Azimuth College Rankings — Technical Methodology Appendix (v3)

Azimuth's technical methodology describes how we transform raw federal datasets into the four pillar scores (Return, Affordability, Access, and Economic Mobility) and the composite ranking. The goal is to evaluate how colleges support students—especially low-income and first-generation students—using publicly available, consistently measured indicators.

This appendix provides detailed explanations of the statistical steps, fallback rules, weighting choices, and institutional coverage decisions used in Azimuth Methodology v3.

The public-facing version is available at /methodology.

Azimuth ranks institutions where comparisons are valid and data coverage is sufficient.

2.1 Included

• Public, four-year, degree-granting institutions
• Private nonprofit, four-year institutions
• Institutions with enough data to calculate stable scores across the four pillars

2.2 Excluded (for now)

• For-profit institutions
• Two-year colleges and certificate-only institutions

Rationale:

These institutions serve different functions, student populations, and missions. Federal reporting requirements also differ, and long-run outcome measures are not directly comparable. We expect to revisit these groups in future iterations when comparative methods stabilize.

3.1 College Scorecard (April 2025 Release)

Primary source for:

• Earnings at 6-, 8-, and 10-year windows
• Program-level earnings (CIP-coded)
• Net price by income band
• Federal loan and Parent PLUS borrowing
• Repayment and default indicators
• Graduation rates
• Pell share, first-generation indicators, and other demographic variables

Where multiple windows exist, we select the most recent 10-year measure. If missing, we fall back to 8-year, then 6-year.

3.2 CAP Longitudinal Loan Repayment Dataset

Built via FOIA and published by the Center for American Progress, this dataset tracks a single federal borrower cohort (entered repayment in 2012) across five years.

It assigns borrowers to:

• Paid in full
• Paid down / on track
• No progress / flat balance
• Delinquency
• Default
• Deferment / forbearance

This dataset enables long-run repayment assessment not available in standard Scorecard default data.

Reference: Can We See a College's Long-Term Default Rate? (CAP).

3.3 Mobility and Context Data

We incorporate Carnegie-style mobility indicators and state-level cost-of-living indices. Cost-of-living is only used to contextualize post-graduation earnings—not tuition or net price.

Each pillar uses the following general method:

1. Identify core indicators
2. Apply consistent fallback rules
3. Standardize to a common scale
4. Combine using transparent weights
5. Apply mild outlier controls

Below are details for each pillar.

4.1 College Investment (ROI)

Goal

Capture both absolute earnings and how much colleges improve earnings relative to expectation.

Components

A. Absolute Earnings

• Use median earnings 10 years after entry
• Fallbacks: 8-year → 6-year
• Ensures institutions with missing windows remain comparable

B. Earnings Beyond Expectations

We estimate expected earnings based on characteristics of similar students nationwide.

We then compute:

earnings beyond expectations = actual earnings – predicted earnings

This isolates the college's contribution to long-term earnings.

Explainable factors include:

• Student demographic mix
• Regional context
• Preparation proxies where provided in Scorecard

C. Earnings Trajectory

Where available, we evaluate change between 6-, 8-, and 10-year windows. Institutions with positive earnings momentum score higher than those where outcomes plateau.

D. Repayment Behavior (CAP Dataset)

We focus on:

• Share who pay down balances
• Share who pay in full
• Share in delinquency or default
• Five-year repayment health overall

This reduces reliance on short-term default metrics.

Standardization & Scoring

All metrics are standardized (z-scored) and combined using weights reflecting empirical stability and interpretability. Outliers are gently limited (soft-winsorized) to prevent domination by extreme values.

Goal

Reflect what college costs after grant aid and how manageable debt is.

Components

A. Net Price by Income Level

We use three bands:

• Low-income
• Middle-income
• Higher-income

Each is compared against national peers.

B. Student Debt

We evaluate typical undergraduate loan amounts. We incorporate distributional indicators where available.

C. Parent PLUS Borrowing

We use average Parent PLUS volume to capture household risk beyond student loans.

D. Debt Sustainability

Repayment trajectories inform whether debt levels appear manageable.

Standardization

Net price and debt indicators are normalized relative to national distributions. High list price is not penalized; only net price matters.

Goal

Reward institutions that create broad educational opportunity.

Components

• Pell share
• First-generation student share (where available)
• Transfer student pathways
• Scale of impact: number of low-income students actually served
• Resource context (serving many Pell students despite modest institutional wealth)

Standardization

Access metrics are adjusted for institutional type and context when necessary.

Goal

Measure upward mobility specifically for low-income students.

Components

A. Low-Income Earnings

Median earnings 10 years after entry for low-income students.

B. Earnings Uplift

Comparison of low-income earnings to expected values and/or high-income peers.

C. Low-Income Graduation Rates

Degree completion is a driver of mobility.

D. Combined Mobility Index

Aggregates:

• Outcomes
• Access
• Number of low-income students served

Mobility Quadrants (Interpretive Layer)

Institutions are grouped into:

• Mobility Engines
• Selective Achievers
• Opportunity Builders
• Under-Resourced Institutions

This helps contextualize institutional profiles, not rank them.

5.1 Z-Scoring

Each metric is transformed to a mean-zero, unit-scale measure to enable equal-weight combination.

5.2 Outlier Control (Soft Winsorization)

Extremely high or low values are gently limited so no single institution dominates.

5.3 Fallback Rules

If a metric is missing but a close substitute exists (e.g., 10-year → 8-year earnings), we use the fallback.

5.4 Neutral Imputation

If no fallback exists, we use the national median for that metric. This avoids penalizing institutions simply for reporting gaps.

5.5 Coverage Requirement

Institutions must have enough data across pillars to support stable rankings. If not, they are excluded from the publish set.

Weights:

• 30% Mobility
• 30% Return
• 20% Access
• 20% Affordability

Each pillar score is normalized before combination.

Alternative Lenses

• Outcome Lens – heavier weight on Return
• Affordability Lens – heavier weight on net price/debt
• Mobility Lens – heavily weights Economic Mobility + Access

• Data describes past cohorts (2015–2018 entry), not guaranteed future outcomes
• Earnings depend on field of study, labor markets, and student choices
• Some institutions report more completely than others
• Rankings summarize complex realities and should be interpreted with nuance

We recommend:

• Reviewing institution pages
• Examining program-level pages when available
• Considering non-quantitative factors (fit, support, geography, etc.)

This specification describes Azimuth Methodology v3, aligned to the April 2025 Scorecard release.

Future updates may revise:

• Component weights
• Mobility construction
• Treatment of missing data
• Enrollment and cohort definitions
• Program-level modeling

Feedback is welcome at: info@collegeazimuth.com

Data sources:

• U.S. Department of Education — College Scorecard (April 2025 Release)
• Center for American Progress — Five-Year Loan Repayment Trajectory Dataset
• Federal demographic and earnings indicators
• Independent analysis and modeling by Azimuth

Azimuth is not affiliated with the U.S. Department of Education or the Center for American Progress.