Can Historical Accounting Information Be Used to …¹€อกสาร ADV 30...1 Can Historical Accounting Information Be Used to Predict Future Stock Returns? Pornthep Tantipanichkul

1

Can Historical Accounting Information Be Used to Predict Future Stock Returns?

Pornthep Tantipanichkul PricewaterhouseCoopers Thailand

Bangkok, Thailand

and

Somchai Supattarakul, Ph.D.* Thammasat Business School

Thammasat University Bangkok, Thailand

ABSTRACT Prior research documents a relationship between accounting information and future

stock returns. Specifically, Piotroski (2000) and Mohanram (2005) demonstrate that the investor-friendly composite scores constructed mainly based on historical accounting information can be used to predict future stock returns for high and low book-to-market (BM) firms, respectively. This paper employs two investor-friendly composite scores -- VSCORE and GSCORE, used in Piotroski (2000) and Mohanram (2005), respectively. The composite scores are the sum of binary scores (1 or 0) marked from each individual financial measure. VSCORE represents financial measures related to profitability, leverage/liquidity, and operating efficiency while GSCORE represents profitability measures and growth-oriented measures related to earnings stability, growth stability, capital expenditures and advertising expenses intensity.

This paper provides empirical evidence for listed firms in the Stock Exchange of Thailand (SET) and the Market for Alternative Investment (mai) during 1994 to 2008 suggesting that a portfolio of stocks with higher composite scores (both VSCORE and GSCORE) earn higher one-year and two-year ahead market-adjusted returns without additional risk and that a zero-investment portfolio of longing high VSCORE (GSCORE) stocks and shorting low VSCORE (GSCORE) earn significant positive future market-adjusted returns. In addition, this paper documents that a portfolio of stocks with higher VSCORE earn higher future market-adjusted returns without additional risk for both subsamples of high and low BM firms although VSCORE is used for only high BM firms in Piotroski (2000). Similarly, our empirical results suggest that a portfolio of stocks with higher GSCORE earn higher future market-adjusted returns without additional risk for both high and low BM firms although GSCORE is used for only low BM firms in Mohanram (2000). Keywords: Historical Accounting Information, Future Stock Return, Book-to-Market Ratio * Corresponding author: Thammasat Business School, Thammasat Univesity, 2 Prachan Rd., Pranakorn, Bangkok, 10200 Thailand. email: [email protected]

2

Can Historical Accounting Information Be Used to

Predict Future Stock Returns? 1. INTRODUCTION

One of the most controversial issues in today’s investment world is the challenge posed to the value of fundamental analysis as a reliable tool to reach profitable investment decisions. Despite it being supported by numerous academic research as a useful means of stock trading, the fundamental analysis has raised many questions relating to the efficient market hypothesis (EMH). According to EMH, one cannot exploit both the historical and publicly available information to gain profits if a stock market is semi-strong form efficient. Specifically, if the stock market is efficient, no profitable trading strategy can be formed based on published financial statements. However, the fact that (1) numerous academic researchers find that the fundamental analysis is a useful tool to predict future earnings and stock returns; and (2) financial ratios have long been employed by investors and financial analysts for fundamental analysis, have raised a question relating to the usefulness of accounting information to predict future stock returns. This question may have been extensively addressed in developed countries, but little has been done on emerging markets, and even if there have recently been some findings, the results are neither solid nor reliable due to the limited numbers of samples. Therefore, this paper aims at examining whether historical accounting information can be used to predict future stock returns for Thai stock markets.

Lev and Thiagarajan (1993) document that financial signals have predictive power in explaining contemporaneous stock returns of U.S. firms and Abarbanell and Bushee (1998) show that forming investment portfolios by longing high-score stocks and shorting low-score stocks based on fundamental signals suggested by Lev and Thiagarajan (1993) yields significant positive returns. In addition, empirical results in Piotroski (2000) and Mohanram (2005) suggest that a portfolio with higher composite scores earn higher future returns for high and low book-to-market (BM) firms in U.S. markets, respectively. The composite score in Piotroski (2000) is constructed based traditional financial measures in three aspects: profitability, leverage and liquidity, and operating efficiency while the composite score in Mohanram (2005) is constructed based on traditional profitability measures as well as growth-oriented measures.

In Japan, Nguyen (2003) constructs a simple financial score for each sample firm and finds that the financial scores exhibit a strong correlation with contemporaneous and future market-adjusted returns. In Thailand, Sukanjanapong (2007) documents that historical financial ratios can be used to form profitable stock portfolios, particularly in the small low BM stocks.

This paper empirically examines whether investor-friendly composite scores constructed based on historical accounting information can help investors earn excess future returns for listed firms in the Stock Exchange of Thailand (SET) and the Market for Alternative Investment (mai) during 1994 to 2008. Consistent with Piotroski (2000) and Mohanram (2005), this paper employs simple, yet comprehensive sets of financial measures to construct two composite scores -- “Value Stock” score (VSCORE) and “Growth Stock” score (GSCORE). The composite scores are the sum of binary scores (1 or 0) marked from each individual financial measures. VSCORE represents nine financial measures suggested by Piotroski (2000). These financial measures include signals related to profitability, leverage/liquidity, and operating efficiency. GSCORE represents seven financial measures suggested by Mohanram (2005). These measures include signals related to earnings stability,

3

growth stability, and capital expenditures and advertising expenses intensity, in addition to traditional profitability signals.

Our empirical results indicate that both firms with higher composite score (either VSCORE or GSCORE earn higher one-year and two-year market-adjusted buy-and-hold returns that do firms with lower composite score without additional risk and that a zero-investment portfolio of longing high score stocks and shorting low score stocks earn significant positive market-adjusted returns. Specifically, a zero-investment portfolio based on VSCORE (GSCORE) earns one-year and two-year ahead market-adjusted returns of 11.98% (14.94%) and 31.14% (30.67%), respectively. This suggests that historical accounting information can be used to predict future stock returns.

Piotroski (2000) suggests that his composite score is appropriate for high BM firms while Mohanram (2005) argues that his composite score is appropriate for low BM firms. This paper then also examines whether VSCORE and GSCORE are associated with future stock returns for subsamples of high and low BM firms. We classify low BM stocks as firms with BM ratio below 30th percentile while firms with BM ratio above 70th percentile are classified as high BM stocks. Consistent with results for our full sample, our empirical results for both subsamples of high and low BM firms, show that portfolios of stocks with higher VSCORE and GSCORE earn higher one-year and two-year market-adjusted buy-and-hold returns than do those with lower VSCORE and GSCORE without additional risk and zero-investment portfolios of longing high score stocks and shorting low score stocks earn significant positive market-adjusted returns. Specifically, a zero-investment portfolio based on VSCORE (GSCORE) earns one-year and two-year ahead market-adjusted returns of 18.70% (16.81%) and 31.56% (16.72%), respectively, for a subsample of high BM firms and of 11.44% (23.92%) and 16.57% (8.01%), respectively, for a subsample of low BM firms.

Our empirical results contribute to the literatures on the usefulness of historical accounting information in predicting future stock returns. While prior research finds that financial ratios are associated with future stock returns, our study, together with Piotroski (2000) and Mohanram (2005), provide empirical evidence suggesting that the investor-friendly composite scores constructed based mainly on historical accounting information can be used to choose stocks to invest to earn positive abnormal returns and they can be applied for not only high or low BM firms, but also for all firms. Moreover, our results contribute to the literatures on the efficient market hypothesis. Specifically, our results that investors can use publically available, historical accounting information to choose stocks and earn abnormal stock returns seem to suggest that Thai stock markets are not semi-strong form efficient.

The next section of this paper discusses literature review. Section 3 discusses a construction of composite scores, stock return calculation as well as sample selection and data collection. Section 4 present empirical results. Finally, section 5 concludes the paper. 2. LITERATURE REVIEW 2.1. The Book-to-Market Effect

A large number of academic research demonstrates that book-to-market (BM) ratio is strongly positively associated with future stock returns.1 Chan, Hamao, and Lakonishok (1990) document that BM ratio, along with earnings-to-price ratio, among others, exhibits an important role in explaining future stock returns in Tokyo Stock Exchange. In the U.S. stock markets, low (high) BM firms generally earn significant negative (positive) returns. Chen and Zhang (1998) also explore the relationship between BM ratio and stock returns from both

1 Book-to-Market ratio is defined as a firm’s book value of equity divided by its market capitalization.

4

developed and emerging markets during 1970 – 1993, and find that BM ratio is highly positively correlated to stock returns in the United States, Japan, Hong Kong and Malaysia, while the relationship is not observed in Thailand and Taiwan.

Although Fama and French (1992) and Lakonishok, Shleifer and Vishny (1994) show that a portfolio of high BM firms outperforms that of low BM firms, they provide two different explanations, namely, risk and mispricing explanations, respectively. Fama and French (1992) explain that the basic argument underlying risk-based concept is simply the fact that different types of stocks are exposed to different amount of systematic risk; and therefore, carry different expected returns. Specifically, they show that the variation of cross-sectional stock returns can be explained by two different factors, namely BM ratio and firm size. They claim that bankruptcy risk or financial distress risk is represented by BM ratio, while firm size acts as a proxy of liquidity risks. High BM ratio means the market judges firm’s prospects to be poor relative to the entire market, so BM ratio may capture financial distress effect. Thus, high BM firms are likely to have greater bankruptcy risks; and hence, higher excess returns in compensation for higher additional risk. Nevertheless, this explanation is less valid for low BM firms, since it is contrary to the fact that low BM stocks are more risky than the stock market as a whole; and therefore, should generate high returns.

Alternatively, Lakonishok, Shielfer and Vishny (1994) argue that there is little evidence that high BM stocks are fundamentally riskier. They claim that high BM stocks produce superior returns because typical investors consistently overestimate future growth of low BM stocks relative to high BM stocks. In other words, investors are extremely pessimistic (optimistic) about high (low) BM stocks as they tie expectations of future growth to past bad (good) growth/earnings; hence, they put excessive weight on the recent past for prediction of future returns. They oversell the stocks that have recently performed poorly and overbuy the stocks that have performed well. Therefore, these stocks are either underpriced and have a high BM, or overpriced and have a low BM. This mispricing explanation implies that typical investors make systematic errors in predicting future growth earnings of stocks; therefore, one can exploit the mistakes of typical investors by purchasing high BM stocks and shorting low BM stocks. This is a common judgment error and may explain the investor preference of low BM stocks (growth stocks) over high BM stocks (value stocks). Their empirical evidence also suggests that institutional investors prefer low BM stocks over high BM stocks, and are willing to pay them at a premium price because they represent prudent investments. LaPorta (1996) also supports this mispricing explanation.

Investors are often the victims of the mispricing effect. They often estimate firm’s future prospect from past performance while ignoring the tendency of corporate profit growth to revert to the mean. Fuller, Huberts, and Levinson (1993) explain that earnings growth rates tend to revert to the mean quickly because of the nature of the capital markets. They find that, although earnings per share (EPS) growth rate of high price-to-earnings (PE) group substantially exceeds that of low PE group in the first year of portfolio formation, it converges closely to the mean after only 4 years. Stated differently, investors are misled by past growth and overlook the nature of business competition. Industries which are experiencing the high growth tend to attract heavy competition by other firms. This competitive process eventually results in lower growth rate and lower returns. Instead, industries with low growth rate attract less capital from the market. Therefore, in order to survive in the competition, management tries to achieve higher earnings by operating more efficiently.

Surprisingly, in an investment world, several brokerage houses do not recommend their clients to buy high BM stocks (value stocks). Stickel (1998) finds that analysts prefer recommending firms with recent strong performance (low BM stocks or growth stocks) because they anticipate high BM stocks to continuously underperform the market in the near

5

future and they recognize the profits from the strategy that depends on purchasing low BM stocks. This is consistent to the mispricing concept discussed by Lakonishok, Shielfer and Vishny (1994) and LaPorta (1996). 2.2. Fundamental Analysis

Lev and Thiagarajan (1993) introduce 12 financial signals widely used in analyst’s reports, and find that most fundamental signals have predictive power in explaining contemporaneous stock returns of US firms. Abarbanell and Bushee (1998) show that forming investment portfolios by longing high-score stocks and shorting low-score stocks based on 9 fundamental signals suggested by Lev and Thiagarajan (1993) yields significant positive returns.

Piotroski (2000) applies fundamental analysis to develop investment strategy for high book-to-market (BM) firms in US markets. He observes that although high BM firms earn high future stock returns, these high stock returns only come from a few firms suggesting that BM ratio alone might sometimes not be adequate to identify good quality stocks in which investors should invest. Hence, a binary score of financial ratios is given to each firm, with 1 indicating that firms possess strong financial status in each of these 4 aspects: profitability, operating efficiency, liquidity, and leverage, and with 0 otherwise. Firms are then ranked by total binary scores. He indicates that a simple strategy of separating winners from losers by using basic financial ratios has the ability to earn large future excess returns. Further, since weak fundamental firms, on average, generate negative excess returns, an investment strategy that buys strong fundamental firms and shorts weak fundamental firms can earn a large magnitude of positive return.

In contrast, Mohanram (2005) documents that one can also apply a fundamentals driven strategy, appropriately modified by other measures specifically for growth firms such as the stability of earnings, sales growth, intensity of R&D, capital expenditures, and advertising, on a sample of low BM stocks in US markets to separate winners from losers, though a large portion of returns is conditioned by the investor’s ability to short sell stocks.

In Japan, Nguyen (2003) constructs a simple financial score for each sample firm and finds that the financial scores exhibit a strong correlation with market-adjusted returns in the current and the following periods, though the longer the holding period, the lower the returns. In Thailand, Sukanjanapong (2007) documents that using simple-to-execute historical financial ratios to form stock portfolios can provide significant positive market-adjusted returns, particularly in the small low BM stocks. 3. RESEARCH METHODOLOGY 3.1. Construction of Composite Scores

Consistent with Piotroski (2000) and Mohanram (2005), this paper uses simple, yet comprehensive sets of financial ratios to predict future stock returns. Specifically, this paper constructs two composite scores (“Value Stock” score and “Growth Stock” score) based on financial signals. A realization of each financial signal is classified as either good or bad depending on its implication to stock future returns, with 0 and 1 score representing bad and good implication, respectively. The composite scores are the sum of binary scores (1 or 0) marked from each individual financial signals.

The “Value Stock” composite score (VSCORE) represents nine traditional financial measures suggested by Piotroski (2000). These financial measures include signals related to profitability, leverage/liquidity, and operating efficiency. Another composite score is the “Growth Stock” score (GSCORE), representing financial measures suggested by Mohanram

6

(2005). These measures include signals related to earnings stability, growth stability, capital expenditures intensity, and advertising expenses, in addition to traditional financial signals.

VSCORE and GSCORE are the sum of all binary scores from each individual financial signal as follows:

Signals VSCORE GSCORE Traditional Fundamental Signals (i) Profitability Signals

bROA (ROA > Industry Median) b∆ROA (∆ROA > 0) bCFROA (CFROA > Industry Median) bACCRUAL (CFROA > ROA)

(ii) Leverage/Liquidity Signals b∆LEV (∆LEV < 0) b∆LIQ (∆LIQ > 0) bEQOFF (Register to issue common equity)

(iii) Operating Effiency Signals b∆OPM (∆OPM > 0) b∆TATO (∆TATO > 0)

Growth-Oriented Fundamental Signals bVARROA (VARROA < Industry Median) bVARSG (VARSG < Industry Median) bCAPEX (CAPEX > Industry Median) bADV (ADV > Industry Median)

Since there are nine signals for VSCORE and seven signals for GSCORE, the

composite score ranges within zero to nine (0 – 9) for VSCORE and within zero to seven (0 – 7) for GSCORE. All signals are discussed as follows. 3.1.1. Traditional Fundamental Signals

This paper implements all nine fundamental signals used in Piotroski (2000). These nine signals are divided into three categories: profitability, liquidity/leverage, and operating efficiency.

(i) Profitability Signals

Albeit of its rising stock price in the previous period, growth firms are very likely to experience low earnings and negative cash flow; consequently, any firm currently generating positive cash flows or more profits than its counterparts displays a signal of improving profitability and should earn a score of 1. For value firms, given the poor historical earnings performance, any firms generating positive profits or cash flows are demonstrating stronger financial health in the future, with positive cash flows showing improving flow of internal funds injected in operating activities, while positive earnings representing higher margins and/or improvement of cost control.

In total, there are four signals for the profitability aspects: ROA, ∆ROA, CFROA, and ACCRUAL. ROA is defined as operating income divided by total assets. Operating income is applied here to avoid non-recurring earnings of any firm. A binary score for ROA (bROA) equals 1 if a firm’s ROA is greater than industry-median ROA and 0 otherwise.2

2 We use median ROA instead of mean ROA to avoid possible extreme values. Median is also applied to other signals where applicable.

7

Furthermore, being profitable is also measured by an increasing trend of profitability. Firms that exhibit a growing trend of profits are more likely to achieve higher future returns.3 Therefore, a binary score for ∆ROA (b∆ROA) equals 1 if a ∆ROA is positive, 0 otherwise.

CFROA is defined as a firm’s cash flows from operations divided by total assets. Since analysts generally use operating cash flows to predict firm’s financial position, in addition to earnings, a binary score for CFROA (bCFROA) equals 1 if a firm’s CFROA is greater than industry-median CFROA and 0 otherwise. The next signal, ACCRUAL, is a bit more complicated since the relationship between earnings and cash flow need to be simultaneously considered. According to Bernard (1994), the importance of accounting returns and cash flows, as well as their relations to each other, needs rigorous attention when assessing the future prospects of a firm. Furthermore, Sloan (1996) demonstrates that firms with greater accrual component in their earnings generally underperform in the future due to their lower quality of earnings. In other words, if earnings is greater than cash flow from operations, it may suggest a bad signal about future profitability; thus, a binary score for ACCRUAL (bACCRUAL) equals 1 if CFROA > ROA and 0 otherwise. (ii) Leverage/Liquidity Signals

The next three signals are ∆LEV, ∆LIQ, and EQOFF. These signals are included in the composite scores to capture firm’s capital structure and its ability to serve short-term debt obligation. ∆LEV is a change in financial leverage measured by firm’s total interest-bearing debt divided by common equity. A binary score for ∆LEV (b∆LEV) equals 1 if a firm’s ∆LEV is negative and 0 otherwise. A decrease in financial leverage is viewed as a positive signal because it demonstrates the firm’s ability to service existing debt obligations. Also, as suggested by Myers and Majluf (1984), by raising external capital, the firm is signaling its inability to generate sufficient internal funds for future operations. Besides, an increase in long-term liabilities may pose more challenges and extra constraints to the firm’s financial flexibility, in addition to its current covenants. This is especially true for high BM firms, which generally experience poor performance recently; however, if they are able to decrease their leverage, this might signal that they are starting to have more capabilities to handle their financings.

∆LIQ is a change in liquidity measured by a firm’s current assets divided by current liabilities (a.k.a. current ratio). A binary score for ∆LIQ (b∆LIQ) equals 1 if ∆LIQ is positive and 0 otherwise, as an improvement in liquidity should imply that a firm is able to meet its short-term debts. A binary score for EQOFF (bEQOFF) is equal to 1 if a firm “registers” to issue common equity in the year before construction of portfolio (even if a firm does not issue an equity in that given year) and 0 otherwise.4 This signal really holds true in high BM firms. The fact that these firms are willing to issue equity even when their stock prices are likely to be depressed in the future highlights the poor condition of these firms. (iii) Operating Efficiency Signals

The last two signals relate to operating efficiency which is the firm’s ability to generate returns from its asset base. Piotroski (2000) applies Dupont framework in which ROA can be disaggregated into two components: operating profit margin and total asset turnover. ∆OPM is a change in a firm’s operating profit margin measured by the firm’s operating profit divided by operating revenues, and ∆TATO is a change in a firm’s total asset

3 Note that even if the firms have negative ROA (losses) in the previous fiscal period, but if they show an improving trend, they are potentially more likely to be profitable in the future. 4 Common equity issuance refers to equity transactions between firms and investors that involve a firm receiving cash. Examples of these transactions include public offerings, private placement, pre-emptive rights for current stockholders, and exercises of warrant, convertible debentures. However, IPO and ESOP exercise are excluded.

8

turnover measured the firm’s operating revenues divided by total assets. A binary score for ∆OPM (b∆OPM) equals 1 if ∆OPM is positive and 0 otherwise, and a binary score for ∆TATO (b∆TATO) is equal to 1 if ∆TATO is positive and 0 otherwise. 3.1.2. Growth-Oriented Signals

It is still an open question whether the above traditional fundamental signals work well with growth firms. Since they are at their early stage of business, these typical signals, especially total asset turnover ratio, earnings and cash flow-related signals, may not possess much ability to effectively separate winner firms from losers; while other measures such as liquidity and leverage may need to be placed with extra importance since these firms may not necessarily be profitable.

As demonstrated by Mohanram (2005), the traditional signals are unsuccessful for growth firms; therefore, we also include growth-oriented signals that may be more useful to identify future winners. Such signals include earnings stability, growth stability, R&D intensity, capital expenditures, and advertising expenses.

Earnings stability is defined as difference between a firm’s ROA in two given years. As suggested by Mohanram (2005), earnings stability may help distinguish between firms with good future prospect and firms that are overvalued because of the excitement or hype of market. Furthermore, Barth, Elliot, and Finn (1999) show that investors ultimately reward firms with earnings stability, as they are more likely to have stable performance in the future. Thus, a binary score for VARROA (bVARROA) equals 1 if a firm’s earnings variability is less than industry median and 0 otherwise.

Firms that show consistent growth are more likely to exhibit stable future growth; and thus, making them less vulnerable to underperform in the future. This is also consistent with the mispricing explanation by Lakonishok, Shielfer, and Vishny (1994) and LaPorta (1996). They conclude that low BM firms represent stocks that are attractive to investors simply due to recent strong performance, which then lead to the formation of “too optimistic” expectations about future performance. LaPorta (1996) further refers to this phenomenon as “naïve extrapolation” in his paper. Consequently, a binary score for VARSG (bVARSG) equals 1 if a firm’s variability in sales growth is less than industry median and 0 otherwise.

Many studies also examine growth firms such as technology firms or R&D intensive firms. Specifically, Lev and Sougiannis (1996) study the value relevance of R&D and find that R&D intensive firms earn excess returns in future periods. Chan, Lakonishok, and Sougiannis (2001) confirm this and also find that advertising expenses are associated with excess returns in the future. Thus, R&D, capital expenditures, and advertising expenses play an important role in growth stocks. High level of these measures highlights a firm’s future investment, and is a reliable signal for future growth, in spite of depressing today’s cash flow and book value through lower earnings (R&D and advertising). Referred by Mohanram (2005) as accounting conservatism, these unrecorded intangibles are effective in separating winners from losers because good prospect firms may have a low BM ratio simply due to accounting reasons rather than overvaluation or hype of investors. Hence, a binary score for CAPEX (bCAPEX) and ADV (bADV) equals to 1 if a firm’s capital expenditure and advertising expense, all divided by total assets, are greater than industry median, respectively. However, we do not adopt R&D in our investigation due to the limited disclosures.5

5 There are only 2 firms between 1997-2009 in SET and mai that voluntarily discloses its R&D expenses, both of which are in the energy sector.

9

3.2. Stock Return Calculation Raw return of each firm in each year is calculated as a buy-and-hold strategy.6 We

calculate returns for one year and two consecutive years starting from the beginning of fourth month after the fiscal year end and ending at the end of third month after the following one (two) fiscal year(s).7 Return of each portfolio formed based on the composite scores is calculated by equally weighted all raw returns in the portfolio. Market-adjusted return (MAR) is also calculated by subtracting market returns from portfolio returns over the corresponding period. Market return is simply computed using value-weighted approach.8 We apply buy-and-hold returns throughout the paper as Blume and Stambaugh (1983) state that the buy-and-hold strategy has an advantage in explaining the portfolio performance since it does not require frequent portfolio rebalancing which leads to higher transaction costs. Therefore, this strategy is more likely to make large profits for investors.

3.3. Sample Selection and Data Collection

All historical financial statement data, stock price, market capitalization, and trading volume are obtained from Datastream database during 1994 to 2008. Equity issuance data are obtained from SETSMART, a sophisticated database consisting of all important news for each Thai public listed company. Advertising expense data are extracted from a firm’s notes to financial statement. The final sample consists of 425 firms listed in both the Stock Exchange of Thailand (SET) and the Market for Alternative Investment (mai) during 1994 to 2008. The sample excludes firms in the banking, finance and securities, and insurance sector, as well as property funds, and companies under rehabilitation since they require different framework for financial statement analysis.9 Since a large numbers of growth firms, especially those in mai, have recently gone public, we do not eliminate firms which insufficient prior data is unavailable. However, because the distribution of stock returns is largely influenced by outliers, we apply the trimming procedures to dispose extreme values at 1st and 99th percentile. In order to categorize sample stocks into low or high BM group, we calculate BM ratios for every firm at fiscal year-end and divide the sample into percentiles accordingly.10 Since this paper focuses not only the entire stock population, but also a group of high and low BM stocks, we classify low BM stocks as firms with BM ratio below 30th percentile while firms with BM ratio above 70th percentile are classified as high BM stocks. Total final observations for the entire sample, high BM stocks, and low BM stocks contain 3,565 firm-years, 1,051 firm-years, and 1,070 firm-years, respectively.

6 Buy-and-hold returns are calculated as the difference of ending and beginning stock price plus dividend per share (if any) and divided by the beginning stock price. They capture both the capital gain yield and dividend yield. 7 For example, for the fiscal year end of December 31, 2000, the one-year and two-year future return period starts on April 1, 2001 and ends on March 31, 2002 and March 31, 2003, respectively. Moreover, this fourth month may not necessarily be April, as it depends on firm’s accounting period. 8 Guay (2000) suggests that the use of value-weighted approach to compute market-adjusted returns in high BM stocks may contaminate the benefits of empirical results. Given that high BM firms tend to be relatively small, an equally-weighted market-adjusted return, which receives equal weights from every firm, may seem more appropriate. However, we rely on market-adjusted return throughout our paper as (1) our samples include both high and low BM firms, and (2) to allow for consistency. 9 Property funds are excluded as they themselves are simply listed in the stock market for ease of investor’s transferability, and hence their business nature and income are similar to the owner of the fund. Therefore, inclusion of these property funds might cause redundancy and autocorrelation of sample. 10 For simplicity, we only employ BM ratio in the classification process despite the support from some scholars to include more than one variable (Calderwood, 1995). For consistency, we use accounting fiscal year end for BM calculation.

10

4. EMPIRICAL RESULTS 4.1. Descriptive Statistics

Table 1 presents the descriptive statistics of the financial characteristics of the sample firms. Descriptive statistics are presented for the full sample, as well as high and low BM firms. The means for most financials are greater than the medians, indicating the presence of some very large firms. Surprisingly, based on medians, high BM firms have far fewer assets and sales than low BM firms because low BM firms tend to spend large amount of investment in capital expenditures, making their asset bases extremely large.

==========================

Insert Table 1 here. ==========================

Low BM firms, based on medians, generate more operating income and cash flows

from operating activities than do high BM firms. Their earnings and cash flows are also more volatile than those of high BM firms (as indicated by high standard deviations). This is consistent with our expectation that low BM firms are generally stocks that experience either extremely high or extremely low earnings; and thus, their earnings should be more volatile. Also, the fact that their earnings are more vulnerable to manipulation may result in higher earnings volatility. This suggests that firms with low BM group are hardly homogenous, and any strategy that can effectively separate good low BM stocks from bad low BM stocks is likely to be profitable.

Consistent with Fama and French (1995) and Piotroski (2000), high BM firms earn lower mean (median) ROA than do low BM firms. This may partly be due to the fact that high BM portfolio consists of a vast majority of poor performing firms. Low BM firms also grow at a much faster rate with a median annual growth at around 14.09% as opposed to high BM firms which grow at only 4.24% annually. As expected, low BM firms have much greater capital expenditure intensity than high BM or universe firms.

Table 1 also presents the descriptive statistics of stock returns. Low BM firms earn mean positive market-adjusted returns (MAR) of 1.03% and 6.35% for the first and second year, respectively while high BM firms earn 15.65% and 35.02%, respectively. Consistent with Lakonishok, Shleifer, and Vishny (1994), returns of high BM firms are much more pronounced. The 25th percentile shows that low BM firms have MAR of -33.10% and -40.98% for the one and two-year buy-and-hold strategy (-21.75% and -20.75% for high BM firms). In both windows, medians MAR for low BM firms are negative, but positive for high BM firms. Returns on both low and high BM firms in the 75th percentile, however, change from negative to positive values. The result can be inferred that any strategy that can choose which firms to end up in the two tails of the distribution and tries to reach profitable returns from the differences is likely to produce fruitful returns. As expected, high BM firms yield more positive proportion of returns than low BM firms.

4.2. Correlation Analysis

Table 2 presents the correlations between the financial measures (in a binary score) as well as the one-year and two-year buy-and-hold market-adjusted returns (MAR) for all firms. We present both Pearson’s correlation and Spearman rank-order’s correlation as our sample consists of both ordinal and ratio scale. In addition to positive correlations between increasing profitability (bROA) and increasing profit margin and turnover (b∆OPM and b∆TATO), denoting the evidence of Dupont ROA-disaggregation framework, there is also a positive relationship between the earnings-based and cash-flow based measures of profits (bROA and bCFROA). Additionally, the fact that profitable firms (bROA) or firms with

11

increasing profitability (b∆ROA) are likely to have stable sales growth (bVARSG) and stable earnings growth (bVARROA) supports our preconception about profitability.

==========================

Insert Table 2 here. ==========================

Strangely, signals about improving profitability (b∆OPM) and advertising intensity

(bADV) show negative correlations with one-year MAR. This may imply that firms with large portion of advertising expenses have lower subsequent returns compared to firms with less advertising spending. The interpretation here is inconsistent with the fact that high level of advertising expenses highlights firm’s future investment, and is a reliable signal for future growth. However, signals relating to actions that may depress current earnings and book value, but boost future growth – capital expenditures (bCAPEX) and advertising expenses (bADV), have weak correlations between each other, which compares favorably with Mohanram (2005).

Significant correlations between the composite scores and subsequent market-adjusted returns (MAR) provide evidence of return predictability based on past financial measures. With one-year MAR, which is corresponded to a four-month lapse after accounting period, correlations for all composite scores are significantly positive, indicating that returns are predictable based on a combination of financial information that is available at the time of portfolio construction. However, the correlations between the composite scores and returns decrease when the investment horizon is lengthened to 2 years. Also, the strength of the correlations becomes less significant. One possible reason is that the information contained in the score has already been integrated into stock prices. 4.3. Composite Scores and Future Stock Returns

The primary methodology employed in this paper is consistent with Piotroski (2000) and Mohanram (2005). As discussed earlier, there are two composite scores: VSCORE and GSCORE. These scores are constructed in such a way that higher scores imply stronger firm performance and in turn high future stock returns. Therefore, firms with high (low) composite scores are expected to have high (low) future stock returns.

We first examine whether VSCORE and GSOCRE are positively associated with future stock returns for all sample firms. Specifically, we empirically examine whether firms with higher VSCORE (GSCORE) earn higher future market-adjusted stock returns and whether a zero-investment portfolio of longing high VSCORE (GSCORE) stocks and shorting low VSCORE (GSCORE) earn positive future market-adjusted stock returns. Our empirical results are discussed in section 4.3.1. In addition, we examine whether VSCORE and GSCORE can also be used to form profitable portfolios for subsamples of high and low BM firms. The empirical results are discussed in section 4.3.2. 4.3.1. Composite Scores and Future Returns for All Sample Firms

Table 3 and 4 demonstrate portfolios of all sample firms from each composite score, VSCORE and GSCORE, respectively. In each table, Panel A shows the one-year investment horizon while panel B shows the two-year investment horizon. VSCORE (GSCORE) ranges from 0 to 9 (0 to 7) since it is constructed based on nine (seven) financial measures. For VSCORE (GSCORE), the high score group consists of stocks with scores of 5, 6 and 7 while the low score group consists of stocks with scores of 0 and 1.

Our results presented in the tables show that higher score firms significantly outperform lower score firms in both one and two year after portfolio formation.

12

Specifically, for VSCORE, the mean (median) of one-year MAR for the high and low group is 11.98% (4.08%) and 1.98% (-11.58%), respectively, producing a significant return difference (High – Low) of 10.00% (15.66%). Similarly, the mean (median) of two-year MAR for the high and low group is 31.14% (14.70%) and 7.23% (-8.82%), producing a significant return difference (High – Low) of 23.90% (23.52%).

Consistently, for GSCORE, the mean (median) of one-year MAR for the high and low group is 14.94% (7.30%) and 5.34% (-6.16%), respectively, resulting in a significant return difference (High – Low) of 9.59% (13.46%). Similarly, the mean (median) of two-year MAR for the high and low group is 30.67% (12.50%) and 13.30% (-4.19%), resulting a significant return difference (High – Low) of 17.37% (16.69%). The results suggest that VSCORE and GSCORE constructed based on historical accounting information can be used to predict future stock returns and a zero-investment portfolio of longing high score stocks and shorting low score stocks earn significant positive future returns.

==========================

Insert Tables 3 and 4 here. ==========================

4.3.2. Composite Scores and Future Returns for Subsamples of High and Low BM Firms

Table 5 and 6 report portfolios of a subsample of high BM formed based on VSCORE and GSCORE, respectively. Panel A (Panel B) of each table shows the one-year (two-year) investment horizon. Similar to results for all sample firms, our results for a subsample of high BM firms indicate that higher score firms earn more positive subsequent abnormal returns than do lower score firms. Specifically, for VSCORE, the mean (median) of one-year MAR for the high and low group is 24.09% (12.97%) and 5.39% (-3.82%), respectively. Consequently, a mean (median) return difference (High – Low) is 18.70% (16.78%), respectively. Similarly, for two-year MAR, the mean (median) for the high and low group is 50.27% (35.11%) and 18.71% (3.89%). As a result, a mean (median) return difference (High – Low) is 31.56% (31.22%). Results for VSCORE are consistent with Piotroski (2000).

For GSCORE, the mean (median) of one-year MAR for the high and low group is 26.08% (18.49%) and 14.64% (1.64%), respectively, resulting in a significant return difference (High – Low) of 11.44% (16.85%) and the mean (median) of two-year MAR for the high and low group is 46.40% (32.42%) and 22.48% (3.38%), resulting a significant return difference (High – Low) of 23.92% (29.04%). Consistent with results for all sample firms, our results for high BM firms suggest that VSCORE and GSCORE can also be used to predict future stock returns and a zero-investment portfolio of longing high score stocks and shorting low score stocks earn significant positive future returns.

==========================


Table 7 and 8 present portfolios of a subsample of low BM formed based on

VSCORE and GSCORE, respectively. Panel A (Panel B) of each table shows the one-year (two-year) investment horizon. Similar to results for all sample firms and a subsample of high BM firms discussed earlier, our results for a subsample of low BM firms reveal that higher score firms earn greater future abnormal returns than do lower score firms. Specifically, the mean (median) of one-year MAR for the high and low VSCORE group is 5.38% (0.92%) and -11.43% (-22.72%), respectively. As a result, a mean (median) return

13

difference (High – Low) is 16.81% (23.64%), respectively. Consistently, for two-year MAR, the mean (median) for the high and low VSCORE group is 8.82% (2.89%) and -7.90% (-20.50%). As a result, a mean (median) return difference (High – Low) is 16.72% (23.40%).

For GSCORE, the mean (median) of one-year MAR for the high and low group is 7.62% (1.71%) and -8.94% (-12.02%), respectively, resulting in a significant return difference (High – Low) of 16.57% (13.73%) and the mean (median) of two-year MAR for the high and low group is 14.31% (4.48%) and 6.29% (-25.69%), resulting a significant return difference (High – Low) of 8.01% (30.17%). Results for GSCORE are consistent with Mohanram (2005).

Consistent with results for all sample firms and a subsample of high BM firms, our results for low BM firms suggest that both VSCORE and GSCORE can also be used to predict subsequent stock returns and a zero-investment portfolio of longing high score stocks and shorting low score stocks earn significant positive future returns. However, it is worth noting that the significant positive return for the zero-investment portfolio mainly results from a short low score firms since these firms earn negative future abnormal returns.

==========================


4.3.4. Additional Tests: Do Greater Returns Come with Higher Risk?

Higher returns for firms with higher scores (either VSCORE or GSCORE) may potentially come with high risks. In other words, high score portfolio may generate high returns just because of a vast majority of high-risk stocks in the portfolio, and the lower returns in the low score portfolio may solely result from a large numbers of low-risk stocks. If so, historical accounting information may not be as useful as we would hope because investors can obtain high returns merely from choosing stocks with high risks. Thus, in this section, we further examine the relationship between portfolio returns formed based on the composite scores and their associated ex-post risks.

This paper employs 3 indicators as a risk proxy: beta, return volatility, and debt-to-equity ratio. These three risk proxies for each portfolio formed based on VSCORE (GSCORE) are reported in the last three columns in tables 3, 5 and 7 (4, 6 and 8) for all sample firms, high BM firms and low BM firms, respectively. We compare all three risk proxies between high and low score groups and find that almost all cases, risk proxies for high score group are significantly lower than those for low score firms. In other words, high score groups are not riskier than low score groups. In summary, portfolios with higher VSCORE and GSCORE earn higher future returns than do portfolios with lower VSCORE and GSCORE, without additional risk.

5. CONCLUSION

This paper shows that a simple accounting-based fundamental-driven strategy on a sample of (1) all firms (2) high BM firms and (3) low BM firms, can effectively earn significant positive future abnormal stock returns. Our sample includes listed firms in the Stock Exchange of Thailand (SET) and the Market for Alternative Investment (mai) during 1994 to 2008. We employ two investor-friendly composite scores -- VSCORE and GSCORE, used in Piotroski (2000) and Mohanram (2005), respectively. The composite scores are the sum of binary scores (1 or 0) marked from each individual financial measure. VSCORE represents traditional financial measures in three areas: profitability, leverage/liquidity, and operating efficiency while GSCORE represents profitability measures

14

and growth-oriented measures related to earnings stability, growth stability, capital expenditures and advertising expenses intensity.

Our empirical evidence suggests that a portfolio of stocks with higher composite scores (both VSCORE and GSCORE) earn higher one-year and two-year ahead market-adjusted returns and that a zero-investment portfolio of longing high VSCORE (GSCORE) stocks and shorting low VSCORE (GSCORE) earn significant positive future market-adjusted returns for all sample firms, high BM firms and low BM firms. Our results with respected to VSCORE for high BM firms are consistent with Piotroski (2000) and our results with respected to GSCORE for low BK firms are consistent with Mohanram (2005).

We also further examine whether higher returns for portfolios with higher scores come with higher risk. We employ three risk proxies: beta, return volatility, and debt-to-equity ratio. Our results show that high score portfolios are not riskier than low score portfolios. Overall, high score portfolios earn more positive abnormal returns than do low score portfolios without additional risk.

Our empirical results contribute to the literatures on the usefulness of historical accounting information in predicting future stock returns. While prior research finds that financial ratios are associated with future stock returns, our study, together with Piotroski (2000) and Mohanram (2005), provide empirical evidence suggesting that the investor-friendly composite scores constructed based mainly on historical accounting information can be used to choose stocks to invest to earn positive abnormal returns and they can be applied for not only high or low BM firms, but also for all firms. Moreover, our results contribute to the literatures on the efficient market hypothesis. Specifically, our results that investors can use publically available, historical accounting information to choose stocks and earn abnormal stock returns seem to suggest that Thai stock markets are not semi-strong form efficient.

15

REFERENCES Abarbanell, J., and B. Bushee, “Abnormal Returns to a Fundamental Analysis Strategy”, The

Accounting Review, January 1998, pp. 1-24. Barth, M., J. Elliot, and M. Finn, “Market Rewards Associated with Patterns of Increasing

Earnings”, Journal of Accounting Research, Autumn 1999, pp. 387-413. Blume, M. E., R. F. Stambaugh, “Biases in Computed Returns: An Application to the Size

Effect,” Journal of Financial Economics, 1983, pp. 387-404. Calderwood, S., “The Positive Bias for Value Investors in U.S. Equities”, Association for

Investment Management and Research, 1995, pp. 4-13. Chan, L., J. Lakonishok, and T. Sougiannis, “Momentum Strategies”, Journal of Finance,

December 1996, pp. 1681-1713. Chan, L., Y. Hamao, and J. Lakonishok, “Fundamental and Stock Returns in Japan”, Journal

of Finance, 1991, pp. 1739-1789. Fama, E., and K. French, “The Cross-Section of Expected Stock Returns”, Journal of

Finance, 1992, pp. 427-465. Fuller, J., L.Huberts, and M. Levinson, “Returns to E/P Strategies, Higgledy-Piggledy

Growth, Analysts’ Forecast Errors, and Omitted Risk Factors”, Journal of Portfolio Management, Winter 1993, pp. 13-24.

Guay, W., “Discussion of Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers”, Journal of Accounting Research, 2001, pp.43-51.

Lakonishok, J., A, Shleifer, and R. Vishny, “Contrarian Investment, Extrapolation and Risk”, Journal of Finance, December 1994, pp. 1541-1578.

LaPorta, R., “Expectations and the Cross Section of Stock Returns”, Journal of Finance, December 1996, pp. 1715-1742.

Lev, B., and R. Thiagarajan, “Fundamental Information Analysis”, Journal of Accounting Research, Autumn 1993, pp. 190-214.

Lev, B., and T. Sougiannis, “Capitalization, Amortization and Value-Relevance of R&D”, Journal of Accounting and Economics, February 1996, pp. 107-139.

Mohanram, P., “Separating Winners from Losers among Low Book-to-Market Stocks using Financial Statement Analysis”, Review of Accounting Studies, 2005, pp. 133-170.

Myers, S., and N. Majiluf. “Corporate Financing and Investment Decisions When Firms Have Information That Investors Do Not Have”, Journal of Financial Economics, 1984, pp. 187-221.

Nguyen, P. “Fundamental Analysis and Stock Returns: Japan 1993-2003”, WBP Financial Integrator, 2003, pp. 193-210.

Piotroski, J. “Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers”, Journal of Accounting Research, Vol.38 Supplement 2000, pp. 1-41.

Sloan, R., “Do Stock Prices Fully Reflect Information in Accruals and Cash Flows about Future Earnings”, The Accounting Review, July 1996, pp. 289-316.

Stickel, S. “Analyst Incentives and the Financial Characteristics of Wall Street Darling and Dogs”, Working Paper, LaSelle University, 1998.

Sukanjanapong, P., “Stock Trading Strategy Based on Historical Financial Ratios”, Independent Study Submitted for the Requirement of Master of Science Program in Finance (International Program), Thammasat University, May 2007.

16

TABLE 1 Descriptive Statistics

Firms that have BM ratios greater than the 30th percentile are classified as high BM firms (value stocks), while firms with BM ratios below the 70th percentile are categorized as low BM firms (growth stocks). Because the distribution of stock returns is heavily influenced by outliers, I apply the trimming procedure to discard extreme values of market-adjusted returns at 1st and 99th percentile accordingly. t-statistic for the mean differences are from independent sample t-test. z-statistic are for median differences using wilcoxon ranked-sum test. */**/*** represents statistical significance using two-tailed tests at 10% / 5% / 1% level.

Variable Mean 10

thPercentile 25

thPercentile Median 75

th Percentile 90

thPercentile S.D. % Positive t-statistics z-statistic

All Firms (3,565 observations)Book to Market Ratio 0.67 0.27 0.52 0.96 1.61 2.63 6.47 96.20%Assets 12,007,549 566,849 1,054,981 2,356,056 7,411,787 24,477,592 39,632,398 100.00%Sales 9,896,154 371,503 789,196 1,955,505 5,328,457 14,883,960 56,413,734 100.00%Operating Income 841,120 -98,787 12,665 105,620 397,861 1,368,657 5,007,001 78.85%Cash Flow from Operating Activities 1,061,407 -127,271 20,288 143,127 542,244 1,901,551 5,519,830 79.83%Return on Assets 5.11% -4.31% 0.82% 5.56% 10.51% 16.16% 15.12% 78.85%Sales Growth 28.66% -18.03% -2.42% 9.03% 23.43% 46.91% 562.57% 71.19%Earnings Growth 1.22% -141.09% -54.90% -12.55% 27.96% 121.23% 1464.73% 40.28%Proportion of Capital Expenditures to Assets 6.02% 0.46% 1.37% 3.68% 8.15% 14.93% 6.88% 99.35%One-Year Market-Adjusted Returns 9.23% -50.32% -26.21% -0.88% 31.23% 71.70% 61.06% 49.42%Two-Year Market-Adjusted Returns 22.99% -57.93% -28.31% 4.38% 47.96% 114.29% 91.70% 53.91%Market Returns 4.27% -41.49% -24.63% 3.56% 16.58% 63.08% 39.14% 53.33%

High BM Firms (1,070 observations)Book to Market Ratio 2.51 1.23 1.52 2.04 2.94 4.00 1.77 100.00% 11.948*** 39.809***Assets 6,520,540 545,025 894,763 1,796,232 4,760,688 13,875,659 16,586,532 100.00% 7.098*** 10.078***Sales 4,182,841 279,953 554,964 1,364,810 3,186,462 7,750,977 11,876,923 100.00% -4.671*** -12.143***Operating Income 119,377 -96,918 -5,812 40,845 133,147 443,739 778,138 71.78% -6.912*** -12.846***Cash Flow from Operating Activities 269,539 -104,488 12,309 90,180 265,995 741,083 1,096,341 79.16% -7.325*** -9.818***Return on Assets 2.13% -4.63% -0.47% 2.80% 6.11% 8.99% 7.77% 71.78% -9.501*** -13.960***Sales Growth 12.76% -19.75% -6.76% 4.24% 16.13% 34.96% 91.10% 61.12% -2.868*** -10.731***Earnings Growth -52.72% -209.49% -81.97% -25.66% 30.37% 151.87% 1433.24% 35.14% -2.010** -5.085***Proportion of Capital Expenditures to Assets 4.40% 0.35% 0.98% 2.48% 5.73% 10.70% 5.32% 99.53% -9.957*** -9.423***One-Year Market-Adjusted Returns 15.65% -43.90% -21.75% 4.23% 36.48% 80.88% 62.39% 55.23% 5.544*** 6.319***Two-Year Market-Adjusted Returns 35.02% -49.21% -20.75% 10.89% 57.64% 135.42% 101.47% 59.67% 2.143** 7.506***

Low BM Firms (1,051 observations)Book to Market Ratio -1.70 -0.67 0.21 0.36 0.51 0.62 11.30 87.25%Assets 20,667,559 614,219 1,303,270 3,795,154 13,037,305 50,988,880 62,457,064 100.00%Sales 18,360,021 502,798 1,039,484 2,961,030 8,488,067 29,639,010 97,632,781 100.00%Operating Income 1,984,362 -181,865 12,280 221,083 891,950 3,233,419 8,708,431 77.26%Cash Flow from Operating Activities 2,442,109 -162,673 19,733 276,993 1,228,850 4,560,784 9,549,175 78.21%Return on Assets 7.02% -6.32% 0.83% 7.89% 14.54% 21.70% 14.77% 77.26%Sales Growth 27.89% -17.13% 2.18% 14.09% 30.88% 57.18% 145.11% 77.35%Earnings Growth 82.84% -118.37% -44.80% -6.39% 30.47% 104.44% 1129.75% 44.81%Proportion of Capital Expenditures to Assets 7.35% 0.48% 1.66% 4.74% 10.32% 7.35% 8.04% 99.33%One-Year Market-Adjusted Returns 1.03% -58.82% -33.10% -6.16% 24.03% 59.02% 58.95% 43.39%Two-Year Market-Adjusted Returns 6.35% -72.60% -40.98% -7.28% 32.54% 83.48% 81.97% 45.07%

17

TABLE 2 Correlations Analysis

The below table provides correlations between a binary score of each financial signal, both traditional and growth-oriented, the VSCORE, GSCORE, and future one-year MAR for 3,565 firms. Coefficients above diagonal are Spearman rank-order, and those below diagonal are Pearson. */**/*** represents statistical significance using two-tailed tests at 10% / 5% level.

1-YR MAR 2-YR MAR bROA b∆ROA bCFROA b∆CFROA b∆OPM b∆TATO b∆LEV b∆LIQ bEQOFF bVARROA bVARSG bCAPEX bADV VSCORE GSCORE

1-YR MAR 0.581** 0.054** 0.016 0.070** 0.076** -0.006 0.029 0.045** 0.087** 0.026 0.025 -0.005 0.020 0.000 0.089** 0.056**2-YR MAR 0.0326** 0.042* 0.026 0.060** 0.064** 0.002 0.058** 0.051** 0.043* 0.015 0.034* -0.010 -0.015 -0.082** 0.084** 0.033*bROA 0.016 0.004 0.173** 0.370** -0.128** 0.096** 0.146** 0.716** 0.113** 0.171** 0.284** 0.194** 0.147** 0.017 0.658** 0.565**b∆ROA 0.039* 0.048** 0.173** 0.061** -0.068** 0.138** 0.151** 0.141 0.027* 0.037** 0.381** 0.260** -0.075** 0.003 0.401** 0.560**bCFROA 0.030 0.010 0.370** 0.061** 0.356** 0.076** 0.236** 0.329** 0.110** 0.162** 0.149** 0.010 0.196** 0.034* 0.666** 0.375**b∆CFROA 0.041* 0.036 -0.128** -0.068** 0.356** 0.007 0.177** -0.123** 0.057** 0.011** -0.016 -0.117** 0.051** 0.005 0.291** 0.206**b∆OPM -0.019 0.020 0.096** 0.138** 0.076** 0.007 0.305** 0.073** -0.076** 0.044** 0.048** 0.013 -0.090** 0.002 0.413** 0.070**b∆TATO 0.017 0.021 0.146** 0.151** 0.236** 0.177** 0.305** 0.133** -0.034* 0.045** 0.096** -0.013** -0.083** 0.004 0.536** 0.153**b∆LEV 0.004 0.000 0.716** 0.141** 0.329** -0.123** 0.073** 0.133** 0.087** -0.073** 0.252** 0.147** 0.128** 0.027 0.564** 0.431**b∆LIQ 0.066** -0.012 0.113** 0.027 0.110** 0.057** -0.076** -0.034* 0.087** 0.035* 0.084** -0.026 0.005** 0.044** 0.230** 0.093**bEQOFF 0.007 0.004 0.171** 0.037* 0.162** 0.011 0.044** 0.045** -0.073** 0.035* 0.049** 0.107** 0.120** 0.033 0.343** 0.167**bVARROA 0.005 -0.008 0.284** 0.381** 0.149** -0.016 0.048** 0.096** 0.252** 0.084** 0.049** 0.198** -0.017 0.018 0.319** 0.619**bVARSG 0.009 0.002 0.194** 0.260** 0.010 -0.117** 0.013 -0.013 0.147** -0.026 0.107** 0.198** 0.112** 0.008 0.145** 0.556**bCAPEX 0.003 -0.018 0.147** -0.075 0.196** 0.051** -0.090** -0.083** 0.128** 0.005 0.120** -0.017 0.112** 0.038 0.100** 0.396**bADV -0.012 -0.048** 0.017 0.003 0.034* 0.005 0.002 0.004 0.027 0.044** 0.033 0.018 0.008 0.013 0.038* 0.158**VSCORE 0.043** 0.033* 0.651** 0.401** 0.657** 0.305** 0.420** 0.536** 0.560** 0.235** 0.353** 0.322** 0.147** 0.099** 0.038* 0.637**GSCORE 0.035* 0.014 0.562** 0.559** 0.381** 0.229** 0.071** 0.157** 0.427** 0.093** 0.170** 0.614** 0.553** 0.408** 0.172** 0.647**

18

TABLE 3 Returns to an Investment Strategy Based on VSCORE for All Sample Firms

VSCORE combines a binary score (0 or 1) of each of nine traditional fundamental signals. Beta, Return Volatility (RV), and debt-to-equity ratio (DE) are proxied for risk. Numbers of observations for 2-year MAR are fewer as we do not include returns for firms in 2008. t-statistic for the mean differences are from independent sample t-test. z-statistic are for median differences using wilcoxon ranked-sum test and can be interpreted similar to t-statistic. Extreme values of MAR at 1st and 99th percentile are eliminated. */**/*** represents statistical significance using two-tailed tests at 10% / 5% / 1% level. Panel A: One-Year Ahead Market-Adjusted Returns (1-YR MAR)

VSCORE N Mean 10% 25% Median 75% 90% % Positive Beta S.D. DE

0 8 -7.46% -41.19% -36.23% -27.89% 9.43% 56.83% 25.00% 0.76 0.69 1.85 1 82 3.35% -55.90% -34.76% -16.81% 16.23% 97.93% 36.59% 0.55 0.54 1.27 2 329 1.87% -62.33% -34.71% -10.15% 28.08% 68.08% 42.55% 0.51 0.51 0.85 3 486 10.20% -52.16% -28.78% -5.45% 29.51% 81.01% 45.68% 0.54 0.54 0.72 4 575 7.82% -51.50% -25.24% -2.45% 25.62% 66.96% 47.65% 0.46 0.48 0.59 5 606 7.69% -51.18% -30.51% -2.30% 30.71% 63.81% 48.18% 0.46 0.46 0.44 6 582 11.53% -40.19% -19.87% 2.98% 33.65% 66.82% 53.61% 0.42 0.42 0.43 7 454 12.03% -48.72% -22.82% 4.51% 35.28% 71.62% 54.41% 0.38 0.39 0.29 8 322 10.34% -47.38% -23.11% 2.06% 33.32% 65.82% 76.71% 0.39 0.39 0.20 9 135 15.87% -43.91% -19.10% 12.03% 42.96% 85.86% 57.78% 0.40 0.38 0.17

All 3,579 9.11% -50.55% -26.30% -0.91% 31.12% 71.66% 49.34% 0.44 0.45 0.47 High (8,9) 457 11.98% -47.00% -22.84% 4.08% 34.76% 76.16% 54.05% 0.39 0.39 0.19 Low (0,1,2) 419 1.98% -61.03% -34.92% -11.58% 27.43% 69.74% 41.05% 0.53 0.52 0.95

High-Low 10.00% 15.66% 13.00% -0.14 -0.13 -0.75 t-statistic / z-statistic 2.479*** 4.263*** -3.622*** -7.941*** -11.969***p-value 0.013 <0.000 <0.000 <0.000 <0.000

Panel B: Two-Year Ahead Market-Adjusted Returns (2-YR MAR)


0 8 -20.37% -63.14% -44.32% -16.33% 8.83% 13.53% 37.50% 0.83 0.99 1.85 1 75 0.87% -68.57% -45.47% -19.29% 11.06% 106.06% 36.00% 0.62 0.81 1.30 2 292 9.63% -66.55% -36.70% -4.51% 28.28% 95.19% 45.21% 0.49 0.80 0.86 3 434 17.72% -69.15% -36.58% -1.85% 44.44% 102.83% 48.62% 0.52 0.79 0.69 4 517 22.73% -60.83% -33.63% -0.02% 40.45% 113.47% 49.90% 0.47 0.72 0.61 5 528 21.80% -54.80% -27.89% 3.02% 45.46% 114.59% 53.22% 0.44 0.68 0.46 6 532 29.14% -49.22% -18.29% 9.81% 52.60% 119.52% 59.96% 0.40 0.61 0.43 7 398 30.81% -53.66% -22.86% 9.88% 52.02% 117.58% 60.30% 0.39 0.56 0.30 8 284 24.17% -55.27% -24.24% 12.38% 54.63% 109.25% 58.45% 0.41 0.58 0.20 9 112 48.81% -38.29% -11.07% 23.06% 79.64% 155.63% 68.75% 0.32 0.58 0.16

All 3,180 23.20% -57.95% -28.31% 4.38% 47.96% 114.29% 53.90% 0.44 0.67 0.49 High (8,9) 396 31.14% -50.05% -18.67% 14.70% 60.63% 128.96% 61.36% 0.37 0.58 0.20 Low (0,1,2) 375 7.23% -66.40% -40.29% -8.82% 27.56% 99.05% 43.20% 0.51 0.80 0.97

High-Low 23.90% 23.52% 18.16% -0.15 -0.22 -0.77 t-statistic / z-statistic 4.133*** 5.466*** -3.643*** -10.089*** -10.962***p-value <0.000 <0.000 <0.000 <0.000 <0.000

19

TABLE 4 Returns to an Investment Strategy Based on GSCORE for All Sample Firms

GSCORE combines a binary score (0 or 1) of each of seven traditional and growth-oriented fundamental signals. Beta, Return Volatility (RV), and debt-to-equity ratio (DE) are proxied for risk. Numbers of observations for 2-year MAR are fewer as we do not include returns for firms in 2008. t-statistic for the mean differences are from independent sample t-test. z-statistic are for median differences using wilcoxon ranked-sum test and can be interpreted similar to t-statistic. Extreme values of MAR at 1st and 99th percentile are eliminated. */**/*** represents statistical significance using two-tailed tests at 10% / 5% / 1% level. Panel A: One-Year Ahead Market-Adjusted Returns (1-YR MAR)

GSCORE N Mean 10% 25% Median 75% 90% % Positive Beta SD DE

0 124 5.70% -55.87% -36.34% -16.89% 27.57% 81.22% 37.10% 0.54 0.54 0.69 1 476 5.25% -58.40% -32.38% -4.15% 27.40% 71.18% 45.17% 0.50 0.51 0.67 2 766 9.00% -49.47% -25.86% -3.26% 28.81% 67.02% 47.65% 0.45 0.51 0.55 3 792 9.70% -49.66% -24.81% 0.17% 31.10% 69.39% 50.13% 0.43 0.45 0.52 4 735 7.28% -51.36% -26.57% -1.11% 29.51% 64.24% 48.57% 0.43 0.43 0.38 5 437 15.19% -45.91% -23.18% 4.55% 35.90% 78.29% 53.55% 0.42 0.39 0.37 6 269 14.49% -39.71% -15.39% 11.04% 34.92% 78.46% 59.85% 0.40 0.36 0.34 7 14 15.55% -24.23% -15.06% 4.39% 38.50% 63.48% 50.00% 0.61 0.37 0.51

All 3,613 9.38% -50.36% -26.27% -0.99% 31.14% 71.71% 49.32% 0.44 0.45 0.48 High (5,6,7) 720 14.94% -43.58% -21.13% 7.30% 35.76% 78.35% 55.83% 0.42 0.38 0.36 Low (0,1) 600 5.34% -58.39% -34.52% -6.16% 27.43% 72.54% 43.50% 0.52 0.51 0.67

High-Low 9.59% 13.46% 12.33% -0.11 -0.13 -0.31 t-statistic / z-statistic 2.774*** 4.891*** -2.372** -8.834*** -6.123***p-value 0.006 <0.000 0.018 <0.000 <0.000



0 116 13.81% -67.72% -37.55% -12.23% 29.71% 116.62% 43.10% 0.55 0.79 0.64 1 427 13.16% -61.86% -36.35% -2.32% 37.90% 91.37% 48.71% 0.50 0.77 0.68 2 696 22.79% -63.23% -30.81% 0.58% 41.96% 113.77% 50.57% 0.44 0.75 0.55 3 711 28.12% -56.08% -27.28% 6.06% 50.17% 125.33% 54.29% 0.42 0.68 0.54 4 642 19.98% -60.14% -27.51% 5.69% 44.90% 107.39% 55.92% 0.45 0.62 0.39 5 382 33.50% -45.70% -20.55% 11.74% 60.53% 132.73% 60.47% 0.43 0.58 0.39 6 236 24.33% -44.77% -22.83% 13.26% 53.14% 113.97% 59.32% 0.37 0.54 0.36 7 9 76.83% -17.73% -16.50% 55.13% 101.87% 189.15% 66.67% 0.82 0.52 0.68

All 3,219 23.34% -58.07% -28.46% 4.15% 48.04% 115.01% 53.81% 0.44 0.67 0.49 High (5,6,7) 627 30.67% -45.09% -20.94% 12.50% 59.52% 126.94% 60.13% 0.42 0.56 0.38 Low (0,1) 543 13.30% -62.62% -36.81% -4.19% 34.31% 95.28% 47.51% 0.50 0.78 0.67

High-Low 17.37% 16.69% 12.61% -0.08 -0.22 -0.30 t-statistic / z-statistic 3.506*** 5.202*** -2.154** -10.661*** -5.484***p-value <0.000 <0.000 0.031 <0.000 <0.000

20

TABLE 5 Returns to an Investment Strategy Based on VSCORE for High BM Firms

VSCORE combines a binary score (0 or 1) of each of nine traditional fundamental signals. Beta, Return Volatility (RV), and debt-to-equity ratio (DE) are proxied for risk. Numbers of observations for 2-year MAR are fewer as we do not include returns for firms in 2008. t-statistic for the mean differences are from independent sample t-test. z-statistic are for median differences using wilcoxon ranked-sum test and can be interpreted similar to t-statistic. Extreme values of MAR at 1st and 99th percentile are eliminated. */**/*** represents statistical significance using two-tailed tests at 10% / 5% / 1% level. Panel A: One-Year Ahead Market-Adjusted Returns (1-YR MAR)


0 4 -10.90% -36.27% -33.34% -27.89% -5.45% 28.07% 25.00% 0.82 0.64 3.28 1 35 12.63% -61.64% -34.68% -5.61% 29.61% 130.48% 45.71% 0.41 0.52 0.86 2 133 3.98% -54.39% -26.63% -2.24% 30.51% 63.09% 48.12% 0.47 0.50 0.60 3 183 15.14% -51.09% -27.07% 2.98% 33.47% 91.04% 51.91% 0.47 0.57 0.65 4 188 20.02% -38.00% -15.97% 4.66% 41.45% 85.41% 58.51% 0.38 0.51 0.51 5 198 17.03% -52.30% -25.43% 1.65% 41.41% 88.32% 52.53% 0.35 0.50 0.33 6 151 17.20% -29.75% -13.43% 12.03% 40.10% 69.49% 64.90% 0.32 0.51 0.38 7 88 17.96% -34.52% -16.04% 4.11% 40.05% 70.79% 54.55% 0.27 0.46 0.43 8 63 22.81% -36.76% -14.94% 4.99% 34.55% 91.99% 76.19% 0.33 0.49 0.22 9 32 26.61% -35.28% -7.50% 22.50% 68.24% 85.98% 65.63% 0.24 0.53 0.24

All 1,075 16.09% -43.83% -21.74% 4.24% 36.75% 81.36% 55.26% 0.38 0.51 0.46 High (8,9) 95 24.09% -36.74% -11.71% 12.97% 40.82% 91.31% 61.05% 0.31 0.50 0.22 Low (0,1,2) 172 5.39% -54.52% -28.81% -3.82% 30.36% 71.38% 47.09% 0.44 0.51 0.65

High-Low 18.70% 16.78% 13.96% -0.13 -0.01 -0.42 t-statistic / z-statistic 2.488** 2.632*** -2.470** -0.722 -5.677***p-value 0.013 0.008 0.014 0.470 <0.000



0 4 -6.84% -29.67% -16.93% -1.26% 8.83% 11.54% 50.00% 1.15 0.91 3.28 1 28 21.37% -56.54% -34.49% 1.15% 48.01% 119.94% 53.57% 0.44 0.78 0.81 2 119 18.95% -51.89% -26.17% 6.10% 55.27% 97.68% 55.46% 0.39 0.77 0.60 3 164 26.51% -51.62% -23.75% 3.54% 41.51% 122.88% 54.27% 0.45 0.83 0.64 4 172 35.71% -54.93% -22.13% 7.87% 55.03% 128.13% 58.14% 0.37 0.73 0.55 5 175 38.23% -45.85% -20.65% 15.57% 52.00% 145.34% 60.57% 0.38 0.74 0.35 6 137 40.38% -42.05% -15.49% 15.96% 65.34% 147.33% 64.96% 0.31 0.69 0.37 7 80 50.50% -46.77% -19.87% 15.61% 58.80% 131.22% 63.75% 0.27 0.64 0.41 8 54 44.51% -34.18% -14.02% 34.83% 91.96% 145.90% 64.81% 0.31 0.68 0.26 9 26 62.22% -48.82% -5.16% 43.44% 100.41% 183.52% 73.08% 0.20 0.83 0.19

All 959 35.04% -49.18% -20.72% 10.89% 57.82% 135.38% 59.65% 0.36 0.75 0.47 High (8,9) 80 50.27% -45.54% -12.63% 35.11% 92.88% 164.99% 67.50% 0.26 0.71 0.23 Low (0,1,2) 151 18.71% -53.27% -27.13% 3.89% 54.82% 110.40% 54.97% 0.41 0.77 0.66

High-Low 31.56% 31.22% 12.53% -0.15 -0.06 -0.43 t-statistic / z-statistic 2.707*** 2.835*** -2.671*** -1.753** -5.239***p-value 0.008 0.005 0.008 0.080 <0.000

21

TABLE 6 Returns to an Investment Strategy Based on GSCORE for High BM Firms



0 56 14.28% -53.14% -35.46% -9.00% 55.32% 115.00% 46.43% 0.51 0.53 0.49 1 207 14.73% -51.26% -28.79% 2.98% 34.78% 88.11% 53.14% 0.44 0.52 0.63 2 285 16.78% -37.73% -17.34% 6.61% 39.33% 73.28% 57.19% 0.38 0.53 0.40 3 244 13.16% -46.18% -21.36% 3.00% 34.86% 78.16% 54.10% 0.35 0.50 0.50 4 170 13.17% -44.39% -25.90% 2.49% 32.37% 76.76% 52.35% 0.37 0.49 0.47 5 88 24.11% -40.38% -16.42% 13.12% 45.83% 101.56% 61.36% 0.32 0.52 0.34 6 30 32.56% -8.63% 5.32% 25.54% 52.83% 83.94% 80.00% 0.27 0.49 0.35 7 3 19.19% -9.82% -7.32% -3.16% 34.52% 57.13% 33.33% 0.29 0.41 0.21

All 1,083 15.92% -43.85% -21.80% 4.30% 36.38% 80.92% 55.31% 0.38 0.51 0.47 High (5,6,7) 121 26.08% -36.68% -8.55% 18.49% 47.89% 94.29% 65.29% 0.31 0.50 0.34 Low (0,1) 263 14.64% -51.37% -30.73% 1.64% 35.11% 89.26% 51.71% 0.45 0.52 0.60

High-Low 11.44% 16.85% 13.58% -0.14 -0.03 -0.26 t-statistic / z-statistic 1.530 3.016*** -2.909*** -1.199 -3.336***p-value 0.127 0.003 0.004 0.230 0.001



0 52 20.84% -49.17% -20.97% -1.28% 38.13% 134.00% 48.08% 0.50 0.76 0.49 1 187 22.94% -56.20% -24.79% 4.06% 48.92% 92.32% 56.68% 0.40 0.79 0.67 2 256 35.50% -50.54% -20.83% 9.24% 54.33% 140.06% 60.55% 0.37 0.78 0.39 3 225 42.87% -46.17% -19.47% 11.28% 63.41% 160.34% 60.44% 0.33 0.74 0.51 4 142 33.44% -55.02% -23.21% 10.90% 57.20% 125.29% 57.75% 0.37 0.67 0.45 5 81 41.50% -41.71% -18.76% 21.97% 85.22% 149.06% 64.20% 0.30 0.75 0.35 6 25 50.46% -21.87% 9.13% 35.18% 92.36% 118.63% 76.00% 0.19 0.62 0.33 7 2 194.25% 82.95% 124.69% 194.25% 263.82% 305.55% 100.00% 0.31 0.96 0.61

All 970 34.91% -49.84% -20.97% 10.85% 57.91% 135.78% 59.48% 0.37 0.75 0.49 High (5,6,7) 108 46.40% -28.95% -14.53% 32.42% 90.80% 141.67% 67.59% 0.27 0.73 0.34 Low (0,1) 239 22.48% -53.46% -24.79% 3.38% 47.29% 96.84% 54.81% 0.43 0.78 0.61

High-Low 23.92% 29.04% 12.78% -0.17 -0.05 -0.27 t-statistic / z-statistic 2.479** 3.096*** -3.515*** -1.584 -3.712***p-value 0.014 0.002 <0.000 0.113 <0.000

22

TABLE 7 Returns to an Investment Strategy Based on VSCORE for Low BM Firms

VSCORE combines a binary score (0 or 1) of each of nine traditional fundamental signals. Beta, Return Volatility (RV), and debt-to-equity ratio (DE) are proxied for risk. Numbers of observations for 2-year MAR are fewer as we do not include returns for firms in 2008. t-statistic for the mean differences are from independent sample t-test. z-statistic are for median differences using wilcoxon ranked-sum test and can be interpreted similar to t-statistic. Extreme values of MAR at 1st and 99th percentile are eliminated. */**/*** represents statistical significance using two-tailed tests at 10% / 5% / 1% level. Panel A: One-Year Ahead Market-Adjusted Returns (1-year MAR)


0 1 -48.11% -48.11% -48.11% -48.11% -48.11% -48.11% 0.00% 0.39 0.50 0.97 1 23 5.92% -52.01% -35.60% -22.46% 23.77% 101.68% 34.78% 0.76 0.58 1.47 2 84 -15.74% -75.16% -47.63% -22.72% 11.56% 45.16% 33.33% 0.49 0.53 1.23 3 138 -0.03% -58.64% -36.30% -9.75% 15.70% 67.90% 36.96% 0.55 0.54 0.66 4 143 2.18% -63.26% -38.03% -9.90% 17.99% 58.49% 40.56% 0.59 0.48 0.63 5 155 -3.07% -51.18% -35.03% -8.29% 21.19% 53.44% 41.94% 0.65 0.47 0.56 6 192 7.02% -47.65% -24.94% -4.98% 29.74% 66.55% 45.31% 0.59 0.39 0.46 7 156 -1.82% -71.75% -36.69% 0.02% 20.59% 53.30% 50.00% 0.60 0.36 0.23 8 117 4.56% -53.43% -27.51% -3.05% 28.96% 52.56% 66.67% 0.59 0.43 0.24 9 46 7.45% -52.36% -25.11% 9.95% 34.20% 75.82% 58.70% 0.42 0.32 0.09

All 1,055 0.51% -59.49% -33.62% -6.16% 23.91% 58.88% 43.32% 0.58 0.42 0.48 High (8,9) 163 5.38% -52.71% -26.97% 0.92% 30.94% 58.65% 50.31% 0.52 0.38 0.19 Low (0,1,2) 108 -11.43% -73.02% -46.08% -22.72% 13.53% 49.03% 33.33% 0.53 0.54 1.23

High-Low 16.81% 23.64% 16.97% -0.01 -0.16 -1.04 t-statistic / z-statistic 2.472** 3.210*** -0.266 -4.995*** -5.841***p-value 0.014 0.001 0.790 <0.000 <0.000

Panel B: Two-Year Ahead Market-Adjusted Returns (2-year MAR)


0 1 -62.78% -62.78% -62.78% -62.78% -62.78% -62.78% 0.00% 0.50 0.79 0.97 1 23 -18.08% -71.62% -52.35% -40.79% -1.69% 53.28% 26.09% 0.84 0.85 1.47 2 76 -4.10% -84.84% -50.31% -17.90% 10.50% 80.99% 34.21% 0.55 0.85 1.15 3 128 6.52% -81.79% -53.25% -17.30% 43.03% 80.39% 39.84% 0.60 0.82 0.60 4 132 3.96% -78.67% -47.75% -14.93% 26.75% 86.72% 37.88% 0.64 0.74 0.63 5 136 7.81% -61.00% -37.72% -1.42% 40.10% 81.42% 49.26% 0.72 0.73 0.65 6 175 16.67% -67.45% -28.26% 2.00% 31.62% 110.75% 52.00% 0.54 0.62 0.48 7 129 1.03% -68.98% -37.25% -2.19% 37.34% 69.02% 48.06% 0.69 0.51 0.26 8 102 3.64% -60.68% -35.32% -4.25% 32.39% 60.99% 45.10% 0.70 0.57 0.32 9 35 23.90% -41.87% -13.00% 13.40% 54.15% 93.91% 65.71% 0.45 0.55 0.11

All 937 6.28% -72.59% -41.00% -7.57% 32.50% 83.44% 45.04% 0.65 0.65 0.52 High (8,9) 137 8.82% -56.83% -30.85% 2.89% 33.25% 72.70% 50.36% 0.61 0.57 0.21 Low (0,1,2) 100 -7.90% -78.02% -53.96% -20.50% 8.67% 77.55% 32.00% 0.67 0.84 1.18

High-Low 16.72% 23.40% 18.36% -0.06 -0.27 -0.97 t-statistic / z-statistic 1.738* 3.203*** -0.457 -6.600*** -4.726***p-value 0.083 0.001 0.648 <0.000 <0.000

23

TABLE 8 Returns to an Investment Strategy Based on GSCORE for Low BM Firms


GSCORE N Mean 10% 25% Median 75% 90% % Positive Beta S.D. DE

0 33 -5.20% -60.44% -49.89% -24.07% 2.91% 77.63% 27.27% 0.39 0.57 0.95 1 92 -10.28% -72.17% -47.37% -15.09% 6.26% 38.51% 31.52% 0.50 0.52 0.61 2 188 0.36% -63.88% -33.16% -8.90% 13.57% 57.02% 39.89% 0.53 0.52 0.59 3 222 2.27% -59.18% -34.08% -6.26% 26.93% 60.41% 42.34% 0.64 0.45 0.62 4 233 -2.47% -57.19% -35.99% -10.95% 22.10% 51.09% 41.63% 0.59 0.42 0.34 5 162 8.66% -44.39% -25.53% 0.39% 31.38% 70.41% 50.62% 0.58 0.36 0.32 6 128 5.61% -53.56% -25.05% 3.19% 32.07% 68.18% 53.91% 0.56 0.35 0.45 7 9 17.57% -23.25% -16.25% 11.95% 31.32% 57.17% 55.56% 0.86 0.36 0.68

All 1,067 1.09% -59.00% -33.62% -6.16% 23.91% 59.04% 43.11% 0.58 0.42 0.48 High (5,6,7) 299 7.62% -46.55% -24.77% 1.71% 31.84% 70.65% 52.17% 0.58 0.35 0.40 Low (0,1) 125 -8.94% -70.07% -48.11% -12.02% 5.88% 51.18% 30.40% 0.48 0.56 0.69

High-Low 16.57% 13.73% 21.77% 0.10 -0.21 -0.29 t-statistic / z-statistic 2.707*** 4.577*** 1.852* -3.964*** -0.916p-value 0.007 <0.000 0.064 <0.000 0.360


GSCORE N Mean 10% 25% Median 75% 90% % Positive Beta S.D. DE

0 31 11.86% -72.89% -53.15% -25.90% 11.05% 112.67% 32.26% 0.47 1.05 0.93 1 82 4.09% -76.31% -50.20% -24.86% 25.67% 87.22% 40.24% 0.64 0.84 0.59 2 173 8.05% -80.44% -46.67% -10.82% 29.28% 85.26% 39.88% 0.63 0.76 0.58 3 197 13.73% -72.96% -38.55% -4.15% 44.53% 89.68% 46.70% 0.64 0.70 0.65 4 211 -3.69% -70.15% -41.65% -11.37% 22.66% 57.76% 42.65% 0.69 0.66 0.34 5 140 17.96% -50.15% -28.10% 6.14% 47.54% 90.86% 55.00% 0.63 0.55 0.37 6 109 8.56% -59.21% -30.12% 0.88% 38.55% 85.50% 50.46% 0.64 0.51 0.52 7 6 33.51% -18.52% -17.03% -1.11% 68.98% 120.16% 50.00% 0.84 0.53 0.80

All 949 8.09% -72.58% -40.83% -7.00% 32.88% 84.66% 45.21% 0.64 0.66 0.51 High (5,6,7) 255 14.31% -54.97% -28.71% 4.48% 41.39% 91.69% 52.94% 0.65 0.54 0.45 Low (0,1) 113 6.29% -76.14% -52.21% -25.69% 20.27% 110.35% 38.05% 0.56 0.90 0.59

High-Low 8.01% 30.17% 14.89% 0.09 -0.36 -0.14 t-statistic / z-statistic 0.666 3.622*** 1.581 -6.623*** -0.056p-value 0.506 <0.000 0.114 <0.000 0.955

Documents

Can Historical Accounting Information Be Used to …¹€อกสาร ADV 30...1 Can Historical Accounting Information Be Used to Predict Future Stock Returns? Pornthep Tantipanichkul