Is there a single variable based on accounting data that reliably captures expected returns of individual stocks? In their October 2018 paper entitled “A Fundamental Factor Model”, Stephen Penman and Julie Zhu construct and test a fundamental expected returns factor based on array of accounting inputs, encompassing earnings, book value and items that sum to these income statement and balance sheet totals. They focus on a robust version of this factor incorporating eight of these inputs (ER8), but consider simpler versions relying on only four (ER4) or two (ER2) inputs. They calculate a premium based on a portfolio that is each month long (short) the equally weighted stocks of firms ranked in the top (bottom) three tenths, or deciles, of the fundamental factor. They update fundamentals yearly three months after firm fiscal year ends from numbers published in annual financial statements. In terms of smart beta terminology, their approach replaces market capitalization weights with fundamentals weights. Using monthly returns and annual financial statements for a broad sample of non-financial U.S. common stocks during April 1981 (or June 1975 or April 1966 for simplified factors) through December 2015, they find that:
- ER8 is related to earnings growth expectations and risks associated with those expectations.
- ER8 relates slightly negatively to the market factor, strongly positively to the conventional HML (high minus low book-to-market ratio) factor and positively to investment factors used in other models.
- ER8 works well in explaining contemporaneous cross-sectional stock returns and, moreover, generates significant alphas relative to returns explained by the most widely used models that employ three, four or five factors (indicating incremental information).
- ER8 works well in predicting next-month cross-sectional stock returns. Specifically:
- Excluding microcaps (bottom 20% of NYSE market capitalizations), ER8 generates average monthly gross premium 0.57% and annualized gross Sharpe ratio 0.73, among the highest of widely accepted stock factors.
- Using simplified versions ER4 and ER2 lowers average monthly gross premiums to 0.48% and 0.36% and annualized gross Sharpe ratios to 0.61 and 0.37, respectively.
- Including microcaps boosts ER8 average monthly gross premium to 0.793% and annualized gross Sharpe ratio to 1.17.
- Since ER8 has approximately zero market beta, the 2-factor market-ER8 model of stock returns combines a mean-variance efficient portfolio with a portfolio that hedges consumption and investment risks.
In summary, evidence indicates that a robust fundamental expected returns factor based on eight accounting inputs may be useful and comprehensive for screening U.S. stocks.
Cautions regarding findings include:
- Reported factor premiums are gross, not net. Accounting for the trading frictions associated with factor portfolio rebalancing and shorting costs would reduce these premiums. Moreover, shorting of some stocks as specified may be costly/infeasible due to lack of shares to borrow. Equal weighting of ERx factor portfolios exacerbates this concern.
- The approach is beyond the reach of many investors, who would bear fees for delegating to an investment/fund manager.
