DAAF-Contribution-Community/daaf / .claude/skills/education-data-source-scorecard
education-data-source-scorecard
VerifiedCollege 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.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 22.3 KB |
| references/completion-rates.md | markdown | 10.5 KB |
| references/data-quality.md | markdown | 11.5 KB |
| references/debt-repayment.md | markdown | 11.6 KB |
| references/earnings-data.md | markdown | 11.6 KB |
| references/field-of-study.md | markdown | 12.4 KB |
| references/population-coverage.md | markdown | 9.5 KB |
| references/variable-definitions.md | markdown | 25.3 KB |
references/field-of-study.md
View on GitHub# Field of Study Data
College Scorecard provides program-level data on earnings and debt by field of study (CIP code), enabling analysis of outcomes at the major/program level.
> **Portal boundary:** Field-of-study data is **not mirrored by the Education Data Portal**. The six Portal Scorecard families are institution-level earnings, default, repayment, institutional characteristics, and two student-body families. For current program-level data, use the stable official page `https://collegescorecard.ed.gov/data/` and choose **Most Recent Data by Field of Study**, or use the documented Scorecard API. Do not rely on a date-stamped ZIP as the sole route. Variable names below describe direct Scorecard products and may differ by release; verify the current file/API documentation before use.
## Overview
Field of study data links outcomes to specific academic programs:
| Aspect | Detail |
|--------|--------|
| Granularity | By CIP code + credential level + institution |
| Earnings timing | Post-completion (not post-entry like institution-level) |
| Coverage | Completers with Title IV aid |
| Suppression | High - many programs below threshold |
## Data Structure
### Key Identifiers
| Variable | Description |
|----------|-------------|
| `UNITID` | Institution ID |
| `CIPCODE` | Classification of Instructional Programs code |
| `CIPDESC` | CIP description |
| `CREDLEV` | Credential level |
| `CREDDESC` | Credential description |
### CIP Code Levels
| Level | Format | Example |
|-------|--------|---------|
| 2-digit | XX | 11 (Computer Science) |
| 4-digit | XX.XX | 11.01 (Computer Science General) |
| 6-digit | XX.XXXX | 11.0101 (Computer Science) |
### Credential Levels
> **Direct-source encoding:** These credential codes apply to the direct field-of-study product described here. Verify them against the current Scorecard file/API documentation; this is not a Portal dataset.
| Code | Description |
|------|-------------|
| 1 | Undergraduate Certificate or Diploma |
| 2 | Associate's Degree |
| 3 | Bachelor's Degree |
| 4 | Post-baccalaureate Certificate |
| 5 | Master's Degree |
| 6 | Doctoral Degree |
| 7 | First Professional Degree |
| 8 | Graduate/Professional Certificate |
## Important Difference: Post-Completion Timing
**Critical:** Field of study earnings are measured from **completion**, not entry.
| Level | Institution Earnings | Field Earnings |
|-------|---------------------|----------------|
| Reference point | Entry (estimated entry year at the institution) | Completion |
| Includes non-completers | Yes | No |
| Time frame meaning | "6 years after entry" | "x years after completion"; the FoS technical documentation (Version: October 2026) describes 1-7 years (`EARN_MDN_HI_1YR`/`EARN_MDN_1YR`, `EARN_MDN_HI_2YR`, `EARN_NE_MDN_3YR`, `EARN_MDN_4YR`/`5YR`/`7YR`) plus national 4/5/7-year values (`EARN_MDN_xYR_NAT`); check which are populated in your release |
### Field-Level Time Frames
| Variable | Description |
|----------|-------------|
| `EARN_MDN_HI_1YR` | Median earnings, 1 year post-completion |
| `EARN_MDN_HI_2YR` | Median earnings, 2 years post-completion |
| `EARN_COUNT_WNE_HI_1YR` | Count earning 1 year post-completion |
| `EARN_COUNT_WNE_HI_2YR` | Count earning 2 years post-completion |
## Field-Level Variables
### Earnings
| Variable | Description |
|----------|-------------|
| `EARN_MDN_HI_1YR` | Median earnings at 1 year |
| `EARN_MDN_HI_2YR` | Median earnings at 2 years |
| `EARN_NE_MDN_3YR` | Median earnings at 3 years (if available) |
| `EARN_COUNT_WNE_HI_*` | Count with positive earnings |
### Debt
| Variable | Description |
|----------|-------------|
| `DEBT_ALL_STGP_ANY_MDN` | Median debt, all sources |
| `DEBT_ALL_STGP_EVAL_MDN` | Median debt, evaluated programs |
| `DEBT_ALL_PP_ANY_MDN` | Median debt, parent PLUS included |
### Completion Counts
| Variable | Description |
|----------|-------------|
| `NUM_AWARDS` | Number of awards conferred (illustrative variable name; verify in the data dictionary) |
| `IPEDSCOUNT1` | IPEDS completions count |
| `IPEDSCOUNT2` | IPEDS completions, alternate count |
### Program Status
| Variable | Description |
|----------|-------------|
| `MAIN` | Main campus indicator |
| `BRANCH` | Branch campus indicator (illustrative variable name; verify in the data dictionary) |
| `DISTANCE` | Distance education indicator for the field of study (not reported to IPEDS / no / some / all credentials completable via distance education; check value codes in the data dictionary) |
## Suppression in Field Data
### High Suppression Rates
Field-level data has much higher suppression than institution-level:
(Illustrative, unsourced ranges — measure suppression in your extract.)
| Data Type | Typical Suppression |
|-----------|-------------------|
| Institution-level earnings | 5-15% |
| 2-digit CIP earnings | 20-40% |
| 4-digit CIP earnings | 40-60% |
| 6-digit CIP earnings | 60-80% |
### Why So Much Suppression
1. **Small programs** - Many programs graduate <30 students/year
2. **Credential splitting** - Same field, different credentials
3. **Campus splitting** - Main vs branch
4. **Working requirement** - Must be working to have earnings
### Checking Suppression
```python
import polars as pl
# Count suppressed/missing records
# Note: Field-of-study data is from Scorecard bulk downloads, not Portal mirrors.
# Suppression may use null, -3, "PS", or (pre-May-2024 files) "PrivacySuppressed" depending on file format.
suppressed = df.filter(
pl.col("earn_mdn_hi_2yr").is_null() |
(pl.col("earn_mdn_hi_2yr") == -3)
)
suppression_rate = suppressed.height / df.height
print(f"Suppression/missing rate: {suppression_rate:.1%}")
# Programs with valid data (positive values only)
with_data = df.filter(
pl.col("earn_mdn_hi_2yr").is_not_null() &
(pl.col("earn_mdn_hi_2yr") > 0)
)
print(f"Programs with valid earnings data: {with_data.height}")
```
```r
# Count suppressed/missing records
suppressed <- df |> filter(is.na(earn_mdn_hi_2yr) | earn_mdn_hi_2yr == -3)
suppression_rate <- nrow(suppressed) / nrow(df)
cat(sprintf("Suppression/missing rate: %.1f%%\n", suppression_rate * 100))
# Programs with valid data (positive values only)
with_data <- df |> filter(!is.na(earn_mdn_hi_2yr), earn_mdn_hi_2yr > 0)
cat(sprintf("Programs with valid earnings data: %d\n", nrow(with_data)))
```
## Common CIP Codes
### High-Earning Fields (Typically)
| CIP | Field | Typical Earnings |
|-----|-------|------------------|
| 11 | Computer Science | High |
| 14 | Engineering | High |
| 52 | Business | Medium-High |
| 51 | Health Professions | Medium-High |
### Lower-Earning Fields (Typically)
| CIP | Field | Typical Earnings |
|-----|-------|------------------|
| 50 | Visual/Performing Arts | Lower |
| 54 | History | Lower |
| 38 | Philosophy/Religion | Lower |
| 23 | English | Lower |
### Fields with High Grad School Rates
FoS earnings exclude graduates who were enrolled in school during the measurement year, and the CY2017/CY2018 and CY2018/CY2019 calculations also excluded graduates who received a higher-level credential than the credential level of the field of study measured (FoS technical documentation, October 2026, footnote 4). In fields where many graduates go on to graduate school, the earnings cohort is therefore a selected subset. The documentation does not state the direction of the resulting bias. Illustrative fields (graduate-school rates not sourced here):
| CIP | Field |
|-----|-------|
| 26 | Biology |
| 40 | Physical Sciences |
| 42 | Psychology |
| 45 | Social Sciences |
## Analytical Considerations
### Comparing Fields
**Valid comparisons:**
- Same credential level (Bachelor's to Bachelor's)
- Same institution or similar institution type
- Fields with similar grad school rates
**Invalid comparisons:**
- Bachelor's vs Associate's earnings
- Fields with high vs low grad school attendance
- Suppressed vs non-suppressed programs
### Using CIP Aggregations
| Approach | Pros | Cons |
|----------|------|------|
| 2-digit CIP | More data, less suppression | Heterogeneous |
| 4-digit CIP | More specific | More suppression |
| 6-digit CIP | Most specific | Highest suppression |
### Aggregating Across Institutions
To reduce suppression, aggregate across similar institutions:
```python
import polars as pl
# Aggregate 2-digit CIP earnings across public 4-years
# This example assumes the direct file has been normalized to lowercase and
# its documented control code 1 means public; verify the current release
field_summary = (
df.filter(pl.col("control") == 1) # Public (integer code)
.filter(pl.col("credlev") == 3) # Bachelor's (integer code)
.filter(pl.col("earn_mdn_hi_2yr") > 0) # Valid earnings only
.group_by("cipcode")
.agg([
pl.col("earn_mdn_hi_2yr").median().alias("median_earnings"),
# num_awards: illustrative variable name; verify in the data dictionary
pl.col("num_awards").sum().alias("total_completers")
])
)
```
```r
# Aggregate 2-digit CIP earnings across public 4-years
# This example assumes the direct file has been normalized to lowercase and
# its documented control code 1 means public; verify the current release
field_summary <- df |>
filter(control == 1, credlev == 3, earn_mdn_hi_2yr > 0) |>
group_by(cipcode) |>
summarise(
median_earnings = median(earn_mdn_hi_2yr),
# num_awards: illustrative variable name; verify in the data dictionary
total_completers = sum(num_awards)
)
```
## Gainful Employment Metrics
Field-level data supports accountability metrics:
### Debt-to-Earnings Ratio
| Metric | Formula |
|--------|---------|
| Annual earnings | Median earnings post-completion |
| Annual debt payment | Estimated from median debt |
| D/E ratio | Annual payment / Annual earnings |
### Earnings Threshold
| Metric | Description |
|--------|-------------|
| Earnings > HS grad | Program earnings exceed HS-only baseline |
| Earnings > federal poverty | Program earnings above poverty level |
### Program-Level Accountability
Used for:
- Gainful Employment regulations
- Consumer disclosures
- Institutional accountability
## Recommended Practices
### For Field-Level Analysis
1. **Start with 2-digit CIP** for broader coverage
2. **Report suppression rates** prominently
3. **Note post-completion timing** differs from institution level
4. **Acknowledge grad school effects** on certain fields
5. **Compare same credential levels** only
### For Consumer Information
```
Good: "Computer Science graduates at School X earn $75,000
two years after graduation"
Better: "Computer Science bachelor's degree graduates at School X
who received federal aid and had earnings during the year earn
a median of $75,000 two years after graduation"
Best: "Computer Science bachelor's degree graduates at School X
who received federal aid earn a median of $75,000 two years
after graduation. This excludes graduates who were enrolled in
school during the measurement year or had no earnings that year.
Based on XX completers in the cohort."
```
### Dealing with Suppression
| Strategy | Use Case |
|----------|----------|
| Aggregate to 2-digit CIP | Broad field analysis |
| Aggregate across years | Increase sample size |
| Aggregate across similar schools | National field estimates |
| Note as missing | When suppression high |
## Linking to Institution Data
Join field data to institution characteristics:
```python
import polars as pl
# Merge field data with institution data
# Assumes direct field and institution files were normalized to lowercase before joining
combined = field_df.join(
institution_df.select(["unitid", "inst_name", "control", "state_abbr"]),
on="unitid",
how="left"
)
```
```r
# Merge field data with institution data
# Assumes direct field and institution files were normalized to lowercase before joining
combined <- field_df |> left_join(
institution_df |> select(unitid, inst_name, control, state_abbr),
by = "unitid"
)
```
## Limitations Summary
| Limitation | Impact |
|------------|--------|
| High suppression | Many programs have no data |
| Post-completion timing | Not comparable to institution-level entry cohorts |
| Completers only | Excludes program non-completers |
| Title IV only | Same selection bias as institution level |
| Earners only | Excludes graduates enrolled in school during the measurement year and those with no earnings that year (no earnings ≠ unemployed; the CY2018/CY2019 three-year measure included non-workers) |
| Early-career earnings | 1-7 years after completion depending on the measure (FoS technical documentation, October 2026) |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 198k | Updated | Claude CodeCodex | Verified | Install |
| 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: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| 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: 194k | Updated | Claude CodeCodex | Verified | Install |