Mobile Edition
Print
Get the Mag
Weekly UpdatesABSTRACT
This paper investigates the anomalous relationship between negative earnings and stock prices for the high-tech sector. We obtain evidence rejecting the claim that including book value in the valuation specification eliminates the anomalous relationship. Test results indicate that sales revenues are more value relevant than reported negative earnings in the valuation of high-tech loss firms.
INTRODUCTION
The negative price-earnings relation for firms that report losses, as reported by Jan and Ou (1995) and documented by Burgstahler and Dichev (1997) and Kothari and Aimmerman (1995), raises questions about the validity of the assumption of a positive and homogeneous relation between price and earnings, as expressed by the simple earnings capitalization model. A negative coefficient on earnings means that the more negative is a firm's earnings per share, the higher is its stock prices, which makes no economic sense at all. Collins et al. (1999) hypothesize and find that including book value of equity in the simple earnings capitalization model eliminates the negative relation. They explain that book value of equity is a proxy for expected normal future earnings, which is especially important for loss firms regarding valuation. The omission of book value in the simple earnings capitalization model renders the model mis-specified, introducing a positive bias to the coefficient on earnings for profit firms and a negative bias to the coefficient on earnings for loss firms.
We investigate the anomalous negative price-earnings relation for firms that report losses in the high-tech sector during the 1990 to 1999 ten-year period, a period when high-tech industry enjoys unprecedented growth. High-tech companies do not usually report profits or positive operating cash flows. They usually have big investments in intangible assets and large R&D expenditures, which, according to SFAS #2, must be fully expensed (Lev and Sougiannis, 1996). As such, earnings are not really as important and book value can hardly be a measure of a firm's true wealth (Barron et al., 2002).
We hypothesize and find the positive coefficient on earnings for high-tech profit firms and the anomalous negative coefficient on earnings for high-tech loss firms. We hypothesize and find that including book value of equity in the simple earnings capitalization model does not eliminate the negative price-earnings relation for loss firms. We hypothesize and find evidence suggesting that sales are more value relevant than earnings or book value for high-tech loss firms (Davis, 2002, Jahnke, 2000).
Our study contributes to the current literature by providing further evidence of the anomalous relation between price and negative earnings for high-tech loss firms. We demonstrate the persistence of the anomaly and rejecting the claim that the inclusion of book value of equity in the model eliminates the anomaly, for the high-tech sector. We also provide evidence of the value relevance of sales for high-tech firms.
Section II discusses the sample selection and data. Section III presents the testing results of the negative price-earnings relation for high-tech loss firms. Section IV provides the empirical evidence of regressing stock price on earnings, book value, and sales. Section V is a summary and conclusion.
HIGH-TECH SAMPLE AND SUMMARY STATISTICS
Our sample for the high-tech companies involves the drug, computer, networking and telecommunication industries. Table 1 is a description of the industries in the sample in the three-digit SIC code.
We incorporate all firm-year observation during the 1990 to 1999 period in the 2000 Standard and Poor's COMPUSTAT CD-ROM active and research databases. The research file is included in the study to mitigate survivorship bias. All variables in this study are measured on a per share basis. Firm-year observations are eliminated of which (1) December is not the fiscal year end, (2) stock price three months after the fiscal year end is missing or negative, (3) earnings per share data is missing, (4) beginning of year book value of equity is missing, (5) sales per share is missing or negative, and (6) the total number of common shares outstanding decreases 1/3 from the previous year as it is suspect of a reverse stock split to maneuver earnings per share.
The selection process results in a total of 8,808 usable high-tech firm-year observations from 1990 to 1999. Of all the usable firm-year observations, 4,671 firm-year observations with negative earnings are grouped in the loss firm category, while 4,137 firm-year observations with positive earnings are grouped in the profit firm category. The sample selection and data treatment of this study follows Collins et al. (1999) with some adjustments. First, this paper employs only December fiscal year-end firms, to mitigate temporal effect of the stock market fluctuation on a company's stock price. Secondly, this study deletes firm-year observations with negative cum-dividend price or negative sales per share because negative price and sales do not make economic sense. Thirdly, reverse stock split distorts earnings per share number, and so deleted are any observation of which the common stock outstanding decreases by one third over the previous year.
Of the ten-year period from 1990 to 1999 used in this study, an average 53 percent of high-tech companies per year reported losses. Compared with profit firms, these loss firms have lower stock price (mean $13.349 for loss firms vs. $18.909 for profit firms), lower beginning of year book value of equity per share (mean $1.073 for loss firms vs. $4.519 for profit firms), and lower sales revenues (mean $4.816 for loss firms vs. $10.413 for profit firms). Note that the number of high-tech firms increase drastically from 447 in 1990 to 1,396 in 1999, more than three-fold. Note also that there is an almost monotonous increase in the frequency of high-tech firms reporting losses in the sample ten years. In 1990, out of 447 high-term firm-year observations, 191 or 42.7 percent have negative earnings. In 1999, out of 1,396 high-term firm-year observations, 898 or 64.3 percent have negative earnings, an increase of 50 percent over the ten-year period.
SIMPLE EARNINGS CAPITALIZATION MODEL
We use Jan and Ou's (1995) simple earnings capitalization model to study the price-earnings relation for our sample.
[P.sub.t] = [alpha] + [beta][X.sub.t] + [[epsilon].sub.t] (1)
Where [P.sub.t] is cure-dividend price of the firm's stock price three months after the end of the fiscal year t plus its dividend per share for year t. [X.sub.t] is the bottom-line earnings including discontinued operations, extraordinary items, and accounting changes, as in Collins et al. (1999). Jan and Ou (1995) report that the estimated coefficient on earnings is positive and significant for each of their 19 years for profit firms. Of their 19 years for loss firms, the estimated coefficient on earnings is negative for all 19 years, and significant as well for all but one year. Collins et al. (1999) find similar results for profit firms and report that for loss firms the estimated coefficient on earnings is negative for all, and significant as well for all but 1978 and 1979, of their 18 years over the 1975-1992 period. The negative coefficient on earnings is robust to various sensitivity checks. They also find that the estimated coefficient on earnings is positive and significant for each of their 18 years for all firms combined.
The test results of this study confirm the findings of an anomalous negative price-earnings relation for loss firms by Jan and Ou (1995) and Collins et al. (1999), as reported in Table 2. For high-tech loss firms, the estimated coefficient on earnings is negative in all 10, and both negative and significant in 9, of the 10 years over the 1990-1999 period. The mean of the estimated coefficient on earnings is -2.498 with a t-value of-9.90, significant at the 1 percent level. For profit firms, the estimated coefficient on earnings is both positive and significant for each of the 10 years from 1990 to 1999. The mean of the estimated coefficient on earnings is 9.730 with a t-value of 17.57, significant at the 1 percent level. For all firms combined, the estimated coefficient on earnings is mixed. The estimated earnings coefficient is both negative and significant for each of the first 6 years from 1990 to 1995, but positive and significant for each of the remaining 4 years from 1996 to 1999. The mixed results of the estimated coefficient on earnings is another indication that the assumption that there is a positive and homogeneous relation between price and earnings is doubtful, even within the same industry group.
EARNINGS, BOOK VALUE, AND SALES
Collins et al. (1999) claim that the simple earnings capitalization model is mis-specified because of the omission of value relevant variables. They report that including book value of equity in the valuation model eliminates the negative relation, suggesting that the anomaly is due to model misspecification. The following is their model of including book value in the regression of price against earnings and book value:
[P.sub.t] = [alpha] + [beta] [X.sub.t] + [gamma] [V.sub.t-1] + [[epsilon].sub.t] (2)
Where [P.sub.t] is cure-dividend price of the firm's stock price three months after the end of the fiscal year t plus its dividend per share for year t. [X.sub.t] is the bottom-line earnings per share including discontinued operations, extraordinary items, and accounting changes, and [V.sub.t-1] is the beginning of year book value per share at year t. Collins et al. (1999) present a detailed explanation in the appendix of how they derive the above equation based on Ohlson's (1995) abnormal earnings and end-of-period book value of equity model.
FEED