Is the S&P 500 Cheap Now? Bonds and CAPE Disagree

After a 2.7 percent weekly drop in the S&P 500 and a 12 percent semiconductor rout, FactSet's November 6 forward price-earnings ratio of 17.7 sits below five- and ten-year averages yet implies an earnings yield of only 5.65 percent—roughly 0.79 point above a 4.86 percent 10-year Treasury—while Shiller CAPE at 38.91 remains more than double its century-long mean, so the index is cheaper on optimistic forward profits than on normalized earnings or versus risk-free bonds.

Headlines ask whether the correction made stocks a buy. The honest answer depends on which denominator you trust: analysts' next-twelve-month earnings, or a decade of inflation-adjusted profits smoothed through booms and busts. Those two books do not agree—and the bond market is siding with the skeptical one.

The week that ended November 6 felt like a crash in one corner of the market and a shrug in the rest. The S&P 500 fell 2.7 percent, the Nasdaq Composite 3.4 percent, and the semiconductor complex roughly 12 percent, yet Friday still managed a 0.31 percent bounce to 7,376.86. The Nasdaq sat nearly 12 percent below its October 6 peak; VIX closed at 25.64. That is a sharp, concentrated AI and technology shock—not, by itself, evidence of an economy-wide collapse.

What investors mean when they ask whether stocks are cheap is rarely one number. It is which claim on future cash flow they are willing to underwrite.

The forward-multiple case for “cheaper”

FactSet’s November 6 Earnings Insight put the S&P 500 at 17.7 times forward twelve-month earnings. That multiple is below the index’s five-year average of 19.8 and its ten-year average of 19.1, and it is down from 19.0 at September 30—so the pullback did compress headline valuations on the yardstick desks quote every day.

The blended third-quarter earnings growth rate had risen to 31.4 percent year over year as more companies reported, with 82 percent of reporters beating EPS estimates. October payrolls added 165,000 jobs and unemployment printed 4.1 percent, a labor tape Culled covered alongside Friday’s split market reaction. High-yield credit spreads were 3.48 percentage points over Treasuries on November 5—stressed in spots, but not a systemic freeze. The immediate catalyst was political: midterm results and local opposition raised the odds that data-center construction slows, repricing a crowded growth trade Culled had already mapped as duration clearing.

On forward earnings alone, the market looks less demanding than it did six weeks ago.

Why the 10-year comparison is thinner than it looks

Translate that 17.7 multiple into an implied earnings yield—earnings divided by price—and you get about 5.65 percent. The 10-year Treasury finished November 6 near 4.86 percent, down from 4.94 percent on October 30 as growth fears competed with still-solid labor data. Equities therefore offer only 0.79 percentage point of extra yield over a default-risk-free claim on the U.S. government.

Historically, investors have demanded a far wider spread to hold volatile equity cash flows instead of coupons. When the spread is this narrow, “cheap versus bonds” really means “only slightly less expensive than holding Treasuries,” not a clearance-sale margin of safety. If October CPI, due November 10, runs hot alongside 165,000 payrolls, the Federal Reserve’s higher-for-longer path can lift the 10-year faster than analyst EPS revisions raise the denominator—widening the bond alternative without moving the index price.

That is the comparison you are intuitively right to focus on: not whether P/E fell, but whether the equity risk premium still pays you for AI concentration, election variance, and the chance that forward estimates are themselves cyclically inflated.

CAPE says the calm denominator never arrived

The cyclically adjusted price-earnings ratio—Shiller CAPE—smooths ten years of inflation-adjusted earnings. Multpl recorded 38.91 on November 6, against a long-run mean of 17.45 and within reach of the 44.19 December 1999 peak. On this gauge, the 2026 drawdown barely dented structural richness: CAPE has spent most of the year above 38.

The tension is not mysterious. Forward P/E uses optimistic next-twelve-month profits anchored in an AI capex boom; CAPE still remembers the decade’s profit mix before that boom fully arrived. When the two diverge this wide, “cheap” on forward earnings can coexist with “historically expensive” on normalized earnings—exactly the pattern that preceded prior debates about whether capex curves can rhyme with 2000 even when consumer cash flow does not.

Research desk with spreadsheet on laptop and a long-horizon valuation chart pinned beside stacked quarterly binders at dusk

Sensitivity beats a single verdict

Valuation is a joint statement about price and the earnings path. Hold price constant and shave 10 percent off projected profits, and 17.7 times forward earnings becomes 19.7 times—back above recent averages without a single tick of index movement. Hold earnings constant and re-rate to 15 times, and the arithmetic implies on the order of 15 percent downside to align with a more skeptical cash-flow story.

Neither scenario is a forecast; both show why a binary “cheap or not” headline misleads. The market is cheaper than September on optimistic forward profits; it is not plainly cheap on normalized earnings or relative to bonds unless you believe 30-plus percent earnings growth persists and AI regulation fears stay confined to multiples, not guidance.

What would change the call

Watch three observables, not narrative.

First, whether FactSet’s forward EPS estimate for calendar 2026 holds near 33 percent growth or starts drifting down as data-center permit risk shows up in hyperscaler capex commentary—the same channel that made Nvidia’s clean quarter insufficient to clear doubt.

Second, whether high-yield spreads push meaningfully above 3.5 points and stay there. Stable earnings with lower multiples and contained credit would read as a valuation reset inside a growth factor; falling estimates plus wider spreads would read as fundamental repricing.

Third, whether the earnings-versus-Treasury spread re-widens. A return toward 1.5 to 2 percentage points of equity premium, with earnings intact, would make the forward-multiple case more convincing. A flat index with rising 10-year yields would make bonds the easier call even at 17.7 times.

The base case after one violent week is choppy differentiation: relief rallies inside an AI unwind, not a confirmed recession. Permanent economy-wide repricing is not established. But cheap? Only if you trust the forward earnings spike completely—and even then, the government bond next door is only 79 basis points away.