Upload
dangdan
View
216
Download
1
Embed Size (px)
Citation preview
Prediction of Future Earnings in an Emerging Market by Fundamental Analysis:
Evidence from China A-share Market
Yu Xin M.Phil. Candidate
A Thesis Submitted in Partial Fulfilment of the Requirements for the Degree of Master of Philosophy in Accountancy
• T h e Chinese University of Hong Kong August 2002
k( 1 ] B ffi )i|
Prediction of Future Earnings in an Emerging Market by Fundamental Analysis: Evidence from China A-share Market
Abstract This paper investigates two issues related to the predictive ability of
fundamental information (financial statement information) in China A-share market The first one is what kind of information in financial statements has predictive ability of future earnings. This question is tested by the association between one-year-ahead earnings changes and current fundamental items. Using a sample of all available A-share observations in Shanghai and Shenzhen Stock Exchanges from 1993 to 2000, I document evidence that three fundamental items useful to predict the future earnings in China A-share market are unexpected operating income, above industry average increase in fixed assets and selling and administrative expenses. The second research question is whether the predictive ability of the financial statement information is used by investors and reflected in the current stock price. I use return model to investigate the incremental information content of fundamental signals. Results indicate that the non-earnings fundamental signals have no incremental information content beyond earnings, indicating that the market cannot use the non-earnings accounting information to value the firms. I further examine whether fundamental analysis is affected by two factors: losses and earnings persistence. The current earnings of loss firms have less predictive ability about future earnings and less information content than firms with positive earnings, indicating that investors in China differentiate positive earnings from losses. Opposite to the expectation, fundamental signals of firms with higher proportion of operating income have less predictive information about future earnings and are less value-relevant than firms with lower proportion of operating income. Keywords: prediction of future earnings; fundamental analysis
從基本面分析看發展中市場的盈利預測:中國A股市場的證據
摘 要
本文對中國A股市場財務報表信息的預測能力進行了探討。具體研究了
兩個問題:一,財務報表中哪些信息對未來盈利有預測作用?上海与深圳證券
交易所1993年到2000年的數據表明,當期未預期營業收入,管理与銷售費用和
相對于行業水平的固定資產變化這三個指標与下一年度收入變化有顯著的相關
性。二,投資者是否使用這些具有預測作用的信息來衡量企業價值?結果表
明,基本面信息中的盈利對投資者具有價值相關性,非盈利的基本面信息不具
有價值相關性。本文進一步研究了齡損和收益持續性對基本面信息的預測能力
和價值相關性的影響。實證結果表明,与現有的關於成熟市場的研究結果相
同,齡損企業的當期盈利具有較少的價值相關性。預測能力結果也表明磨損企
業的盈利与下一年盈利變化有較弱的相關性,體現了賠損的暫時性。收益持續
性研究表明,与現有研究結果相反,營業收入所占比例小的公司其基本面信息
中含有較多的預測信息,也具有較強的價值相關性。
關鍵字:盈利預測,基本面分析
Prediction of Future Earnings in an Emerging Market by Fundamental Analysis: Evidence from China A-share Market
1. Introduction This paper investigates two issues related to the ability of financial statement
information to predict the future earnings of listed Chinese firms. The first is what kind of information in financial statements has predictive ability with regard to future earnings. The second is whether investors use the predictive ability of financial statement information in making valuation decisions, i.e., whether such information is reflected in share prices.
Financial statements are useful to investors mainly because they help predict future earnings. Why is the historical cost-based financial statement information useful? The following two perspectives offer different explanations. The information perspective assumes that the good or bad news contained in financial statements will persist into the future. Investors use current fundamental information to predict future earnings or components of earnings, including cash flow. Subsequently, the prediction of earnings is used to predict future investment returns, the investors' ultimate interest. Beaver (1989) describes a three-link process: (1) current earnings are useful for predicting future earnings, (2) future earnings are an indicator of future dividend-paying ability, and (3) expected future dividends are discounted to the present to infer equity value.
In contrast to the information perspective, the fundamental perspective proposes that the book value of equity and earnings are relevant valuation attributes, not merely signals about other attributes. Ou (1990) asserts that a two-link process establishes the relation between non-earnings data and the firm value. The "predictive information link" ties current financial data to future earnings, while the "valuation
1
link" connects predicted future earnings to firm value. Thus, the current fundamental signals are connected to firm values through future earnings. Under this perspective, fundamental analysis helps explain variations in return because of its valuation attributes.
Under both the information and the fundamental perspectives, the role of accounting information in predicting future earnings is one of the most important issues. However, fundamental analysis is rare in emerging markets. Although the predictive ability of financial statements in developed markets has been tested (e.g., Ou and Penman 1989; Lev and Thiagarajan 1993; Abarbanell and Bushee 1997), it is yet to be examined in emerging markets. Therefore, the first objective of this study is to determine whether accounting information in emerging markets have predictive ability, and if so,through which fundamental signals.
The second objective is to investigate the value relevance of fundamental signals in the emerging Chinese market, i.e., whether the market prices information that predicts future earnings. A number of studies have examined the efficiency of the Chinese market in valuing contemporary earnings by investigating the value-relevance of summary accounting information, such as annual earnings (e.g., Haw, Qi and Wu, 1999), interim earnings (e.g., Ma, 2000), and cash flows (e.g., Haw, Qi and Wu, 2000; Li, 2001). In this paper, using the predictive signals derived from the first research question, I provide further evidence on the efficiency of China's A-share market in terms of additional financial statement information.
To achieve the first objective, I search for fundamental signals that have predictive ability by investigating their association with earnings changes for the following year through both ordinary least squares (OLS) and logistic regressions. Using observations from 1993 to 2000, I examine the predictive ability of 9 variables.
2
They are three earnings components (unexpected operating income, unexpected investment income, and unexpected non-operating income) and six non-earning fundamental signals (Accounts receivable, Inventory, Gross margin, Selling and administrative expenses, Fixed assets and Long-term debt). The results indicate that unexpected operating income, selling and administrative expenses and fixed assets are negative indicators of one-year-ahead earnings change.
For the second research objective, I use the return model to investigate the value relevance of fundamental signals. I find that the investors in Chinese stock market cannot use non-earnings fundamental information to value the firms. I also test how losses and earnings persistence affect the fundamental analysis. Consistent with the conclusions from matured markets, the fundamental signals of loss firms are more value relevance. It is mainly due to the transitory nature of losses indicated by predictive tests. Opposite to the previous studies, earnings with higher proportion of operating income are less associated with future earnings and less value relevant than those with lower proportion of operating income. It may due to the fact that the majority of listed firms in China use investment income and non-operating income to boost accounting earnings so that Chinese investor perceive these items as more permanent than theoretically defined.
This paper makes three contributions to the accounting research. Firstly, this paper is the first to investigate signals that may have predictive information in the emerging capital market of China. Secondly, it extends the existing literature on the usefulness of non-earnings fundamental information in the Chinese stock market by determining whether the market can price non-earnings financial signals associated with future earnings. Finally, the findings in this study may have implications for investors by shedding light on the understanding of the quality of earnings. The
3
quality of earnings here refers to positive vs. negative earnings and earnings persistence measured by proportion of operating income.
The rest of this paper proceeds as follows. The next section discusses institutional background. Section 3 reviews related literature. The hypotheses and research design are presented in section 4,followed by sample description in section 5. The empirical results are presented in section 6. Section 7 exams the impact of two factors on fundamental analysis: positive vs. negative earnings and earnings persistence. Sensitivity tests are reported in section 8 and section 9 concludes this study.
2. Institutional Background 2.1 The lack of fundamental analysis in Chinese stock market
There may be three reasons for the lack of fundamental analysis in the Chinese stock market. The first is the shortage of financial analysts. Financial analysts do not follow A-shares of listed Chinese firms and no earnings forecasts are reported to the market. It is difficult for researchers to get predictive information. At the same time, investors are unsophisticated and functional fixation is a pervasive phenomenon in the Chinese stock market (Zhao and Wang, 1999). The market may not price earnings components with different degree of persistence accordingly.
The second reason concerns listing regulations. Until recently, Chinese government has looked upon the securities market primarily as a means to tap Chinese individual savings for the benefit of favored state-owned enterprises (SOEs) (Zaloom and Liu, 1999). Access to the market has been strictly controlled. In such circumstances, most listed firms are not managed on a profit-maximizing basis or with attention to fundamentals, but on how to obtain rights issues or avoid being
4
suspended or delisted. The regulations regarding rights issues and special treatment are based primarily on accounting numbers, especially on net profit. The China Securities Regulatory Commission (CSRC) has issued several guidelines since 1993 to define the qualification of rights issuing firms. In 1996, the CSRC required rights issuing firms to have at least 10% return on equity (ROE) in each of the last three years (CSRC 17, 1996). A revised guideline issued in early 1999 reduced the ROE from 10% to 6% for each of the last three years, but the average for the three years must still exceed 10%. Another regulation relates to share suspension and delisting. The CSRC may suspend and terminate the listing of a firm's stock if the firm has losses for three consecutive years. Manipulation is a serious problem in the Chinese stock market because of the need to meet the criteria of regulations based on accounting numbers (Chen and Yuan, 2000; Haw, Qi and Wu, 1999). It affects the usefulness and reliance of fundamental analysis.
Finally, information disclosure is limited. Most listed Chinese firms fulfill only the minimum disclosure requirements mandated by the CSRC (Haw,Qi and Wu, 2000). Earnings forecasts or voluntary information are seldom provided. The information in financial statements may not adequately reflect a firm's economic situation. Although these factors affect the usefulness of fundamental information to investors, it is unclear whether investors use such information, or indeed what kind of information they use at all.
2.2 Earnings components of listed Chinese firms According to Accounting Systems for Shareholding Company (promulgated by
Ministry of Finance, PRC, 1998), the income statement must disclose the operating
5
income, investment income and non-operating income separately. An example of a typical income statement is presented in Table 1.
Insert Table 1 here Among the three main components, the persistence of operating income is
expected, in theory, to be higher than other components. Chen and Yuan (2000) indicated that in China, earnings management is often achieved by transactions not related to core business. Haw et al. (1998) and Chen (2001) documented that ST firms rely on non-core items to increase their net earnings.
The proportion of earnings components is presented in Table 2. Insert Table 2 here
The median of operating income proportion is consistent year by year and changes around 80% during 1994 to 1999. The proportion of investment income is large in early 3 years and decreased in late years. The median of non-operating income is less than 0.4% in all the 6 years. It suggests that non-operating income does not occupy a significant proportion. However, some of investment income and non-operating income is seriously related with the firms' economic situation, such as the rights issuing (e.g., Chen and Yuan, 2000), avoiding suspension and termination of the listing (e.g., Haw et al., 1998, and Chen, 2001). Meanwhile, the sophistication level of Chinese investors remains a controversial issue. Whether they can distinguish the earnings of different quality needs to be further explored.
3. Literature Review 3.1 Studies on predictive ability of fundamental items
6
In this paper, the information set for the prediction of future earnings is limited to accounting data, divided into earnings (including earnings components) and non-earnings information.
Ou and Penman (1989) identify financial statement items that project future earnings by using a statistical search method. These items are evaluated on their ability to predict the direction of one-year-ahead earnings. No conscious attempt is made to assess predictive ability on the basis of what the authors think should work or what has been observed to work from experience. The 68 possible predictors are assessed after a survey of financial accounting and financial analysis texts. In contrast to Ou and Penman's data-driven approach, Lev and Thiagarajan (1993) identify 12 accounting-related fundamental signals appealing to economic intuition.
Insert Table 3 here Lev and Thiagarajan (1993) provides empirical evidence that analysts draw explicit inferences about the persistence of earnings by testing the value relevance of the 12 fundamental signals. These signals are calculated so that the association between each one of them and current returns is negative. For example, in the case of Inventory, an increase that outstrips sales is predicted to indicate bad news for earnings and vice versa. Disproportionate (to sales) increases in Accounts receivable convey negative signals. Analysts perceive decreases relative to industry average level in capital expenditures and R&D intensities negatively. LIFO earnings are regarded as more sustainable or closer to "economic earnings" than FIFO earnings. The use of the LIFO inventory method is considered as a positive signal, depicted by a dummy variable. Auditors' qualifications are identified by an indicator variable. A reduction in the effective tax rate is perceived to reflect less persistence and thus indicates poor future performance. A relative (to sales) decrease in order backlog (defined as the dollar
7
amount of firm unfilled orders at year-end) may suggest unrealized sales were recorded in the current period, thus a negative signal of future earnings. The empirical results indicate that the coefficients of the Gross Margin, S&A Expenses, Inventory and Order Backlog are significantly negative in most years. Abarbanell and Bushee (1997) have examined the relationship between 9 of the fundamental signals in Lev and Thiagarajan's full sample (1993) and future earnings changes. Based on the direct association between individual signals and future earnings changes, they argue that there is an economic justification for analysts and investors to rely on numerous financial statement signals to assess a firm's future performance.
Sougiannis (1994) investigates the extent to which R&D expenditures impact future earnings. Kerstein and Kim (1995) test the incremental information content of capital expenditures. Amir and Lev (1996) and Lev (1997) argue that financial information is of limited value to investors when valuing service and technology-based firms that invest heavily in intangibles. To solve this problem, they investigate information about intangible assets in financial statements. Bernard and Noel (1991) examine the ability of inventory disclosures to predict future sales and earnings. Stober (1993) extends the work of Bernard and Noel (1991) by exploring the usefulness of accounts receivable disclosures. Hodgson and Clarke (2000a) investigate the impact of leverage on earnings and return association by an arctan (inverse tangent) specification as applied by Freeman and Tse (1992).
3.2 Studies on value relevance of financial statement items A large body of research demonstrates that accounting numbers and, in
particular, earnings have information content. Yet earnings appear to explain only a small fraction of the total variation in returns. Lev (1989) summarizes the findings of
8
a number of studies employing earnings response coefficient (ERCs) models which appeared in the three major accounting research journals over the period 1980-1988, and finds that the correlations between earnings and stock return is generally very low (r2 of two to five percent). A number of previous studies have tried to offer explanations for the weak association between earnings and returns by obtaining more information to evaluate the earnings persistence. 3.2.1 Models used in value-relevance studies of financial items
Researchers in accounting often choose between return models, in which returns are regressed on scaled earnings variables, and price models, in which stock prices are regressed on earnings per share. Several papers discussed the conceptual advantage and disadvantages of price and return models. Kothari and Zimmerman (1995) argued that price models did not measure information arrival over a period. In an efficient market, the impact of information over a period is measured by stock returns. In this paper, I use the return model because my study focuses on the predictive ability and value relevance of accounting data over one year.
Ever since Ball and Brown (1968) used a simple nonparametric method and focused on the question of whether accounting earnings are associated with returns, subsequent studies have refined the methodology for both theoretical measurement and statistical techniques. An important and popular refinement is a linear statistic model that uses the unexpected earnings variables as a regressor to explain risk-adjusted returns. Lev and Thiagarajan (1993) expand the traditional return-earnings regression by including independent variables of twelve fundamental signals, which help analysts to forecast the profit of a firm by providing the whole picture of a firm's financial status and growth. Since then, various accounting variables have been added to this model, and their "incremental information content" is assessed by testing the
9
statistical and economic significance of their coefficients, e.g., Sougiannis (1994); Kerstein and Kim (1995); Amir and Lev (1996).
However, the assumption of a linear relation between stock returns and earnings has been challenged by many researchers, e.g., Cheng, Hopwood and Mckeown (1992); Freeman and Tse (1992); Das and Lev (1994). Kerstein and Kim (1995) examine the incremental information content of capital expenditures with a nonlinear regression including interaction variables between unexpected capital expenditures and growth factors. Ali (1994) investigates the value relevance of working capital and cash flow from operations by allowing for nonlinear relations between returns and each of three performance variables (earnings, working capital and cash flows). He divides the sample into groups with high or low levels of earnings, working capital and cash flows and uses dummy variables to indicate to which group the firm belongs. 3.2.2 Change versus level of earnings
Many studies indicate that earnings persistence is related to the magnitude of earnings and unexpected earnings. Brooks and Buckmaster (1976) suggest that extreme values of unexpected earnings primarily reflect transitory surprise. The common practice of assessing unexpected earnings in emerging market assumes that earnings follow a random walk, e.g., Haw, Qi and Wu (2000). The random walk model assumes that earnings innovations are permanent and will persist into the future. Easton and Harris (1991) combine the book value valuation model and earnings valuation model and conclude by using a time series sample that both the current earnings level and earnings-change variables play a role in security valuation. They further indicate that the current level of earnings is a better proxy for unexpected earnings when earnings are transitory. Ali and Zarowin (1992) also determine that for
10
firms with permanent earnings in the previous period, the incremental explanatory power of earnings level variables in return-earnings regressions is small. However, firms with transitory earnings have a much greater increase in the earnings response coefficient as a result of including the earnings level variables. These studies suggest that earnings level variables should be better independent variables than earnings change variables because of the predominantly transitory earnings of Chinese listed firms.
3.3 Other factors affecting fundamental analysis 3.3.1 Positive versus negative earnings
Hayn (1995) investigates the information content contained in losses. Evidence in her study shows that returns have a lower response to losses than to profit and that the inclusion of losses in the sample has lowered the overall slope coefficient for earnings. She concludes that the lower earnings response coefficients (ERCs) for losses are due to the existence of an implicit option held by shareholders to liquidate the firm. Alternatively, Guay et al. (1996) indicates that losses reported by some firms are due to managers' taking a ‘big bath', where the losses are temporary and are expected to reverse to profitability in future periods. Basu (1997) also attributes the lower ERCs of losses to the transitory nature of negative earnings and shareholders' expectation of earnings reversal. 3.3.2 Earnings persistence
Using a time series of past earnings relations to forecast future earnings has a long history. Kormendi and Lipe (1987) and Easton and Zmijewski (1989) characterize earnings persistence as a stationary, firm-specific phenomenon and estimated parameters of earnings persistence from time-series data on earnings.
11
Ramakrishnan and Thomas (1991) distinguish three types of earnings events: (1) permanent, and expected to persist indefinitely; (2) transitory, which affect earnings in the current year but will not affect future years; and (3) price irrelevant, with no impact at all. Most of the studies of earnings persistence take the bottom line of earnings excluding extraordinary items as variables under the assumption that the effect of transitory and price irrelevant events with regard to earnings is normally distributed and averaged to zero, e.g., Ou and Penman, 1989; Lev and Thiagarajan, 1993. Other studies examine the impact of earnings persistence by decomposing earnings into recurring versus non-recurring items. Elliott and Hanna (1996) report that investors place less value on special items than earnings before special items. Collins, Maydew and Weiss (1997) document a shift in value-relevance from earnings to book values over a 40-year period due to an increase in non-recurring items.
Chinese GAAP requires the non-recurring earnings to be reported separately from operating income. Previous studies indicate that the qualities of earnings components differ greatly, because non-core profits are a much more convenient means of earnings management for Listed Chinese firms than for those in the West; e.g., Haw, Qi, Wu and Zhang, (1999); Chen and Yuan (2000). This is because most listed firms' majority shares are held by a parent SOE. As a result it is easy for them to create profit by "asset restructuring", which takes the form of selling fixed assets or equity investment to parent firms. Listed Chinese firms manipulate profits from such related-party transactions (Chen and Yuan, 2000). Thus, it is reasonable to assume that profit from operating activities is more persistent than profit from non-core items, and that the earnings of firms with a high proportion of profit from core operating activities are more persistent and predictable than those of firms whose profit is largely derived from non-core operating activities.
12
3.4 Studies on financial statement information in the Chinese stock market Most of the studies on Chinese capital markets focus on the value-relevance of
summary accounting information; that is, annual earnings, interim earnings, and cash flows. Haw, Qi and Wu (1999) find earnings under the People's Republic of China Accounting Standards (PRC-GAAP) are value-relevant to A-share investors. They document that good news firms release their annual reports earlier than bad news firms, and loss firms release their annual reports the latest. Ma (2000) finds the interim earnings information is value-relevant to Chinese investors. Li (2001) finds that cash flow information has incremental information content on earnings. Only a few studies have investigated other information in financial reports. Chen, Su and Zhao (2000) have tested the value-relevance of an auditor's opinion. Chen, Chen and Su (2001) examine the value-relevance of earnings and book value with respect to four factors including positive vs. negative earnings, firm size, earnings persistence and liquidity of stock. Their findings indicate that investors in China differentiate positive earnings from negative earnings but do not distinguish earnings with more permanent components from those with less permanent components. Earnings and book value are more value-relevant for firms whose stocks are more liquid as measured by higher public share holdings.
4. Hypotheses Development and Research Design 4.1 Predictive ability of earnings components and fundamental items
The aim of the first part of this study is to determine what information in financial statements has predictive ability. The following hypothesis is designed to test the predictive ability of fundamental items. It is presented in the null form.
13
Hypothesis 1: There is no relationship between selected fundamental signals and future earnings change.
Fundamental items include two types: earnings components and non-earnings items. As I explained before, the earnings components are assumed to have different qualities, and thus comprise different predictive ability.
In the first stage, I select possible financial items on the basis of studies of Ou and Penman (1989) and Lev and Thiagarajan (1993). Following Lev and Thiagarajan (1993), Table 4 shows the definition and calculation of fundamental signal candidates.
The first three items are components of earnings and the rest six items are non-earnings fundamental items.
Insert Table 4 here 4.1.1 Unexpected Operating Income
Unexpected operating income is defined as: [Operating incomct - E(Operating incomet)]/ Market value of equityt
Where E(.) denotes market expected value.; Market value of equityt is the firm's market value of total equity at the beginning of year t. Similar to most of the previous Chinese studies, e.g.. Haw, Qi and Wu, 2000, a random walk model is used to proxy for the market expectation,
E(Operating incomet) = Operating incomet.i. I deflate the unexpected income by market value of equity at the beginning of the year to consist with the earnings measurement in return models in value relevance test. Unexpected operating income is measured as:
(Operating incomet - Operating incomet.i)/Market value of equityt It is expected to be negatively associated with earnings changes of the following year. Penman (1992) provides evidence to support mean reverting in the indicators of earnings persistence. Thus, I expect that the earnings increase cannot persist into the
14
next year. The following two signals of earnings components are measured similarly to operating income. 4.1.2 Unexpected Investment Income
It is unclear if increases in investment income are good news or bad news in the Chinese capital market. Previous studies indicate that Profit from Assets-Reconstruction (a kind of related-party transaction) is the main element of investment income, especially for loss-making firms seeking to extricate themselves from negative financial situations. If assets-reconstruction can improve the profitability of firms, it is good news. However, if it used to manipulate current profit, it is a bad news. 4.1.3 Unexpected Non-operating income (expenses)
The increase in non-operating income may indicate a decrease of earnings in the next year. Mangers can manipulate this part of profit with relative ease by sales of assets or equity investment. Most of the non-operating income is "paper-profit", which is not part of a firm's actual income and is mainly used to avoid sharp declines in net profit or avoid losses. Thus, increase of non-operating income is often associated with operation difficulties. Even if the non-operating income results in positive cash flow, it is transitory. If extraordinary items result in profit, it follows that in subsequent years there will be fewer such extraordinary items. From this perspective, an increase of non-operating income may indicate decreases of profits from extraordinary items and net profit in the future.
The following 6 fundamental items are non-earnings signals related to the future operating income. They are designed to be positively associated with future earnings change. 4.1.4 Accounts receivable
15
Referring to Lev and Thiagarajan (1993), I compute for each sample firm and year the following accounts receivable signal:
A Sales- A AR
Where A denotes annual percentage change. The annual percentage change in
accounts receivable is defined as: [AR-E(AR)]/E(AR)
Where E(.) denotes expected value, calculated by the random walk model. Thus, the annual percentage change in accounts receivable is measured as:
(ARt- ARt-i)/ARt.i The percentage change of sales is defined in the same way. The following 3 signals are measured similar to the accounts receivable signal. A disproportionate (to sales) decrease in accounts receivable is a positive signals. It may suggest that it is easy to sell the firm's products and receive cash. It also suggests an increasing likelihood of future earnings increase from decreases in receivables' provisions. 4.1.5 Inventory
This signal is measured as A Sales- A Inv.
Sales increases that outrun inventory are frequently considered a positive signal because they indicate an increasing demand for products. They also suggest decreases in obsolete items and less overhead cost absorption, which is good news for current and future earnings. There are many inventory-holding motives, such as smoothing production in the face of fluctuating sales, minimizing stock-out costs, and speculating or hedging against future price movements. An inventory increase might sometimes convey a positive rather than a negative signal. However, when inventories increase more than sales, it may result from an unexpected sales decrease, loss of
16
inventory control, or growth of obsolete inventory items, all of which reflect negatively on future earnings. Previous studies, e.g., Lev and Thiagarajan (1993), also indicate that inventory increases that outrun cost of sales increases are considered a negative signal. 4.1.6 Gross margin
This signal is measured as A GM-A Sales
Gross margin increases that outrun increases in sales is expected to be positive signals for future earnings. This item indicates the relation between the firm's input and output prices. This relation is driven by underlying factors, such as intensity of competition and operating leverage (the relation between fixed and variable expenses). Variations in these underlying factors, indicated by changes in gross margin, obviously affect the firm's performance and are therefore informative with respect to earnings persistence and firm values.
4.1.7 Selling and Administrative Expenses This signal is measured as
A Sales- A S&A
A disproportion (to sales) decrease of S&A is considered a positive signal, which suggests a more efficient managerial cost control. 4.1.8 Fixed assets
I use an industry benchmark for the investment in fixed assets. This signal is defined as the annual percentage change in the firm's items minus the annual percentage change in the corresponding industry.
A Firm's Fixed assets - A Industry's Fixed assets
A disproportionate increase of the firm's fixed assets to the industry average assets is a negative signal of one-year-ahead earnings change. New capital projects do not
17
usually affect earnings immediately in the following year but the related depreciation in the following year increases to reduce the firm's earnings (Abarbanell and Bushee, 1997). 4.1.9 Long-term debt
Several studies investigate whether firm financial leverage influences the earnings performance. There are debates in literature. On one hand, high leverage increases exposure to interest rate risk, which may result in more volatile and less persistent cash flows. On the other hand, as leverage increases, management may be expected to strive to maintain consistent and stable cash flows so as to reduce the probability of a liquidity crisis resulting in loan default costs. Consistent with prior studies, e.g., Hodgson and Clarke (2000),I use the industry mean to proxy for an optimal change of LT debt. Similar to fixed assets, Debt-equity is measured as:
A Firm's long-term debt - A Industry's long-term debt
Hodgson and Clarke (2000) indicates that earnings are likely to be more transitory when firms are highly levered, because of greater exposure to interest rate volatility or accruals manipulation to reduce proximity to debt covenants. They also find that greater incremental information of cash flows for above optimal leverage firms because of an increasing likelihood of earnings manipulations. At the same time, just as the investment in fixed assets, a time lag between long-term debt and earnings from new projects financed by the debt is also possible. Thus, I expect the larger increase of long-term debt than industry average level is a negative signal to one-year-ahead earnings change.
The second stage of exploring connections between financial items and future earnings comprises an assessment of the predictive ability of candidate signals. By
18
utilizing annual report data from the Chinese stock market for each year from 1993 to 2000, I analyze earnings changes between the current year and the next year. Each descriptor is included as an explanatory variable in both univariate and multivariate logistic earnings prediction models, in which the dependent variable is the direction of operating profit changes in the following year. To be consistent with Ou and Penman (1989), Lev and Thiagarajan (1993), the dependent variable in the model is the change in earnings before extraordinary items. Chinese GAAP does not clearly define extraordinary items. Previous studies (e.g., Haw, et al., 1998; Chen, 2001) indicate that Chinese managers have considerable discretion in manipulating earnings through various line items below operating income. At the same time, all the six fundamental signals selected in the first stage are all related with the core operation. Thus, I use the change of operating income as dependent variable. Ou and Penman (1989) define the earnings variables in year t+1 as epSt+i 一 epst -driftt+i to account for firm-specific trends. The drift term refers to the mean earnings-per-share change over the four years prior to year t+1. Abarbanell and Bushee (1997) use the one-year-ahead earnings change, measured as EPSt+i — EPSt as dependent variable. In this study, I use the one-year-ahead earnings change to account for firm-specific trends. The assumption is that the drift is zero. Since earnings forecasts are seldom disclosed for the A-shares of listed Chinese firms, most studies about the Chinese stock market (e.g., Haw, Qi and Wu, 1999)) use the random walk model to calculate unexpected earnings. The assumption that the drift is zero is consistent with the assumptions of the random walk model.
FECt+i = ao,t +ai’tAOIt+a2’tAIIt+a3,tAEIt + f j " F S “ t +St (1) ;=1
Where,
19
FECt+1 =1 if the change in operating income of the following year is positive; and 0 otherwise;
AOIt = Unexpected operating income in year t;
Allt =Unexpected investment income in year t;
AElt = Unexpected non-operating income in year t;
FSj,t = Non-earning fundamental signals in year t.
The earnings variable of the following year is specified as a binary outcome; that is, an earnings increase or earnings decrease. Financial statement descriptors published in any given annual report are selected on the basis of their ability to predict the direction of the annual earnings change in the following year. There is a loss of information in the binary specification. However, given outliers common to accounting data, estimation with dollar magnitude might produce parameter estimates that perform poorly in out-of-sample predictions because of estimation errors.
In order to fully utilize earnings information, I follow Abarbanell and Bushee (1997) and run regressions between the change in operating income of year t+1 and fundamental signals in year t for each year.
6
FEt+1 = ao,t +ai,tAOIt+a2,tAIlt+a3,tAEIt + ^ P j jFS . , +s t (2)
FEt+i == (Operating income of year t+1 - Operating income of year t)/ Market value of total equity at the beginning of year t; AOIt = Unexpected operating income in year t;
Allt =Unexpected investment income in year t;
AEIt = Unexpected non-operating income in year t;
FSj’t = Non-earning fundamental signals in year t.
20
Just as Ou and Penman (1989) have suggested, limiting an investigation to one-year-ahead earnings and disregarding information about earnings more than one year ahead produces a conservative bias to tests; that is, towards the null hypothesis that financial statement items have no predictive information content. However, the future performance of Listed Chinese firms is not easily predicted. This is especially true in the long term due to the lack of information disclosure, large operational transfers, and a short history of both stock market and listed firms (which have fluctuated widely as a consequence of government policy and macroeconomic factors). Hence, using one-year-ahead earnings to predict future earnings produces a bias, but it is smaller for the Chinese market than in mature markets.
4.2 Value relevance of earnings components and fundamental signals To test the market reaction to signals that are associated with future earnings, I
estimate the return models for each year. Lev and Thiagarajan (1993), and Abarbanell and Bushee (1997) both use the change of earnings in the return-earnings regression models. However, Easton and Harris (1991) demonstrate that both the level of earnings and unexpected earnings are associated with stock price. They also indicate that in situations where earnings/price ratios revert immediately to a mean, current earnings levels have more explanatory power for unexpected security returns than do earnings changes. Earnings' mean-reverting (non-persistent) is common in the Chinese stock market. In this study, I include both levels and changes of earnings in the return model.
21
I investigate the relation between fundamental signals and contemporaneous security returns to test if these fundamental signals convey value-relevant information. The null hypothesis is presented as follows: Hypothesis 2: There is no relationship between annual stock returns and fundamental signals.
To examine the incremental value-relevance over earnings of selected fundamental signals, three cross-sectional regressions are investigated. The first is a conventional return-earnings regression.
RETt= ao,t + ai,t Et + ot2,t AEt + |it (3) Where, RETt = Market-adjusted annual return in year t, cumulative monthly returns from the beginning of May of year t to the end of April of year t+1, adjusted for dividends and stock rights; Et = the annual earnings in year t, scaled by the market value of equity at the beginning of year t. AEt = the annual change of earnings in year t, scaled by the market value of equity at the beginning of year t-1.
This regression is a benchmark. The next regression includes all the non-earning fundamental signals as explanatory variables. Consistent with previous studies, I test the incremental information content of fundamental signals beyond earnings.
RETt= ao,t + ai,tEt + a2,tAEt + ^ P j .FS j , + l t (4) ;
Where,
22
RETt = Market-adjusted annual return in year t, cumulative monthly returns from the beginning of May of year t to the end of April of year t+1, adjusted for dividends and stock rights; El = the annual earnings in year t, scaled by the market value of equity at the beginning of year t.
AEt = the annual change of earnings in year t, scaled by the market value of equity at
the beginning of year t. FSj,t = One of the six fundamental signals of year t,
Model (4) is used to test the value-relevance of fundamental signals. It indicates the incremental information of signals beyond earnings.
In model (5) I include the three earnings' components instead of earnings: 6
RETt=ao,t+ai,tOIt+a2,tIIt+a3,tEIt+a4,tAOIt+a5,tAIIt+a6,tAEIt + P j j F S ” (5)
Where,
RETt = Market-adjusted annual return in year t, cumulative monthly returns from the beginning of May of year t to the end of April of year t+1, adjusted for dividends and stock rights; Olt = Operating income in year t, scaled by the market value of equity at the beginning of year t; lit =Investment income in year t, scaled by the market value of equity at the beginning of year t; Elt 二 Non-operating income in year t, scaled by the market value of equity at the beginning of year t;
AOlt = Unexpected operating income in year t,
Allt =Unexpected investment income in year t,
23
AEIt = Unexpected non-operating income in year t, FSj’t = One of the nine fundamental signals of year t.
This model tests whether decomposing the earnings can increase the explanatory power of earnings to returns. Model (4) is the benchmark against which model (5) is evaluated. Model (5) is also used to test how investors in the Chinese stock market price the earnings components. If the market is efficient enough to utilize the earnings quality for share valuation, we expect higher coefficients of Operating profit than those of other earnings components in equation (5). F-tests are used to compare the equality of coefficients of operating income and coefficients of other earnings components.
5. Sample Selection The predictive ability of financial statement descriptors is assessed using the
annual report data of A-share companies listed on the Chinese stock market from 1993 to 2000. Financial data and trading data are collected from the Chinese Stock Market and Accounting Research (CSMAR) database.
The A-share firms included in the full sample meet the following criteria. 1. All have published annual reports from 1993 to 2000. 2. All have traded continuously and thus have stock prices for the period 1993-2000.
3. None belongs to the banking industry.
Insert Table 5 here As shown in Table 5, the full sample includes 4784 firm years. Among them,
13 firm years have negative Net sales and are deleted from the sample. Financial data
24
from 1993 are used to calculate the change value of 1994. 176 firm years of 1993 are excluded from the full sample. 747 firm years are first-listed years and have no change numbers to calculate signals. 1054 firm years of 2000 are excluded from the prediction model because they lack earnings change for the next year. Three firm years have unusual proportion of operating/non-operating income and 44 firm years with no data regarding market price are also excluded. The final sample includes 2747 firm years. The sample is divided into 5 industries according to CSMA database, Manufacturing, Retail, Real estate, Utilities and Conglomerates. The industry data of Fixed Assets and LT Debt is calculated according to this 5-industry classification.
6. Empirical Results 6.1 Descriptive statistics
The year-by-year descriptive statistics of the variables in three prediction models are reported in panel A of Table 6.
Insert Table 6 here Panel B presents the descriptive statistics of variables in value-relevance tests. The one-year-ahead change of operating income has very large standard deviation. The mean is much larger than media in each year, indicating that a few extreme large outliers exist. This may cause some bias in the regression results. I'll test it further in the sensitivity tests by delete large outliers.
The correlation among the signals is reported in Table 7. Insert Table 7 here
GM is positively correlated with change of operating income, which makes sense due to the positive relation between gross margin and operating income. S&A is positively correlated with change of operating income in all the 6 years, which is opposite to the
25
intuition. However, the correlation coefficients are relatively small. It does not indicate a multicollinearity problem.
6.2 Results of predictive ability tests Table 8 presents the results of relation between future operating income
changes and current earnings components changes and fundamental signals, tested by the univariate-logistic model, multi-logistic model and regression model.
Insert Table 8 here Some of the yearly coefficients show consistent and significant signs. The significant coefficients reject hypothesis 1. For each year, if the signs of two or three models are the same, I assign this sign as the sign of that year.
The current unexpected operating income is negatively associated with the one-year-ahead change of earnings and the coefficients are significant in all the years and the pool sample, consistent with the prediction. It indicates that the mean reverting problem of operating profit is serious and common in Chinese stock market. The listed firms cannot keep the current operating profit level. Investment income is positively associated with future operating income changes. Non-operating income is negatively associated with one-year-ahead operating income change in 4 years, consistent with prediction. However, the coefficients of investment income and non-operating income are not significant. Just as predicted, the coefficients of Fixed assets are negatively significant in 4 years and in pool. Another significant signal is S&A expense, which is negatively associated with one-year-ahead change of earnings in 5 out of 6 years, opposite to the expectation. It may due to the positive correlation between S&A and current operating profit. The coefficients of LT Debt are negatively, consistent with expectation. However, the coefficients are not significant.
26
The other three fundamental signals, AR, Inv. and GM, don't show consistent signs as predicted.
Overall, three signals show consistent and significant signs as prediction. They are Unexpected operating income, S&A and Fixed assets. The coefficients of Non-operating income and LT Debt have the same sign as predicted, but are not significant.
6.3 Results of value relevance tests The value relevance tests results are presented in Table 9.
Insert Table 9 here Panel A presents the regression results of return on earnings level and earnings changes. It indicates that both change and level of earnings are value relevant in 1994 and 1997. In 1995 and 1996,only level of earnings is value-relevant. In 1998 and 1999, only change of earnings has information content. The same pattern is disclosed in Panel B. Previous studies (Easton and Harris, 1991; Ali and Zarowin, 1992) indicated that for firms with permanent earnings in the previous period, the incremental explanatory power of earnings level variable in return-earnings regressions is small while the current level of earnings is a better proxy for unexpected earnings when earnings are transitory. The explanatory power of earnings level and changes disclosed in Panel A and B suggests that the earnings permanence of Chinese listed firms has been increasing.
Panel B display the results of return on earnings and fundamental signals. Compared with the adj. R^ of Panel A, we can find that the fundamental signals have incremental explanatory power beyond the current-year earnings and earnings changes. In every year,the adj. R? of Panel B is larger than that of Panel A. The F-test
27
indicates that the combined contributes of fundamental signals and earnings to the explanation of return variability are statistically significant. The coefficients of 6 fundamental signals are not consistent, indicating that the market does not value the non-earnings fundamental signals. Although S&A and Fixed assets have predictive information, they are not value relevant.
Panel C of Table 9 presents the results of decomposing the earnings into three components. Comparison of the adj. R? in Panel C with those in Panel B indicates that decomposing earnings contributes to the explanation of excess return variance. In almost every year, the adj. R^ in Panel C is larger than that in Panel B. In most of the years, the earnings components show positive coefficients. The coefficients of Unexpected operating income are significantly positive in most of the years, although predictive ability tests indicate that Unexpected operating income is negatively associated with one-year-ahead earnings. It suggests that the market cannot distinguish the temporary and relatively permanent profit.
In summary, three signals. Unexpected operating income, S&A and Fixed assets are negatively associated with one-year-ahead earnings changes but the non-earnings fundamental signals are not value relevant.
7. Impact of Two factors: Loss & Earnings Persistence In this section, I analyze the impact of losses and earnings persistence on the
predictive ability and value relevance of fundamental items. Consistent with predictive ability tests, losses here refer to negative operating income. Earnings persistence is measured by proportion of operating income in pre-tax income. I divide
28
the sample into two groups based on each of the two factors. Dummy variables are employed to denote a firm's membership in each group.
Firstly, the sample is split between companies reporting positive and negative operating income. The dummy variable (Di) equals to 1, if the firm reports positive net earnings, and 0 if the firm reports negative net earnings. Secondly, the sample is split into high versus low permanent earnings group based on the median of proportion of operating income to pre-tax earnings. The dummy variable (D2) equals to 1 if a firm belongs to the group with higher proportion of operating income. Regression (1), (2) and (5) are modified to combine operating income and losses, operating income and earnings persistence addictively.
6
FECt+,/FEt+, = ao,t "h iiDi -+7120,x AOIt + a , . t A O I t + a 2 , t A I I t + a 3 , t A E I t + St (6a) ;=1
6
FECt+i/FE…=ao.t -^2x02 +T22D2 x AOIt +auAOIt+a2.tAIIt+a3,tAEIt + Y a P j � t F S j ’ t + St (6b)
RETt=ao,t+yiiDi +ai,tOIt- i2DixOIt +a2,tAOIt-+7i3Di x AOIt +a3,tIIt+a4,tAIIt +a5,tEIt (7a)
;=1
RETt=ao,t+Y2iD2 +a,,tOIt-Hy22D2xOIt +a2,tAOIt+Y23 D2 x AOIt +a3,tIIt+a4,tAIIt +a5,tEIi
J,FS J, + (7a) y=i
Where, FECt+i =1 if the change in operating income of the following year is positive; and 0 otherwise; FEt+i = (Operating income of year t+1 - Operating income of year t)/ Market value of total equity at the beginning of year t;
29
RETi = Market-adjusted annual return in year t, cumulative monthly returns from the beginning of May of year t to the end of April of year t+1, adjusted for dividends and stock rights; Olt = Operating income in year t, scaled by the market value of equity at the beginning of year t; lit =lnvestment income in year t, scaled by the market value of equity at the beginning of year t; Elt = Non-operating income in year t, scaled by the market value of equity at the beginning of year t; AOlt = Unexpected operating income in year t; Allt =Unexpected investment income in year t; AEIt = Unexpected non-operating income in year t; FSj,t = Non-earning fundamental signals in year t. Di = 1 for positive operating income firm, 0 for negative operating income D2 =1 for firms with proportion of operating income greater than median, 0 otherwise. 1 expect Yii,Y12 in (6a) and yii, Y12, Y13 in (7a) to be positive, indicating that fundamental signals of firms reporting positive operating income have greater predictive ability of future earnings and more information content. 丫21, In in (6b) and Y21, Y22, Y23 in (7b) are also expected to be positive, indicating that both the predictive information and value relevance of fundamental signals in high group are greater than those in low group. 7.1 Positive versus negative operating earnings
Table 10, panel A presents the results of predictive ability tests after including the dummy variable assessing the impact of losses.
Insert Table 10 here
30
As expected, yn are significantly positive from 1996 to 1999 and in full sample, indicating that the unexpected operating income of loss firms has less association with next year earnings change. It is consistent with the transitory nature of negative earnings and earnings reversal. The results of value-relevance tests are presented in Table 10, panel B. Consistent with the predictive tests, the coefficients of Y12 are significantly positive from 1996 to 1999 and in the full sample, yo are significantly positive in 1995,1996 and the full sample. It indicates that the earnings of loss firms have less information content than those of firms reporting positive earnings in year 1995 to 1999. Yii in panel A and B of Table 10 are not significant and the signs are not consistent year by year, indicating that losses do not affect the predictive ability and value relevance of fundamental signals excluding operating income. 7.2 High versus low permanent earnings
The results of impact of earnings persistence are presented in Table 11. Insert Table 11 here
Opposite to the expectation, the coefficients of 721 and 722 in panel A are negative in most of the years, indicating that firms with lower permanent earnings have more predictive information in current fundamental items than firms with higher permanent earnings. It is opposite to the expectation that the earnings persistence of firms with higher proportion of operating earnings is higher than that of firms with lower proportion. The 722 in panel B are significantly negative in most years, indicating that investors in China price the earnings with lower permanent components higher than those with less permanent earnings as measured by proportion of operating income. One possible explanation to the surprising results is that Chinese investors perceive the investment income and non-operating income as more permanent than
31
theoretically defined due to the fact that the majority of listed firms in China use these items to boost accounting earnings.
8. Sensitivity Tests In order to test the robustness of the findings, I try alternative ways to estimate
the variables used in this study. First, I substitute the change of pre-tax earnings for the change of operating income in the prediction models. The results of predictive information are weaker than those reported in Table 8 and Table 10. No fundamental non-earning signals are significantly associated with one-year-ahead pre-tax earnings changes, indicating that the changes of net earnings are largely due to changes in non-core operating income. Second, I deflate the variables by total assets instead of total market value. The results are similar to those shown in Table 8 to Table 11.
Descriptive statistics in Table 6 indicate that the one-year-ahead change of operating income is positively skewed. I delete top 5% of the one-year-ahead change of operating income. The six-year average standard deviation, mean and media are 78.31, -6.58 and -6.92, respectively. The results of both predictive tests and value relevance tests are similar with those of full sample. Firstly, in the predictive tests, the coefficients of S&A are not significant. The coefficients of Unexpected operating income and Fixed assets are significantly negative, similar to the results of full sample. Secondly, in the value relevance tests, the non-earning fundamental signals have no incremental information content beyond earnings, consistent with the results of full sample. Finally, the two factors, losses and earnings persistence, have the same impact on the samples deleted outliers as on the full sample.
32
9. Conclusions This study provides an empirical examination of whether the current
fundamental information has predictive information for Chinese listed firms and whether domestic investors in Chinese stock market perceive such information and value the stock price with it. Using a sample of all available A-share observations in Shanghai and Shenzhen Stock Exchanges from 1993 to 2000, I obtain evidence of predictive information contained by fundamental signals. The results indicate that the fundamental analysis is useful to predict the future earnings in China A-share market. ) Unexpected operating income, Fixed assets and S&A are negative signals of one-year-ahead earnings changes. However, the market does not value the stock prices with such predictive information. The significantly negative association between current unexpected operating income and one-year-ahead operating income change indicates that the mean reverting problem of operating income is serious in Chinese stock market. It may indicate a poor earnings persistence of Chinese listed firms.
I further examine whether predictive ability and value relevance change are affected by losses and earnings persistence. Consistent with previous studies, the predictive ability and the value relevance of current fundamental items of firms with profit are greater than those of loss firms. As to the earnings persistence, the empirical results provide evidence that earnings of firms with lower proportion of operating earnings is more persistent and value relevant than earnings of firms with higher proportion of operating earnings. One possible explanation is that Chinese investor perceive the investment income and non-operating income as more permanent than theoretically defined due to the fact that the majority of listed firms in China use these items to boost accounting earnings.
33
By comparing with Lev and Thiagarajan (1993) and Abarbanell and Bushee (1997), the following differences of fundamental analysis can be concluded between China and US markets. Firstly, the fundamental signals in US market have broader extent because of higher information disclosure level. Secondly, fundamental signals, such as Inventory, Gross margin, have both predictive ability and value relevance in US market. However, the investors in Chinese stock market cannot value the stocks by fundamental signals, which are related with future earnings. It is mainly due to the short history of Chinese stock market and less sophistication of Chinese investors. Finally, previous studies about matured markets indicate that more persistent earnings have more value relevance. However, this study shows opposite results in China. It may due to the fact that the majority of listed firms in China use investment income and non-operating income to boost accounting earnings so that Chinese investor perceive these items as more permanent than theoretically defined.
The main contribution of this study is to investigate the prediction of future earnings in China stock market for the first time. Although the predictive ability of fundamental signals is limited, this paper does find some useful information shedding light on the prediction of future earnings. Current change of operating income and Fixed assets are indicators of one-year-ahead earnings change. The investor may consider this information when making investment decisions.
34
References Abarbanell, J.S., and Bushee, B.J., "Fundamental analysis, future earnings and stock
prices", Journal of Accounting Research 35, No.l (Spring 1997), 1-24 Abarbanell, J.S., and Bushee, B.J., "Abnormal returns to a fundamental analysis
strategy", The Accounting Review 73,No.l (January 1998), 19-45 Ali, A., "The incremental information content of earnings, working capital from
operations and cash flows", Journal of Accounting Research 32, No. 1 (Spring 1994),61-74
Ali, A., and Zarowin, P., “ The role of earnings levels in annual earnings-returns studies", Journal of Accounting Research 30, No.2 (Autumn, 1992), 286-297
Amir, E., and Lev, B., "Value-relevance of non-financial information: the wireless communications industry", Journal of Accounting and Economics 22 (1996), 3-30
Barth, M.E., "Fair value accounting: evidence from investment securities and the market valuation of banks", The Accounting Review 69’ No.l (January 1994), 1-25
Basu, S.,"The conservatism principle and the asymmetric timeless of earnings". Journal of Accounting and Economics 24 (1997), 3-37
Bauman, M. P., "A review of fundamental analysis research in accounting". Journal of Accounting Literature 15 (1996), 1 -3 3
Bernard, V. and Noel, J., "Do inventory disclosures predict sales and earnings". Journal of Accounting, Auditing and Finance (Spring 1991) 145-182 Brown, S., Lo, K.’ and Lys, T.,"Use ofR^ in accounting research: measuring changes
in value relevance over the last four decades", Journal of Accounting and Economics 28 (1999), 83-115
Chen, C., "An empirical analysis of mandatory valuation allowance change in China: an accounting choice and valuation study", working paper, 2001
Chen, C.J., Su, X., and Zhao, R., "An emerging market's reaction to initial modified audit opinions: evidence from the Shanghai stock exchange", Contemporary Accounting Research 17, No. 3 (Fall 2000), 429-455
Chen, K.,and Yuan, H., "Earnings management and capital resources allocation: evidence from China's accounting-based regulation of rights issue", working paper, (2000)
Chen, R., "Earnings Management of ST Firms in China A-share Market", working paper, (2001)
35
Cheng, C.S., Hopwood, W. S., and McKeown, J. C.,"Non-linearity and specification problems in unexpected earnings response regression model", The Accounting Review 67, No.3 (July 1992), 579-598
Collins, D.W., Maydew, EX., and Weiss, I.S., "Changes in the value-relevance of earnings and book values over the past forty years", Journal of Accounting and Economics 24 {\991\ 39-67
Dechaw, P.M, "Accounting earnings and cash flows as measures of firm performance: the role of accounting accruals", Journal of Accounting and Economics 18 (1994), 3-42
Easton, P.D., and Harris, T.S., "Earnings as an explanatory variable for returns". Journal of Accounting Research 29, No. 1 (Spring 1991), 19-36
Elliott, J., and Hanna, J., "repeated accounting write-offs and the information content of earnings". Journal of Accounting Research 34 (1996), 135-155
Freeman, R.N. and Tse, S. Y., "A nonlinear model of security price responses to unexpected earnings", Journal of Accounting Research 30, No. 2 (1992), 185-209
Guay, W., Kothari, S. and Watts, R., "A market-based evaluation of discretionary-accrual models", Journal of Accounting Research 34, (1996), 83-105
Haw, I.M., Qi, D., and Wu, W.,"The nature of information in accruals and cash flows in an emerging capital market: the case of China", Working paper, 2000
Haw, I.M., Qi, D.,and Wu, W., "Timeliness of annual report releases and market reaction to earnings announcements in an emerging capital market: the case of China", Journal of International Financial Management and Accounting 11 (2000), 108-131 ‘
Haw, I.M., Qi, D.,and Wu, W., "Value relevance of earnings in an emerging capital market: the case of A-shares in China", Pacific Economic Review 4, No. 3 (1999), 337-347
Haw, I.M., Qi, D., Wu, W., and Zhang, W.G., "Earnings management of listed firms in response to security regulations in China's emerging capital market", working paper, (1998)
Hayn, C., "The information content of losses", Journal of Accounting and Economics 20 0995), 125-153
Hodgson, A., and Clarke, P.S., "Accounting variables and stock returns: the impact of leverage". Pacific Accounting Review 12, No. 2 (December 2000a), 37-64
Hodgson, A., and Clarke, P.S., "Earnings, cashflows and returns: Functional relations and the impact of firm size", Accounting and Finance 40 (2000b), 51-73
36
Kerstein, J., and Kim, S., "The incremental information content of capital expenditures", The Accounting Review 70,No. 3 (July 1995), 513-526
Kormendi, R.,and Lipe, R.,"Earnings innovations, earnings persistence and stock returns", Journal of Business 60, No. 3 (1987), 323-345
Kothari, S.P., and Zimmerman, J.L., "Price and return models", Journal of Accounting and Economics 20 (1995), 155-189
Lev, B., and Thiagarajan, R.,"Fundamental information analysis", Journal of Accounting Research 3J, No. 2 (Autumn 1993), 190-215
Li, X., "Value-relevance of cash flow information in Chinese capital market: a further investigation", working paper, (2001)
Maddala, G.S., "A perspective on the use of limited-dependent and qualitative variables models in accounting research", The Accounting Review 66, No. 4 (October 1991),788-807
Ministry of Finance of PRC, Accounting Systems for Shareholding Companies: Accounts and Statements, 1998
Ou, J. A., and Penman, S.H., "Financial statement analysis and the prediction of stock returns". Journal of Accounting and Economics 11 (1989), 295-329
Penman, S.H., "Financial statement information and the pricing of earnings changes,,, The Accounting Review 67, No.3 (July 1992), 563-577
Sin, S., and Watts, E., "The information content of losses: shareholder liquidation option and earnings reversals", Australian Journal of management 25, No. 3, (December 2000), 327-338
Skinner, D. J., "The investment opportunity set and accounting procedure choice". Journal of Accounting and Economics 16 (1993), 407-445
Sougiannis, T., "The accounting based valuation of Corporate R&D", The Accounting Review 69, No.l (January 1994), 44-68
Stober, T., "The incremental information content of receivables in predicting sales, earnings, and profit margins", Journal of Accounting, Auditing and Finance, (Fall 1993), 447-474
Subramanyam, K.R., "The pricing of discretionary accruals", Journal of Accounting and Economics 22 (1996), 249-281
Teets, W.R., and Wasley, C.E., "Estimating earnings response coefficients: pooled versus firm-specific models", Journal of Accounting and Economics 21 (1996), 279-295 “ Scott, W.R., Financial Accounting Theory, Prentice Hall (1998)
37
Zaloom, A., and Liu, H. C., "A legal framework for Securities", The China Business Review (May-June 1999), 26-31 Zhao, Y.L., and Wang, Z. T.’ "An empirical study of the phenomenon of securities market's 'Fundational Fixation' in China", Review of Economics 9 (1999),56-63
38
Table 1 Income Statement
(Prepared under the Ministry of Finance of PRC (1998) Accounting System for Shareholding Companies: Accounting and Statements)
I. Sales Minus: Sales discount
Net Sales Minus: Cost of goods sold
Tax for sales II.Profit from core operations
Plus: Profit from other operations'' Minus: Allowance for inventory
Selling expenses Administrative expenses Financial expenses
III.Operating income Plus: Investment income^
Subsidized income Non-operating income'
Minus: Non-operating expenses IV.Profit before tax
Minus: Income tax Minority interest
V. Net profit a: Profit from other operations includes rental income, income from selling raw materials, income from consignment, etc. b: Investment income includes short-term, long-term debt and long-term equity investment income and loss. c: Non-operating income (expenses) includes gain/loss incurred in activities irrelevant to the business operations, such as disposal of fixed assets, debt restructuring, assets revaluation, donation and fines.
39
Table 2 Descriptive Statistics of Earnings Components
Panel A: Descriptive statistics of proportions of operating income
1994 1995 1996 1997 1998 1999
N 161 269 290 505 711 811
Mean 65.80% 35.70% 32.00% 44.50% 57.59% 51.65%
Std Dev 0.44 2.58 2.32 2.04 1.51 1.73
Max 146% 320% 1367% 1729% 1207% 1081%
Q3 94% 95% 97% 95% 97% 97%
Med 75% 80% 77% 83% 84% 85%
QI 51% 49% 36% 50% 61% 60% I
Min -243.00% -30% -2050% 2521% 2218% 1978% !
1
Panel B: Median proportions of main components of earnings 1994 1995 1996 1997 1998 1999 :
1
Operating Income 75.00% 80.00% 77.00% 83.00% 84.00% 85.00% \ Investment ; Income 20.00% 15.00% 15.00% 9.80% 4.12% 3.48% Non-operating Income 0.30% 0.02% 0.10% 0.17% 0.21% 0.39%
40
Table 3 Definition of Signals in LT's Paper
Signals Measured as
1. Inventory A' Inventory - A Sales^
2. Accounts Receivable A Accounts Receivable- A Sales
3-4 Capital Expenditure, R&D A Industry capital expenditures or R&D-
A Firm capital expenditures
5. Gross Margin A Sales- A Gross margin
6. Sales and Administrative Expenses A S&A - A Sales
7. Provision for Doubtful Receivables A Gross receivables- A Doubtful receivables
8. Effective Tax PTE^ (Tt.i -Tt)
9. Order Backlog A Sales- A Order Backlog :
10. Labor Force Annual percentage change in sales-per-employee
11. UFO Earnings 0 for UFO; 1 for FIFO | I 12. Audit Qualification 1 for Qualified; 0 for Unqualified
1. A: Percentage annual change 2. Sales: Net sales 3. PTE: Changes in pretax earnings
41
Tabl
e 4
Def
initi
on a
nd M
easu
rem
ent
of E
arni
ngs
Com
pone
nts
and
Fund
amen
tal
Sig
nals
Sig
nals
—
M
easu
red
as
Pre
dict
ions
1 Unex
pected
Ope
ratin
g In
com
e (O
pera
ting i
ncome
t - Op
eratin
g inco
met-i)
/Mark
et val
ue of
equity
t -
2 Une
xpec
ted
Inve
stm
ent I
ncom
e (In
vest
men
t inc
omet
- Inv
estme
nt inc
omet.i
)/Mark
et val
ue of
equity
t +/-
3 U
nexp
ecte
d N
on-o
pera
ting
Inco
me
(Non
-ope
ratin
g in
com
et -
Non-o
perati
ng inc
omet-
i)/Mark
et val
ue of
equity
t -
4 Ac
coun
ts R
ecei
vabl
e Sa
les?
- A
AR
^ +
5 In
vent
ory
A Sa
les-
A In
v.''
+
6 G
ross
Mar
gin
AG
M-A
Sal
es
+
7 Se
lling
and
Adm
inis
trativ
e Ex
pens
es
A Sa
les
- A S
&A
+
8 Fi
xed
Asse
ts
A Fi
rm's
fixed
ass
ets
- A In
dust
ry's
fixed
ass
ets
-
9 Lo
ng-te
rm D
ebt
A Fi
rm's
long
-term
deb
t - A
Indu
stry
's lo
ng-te
rm d
ebt
-
1. A: p
ercent
age an
nual
chang
e 2.
Sales:
net s
ales
3 Ac
counts
recei
vable:
net a
ccoun
ts rec
eivabl
e; in
some f
irm ye
ar wh
en the
net a
ccoun
ts rece
ivable
was
not re
porte
d, it i
s meas
ured b
y acco
unts
receiv
able +
other
recei
vable
— all
owanc
e for
bad de
bt 4
Invent
ory: it
is me
asured
by In
ventor
y — A
llowa
nce fo
r valu
e decr
ease
The f
irst th
ree si
gnals
are ch
anges
of ear
nings
compo
nents;
the la
st 6 sig
nals a
re fun
dament
al sig
nals.
42
Tabl
e 5
Sam
ple
Des
crip
tion
Pane
l A: S
ampl
e se
lect
ion
N
Full
sam
ple
from
199
3 to
200
0 47
84
Less
: Ne
gativ
e ne
t sal
es
(13)
Fi
rm y
ear o
f 199
3^
(176
) Fi
rst y
ear l
iste
d, h
avin
g no
cha
nge
num
bers
(7
47)
Firm
yea
r of 2
000^
, hav
ing
no e
arni
ngs
chan
ge o
f nex
t yea
r (1
054)
Fi
rm y
ear w
ith u
nusu
aP o
pera
ting/
non-
oper
atin
g pr
ofit
⑶
2791
Pa
nel B
: Sam
ple
used
in s
earc
h of
fund
amen
tal s
igna
ls a
nd v
alue
-rele
vanc
e te
stin
g Le
ss:
Firm
yea
r hav
ing
no m
arke
t pric
e (4
4)
Sam
ple
used
in s
earc
h of
fund
amen
tal s
igna
ls a
nd v
alue
-rele
vanc
e te
stin
g 27
47
Firm
yea
r of 1
994
1G1
Firm
yea
r of 1
995
^69
Firm
yea
r of 1
996
^90
Firm
yea
r of 1
997
Firm
yea
r of 1
998
711
Firm
yea
r of 1
999
^~
Not
e:
1 Th
e fin
anci
al s
tate
men
t dat
a of
199
3 is
used
to c
alcu
late
the
chan
ge v
alue
of 1
994.
2.
The
dat
a of
200
0 is
used
to c
alcu
late
the
futu
re e
arni
ngs
chan
ge o
f yea
r 19
99.
3' O
pera
ting/
non-
oper
atin
g pr
ofit
is ab
ove
100
times
larg
er th
an p
re-ta
x pr
ofit
43
Table 6 Descriptive Statistics of Fundamental Signals and Future Earnings
Panel A: Variables in predictive ability tests N Mean Sta.Dev. Median Q3
1994 One-year-ahead operating income change 161 44.3 260.95 -27.64 -2.49 37.83 One-year-ahead pretax profit change 161 -10.10 21.01 -21.28 -8.38 3.42 Operating income 161 -0.93 20.06 -11.81 -0.23 9.37 Investment income 161 3.37 9.38 -0.71 1.48 7.69 Non-operating income 161 1.75 5.87 -1.14 0.56 3.32 AR 161 -1.12 6.06 -0.76 -0.14 0.27 Inv. 161 -0.92 9.52 -0.43 -0.11 0.28 GM 161 3.15 7.61 -0.01 1.62 3.76 S&A 161 0.72 1.32 -0.22 0.91 1.29 Fixed Assets 161 0.000001 1.15 -0.34 -0.13 0.21 LT Debt 161 0.000001 1.51 -0.53 -0.25 0.13 1995 One-year-ahead operating income change 269 47.3 396.14 -39.35 -2.10 48.9 One-yeai-ahead pretax profit change 269 -1.74 66.82 -14.07 1.12 13.41 Operating income 269 -9.78 43.91 -23.91 -6.78 4.89 Investment income 269 -4.00 15.62 -5.92 -0.70 1.41 Non-operating income 269 -0.40 9.17 -2.08 -0.02 1.38 AR 269 -0.45 3.34 -0.46 -0.08 0.30 Inv. 269 -0.27 1.91 -0.43 -0.12 0.15 GM 269 -0.56 0.78 -0.93 -0.69 -0.12 S&A 269 -0.05 0.57 -0.10 0 0 Fixed Assets 269 0.00 1.99 -0.36 -0.16 0.07 LT Debt 269 -0.00001 9.10 -1.53 -1.12 -0.14 1996 One-year-ahead operating income change 290 54.23 358.16 -55.26 -8.18 71.35 One-year-diead pretax profit change 290 7.40 100.61 -15.36 5.20 36.21 Operating income 290 -12.65 56.08 -24.80 -4.67 8.82 Investment income 290 5.64 32.47 -2.68 0.54 7.78 Non-operating income 290 1.07 13.21 -1.39 -0.11 1.32 AR 290 -0.52 2.56 -0.54 -0.15 0.14 Inv. 290 -0.11 0.73 -0.29 -0.02 0.17 GM 290 0.05 1.17 -0.21 -0.03 0.09 S&A 290 -0.11 0.45 -0.27 -0.10 0.10 Fixed Assets 290 0.00001 0.39 -0.19 -0.08 0.08 LT Debt 290 0.00001 2.11 -0.61 -0.33 -0.04 1997 One-year-ahead operating income change 505 37.24 274.77 -31.16 -1.35 41.70 Oiie-year-ahead pretax profit change 505 -2.98 69.91 -17.78 0.86 13.58 Operating income 505 6.27 35.18 -5.53 4.13 15.69 Investment income 505 1.12 14.69 -2.45 0.00 3.59 Non-operating income 505 -0.33 7.01 -2.00 -0.14 0.48 AR 505 -2.65 44.50 -0.41 -0.11 0.22 Inv. 505 -0.30 3.56 -0.29 -0.01 0.26 GM 505 -0.04 0.89 -0.15 -0.02 0.15 S&A 505 0.03 1.10 -0.21 -0.01 0.19 Fixed Assets 505 0 1.84 -0.33 -0.19 0.06 LT Debt 505 0.00001 10.71 -1.51 -1.13 -0.29 1998 One-year-ahead operating income change 711 19.52 148.26 -25.63 -0.64 47.29 One-year-ahead preUx profit change 711 3.35 56.91 -11.47 0.003 12.5 Operating income 711 -3.92 42.01 -14.51 -0.62 9.11 Investment income 711 -0.49 15.38 -3.10 -0.04 1-52
44
N Mean Sta.Dev. Q1 Median Q3 Non-operatiiig income 711 -0.22 12.21 -1.70 -0.12 0.65 AR 711 -0.51 2.73 -0.45 -0.10 0.20 Iiiv. 711 -0.13 10.22 -0.41 -0.06 0.20 GM 711 -0.96 26.76 -0.19 -0.03 0.12 S&A 711 0.18 0.84 -0.06 0.25 0.51 Fixed Assets 711 -0.000001 1.34 -0.40 -0.27 0.08 LT Debt 711 0.000001 14.59 -1.75 -1.48 -0.48 1999 One-year-ahead operating income change 811 20.81 134.96 -29.89 -1.89 37.34 One-year-aliead preUix profit change 811 3.43 64.51 -9.48 2.18 13.91 Operating income 811 1.08 55.18 -10.21 0.55 11.32 Investment income 811 0.94 12.52 -1.78 0.00 1.84 Noii-operating income 811 0.41 10.88 -1.15 -0.03 0.81 AR 811 -4.02 25.20 -0.79 -2.10 -0.16 Inv. 811 0.27 7.34 -0.24 0.04 0.31 GM 811 -0.44 13.65 -0.18 -0.02 0.13 S&A 811 0.14 7.50 -0.41 -0.11 0.15 Fixed Assets 811 0.000001 1.01 -0.25 -0.15 0.06 LT Debt 811 0.00001 18.00 -2.22 -1.72 -0.44
Panel B: Additional variables in value relevance tests N Mean Sta.Dev. Q1 Median Q3
1994 Market adjusted return 161 -0.02 0.29 -0.23 -0.11 0.09 Net profit 161 36.01 17.00 25.00 34.20 46.51 Change of net profit 161 3.73 225.01 -2.91 5.61 12.37 Level of operating income 161 30.89 24.22 14.18 27.36 41.61 Level of investment income 161 10.40 10.81 1.92 7.01 16.26 Level of non-operating income 161 0.94 6.43 -0.72 0.06 2.65 1995 Market adjusted return 269 0.01 0.49 -0.26 -0.09 0.14 Net profit 269 40.11 51.30 13.10 35.82 53.21 Change of net profit 269 -14.52 39.08 -25.11 -10.11 3.78 Level of operating income 269 38.91 62.31 6.68 25.62 56.20 Level of investment income 269 9.91 13.82 1.61 4.82 14.55 Level of non-operating income 269 1.12 8.48 -0.65 0.17 2.86 1996 Market adjusted return 290 -0.08 1.32 -0.88 -0.36 0.32 Net profit 290 39.20 90.08 8.00 37.11 70.48 Change of net profit 290 -6.12 63.51 -16.28 1.12 11.32 Level of operating income 290 29.15 84.25 0.09 25.56 60.11 Level of investment income 290 17.12 39.84 1.81 7.29 20.78 Level of non-operating income 290 2.33 15.11 -0.97 -0.0002 1.81 1997 Market adjusted return 505 0.01 0.43 -0.28 -0.11 0.18 Net profit 505 25.70 41.51 17.11 30.12 46.51 Change of net profit 505 5.47 45.36 -4.23 3.29 14.28 Level of operating income 505 23.61 40.41 8.38 26.01 44.58 Level of investment income 505 7.60 15.40 0.42 3.28 10.32 Level of non-operating income 505 1.22 6.79 -0.37 0.0003 1.47 1998 Market adjusted return 711 0.004 0.34 -0.20 -0.08 0.11 Net profit 711 17.32 54.01 13.06 26.71 37.11 Change of net profit 711 -3.94 47.32 -10.00 0.81 7.32 Level of operating income 711 15.91 51.06 5.23 24.00 37.78 Level of investment income 711 4.78 18.88 0.00 1.14 6.24 Level of non-operating income 711 1.19 9.31 -0.45 0.01 1.13
45
N Mean Sta.Dev. Q1 Median Q3 1999 Market adjusted return 811 -0.03 0.70 -0.45 -0.22 0.15 Net profit 811 17.31 66.88 11.91 24.46 37.11 Change of net profit 811 1.95 59.65 -8.01 0.65 8.62 Level of operating income 811 16.30 66.45 5.00 21.70 38.68 Level of investment income 811 4.12 9.51 0.00 0.92 5.38 Level of non-operating income 811 1.44 8.80 -0.29 0.05 1.32
46
Table 7 Pearson Correlation Coefficients of Signals
01 II NI AR Inv. GM S&A FA LT Debt
Operating income 1.00 -0.28*** -0.41*** -0.10 -0.02 0.14* 0.21** -0.04 -0.09 Investment income 1.00 0.00 0.05 0.31*** 0.04 0.03 -0.02 0.04 Non-operating income 1.00 0.00 0.04 -0.15* -0.31*** -0.07 -0.05 AR 1.00 -0.01 -0.01 -0.01 -0.08 0.01 Inv. 1.00 0.07 -0.07 0.03 0.01 GM 1.00 0.30*** -0.01 -0.05 S&A 1.00 0.09 -0.15* Fixed Assets 1.00 -0.04 LT Debt 1.00 1995 Operating income 1.00 -0.2** -0.04 -0.2** 0.09 0.16** 0.16* 0.02 0.04 Investment income 1.00 -0.10* 0.27*** -0.08 -0.2** -0.08 0.01 0.01 Non-operating income 1.00 0.02 0.01 -0.04 0.13* -0.02 -0.05 AR 1.00 0.08 -0.27*** 0.09 0.01 0.02 Inv. 1.00 -0.1 0.12* 0.02 0.02 GM 1.00 0.06 -0.01 -0.04 S&A 1.00 0.01 0.02 Fixed Assets 100 0.02 LT Debt 100 1996 Operating income 1.00 0.04 0.11* -0.2** 0.04 0.13* 0.35*** 0.2** 0.08 Investment income 1.00 -0.03 0.00 -0.04 0.19** -0.03 0.08 0.12 Non-operating income 1.00 0.05 0.06 0.03 0.07 0.11* 0.09 AR 1.00 0.00 -0.01 0.05 0.01 0.01 Inv. 1.00 -0.02 0.15* 0.02 -0.02 GM 1.00 -0.05 0.3*** -0.02 S&A 1.00 0.04 -0.01 Fixed Assets l.()() -().2** LT Debt 100 1997 Operating income 1.00 0.1* 0.13** 0.01 0.03 0.08* 0.34*** 0.1* -0.02 Investment income 1.00 0.02 0.00 0.03 0.04 0.12** -0.02 -0.03 Non-operating income 1.00 0.02 0.03 0.09* 0.08* -0.02 -0.02 AR 1.00 0.00 -0.01 0.01 -0.01 0.00 Inv. 1.00 -0.1* 0.17*** 0.04 0.00 GM 1.00 -0.41*** -0.1* 0.00 S&A 1.00 0.03 -0.02 Fixed Assets 1.00 -0.06 LT Debt 1.00 —
47
01 II NI AR Inv. GM S&A FA LT Debt 1998 Operating income 1.00 0.16*** 0.05 -0.01 0.00 0.18*** 0.4** 0.14*** 0.03 Investment income 1.00 -0.04 0.02 0.00 -0.01 0.08 0.08* -0.03 Non-operating income 1.00 -0.01 0.00 0.52*** -0.07* 0.04 0.00 AR 1.00 -0.01 -0.06 -0.13*** -0.05 0.00 Inv. 1.00 0.00 -0.02 -0.03 0.00 GM 1.00 -0.16*** 0.01 0.01 S&A 1.00 0.05 0.05 Fixed Assets 1.00 0.14*** LT Debt 1.00 1999 Operating income 1.00 0.12** 0.2*** 0.03 0.06* -0.03 0.12*** 0.03 0.06* Investment income 1.00 0.11** 0.00 0.03 -0.02 0.04 -0.08* 0.00 Non-operating income 1.00 0.05 0.1** -0.11** 0.12** -0.02 0.08* AR 1.00 0.01 -0.01 0.02 0.01 0.00 Inv. 1.00 -0.16** 0.28* 0.02 0.00 GM 1.00 -0.15*** -0.01 0.02 S&A 1.00 0.05 -0.01 Fixed Assets 1.00 0.02 LT Debt 1.00 ***(**)(*) p-value is less than 0.001(0.01)(0.1).
48
Table 8
Resul
ts of
Predi
ctive
Abilit
y Test
s 6
6
Logis
tic mo
del: F
EQ+i
= ao’t
+(Xi,tA
OIt+a2
,tAIIt+
a3’tAE
It + ^
"乂 ^F
S^, +
8t Re
gressi
on mo
del: F
Et+i =
oto.t +
ai,tAO
It+a2’
tAIIt+
oc3’tA
EIt +
^ 人,F
iS",, +S
t ;
=1
‘
‘
/=!
Fixe
d A
^ ^
^ ^
|ny.
G
M
S&A
Asse
ts
LT D
ebt
1994
Uni
varia
te-lo
gist
ic m
odel
-0.
018*
0.
016
^^
-0.0
28
0.03
2 ^
^ -0
.12
Mul
ti-lo
gist
ic m
odel
-0
.02*
-
0.01
6 +
0.02
1 +
-0.0
79
- -0
.041
-
0.04
7 +
0.11
3 +
0.09
4 +
-0.0
9 -
Reg
ress
ion
-0.5
12
0.21
3 0.
668
0.02
3 -3
.46
0.29
3 17
.35
-1.0
23
-3.3
8 19
9 5
Uni
varia
te-lo
gist
ic m
odel
-0.
019*
**
0.00
3 -0
.031
* -0
.017
-0
.049
-0
.235
-0
.09
-0.0
26
-0.0
41
Mul
ti-lo
gist
ic m
odel
-0
.022
***
- -0
.01
+ -0
.051
* -
-0.0
87
- -0
.032
-
0.32
6*
- 0.
426
- -0
.015
-
-0.0
45
-R
egre
ssio
n -1
.027
* 1.
828
0.69
2 -2
.12
10.1
3 -1
2.29
-3
.66
-3.1
6*
-0.2
9 19
96
Uni
varia
te-lo
gist
ic m
odel
-0.
015*
**
-0.0
04
-0.0
15
-0.0
07
0.09
5 -0
.07
-0.9
88**
-1
.105
**
-0.1
21
Mul
ti-lo
gist
ic m
odel
-0
.013
***
- -0
.004
-
0.01
3 -
-0.0
52
- 0.
245
+ 0.
067
+ -0
.604
-
-0.6
75*
- -0
.058
-
Reg
ress
ion
-1.0
46*
0.11
9 -1
.507
-7
.69
63.2
4*
33.2
* 37
.63
-15.
1**
-4.5
9 19
97
Uni
varia
te-lo
gist
ic m
odel
-0.
016*
**
0.00
2 -0
.008
-0
.009
-0
.016
0.
025
-0.2
81*
-0.1
96*
0.00
3 M
ulti-
logi
stic
mod
el
-0.0
15**
* -
0.00
5 +
-0.0
01
- -0
.012
-
-0.0
11
- 0.
069
+ -0
.043
-
-0.1
57*
- 0.
004
+ R
egre
ssio
n 0.
222
-0.7
25
-0.3
07
-0.3
79
-4.1
65
-4.8
9 -2
2.37
-5
.99
-0.3
25
1998
U
niva
riate
-logi
stic
mod
el -
0.02
***
-0.0
04
-0.0
01
0.02
1 0.
002
-0.0
08
-0.5
34**
* -0
.034
-0
.028
M
ulti-
logi
stic
mod
el
-0.0
18**
* -
0.00
2 +
0.00
3 -
0.03
3 +
0.00
1 +
-0.0
01
- -0
.239
* -
-0.0
51
- -0
.026
-
Reg
ress
ion
-0.8
27**
* 0.
902*
-0
.674
1.
594
0.00
7 0.
028
-8.7
67
2.22
6 -0
.141
19
99
Uni
varia
te-lo
gist
ic m
odel
-0.
008*
**
0.00
7 0.
016*
-0
.002
-0
.016
0.
003
-0.0
72*
-0.4
**
0.01
M
ulti-
logi
stic
mod
el
-0.0
1***
-
0.00
8 +
0.02
5**
+ -0
.002
-
-0.0
01
- -0
.009
-
-0.0
26
- -0
.37*
-
0.01
3 +
Reg
ress
ion
-0.7
08**
* -0
.404
0.
891*
-0
.002
-1
.65
-1.0
6 -0
.364
-3
.302
0.
072
Pool
U
niva
riate
-logi
stic
mod
el -
0.01
4***
-0
.001
-0
.000
3 -0
.003
0.
001
-0.0
03
-0.1
47**
* -0
.116
**
-0.0
003
Mul
ti-lo
gist
ic m
odel
-0
.014
***
- 0.
002
+ 0.
005
- -0
.003
-
0.00
1 +
-0.0
01
- -0
.013
-
-0.0
73*
- 0.
001
-R
egre
ssio
n -0
.743
***
0.24
3 -0
.167
-0
.209
0.
002
-0.0
68
-0.0
34
-3.4
7 -0
.057
49
FECt+1
=1 if t
he ear
nings
change
of the
follow
ing ye
ar is p
ositive
; and 0
other
wise;
FEt+i
= (Ear
ning o
f year t
+1 -
Earnin
gs of y
ear t)/
Earni
ngs of
year t
. AO
It = U
nexpec
ted op
eratin
g inco
me in
year
t; All
t =Un
expect
ed inv
estme
nt inc
ome i
n year
t; AE
It = U
nexpec
ted no
n-oper
ating
incom
e in y
ear t;
FSj,t =
Non
-earni
ngs f
undam
ental
signal
s in ye
ar t;
***(**
)(*) Fo
r logis
tic mo
dels, C
hi-squ
are is
signif
icant
at 0.00
1(0.01
)(0.1)
level
For re
gressio
n mode
l, t-te
st is s
ignific
ant at
0.001(
0.01)(
0.1) le
vel
50
Table 9 Results of Value-relevance Tests
Panel A: Return on earnings RETt= ao,t + ai,t Et + a2,t AEt + Ht
N Intercept E ^ Adj- R^ F-value 1994 163 0.118 -0.005 0.009 0.1213 11.84
(2.07)* (-3.1)** (4.83)***
1995 269 -0.123 0.003 -0.0007 0.0774 12 (-2.29)* (3.72)*** (-0.68)
1996 291 -0.35 0.008 0.0014 0.3785 74.4 (-3.14)*** (5.44)*** (0.67)
1997 505 0.019 -0.001 0.0051 0.1299 33.61 (0.79) (-1.92)* (7.01)***
1998 711 0.024 -0.0006 0.0021 0.0444 15.07 (1.4) (-1.16) (3.91 广
1999 813 -0.023 0.0004 0.002 0.0422 15.64 (-0.82) (0.66) (3.48)***
Pool 2747 -0.067 0.002 0.002 0.1123 174.75 (-4.89)*** (8.05)*** (6.49)***
Panel B: Return on earnings and fundamental signals 6
RETt= ao.t + ai.tEt + a2.tAEt + + ^ ^ Fixed
N Intercept E A E AR Inv. GM S&A Assets LT Debt AcH-R^ F-value 1994 163 0.003 -0.003 0.007 0.001 0.001 -0.006 0.088 0.008 0.0002 0.1852 5.46
(0.05) (-1.54) (3.61)*** (0.4) (0.6) (-1.97)* (4.11)*** (0.45) (0.43) 1995 269 -0.032 0.002 0.0002 0.014 0.015 0.079 -0.185 -0.007 -0.0001 0.1139 5.21
(-0.5) (2.72)** (0.22) (1.21) (1.04) (1.98)* (-3.37)*** (-0.51) (-0.15) 1996 291 -0.365 0.009 0.001 0.019 0.073 0.072 -0.52 -0.009 -0.000001 0.4 21.12
(-3.4)*** (5.72)*** (0.42) (0.75) (0.72) (1.19) (-2.92)** (-0.05) (-1.98)* 1997 505 0.0366 -0.002 0.004 0.0004 0.001 0.085 0.121 0.013 -0.0000001 0.191 13.9
(1.48) (-2.4)* (5.32)*** (1.1) (0.27) (3.61)*** (5.84)*** (1.39) (-0.48) 1998 711 0.031 -0.0004 0.002 0.002 0.0001 -0.001 -0.015 -0.003 -0.0000001 0.0509 5.06
(1.66) (-0.82) (4.14)*** (0.49) (0.67) (-2.72)** (-0.84) (-0.28) (-1.25) 1999 813 0.018 0.0001 0.002 -0.001 -0.009 0.006 0.024 -0.018 -0.0000001 0.0589 6.2
(0.58) (0.1) (2.72)** (-0.58) (-0.45) (0.81) (1.16) (-0.73) (-3.81)*** Pool 2747 -0.066 0.002 0.002 0.001 0.0001 -0.003 -0.003 -0.001 -0.0004 0.1145 45.38
(-4.9)*** (8.25)*** (6.66)*** (1.21) (0.35) (-3.17)** (-0.94) (-0.14) (-0.48) .
51
Panel
C: R
eturn
on Ea
rning
s Com
ponen
ts and
Non
-earni
ngs S
ignals
6
RETt
=ao,t
+oci,
tOIt+
a2,
tIIt+o
c3,
tEIt+
a4,tA
OIt+o
c5,
tAIIt
+a6,
tAEI
t +
2]
厂+
M-
M Fi
xed
N
Inte
rcep
t 01
II
EI
AP
I A
ll A
EI
AR
Inv.
G
M
S&A
Asse
ts L
T De
bt
Adj-
R F
-val
ue
1994
163
0.
032
-0.0
02
-0.0
04
-0.0
11
0.00
6 0.
004
0.01
7 0.
002
0.00
1 -0
.006
0.
086
0.01
7 0.
0002
0.
1846
3.
96
(0.4
2)
(-1.
68)*
(-
1.63
) (-
2.01
)* (
3.36
)**
(1.5
) (2
.65)
** (
0.57
) (0
.57)
(-
1.94
)*
(3.8
)***
(0
.9)
(0.6
4)
1995
269
-0
.039
0.
002
0.00
3 -0
.002
0.
001
0.00
05
-0.0
06
0.01
4 0.
013
0.07
-0
.183
-0
.008
-0
.000
1 0.
1368
4.
46
(-0.
58)
(2.6
3)**
(1
.25)
(-
0.53
) (0
.73)
(0
.22)
(-
1.78
)* (
1.25
) (0
.86)
(1
.74)
(-
3.35
)***
(-0.
61)
(-0.
15)
1996
291
-0
.626
0.
008
0.01
9 0.
009
-0.0
02
-0.0
04
0.02
3 -0
.006
0.
052
0.06
-0
.434
-0
.083
-0
.000
001
0.56
95
27.5
7 (-6.6)***
(7.43)*** (5.25)*** (1.38)
(-1.1)
(-0.95)
(3.04)** (-0.25) (0.6)
(1.17)
(-2.77广(-0.52) (-2.6)**
1997
505
0.
041
-0.0
02
-0.0
04
0.00
6 0.
004
0.00
6 -0
.003
0.
0004
0.0
02
0.07
9 0.
115
0.01
1 -0
.000
0001
0.2
065
10.4
7 (1
.5)
(-2.
56)*
(-
2.42
)*
(1.5
8)
(5.3
2)**
* (3
.45)
***
(-0.
89)
(1.0
4)
(0.4
4)
(3.3
8)**
* (5
.5)*
**
(1.1
8)
(-0.
22)
1998
711
0.
042
-0.0
01
0.00
05
0.00
3 0.
003
0.00
1 0.
004
0.00
2 0.
0001
-0.
002
-0.0
16
-0.0
05
-0.0
0000
01 0
.072
4 4.
94
(2.0
8)*
(-2.
21)*
(0
.61)
(1
.22)
(4
.48)
***
(1.0
9)
(2.3
5)*
(0.3
3)
(0.6
6)
(-3.
5)**
* (-
0.9)
(-0
.49)
(-1
.4)
1999
813
-0
.012
-0
.000
1 0.
01
-0.0
04
0.00
2 -0
.002
0.
009
-0.0
01
-0.0
06
0.00
5 0.
019
-0.0
27
-0.0
0000
1 0.
072
5.3
(-0.34)
(-0.19)
(2.87广
(-0.66)
(2.32)*
(-0.74)
(1.7)*
(-0.49) (-0.31) (0.67)
(0.9)
(-1.09) (-3.91)***
Pool
27
47 -
0 11
0
002
0.00
6 0.
007
0.00
1 0.
003
0.00
5 0.
001
0.00
01 -
0.00
4 -0
.005
-0
.003
-0
.000
5 0.
1611
44
.94
(-7.8
9) (7
.45)*
** (
7.61)
***
(4.02
)***
(4.34
)***
(3.4
8)**
* (3
.01)
** (
1.04
) (0
.27)
(-4.64
)***
(-1.
66)*
(-0.3
4) (-0
.56)
RETt
= Mark
et-ad
justed
annu
al ret
urn in
year
t, cum
ulativ
e mon
thly r
eturns
from
the b
eginn
ing of
May
of yea
r t to
the en
d of A
pril o
f year
t+1,
adjust
ed for
divid
ends a
nd sto
ck rig
hts;
Et 二
the a
nnual
earni
ngs in
year
t, scal
ed by
the m
arket
value
of equ
ity at
the b
eginni
ng of
year t.
AE
t = t
he ann
ual ch
ange o
f earn
ings in
year
t, scal
ed by
the m
arket
value
of equ
ity at
the b
eginni
ng of
year t-
1. Olt
= Op
eratin
g inco
me in
year
t, scal
ed by
the m
arket
value
of equ
ity at
the b
eginni
ng of
year t;
lit
=Inves
tment
incom
e in y
ear t,
scaled
by th
e mark
et val
ue of
equity
at th
e begi
nning
of yea
r t;
Elt =
Non-o
perat
ing in
come i
n year
t, sca
led by
the m
arket
value
of equ
ity at
the b
eginni
ng of
year t;
AO
It 二 U
nexpec
ted op
eratin
g inco
me in
year
t; All
t ^Un
expe
cted i
nvest
ment
incom
e in y
ear t;
AEIt =
Unex
pected
non-o
perati
ng inc
ome i
n year
t; FS
j,t =
Non
-earni
ngs f
undam
ental
signal
s of y
ear t;
***(**)
(*) T-t
est is
signif
icant
at 0.0
01(0.
01)(0
.1) le
vel
Table
10
Resul
ts of
Impa
ct of
Losse
s on F
unda
menta
l Ana
lysis
6
Panel A
: FEC
t+i/FE
t+i = a
�’t 4711D丨
+yi2D
,x AO
It +(Xi’
tAOIt+a
2,tAII
t+a3,tAE
It 十]
+ £
t ‘
‘ Yi
i Y1
2 Fi
xed
(Di)
(Dix
AO
I) AP
I A
ll AE
I AR
In
v.
GM
S&
A As
sets
LT
Deb
t 19
94
Mul
ti-lo
gist
ic m
odel
0.4
50
-0.0
29
-0.0
18
0.01
7 0.
024
-0.0
84*
-0.0
42
0.05
0 0.
115
0.10
5 -0
.010
Reg
ress
ion
-47.
08
-2.3
19
1.41
5 -0
.127
1.
032
-0.2
60
-3.4
76
0.54
4 16
.98
-1.7
37
-1.9
76
1995
M
ulti-
logi
stic
mod
el -
0.38
6 -0
.020
-0
.014
* -0
.013
-0
.051
* -0
.095
-0
.028
-0
.321
0.
543
-0.0
18
-0.0
46
Reg
ress
ion
-22.
15
-1.1
45
-0.3
39
1.60
5 0.
628
-2.6
19
10.7
4 -8
.701
-3
.026
-3
.418
-0
.336
19
96
Mul
tNog
istic
mod
el -
0.10
6 0.
028*
-0
.006
-0
.004
-0
.014
-0
.093
0.
279
0.10
9 -0
.689
* -0
.680
* -0
.056
R
egre
ssio
n 10
.57*
-0
.188
-0
.109
* 0.
025
-0.1
45
-1.4
24
6.08
8 3.
318*
2.
973
-14.
94**
-0
.451
19
97
Mul
tNog
istic
mod
el -
0.21
6 0.
038*
* -0
.041
* 0.
005
0.00
1 -0
.012
-0
.016
0.
130
-0.0
06
-0.1
39
0.00
2 R
egre
ssio
n 14
.14
2.54
3**
-1.5
54*
-0.7
05
0.03
5 -0
.406
-4
.569
1.
294
-19.
92
-5.9
66
-0.4
95
1998
M
ulti-
logi
stic
mod
el -
0.24
4 0.
033*
**
-0.0
36**
* 0.
004
0.00
2 0.
03
0.00
1 0.
003
-0.1
55
0.05
6 -0
.023
R
egre
ssio
n 20
.60
0.77
4*
-1.2
62**
* 0.
981*
-0
.478
1.
504
-0.0
001
0.08
2 -6
.243
1.
012
-0.1
55
1999
M
ulti-
logi
stic
mod
el -
0.55
7**
0.02
6***
-0
.024
***
0.00
4 0.
023*
-0
.002
-0
.045
-0
.002
0.
032
-0.2
90*
0.00
9 R
egre
ssio
n -2
2.72
* 1.
933*
**
-1.4
37**
* -0
.705
* 0.
645
0.08
8 -7
.672
* -1
.664
4.
205
-1.7
24
-0.1
10
Pool
M
ulti-
logi
stic
mod
el -
0.43
5***
0.
017*
**
-0.0
18**
* 0.
001
0.00
3 -0
.003
0.
001
0.00
1 -0
.006
-0
.064
* 0.
001
Reg
ress
ion
5.88
77
1.11
9***
-1
.238
***
0.23
9 -0
.114
-0
.224
-0
.006
0.
033
0.10
0 -3
.629
__-0
.113
_ ***
(**)(*
) For
logist
ic mo
dels,
Chi-s
quare
is si
gnifi
cant a
t 0.00
1 (0.0
1)(0.1
) leve
l Fo
r reg
ressio
n mod
el, t-
test is
sign
ifican
t at 0
.001(0
.01)(0
.1) le
vel
53
6
Panel B
: RET
t=ao,t+
7iiDi +
ai,tOIt
-Hyi2D
ixOIt +
a2.tAO
It+Yi3
Dj x A
OIt +a
3,tIIt+a
4.tAIIt
+a5.tEI
t +a6,t
AEIt
+ M-
^
Inte
r- Yn
丫
12
Yi3
Fixe
d Ad
j.-
F-cep
t (Di
) 01
(Di*O
I) AP
I (Di
*AOI)
n AH
EI
AEI
AR
Inv.
GM
S&A
Asset
s LT
Debt R
val
ue 199
4 -0.
042
0.150
-0.00
1 -0.
002
0.004
-0.00
2 -0.
005
0.005
-0.01
3 0.0
18 0.0
02 0.0
01 -0.
006
0.071
0.014
0.038
0.2471
4.50
(-0.50) (1.64) (-0.35)
(-0.63)
(1.80)*
(-0.61)
(-1.77)* (1.69)* (-2.41)* (3.17广(0.61) (0.59) (-2.02)* (3.20广(0.75)
(2.75)**
1995
-0.10
6 0.0
20 0.0
01 -0.
002
-0.00
1 0.0
10 0.0
02 0.0
01 0.0
01 -0.
005
0.006
0.007
0.037
-0.18
8 -0.
006
-0.00
3 0.2
469
6.86
(-1.51)
(0.24)
(1.41)
(-1.24)
(-1.19)
(4.82)***
(0.98) (0.75)
(0.25)
(-1.74)* (0.65)
(0.52) (1.02)
(-3.87)***(-0.43) (-1.04)
1996
-0.85
7 0.0
26 0.0
05 0.0
04 -0.
005
0.013
0.016
-0.00
1 0.0
14 0.0
18 0.0
49 -0.
017
0.027
-0.06
3 0.0
11 0.0
14 0.6
441 3
5.87
(-9.31)***(0.20) (5.01)*** (2.35)*
(-3.03)**(3.54)*** (2.81)** (5.54)***(0.33)
(2.81 广
(3.12)** (-0.26) (0.62)
(-0.55)
(0.09)
(0.61)
1997
-0.05
6 -0.
026
-0.00
4 0.0
06 0.0
03 0.0
01 -0.
002
0.007
0.007
-0.00
3 0.0
003
0.000
6 0.0
76 0.1
16 0.0
04 -0.
0003
0.2657
13.1
6 (-1.15)
(0.44)
(-2.97)**(3.86)***(1.47)
(0.31)
(-1.17)
(4.54)*** (1.81)*
(-0.68)
(0.70)
(0.12) (3.52)*** (5.97)*** (0.39) (-0.20)
1998
-0.02
1 -0.
034
-0.00
2 0.0
04 0.0
02 0.0
01 0.0
004
0.002
0.005
0.004
0.002
0.001
-0.00
3 0.0
02 -0.
011
0.0004
0.13
02 8.0
9 (-0.72)
(-0.90)
(-2.58)**(3.82)***(2.06)*
(0.59)
(0.55)
(1.79)*
(2.16)* (2.89广
(0.38)
(0.48)
(-3.94)***(0.10)
(-1.18)
(0.50)
1999
-0.18
7 0.1
52 -0.
001
0.005
0.003
-0.
002
0.012
-0.00
1 0.0
02 0.0
05 0.0
01 -0.
004
-0.00
1 0.0
03 -0.
030
0.001
0.1057
7.38
(-3.32广*(2.08)* (-1.13) (2.86)** (1.59)
(-0.83)
(0.44)
(3.96)***
(-0.20
) (0.4
4) -1.
51 (-0
.22)
(-0.17)
(0.16)
(-1.29)
(0.72)
Pool
-0.24
2 0.0
54 0.0
01 0.0
03 -0.
001
0.004
0.006
0.004
0.008
0.006
0.000
3 0
-0.00
4 -0.
005
-0.01
0 0.0
003
0.2229
53.5
3 (-1
0.2)**
*(1.73
)* (2.7
6)** (
6.00)*
**(1.5
8) (5.1
1)** (
4.63)*
**(8.0
4)***(
4.16)*
**(4.6
3)***(
3.97)*
* (-0-
02) (-
4.49)*
**(-1.
55)
(-1.20
) (0.
44)
FEQ+
1 =1 i
f the c
hange
in op
eratin
g inco
me of
the f
ollow
ing ye
ar is p
ositiv
e; and
0 oth
erwise
; FE
t+i = (
Opera
ting i
ncome
ofve
ar t+1
- Op
eratin
g inco
me of
year
t)/ Ma
rket v
alue o
f total
equit
y at th
e begi
nning
of ye
ar t;
RETt
= Mark
et-adj
usted
annual
retur
n in y
ear t,
cumula
tive m
onthl
y retu
rns fro
m the
begin
ning o
f May
of year
t to t
he end
of Ap
ril of y
ear t+
1, adj
usted
for di
vidend
s and
stock
rights
; Olt
= Op
eratin
g inco
me in
year
t, scal
ed by
the m
arket
value
of equ
ity at
the b
eginn
ing of
year
t; lit
=Inves
tment
incom
e in y
oar t,
scaled
by th
e mark
et val
ue of
equity
at th
e begi
nning
of ye
ar t;
Elt =
Non-o
perati
ng inc
ome i
n year
t, sca
led by
the m
arket
value
of equ
ity at
the b
eginn
ing of
year
t; AO
It 二 U
nexpec
ted op
eratin
g inco
me in
year
t; All
t =Un
expect
ed inv
estme
nt inc
ome i
n year
t; AE
It = U
nexpec
ted no
n-oper
ating
incom
e in y
ear t;
FSj,t =
Non
-earni
ng fu
ndam
ental
signal
s in y
ear t.
Di = 1
for p
ositiv
e oper
ating
incom
e firm
, 0 fo
r nega
tive o
perati
ng inc
ome
***(**
)(*) T-
test is
sign
ificant
at 0.0
01(0.
01)(0
.1) le
vel
Table
11
Resul
ts of
Impa
ct of
Earni
ngs P
ersist
ence o
n Fun
dame
ntal A
nalys
is 6
Panel A
: FEC
t+i/FE
t+,= Oo
.t "T2iD
2 +722
D2 x
AOIt +
ai.tAO
It+a2.
tAIIt+
a3,tAE
It +2]
瓜"
+ St
^ 72
1 111
Fi
xed
(D2)
(D
2X
AO
I) A
PI
All
AEI
AR
Inv.
G
M
S&A
Asse
ts
LT D
ebt
1994
M
ulti-
logi
stic
mod
el -
1.47
6***
-0
.013
-0
.010
-0
.011
-0
.013
-0
.094
* -0
.032
0.
035
0.13
9 0.
034
-0.1
93
Reg
ress
ion
7.83
8 -2
.208
0.
776
0.49
4 1.
416
-0.1
53
-3.5
44
0.35
1 13
.63
-2.5
46
-3.6
32
1995
M
ulti-
logi
stic
mod
el -
0.86
6**
-0.0
11
-0.0
30**
-0
.012
-0
.052
**
-0.0
91
-0.0
21
-0.4
05*
0.48
0 -0
.005
-0
.052
Reg
ress
ion
-78.
01
-0.1
76
-0.7
91
1.69
4 0.
438
-1.8
81
12.0
3 -1
3.90
-1
.816
-2
.973
-0
.244
19
96
Mul
ti-lo
gist
ic m
odel
-0.
875*
* -0
.013
* -0
.004
-0
.007
-0
.019
-0
.079
0.
284
0.03
0 -0
.645
* -0
.687
* -0
.050
Reg
ress
ion
3.80
3 -1
.269
-0
.029
-0
.101
-0
.199
-0
.898
6.
255*
2.
938
3.62
3 -1
5.36
**
-0.3
67
199 7
M
ulti-
logi
stic
mod
el -
0.85
1***
-0
.022
**
-0.0
05*
-0.0
03
-0.0
14
-0.0
14
-0.0
24
0.13
0 0.
088
-0.1
24
0.00
6 R
egre
ssio
n 2.
105
-0.7
50
0.61
1 -0
.832
-0
.701
-0
.375
-4
.219
-5
.517
-2
1.63
-5
.488
-0
.376
19
98
Mul
ti-lo
gist
ic m
odel
-0.6
81**
* -0
.022
***
-0.0
07
-0.0
02
-0.0
01
0.02
7 0.
001
-0.0
03
-0.2
40*
0.05
4 -0
.023
R
egre
ssio
n -1
9.89
* 0.
731*
-0
.262
0.
673*
-0
.576
1.
404
-0.0
03
-0.1
44
-9.7
52
2.47
9 -0
.102
19
99
Mul
ti-lo
gist
ic m
odel
-0.
422*
* -0
.019
***
0.00
1 0.
002
0.01
9*
-0.0
02
-0.0
06
-0.0
07
-0.0
19
-0.3
30*
0.00
7 R
egre
ssio
n -4
.996
-0
.618
**
-0.1
68
-0.6
01
0.57
9 0.
102
-2.3
46
-1.3
14
-0.3
95
-3.6
63
-0.0
37
Pool
M
ultH
ogis
tic m
odel
-0.7
38—
-0
.017
***
-0.0
04*
-0.0
03
-0.0
01
-0.0
03
0.00
1 -0
.002
-0
.021
-0
.070
* 0.
002
Reg
ress
ion
-15.
31*
-0.7
08**
-0
.189
0.
072
-0.3
63
-0.2
12
-0.0
01
-0.1
67
-0.4
08
-3.3
32__
-0.1
17
***(**
)(*) Fo
r logis
tic mo
dels, C
hi-squ
are is
signif
icant
at 0.00
1(0.01
)(0.1)
level
For re
gressio
n mode
l, t-te
st is s
ignific
ant at
0.001(
0.01)(
0.1) le
vel
55
6
Panel
B: RE
Tt=ao,
t+T2iD
2 +a,.t
OIt+y2
2D2xO
It +a2,
tAOIrf72
3 D2 x
AOIt +
a3,tIIt
+a4.tA
IIt +cx
5,tfilt +
a6,tAE
It + 工
fijj^S
- , + \
i j=i
‘ ‘
Inte
r- Y2
1 Ji
i Y2
3 Fi
xed
Adj.-
F-
cept
(D2)
01 (D2
*0I) A
PI (D2
*AOI)
n All
EI
AEI
AR
Inv.
GM
S&A
Asset
s LT
Debt R
val
ue 199
4 -0.
082
0.145
0.002
-0.
005
0.003
0.0
04 -0.
003
0.004
-0.01
0.015
0.0
02 0.0
003
-0.00
5 0.0
86 0.0
17 0.0
37
0.238
8 4.3
5 (-0
.79)
(1.36)
(0.
63)
(-1.69
)* (1.
22)
(1.19
) (-1
.09)
(1.45
) (1.
86)*
(2.45)
* (0.
55)
(0.14
) (-1
.58)
(3.89)*
** (0.
90)
(2.72)
** 19
95 -0.
169
0.226
0.004
-0.00
4 -0.
001
0.004
0.004
-0.00
1 -0.
001
-0.00
5 -0.
002
0.015
0.074
-0.18
5 -0.
005
-0.00
3 0.1
794
4.91
(-2.01
)* (2.
43)*
(2.63)
** (-2
.00)*
(-0.54
) (1.
70)*
(1.71
) (-0
.54)
(-0.34
) (-1
.48)
(-0.27
) (1.
00)
(1.96)
* (-3
.68)**
*(-0.4
0) (-1
.17)
1996
-0.69
6 0.0
12 0.0
08
0.001
-0.00
1 -0.
002
0.017
-0.
002
0.014
0.018
0.0
04 0.0
08
0.041
-0.16
2 0.0
55
0.017
0.5
528
24.82
(-6.39)磁
(0.08) (4.27)*** (0.28)
(-0.23)
(-0.71)
(2.50)*
(5.18)***(-0.54) (2.50)*
(0.17)
(0.11)
(0.83)
(-1.29)
(0.38)
(0.68)
1997
-0.01
5 0.1
12 -0.
001
-0.00
3 0.0
02 0.0
05 -0.
002
0.007
0.006
-0.00
4 0.0
004
0.000
8 0.0
77 0.1
14 0.0
12 -0.
0001
0.21
61 10
.26
(-0.47
) (2.
30)*
(-0.69
) (-2
.27)*
(3.00)
** (3.
09)**
(-1-32
) (4.
07)***
(1.6)
(-1.03
) (0.
98)
(0.17
) (3.
46)***
(5.66
)*** (
1.30)
(-0.08
) 199
8 -0.
008
0.05
0.001
-0.00
2 0.0
01 0.0
03
0.001
0.001
0.006
0.003
0.0
01 0.0
01 -0.
003
-0.01
8 -0.
005
0.000
4 0.0
757
4.88
(-0.35
) (1.
54)
(0.97
) (-2
.39)*
(1.38
) (2.
49)*
(0.7)
(0.95
) (2.
53)*
(1.82)
* (0.
22)
(0.65
) (-4
.41)**
*(-1.0
8) (-0
.53)
(0.48
) 199
9 -0.
127
0.069
0.003
-0.
002
0.004
-0.00
3 0.0
1 -0.
001
0.002
0.002
0.0
01 -0.
007
0.002
0.011
-0.02
5 0.0
001
0.066
0 4.8
2 (-2
.89)**
(1.20)
(1.
94)
(-1.48
) (3.
44)***
(-2.09
)* (0.
52)
(3.22)
** (-0
.35)
-0.52
(0.96
) (-0
.40)
(0.28
) (0.
67)
(-1.04
) (0.
11)
Pool
-0.15
5 0.0
77 0.0
03
-0.00
2 0.0
02 -0.
001
0.007
0.003
0.0
07 0.0
05
0.001
0 -0.
005
-0.00
7 -0.
004
0.000
4 0.16
62 37
.49
(-7.8)
***(2.
83)**
(6.26)
***(-2
.92)**
(4.06)
*** (-
1.02)
(3.75)
***(7.
75)***
(3.05)
** (3.7
5)***
(107)
(0.30
) (-5
.36)**
*(-2.1
0)* (
-0.49
) (0.
42)
FEQ+
i =1 i
f the c
hange
in op
eratin
g inco
me of
the f
ollow
ing ye
ar is p
ositiv
e; an
d 0 ot
herwi
se;
FEt+i
= (Op
eratin
g inco
me of
vear
t+1 —
Oper
ating
incom
e of y
ear t)/
Mark
et val
ue of
total
equity
at th
e beg
inning
of ye
ar t;
RETt
= Mark
et-adj
usted
annu
al ret
urn in
year
t,cum
ulative
mon
thly r
eturns
from
the b
eginn
ing of
May
of year
t to t
he end
of A
pril o
f year
t+1, a
djuste
d for
divide
nds a
nd
stock
rights
; Olt
= Op
eratin
g inco
me in
year
t, scal
ed by
the m
arket
value
of equ
ity at
the b
eginn
ing of
vear
t; lit
=Inve
stmen
t incom
e in y
ear t,
scaled
by th
e mark
et val
ue of
equity
at th
e beg
inning
ofve
ar t;
Elt =
Non-o
perat
ing in
come i
n year
t, sca
led by
the m
arket
value
of equ
ity at
the b
eginn
ing of
year
t; AO
It = U
nexpec
ted op
eratin
g inco
me in
year
t; All
t =Un
expect
ed inv
estme
nt inc
ome i
n year
t; AE
It = U
nexpec
ted no
n-oper
ating
inco
me in
year
t; FS
j,t = N
on-ea
rning
ftmd
ament
al sig
nals i
n year
t. D2
=1 fo
r firm
s with
prop
ortion
of op
eratin
g inco
me gr
eater
than m
edian,
0 oth
erwise
. ***
(**)(*
) T-te
st is
signif
icant
at 0.0
01(0.
01)(0
.1) le
vel
I
CUHK L i b r a r i e s
ll_l__lllllll • oaTSSbflt,