: 20
  • learn rate: 0.1813
  • l1 regularization: 1.497e-09
  • l2 regularization: 4.699
  • min relative progress: 0.01
  • early stop: TRUE
  • input label columns: benefits_enrolled_in_chip
  • data split method: AUTO_SPLIT
  • num trials: 12
  • max parallel trials: 4
  • hparam tuning objectives: ROC_AUC
  • enable global explain: TRUE
  • auto class weights: FALSE
  • max tree depth: 3
  • subsample: 1
  • min split loss: 0
  • 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

Features

Importance

Interaction terms are split evenly between the features they combine.

Sourcing

Revelio Labs

  • current position salary

State voter files (via TargetSmart)

  • age
  • registered democrat
  • registered republican

TargetSmart

  • abortion issue motivation score
  • child tax credit support score
  • children present score
  • college graduate score
  • double negative gop score
  • harris support score
  • income rank
  • midterm general turnout score
  • paid leave support score
  • partisan score
  • progressive tax support score
  • race: aapi
  • race: black
  • race: white
  • working class score

U.S. Census American Community Survey

  • block group % aapi
  • block group % some college

Verisk (via TargetSmart)

  • mobile home
  • net worth
  • residence tenure
  • residence year built

Implementation

Prediction distribution

statistic value
mean 0.06
std dev 0.07
min 0.02
25th 0.02
median 0.03
75th 0.09
max 1

Segments

segment lowest score highest score share of records
Most likely 0.17 1 5.7%
Somewhat likely 0.14 0.16 5.7%
Least likely 0.02 0.13 88.6%

External validity

Our model indicates that 6% of people with scored records have a child enrolled in CHIP. CMS Medicaid/CHIP Enrollment and Census ACS 2023 place household CHIP enrollment at roughly 2–3% of U.S. adults.