is employed

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 employed.

Trained on survey data, it uses demographic features, voter registration tenure, and neighborhood employment patterns to estimate employment status.

Organizations can use these scores to identify unemployed or underemployed voters for outreach around job training, workforce development, and unemployment benefits.

  • approach: deep learning
  • training rows: 11,267
  • last updated: 2026-04-29 05:36:22

Validation

metric value
area under ROC curve 0.83
area under precision-recall curve 0.85
accuracy 0.76
precision 0.78
recall 0.76
f1 score 0.77
mean log loss 0.50
gini coefficient 0.67
brier skill score 0.33

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
10 139 0.94 0.91 0.03 1.75
9 138 0.86 0.87 0.01 1.60
8 138 0.79 0.82 0.03 1.47
7 138 0.79 0.74 0.05 1.47
6 138 0.59 0.62 0.03 1.10
5 138 0.49 0.46 0.02 0.90
4 138 0.38 0.35 0.02 0.70
3 138 0.28 0.27 0.01 0.52
2 138 0.17 0.20 0.04 0.31
1 138 0.09 0.13 0.04 0.17

Architecture

This model is a deep neural network trained with TensorFlow.

Network structure

layer units activation
↓ dense 128
↓ dropout (0.20 rate)
↓ dense 76

Hyperparameter configuration

  • loss: binary_crossentropy
  • optimizer: adam
  • batch size: 32
  • activation: sigmoid
  • label smoothing: 0.01
  • learning rate: 0.002

Features

Importance

Sourcing

Attom Property Data API

  • total assessed value

Revelio Labs

  • avg salary differential
  • avg tenure (days)
  • current position salary
  • current position tenure (days)

State voter files (via TargetSmart)

  • age

TargetSmart

  • children present score
  • climate change concern score
  • cord cutter score
  • early vote timing score
  • evangelical score
  • jewish score
  • marriage equality support score
  • midterm general turnout score
  • non-religious score
  • off-year general turnout score
  • paid leave support score
  • presidential general turnout score
  • race: aapi
  • race: latino
  • working class score

U.S. Census American Community Survey

  • block group % age 18–34
  • block group % female
  • block group households

Verisk (via TargetSmart)

  • self-employed

Implementation

Prediction distribution

statistic value
mean 0.67
std dev 0.27
min 0.05
25th 0.45
median 0.82
75th 0.88
max 0.98

Segments

segment lowest score highest score share of records
Most likely 0.87 0.98 32.6%
Somewhat likely 0.65 0.86 33.7%
Least likely 0.05 0.64 33.8%

External validity

Our model indicates that 67% of people with scored records are currently employed. As of April 2026, ~162M U.S. adults were employed — roughly 61% of the U.S. adult population (2026 BLS Employment Situation).