MSc in Finance and Financial Information Systems. Dissertation EARNINGS RESPONSE COEFFICIENTS IN THE GREEK MARKET

Transcription

1 MSc in Finance and Financial Information Systems Dissertation EARNINGS RESPONSE COEFFICIENTS IN THE GREEK MARKET by Karakoltsidis Nikolaos Supervisor: Dr Dimitrios I. Maditinos April 2010

2 ACKNOWLEDGMENTS I wish to express my sincere gratitude to Professor Dimitrios Maditinos, my supervisor, for his unwavering guidance, patience and endless support throughout the writing of the present dissertation. I also wish to thank Professor Dimitrios Kousenidis for his valuable comments and guidance during the writing of the dissertation. I would like to extent my gratitude to the academic and administration staff of Technological Educational Institute (T.E.I.) of Kavala, all the people who helped me along the way with my M.Sc. studies in the field of Finance and Financial Information Systems. Special thanks is reserved for my parents. Thank you both for your love, support and encouragement over the years of my studies. 2

3 ABSTRACT The relationship between stock prices or returns and earnings has been a fundamental issue in accounting and finance research since three decades. The present dissertation attempts to discover the relationship between accounting data and market price returns of the companies listed on the Athens Stock Exchange (ASE). To achieve it, the relevant literature was studied and publicly available financial data of the listed companies in the ASE during was collected and analysed. The data sample consists of 245 companies and varies from 2166 to 1441 firm-year observations. The research methodology was based on Kothari and Sloan (1992) and was investigated whether the level of earnings divided by price at the beginning of the stock return period is associated with returns in the context of prices lead earnings using annual and quarterly data. The primary model stimulated the present dissertation relies on the idea that stock return over a period reflects the market s revision in expectation of future earnings. Accounting earnings over the same period, however, do not. Kothari and Sloan (1992) presented a method that uses leading period returns to control for the biased coefficients of return models. Kothari and Sloan extended the return measurement window to several years before the fiscal year of interest, and this resulted in higher earnings response coefficients. This dissertation employs their model in order to test if there is statistically significant relationship between returns and accounting data on the Greek market. Cross-sectional regression analysis indicates that there is a significant relationship between earnings and returns for measurement windows of one year and longer. Similar results found in the case of cumulative model where earnings are aggregated up to four years, however, the relationship in the short measurement window up to three quarters resulted in low earnings response coefficients.. Keywords: Earnings response coefficients, Kothari and Sloan (1992), Stock returns, Athens Stock Exchange. 3

4 TABLE OF CONTENTS Chapter 1 Introduction Introduction Objectives of the Dissertation Outline of the Dissertation...7 CHAPTER 2 Literature Review Introduction Basic Concepts Results of other studies Determinants of earnings response coefficient The use of coefficient of variation and the low explanatory power Return and price models Length of the response window Empirical results of earnings response coefficient Introduction Results of alternative studies Results of European stock markets Results of Non-European stock markets Results of Greek stock market Summary...25 CHAPTER 3 Research Methodology Introduction Other studies Methodology Specification Summary...30 CHAPTER 4 Empirical Results Introduction Description of the data

5 4.3 Descriptive estimates First Model Results Intertemporal Results Annual Results Intertemporal Quarterly Results Quarterly Results Cumulative Model Results Intertemporal Cumulative Model Results Annual Cumulative Model Results Summary...47 Chapter 5 Concluding Remarks and Recommendations Introduction Summary of the research Suggestions for further research...49 References

6 Chapter 1 Introduction 1.1 Introduction A fundamental issue of economics, finance, and accounting involves the relation between a firm's reported earnings and its stock returns (Kormendi & Lipe, 1987). Standard valuation models assume that price is the discount present value of future expected dividends or future cash flows. It is commonly assumed that, over long periods, reported accounting earnings are directly related to futures dividends and cash flows. Since Ball and Brown (1968), numerous studies have been trying to identify whether reported earnings contain information used by the market in assessing the value of a firm's common stock. Later studies explicitly associated the response of stock returns to earnings by the introduction of the earnings response coefficient (ERC). Earnings response coefficient studies examine the explicit relationship between prices and earnings as implied by finance valuation models. In general, empirical studies concluded that information provided by accounting earnings is relevant to valuation. Therefore, given that earnings contain useful information, it is important to investigate what is the economic nature of the information in reported earnings, and how does it relate to firm valuation. There are lots of studies that suggest, and empirically test, that stock prices lead earnings. That is, prices contain information about future earnings that is not reflected in the past timeseries of earnings. This dissertation attempts to test this relationship in the context of Greek market. 1.2 Objectives of the Dissertation The general objective of this dissertation is to analyse the relationship between accounting data and market price returns of the companies listed on Athens Stock Exchange (ASE). More specific, this study analyses if there is statistically significant earnings response coefficient of companies of ASE using individual year cross-sectional and intertemporal regression analysis. The theoretical platform is based on the previous studies of Kothari and Sloan (1992) and Jindrichovska (2001). The major findings of this 6

7 dissertation may contribute to a variety group of people such as investors, corporations, regulators, educators and researchers. 1.3 Outline of the Dissertation The present dissertation is organised as follows: The next chapter (chapter 2) consists of a literature review discussing various relevant issues in earnings response coefficient research. The purpose of this is to provide the basic theoretical and a detailed review of the earnings response coefficient (ERC). Moreover, it presents empirical foundation for the dissertation of other studies and their implications. Chapter 3 examines research design issues and provides the research method with a detailed discussion of the model employed. Chapter 4 contains the description of the data and the core results from the statistical analysis, with a focus on answering the research hypothesis developed in previous chapter. Chapter 5 discusses and summarises the findings of the study including limitation of the results and suggestions for future research. 7

8 CHAPTER 2 Literature Review 2.1 Introduction This chapter reviews several areas of the earnings response coefficient literature and present studies and empirical results of this field. The chapter is concluded with a summary of the main issues raised in the chapter. 2.2 Basic Concepts According to Maggina (2009) common interest of both accounting and finance scholars and users of financial statements (primarily, financial analysts and investors) is to obtain concrete and increased knowledge of the association between accounting earnings and stock prices for a more or less predictive ability explanation. The efficient market hypothesis (EMH) is commonly used to base accounting studies regarding earnings prices associations. The economics literature argues that, in a free market economy with perfect competition, price is determined by (1) the availability of the product (supply) and (2) the desire to possess this product (demand), then, the price of a product/asset is determined by a market equilibrium based on the purchasers knowledge of relevant information about a product/asset. However, in the security markets, two issues are involved among others: the information about a company that is valuable to an investor and the form of corporate disclosure and its understandability. Based on these two issues, three separate forms of the efficient market hypotheses were developed: the weak form, the semi-strong form and the strong form. Consequently, the efficient market hypothesis has implications for the development of accounting theory and practice. Some critics of accounting have argued that the lack of uniformity in accounting principles has allowed corporate managers to manipulate earnings and mislead investors (Ball and Brown, 1968). This argument is based on the assumption that accounting reports are the only sources of information on a business organization. The results of efficient market hypothesis research suggest that stock prices are not determined solely by accounting reports. This conclusion has led researches to investigate how accounting earnings are related to stock prices. 8

9 The results of these investigations imply that accounting earnings are correlated with securities returns. Other accounting research relies on research findings that support the efficient market hypothesis to test market perceptions of accounting numbers and financial disclosures. This research is rested on the premise that an efficient market implies that the market price of a firm reflects the consensus of investors regarding the value of the firm. Thus, if accounting information and other financial disclosure items that affect firm value, then they should be reflected in firms security prices. In the late 1980s, researchers started to investigate a new area, the earnings response coefficient (ERC). Theoretically, the earnings response coefficient is defined as "the price change induced by a one-dollar change in current earnings" (Collins and Kothari, 1989), and typically measured as a slope coefficient in a regression of stock return on unexpected earnings. While the common earnings/price studies concentrate on market reactions to earnings announcements, the earnings response coefficient studies are more interested in the nature of the information in reported earnings, and how it relates to firm valuation (Kormendi and Lipe, 1987). The commonly estimated univariate returns / earnings regression models generally have the form UR it = a +bux it +e it (2.1) where UR it is the unexpected return, UX it represents the unexpected earnings, and e it is a random disturbance term assumed to be normally distributed with a zero mean and variance σ 2. The coefficient b refers to the ERC that is, the increment to the stock price of firm i caused by a unit increase in its earnings. The constant term is assumed to be stationary over time. The strength of the price earnings relation (or the usefulness of earnings information to investors) is measured in terms of the magnitude of the estimated ERC and the correlation between returns and earnings. Expected return can be estimated in several ways, for instance by using the market model or the Fama and French threefactor model (Fama & French, 1992, 1993). The unexpected stock return is regressed on unexpected earnings. Unexpected earnings are the difference between total earnings and a measure of expected earnings. Expected earnings can for instance be calculated 9

10 from analysts forecasts (Easton & Zmijewski, 1989; Freeman & Tse, 1992) or from timeseries models of earnings (Ahmed, 1994; Kormendi & Lipe, 1987). 2.3 Results of other studies There are many approaches of how accounting data affects market price returns. Economic theory has given too much attention in this subject and there are many useful results by previous theoretical and empirical research. As the purpose of this dissertation is to investigate the relationship of accounting data on market price return in the Greek market it is necessary to be represented the work of several researchers at the past. Firstly, Kothari in 2001 represented a review empirical research on the relationship between capital markets and financial statements. He uses an economics-based framework of demand for and supply of capital markets research in accounting to organise the paper. The main sources of demand for capital markets research are fundamental analysis and valuation, tests of market efficiency, the role of accounting in contracts and in the political process, and disclosure regulation. In summarizing past research, he critiques existing research as well as discuss unresolved issues and directions for future research. In addition, he offers a historical perspective of the genesis of important ideas in the accounting literature, which have greatly influenced future accounting thought in the area of capital markets research. Among the various treatments of the relation between accounting earnings and stock prices, the seminal work of Ball and Brown (1968) is precedent. They documented a positive statistical association between earnings surprises' and stock returns around earnings announcements. They concluded that accounting earnings revealed 'useful' information to the market. However, the question of how accounting earnings give useful information is still being understood. Results also show that the usefulness of accounting earnings is mitigated by information contained in current dividends. Since Ball and Brown (1968), a voluminous body of research has examined the role of accounting earnings in financial markets. Their work indicates that accounting earnings and some of its components exhibit an information content drawn through stock prices. Earnings surprises are defined as actual earnings minus earnings expected by the market. Ball and Brown used the change in earnings and the change in earnings after removing the effect of the change in a market index of earnings as proxies for earnings surprise. Other 10

11 studies also use explicit theoretical models. A widely accepted model (Beaver, 1989) uses three links to describe the relationship between current earnings and future dividend-paying ability. The three links are: (1) a time-series link that relates current accounting earnings to future earnings; (2) a dividend-earnings link that relates future earnings to future dividends; and (3) a valuation-link that relates stock price to future dividends. These three links imply that current accounting earnings (unexpected earnings) are related to stock price (stock returns). They also imply that the price earnings (or the returns-unexpected earnings) relation will vary cross-sectionally according to the time-series behavior of earnings. In influential articles, Easton and Zmijewski (1989), Collins and Kothari (1989), and Kormendi and Lipe (1987) empirically test the last implication. They document that the price-earnings relation varies crosssectionally according to the time-series behavior of accounting earnings. Specifically, they show that slope coefficients in the price-earnings regression are positively related to the time-series parameter of accounting earnings. Their finding provides interesting insights on the type of information contained in accounting earnings. It has also paved the way for a large volume of research analysing earnings response coefficients. After these studies a large body of research has been analyzing the market reaction to accounting data, and a formal theory regarding this relation was first developed by Ohlson (1995) Determinants of earnings response coefficient There are many studies that describe the relationship between earnings and stock returns. However, it is not possible to give a general answer to how sensitive stock returns are to earnings or changes in earnings. This sensitivity, the ERC, depends on many factors. According to Brown (1994) there are many factors effect the magnitude of the ERC such as the cost of capital, leverage, growth opportunities, the risk free rate of interest, systematic risk of firm assets, operating risk, earnings persistence, earnings predictability, firm size, industry classification, and the level of uncertainty. He implied that each of these factors effects the determination of the ERC and that each should be taken into consideration when calculating it. Collins and Kothari (1989) examined persistence, growth and the risk free rate of interest as determinants of the ERC. Persistence has been found to affect the ERC in a positive 11

12 way; the more persistent the earnings, the more responsive (larger change) prices are, and therefore the larger the ERC. As far it concerns growth, they argued that current growth opportunities are probable to be weakly reflected in current earnings and hence the degree to which current earnings provide information in respect to growth opportunities, the greater the expected change in price. The significance of the risk free rate relates to its contribution to the discount rate. When the risk free rate increases, so too does the discount rate, consequently, reducing the discounted value of future earnings and therefore reducing the ERC. Another factor that has been argued to be a determinant of the ERC is risk. According to Easton and Zmijewski (1989) as systematic risk or volatility increases, so does the equity discount rate. As a result, the greater the risk, the lower the present value of expected earnings and the lower the ERC. Cho and Jung (1991) mentioned a variety of determinants for ERCs consisting of earnings persistence, beta risk values, risk free interest rates, and industry effects such as barriers to entry and financial leverage. Furthermore, they found that factors such as growth, earnings predictability, firm size, industry effects such as product type, operating leverage, and growth did not effect the determination of the ERC. Likewise, Jeter and Chaney (1992) hypothesised that industry classification, systematic risk, leverage, profitability, and earnings volatility were the main factors that influence the ERC. Though, they discovered that only profitability, industry classification, and leverage could be statistically related to the magnitude of the ERC. Thus, there are various factors that have been shown to influence the magnitude of ERCs and these issues should be considered when planning research in this area The use of coefficient of variation and the low explanatory power In regression analysis, the coefficient of variation (the explanatory power or simply R 2 ) is a measure of the proportion of the variance in the dependent variable explained by the independent variable(s). If stock price or returns are regressed on accounting variables, R 2 is a measure of how much of the variation in stock prices/returns is explained by the accounting variables analysed. Thus, explanatory power is a measure of value relevance. The explanatory power from different samples is often compared to study if value relevance differs between the samples. (Beisland, 2008) 12

13 In financial accounting research, there are many studies that examine the relation between accounting earnings and stock prices. A common finding of these studies is that the estimated relationship between stock returns and accounting earnings remains surprisingly low (see, e.g., Lev, 1989). According to Kothari (2001) at least four hypotheses explain the observed low magnitudes of earnings response coefficients: (a) prices lead earnings, (b) inefficient capital markets, (c) noise in earnings and deficient GAAP and (d) transitory earnings. (a) Prices lead earnings. In an important study, Beaver et al. (1980) developed the idea that the information set reflected in prices is richer than that in the past time series of accounting earnings. They derived a relationship between changes in share prices and changes in earnings by expanding the information upon which earnings are conditioned to include variables other than the prior earnings history. They recognized that, in general, the assessed distribution of future earnings conditional upon past earnings will differ from the assessed distribution of future earnings conditional upon past earnings and past prices. This will occur if prices convey information about future earnings that is not conveyed by the past earnings. They characterised earnings as a combination of two processes: one which reflects the impact of earnings on events that also affect prices, whilst, the other (called garbling ) deems the impact of earnings on events with no effect on security prices. In this fashion, changes in prices are dependent on changes in expectations concerning future earnings and consequently, share prices may incorporate information about future earnings not reflected in the past time series of earnings. Kothari (1992) and Kothari and Sloan (1992) also examined the strength of the priceearnings relationship. On the contrary to previous studies where the price-earnings regression in levels includes prices and earnings, they deflated the earnings by the beginning-of-the year share price. They explored the price-earnings regressions when prices lead earnings. That is, prices contain information about future earnings that is not reflected in the past time-series of earnings. They found that by using the earnings forecasts incorporated in the security prices, they were able to reduce the measurement error bias caused by a noisy proxy for the market expectation of future earnings and also attained a better explanatory power of the price-earnings regression. 13

14 According to Kothari (2001) one inference of prices leading earnings is that although annual earnings time-series properties are rationally described as a random walk and therefore successive earnings changes are unpredictable using the information in past time series of earnings, the information set reflected in prices include information about future earnings changes. That is, from the viewpoint of the market, successive annual earnings changes are not unpredictable. (b) Inefficient capital markets The second issue relates to inefficient capital markets. The value relevance literature assumes some form of informational efficiency within the market in order to assess the impact that this information has. Nevertheless, if the market fails to correctly interpret and act upon this information, or act in an untimely manner, then this will reduce the strength of the ERC and the associated explanatory power of the regression model. This has been rather overcome by the use of long window association studies such as Easton et al. (1992). (c) Noise in earnings and deficient GAAP Another reason that may lead to low ERC s is noise in earnings and poor quality earnings because of deficient accounting standards. This argument implies that the accounting earnings figure is not the true earnings of a given firm due to issues such as discretionary accounting policy choice, accounting adjustments, and so on. Therefore, the earnings figure includes a noise component that biases the ERC downward (Beaver et al., 1980). According to Foster (1986) there are some reasons for the timing difference between the concept of earnings implicit in security price and the concept of earnings implicit in GAAP. First, accounting earnings are relied on accounting notions such as realization and conservatism. With these notions, there are restrictions on both the scope and the timing with which events are reflected in the reported earnings. Second, the time focus of GAAP earnings series is for a past period, while the time focus of the earnings concept implicit in security prices is for a future period. This difference can cause two types of divergence. First, past earnings are reflected in reported generally accepted accounting principles earnings but are not expected to occur in the future, and second, events that did not occur in the past but are expected to occur in the future (p.438) 14

15 (d) Transitory earnings Getting into the field of transitory earnings, we should first explain the meaning of it. The assumption of transitory earnings means that the earnings change is expected to be non-permanent, which is a departure from a random walk assumption (Luberrink, 2000). There is wide literature documenting smaller earnings response coefficients on transitional earnings as proxied for by non-recurring items reported in financial statements (Hayn 1995, Ramakrishnan & Thomas 1998, etc.). The interpretation is that markets do not expect extreme earnings changes (positive and negative) to be permanent, so price adjustments are smaller. As a result, there is a negative correlation between the absolute magnitude of the earnings change and the possibility that it is permanent. That means a non linear, S-shaped relation between stock returns and accounting earnings. According to Easton et al. (2000) the changes in earnings response coefficients generally stem from the degree in permanence in earnings and/or the accounting recording lag and that the failure to recognize that both these effects impact the earnings coefficient may lead to misleading inferences. For instance, since accounting recording lag is the cause for prices leading earnings, it can be inferred that a low earnings response coefficient may reflect either transitory earnings and/or a great effect of prices leading earnings. Lubberink (2000) stated that low earnings response coefficients are often viewed as a result of low quality of financial statements, while low ERC can be the outcome of investors anticipations (prices leading earnings) that are not properly captured by the association model. There is also a lot of literature with a special regard to the transitory nature of losses. Hayn (1995) noted that losses are very important when estimating return earnings relation, because they are not expected to continue for ever, since shareholders have a liquidation option. She found that the information content of losses is limited in the US market. When loss observations excluded, the association between returns and earnings become much stronger. This is supported by Finnish data by Martikainen et al. (1997) and Kallunki & Martikainen (1997). 15

16 2.3.3 Return and price models Kothari and Zimmerman (1995), provide an explanation of why return models are commonly preferred to price models. The authors used several econometric models to explain that price models earnings response coefficients are less biased. Nevertheless, return models have less serious econometric problems than price models. In some studies the combined use of both price and return models may be useful. The logic is straightforward. Current earnings include both a surprise component and an expected component. The last is referred to as a stale component by Kothari and Zimmerman. They argue that this stale component is irrelevant in explaining current return and therefore constitutes an error in the independent variable. The result is that the slope coefficient in the return specification is biased towards zero. The price specification does not suffer from this problem because the stock price reflects the cumulative information content of both components. Current earnings are, however, uncorrelated with the information about future earnings contained in the current stock price (see also Liu and Thomas, 2000). Consistent to Kothari and Zimmerman (1995) were the results by Martikainen et al. (1997) and Dumontier and Labelle (1998) in Finland and France respectively. Using the published earnings numbers and the adjusted earnings numbers, estimated according to the recommendations by the Finnish Committee for Corporate Analysis (COC), they deduced that the return model specification yields higher and more significant ERC s compared to differenced model specification under the adjusted accounting data. Dumontier and Labelle (1998) found similar results regarding the ability of the return model to evaluate the return-earnings relation by providing highly significant ERC s. Moreover, their tests show that a cross-sectional data cumulating procedure could yield a large increase in the explanatory power of the return model and the significance of the earnings response coefficient. Ohlson and Schroff (1991) presented an analytical study of the issue of earnings levels versus earnings changes. In this study they argued against the common use of earnings changes in regressions. They confirmed that if investors use other information than information in earnings and dividends alone, there is a reason to prefer one specification over the other. They showed that a levels specification is appropriate in this case. 16

17 Easton and Harris (1991) empirically investigated the issue of levels versus changes. They applied a cross-sectional approach that uses a one year buy and hold return window that is regressed on both levels and changes of earnings. They concluded that the earnings levels variable is more often significant than the earnings change variable. They claimed that earnings levels outperform earnings changes in association studies. Results from separate univariate regressions support this claim as regressions using only levels of earnings show higher R 2 's than do those using only changes in earnings. Kothari (1992) presented analytical support for the levels specification, but now in a setting where returns lead earnings and Kothari and Sloan (1992) empirically supported the analysis. They assumed that stock returns over a period anticipate future earnings, while accounting earnings do not. They acknowledged that ERCs from return specifications are biased downwards. Kothari and Sloan offered a solution to this problem. They reduced the bias by using a return measurement interval that includes a leading time period in addition to the current time period. Despite that different methods are used to arrive to the same result, the research discussing the choice of levels versus changes of earnings in association studies favors the use of levels. Moreover, Kothari and Sloan (1992) demonstrated that including leading returns results in higher explained variation. This makes their approach a very powerful tool to assess accounting numbers. Additionally, this approach is more attractive to investors and analysts, because they can use available information to forecast future earnings Length of the response window Another group of studies examine how the strength of the earnings-returns relationship differs over different time intervals (Easton et al., 1992; Dechow, 1994). These studies theorise that accounting information suffers from timing and matching problems that are more severe over short-term intervals and dissolved as the interval is extended (Dechow, 1994). Scholars have documented that the contemporaneous linear relation between annual stock returns and accounting earnings has declined over time (Ryan and Zarowin 2003, Ely and Waymire, 1999) arising many speculations. In their study, Ryan and 17

18 Zarowin (2003) examined two reasons for the declining contemporaneous linear relation between annual stock returns and accounting earnings over the past thirty years. First, earnings increasingly reflect news with a lag relative to stock prices and second earnings increasingly reflect good and bad news in an asymmetric manner. They found that annual earnings have a weaker relationship with current price changes and a stronger relationship with lagged price changes over time. They hypothesised and found that annual earnings reflect current positive price changes less strongly and current negative price changes more strongly over time. They also found that asymmetry as regards lagged price changes is increasingly important over time. What is quite remarkable is that, since about 1980, the aggregation of earnings over a four-year window increasingly does less to reduce the importance of lags and asymmetry. Easton et al. (1992) showed (by aggregating earnings and investment outcomes over periods of up to ten years) that, over long intervals, the contemporary relation between aggregated earnings and stock prices grows stronger. The return-earnings association over shorter intervals is low because some economic events that cause revisions in the market s expectation about earnings are not captured in current earnings, or some past economic events are reflected in current earnings. Over longer intervals, however, the impact of a greater fraction of economic events is captured by the earnings, thereby yielding a stronger contemporaneous correspondence between longer interval returns and earnings. An alternative approach concerning the choice of measurement windows is Collins et al. (1994). Basically, they extended only the earnings window into the future: future earnings are regressed on current returns. They incorporated up to three future earnings years in their returns-earnings regressions, which increased the returns-earnings association considerably. Collins et al. found levels of explained return association that are up to six times higher when compared to regressions that only use contemporaneous earnings.. An issue arising both in Easton et al. (1992) and Collins (1994) studies is that they used realized lagged earnings in the regressions. Incorporating future realized earnings values increased the explanatory power and the slope coefficients. Though both studies only made use of earnings realized over a rather long period. Although the results 18

19 showed a relation between earnings and returns over the long term, it would be more interesting if one could forecast future earnings using current, available information. Kothari and Sloan (1992) presented a methodology that resolves this issue. Their research design, based on Kothari (1992), showed that earnings are anticipated by investors at least some periods ahead. Kothari and Sloan found high explanatory power (R 2 ) by including three leading period returns. 2.4 Empirical results of earnings response coefficient Introduction The following section represents results of empirical studies that used different method approaches and measures from the traditional ERC literature reviewed above. Furthermore, it represents results of European, non-european and Greek studies as well Results of alternative studies Beaver et al. (1980) explore the earnings-returns relationship from the reverse perspective. They used stock prices as the predictor of earnings and argued that this process might lead to more robust and accurate earnings forecasting models than the previously used random walk models. They reported that it is not likely that earnings follows a random walk simply because of its complexity and number of components involved. Easton and Harris (1991) used a different method and examined earnings as an explanatory variable for return and confirmed a relationship between the level of earnings (scaled by price) and stock returns at the beginning of the period. The main difference in this study is that it incorporated the level of and change in earnings rather than just the change in earnings. It was suggested that this would overcome the measurement error inherent in estimating unexpected earnings. The results showed that 19

20 the earnings levels and changes were both significant in most cases. This showed that both of these play an important role in stock price valuations. Scholars have also recognised that the relation between returns and stock prices varies across firms and over time. For example, Brown et al. (1987) showed that unexpected earnings that are based on financial analysts earnings forecasts provide a better explanation for abnormal returns than other proxies. They claimed that the measurement error in unexpected earnings may change on the basis of the firm s peculiarities and that in general; large companies tend to release information and earnings figures to market earlier and more frequently than small companies. Moreover, Kormendi and Lipe (1987) gave an explanation why a similar increase in unexpected or surprise announcement of accounting earnings in two firms might result in different stock price reactions. They showed that the impact of a current earnings surprise on stock prices depended on how much this surprise is expected to persist in the future. There is also a growing number of studies that use nonfinancial measures additionally to traditional financial variables to strengthen the estimated returns earnings relation. Amir and Lev (1996) argued that in fast-changing technology-based industries, the usefulness of financial information (earnings, book values, and cash flows) is limited, as firms in this sector invest heavily in intangibles such as R&D, franchise and brand development, which in turns leads to low or even negative earnings. Thus, the current earnings of these firms are transitory by nature and appear unrelated to market values. Using a sample of companies of the U.S cellular industry, they reported that the financial information is largely value irrelevant in returns earnings regressions, while the nonfinancial information is highly value relevant. Demers and Lev (2001) applied a factor analysis on financial and nonfinancial measures in order to examine the value drivers of the Internet firms. They reported that the nonfinancial Web traffic metrics are positively related to stock prices of the Internet firms even after the market correction in early In addition, they found that research and development expenditures of the Internet firms are regarded as a value-increasing activity by investors. Hand (2001) provided a analysis of the determinants of Internet stock valuation and he found that there are similarities in the cross-sectional pricing of Internet and non-internet firms during the 2000, concluding that the pricing of Internet stocks was not entirely irrational at that time. Trueman et al. (2000) found that gross 20

21 profits are significantly positively related to stock prices of the Internet firms when decomposing the net income into its components. In addition, they reported that the Web traffic measures have incremental importance with respect to net income and its component when explaining the prices of the Internet stocks Results of European stock markets Donnelly and Walker (1995) investigated and show that the extent to which prices lead earnings is reflected in differences in earnings response coefficients between models that allow for prices leading earnings and those that do not. Using the methodology proposed of Kothari and Sloan (1992) and a sample of UK companies they supported the view that increases in the ERCs realized by incorporating leading prices in the return earnings relation were not specious. Jermakowicz and Gornik-Tomaszewski (1998), studied the association between stock returns and annual earnings of 52 firms for the period of in the emerging stock market of Poland using a theoretical model which expresses price as a multiple of earnings developed by Ohlson (1995). They evidenced a significant association between stock returns and accounting earnings and they concluded that the annual earnings are an important element of the firms valuation process. The findings compare closely with the results found in more mature capital markets. Jindrichovska (2001) following the methodology of Kothari and Sloan (1992) investigated if there is a statistically significant relationship between returns and accounting data on the Czech market. She used accounting earnings and stock prices over the period of and supported the evidence from previous studies such as Kothari and Zimmerman (1995) and Kothari and Sloan (1992) that stock prices lead earnings. The results of firm-specific and pooled regression models suggested that the relation is statistically significant for measurement windows of one year and longer and reported that one-leading-year returns are as important as contemporaneous returns in terms of their sensitivity to annual earnings changes. Prices do not lead accounting data in the short run, which may be caused by the unreliable quarterly results in the Czech. Jarmalaite (2002) following the methodology of Kothari and Sloan (1992) analysed the relationship between accounting data and stock price returns in the stock markets of 21

22 Lithuania, Latvia and Estonia. He investigated both the contemporaneous one-period return-earnings relation and leading period returns in regression model of the main and secondary list securities over a 5-year period ( ). Results from this study suggested that the relationship between returns and earnings in Latvia seems to be very is similar to Estonia with Lithuania showing the weakest and Estonia showing the highest value relevance. Moreover, it is found that the occurrence of losses decreased the observed returns-earnings relation in all three Baltic countries Results of Non-European stock markets Al-Qenae et al. (2002) tested the relationship between share returns and earnings and macroeconomic variables, as gross national product, interest rates and the inflation rate for share prices within the prices leading earnings framework. They found evidence supporting the phenomenon of prices leading earnings for the Kuwait Stock Exchange (KSE) after controlling for basic macroeconomic variables, That is, the estimated earnings response coefficient is found to be significant to the leading periods and it increased when more leading periods were included. The results suggest that prices anticipate earnings and hence provide useful information to KSE investors and there are consistent with those of developed markets of the US (Kothari and Sloan, 1992) and the UK (Donnelly and Walker, 1995). Hisao Kai (2002) examined the information content of earnings for investors by using the liquidation option hypothesis and investigated the ERC changes in relation to firm's default risk in Tokyo Stock Exchange. He used three measurements of firm's default risk. Firm s debt ratio, an index based on earnings by classifying the firms into three categories and a classification of positive and/or negative earnings. He hypothesized and found that as the firm s debt ratio decreases, the relation between abnormal returns and unexpected earnings increases. The results are similar in the case of index, as abnormal returns unexpected earnings relation increases as the default risk decreases. These results are consistent to liquidation option hypothesis. However, the relation between abnormal returns and unexpected earnings had a negative sign in the case of positive/ negative earnings. 22

23 Myring et al. (2003) examined the existence of a short-term market reaction to unexpected earnings in Australia and the United States using data from 1987 to They used analysts forecasts as an indication of updated earnings expectations and examined the incremental explanatory power over the change in earnings per share. The results indicated that both the Australian and the United States markets react quite quickly to earnings releases. Similar results reported by Liu and Thomas (2000), who investigated the role of the analysts earnings forecasts in explaining returns in Sweden. Suwardi (2009) explored the relationship between accounting numbers and share prices of firms of Indonesia for the period , using dynamic modelling additionally to the cross sectional analysis. The results show that the dynamic specification that models equilibrium process between market and book values can often be identified using accounting regressors. In Jakarta Stock Exchange (JSX) compared to US studies that use similar models estimated, the book value of net assets appear to have a stronger relationship with market value. Azam (2009) studied the use of firm-specific versus pooled cross-sectional regression estimation procedures on applying the price and return models (Kothari and Zimmerman, 1995) to describe the relationship between stock prices or returns and accounting earnings. Using random samples of firms he found that the mean of the firm specific coefficients was, on average, two times larger than the corresponding coefficient estimated with a pooled cross-sectional regression methodology. The empirical results provided the evidence that there is a great impact on stock prices or returns due to accounting earnings in the Karachi Stock Exchange over the period of 2000 and Chien et al. (2010) explored the effect of investors moods on the way they interpret and respond to quarterly earnings announcements worldwide and whether this mood effect is asymmetric when firms announce good and bad earnings news. Furthermore, they investigated the role of institutional factors in decreasing the mood effect. Using 20 countries (including Greece) from 1990 to 2006, they found that market response to earnings announcements is connected to investor mood, which is dependent on weather-related factors such as cloudiness. Also, they concluded that the degree to which an investor s reaction to a firm s quarterly earnings announcements is affected by 23

24 mood is less for higher financial information transparency or common-law system countries Results of Greek stock market Research for the Greek stock market is not too limited.niarchos and Georgakopoulos (1986) examined the effect of annual corporate profit reports on the share prices of Athens Stock Exchange, and found a slow and gradual adjustment of stock returns to new information and concluded that the Greek stock market is inefficient. Kavussanos and Dockery (1995) also tested the efficiency of the Greek market by performing unit root tests on panel data and found similar results. The results are quite different with those of Panas (1990), who found an efficient ASE through univariate tests on monthly data of ten individual stocks. Kahya et al. (1993) showed that earnings growth rates are predictable using past earnings growth rates, stock price returns are predictable using past earnings growth rates as well as stock price returns, and firm size has no incremental explanatory power with respect to equity prices. Kousenidis (2005) employed the Easton and Harris (1991) model on a cross sectional basis to test for the information content of earnings for stock returns on the Athens Stock Exchange (ASE) over the period from 1992 to Moreover, the model is adjusted to account for the size effect while the life-cycle effect is examined by grouping sample firms into three different portfolios according to life-cycle stages. He found that the explanatory power of earnings for contemporaneous stock returns is not significant. Moreover when the regressions are adjusted to account for size, supported the hypothesis that firm-size is a strong factor in explaining the returns/earnings relation. Though, the results are not able to maintain the hypothesis that the information content of earnings for stock returns is different according to the stage of the firm s life-cycle. Papadaki and Siougle (2007) investigated the association between price and earnings across firms listed in the ASE for the period from 1985 to 1996 by dividing the sample into those firms that reported losses and those that reported profits. The regression 24

25 results confirmed a negative price-earnings relation for loss firms and a positive priceearnings relation for profit firms. Maditinos et al. (2007) investigated whether the level of EPS and EPS changes, the level of return on investment (ROI) and ROI changes, and the level of return on Equity (ROE) and ROE changes, are related to stock market returns in Athens Stock Exchange using the Easton and Harris (1991) model. The sample period spans 10 years, from 1992 to 2001 with a total of 977 year-observations. They used both relative and incremental information content approaches to test this association. The results provided evidence that there is an association between EPS and stock market returns. The results concerning ROI and ROE were not significant. As far it concerns the incremental information content approach results showed that the pair wise regression of EPS and ROI best explains the stock market returns in Greece, compared to the results provided by the combinations of EPS and ROE, and ROI and ROE. Dimitropoulos and Asteriou (2009) concluded that the price model produces less biased ERCs than the return model but it has econometric problems. They investigated the relationship between published accounting earnings prepared under the Greek GAAP and stock returns for a sample of 105 Greek companies during the period The results showed the increased ability of the price and return models to explain better the earnings-return relationship by providing highly significant earnings response coefficients. On the contrary, compared to the previous two models the changes and deflated models proved to be insufficient. These results are consistent to Vafeas et al. (1998), Kothari and Zimmerman (1995), Martikainen et al. (1997) and Dumontier and Labelle (1998). 2.5 Summary This chapter examined a variety of issues of the earnings response coefficient literature. This research developed from the early work of Ball and Brown (1968) and has become one of the most popular issues in the accounting literature. As it has been presented above the choice of the correct regression model is very important and in the case of prices leading earnings Kothari and Sloan (1992) provided analytical methodology, which is repeated in this dissertation, for the return model and a method that uses leading period returns to control for biased coefficients of return models. 25