Fundamental Valuation

Is the evolving set of artificial intelligence (AI) platforms based on large language models interesting for selection of stocks that are deeply undervalued? Are they monolithic, or diverse? As a simple exploration, we pose to each of Grok, ChatGPT, Claude, Perplexity and Gemini the following prompt regarding undervalued U.S. stocks:

Adopt the persona of a deep value investor seeking the most undervalued publicly listed U.S. stocks of any size. List the three stocks most attractive to you as medium-term holdings. Do not provide any explanations.

We then compare and contrast results from AI panel members. Using responses to the prompt as posed at the end of June 2026, we find that: Keep Reading

Can artificial intelligence (AI) in the form of large language models (LLM) improve upon existing methods to explain stock returns around earnings announcements? In the June 2026 version of their paper entitled “Assessing the Benefits of Optimized Agentic AI Systems for Asset Pricing”, Ralph Koijen and Bradford Levy employ a a real-time, out-of-sample benchmark for evaluating optimized LLMs while avoiding lookahead bias and market adaptation effects. The benchmark measures how well AI systems explain stock returns around earnings announcements using only information available at announcement time (especially the announcement text). Optimized means experimenting with LLM prompts to improve results, such as guiding LLMs to assess earnings relative to expectations. Using earnings call transcripts, analyst consensus earnings and associated daily returns for U.S. stocks during the fourth quarter of 2025 (1,849 earnings announcements), they find that: Keep Reading

Under what conditions does equity market time series momentum (TSMOM) work and not work? In their June 2026 paper entitled “Boundaries of Time Series Momentum”, Matti Suominen and Erik Hjalmarsson examine performance of equity market TSMOM across ranges of three valuation metrics: cyclically adjusted price-to-earnings ratio (CAPE), dividend yield and term spread (difference between long-maturity and short-maturity Treasury instrument yields). They specify TSMOM as long (short) the market when past market return in excess of the risk-free rate over a specified lookback interval is positive (negative). They specify a Boundaries variable for predicting TSMOM performance as follows:

1. Scale each of the CAPE, dividend yield and term spread to values between -1 and 1 as follows:
   1. Subtract from its 12-month average the past 10-year or 20-year minimum observation and divide the difference by the past 10-year or 20-year range (maximum minus minimum).
   2. Multiply results by two and subtract one.
2. Compute a Boundaries variable as the square of the scaled term spread plus the square of scaled CAPE or scaled dividend yield.

For a given equity market, they construct a TSMOM index as an equal-weighted average of 25 time series momentum strategies, with lookback and investment intervals of 1, 3, 6, 9 or 12 months. They then explore how index returns interact with the Boundaries variable. Using the specified inputs and stock index returns during July 1927 through December 2024 for the U.S. and during January 1989 through December 2024 for a 20-country international sample, they find that: Keep Reading

The conventional Cyclically Adjusted Price-Earnings ratio (CAPE), or P/E10, divides current real S&P 500 Index level by average annual aggregate real index earnings as reported over the prior 10 years. Is there a more useful way to aggregate stock-level information? In the June 2026 revision of their paper entitled “CAPE Ratios and Long-Term Returns”, Rui Ma, Ben Marshall, Nhut Nguyen and Nuttawat Visaltanachoti introduce Component CAPE, for which they each year:

1. Calculate the CAPE ratio for each constituent stock based on either five or 10 years of past earnings.
2. Weight individual stock CAPE ratios either by market capitalization (value) or by earnings.

They measure CAPE predictive performance via constant slope regression to calculate out-of-sample (OOS) R-squared, which relates the mean squared error of 10-year return predicted by CAPE to that predicted by historical average return. They start CAPE calculations in 1964, with OOS predictions commencing in 1974. To assess economic value of predictions, they calculate the certainty equivalence return (CER) for investors with power or quadratic risk aversions of a stocks/Treasury bills allocation strategy based on CAPE signals. Using S&P 500 Index level and aggregate earnings data and annual price and earnings data for individual S&P 500 components during 1955 through 2024, they find that:

Keep Reading

The home page, Simple Asset Class ETF Momentum Strategy (SACEMS) and Simple Asset Class ETF Value Strategy (SACEVS) now show preliminary positions for July 2026. SACEMS rankings probably will not change by the close. SACEVS allocations are unlikely to change by the close.

A subscriber commented and asked:

“I have been wondering whether Morningstar’s estimate of ‘fair value’ for stocks has any relationship to actual subsequent returns. For instance, when the fair value is more than double the market price, is the return over the next year substantially greater than for stocks where the fair value is less than half the market price? I haven’t found anything like this on your site, only an assessment of mutual fund ratings which is a very different matter.”

As an alternative, tractable investigation, we hand collect end-of-month Morningstar price relative to Fair Market Value (P/FMV) for the overall stock market from their 5-year historical chart (extended backward by a prior collection). We then relate those P/FMVs to monthly returns for SPDR S&P 500 ETF (SPY) over the same period. Using monthly Morningstar P/FMVs and SPY returns during May 2020 through May 2026, we find that: Keep Reading

Is the evolving set of artificial intelligence (AI) platforms based on large language models interesting as niche fundamental analysts? Are they monolithic, or diverse? As a simple exploration, we pose to each of Grok, ChatGPT, Claude, Perplexity and Gemini the following two prompts regarding projected aggregate revenue and net profit for major AI platforms:

1. Applying all available training data and real-time updates, please project the aggregate revenue for major AI platforms year-by-year over the next five years (2027-2031). Your projection should account for realistic supply (including limits on offeror buildout of capabilities and global competition) and demand (including limits on customer diversion of funds and prospects for ROI) and not just rely on offeror forecasts. Do not provide any explanations.
2. Now repeat this exercise for aggregate net profit. Do not provide any explanations.

We then compare and contrast results from AI panel members. Using responses to these prompts as posed at the end of May 2026, we find that:

Keep Reading

Do the record market capitalizations of firms focused on artificial intelligence (AI) reasonably reflect future profit growth, after accounting for huge investments in data centers and ongoing costs to train and deploy AI models? In his May 2026 paper entitled “Valuation and Accounting Issues in the AI Ecosystem”, Andreas Haaker examines valuation and accounting problems associated with the AI boom, with focus on capital market narratives, accounting practices and free cash flow generation. Based on AI firm stock prices, their reports/projections and general accounting principles, he concludes that:

Keep Reading

Are prediction markets better at forecasting firm earnings than professional analysts? In their April 2026 paper entitled “Beating the Earnings Game: Why Do Prediction Markets Outperform Professional Analysts?”, Daniel Rabetti, Jiaqi Shao and Che Zhang investigate whether and, if so, why a blockchain-based prediction market such as Polymarket outperforms professional analysts in forecasting U.S. stock earnings. The earnings predictions of this market are public and unchangeable contracts, taking the form:

“Will [Company] beat earnings for [Quarter] [Fiscal Year]?”

relative to analyst consensus as of contract creation date. Using data for 469 Polymarket firm-quarter earnings beat contracts, corresponding analyst earnings forecast data and associated daily stock prices during September 2025 through February 2026, they find that:

Keep Reading

Will the relatively high U.S. stock valuation ratios observed over the past few decades revert, or are they persistent artifacts of fundamental shifts in the U.S. economy? In their January 2026 paper entitled “A Macroeconomic Perspective on Stock Market Valuation Ratios”, Andrew Atkeson, Jonathan Heathcote and Fabrizio Perri examine the interplay between economic data (share of labor in corporate output and corporate investment/capital base) and stock market valuation ratios. They derive aggregate U.S. corporate value from the Integrated Macroeconomic Accounts (IMA). Their measure of enterprise value differs from stock market capitalization in two ways:

1. It is insensitive to the mix of debt and equity used for firm financing.
2. It includes estimated value U.S. subsidiaries of foreign multinationals and excludes estimated value of the foreign subsidiaries of U.S. multinationals. Thus, it measures the value of entities filing U.S. corporate tax returns.

Using U.S. economic and corporate valuation data during January 1952 through September 2025, they find that: Keep Reading