Key Takeaways
- • Andreasen and Bro, in a Review of Finance paper (2026), propose a present-value model that splits stock prices into expected future dividends, expected future returns, and a rational bubble, using survey-based expectations alongside market data.
- • Using monthly U.S. data from January 1971 through December 2024, the estimated bubble accounts for 61% of the variation in the S&P 500 price-dividend ratio, while expected returns account for another 42% (the authors note these are not additive shares).
- • Positive bubble shocks are followed by lower unemployment, higher industrial production and longer working hours, with only a modest, generally insignificant inflation response.
- • Positive bubble shocks are followed by higher policy rates, peaking at roughly 48 basis points eight months after a one-standard-deviation shock, alongside narrower corporate bond spreads.
- • The model estimates the bubble at nearly 25% of the price-dividend ratio in December 2024, which the authors tie to possible investor enthusiasm for artificial intelligence; they stress this is a model-based estimate, not proof of overvaluation.
Summary
A paper by Elisabetta Basilico, PhD, CFA, published on Alpha Architect on October 19, 2026, summarizes research by Andreasen and Bro, published in the Review of Finance in 2026, that seeks to separate fundamental drivers of stock prices from speculative bubbles. The authors argue that a high price-dividend ratio alone does not establish a bubble, since investors may rationally pay more for stronger expected dividend growth or lower required returns. Their approach augments a present-value framework with survey-based expectations about dividend growth and stock prices to isolate a non-fundamental, rational bubble component.
Applied to monthly U.S. data from January 1971 through December 2024, the model attributes 61% of the variation in the S&P 500 price-dividend ratio to the estimated bubble, with expected returns accounting for a further 42%. Expected dividend growth contributes little over the full sample. Removing the estimated bubble substantially reduces excess volatility in valuations and eliminates the explosive dynamics detected in the unadjusted ratio. The estimated bubble tracks several historical episodes: roughly 10% of the price-dividend ratio during the early-1970s Nifty Fifty boom, about 2% in the 1980s, 27% in January 2001 during the dot-com era, and approximately 19% at the November 2021 COVID-era peak. In December 2024 it reached nearly 25%, which the authors associate with possible enthusiasm for artificial intelligence.
The authors also examine macroeconomic effects using a structural vector autoregression, with bubble innovations as an instrument. Positive bubble shocks are followed by lower unemployment, higher industrial production and increased working hours. The authors suggest two channels: rising household wealth supporting spending, and higher equity values easing corporate financing constraints. Inflation rises only modestly. Monetary policy also responds, with the policy rate peaking at about 48 basis points eight months after a positive one-standard-deviation shock, while the long-term/short-term Treasury spread declines and corporate bond spreads narrow. The paper does not establish that policymakers respond to the bubble directly rather than to its economic effects, and it does not show that negative shocks would produce symmetric effects.
For practitioners, the paper's guidance centers on avoiding exclusive reliance on conventional valuation ratios, recognizing that bubbles can build and deflate gradually so identification does not signal timing, and incorporating bubble risk into scenario analysis, concentration review and capital market assumptions. The authors' framework is presented as a model-based historical estimate and the accompanying results are hypothetical and not indicative of future performance.
Applied to monthly U.S. data from January 1971 through December 2024, the model attributes 61% of the variation in the S&P 500 price-dividend ratio to the estimated bubble, with expected returns accounting for a further 42%. Expected dividend growth contributes little over the full sample. Removing the estimated bubble substantially reduces excess volatility in valuations and eliminates the explosive dynamics detected in the unadjusted ratio. The estimated bubble tracks several historical episodes: roughly 10% of the price-dividend ratio during the early-1970s Nifty Fifty boom, about 2% in the 1980s, 27% in January 2001 during the dot-com era, and approximately 19% at the November 2021 COVID-era peak. In December 2024 it reached nearly 25%, which the authors associate with possible enthusiasm for artificial intelligence.
The authors also examine macroeconomic effects using a structural vector autoregression, with bubble innovations as an instrument. Positive bubble shocks are followed by lower unemployment, higher industrial production and increased working hours. The authors suggest two channels: rising household wealth supporting spending, and higher equity values easing corporate financing constraints. Inflation rises only modestly. Monetary policy also responds, with the policy rate peaking at about 48 basis points eight months after a positive one-standard-deviation shock, while the long-term/short-term Treasury spread declines and corporate bond spreads narrow. The paper does not establish that policymakers respond to the bubble directly rather than to its economic effects, and it does not show that negative shocks would produce symmetric effects.
For practitioners, the paper's guidance centers on avoiding exclusive reliance on conventional valuation ratios, recognizing that bubbles can build and deflate gradually so identification does not signal timing, and incorporating bubble risk into scenario analysis, concentration review and capital market assumptions. The authors' framework is presented as a model-based historical estimate and the accompanying results are hypothetical and not indicative of future performance.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.