Data Sources

What it is, where it comes from, and what’s wrong with it.

Every chart on this site is derived from one of the datasets below. This page documents what each dataset is, who publishes it, how it gets from the source to this site, and what known limitations you should keep in mind before drawing any strong conclusions from it.

Transparency about methodology is not optional here. See the constitution if you want the full doctrine.

ABS-EE Autoloan Panel

Publisher: SEC / Office of Structured Finance
Source: SEC EDGAR ABS-EE filings
Update cadence: Monthly (new filings processed as they appear on EDGAR)
R2 prefix: abs_ee/curated/asset_class=autoloan/
Pipeline: bens-data-lake/pipelines/scripts/ingest_autoloan_panel_incremental.py

What it is

Under SEC Regulation AB II, issuers of certain auto loan asset-backed securities are required to file loan-level performance data (Form ABS-EE) monthly. The filing contains data on every loan in the pool: current balance, days past due, payment status, origination date, and other servicer-reported fields.

The data lake pipeline downloads these filings from EDGAR’s master index, parses the XML schema, and writes structured Parquet files partitioned by filing month.

Key variables

Variable	Description
`filing_date`	Month-end date of the servicer report
`days_past_due`	Days since last payment, as reported by servicer
`delinquency_status`	Bucketed: `current`, `30-59`, `60-89`, `90+`, `default`
`current_balance`	Outstanding principal balance at filing date (USD)
`origination_date`	Date the loan was originated
`loan_type`	Vehicle type (new, used, lease) where available

Known limitations

Coverage: Only loans in filed ABS trusts. Bank-held auto loans and captive finance loans not securitized are excluded.
Servicer definitions: “Days past due” is servicer-reported. Definitions vary slightly across issuers.
Filing gaps: Some issuers file late or amend prior filings. The pipeline processes filings as they appear; amendments may not be retroactively applied.
Universe size: Varies by month depending on which issuers are active. Check the loan count chart on the dashboard to assess coverage.

SLOOS — Senior Loan Officer Opinion Survey

Publisher: Federal Reserve Board
Source: Federal Reserve SLOOS release
Update cadence: Quarterly (typically released ~2 weeks after survey close)
R2 prefix: sloos/curated/source=fed_sloos_table_1/
Pipeline: bens-data-lake/pipelines/scripts/ingest_sloos_tables_to_r2.py

What it is

The Federal Reserve surveys senior loan officers at approximately 80 large domestic banks and 24 US branches of foreign banks quarterly. The survey asks whether lending standards have tightened, eased, or remained unchanged over the past three months, and whether loan demand has increased or decreased.

Results are reported as a net percentage: (% tightening) − (% easing). Positive values mean more banks are tightening than easing. Negative values mean more banks are easing.

The pipeline scrapes the HTML tables from the Federal Reserve’s release page and writes Parquet to R2.

Key variables

Variable	Description
`survey_date`	Quarter-end date of the survey
`net_pct_tightening`	Net % reporting tighter standards (positive = more tightening)
`loan_type`	C&I (large/small firm), CRE, consumer (auto, CC, other)
`question_label`	Federal Reserve question text as published

Known limitations

Survey-based: Measures opinion and intent, not actual loan volumes or prices. A bank saying it’s tightening standards is not the same as measuring tighter standards.
Aggregation: Net percentage masks the distribution. A 20% net tightening reading could be 30% tightening / 10% easing or 60% tightening / 40% easing. Different situations.
Definition changes: The Fed has revised question wording over the years. Long historical comparisons should account for definitional shifts.
Response rate: Not all surveyed banks respond in every quarter.

FRED Macro Bundle

Publisher: Federal Reserve Bank of St. Louis (FRED)
Source: FRED API
Update cadence: Daily (FRED series updated as new data releases)
R2 prefix: macro/curated/source=fred_api/
Pipeline: bens-data-lake/pipelines/scripts/ingest_fred_series_bundle_to_r2.py

What it is

A curated bundle of macroeconomic time series fetched from the FRED API. Includes financial stress indicators, credit spreads, and delinquency rates. The pipeline fetches only series that have updated since the last run (--skip-if-unchanged).

Included series (non-exhaustive):

Series ID	Description
`NFCI`	Chicago Fed National Financial Conditions Index
`DRTSCIS`	Net % banks tightening C&I loan standards (from SLOOS, on FRED)
Various	Credit spreads, consumer delinquency aggregates

Status

The FRED macro dataset is present in the data lake but not yet surfaced on this site. It will appear in v2 of the dashboard. The R2 data is there if you want to query it directly: macro/curated/source=fred_api/*.parquet.

Known limitations

FRED series definitions, vintage dates, and revisions vary by source agency.
Some series are seasonally adjusted (SA); others are not (NSA). Check the units field in the Parquet schema.

Last updated: 2026-04-25. For data pipeline documentation, see bens-data-lake.

--- title: "Data Sources" subtitle: "What it is, where it comes from, and what's wrong with it." toc: true --- Every chart on this site is derived from one of the datasets below. This page documents what each dataset is, who publishes it, how it gets from the source to this site, and what known limitations you should keep in mind before drawing any strong conclusions from it. Transparency about methodology is not optional here. See the [constitution](https://github.com/benuehlinger/bens-field-notes/blob/main/.specify/memory/constitution.md) if you want the full doctrine. --- ## ABS-EE Autoloan Panel **Publisher**: SEC / Office of Structured Finance **Source**: [SEC EDGAR ABS-EE filings](https://www.sec.gov/cgi-bin/browse-edgar) **Update cadence**: Monthly (new filings processed as they appear on EDGAR) **R2 prefix**: `abs_ee/curated/asset_class=autoloan/` **Pipeline**: `bens-data-lake/pipelines/scripts/ingest_autoloan_panel_incremental.py` ### What it is Under SEC Regulation AB II, issuers of certain auto loan asset-backed securities are required to file loan-level performance data (Form ABS-EE) monthly. The filing contains data on every loan in the pool: current balance, days past due, payment status, origination date, and other servicer-reported fields. The data lake pipeline downloads these filings from EDGAR's master index, parses the XML schema, and writes structured Parquet files partitioned by filing month. ### Key variables | Variable | Description | |---|---| | `filing_date` | Month-end date of the servicer report | | `days_past_due` | Days since last payment, as reported by servicer | | `delinquency_status` | Bucketed: `current`, `30-59`, `60-89`, `90+`, `default` | | `current_balance` | Outstanding principal balance at filing date (USD) | | `origination_date` | Date the loan was originated | | `loan_type` | Vehicle type (new, used, lease) where available | ### Known limitations - **Coverage**: Only loans in filed ABS trusts. Bank-held auto loans and captive finance loans not securitized are excluded. - **Servicer definitions**: "Days past due" is servicer-reported. Definitions vary slightly across issuers. - **Filing gaps**: Some issuers file late or amend prior filings. The pipeline processes filings as they appear; amendments may not be retroactively applied. - **Universe size**: Varies by month depending on which issuers are active. Check the loan count chart on the dashboard to assess coverage. --- ## SLOOS — Senior Loan Officer Opinion Survey **Publisher**: Federal Reserve Board **Source**: [Federal Reserve SLOOS release](https://www.federalreserve.gov/releases/sloos/) **Update cadence**: Quarterly (typically released ~2 weeks after survey close) **R2 prefix**: `sloos/curated/source=fed_sloos_table_1/` **Pipeline**: `bens-data-lake/pipelines/scripts/ingest_sloos_tables_to_r2.py` ### What it is The Federal Reserve surveys senior loan officers at approximately 80 large domestic banks and 24 US branches of foreign banks quarterly. The survey asks whether lending standards have tightened, eased, or remained unchanged over the past three months, and whether loan demand has increased or decreased. Results are reported as a **net percentage**: (% tightening) − (% easing). Positive values mean more banks are tightening than easing. Negative values mean more banks are easing. The pipeline scrapes the HTML tables from the Federal Reserve's release page and writes Parquet to R2. ### Key variables | Variable | Description | |---|---| | `survey_date` | Quarter-end date of the survey | | `net_pct_tightening` | Net % reporting tighter standards (positive = more tightening) | | `loan_type` | C&I (large/small firm), CRE, consumer (auto, CC, other) | | `question_label` | Federal Reserve question text as published | ### Known limitations - **Survey-based**: Measures opinion and intent, not actual loan volumes or prices. A bank saying it's tightening standards is not the same as measuring tighter standards. - **Aggregation**: Net percentage masks the distribution. A 20% net tightening reading could be 30% tightening / 10% easing or 60% tightening / 40% easing. Different situations. - **Definition changes**: The Fed has revised question wording over the years. Long historical comparisons should account for definitional shifts. - **Response rate**: Not all surveyed banks respond in every quarter. --- ## FRED Macro Bundle **Publisher**: Federal Reserve Bank of St. Louis (FRED) **Source**: [FRED API](https://fred.stlouisfed.org/) **Update cadence**: Daily (FRED series updated as new data releases) **R2 prefix**: `macro/curated/source=fred_api/` **Pipeline**: `bens-data-lake/pipelines/scripts/ingest_fred_series_bundle_to_r2.py` ### What it is A curated bundle of macroeconomic time series fetched from the FRED API. Includes financial stress indicators, credit spreads, and delinquency rates. The pipeline fetches only series that have updated since the last run (`--skip-if-unchanged`). **Included series (non-exhaustive)**: | Series ID | Description | |---|---| | `NFCI` | Chicago Fed National Financial Conditions Index | | `DRTSCIS` | Net % banks tightening C&I loan standards (from SLOOS, on FRED) | | Various | Credit spreads, consumer delinquency aggregates | ### Status The FRED macro dataset is present in the data lake but **not yet surfaced on this site**. It will appear in v2 of the dashboard. The R2 data is there if you want to query it directly: `macro/curated/source=fred_api/*.parquet`. ### Known limitations - FRED series definitions, vintage dates, and revisions vary by source agency. - Some series are seasonally adjusted (SA); others are not (NSA). Check the `units` field in the Parquet schema. --- *Last updated: 2026-04-25. For data pipeline documentation, see [bens-data-lake](https://github.com/benuehlinger/bens-data-lake).*