Who Watches the Watchmen?

Private credit is a $1.7 trillion market, and almost none of it is marked to market. When a BDC holds a portfolio of private loans, the BDC itself decides what those loans are worth. The marks determine the borrowing base, which determines how much the BDC can borrow from banks.

The Blue Owl saga (2025–2026) illustrates what happens when the marks are questioned:

• Nov 2025: Blue Owl terminated a planned merger between OBDC II and OBDC, which was trading at a 20% discount to NAV
• Mar 2026: Saba Capital launched a hostile tender offer for OBDC II at a 33% discount to NAV
• Meanwhile: Blue Owl is selling $1.4B of assets at 99.7 cents on the dollar to fund a 30% capital return
• Are the assets worth par, or is the equity worth 67 cents?

Who can independently verify these marks?

• SEC — requires quarterly statements but doesn't audit individual loan valuations
• Rating agencies — assess the BDC at entity level, not loan-by-loan
• Banks lending to BDCs — the primary parties with both contractual authority and financial incentive to challenge marks

This dashboard extracts and maps those contractual provisions from SEC EDGAR filings.

BDCs are regulated investment companies that must file 10-K, 10-Q, and 8-K on SEC EDGAR — regardless of whether they are publicly traded, non-traded, or private.

• Credit agreements are public. Filed as EX-10.x exhibits to 8-K or 10-K filings.
• Coverage is near-universal. 281 BDCs, covering 97% of currently active BDCs.
• Full history available. EDGAR filings back to the early 2000s.

• Step 1 — BDC identification: Search EDGAR for N-2 and 10-K filings mentioning "business development company." Filter false positives using SIC codes.
• Step 2 — Agreement download: Search EDGAR full-text search API for "credit agreement," download all EX-10.x exhibits.
• Step 3 — LLM extraction: Each HTML exhibit split at ARTICLE boundaries, processed through four Gemini 2.5 Flash prompts:

• Deal Terms (borrower, admin agent, facility size, spread, benchmark rate)
• Valuation Mechanics (appraisal rights, testing frequency, override thresholds)
• Advance Rates (first-lien, second-lien, unitranche)
• Covenant Protections (asset coverage, leverage, concentration limits)

Caveats: We capture only agreements filed as separate exhibits. LLM extraction is imperfect. We observe whether the right exists, not whether it is exercised.

The administrative agent is the lead bank managing the credit facility:

• Processes drawdowns and repayments
• Monitors covenant compliance
• Exercises valuation rights (if included)

An independent valuation right lets the bank appoint a third-party appraiser. The "5/20 rule":

• <5% gap: No change
• 5–20% gap: Two valuations averaged
• >20% gap: Third appraiser; all three averaged
• Lower marks shrink the borrowing base, potentially forcing asset sales

We test whether facility terms differ systematically between agreements with and without independent valuation rights. For each outcome variable Y, we estimate:

Y_i = α + β · ValuationRights_i + γ_t + δ_b + ε_i

where γ_t are year fixed effects and δ_b are bank (administrative agent) fixed effects. The coefficient β captures the difference in the outcome variable associated with having valuation rights.

Advance rate = the percentage of a loan's value that counts toward the borrowing base. A 75% advance rate on a $100M loan adds $75M to the borrowing base. Higher advance rates mean more leverage. First-lien loans (senior secured) get higher rates than second-lien (junior) loans.

We estimate a linear probability model of the determinants of independent valuation rights:

ValuationRights_i = α + β₁ · log(FacilitySize_i) + β₂ · Amendment_i + β₃ · Spread_i + γ_t + δ_b + ε_i

The dependent variable equals 1 if the agreement grants the lender an independent valuation right. We progressively add controls, year FE, spread, and bank FE to test robustness.