Similar new-loan spreads mask meaningful differences in sector risk.
A benchmark built from SEC filings has an expected 7% to 8% more spread, net of expected losses, per unit of risk than business development companies’ (BDCs) actual sector mix.
Because direct loans are not rebalanced, origination determines sector allocation.
Ask an equity manager for sector weights against the index, and you get a table in thirty seconds.
Ask a direct lending manager and you get a pause, then a fair objection: against what index?
Direct lending, the largest part of private credit, means funds lending straight to midsized companies. It has no common sector benchmark, so a fund's sector exposure largely reflects whatever deals arrive.
Two numbers could guide the choice. The spread a loan pays over its floating reference rate is fixed at origination; the value a fund reports for the loan each quarter moves with the borrower's credit.
The problem is competition has compressed new loan spreads across sectors, so they say little about sector risk; the values of older loans say more. A book assembled by deal flow is still a sector bet; the only choice is whether it was deliberate.
In the first edition of a recent publication series, the Credit BDC Sector Monitor1, Cardo AI looks at the loan-by-loan schedules that 53 BDCs file with the US Securities and Exchange Commission (SEC), sorts each loan into one of ten sectors (such as financial services, healthcare, media, consumer), and establishes a common reference point from them. BDCs are the only direct lenders required to report every loan they hold, on comparable terms, in public filings.
This article examines what the authors learned from building the monitor and how LPs and GPs can apply those insights.
Upgrowth Has Moved to Funds Without a Market Price
Private credit has grown from roughly $250 billion after the global financial crisis to an estimated $2.6 trillion, according to research from the CFA Institute Private Credit: Market Structure, Fund Design, and Retail Access2. This growth has increasingly come through funds sold to individual investors.
The main investment channel is the business development company (BDC), a US fund that must report every loan it holds, and its value, in quarterly US Securities and Exchange Commission (SEC) filings.
Listed BDCs trade on an exchange; non-traded BDCs are bought and redeemed at net asset value (NAV) set from the manager's own loan values.
Retail access is not the problem, missing prices are. For the fastest-growing part of the market, the loan-level filings are the only public view of how a portfolio is built.
Among the 72 BDCs which consistently submit filings on the SEC’s Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system, non-traded net assets grew from $32 billion in early 2023 to $116 billion by the end of2025, and listed ones from $42 billion to $55 billion (Figure 1).
This means there is limited data on the majority of the funds holding the loans.
Market prices provide transparency. When investors doubt a listed BDC's loan values or sector bets, its shares fall below NAV for all to see. A non-traded BDC's NAV follows a valuation policy under board oversight but is never tested by trading.
The bias therefore falls on manager selection: across the 53 BDCs whose loan schedules we parse, spreads run 321 to 728 basis points.
Beneath it sits the sector decision. The case for making it deliberately rests on two filed numbers that no longer agree. In 2025, the last full origination year, median new-loan spreads in all ten sectors sat inside a 50-basis point band, 475 to 525. The values on seasoned loans disagree.
Each quarter a manager estimates what every loan is worth; that estimate, over the loan's outstanding principal, is the nearest thing this market has to a price.
Among seasoned loans, those held on a fund's books for at least four quarters, 15% to 16% of media and transportation loans are valued below 85 cents on the dollar, against 8% to 9% in financial services and technology. A newly originated loan is almost always carried near par, so a fast-growing sector looks safer than it is until its book has aged.
What matters for allocation: sectors still differ in risk, no longer in price, and composition is the only tool left.
Old Tools, Correct Market
Private markets are called too opaque for Markowitz’s Modern Portfolio Theory. With no price history, the argument runs, one cannot estimate the expected returns, volatilities and correlations that mean-variance optimization requires, the opposite holds.
Several standard tools in the quantitative toolkit were built precisely for short, noisy samples, and the sparser the data, the stronger the case for using them rather than naïve estimates.
Three of these filters apply to private credit.
Ledoit-Wolf shrinkage pulls a crowded sample correlation matrix toward a simpler, more stable structure; it exists because short samples produce unreliable correlation estimates, and private credit is a permanently short sample.
Black-Litterman starts from the market's own allocation, here the aggregate sector weights reported in BDC filings, and moves weight away from it where a specific view is strong enough to justify the move, exactly what noisy inputs demand.
Stress-amplified correlations, calibrated on 2020 and 2022, answer the question that matters: how much diversification survives a bad year.
These borrowed tools work in private credit for a structural reason. A direct loan runs five-to-seven years and self-amortizes; nobody rebalances it. Sector exposure is fixed the day a loan closes: allocation is decided before the money leaves.
From Filings to Weights
Figure 2 shows the machinery; healthcare shows what it does. Its new loans pay a near-average spread, but its reported values carry one of the smallest discount tails, so expected losses are low and its excess return, spread minus expected losses, ranks among the best per unit of risk. The market holds 19% there; the optimizer lifts it toward 26%.
Four current-quarter signals then adjust the model's inputs, not its weights: new-issue pricing, its dispersion, origination tails and mark momentum.
The study shows that, based on Q1 2026 optimization, weights were trimmed by five points in the technology sector on softer new-loan pricing and slipping values, keeping it as the book’s best diversifier. Trimmed, not cut, for stated reasons.
What It Changes
Compared with the aggregate portfolio composition reported in SEC filings, based on the modeling exercise performed in Q1 2026, the benchmark produced 7% to 8% more excess return per unit of risk, a relative improvement in the ratio from 18.5 to 19.8 for listed BDCs and from 24.8 to 26.8 for non-traded BDCs, while maintaining the same overall spread. These are model estimates, not realized returns.
The result held across both listed and non-traded portfolios. In modeled stress scenarios ranging from mild deterioration to conditions comparable with the global financial crisis, the benchmark also produced smaller losses. In the most severe scenario, 15.9% of the book against 17.2% for the listed filed book, and 15.7% against 16.9% for the non-traded one. It achieved this result by overweighting healthcare and business services and underweighting consumer, financial services, technology, and transportation.
What This Means
For LPs:
Ask for sector weights against a stated reference. A GP who cannot compare is not managing the exposure.
Read concentration in falling-value sectors as a pricing question: the market pays nothing extra for media or transportation risk.
For GPs:
When a model moves a weight, demand the reason, as the Cardo AI study shows for technology.
Publish your sector mix, with reasons. LP data is improving faster than most GPs assume; funds that answer first will own the conversation.
Full methodology and results: Credit BDC Sector Monitor, Q1 2026, Cardo AI. https://cardoai.com/reports/credit-bdc-sector-monitor/
If you liked this post, don’t forget to subscribe to the Enterprising Investor.
All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer.
Image credit: ©Getty Images