Market indicators

The Buffett indicator

The Buffett indicator is the ratio of the total market value of publicly traded U.S. equities to the country’s nominal gross domestic product, expressed as a percentage. It takes its name from Warren Buffett, who in a 2001 Fortune article described the ratio of market value to GNP as “probably the best single measure of where valuations stand at any given moment.” The underlying idea is that, over long horizons, the aggregate value of corporate equity should bear some relationship to the size of the economy that generates corporate revenue.

The chart plots market value of U.S. corporate equities ÷ nominal GDP × 100, quarterly. The numerator is the market value of corporate equities from the Federal Reserve’s Financial Accounts of the United States (the “Z.1” release); the denominator is nominal GDP from the Bureau of Economic Analysis, seasonally adjusted at an annual rate. Both series are quarterly, and the Z.1 is published roughly ten weeks after quarter-end, so the latest reading lags the calendar by about one quarter.

Several caveats apply when comparing the ratio across decades. Buffett’s original formulation used GNP, which includes net income U.S. residents earn abroad, while most modern versions — including this one — use GDP, which measures production inside U.S. borders; the choice shifts the level slightly. More significantly, the numerator reflects the worldwide operations of listed companies while the denominator counts only domestic output: as U.S.-listed firms have earned a growing share of their revenue overseas, the ratio has drifted upward for reasons unrelated to valuation. The indicator also takes no account of interest rates, tax policy, or changes in how much of the corporate sector is publicly listed.

Because the ratio has drifted upward over the decades, the chart also shows an exponential trend line fitted to the full 1950–present series — a least-squares regression on the natural logarithm of the ratio. It describes where a reading sits relative to the series’ own history rather than against a fixed threshold, and it is recomputed from the data on the page, so it updates whenever the data does.

The Shiller CAPE ratio

The cyclically adjusted price-to-earnings ratio, or CAPE, divides the real (inflation-adjusted) price of the S&P 500 by the average of its real earnings per share over the previous ten years. The measure was developed by John Campbell and Robert Shiller in the late 1980s. Averaging a decade of earnings smooths the sharp swings in profits that occur within a single business cycle, which can make a one-year P/E look low at cyclical earnings peaks and high at earnings troughs.

The calculation is CAPE = real price ÷ average real EPS over the trailing 120 months, with both the index price and reported (GAAP) earnings per share deflated by the consumer price index. The series here is built from Robert Shiller’s long-run monthly dataset, which splices the S&P 500 and its predecessor indexes back to 1871 and uses the monthly average of daily closing prices.

The main limitations are definitional. Accounting standards have drifted over the sample: rules adopted since 2001 — goodwill impairment and mark-to-market write-downs among them — make reported earnings fall harder in modern recessions than they did under older standards, which depresses the ten-year average and raises measured CAPE relative to earlier eras. The ten-year window itself is a convention rather than a finding; shorter or longer windows produce different levels. And like any price-to-earnings measure, CAPE does not adjust for prevailing interest rates, payout policy, or changes in index composition.

Forward price-to-earnings

The forward price-to-earnings ratio divides the current index level by the consensus of analyst estimates of earnings per share over the next twelve months — forward P/E = price ÷ expected 12-month EPS. Unlike the trailing ratio, whose denominator is realized, published earnings, the forward ratio’s denominator is a forecast: it reflects what analysts collectively expect, and it is only as good as those expectations.

The series here is the aggregate forward P/E of S&P 500 firms as published by the Federal Reserve Board in its semiannual Financial Stability Report, which reports month-end values derived from LSEG’s I/B/E/S consensus estimates. The Fed’s published history begins in January 1990 and is republished in full with each edition, so the series lags the calendar by a few months between reports; the boxed figure above the chart shows the most recent weekly reading from FactSet for comparison.

The axis below starts at 1950 for comparability with the other two charts, but no data is shown before 1990 — and none can be: systematic collection of consensus estimates only began with the I/B/E/S database in 1976, and the long estimate histories built since then are proprietary datasets maintained by LSEG and FactSet. The Federal Reserve’s figure data is the longest publicly available span.

The equity yield gap

The equity yield gap is the difference between the earnings yield of the S&P 500 — trailing twelve-month earnings per share divided by price, the inverse of the trailing P/E — and the yield on the 10-year U.S. Treasury note. It compares what a dollar invested in the index currently earns in corporate profits with what the same dollar earns in government-bond interest. The gap can be negative: when the Treasury yield exceeds the earnings yield, the bond pays more per dollar than the index earns, and the chart’s solid horizontal line marks zero.

The chart plots (trailing 12-month EPS ÷ price) × 100 − 10-year Treasury yield, monthly, in percentage points. Earnings and prices are from Robert Shiller’s long-run dataset (monthly average of daily closing prices, reported GAAP earnings per share); the Treasury yield is the 10-year constant-maturity rate published by the Federal Reserve (FRED series GS10). Because reported earnings arrive with a lag, the series ends at the last Shiller month with earnings data.

The measure’s best-known limitation is that it compares unlike quantities: the earnings yield is a claim on profits that tend to grow with inflation, while the bond yield is nominal and fixed — a mismatch at the center of the academic critique of the so-called “Fed model.” The gap also embeds no adjustment for the different risk of the two assets, and its trailing-earnings numerator swings with the profit cycle, so the gap can widen in recessions because earnings collapsed rather than because bonds became less attractive.

Tobin’s Q

Tobin’s Q is the ratio of the market value of a firm — here, the aggregate U.S. nonfinancial corporate sector — to the replacement cost of its assets. The measure was proposed by James Tobin in 1969. The underlying idea is arithmetic: if the market values corporate capital above what it would cost to rebuild it, new investment is rewarded; if below, buying existing capacity through the stock market is cheaper than building it.

The chart plots market value of corporate equities ÷ corporate net worth, quarterly: the equity liability of nonfinancial corporate business (FRED series NCBEILQ027S) divided by the sector’s net worth measured at market and replacement value (FRED series TNWMVBSNNCB), both in millions of dollars from the same vintage of the Federal Reserve’s Financial Accounts (Z.1). Like the Buffett indicator, it is published roughly ten weeks after quarter-end, so the latest reading lags the calendar by about one quarter.

The denominator is the measure’s weak point. Replacement cost is estimated, not observed, and it captures intangible assets — software, brands, research — poorly, which matters more as the corporate sector has shifted toward intangible-heavy business models; that shift pushes measured Q upward for reasons unrelated to the price of any individual asset. The Z.1 data is also revised, sometimes substantially, and the theory gives no fixed level at which Q should settle.

Margin debt to GDP

Margin debt is money that investors borrow from their brokers against securities they hold, most commonly to buy more securities. Scaling the outstanding total by nominal GDP turns a dollar amount that naturally grows with the economy into a comparable ratio across decades. It is a measure of one form of investor leverage rather than of valuation: it says how much borrowed money is financing securities positions, not whether prices are high or low relative to fundamentals.

The chart plots broker-dealer margin loans ÷ nominal GDP × 100, quarterly, in percent. Margin loans are the amounts security brokers and dealers report as receivable from customers in the Federal Reserve’s Financial Accounts (Z.1, via FRED); nominal GDP is from the Bureau of Economic Analysis (FRED series GDP), seasonally adjusted at an annual rate.

The series captures only one channel of leverage. It excludes securities-based loans made by banks, leverage embedded in options and futures, and borrowing by hedge funds through prime brokerage and repo, all of which have grown relative to classic margin accounts. Reporting definitions have also shifted over the decades as the brokerage industry consolidated and rules changed, and the ratio tends to move with prices themselves, since the collateral backing the loans is the same market being measured.

CBOE Volatility Index (VIX)

The VIX measures the volatility of the S&P 500 that option prices imply over the next 30 days, expressed in annualized percentage points. It is computed by Cboe from the mid-quotes of a strip of near-term S&P 500 index options across strike prices — in outline, VIX = 100 × √(annualized 30-day option-implied variance). Because option buyers pay more when they expect larger price swings, the index summarizes what hedging insurance costs at each moment.

The chart plots the closing value on the last trading day of each month, from the daily VIXCLS series on FRED. Cboe introduced the index in 1993 and, after a methodology revision in 2003, published a back-computed history that begins in January 1990. The axis below starts at 1950 for comparability with the other charts, but no data exists before 1990.

The VIX is a measure of expected volatility, not of market direction or valuation, and the expectation is not a forecast that must be borne out — implied volatility has historically tended to run above subsequently realized volatility. Month-end sampling also misses spikes that begin and fade within a month, so the plotted series understates the index’s intramonth extremes.

Relative Strength Index (RSI)

The Relative Strength Index is a momentum oscillator introduced by J. Welles Wilder Jr. in 1978. It compares the size of recent gains with the size of recent losses and maps the result onto a fixed 0–100 scale: RSI = 100 − 100 ÷ (1 + RS), where RS = average gain ÷ average loss over the previous 14 periods, with each average updated by Wilder’s smoothing (each new period carries 1/14 weight). Readings rise when advances dominate declines and fall when declines dominate; 70 and 30 are the commonly used upper and lower reference levels, shown as dashed lines on the chart.

The series here applies the standard 14-period calculation to the monthly average S&P 500 price in Robert Shiller’s dataset, so each period is a month and the first plotted value falls in March 1951 — the fifteenth month of data from January 1950. Most practitioner use of RSI is on daily prices with a 14-day window; at a monthly frequency the oscillator moves far more slowly and visits its extremes far less often than the daily version.

RSI is computed from prices alone: it carries no information about earnings, rates, or any other fundamental. The 14-period window and the 70/30 reference levels are conventions from Wilder’s original book rather than estimated thresholds, and applying the formula to monthly averages of daily closes — rather than to closing prices themselves — further smooths the input before the oscillator sees it.

Market concentration (HHI)

The Herfindahl–Hirschman Index measures how evenly a total is divided among its parts. Applied to the S&P 500, it is the sum of the squared index weights of the constituents, scaled by 10,000 by the usual convention: HHI = 10,000 × Σ wi², where wi is a constituent’s share of total index market value. If all 500 names carried equal weight the index would read 20; a single name carrying all the weight would read 10,000. Because squaring emphasizes the largest weights, the measure rises when market value gathers in a handful of companies and falls when it spreads out.

No institution publishes a long HHI of S&P 500 weights as data, so the series here is computed from public records: the portfolio disclosures that the SPDR S&P 500 ETF Trust (SPY) files with the SEC. SPY replicates the index in full — it holds every constituent at index weight — so its filed portfolio weights stand in closely for index weights. Each plotted point is one filing: the schedule of investments in the trust’s N-30D shareholder reports (December 1995 through 2019), and the exact percentage-of-assets values in its Form N-PORT portfolio reports thereafter. Weights are taken over common-stock positions only — cash and money-market holdings are excluded — and renormalized to sum to one before squaring.

The series inherits the filing calendar rather than a monthly one: reports are annual through 2009, semiannual through early 2019, and quarterly since September 2019, so the early decades carry one or two points per year and moves between report dates are not visible. The first available schedule is dated December 1995, which is where the series begins; longer constituent-weight histories exist only in proprietary index datasets. The measure describes concentration within the S&P 500 — how its own market value is distributed across its 500 members — not the concentration of the whole U.S. equity market.