Fundamental trend following uses moving averages of past fundamental data, such as valuation metrics or economic indicators, to predict future fundamentals, analogously to the conventions in price or return trend following. A recent paper shows that fundamental trend following can be applied to equity earnings and profitability indicators. One approach is to pool fundamental information across a range of popular indicators and to sequentially choose lookback windows for moving averages in accordance with past predictive power for returns. The fundamental extrapolation measure predicts future stock returns positively and would historically have generated significant profits. Most importantly, fundamental trend following returns seems to have little correlation with price trend following returns, supporting the idea that these trading styles are complementary.

Huang, Dashan, Huacheng Zhang, Guofu Zhou, and Yingzi Zhu, (2020), “Fundamental Extrapolation and Stock Returns”

The below are quotes from the above paper. Cursive text and text in brackets have been added for clarity and to interpret the research findings for the purpose of trend following.
The post ties up with this site’s summary on information inefficiency of financial markets.

What is fundamentals extrapolation?

“Fundamental extrapolation has received increasing attention in behavioral finance…Fundamental extrapolation…explains a number of stylized facts such as risk premium, price momentum, long-run reversal, value premium, and bubbles.”

“[The fundamental extrapolation] approach closely follows price extrapolation, which has a long history in finance. Price moving averages are widely used in…trading systems that attempt to predict future prices based on the past. Similarly, we use moving averages of past fundamentals to predict future fundamentals. Our approach is purely fundamental because we use only fundamental information…Our approach can be easily applied to all stocks in the US and to global stock markets. It can also be applied to bond and other asset classes.”

Why do investors extrapolate? From the behavioral finance point of view, there are two categories of economic explanations: mechanisms based on concepts from psychology and mechanisms based on bounded rationality. To the best of our knowledge, the possible drivers of fundamental extrapolation include representativeness heuristic, information diffusion, the law of small number, natural expectation, experience effect, diagnostic expectation, and fading memory.”

Extrapolating fundamentals for the equity market

“We assume that investors use a moving average of past fundamental values, which captures the past trend, to do fundamental extrapolation to forecast the future values.”

We consider 10 popular fundamental variables [for stocks]…return on equity, return on assets, earnings per share, earnings to price ratio, operating profitability, cash-based operating profitability to assets, earnings surprise, gross profitability to assets, revenue surprise, and net payout ratio. These fundamental variables are earnings- and profitability-related, and are relevant to investors to use for valuation purposes.”

“We assume, similar to price extrapolation, that investors use the average of the most recent L observations to extrapolate future values. However, since there is no theoretical guidance as to which L investors may use in practice, we assume that investors simply determine this by running regressions to see which L works best in the past. Note that we run regressions of returns on fundamentals. This is because investors are interested in how their extrapolated fundamentals help them to obtain better extrapolated returns, i.e., the primary goal here is investment performance… A simple way of choosing the optimal window L to select the one that yields the maximum adjusted R2 [coefficient of determination].”

“Instead of assuming that investors use some arbitrary lag of fundamentals to form their stock return expectations, we allow them to select the lag and to pool information across fundamental variables. Although our approach seems slightly more complex than just performing regressions, it is at the same level of complexity and no more complex than a typical theoretical model. Indeed, our approach is really intuitive. As investors do not know ex-ante which lag of the moving averages they should rely on, the natural way is to see which one does better in the past regressions.

“The firm-level fundamental extrapolation measure, or fundamental extrapolated return… has four desirable properties.

  • First, it is data-driven and based on the past predictive power, and can adapt to the time-varying predictability of the fundamentals…
  • Second, it is entirely dependent on past data and does not suffer from any looking-forward bias.
  • Third, it accommodates any number of fundamental variables, instead of using only one as common in the literature.
  • Finally, it is based on pure fundamental variables, and does not use survey/expectation data that may contain information beyond firm fundamentals.”

Empirical findings

“We show that our fundamental extrapolation measure predicts future stock returns positively…The predictive power…on future stock returns is confirmed by Lewellen’s regression slope test and is robust to a number of different set-ups and to international stock markets.”

“A strategy based on fundamental extrapolation measure generates…economically and statistically significant profits in the stock market as a whole.”

“At the end of each month, we sort stocks into five groups based on fundamental extrapolation…Within each group, we construct a value-weighted portfolio and hold it for one month. The average returns monotonically increase in the fundamental extrapolation measure, from 0.14% for the low fundamental extrapolation portfolio to 0.94% for the high fundamental extrapolation portfolio, suggesting that a strategy that buys the high fundamental extrapolation portfolio and sells the low fundamental extrapolation portfolio earns a monthly average return of 0.80% (t-value = 5:02). [The figure below] shows that if one investor invests $1 in the high fundamental extrapolation portfolio at the beginning of June 1975 and rebalances it on a monthly basis, he would earn $475 at the end of December 2018. By contrast, if he invests $1 in the low fundamental extrapolation portfolio or the market portfolio, he would make only $6 or $106.”

Fundamentals-based versus conventional trend following

“We compare fundamental extrapolation with price extrapolation and find that fundamentals matter more after all, although they are complementary to each other in extracting return information.”

“The well-known price extrapolation that yields the popular momentum factor, has a lower average return [0.7% per month versus 0.8% for an equivalent fundamental extrapolation strategy] with the same value-weighting over the sample period. Moreover, while price momentum has high crash risk with its two worst monthly returns being -39% and -32%, fundamental extrapolation seems more crash-proof, with two worst monthly returns only at -16% and -14%…The two strategies have a low correlation of 0.13, and combining them generates a 1.48% average return per month, roughly equal to 0.8% plus 0.7%.”

“Thus, in practice, one investor can improve her investment performance by buying high fundamental extrapolation winner stocks and selling low fundamental extrapolation loser stocks.”

Theoretical explanations

“We show that fundamental extrapolation has dual effects on stock price: a cash flow effect and a discount rate effect. The former pushes stock price up relative to its fundamental value, whereas the latter increases the representative investor’s expected volatility and depresses today’s stock price. Our empirical results suggest that the discount rate effect dominates the cash flow effect overall.”

“On the one hand, fundamental extrapolation with high past cash flows makes the investor optimistic about future cash flows and pushes the asset’s price up relative to the current fundamental value, which generates a cash flow effect and predicts a low future return. On the other hand, the extrapolative expectation about future cash flows also increases the investor’s expected volatility and depresses today’s stock price, which generates a discount rate effect and predicts a high future return.”

N.B.: A simpler but plausible explanation of a positive relation between fundamental trends and future returns are limits of information efficiency (view summary article here). For most investors there are too many fundamental indicators to monitor in real-time. Meanwhile, algorithmic trading systems to this day pay much more attention to price trends than to fundamental trends.