between asset growth and future stock returns at the cross-sectional level is referred to as the asset
While both rational and mispricing explanations of the asset growth anomaly have been proposed,
much of the recent evidence in the literature has been focused on rational explanations via q-theory.
The latter is largely inconclusive,
and mispricing as an explanation has received considerably less
attention. Given the mixed evidence to date, the purpose of this paper is to extend our understanding of
the drivers of this phenomenon by focusing on investor overreaction as an explanation of the asset
growth anomaly. Such a conjecture has its origin in the anomaly literature (Cooper et al., 2008).
Extending their finding, if the asset growth anomaly is driven by investor overreaction to asset growth
news, it follows that growth firms will be more likely to experience such an anomaly, as they should
generate more growth news to be extrapolated by investors.
We therefore develop two hypotheses concerning investor reaction to asset growth. First, if
overreaction to growth explains the anomaly, then growth firms should show a stronger asset growth
anomaly than mature firms.
Second, the behavioral bias of representativeness suggests that investors
overreact more to a series of events, and therefore, if overreaction is the explanation, then firms
experiencing a longer series of high (low) asset growth should experience lower (higher) subsequent
returns; thus, the anomaly would be stronger for firms experiencing a sequence in their growthpattern.
It is often argued that investors extrapolate a sequence of same-signed news, which results in an
overreaction to the information (see, e.g., the theoretical model of Barberis, Shleifer, & Vishny, 1998).
To test the first hypothesis we employ three growth-type proxies: retained earnings scaled by firm
total assets, the dividend-to-income ratio, and cash flow scaled by firm total assets. Growth firms tend
to have lower retained earnings, dividends, and cash flow than mature firms (see Anthony & Ramesh,
1992; DeAngelo, DeAngelo, & Stulz, 2006; Dickinson, 2011). Using US data from 1963 to 2011, we
show, by comparing the hedged returns of asset growth-sorted portfolios for growth and mature firms,
that growth firms demonstrate a stronger asset growth anomaly. Sorting on asset growth within growth
type yields an annualized return difference between growth and mature firms of 20%, 18%, and 13%,
respectively, for the three proxies of growth type −retained earnings, dividend-to-income ratio, and
cash flow. The Fama–MacBeth regression results provide further support: for example, they show that
The rational explanation of the asset growth anomaly relies on the q-theory model, which studies the investment–return
relationship from a production-based asset pricing or firm optimal investment standpoint (e.g., Chen & Zhang, 2010;
Cochrane, 1991, 1996; Li & Zhang, 2010; Li, Livdan, & Zhang, 2009). The basic argument is that firms with low discount
rates (expected returns) have high net present values and high investment, whereas firms with high discount rates have low
net present values and low investment. Li and Zhang (2010) show that limits to arbitrage dominates q-theory in explaining
the asset growth anomaly. Watanabe et al. (2013) favor the optimal investment explanation by using global stock markets;
they find that the asset growth anomaly is stronger in more advanced markets where stocks are more efficiently priced.
Finally, Lam and Wei (2011) present evidence to support both limits to arbitrage and q-theory.
Firm life cycle has been shown to be a useful dimension in understanding the cross-sectional variation of corporate finance
decisions and accounting ratios. For example, Hirsch and Walz (2011) show that firms in different life-cycle patterns make
different financing decisions and that these decisions interact with future growth and development decisions. Hribar and
Yehuda (2015) study the mispricing of accrual and cash flow information by the stock market in different firm life-cycle
According to Barberis et al. (1998), after a trend of good or bad information, representativeness causes investors to
overreact to information and push the price too high. Hong and Stein (1999) argue that momentum traders make decisions
conditional on past price change; that is, they push stock prices higher (lower) when there is an up (down) trend.
Alti and Tetlock (2014) use a structure model approach to study the influence of behavioral biases on asset prices. They
also identify over-extrapolative belief as the main cause of mispricing.
CAI ET AL.