Do Expected Stock Returns Wear a CAPE? (EP.146)

Episode Summary

Executive Summary: This episode mixed community updates, market/news commentary, and a deep dive into forecasting stock returns. The hosts reviewed Archegos fallout, Madoff, index-fund criticism, and a sensational private credit pitch, then analyzed long-run equity return data from Dimson, Marsh, and Staunton. Their conclusion: historical equity premiums are heavily shaped by survivorship, repricing, and luck, so forward-looking expectations should be more conservative and may incorporate a modest earnings-yield signal, but without overconfidence in predictability.

Main Topics: Community updates and listener feedback: The hosts discussed positive reviews, thoughtful reactions to provocative episodes, merch updates, Instagram launch, Talking Sense cards, and upcoming guests such as Paul Merriman, Robert Novi Marks, Brad Cornell, and Larry Swedroe’s community participation. Archegos, Madoff, and market-manipulation headlines: They revisited Archegos’ collapse and its knock-on losses at major banks, reflected on Bernie Madoff’s legacy and investor recoveries, and highlighted the bizarre Hometown International deli valuation story as a symbol of speculation. Critique of index-fund concentration fears: The hosts commented on an Atlantic article warning that index funds may concentrate too much power, contrasting it with Adriana Robertson’s view that there is insufficient evidence of systemic harm to liquidity, price efficiency, or governance. Long-run equity risk premium research: A major segment examined why historical stock returns may overstate future expected returns. The discussion focused on survivorship bias, changing discount rates, and Dimson-Marsh-Staunton global return data to estimate a more realistic equity premium. Predictability, CAPE, and bootstrap simulations: They debated whether valuation ratios and historical returns meaningfully predict future returns, noting mixed evidence and out-of-sample problems. Their simulations compared rolling history, country outcomes, bootstraps, and a modest earnings-yield-based predictability model. Bad advice of the week: private credit marketing: They criticized the framing of a private credit product marketed as having ‘bond-like stability’ and ‘equity-like returns,’ arguing that exclusivity, leverage language, and lockups can obscure risk and mimic old hedge-fund-style sales tactics.

Key Arguments: Historical U.S. stock returns are not a reliable standalone guide for future expectations because the U.S. experienced survivorship bias and unusually favorable long-run outcomes. Equity returns are partly driven by changing discount rates and valuation multiples; if those repriced upward in the past, future returns should not assume the same boost. Dimson, Marsh, and Staunton’s global data provide a better basis for forward-looking estimates because it includes failed and underperforming markets. The historical equity premium should be decomposed into yield, dividend growth, and price-multiple expansion; only the first two are reasonable components to extrapolate. Long-horizon return predictability looks compelling in-sample but weakens or disappears out-of-sample, so CAPE-style timing should be treated cautiously. Bootstrap simulations may look too wide, but their distribution aligns with the real dispersion seen across individual countries, suggesting tail risk should not be dismissed. A small contribution from current earnings yield is economically intuitive, even if the statistical evidence for predictability is mixed. Marketing phrases like ‘bond-like stability, equity-like returns’ often imply a free lunch and can understate the risks of illiquidity, leverage, and complexity.

Data Points: Archegos credit losses at Credit Suisse: $4.7 billion US - Hit taken by one lender after Archegos could not meet margin requirements. Archegos loss at Morgan Stanley: almost $1 billion - Another major bank reported a substantial hit from the Archegos collapse. Archegos loss at Nomura: $1 billion to $4 billion - Estimated loss range discussed for Nomura. Bernie Madoff fraud size: as much as $65 billion - Size of the Ponzi scheme attributed to Madoff. Madoff victims: as many as 37,000 people in over 100 countries - Scale of the defrauded investor base. Hometown International market cap: $101 million US - Valuation of the deli-linked public company discussed as a speculation example. Hometown deli sales over two years: $37,000 - Revenue cited for the single-location deli over the prior two years during COVID. Global equity risk premium (1900-2020): 4.4% geometric over U.S. Treasury bills - Dimson-Marsh-Staunton world index historical premium. U.S. equity risk premium (1900-2020): 5.8% geometric over U.S. Treasury bills - Historical U.S. premium over the same period. World real return (1900-2020): 5.2% geometric - Inflation-adjusted return for the world index. U.S. real return (1900-2020): 6.6% geometric - Inflation-adjusted return for U.S. stocks. World real return 1900-1949: 2.7% annualized - Example of poor realized returns despite initial optimism. World real return 1950-1999: 8.6% annualized - Example of strong realized returns despite widespread pessimism at the start of the period. Return decomposition: dividend yield: about 4.0% - Largest historical contributor to long-run equity returns. Return decomposition: real dividend growth: 0.7% - Historical growth component in the Dimson-Marsh-Staunton decomposition. Return decomposition: multiple expansion: 0.5% - Small contribution from valuation expansion over the full sample. Forward-looking geometric equity risk premium estimate: 3.5% - Dimson-Marsh-Staunton style estimate after adjusting for non-repeatable components. Arithmetic vs geometric adjustment: 1.5% - Difference added back to convert the forward-looking premium into an arithmetic-style input for simulations. World index median 30-year risk premium: 5.4% - Rolling historical World Index 50th percentile outcome. World index median 60-year risk premium: 5.7% - Rolling historical World Index 50th percentile outcome. World index 10th percentile 30-year premium: 1.2% - Rolling historical World Index downside scenario. World index 10th percentile 60-year premium: 4.6% - Rolling historical World Index downside scenario over 60 years. Individual-country 10th percentile 30-year premium: -0.8% - Shows how weak some country outcomes can be over long horizons. Individual-country 10th percentile 60-year premium: 1.6% - Long-horizon downside across country outcomes; matched by bootstrap. Bootstrap 10th percentile 30-year premium: 0.4% - Distribution from 10,000 bootstrap simulations of world returns. Bootstrap 10th percentile 60-year premium: 1.6% - Matched the individual-country 10th percentile over 60 years. Predictability model R-squared: 0.2 - Simulation assumption that 20% of return variation is explained by earnings yield.

Pivotal Quotes: "It's not the notes you play, it's the notes you don't play." — DragonFB (quoted by hosts): Listener review praising the hosts’ restraint during guest interviews. "Do Expected Stock Returns Wear a Cape?" — Benjamin Felix: Playful title for the main segment on expected returns and forecasting. "bond-like stability, equity-like returns" — Transcripted marketing claim from private credit article: The phrase the hosts criticized as misleading and sales-oriented.

Implications: Investors should be wary of extrapolating U.S. history or seductive return stories. Better planning uses conservative long-run assumptions, recognizes wide outcome ranges, and treats valuation signals as modest inputs—not timing tools or free-lunch guarantees.