Behavioral Biases and the Asset Growth Anomaly

Abstract

We find evidence that the asset growth anomaly is due, in part, to investors’ behavioral biases. Two-way sorts based on asset growth and proxies for known behavioral biases (anchoring, recency, nominal price illusion, and lottery-seeking) indicate that the asset growth anomaly is stronger in stocks that investors affected by behavioral biases tend to buy and non-existent or negative in stocks they tend to sell. These results are not explained by limits of arbitrage or investor sentiment and hold in both portfolio analyses and regressions. The evidence suggests that behavioral investors’ attraction to certain stocks drives the asset growth anomaly.

Keywords:

• Asset growth
• Behavioral biases
• Anchoring
• Recency
• Lottery-seeking
• Nominal price

JEL Classifications:

• G12
• G14
• G18

Acknowledgments

The authors appreciate the comments from seminar participants at Financial Management Association annual meetings and Carson College of Business of Washington State University. Zhang acknowledges financial support from the College of Business at California State University, Chico in the form of summer research award. All errors are our own.

Notes

1 See, for example, Titman, Wei, and Xie (2004); Anderson and Garcia-Feijoo (2006); Xing (2008); Cooper, Gulen, and Schill (2008); Fama and French (2008); Polk and Sapienza (2009); Lyndres, Sun, and Zhang (2008); Lipson, Mortal, and Schill (2011); among others. Watanabe et al. (2013) confirm this relation outside of the US market. Wahal (2019), however, shows that there is no strong relationship between asset growth and future returns in the US for the pre-1963 period.

2 The theoretical and empirical studies include, among others, Cochrane (1991, 1996); Liu, Whited, and Zhang (2009); Lyndres, Sun, and Zhang (2008); Li, Livdan, and Zhang (2009); Berk, Green, and Naik (1999); Carlson, Fisher, and Giammarino (2004); Watanabe et al. (2013).

3 This line of research supporting behavioral explanations includes, among others, Titman, Wei, and Xie (2004); Cooper, Gulen, and Schill (2008); Polk and Sapienza (2009); Li and Zhang (2010); Lam and Wei (2011); Lipson, Mortal, and Schill (2011); Stambaugh, Yu, and Yuan (2012); Mao and Wei (2015); Papanastasopoulos (2017).

4 Related research can be found in Shue and Townsend (2017), Baker and Wurgler (2004a, 2004b), Baker, Nagel, and Wurgler (2006), and Hartzmark and Solomon (2017, 2018), Brunnermeier and Julliard (2008), and Svedsäter, Gamble, and Gärling (2007). See also Kumar (2009).

5 Excess returns of the portfolios also significantly differ between the two extreme Ag quintiles. The asset growth anomaly does become weaker when the portfolio returns are value-weighted. The magnitude of the anomaly is smaller than that in Cooper et al. (2008) primarily because our sample excludes the low-priced and mini-cap stocks. Our main conclusions are robust to the sampling procedure.

6 It is worth noting that some of these stock/firm characteristics are not monotonic across the five quintiles. The two extreme quintiles (lowest and highest asset growth) tend to have extreme average values of these characteristics.

7 As discussed earlier, the relationship is not monotonic. One possible explanation is that, compared to firms with no or no unexpected asset growth, those with very low or likely negative asset growth involve greater valuation uncertainty. This nonmonotonicity, however, appears dominated by the general relationship between asset growth and valuation uncertainty, which is what our paper focuses on.

8 In our analyses we use independent two-way sorts. In unreported analyses, we use sequential sorts (first on AG, then on the Bb variable). The findings are quantitatively similar and yield the same conclusion.

9 While our main analyses are based on equal-weighted portfolio returns, those based on value-weighted portfolios yield same conclusions. For example, the monthly alpha for the value-weighted BbSell portfolio (based on RPR) is -0.17% (t=-1.37); the monthly alpha for the value-weighted BbBuy portfolio is 0.64% (t=4.48); the difference in the value-weighted monthly alphas between the BbBuy and BbSell portfolio is 0.81% (t=4.38).

10 The unreported further analyses include a falsification test, an out-of-sample analysis on other anomalies, alternative proxies for asset growth and investments, factors beyond Fama and French (1993), more restrictive samples, etc. For brevity, these results are not reported but are available from the authors upon request.

11 The investor sentiment data are from Professor Jeffrey Wurgler’s website http://people.stern.nyu.edu/jwurgler/. We thank Professor Wurgler for making the data available.

12 The conclusion would be made stronger with an alternative definition of volatility or alternative limits of arbitrage proxies. In unreported analysis, we use market capitalization, for example, as a (possibly lesser) proxy for limits of arbitrage and find strong evidence of the Bb effect among both the lower and higher market capitalization subsamples.