Skip to content

brycewang-stanford/Auto-Empirical-Research-Skills / skills/17-DAAF-Contribution-Community-daaf/dot-claude/skills/education-data-source-scorecard

education-data-source-scorecard

Verified

College Scorecard — post-enrollment outcomes linking aid records to IRS/Treasury earnings. Earnings, loan repayment, debt via six Portal sub-datasets. Use when tax-record-based earnings needed. Tracks only Title IV aid recipients, not all students.

references/earnings-data.md

View on GitHub
# Earnings Data

College Scorecard's earnings data represents actual post-college earnings from IRS W-2 records, making it uniquely valuable for understanding labor market outcomes.

> **Portal dataset:** `scorecard/colleges_scorecard_earnings` (203,066 rows x 33 columns)
>
> **Portal uses lowercase LONG format.** Original Scorecard names like `MD_EARN_WNE_P6` become `earnings_med` filtered by `years_after_entry == 6`. See `variable-definitions.md` for the full mapping.
>
> **Suppression:** Earnings columns use `-3` integer code (NOT null) for suppressed values.

## Data Source: IRS/Treasury

Earnings come from IRS W-2 wage records matched to students via Social Security numbers:

| Aspect | Detail |
|--------|--------|
| Source | IRS W-2 forms (employer-reported wages) |
| Matching | Treasury matches NSLDS records to tax filings |
| Reporting | Aggregated to protect individual privacy |
| Coverage | Employed individuals with W-2 income only |

### What W-2 Earnings Include

- Wages, salaries, tips
- Employer-reported compensation
- Most full-time and part-time employment

### What W-2 Earnings Exclude

| Excluded Income | Impact on Analysis |
|-----------------|-------------------|
| Self-employment (1099) | Entrepreneurs, gig workers missing |
| Business income | Small business owners missing |
| Investment income | Passive income not counted |
| Non-reported income | Cash economy excluded |
| Foreign earnings | Those working abroad missing |

## Cohort Definitions

### Entry Cohort vs Completion Cohort

Scorecard primarily uses **entry cohorts** - students grouped by when they first enrolled:

```
Entry Cohort Example:
- Students who entered in Fall 2008
- 6-year earnings measured in 2014
- 10-year earnings measured in 2018
```

**Important:** Earnings are measured from entry, NOT graduation. This means:
- A student who took 4 years to graduate has earnings measured 2 years post-graduation at the "6-year" mark
- A student who took 6 years to graduate has earnings measured AT graduation at the "6-year" mark

### Entry Cohort Construction

Students included in entry cohorts:
- First-time undergraduates who received Title IV aid
- Includes full-time AND part-time students
- Includes students who transferred in (at receiving institution)
- Includes students who did not complete

### Pooled Cohorts

Many metrics use "pooled" cohorts combining multiple years:
- Increases sample size
- Reduces suppression
- Typical pooling: 2-4 entry years combined

Variable naming convention: `*_POOLED` or `*_POOLED_SUPP`

## Time Horizons

### Available Measurement Points

| Metric | Years After Entry | What It Captures |
|--------|------------------|------------------|
| 6-year earnings | 6 years | Early career, ~2 years post-graduation for 4-year completers |
| 8-year earnings | 8 years | Early-mid career |
| 10-year earnings | 10 years | Mid-career, ~6 years post-graduation for 4-year completers |

### Interpreting Time Horizons

**6-Year Earnings:**
- Most commonly reported metric
- Captures early career earnings
- For 4-year completers: ~2 years post-graduation
- For 2-year completers: ~4 years post-graduation
- For non-completers: earnings regardless of completion status

**10-Year Earnings:**
- Better reflection of career trajectory
- More stable earnings picture
- Fewer students still in graduate school
- But: cohorts are even older (11+ years old when data released)

### Data Vintage and Lag

Earnings data has significant lag:

```
Example for 2024 data release:
├─ 10-year earnings
│   ├─ Entry cohort: ~2014
│   ├─ Measured in: ~2024
│   ├─ Reflects: Students who entered 10+ years ago
│   └─ Labor market: 2024 conditions applied to old cohort
└─ 6-year earnings
    ├─ Entry cohort: ~2018
    ├─ Measured in: ~2024
    └─ Reflects: Students who entered 6+ years ago
```

**Implication:** Earnings data reflects students and programs from years ago. Rapid changes in:
- Program quality
- Curriculum
- Labor market conditions
- Industry demand

...may not be reflected in current data.

## Earnings Metrics

### Median Earnings (Primary Metric)

> **Portal column names are lowercase.** Filter by `years_after_entry` to select time horizon.

| Portal Column | Description | Original Scorecard |
|---------------|-------------|-------------------|
| `earnings_med` | Median earnings (filter by `years_after_entry`) | `MD_EARN_WNE_P6/P8/P10` |
| `earnings_mean` | Mean earnings | `MN_EARN_WNE_P*` |

"WNE" (Working and Not Enrolled) is implicit in the Portal data — these columns only include workers.

Available `years_after_entry` values: **6, 7, 8, 9, 10**

### Percentile Earnings

| Portal Column | Description | Original Scorecard |
|---------------|-------------|-------------------|
| `earnings_pct10` | 10th percentile earnings | `PCT10_EARN_WNE_P*` |
| `earnings_pct25` | 25th percentile earnings | `PCT25_EARN_WNE_P*` |
| `earnings_pct75` | 75th percentile earnings | `PCT75_EARN_WNE_P*` |
| `earnings_pct90` | 90th percentile earnings | `PCT90_EARN_WNE_P*` |

### Earnings by Family Income

Disaggregated by FAFSA-reported family income tercile:

| Portal Column | Family Income Level | Original Scorecard |
|---------------|--------------------|--------------------|
| `earnings_lowinc_mean` | Lowest third ($0-$30,000) | `MN_EARN_WNE_INC1_P*` |
| `earnings_midinc_mean` | Middle third ($30,001-$75,000) | `MN_EARN_WNE_INC2_P*` |
| `earnings_highinc_mean` | Highest third ($75,001+) | `MN_EARN_WNE_INC3_P*` |

### Additional Disaggregations

| Portal Column | Description |
|---------------|-------------|
| `earnings_dep_mean` | Mean earnings, dependent students |
| `earnings_dep_lowinc_mean` | Mean earnings, dependent low-income |
| `earnings_ind_mean` | Mean earnings, independent students |
| `earnings_female_mean` | Mean earnings, female |
| `earnings_male_mean` | Mean earnings, male |

### Working and Enrollment Status

| Portal Column | Description | Original Scorecard |
|---------------|-------------|-------------------|
| `count_working` | Number working and not enrolled | `COUNT_WNE_P*` |
| `count_not_working` | Number not working and not enrolled | `COUNT_NWNE_P*` |
| `earnings_greater_than_25k_pct` | Share earning > $25K | `GT_25K_P*` |

## Working Population Filter

**Critical:** Earnings are only calculated for students who are:
1. Working (have positive W-2 earnings)
2. Not enrolled in postsecondary education

### Who Is Excluded from Earnings Calculations

| Excluded Group | Reason | Impact |
|----------------|--------|--------|
| Graduate school students | Enrolled, not working | Fields with high grad school rates (e.g., biology, pre-law) show higher earnings than they would if grads included |
| Non-employed | No W-2 income | Unemployed, stay-at-home parents, disabled excluded |
| Self-employed only | No W-2 | Entrepreneurs missing |
| Deceased | No W-2 | Small effect |
| Living abroad | No W-2 | May affect certain fields |

### Implications for Analysis

- Fields with high graduate school attendance may appear to have HIGHER earnings (low earners are still in school)
- Institutions/programs with higher unemployment rates are NOT penalized in earnings metrics
- Self-employment heavy fields (arts, consulting) may show lower median earnings

## Earnings Suppression

Earnings are suppressed when privacy thresholds are not met:

### Suppression Rules

| Threshold | Action |
|-----------|--------|
| < 30 students with positive earnings | Earnings suppressed |
| Small cells in disaggregations | Additional suppression |

### Variables Indicating Suppression

> **Portal Encoding:** In the Portal earnings dataset, **`-3` is the primary suppression indicator** for earnings and count columns (NOT null). This was verified empirically — ~24,000 rows have `-3` in `earnings_mean` alone.

| Data Pattern | Meaning | Notes |
|--------------|---------|-------|
| `-3` | Suppressed for privacy | Primary suppression indicator in earnings columns |
| `null` | Missing data | Null count is 0 for `unitid`, `year`, etc. but ~109K for disaggregated columns |
| Positive value | Valid earnings | Actual data |

```python
import polars as pl

# CORRECT: Filter both -3 and null for earnings
valid = df.filter(
    pl.col("earnings_med").is_not_null() &
    (pl.col("earnings_med") != -3) &
    (pl.col("earnings_med") > 0)
)
```

### High Suppression Rates Affect

- Small institutions
- Small programs (by CIP code)
- Disaggregations (by income, race, gender)
- Newer programs without enough cohorts

## Cohort Year Mapping

Understanding which cohort corresponds to which measurement year:

### Institution-Level Earnings

For annual data files (e.g., `merged_2022_23`):

| Variable | Entry Cohort Years (Approx) | Measurement Year |
|----------|---------------------------|------------------|
| 6-year earnings | 2015-16, 2016-17 | 2021, 2022 |
| 8-year earnings | 2013-14, 2014-15 | 2021, 2022 |
| 10-year earnings | 2011-12, 2012-13 | 2021, 2022 |

**Note:** Specific cohort years vary by data release. Always check documentation.

### Data Errata

**Known Issue (2022-2023):** Entry-cohort earnings calculations in some files were calculated with misaligned cohorts. Check the College Scorecard errata document for affected variables.

## Recommended Practices

### Before Earnings Analysis

1. **Identify cohort timing** - Know which entry years are reflected
2. **Check suppression rates** - How many institutions/programs have data?
3. **Note WNE filter** - Earnings exclude non-workers and enrolled students
4. **Understand coverage** - Title IV recipients only

### Reporting Earnings

1. **State population clearly**: "Title IV aid recipients who were working and not enrolled"
2. **Note time context**: "Earnings measured 6 years after enrollment"
3. **Acknowledge lag**: "Reflects students who entered in [year]"
4. **Report suppression**: "X% of programs had suppressed earnings data"

### Comparing Earnings

**Valid comparisons:**
- Same institution type, similar aid populations
- Same field across similar institution types
- Same institution over time (with cohort caveats)

**Invalid comparisons:**
- Scorecard earnings vs BLS occupational wages
- Scorecard earnings vs Census income
- Scorecard vs alumni survey earnings
- Institutions with very different Title IV coverage rates

## Example: Interpreting 6-Year Earnings

```
Institution A: Median 6-year earnings = $45,000

What this means:
- Half of Title IV aid recipients earned more, half less
- Measured 6 years after first enrollment
- Only includes those working and not enrolled
- Excludes students in graduate school
- Excludes unemployed, self-employed, those abroad
- Reflects cohort that entered ~6 years ago
- May not reflect current program quality or labor market
```
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall