Oregon School Assessment

BA+ Signal in High-Poverty Schools: Summary Note

Generated: 2026-02-21

BA+ Signal in High-Poverty Schools: Summary Note
Generated: 2026-02-21

Question
Can we provide clear, plain-language numbers on how strongly school/community BA+ rates relate to proficiency in high-poverty schools, and compare that to low-poverty schools?

Data used
- data/processed/SchoolDataWithAddressesAndCensusSES.csv (ELA)
- data/processed/SchoolDataMathWithAddressesAndCensusSES.csv (Math)
- data/processed/SchoolDataScienceWithAddressesAndCensusSES.csv (Science)

Scope and filters
- School-level analysis, Total Population (All Students).
- Non-charter, non-virtual schools for comparability with prior work.
- Weighted by Number of Participants.
- School rows assembled by district/school ID; Science uses All Grades row where present, ELA/Math aggregate grade rows to school level.

Variable notes
- Outcome: Percent Proficient.
- Poverty: Students Experiencing Poverty (ODE At-a-Glance school-level field).
- BA+: ACS_Ed_BA_or_Higher_Rate (stored in the processed files as proportion values; converted to percent units in these tests).
- Attendance control (for adjusted tests): Attendance_Percent_Regular_Attenders.

What we tested
1) Statewide composition and poverty gradient context
- Statewide share check: the school-population proxy shows Students Experiencing Poverty at about one-third of students.
- Direct assessment subgroup rows added later put tested-participant poverty shares somewhat lower, at about 26.5% to 30.6% depending on subject; this memo still uses the school-level poverty field because the analysis is school-level.
- Statewide school-level poverty-proficiency gradients were quantified to answer: "roughly one-third of students in school-population terms, but how big is the statewide effect?"

2) BA+ strength inside high-poverty schools (bivariate)
- High poverty defined two ways:
  - >=40% Students Experiencing Poverty
  - >=50% Students Experiencing Poverty
- Reported weighted correlation and weighted slope (proficiency points per +10 BA+ points).

3) BA+ strength in high-poverty schools after controls (multivariate)
- Compared:
  - Base model: proficiency ~ poverty + attendance
  - Full model: proficiency ~ poverty + attendance + BA+
- Reported BA+ adjusted slope and added explanatory power (delta R^2).

4) Apples-to-apples low-poverty comparison
- Same model framework, low poverty defined as <20%.
- Also reported middle band 20-40% as a bridge between low and high.

Key findings

A) Statewide composition and effect
- Students Experiencing Poverty is a minority share statewide:
  - About 33.27% poverty vs 66.73% non-poverty (ADM-weighted school-pop proxy).
- Direct subgroup reminder:
  - Later assessment-group merges put tested-participant poverty shares below that proxy level, at roughly 26.5% to 30.6% depending on subject.
- Despite that, statewide poverty gradients are large:
  - Roughly -7.9 (ELA), -7.1 (Math), -6.6 (Science) proficiency points per +10 poverty points (non-charter/non-virtual estimates).

B) BA+ is clearly positive overall, but weaker inside high-poverty bands
- Overall BA+ slopes (all non-charter/non-virtual schools):
  - ELA +5.50 points per +10 BA+
  - Math +5.75 points per +10 BA+
  - Science +4.45 points per +10 BA+

- High poverty >=40% BA+ slopes:
  - ELA +1.64 per +10 BA+ (r=0.191)
  - Math +1.60 per +10 BA+ (r=0.203)
  - Science +1.46 per +10 BA+ (r=0.181)

- High poverty >=50% BA+ slopes:
  - ELA +0.97 per +10 BA+ (r=0.110)
  - Math +1.13 per +10 BA+ (r=0.146)
  - Science +0.96 per +10 BA+ (r=0.122)

C) Low-poverty comparison confirms a stronger BA+ signal
- Low poverty <20% BA+ slopes:
  - ELA +4.83 per +10 BA+ (r=0.585)
  - Math +5.53 per +10 BA+ (r=0.533)
  - Science +4.55 per +10 BA+ (r=0.491)

- Middle band 20-40% BA+ slopes:
  - ELA +2.33
  - Math +3.30
  - Science +1.62

Interpretation:
- BA+ matters in every band.
- The BA+ signal is strongest in low-poverty schools, weaker in middle-poverty schools, and weakest in high-poverty schools.
- In practical terms, high-poverty context appears to dampen (not erase) the BA+ gradient.

D) BA+ still contributes after controlling for poverty and attendance in high-poverty schools
- High poverty >=40% (adjusted BA+ slope; delta R^2):
  - ELA: +0.90 per +10 BA+; delta R^2 = +0.011
  - Math: +1.29 per +10 BA+; delta R^2 = +0.025
  - Science: +0.83 per +10 BA+; delta R^2 = +0.010

- High poverty >=50% (adjusted BA+ slope; delta R^2):
  - ELA: +0.75 per +10 BA+; delta R^2 = +0.007
  - Math: +1.01 per +10 BA+; delta R^2 = +0.017
  - Science: +0.76 per +10 BA+; delta R^2 = +0.009

Interpretation:
- Even after accounting for poverty level and attendance, BA+ remains positive.
- Independent BA+ lift is modest in high-poverty bands, strongest in Math, and smaller in ELA/Science.

Method cautions (keep in narrative)
- These are school-level observational associations, not causal claims.
- BA+ is a community context measure, not a direct measure of parents of tested students.
- Suppression and missingness rules still apply for some ODE fields.

Pitch-ready paragraph (Option 1 voice)
Even in high-poverty schools, places with more college-educated adults tend to do a bit better, but the lift is much smaller than in low-poverty schools. In our 2024-25 statewide school data, a 10-point increase in BA+ is linked to roughly 1.5 points of added proficiency in high-poverty schools, compared with about 5 points in low-poverty schools. Put simply: BA+ still matters in high-poverty communities, but poverty appears to blunt how much that advantage shows up in test outcomes.