Adult BA+ and Student Poverty Both Matter

Short Answer

School poverty explains a large share of school-to-school proficiency differences in ELA, Math, and Science. But adult BA+ does not disappear when student poverty is included in the same model. The cleaner reading is not "BA+ is the single dominant factor." It is: adult BA+ is a strong community-context predictor, and school poverty is a strong indicator of hardship among enrolled students. The two overlap, but neither fully replaces the other.

What We Modeled

The model is intentionally simple: Percent Proficient is predicted from two variables, ACS adult BA+ rate and ODE's school-sourced Students Experiencing Poverty measure. Elementary-only eligibility is classified first: a school's combined observed ELA, Math, and Science tested-grade set must be nonempty and wholly within grades 3-5. Each subject outcome then comes from that subject's single official Total Population All Grades row; grade rows are never pooled. Outcomes and model weights use the exact Scored Performance Denominator; complete Level 1-4 counts must sum to it and proficiency must equal Level 3 plus Level 4. The primary view excludes charter, virtual, and schools identified by the ETL's effective-dated special-enrollment policy overlay.

School-poverty disclosure status and bounds are validated. Exact values must equal both bounds. ODE inequality disclosures such as <5% use their documented interval midpoint for computation (2.5% in that example), require the retained original inequality token, and preserve the legal bounds. These midpoint values are explicit working values, not exact observations.

What Each Measure Adds After the Other

Primary Results

Weaker-Geographic-Linkage Sensitivities

The earlier draft treated statewide All Grades results as primary. This correction makes elementary grades primary because school-site tract SES is more plausibly linked to the enrolled population. Official All Grades and pooled middle-grade models remain useful sensitivities, but their tract linkage is weaker.

How to Read the Table

• Correlation r is the simple weighted association before controls. BA+ is positive because higher BA+ is associated with higher proficiency. Poverty is negative because higher poverty is associated with lower proficiency.
• Standardized beta compares each predictor on a common scale inside the same model. Larger absolute values indicate stronger unique association after the other predictor is held fixed.
• Added R² asks how much one measure improves the model after the other measure is already included.

Interpretation

In the elementary-primary analysis, adult BA+ remains a strong indicator of community educational context, but school poverty carries more additional descriptive explanatory power in 2024-25 Oregon data. The right conclusion is not that one measure makes the other irrelevant. It is that statewide performance patterns are better described with both: one broad community-education measure and one enrolled-student hardship measure.

Related Evidence

• SES explorations summary: a synthesis of income, education, poverty, and other SES proxy checks.
• BA+ association in high-poverty schools: a descriptive banded analysis showing that the BA+ slope remains positive across the reported poverty bands; it does not supply a formal test of slope differences.
• Why poverty can outpredict income: a plain-language explanation of why a school-population poverty field can outperform a tract-level income measure.
• Income vs poverty reassessment: a poverty-aware model comparison using BA+, attendance, income, and Students Experiencing Poverty.
• Two-factor model source script: the commented reproducibility script for the model summarized on this page.
• Two-factor model memo: script-generated scope notes and a rounded result table.
• BA+-residuals vs school poverty summary CSV and companion figure: a residual check showing that high-poverty schools tend to fall below the BA+-only expectation line.
• Residual-check source script: the reproducibility script for the companion residual analysis.