The CAPE Problem Is Really a Concentration Problem

The CAPE Problem Is Really a Concentration Problem

The aggregate Shiller CAPE ratio near 42x is alarming, but it is also incomplete. A single market-level number treats the S&P 500 as if every constituent were equally responsible for the valuation extreme. That is not how the current market is built. The index is heavily capitalisation-weighted, heavily concentrated in a small group of mega-cap technology and platform companies, and heavily influenced by a sector mix that has changed faster than the historical CAPE framework can easily absorb. The result is a valuation signal that is directionally important but mechanically blunt. It tells us that the market portfolio is expensive. It does not tell us that the typical American company is trading at the same extreme.

This distinction is not a semantic escape hatch. It changes how investors should interpret the signal. If an equal-weighted CAPE or an S&P 500 excluding the Magnificent Seven sits materially below the headline index CAPE, then the aggregate multiple is partly a statement about concentration rather than a statement about universal excess. Sector-level evidence already points in that direction. Technology trading near 65x on a CAPE basis is a very different object from the rest of the market trading at far more moderate normalized earnings multiples. The headline 42x figure therefore overstates the valuation of the median or typical listed U.S. company, while still accurately describing the price of the cap-weighted market portfolio that investors actually own through passive exposure.

That final clause matters. The headline may overstate broad-based excess, but it does not become irrelevant. Passive capital is not allocated to the typical company. It is allocated by market capitalisation. If the largest companies are priced for exceptional future profitability, then the index return is priced for those assumptions as well. Investors in cap-weighted U.S. equity beta cannot dismiss concentration as someone else's problem. Concentration is the portfolio. The correct conclusion is subtler: the valuation risk is not evenly distributed across the market, but it is highly relevant for the index because the most expensive and most important companies carry unusual weight.

The Difference Between Market Valuation and Typical-Company Valuation

A market-cap-weighted index answers the question, “What is the value-weighted price of the market?” It does not answer the question, “What is the average valuation of the average company?” Those two questions can diverge dramatically when a small number of firms become very large. In the current market, the largest technology and platform companies account for an outsized share of index capitalisation, profit growth, investor attention, and factor exposure. If those firms trade at elevated normalized multiples, the aggregate CAPE can look extreme even if hundreds of smaller index constituents trade at ordinary or only moderately expensive valuations.

This is why equal-weighted measures matter. An equal-weighted index gives each company the same portfolio weight and therefore describes the valuation experience of the typical constituent more directly. If equal-weighted CAPE is materially lower than cap-weighted CAPE, the market is not uniformly expensive. Instead, a valuation premium is concentrated in the winners. That premium may be rational if the winners have structurally superior margins, stronger balance sheets, more durable growth, and higher returns on invested capital. But it also means that the aggregate market's expected return depends disproportionately on whether those winners can continue to justify their premium.

The same logic applies to an ex-Magnificent Seven index. Removing the largest platform and AI-linked companies would almost certainly lower the normalized market multiple. That exercise does not prove that the remaining market is cheap, but it clarifies where the valuation pressure resides. It is one thing to say that U.S. equities are expensive because every sector has repriced to heroic assumptions. It is another to say that U.S. equities are expensive because a narrow leadership cohort has become so large and so highly valued that it pulls the market aggregate into rare historical territory.

For active managers, this creates both opportunity and career risk. The opportunity is that broad index valuation can mask more reasonable prices outside the dominant cohort. The career risk is that underweighting the dominant cohort can produce severe underperformance if leadership persists. For passive investors, the issue is different: they may believe they own broad diversification, but the valuation risk embedded in the index is increasingly tied to a small set of corporate profit engines. Diversification by line item is not the same as diversification by economic exposure.

Sector Composition and the Technology CAPE

The cited technology CAPE near 65x is the central clue. A sector with asset-light business models, high gross margins, platform economics, global revenue pools, and large buyback capacity should trade at a premium to the old industrial market. Software and platform earnings can scale with lower marginal cost than railroads, utilities, commodity producers, or traditional manufacturers. Network effects can protect profitability. Data advantages and ecosystem lock-in can lengthen competitive advantage periods. These features can justify higher normalized multiples than the long-run market average.

But a 65x sector CAPE is not merely a quality premium. It is an embedded forecast. It says that investors expect technology earnings to be durable, margins to remain high, reinvestment opportunities to remain attractive, and competitive pressure not to destroy excess returns. It also says that the market is willing to capitalize current and trailing profitability at a very high rate, even though the technology sector itself is undergoing an enormous AI-related capital expenditure cycle. The more the sector becomes both the engine of earnings and the object of heavy reinvestment, the more the valuation question turns from “Are these good businesses?” to “Are these businesses good enough to make this price rational?”

Sector composition also complicates historical comparison. The S&P 500 of the 1970s or early 1980s contained more capital-intensive, cyclical, regulated, and lower-margin businesses. The modern index contains more intangible assets, network economics, and global platform rents. A direct comparison between today's CAPE and the long-run mean near 17x therefore mixes two questions: whether the same type of asset is more expensive, and whether the asset mix itself has improved. The answer is likely both. The index is better in some economic respects, and it is also more expensive.

This is why the strongest valuation work should separate composition from price. If the market's CAPE is high because the index owns more high-quality businesses, some premium is justified. If it is high because investors are paying dramatically more for the same quality-adjusted cash flows, the signal is more worrying. In practice, the current market likely contains both effects: genuine structural improvement in the earnings base and genuine overcapitalisation risk in the most celebrated sectors.

The Ten-Year Earnings Window Cuts Both Ways

CAPE's ten-year real earnings denominator is designed to smooth the business cycle, but the current cycle creates a special problem. For mega-cap technology companies, the last ten years may understate today's structural earnings power. Many firms have scaled software, cloud, advertising, subscription, semiconductor, and platform businesses in ways that were less mature a decade ago. If the trailing earnings window includes years before these businesses reached current scale, then the denominator is too low relative to true sustainable earnings. In that case, the CAPE multiple exaggerates expensiveness.

This is the bullish denominator argument. It says that CAPE is backward-looking in a market where business models improved rapidly. Higher-margin software, better monetization, global platform economics, and winner-take-most dynamics did not exist in the same form ten years ago. A simple ten-year average may fail to capture the step-change in profitability. If current margins are structurally higher and durable, normalized earnings should be measured closer to current earning power than to the old decade average. On that view, a high CAPE is partly an accounting artifact of real corporate improvement.

The bearish denominator argument is the mirror image. If today's AI-capex-driven margins prove cyclical rather than structural, or if recent profit strength reflects a temporary scarcity premium in compute, chips, cloud capacity, and digital advertising, then the trailing average may not be conservative at all. It may already embed an optimistic baseline. Heavy AI investment can support near-term revenue for infrastructure suppliers while raising future depreciation, competition, and return-on-capital questions. If the current margin regime is less durable than investors assume, sustainable earnings power may be lower than both current earnings and the market's implicit forecast.

The denominator question is therefore empirical, not theological. CAPE itself cannot decide whether mega-cap margins are permanent. Investors need to watch free cash flow after capital expenditure, incremental returns on invested capital, pricing power, customer adoption, depreciation burdens, and competitive responses. The key is not whether AI is important. It clearly is. The key is whether AI converts into broad, durable, shareholder-level cash flow after the cost of building the infrastructure is included.

Concentration Changes the Risk Distribution

When valuation excess is broad, the market is vulnerable to a broad reset. When valuation excess is concentrated, the market is vulnerable to leadership failure. The latter can be just as important for index investors because the leaders carry so much weight. If a handful of mega-cap firms trade at very high normalized multiples and account for a large share of index earnings growth, then the index becomes sensitive to small changes in assumptions about those firms. A modest reduction in expected margin durability or terminal growth can have an outsized effect on the aggregate index multiple.

This is a duration problem as much as a valuation problem. High-multiple growth companies behave like long-duration assets because more of their value is tied to cash flows far in the future. If real rates rise, if the equity risk premium widens, or if investors shorten the horizon over which they are willing to capitalize growth, the valuation effect can be large. The market does not need to decide that the leading companies are bad. It only needs to decide that the future should be discounted at a higher rate or with less certainty.

Concentration also creates reflexivity. As the largest firms outperform, passive inflows buy more of them. Benchmark-aware active managers are reluctant to underweight them. Options markets amplify attention and sometimes momentum. Corporate buybacks reduce float. The leadership cohort becomes both fundamentally important and mechanically supported. That can extend the regime, but it can also make the eventual adjustment more consequential. The more the index depends on a small number of assumptions, the less diversified it really is.

This is where aggregate CAPE retains its warning power. Even if the typical company is not at a historic extreme, the cap-weighted market is priced so that a concentrated group must continue to deliver. A cheaper median stock does not fully protect an investor whose benchmark exposure is dominated by expensive giants. The aggregate measure may be blunt, but it is blunt in a relevant way: it describes the portfolio many investors actually hold.

Expected Returns Under a Concentrated CAPE

The expected-return implication of a high aggregate CAPE should be adjusted, not discarded. If the headline multiple is elevated mainly because of concentration, long-run market returns depend on two layers of mean reversion: whether the leadership premium compresses, and whether the rest of the market rerates. If the leaders maintain margins and growth while the rest of the market remains reasonably valued, the index can produce acceptable returns despite a high headline CAPE. If the leaders derate and the rest of the market does not offset that through earnings acceleration, index returns can be poor even without broad valuation excess.

A simple thought experiment helps. Imagine a market in which the top cohort represents a very large share of capitalisation and trades at 60x normalized earnings, while the rest trades at 22x. The headline market CAPE can land in the low forties. If the top cohort merely falls to 35x while earnings grow, the index can still face a meaningful valuation drag. The rest of the market may be fairly priced, but its lower weight limits its ability to rescue the aggregate return. This is the arithmetic of concentration.

This also changes how investors should diversify. Owning the S&P 500 plus a technology-heavy growth fund may be less diversified than it appears. Owning equal-weighted U.S. equities, value strategies, international equities, quality companies outside the mega-cap complex, or cash-flow-oriented sectors may provide more genuine exposure diversity. The point is not that equal-weight is automatically superior. Equal-weighting introduces its own sector, size, liquidity, and turnover risks. The point is that investors should know whether their portfolio is implicitly making the same mega-cap margin durability bet in several different wrappers.

For institutions, the right response is to decompose risk budgets by economic exposure rather than by product label. How much of total equity risk depends on the same AI capital expenditure cycle? How much depends on the same cloud and semiconductor supply chain? How much depends on the same low-real-rate duration assumption? How much depends on the same U.S. exceptionalism premium? These questions are more useful than asking whether the headline CAPE is “right” or “wrong.”

What Evidence Would Settle the Debate

The debate about concentration and CAPE will be settled by future margins, not by today's ratio. If mega-cap technology firms continue to convert revenue growth into free cash flow at high incremental margins, if AI investment raises productivity across the economy, and if competition does not erode returns, then the high sector CAPE will look less extreme in retrospect. It may represent a rational capitalization of a structurally superior earnings stream. In that scenario, the aggregate CAPE was high because the index contained more exceptional businesses than it used to.

If, however, AI infrastructure spending produces lower incremental returns, if depreciation and energy costs rise faster than monetization, if customers resist pricing, if open-source models commoditize parts of the stack, or if regulators attack platform rents, then the high CAPE will have been warning about overcapitalized expectations. The same trailing earnings window that may understate structural improvement can also hide the cyclicality of a capital spending boom. The denominator can be too pessimistic or too optimistic depending on what happens next.

The practical monitoring list is clear. Watch free cash flow after capital expenditures, not just operating income. Watch return on invested capital for AI projects, not just revenue growth around AI themes. Watch breadth in earnings growth across the index. Watch whether equal-weighted indices confirm or diverge from cap-weighted leadership. Watch real rates and credit spreads, because valuation concentration becomes more fragile when financing conditions tighten. Watch tax and regulatory pressure, because exceptionally high profit pools often attract political attention.

Investors should also watch revisions. A high multiple can survive if earnings expectations are revised upward quickly enough. It becomes dangerous when estimates stop rising while multiples remain high. The first sign of a concentration problem is often not a market crash; it is a narrowing of earnings surprise, a decline in forward revenue confidence, or a loss of pricing power in the leadership group. By the time the headline index falls, the fundamental evidence may already have been deteriorating for several quarters.

How to Test the Concentration Hypothesis

The concentration hypothesis should be tested with data rather than asserted from market anecdotes. The first test is valuation dispersion. If the headline CAPE is extreme but the median sector, median industry, or median stock is much less extreme, then the aggregate signal is being pulled by the largest and most expensive companies. This does not invalidate the headline number, because the headline number describes the capital-weighted investor experience. But it does change the diagnosis from “everything is expensive” to “the index is expensive because the top of the index is expensive.” That diagnosis has different implications for security selection, factor allocation, and risk budgeting.

The second test is earnings contribution. A concentrated index is less dangerous if the largest firms also contribute a proportionate or greater share of sustainable free cash flow. It is more dangerous if their market-cap weight has risen faster than their normalized earnings power. Investors should compare the top cohort's share of market value with its share of revenue, operating income, free cash flow after capital expenditures, and research-and-development intensity. If market value has grown because cash-flow dominance has grown, concentration may be partly justified. If market value has grown because the capitalization rate applied to those cash flows has expanded, expected-return risk is higher.

The third test is margin persistence. CAPE uses trailing real earnings, but the central debate is forward sustainability. A high technology CAPE can be rational if high margins are persistent and incremental investment earns high returns. It is vulnerable if margins are elevated because of temporary scarcity, pandemic-era demand patterns, unusually low funding costs, or a short-lived AI infrastructure cycle. The relevant evidence is not the level of margins alone. It is the stability of margins through competition, customer bargaining, depreciation, energy cost, and regulatory pressure.

The fourth test is breadth. If the equal-weighted index, smaller companies, cyclical sectors, and non-U.S. equities begin to outperform while the headline CAPE remains high, the market may be rotating away from concentration without requiring an aggregate crash. If the leaders continue to dominate both price and earnings revisions, concentration can persist longer. Breadth is the market's way of revealing whether investors are discovering alternatives to the mega-cap growth complex.

Finally, investors should test valuation against the rate environment. A 65x sector CAPE is easier to defend when real yields are falling and liquidity is ample. It is harder to defend when real rates remain high, credit spreads widen, or fiscal risk raises the term premium. Concentrated growth leadership is highly sensitive to the discount rate because its valuation relies on long-duration cash flows. The concentration hypothesis therefore cannot be separated from macro-finance. It is simultaneously a micro story about dominant firms and a macro story about the price of duration.

Risk Management When the Average Lies

The practical problem with averages is that they invite average responses. If the market CAPE is 42x, one investor may sell the whole index; another may dismiss the signal because many stocks are not expensive. Both responses are too simple. A concentrated average requires a concentrated risk-management response. The investor should ask which parts of the portfolio own the expensive aggregate, which parts own the cheaper median, and which parts unknowingly duplicate the same mega-cap exposure.

One useful exercise is to map portfolio holdings into economic buckets rather than ticker buckets. A mega-cap technology stock, a growth ETF, a quality factor fund, a U.S. large-cap benchmark, and a global developed-market index may all contain the same underlying exposure to platform margins, AI capital expenditure, and low discount rates. Counting them as separate line items creates false diversification. The better question is how much total portfolio risk depends on the same earnings and valuation assumptions.

Another exercise is to run a leadership-compression scenario. Assume the dominant cohort's multiple falls materially while its earnings continue to grow. That is not a crash scenario; it is a valuation-normalization scenario. Then ask what happens to the total portfolio. If the answer is severe underperformance, the portfolio is not merely exposed to U.S. equities. It is exposed to the persistence of a specific leadership premium. That exposure may be acceptable, but it should be intentional.

A third exercise is to identify offsetting assets that are not just cheaper versions of the same trade. Equal-weighted equities may help, but they introduce smaller-company and cyclical exposure. International equities may help, but they introduce currency, governance, and sector-mix differences. Bonds may help if growth disappoints, but they may not help if the shock is inflationary or fiscal. Cash has optionality but creates reinvestment risk. The right hedge depends on the failure mode. If the failure mode is mega-cap multiple compression with stable growth, equal-weight and value may help. If the failure mode is higher real rates, duration-sensitive equities and bonds may both struggle. If the failure mode is margin compression from regulation or taxation, sector and geography matter more.

This is why decomposition is not academic. It determines the hedge. A broad valuation bubble calls for broad beta reduction. A concentration-driven CAPE calls for more precise action: reduce unintended mega-cap overlap, diversify by economic exposure, stress-test leadership compression, and avoid assuming that the index is diversified simply because it contains five hundred names.

Conclusion: The Headline CAPE Is Right, But Not Sufficient

The aggregate CAPE's historically elevated level remains directionally valid. It says that the cap-weighted U.S. equity market offers a low normalized earnings yield and therefore a demanding long-run expected-return setup. That signal should not be ignored. But the source of the signal matters. A meaningful portion of the extreme reading appears attributable to concentration, sector composition, and the very high valuation of technology and mega-cap platform businesses rather than to uniformly excessive pricing across all U.S. equities.

This makes the investment problem more nuanced. The typical company may not be priced at a 42x normalized multiple. An equal-weighted market or an ex-Magnificent Seven market would likely look materially less extreme. Yet passive investors do not own the typical company in typical-company weights. They own a market portfolio in which the most expensive, most profitable, and most debated companies dominate the outcome. The headline CAPE may overstate broad-based excess while still correctly warning about the risk embedded in the actual index.

The decisive question is the durability of mega-cap margins and returns on AI-linked investment. If the current leaders turn heavy capital spending into lasting cash-flow growth, the market can justify more of its premium. If margins prove cyclical, competition rises, real rates stay high, or the AI investment cycle disappoints, the same concentration that lifted the index can become the channel of valuation compression. CAPE cannot answer that empirical question on its own. It can only tell us that the burden of proof is high.

The disciplined response is not to reject CAPE, and not to worship it. It is to decompose it. Separate cap-weighted from equal-weighted valuation. Separate technology from the rest of the market. Separate true structural profitability from trailing-window accounting effects. Separate business quality from price paid. Once that is done, the message becomes sharper: U.S. equities are not uniformly in a bubble, but the index is highly exposed to a concentrated valuation regime whose success requires exceptional fundamentals to remain exceptional.

A Portfolio Lens

From a portfolio lens, the concentration version of the CAPE problem is more actionable than the headline version. The headline version says only that long-run expected returns for the broad market are likely lower than history. The concentration version says where that lower expected return is being manufactured. It is being manufactured in the interaction between high mega-cap weights, high technology multiples, and confidence that today's exceptional margins represent a new steady state rather than a cyclical peak. That is an investable diagnosis. It points toward stress-testing portfolios for mega-cap derating, comparing cap-weighted and equal-weighted exposures, and asking whether factor diversification is genuine or merely another expression of the same growth-duration trade.

It also suggests that valuation discipline should be applied unevenly. There is little analytical value in treating a regional bank, a utility, a medical device company, a software platform, and an AI semiconductor supplier as if they all share the same CAPE problem. They do not. Their earnings cycles, reinvestment needs, regulatory risks, balance sheets, and competitive dynamics differ. The useful question is not whether the market is expensive in the abstract. The useful question is which part of the market is priced for perfection, which part is priced for mediocrity, and which part is misclassified by a headline index statistic.

This is why a high aggregate CAPE should lead to decomposition before liquidation. Investors who simply sell the market may avoid concentration risk but also abandon reasonably priced assets hidden inside the index. Investors who ignore the market-level signal may keep unintended exposure to the most expensive assumptions in the portfolio. The middle path is better: identify the concentrated source of valuation risk, decide how much of it is justified by fundamentals, and size it deliberately rather than inheriting it passively.