Stock return predictability when growth and accrual measures are negatively correlated

Purpose For most companies, growth measures such as asset growth are positively correlated with accrual measures. Just like investment in fixed assets, current accrual represents one form of investment and is an integral part of a firm's business growth. This makes it difficult to distinguish between the growth-based and earnings quality-based interpretations of the accrual effects, because high accruals can represent both high growth and inflated earnings. The purpose of this paper is to add to the literature by examining an issue that has not received much attention: the correlation between asset growth and accruals and its implication on stock return predictability. The authors address the issue using Fama and Macbeth's (1973) cross-sectional regressions that are conditional on the correlations between the two variables. Design/methodology/approach The authors partition firms based on whether the correlation between current accrual and asset growth in the past five years is positive (rho+) or negative (rho-). The authors refer to these two types of firms such as "positive correlation" and "negative correlation" groups. For both groups, the authors examine whether firms with higher asset growth and higher accruals are associated with lower future stock returns. The authors implement Fama and MacBeth's cross-sectional regressions incorporating the effect of correlations between growth and accrual measures. In addition, the authors regress hedge portfolio returns on Fama and French (1993) three-factor and Fama and French (2015) five-factor models to see if the intercepts (a's) from these regressions are significantly different from 0. Findings For each year, the authors partition firms based on whether the correlation between asset growth and current accrual is positive or negative. For both the "positive correlation" and "negative correlation" firms, the authors examine the association between accruals and future stock returns. The authors find that accruals remain strong in predicting future stock returns for both groups. The accrual effects from the "negative correlation" group cannot be attributed to the growth-based hypothesis because for these firms, when accruals are high, growth measures tend to be relatively low and vice versa. The effects are most likely driven by the alternative hypothesis that investors overvalue the accrual part of earnings. Research limitations/implications - There exist a few issues when investors actually implement these strategies. These include liquidity costs, institutional holdings and short sale constraints. Lesmond (2008) concludes that the bulk of the trading profits is derived from the short side of the trade, but that this position suffers from high liquidity costs that reduces institutional holdings with consequent short sale constraints. The net gains after taking into account these issues remain an open question be addressed in the future. Practical implications - The empirical results indicate that investors can do an implementable portfolio strategy of going long for a year on an initially equally weighted lowest asset growth (accrual) decile portfolio and going short for a year on an initially equally weighted highest asset growth (accrual) decile portfolio, which produces significant abnormal returns. The results further show that these abnormal returns can be improved if investors classify stocks into "the positive correlation" and "negative correlation" groups and implement trading similar trading strategies. Originality/value The empirical evidence finds that firm-year observations that exhibit a negative correlation between growth and accrual measures represents a significant 30 percent of the total firm-year observations during the sample period from July 1974 to June 2017. This highlights the necessity to undertake a detailed analysis on the issue. The authors continue to find accrual effects among these groups of firms. Therefore, the accrual effect cannot be attributed to the diminishing marginal return hypothesis. This is the main contribution of the paper.