Information-based accruals strategy.

Introduction

It has been well documented that investors frequently fail to detect the lower persistence of the accrual component of earnings, and thus tend to overreact to the information contained in the accruals (e.g., Sloan, 1996; Subramanyam, 1996; Xie, 2001). However, it is still unclear why investors fail to correctly price accruals. Sloan (1996) argues that the mis-pricing is due to some market participants' fixation on the total amount of reported earnings, without regard for the relative magnitude of the accrual and cash flow components of earnings. Teoh et al. (1998a, 1998b) suggest that managers opportunistically manage earnings before initial public offerings (IPOs) and seasoned equity offerings (SEOs), and investors may have failed to detect this. Francis, LaFond, Olsson, and Schipper (2003) present evidence showing that accounting-based trading anomalies (including accruals strategies) can be justified by rational investors' responses to information uncertainty.

In the paper, we seek to further explore the issue by offering a new explanation for the ability of the accruals to predict future stock returns. If accruals mis-pricing indeed exists, we hypothesize that it should be more conspicuous for firms where information production has been less intense and effective. We emphasize the role of stock trading and information production in accruals mis-pricing for the following three reasons. First, when a firm's stock is traded by many sophisticated (informed) traders, more information about the firm's fundamentals will be impounded upon its stock prices. A more efficient stock market might generate more information for the market participants to detect the differential persistence of the accrual and cash flow components of earnings.

Hence, it helps to reduce the errors when predicting future returns. Second, when there are a lot of institutional investors in a firm's investor base, and when its stock is traded by many informed traders, the firm manager is more likely to be disciplined in his financial reporting practice, which reduces the manager's ability to exercise discretion opportunistically and consequently increases the informativeness of the firm's accounting numbers. This again may lead to a smaller degree of accruals mis-pricing. Third, even if we believe that managers strategically choose accruals to improve the informational value of accounting numbers (Watts and Zimmerman, 1986; Healy and Palepu, 1993), active stock trading and effective information production may substitute for this kind of practice. Thus, it may also help to reduce the discretionary use of accruals and then lessen the degree of accruals mis-pricing.

This paper contributes to the literature on accruals pricing and market efficiency in several ways. First, it provides direct evidence demonstrating the role of stock trading and information production in accruals mis-pricing. Thus, it goes beyond prior literature which primarily focuses on presenting evidence showing that the market overprices accruals or creates abnormal accruals. Second, this paper identifies the bid-ask spread, analysts' forecast error, and institution shareholding as empirically feasible information variables, to measure the intensity and efficiency of information production. It thus complements Barth and Hutton (2003), which stresses the role of financial analysts as information intermediaries. More importantly, using our information variables to cut stocks, we suggest a refined accruals trading strategy that generates sizeable four-factor adjusted abnormal returns. Third, this paper also offers a test on the naive investor hypothesis. If the negative association between accruals and future stock returns is indeed related to earnings fixation by naive investors, then the magnitude of such association should be inversely related to the number of sophisticated (informed) investors. Our empirical results seem to support this hypothesis, given that we present evidence showing that firms with more institutional shareholdings suffer from less accruals mis-pricing.

Information Production and the Association between Accruals and Future Stock Returns

Development of Hypotheses

To the extent that managers manipulate earnings, and earnings fixation causes investors to ignore value-relevant information about the components of earnings and to overestimate the effect of accruals, it seems likely that this effect will be most pronounced for firms where information production about the firms' fundamentals is less intense. When a firm's stock is traded by many informed traders, more information about the firm's fundamentals would be generated in the process of trading. Meanwhile, those informed traders are more sophisticated than merely responding to firm profitability. They are able to detect earnings manipulations and are likely to exert a stabilizing force against temporary mis-pricing in the market. Thus, the accruals mis-pricing, if any, should be smaller for the firms with more intense information production.

In the presence of sophisticated investors and active stock trading, managers, concerned with the disciplining role of the informed traders, may be forced to reduce the subjectivity of accruals. Even though we believe that managers choose accruals to improve the informational value of accounting numbers, active stock trading and intense information production may also provide more information production channels that substitute for the practice of choosing accruals to convey information, and thus reduce the discretionary use of accruals in the earnings reporting process. Based on a totally different line of reasoning, we obtain the same conclusion: the degree of accruals mis-pricing should be smaller for forms with more information production.

To summarize the above discussion, we have the following hypotheses:

H1: The market value relevance of earnings and accruals are higher for firms with more active stock trading and more intense information production (measured by lower bid-ask spread, higher presence of institutional investors, and lower analyst forecast error).

H2: Earnings expectations embedded in share prices more accurately reflect the higher earnings persistence that is attributable to cash flow component of earnings, and the lower earnings persistence attributable to the accrual components for firm with more information production (again, measured by lower bidask spread, higher presence of institutional investors, and lower analyst forecast error).

Measurement of Information Production

To start, we first specify the variables used to proxy for the intensity of information production. In this paper, we propose three stock market-based variables to measure the intensity of information production. We first use a transactions level TAQ database to calculate the bid and ask spread variable--BAS. BAS is measured as the quoted bid-ask spread deflated by stock price, which is defined as the midpoint of the spread. Our second measure of the intensity and effectiveness of information production is based on the earnings estimates provided by the I/B/E/S database. For each firm year, we calculate the analysts' forecast error--FERROR, which is defined as the absolute value of the difference between mean consensus earnings estimates and the actual earnings, normalized by the actual earnings. When a firm's information production process is intense and effective, analysts following the firm's stock are more likely to agree with one another on the firm's earnings prospect. FERROR therefore is a good proxy for the intensity of a firm's information production.

Note that institutional investors are generally considered to be informed traders. Their trading and research on certain stocks surely generate a lot of information. In the paper, we define the percentage of outstanding common shares held by institutional owners at year end--IOWNER--and use it as another measure for the degree of information production.

Data and Measurement of Variables

Sample Selection. We obtain the firm-year observations from the Standard and Poors' Compustat database from 1985 to 2002. We then match the I/B/E/S database with Compustat and CRSP to get the second sample, which covers the period from 1985 to 2002. The institutional investors' information is obtained from the Thomson Financial database. Again we focus on the period from 1985 to 2002. When we match this database with Compustat and CRSP, we obtain 49,694 firm year observations.

Variable Measurement. Following Dechow et al. (1995) and Sloan (1996), the accrual components of earnings are estimated based on depreciation expenses and changes in current assets (CA) other than cash and current liabilities (CL) other than debt and income tax payable (ITP):

Accurals = ([DELTA]CA - [DELTA]Cash) - ([DELTA]CL - [DELTA]Debt - [DELTA]ITP) - Depreciation (1)

where:

* [DELTA]CA = changes in current assets (Compustat item #4)

* [DELTA]Cash = change in cash and short-term investment (Compustat item #1)

* [DELTA]CL = change in current liabilities (Compustat item #5) m #34)

* [DELTA] ITP = change in income tax payable (Compustat item #71)

* Depreciation = depreciation expenses (Compustat item #14).

The cash flow components of earnings are then estimated as:

* CFO = EARN - Accruals (2)

where EARN represents income before extraordinary items (Compustat item #18), and CFO denotes the cash flows. Consistent with the existing literature, accruals and cash flows are deflated by average total assets for meaningful cross-sectional analysis.