Auto Loan Credit

ABS-EE Loan-Level Panel — SEC EDGAR

Delinquency trends from SEC EDGAR ABS-EE filings. Loan-level auto credit data, parsed from servicer reports so you don’t have to.

Data source: SEC EDGAR ABS-EE autoloan panel — loan-level data from asset-backed securities filings. Parsed from raw XML filings via the bens-data-lake pipeline. Coverage varies by issuer. See Data Sources for full methodology.

This is structured finance data — the kind of thing you’d normally pay a Bloomberg terminal to look at, except it’s sitting in SEC EDGAR and nobody told you. ABS-EE filings contain loan-level performance data for auto loan ABS trusts: origination date, balance, days past due, payment status. The good stuff.

---

Delinquency Trends

90+ day delinquency rate across all loans in the panel, by filing month. A number going up here means consumers are falling behind on car payments. Make of that what you will.

import { DuckDBClient } from "npm:@duckdb/duckdb-wasm"
db = DuckDBClient.of({})

// R2 base URL — injected via Quarto site params or environment
r2Base = params["r2-base-url"]

// Query: 90+ day delinquency rate by filing month
delinqData = {
  if (!r2Base || r2Base === "") {
    return null
  }
  try {
    const result = await db.query(`
      SELECT
        filing_date,
        COUNT(*) FILTER (WHERE days_past_due >= 90) * 100.0 / COUNT(*) AS delinq_rate_90plus,
        COUNT(*) AS total_loans,
        COUNT(*) FILTER (WHERE days_past_due >= 30) * 100.0 / COUNT(*) AS delinq_rate_30plus
      FROM read_parquet('${r2Base}/abs_ee/curated/asset_class=autoloan/*.parquet')
      WHERE filing_date IS NOT NULL
      GROUP BY filing_date
      ORDER BY filing_date
    `)
    return result
  } catch(e) {
    return { error: e.message }
  }
}

// Chart: 90+ day delinquency rate
{
  if (!r2Base || r2Base === "") {
    const el = document.createElement("div")
    el.className = "chart-error"
    el.innerHTML = `<span>⚠</span><span>R2 data source not configured. Set <code>r2-base-url</code> in site params.</span>`
    return el
  }

  if (!delinqData || delinqData.error) {
    const el = document.createElement("div")
    el.className = "chart-error"
    el.innerHTML = `<span>⚠</span><span>Data unavailable — ${delinqData?.error ?? "unknown error"}. Check R2 connection.</span>`
    return el
  }

  const maxDate = d3.max(delinqData, d => new Date(d.filing_date))
  const asOfStr = maxDate
    ? maxDate.toLocaleDateString("en-US", { month: "short", year: "numeric" })
    : "unknown"

  return Plot.plot({
    title: "90+ Day Delinquency Rate",
    subtitle: "% of loans 90+ days past due, by filing month",
    width: 800,
    height: 340,
    marginLeft: 55,
    marginBottom: 45,
    x: {
      type: "time",
      label: null,
      tickFormat: d => d.toLocaleDateString("en-US", { month: "short", year: "2-digit" })
    },
    y: {
      label: "Delinquency Rate (%)",
      tickFormat: d => `${d.toFixed(1)}%`,
      grid: true,
      gridColor: "var(--border-color, #D8D3C8)"
    },
    marks: [
      Plot.ruleY([0], { stroke: "var(--border-color, #D8D3C8)" }),
      Plot.line(delinqData, {
        x: d => new Date(d.filing_date),
        y: "delinq_rate_90plus",
        stroke: "#4A6FA5",
        strokeWidth: 2
      }),
      Plot.dot(delinqData, Plot.selectLast({
        x: d => new Date(d.filing_date),
        y: "delinq_rate_90plus",
        fill: "#4A6FA5",
        r: 4
      }))
    ],
    caption: `Source: SEC EDGAR ABS-EE | Units: % of loans | No smoothing applied | As of ${asOfStr}`
  })
}

---

Loan Volume

Total loans in the panel over time. This gives a sense of coverage — more loans means more issuers filing, means the delinquency rate above is a better sample.

{
  if (!delinqData || delinqData.error || !r2Base) return html`<div class="chart-loading"><span>◌</span><span>Waiting for data...</span></div>`

  return Plot.plot({
    title: "Total Loans in Panel",
    subtitle: "Count of loans reported in ABS-EE filings, by month",
    width: 800,
    height: 260,
    marginLeft: 65,
    marginBottom: 45,
    x: {
      type: "time",
      label: null,
      tickFormat: d => d.toLocaleDateString("en-US", { month: "short", year: "2-digit" })
    },
    y: {
      label: "Loan Count",
      tickFormat: d => d >= 1e6 ? `${(d/1e6).toFixed(1)}M` : d >= 1e3 ? `${(d/1e3).toFixed(0)}K` : d,
      grid: true,
      gridColor: "var(--border-color, #D8D3C8)"
    },
    marks: [
      Plot.areaY(delinqData, {
        x: d => new Date(d.filing_date),
        y: "total_loans",
        fill: "#4A6FA5",
        fillOpacity: 0.15
      }),
      Plot.line(delinqData, {
        x: d => new Date(d.filing_date),
        y: "total_loans",
        stroke: "#4A6FA5",
        strokeWidth: 1.5
      })
    ],
    caption: "Source: SEC EDGAR ABS-EE | Units: loan count | As of latest filing month"
  })
}

---

Methodology

What counts as delinquent: A loan is classified as 90+ days past due (days_past_due >= 90) at the servicer-reported filing date. The bucket used here is the most severe — it excludes 30–59 day and 60–89 day buckets.

What’s included: All loans in Hive-partitioned Parquet files under abs_ee/curated/asset_class=autoloan/ on R2. Coverage is limited to issuers that file ABS-EE with the SEC — not all auto ABS trusts are required to do so.

Known limitations: Servicer definitions of “days past due” can vary slightly by issuer. This is the data as filed. No harmonization has been applied across issuers.

For full source documentation, see Data Sources →