Tight Spreads, Loose Logic

Henrik Helsen

Contents

In the Spotlight
Appendix: Testing Credit-Spread/Equity Argument
Data
Regressions
Results
Reading the results
Bibliography

Summary

In a recent note to investors, Goldman Sachs Research recommends overweighting equities and underweighting credit, citing unusually tight spreads as a warning sign for corporate bonds . This bundles two separate claims: a diagnosis about credit markets, and a prescription for equity allocation that does not automatically follow from it.
Tight spreads are conventionally a mid-cycle signal, not a late-cycle one. What actually marks the late stage of a credit cycle is rising leverage and a shift toward lower-quality issuance, both of which appear to be building beneath the surface amid a historic wave of AI-related debt .
I test both halves of Goldman's argument directly. Tight spreads reliably predict subsequent widening across every specification tried. They do not reliably predict stronger equity returns: the result is fragile, and disappears once a handful of large episodes are prevented from dominating the sample.

"We remain tactically neutral across assets for 3m and modestly pro-risk for 12m, with OW equities, N bonds, commodities and cash, and UW credit."

Christian Mueller-Glissmann, Goldman Sachs Research, 2 September 2026

In the Spotlight

Goldman Sachs has some investment advice for you: favour equities over credit. In a note published this week, the bank's portfolio strategy team said it remains overweight equities and underweight credit over the next twelve months, citing tight spreads, late-cycle releveraging, and AI-related debt issuance as headwinds for corporate bonds . This is not a lone view. Several desks are running versions of the same argument, treating today's unusually tight credit spreads as a signal that equities offer better value .

Goldman's argument is really two separate claims stacked on top of each other. The first is diagnostic: credit spreads are unusually tight, meaning investors are being paid very little to take on corporate risk, and that thin compensation is unlikely to last - especially with AI-related debt issuance flooding the market. The second is prescriptive: because credit looks vulnerable, the better place for money right now is equities. The note treats these as one continuous piece of logic. Not so fast. The first claim is a statement about credit markets. The second is a conclusion about a completely different asset class, and it does not follow directly from the first, as I show in the appendix.

There's no doubting that the gap between corporate and Treasury bonds has shrunk over the past few years. A tight spread simply means that compensation has shrunk - which sounds, on its face, like an unambiguously bad deal for lenders. By some measures, spreads are near their tightest levels since the late 1990s . At the same time, the path hasn't been a straight line: spreads jumped sharply in the spring of 2026, on the back of tariff and geopolitical tension, before reverting back down toward multi-decade lows by the time Goldman's note landed in September .

It's also worth being precise about what tight spreads are conventionally supposed to signal, because tight spreads on their own do not mean a cycle is ending. In the standard credit-cycle framework, tight spreads are actually a mid-cycle phenomenon: as an expansion takes hold, corporate balance sheets improve, defaults fall, and investors demand less compensation for lending to companies . What marks the late stage of a cycle isn't the tightness of spreads by itself - it's what's happening alongside it: leverage rising quickly, debt increasingly funding shareholder-friendly activity like buybacks rather than productive investment, and the mix of new issuance quietly shifting toward lower-quality borrowers . Read that way, Goldman's "late-cycle" framing isn't really about the spread number itself. It's a claim that the market hasn't yet priced in a leverage buildup happening in plain sight - with AI-related borrowing as the clearest example, hyperscaler bond issuance alone has roughly doubled year-over-year - and that the tight spread simply reflects the market not having noticed yet.

So does the data back the conclusion Goldman draws from it? I tested this directly, with the methodology and findings in the appendix. In particular, does a tight spread actually predict subsequent widening? And does a tight spread actually predict stronger equity returns? The first holds up pretty well ("robustly", as econometricians like to say), across multiple ways of slicing the data. Tight spreads really do tend to revert. The second does not really survive scrutiny. It looks significant in a naive version of the test, but the result evaporates once you control for the fact that a handful of large episodes, chiefly the 2020 shock and recovery, were doing most of the work. Strip those out, and there's no reliable evidence that a tight spread today tells you anything useful about where stocks are headed.

That leaves the more defensible version of Goldman's argument as something narrower than what's being marketed: not "tight spreads predict equity gains," but "credit's best-case outcome is capped near historic tights, while equity's isn't" - a statement about the shape of possible payoffs, not a forecast. Not everyone reads it the same way, either: some desks argue precisely the opposite, that high-yield credit has historically delivered near-equity returns with lower volatility and faster recoveries after drawdowns, making it a substitute for equity exposure rather than a reason to avoid it .

None of this makes Goldman wrong to worry about credit. The diagnosis (that spreads are unusually tight and this is coinciding with an explosion of AI-related borrowing) is sound, and the data agrees. But the leap from there to "so buy stocks" might just be doing more work than the evidence can support. Caveat emptor.

Appendix: Testing Credit-Spread/Equity Argument

On September 2, 2026, Goldman Sachs Research argued for overweighting equities relative to credit, citing tight credit spreads as a headwind for corporate bonds . This rests on two claims: that a tight spread signals credit is due to weaken, and that this in turn favors equities. I test both directly.

Data

Monthly series, September 2016-August 2026, merged from two FRED series: OAS, the Moody's Baa corporate bond yield minus the 10-year Treasury yield (BAA10Y) , a standard measure of the extra yield investors demand to hold corporate over government debt; and SPX, the S&P 500 index level (SP500) , which FRED restricts to a trailing 10-year window.

Regressions

Two forward-looking outcomes are regressed on the contemporaneous spread level, $\text{OAS}_t$ :

\begin{aligned} \Delta \text{OAS}_{t \to t+6} &= \alpha + \beta_1\, \text{OAS}_t + \varepsilon_t \\ \text{Ret}^{SPX}_{t \to t+6} &= \alpha + \beta_2\, \text{OAS}_t + \varepsilon_t \end{aligned}

The first equation asks whether today's spread predicts how the spread itself moves over the next six months - i.e., whether tight spreads actually tend to widen more often than they shrink, as Goldman's caution on credit implies. The second asks whether today's spread predicts how the S&P 500 performs over the same period. This, after all, is the implicit basis for preferring equities over credit in a tight-spread scenario. To avoid overlapping-window bias, both are estimated by OLS on non-overlapping 6-month windows ( $n=19$ ), and separately on non-overlapping annual windows ( $n=10$ ), so no observation shares data with another.

Results

	6-month windows ( $n=19$ )		Annual windows ( $n=10$ )
Outcome	$\hat\beta$	$p$	$\hat\beta$	$p$
$\Delta$ Spread, forward	$-0.553$	$0.018$	$-0.693$	$0.034$
S&P 500 return, forward	$+10.047$	$0.049$	$+3.038$	$0.784$

Reading the results

The spread-widening result is robust: negative and significant at both the 6-month and annual horizon, meaning a tight spread reliably precedes a widening one, consistent with Goldman's premise. The equity-return result is not: significant, barely, at 6 months, but statistically indistinguishable from zero at the annual horizon. A result that flips between significant and null depending on the sampling window is not a reliable finding. Taken together, the data support the case for caution on credit specifically, but not the further claim that this favors equities over credit.

Bibliography

Mueller-Glissmann, C. (2026). "Goldman Sachs favors equities over credit as late-cycle risks build." Goldman Sachs Research, reported via Investing.com, 2 September 2026.

Lynam, A. and Ablon, Z. (2026). "How AI Debt Is Reshaping Credit Markets." Goldman Sachs Exchanges, Goldman Sachs.

Loomis Sayles (2026). "Unlocking the Credit Cycle."

State Street Investment Management (2025). "Credit spreads signal confidence and risk." Mind on the Market, 24 November 2025.

AllianceBernstein (2024). "High-Yield Opportunity Persists, Despite Tight Spreads."

ETF Trends (2026). "Credit Spreads at Historic Tights: What Now?" 28 January 2026.

Forbes (2026). "What The Credit Spread Signal Is Actually Telling Investors Right Now." 9 April 2026.

Federal Reserve Bank of St. Louis (2026). "Moody's Seasoned Baa Corporate Bond Yield Relative to Yield on 10-Year Treasury Constant Maturity [BAA10Y]." FRED.

S&P Dow Jones Indices LLC (2026). "S&P 500 [SP500]." Retrieved from FRED, Federal Reserve Bank of St. Louis.