Direct Lending Returns Are Simpler Than You Think

What's New

Cliffwater's recent paper "The Arithmetic of Direct Lending" makes the case that private credit's return profile is governed by a small set of observable, measurable inputs rather than the complexity often attributed to it. Using 20+ years of CDLI data, the paper decomposes returns into income, credit gains/losses, and fees/leverage, and shows that even under extreme stress scenarios the asset class produces outcomes within a surprisingly narrow and predictable range.

Why It Matters

At a moment when press scrutiny and BDC redemption headlines are driving fear about private credit, Cliffwater is essentially publishing the counter narrative in plain arithmetic. The paper's core argument is that income dominance, not mark to market volatility, determines long term outcomes. For allocators debating whether to stay the course or reduce exposure, this framework provides a structured way to evaluate downside scenarios without relying on sentiment.

Big Picture Drivers

• Income stability: The CDLI has generated average annual income of 10.93% since inception, with excess income over T bills averaging 9.22%, providing a thick cushion against credit events.

• Return consistency: 48% of the CDLI's 21 calendar years landed within 2 percentage points of the long term average, compared to just 6% of years for the S&P 500 over the past century.

• Mark to market mechanics: Unrealized losses function as forward looking loss reserves that have historically overstated eventual realized outcomes, as demonstrated by the 2008 cycle where early markdowns proved more conservative than actual losses.

• Structural recovery advantages: Low loan to value ratios, strong sponsor involvement, coordinated lender groups, and direct control in restructurings have supported recoveries that remain in line with long term averages.

• Fee dispersion by vehicle: Total fees vary meaningfully across structures, from roughly 2.34% for credit interval funds to 4.68% for non perpetual BDCs, with leverage profiles differing accordingly.

By The Numbers

• 9.65%: CDLI average annual total return since 2005, with only one negative calendar year (2008, at negative 6.5%).

• 1.00%: Average annual realized loss rate since CDLI inception, well below the income cushion.

• 48.8%: First lien recovery rate in 2025, consistent with the 10 year average of 48.6%.

• ~4 to 5%: Implied total return even under a draconian scenario where every software borrower defaults over five years with zero recovery.

• 3.81%: Average total fees and expenses across private debt strategies as a percentage of net assets.

Key Trends to Watch

• Software stress test framing: Software and IT represent roughly 24% of the private debt market, but even total wipeout of the sector over five years would still leave direct lending delivering investment grade level returns due to income dominance.

• Vehicle structure divergence: Perpetual BDCs are running lower fee profiles (approximately 3.05%) than legacy non perpetual structures (4.68%), with credit interval funds lowest of all, creating meaningful dispersion in net outcomes for investors.

• Leverage as fee offset: Higher fee vehicles tend to rely on approximately 1.0x leverage to bridge fee drag, while lower fee structures achieve comparable results at 0.25 to 0.35x, making leverage policy a key differentiator in manager selection.

• Vintage risk concentration: 44% of 2025 non accrual activity in the CDLI traced to 2021 and 2022 vintages originated during ultra low rates, signaling that vintage selection remains a primary risk factor as those cohorts season.

The Wrap

Cliffwater's paper is a deliberately simple rebuttal to the complexity narrative. The math shows that contractual income has overwhelmed credit losses in every period except 2008, and even that year's negative 6.5% return was mild relative to public equity and high yield drawdowns. The real message for allocators: if you believe realized losses will stay anywhere near historical norms (or even multiples of them), the income engine makes direct lending outcomes far more bounded than headlines suggest. The risk is not that the math breaks. The risk is that you abandon the asset class precisely when the math is working most in your favor.