UPPSALA UNIVERSITY Department of Business Studies Master Thesis Spring Semester The value relevance of comprehensive income

Transcription

1 UPPSALA UNIVERSITY Department of Business Studies Master Thesis Spring Semester 2012 The value relevance of comprehensive income Authors: Elena Ringström and Supervisor: Katarzyna Cieslak Date of submission: May 25, 2012

2 Abstract Master thesis: Department of Business Studies at Uppsala University Authors: Elena Ringström and Title: The value relevance of comprehensive income Background: In this study, we look at the effects of the adoption of the revised IAS 1 rules, which has been in effect since January 1, The revised IAS 1 requires that all changes in equity, excluding changes in equity arising from transactions with owners, should be recognized in comprehensive income statement. Revised IAS 1 requires companies to report total comprehensive income that is a sum of net income and other comprehensive income. Total comprehensive income includes all unrealized gains and losses recognized under IFRS. Before the amendment, some of the unrealized gains and losses were shown in a statement of changes in equity but not in the income statement. Research problem: We hope to answer the question whether inclusion of the components of other comprehensive income provides investors with useful information. We investigate if stock prices have an association with the components of other comprehensive income. We investigate how effective are attempts of IASB to increase the relevance of accounting information about corporate income. We hope that results from the study will be of interest to the standard-setter. Research design: In this study, we use data from annual reports and year-end reports for companies listed on the Large and Mid Cap segment at NASDAQ OMX Stockholm and that covers the years 2009 to We use two regression models to test value relevance of components of other comprehensive income. Empirical findings: We have found some evidence that the share price statistically relates to such component of comprehensive income as the change of the fair value of cash flow hedges. This can also be interpreted as that the change of the fair value of cash flow hedges has some value relevance. We also found some evidence that the share price significantly associates with winning cash flow hedging position. We did not find that the share price associates with some other components of other comprehensive income. Keywords: Value relevance, Comprehensive income, Components of other comprehensive income, Revised IAS 1, Fair value. 1

3 Table of contents Abstract... 1 Table of contents Introduction Background Research problem Theory Theoretical framework Income concepts Clean surplus accounting Previous research Method Data collection and sample Research design Empirical analysis Regression model Regression model Discussion and conclusions Discussion Conclusions, limitations and further research References

4 Information is a source of learning. But unless it is organized, processed, and available to the right people in a format for decision making, it is a burden, not a benefit. - William Pollard (n.d.) 1. Introduction 1.1 Background From January 1, 2009, The International Accounting Standards Board (IASB) implemented revised IAS 1 that has changed a presentation of the income statement. According to the revised IAS 1, listed companies have to report on comprehensive income, in addition to net income. Total comprehensive income is a sum of net income and other comprehensive income. It should be noted that total comprehensive income includes all unrealized gains and losses recognized under IFRS regulation (Kabir and Lasward, 2011, p. 270). These unrealized gains and losses mainly arise from the fair value changes in the balance sheet accounts. Comprehensive income is all-inclusive income that is also known as clean surplus income. In contrast, before the amendment, listed companies had to do some certain transactions into equity and show them in the statement of changes in equity instead of the income statement. These transactions were known as dirty surplus components of (non all inclusive) income (Pronobis and Zülch, 2011). Lönnqvist (2011) confirms that, before the amendment, some revaluation changes in fair values were recognized in the statement of changes in equity because of the precautionary principle since these revaluations could be difficult to realize. In this study, we look at the effects of the adoption of the revised IAS 1 standard that was caused by the convergence project between IASB and Financial Accounting Standards Board (FASB). IASB has decided to revise IAS 1 regarding the presentation of comprehensive income similar to make it similar to SFAS 130 that was issued by FASB in 1997 (IFRS Foundation, 2011, p. IN3 A552). FASB introduced the definition of comprehensive income in the Statement of Financial Accounting Concepts No. 5 Recognition and Measurement in Financial Statements of Business Enterprises as the change in equity (net assets) of an entity during a period from transactions and events and circumstances from non-owner sources. Comprehensive income includes all changes in equity during a period except those resulting from investments by owners and distributions to owners (Schroeder, Clark, and Cathey, 2001, p. 111). 3

5 The revised IAS 1 follows FASB s definition of comprehensive income and suggests that all changes in equity, excluding changing in equity arising from transactions with owners, are recognized in comprehensive income statement (IFRS Foundation, 2011, p. IN13 A553 f.). Other comprehensive income includes incomes and expenses that are not recognized in the profit or loss as permitted or required by other IFRSs (IAS 1, 2011, p. 7). Other comprehensive income includes unrealized gains and losses resulting from changes in fair values of assets/liabilities such as: changes in revaluation amount of property, plant and equipment (IAS 16) changes in fair value of intangible assets (IAS 38) actuarial gains and losses on defined benefit plans (IAS 19) gains and losses from translating the financial statements of a foreign operation (IAS 21) changes in fair value of available-for-sale financial instruments (IFRS 9) changes in fair value of a hedging instruments in a cash flow hedge (IAS 39) According to US GAAP, other comprehensive income includes foreign currency items, minimum pension liability adjustments, unrealized gains and losses on certain investments in debt and equity securities. In addition, according to SFAS 130 the components of other comprehensive income can be reported in three ways as below the total for net income in an income statement, a separate statement that starts with net income, or in a statement of changes in equity (Schroeder et. al, 2001, pp ). However, revised IAS 1 does not permit reporting of other comprehensive income in a statement of changes in equity (IFRS Foundation, 2011, p. IN13 A553 f.). It implies that there are some differences in components and presentation of other comprehensive income between IFRS and US GAAP. Consequently, prior research on total and other comprehensive income conducted in Anglo-Saxon countries cannot be fully generalized to reporting of comprehensive income in European countries and according IAS 1. The ideas of clean surplus relation have influenced this new way of reporting (Kanagaretnam, Mathieu, and Shehata, 2009; Smith, 2006, p. 280). Clean surplus relation proposes that gains and losses should not be recognized directly in equity (Sundgren, Nilsson H. and Nilsson S., 2007, p. 37). That can be interpreted as clean surplus relation suggests that all gains and losses (even unrealized) should be recognized in the statement of profit and loss. Before the amendment of IAS 1, some non-owner changes in equity could be recognized in equity which was not consistent with 4

6 clean surplus relationship (Smith, 2006, p. 280). Hence, one can suggest that since the previous way of reporting of income violated clean surplus relation, it was called in the literature as dirty surplus income. For example, before the introduction of comprehensive income, if there was an increase of the carrying amount of item of property, plant and equipment (IAS 16), it was credited to revaluation reserve in equity but was not recognized in the statement of profit and loss. This was because of the precautionary principle in order to separate short-term revaluation fluctuations from operating profit. However, it was a departure from clean surplus relation (Sundgren et. al., 2007, p. 86). After the amendment, if there is a revaluation increase it shall be credited to a revaluation reserve and recognized as other comprehensive income in the statement of comprehensive income. If there is a revaluation decrease of property, plant and equipment items it shall be debited to the revaluation reserve and recognized as a negative number in other comprehensive income to the amount of any credit balance earlier existing in the revaluation reserve relative to the same asset (Melville, 2011, p. 81). This accounting procedure attempts to signal that unrealized revaluation gains in other comprehensive income are not converted to cash and consequently cannot be paid as a dividend (Melville, 2011, p. 81). It can be interpreted in this way that IAS 1 tries to follow clean surplus relation when showing unrealized gains and losses in the income statement as other comprehensive income. At the same time, the separate post of other comprehensive income attempts to point out to the users that these gains and losses are not real enough to recognize them in net income. 1.2 Research problem IASB states that the main purpose of the extension of the income concept was to provide users of financial statements with more relevant information (IAS 1, 2011). FASB confirms that reporting of comprehensive income would help investors and other users of financial reports in evaluating of firms financial performance and their future cash flows (Schroeder et. al., 2001, p. 116). Furthermore, the position of comprehensive income signals to investors that it contains valuable information (Schroeder et. al., 2001, p. 118). The regulator by requiring that comprehensive income should be presented in the income statement wants to communicate to the capital market that the information in comprehensive income can contain significant and relevant information. 5

7 Ball and Brown (1968) found that the content of the company's income number was considerably crucial for investors. The study justified as well that information about companies annual income has a connection with the stock prices (Ball and Brown, 1968). This might also be interpreted in this way that Ball and Brown (1968) showed that the income number was value relevant for investors. Based on the above discussion, an assumption can be made that the inclusion of additional information in the companies income will affect the capital market. Since 2005, all listed companies within EU have to report with accordance to IFRS standards (FAR, 2011). The main change caused by implementation of IFRS was move towards fair value accounting for assets and liabilities (Penman, 2007; Sundgren et. al., 2007). There has been a significant growth in the use of fair value accounting for assets in Sweden since the implementation of IFRS (Sundgren et. al., 2007). One of advantages of fair value accounting appears to be providing of more relevant accounting information to the investors (Penman, 2007). However, the use of fair values without active market prices with the help of valuation models based on expected future cash flows might be subjective and, hence, volatile financial statements (Penman, 2007; Sundgren et. al., 2007). Other comprehensive income adjusts unrealized gains and losses arising from the fair value changes those before the amendment of IAS 1 were booked in a statement of changes in equity (Lönnqvist, 2011). The study of Ball and Brown (1968) was many years ago, and it is appealing to check how much accounting is relevant now, especially in the more turbulent times, when relevance of accounting can be questioned. IASB makes an attempt to enhance the relevance of accounting with the changes in IAS 1. It is of great importance to investigate the success of these attempts by asking how the other comprehensive income components are relevant for investors. This is what the aim of the study is. To the present time, there has been an extensive research conducted on the value relevance of comprehensive income compared to net income in Anglo Saxon countries (Pronobis and Zülch, 2011). The implementation of SFAS 130 in 1997 caused considerable research on the value relevance of comprehensive income (Pronobis and Zülch, 2011). The study of Kanagaretnam et. al. (2009, p. 349) who used data from Canadian companies found that comprehensive income had a stronger explanatory power and association with stock prices and stock returns than net income. This study has also found that the share price and stock returns associated with some components of other comprehensive. On the contrary, the study of Dhaliwal, Subramanyam and Trezevant (1999) established that only one component of other comprehensive income had an association with stock 6

8 returns. However, not much research has been done on value relevance of transactions in equity ( dirty surplus flows) in EU countries including Sweden (Roberts and Wang, 2009). In contrast to Anglo Saxon countries, the issue of addition of other comprehensive income is new in Europe; hence, there has been no research conducted on value relevance of components of other comprehensive income after the amendment of IAS 1. Earnings might be interpreted as one of the most powerful accounting measures that are of interest to investors. In view of capital market research, investors can be seen as primary users of financial statements (Ball and Brown, 1968; Kothari, 2001). Professional and non professional investors who apply fundamental analysis of business valuation widely use such accounting measure as earnings (Barker, 2001; Larsson, 2008; Palepu, Healy, Bernard and Peek, 2007). To evaluate whether inclusion of additional information in the income number affects capital market there might be a need to study if stock prices react to additional accounting information included in the components of other comprehensive income. We hope to answer the question whether reporting on the components of other comprehensive income provides capital market (investors) with useful information. We investigate if stock prices have an association with the components of other comprehensive income. According the prior researches, we expect that some components of other comprehensive income are associated with stock prices. In line with other studies on value relevance of comprehensive income (Kanagaretnam et. al., 2009; Pronobis and Zülch, 2011), as well as the IASB's Conceptual Framework (2011), we suggest investors as primary users of financial statements. From a practical point of view, since we study association between share price and accounting information, we believe that the study might be especially useful for non-professional investors those make business evaluation without professional investment help. Furthermore, we hope that the study might be stimulating for scholars working in the field of accounting research. The results from the study might be also relevant for the standard-setter since it provides empirical evidence on the use of accounting standards in Sweden one of EU countries. In addition, the results from the study of association between the share price and accounting information might be of interest to the listed companies since capital market is a source of their financing. 7

9 2. Theory 2.1 Theoretical framework Income concepts According to Riahi-Belkaoui (2000, pp ), the theory and measurement of business income has always been a prominent area of financial and managerial accounting. In spite of different measures and measurement methods, the concept of income remains the central subject of various interpretations. The classical school of thoughts applies historical - cost principle and defines business income as accounting income. Accounting income is the difference between the realized revenues of the period and the matching historical costs. However, accounting income has been insensitively criticized for its limitations (Riahi-Belkaoui, 2000, p. 390). One of the arguments against accounting income is it does not recognize unrealized increases in values of assets because of the application of historical-cost and realization principle. This allows the recognition of a various gains from the past and present periods and, as a result, the net income does not match, in fact, the income of the current period (Riahi-Belkaoui, 2000, p. 390). Riahi-Belkaoui (2000, pp ) describes that income concept has also been a point of considerable interest to economists. Adam Smith described income as an increase in wealth. Following Smith s ideas, Alfred Marshall connected concept of income to business practices. Fisher, Lindahl and Hicks developed a concept of economic income. According to Van Cauwenberge (2007, p. 21), Hicks described firm s profit as value changes over the level that is needed to maintain the value of firm s capital (net assets). Therefore, Van Cauwenberge (2007) argues that Hicks concept of income as capital maintenance supports the concept of comprehensive income. The concept of capital maintenance suggests that income is recognized after capital is maintained and expenses are recovered (Riahi-Belkaoui, 2000, p. 391). Financial capital maintenance means that the financial (money) amount of company s net assets at the end of the accounting period goes over the financial amount of the net assets at the beginning of the period, except transactions with owners (Schroeder et. al., 2001, p. 64). One can see similarities between definition of 8

10 comprehensive income and content of the concept of financial capital maintenance. Therefore, it can be understood that the idea of financial capital maintenance supports the concept of comprehensive income. In addition, the concept of financial capital maintenance considers that holding gains and losses should be included in income. Holding gains and losses arise from the value of items in the balance sheet changes (Schroeder et. al., 2001, p. 65). Riahi-Belkaoui (2000, p. 398) mentions that holding gains and losses occur when assets and liabilities those a firm owes or holds measured in fair values. Holding gains and losses can be separated in two groups. The realized holding gains and losses concern the items sold or to the liabilities paid and non-realized gains and losses relate to items still held and liabilities still owed. With reason that the concept of financial capital maintenance proposes to include all holding gains and losses in income, one can say that it supports the model of comprehensive income presented by IASB Clean surplus accounting Concept of comprehensive income or all-inclusive income is consistent with the accounting system that is known as clean surplus relation (Christensen and Feltham, 2003, p. 280). The book value of equity is the difference between company s assets and liabilities and can be called the owners surplus. Clean surplus relation means the accounting system that recognizes all changes in the owners surplus and exchanges of cash between the firm s owners as well as earnings (Christensen and Feltham, 2003, p. 280). More exactly, the clean surplus relation suggests that comprehensive income (earnings) is equal to change in equity during a period increased by issues of new shares and reduced by dividends to shareholders (Barker, 2001; Ohlson, 1995): CI t = BV t BV t-1 d t Where: BV t = book value of the firm s equity, date t BV t-1 = book value of the firm s equity date, t-1 CI t = comprehensive income for period (t-1, t) d t = dividends, net of capital distributions, date t 9

11 It implies that since clean surplus relation means that there is a connection between equity (balance sheet) and earnings (the statement of profit and loss), all changes in equity can affect earnings. In this case, it can be logical to recognize all changes in equity in the statement of profit and loss. With reason that earnings are equal to change in equity reduced by a factor of dividends (net of capital distributions) it can be rational to separate changes in equity that comes from transactions with owners (dividends). This principle of aggregation of changes in equity is obviously applied to the model of comprehensive income reporting in revised IAS 1. In addition, Lönnqvist (2011) verifies that since clean surplus relation proposes connection between balance sheet and a statement of profit and loss, on one hand, there is a connection between assets and liabilities between incomes and expenses. Lönnqvist (2011) explains net asset as a difference between a firm s assets and liabilities that is equity in the balance sheet. When following clean surplus relation, if there is an increase in net asset that does not arise from new shares issue it should be recognized as income in a statement of profit and loss. If there is a decrease in the net asset that does not arise from dividends or when company purchases its own stocks, it should be recognized as an expense in a statement of profit and loss. It can be assumed that this way of thinking can be applicable to the revised IAS 1 since it requires recognizing of all changes in equity not arising from transactions with owners in the statement of comprehensive income. 2.2 Previous research Relevance and value relevance of the information in the financial statements is often understood differently. Various user groups of the financial statements can often have different ideas about what information is relevant (Holthausen and Watts, 2001, pp ). The IASB's Conceptual Framework (2011, p. 9) lists various user groups of the financial statements of a listed company and their different information needs. IASB mean that since investors supply equity capital to the company, information is of interest to them (IASB's Conceptual Framework, 2011, p. 10). To enable investors to decide whether to sell, hold or buy shares in the company, the investors need relevant information. However, opinions about relevant information may vary between different groups of investors (Holthausen and Watts, 2001, pp ). 10

12 In view of capital market research, investors are the primary users of financial statements. Capital market research area supposes that the primary aim of the financial statements is to provide a basis for valuation of the companies. It is an extensive research area which origin can be traced back to Ball and Brown (1968) empirical study where they examined the usefulness of income data that existed in the company's accounting (Ball and Brown, 1968, p. 176; Kothari, 2001, p. 106). The study of Ball and Brown (1968) is often associated with the term of value relevance that usually means that stock prices have statistically connection to accounting information. This association indicates that the tested part of the accounting information relates to the information used by investors when they make their investment decisions (Francis and Schipper, 1999, p. 326). An assumption can be made that since the most of research on value relevance of comprehensive income used different methods of statistical association between income and some market information (stock prices, stock returns), this research can be understood as capital market research. As capital market research suggests that financial statements provide investors with information for companies valuation, information about comprehensive income can be of interest to investors. An extensive research on relevance of comprehensive income comparing to net income has been done in different countries, especially with Anglo Saxon accounting tradition. Countries with Anglo Saxon accounting tradition apply accounting standards those have been influenced by US GAAP (Smith, 2006). Chambers, Linsmeier, Shakespeare, and Sougiannis (2007) and Choi, Das, and Zang (2007) found that other comprehensive income is significant in the valuation of a company. Choi et. al. (2007) provided evidence that the stock prices react on the inclusion of other comprehensive income. Kubota, Suda and Takehara (2009) examined the information content of comprehensive income, other comprehensive income and net income for Japanese firms. They concluded that other comprehensive income is useful information although they could not rank between net income and comprehensive income (Kubota et. al., 2009). Kanagaretnam et. al. (2009) that were using data from Canadian companies, found that such components of other comprehensive income as available-for-sale and cash flow hedges were considerably associated with stock prices and stock returns. By contrast, Dhaliwal et. al. (1999) justified that only such component of comprehensive income as marketable securities adjustment was associated with stock returns. The researchers found no evidence that comprehensive income is more strongly associated with stock returns and market value of equity than net income. Pinto (2005) provides empirical evidence that reporting of such 11

13 element of comprehensive income as foreign currency translation adjustments is significantly value relevant. The researcher justifies disclosure of foreign currency translation adjustments when she uses the sample of US multinational companies with considerable investments in foreign countries. Moreover, some of behavioral studies investigated whether the way of presentation of comprehensive income is relevant for investors. The study of Hirst and Hopkins (1998) revealed that financial analysts could detect earnings management on some items only if this information was disclosed in a statement of comprehensive income. Maines and McDaniel (2000) found that the way of presentation of accounting information influenced evaluation of companies performance by non - professional investors. The researchers showed that non - professional investors used rather comprehensive income information in a separate statement of comprehensive income than the same information disclosed as a component of equity. Schroeder et. al. (2001) argue that the results of these studies have shown that the way of comprehensive income reporting can affect investors decision making. Not much research has been conducted in Sweden. Roberts & Wang (2009) examined how differences in institutional factors, in different European countries, including Sweden, affect the value relevance of total hypothetical comprehensive income. The researchers added components from statement of equity to net income and then linked the total income with stock returns. Roberts & Wang (2009) justified that comprehensive income was significant in explaining stock returns. Roberts & Wang (2009) found that total comprehensive income was incrementally relevant in Sweden where equity market plays a crucial role in financing of the companies. However, Roberts & Wang (2009) conducted their study on total comprehensive income. However, the study did not test value relevance of components of other comprehensive income. The reason of lack of research on the relevance of components of other comprehensive income on data from the Swedish companies can be explained by newly revision of IAS 1. Practitioners, standard-setters and academics appear to consider other comprehensive income economically significant (Roberts & Wang, 2009). Reporting of other comprehensive income components corresponds to require from users of financial statements for more visible and useful presentation (Van Cauwenberge, 2007). New accounting information has been included in the corporate income number, although, there is no study on the usefulness of this information for the users in Sweden. Study on the value relevance of other comprehensive income based on data from 12

14 Swedish companies might be needed. Value relevance of accounting information can be tested by the method when stock price relate to book value of equity, earnings and other relevant information (Ohlson, 1995). To evaluate whether inclusion of other comprehensive income was useful for the users in Sweden, study of association between stock prices and book value of equity, as well as earnings, might be useful. 13

15 3. Method 3.1 Data collection and sample In this section, we describe what companies we have selected to investigate and what information we have collected for these companies. We begin with the selection of companies to investigate. Since 2005, all companies that are listed on the stock market within the European Union are obliged to prepare their consolidated financial statements in accordance with IASB's standards 1 (FAR, 2011, pp. 3 and 29-32). In Sweden, there are three securities exchanges and four multilateral trading facilities that are authorized by the Swedish Financial Supervisory Authority (Finansinspektionen, 2012). In these stock exchanges, there are a total of 516 listed companies 2, of which 245 companies are listed on NASDAQ OMX Stockholm (also known as Stockholm Stock Exchange or Stockholmsbörsen). We have chosen to limit selection to the companies that are listed on the Stockholm Stock Exchange. The reason for this limitation is that the companies listed on the other stock exchanges are usually smaller companies with low trading volumes. The companies on the Stockholm Stock Exchange are divided into three size segments. The size segments are Small Cap, Mid Cap and Large Cap. Large Cap contains companies which have a market value of over 1 billion Euros. The Mid Cap segment includes companies, which have a market value between 150 million and 1 billion Euros. Companies with a market value of less than 150 million Euros are contained in the Small Cap segment. Of the 245 companies listed on the Stockholm Stock Exchange, 111 companies belong to the Small Cap segment, 76 companies to the Mid Cap segment, and 58 companies belong to the Large Cap segment (NASDAQ OMX Stockholm, 2012:1). The sample includes companies that belong to Large and Mid Cap since it is more likely that these companies have a wide range of other comprehensive income components. The initial sample includes 134 companies. We look at the market information about the companies three months after the fiscal year end (in the end of March). Therefore, we have excluded companies which fiscal year is dissimilar to the calendar year because they publish their annual reports in different periods. Following Kanagaretnam et. al. (2009) who showed results only for 1 Such as The commission of the European Communities has adopted these standards (FAR, 2011, pp. 3 and 29-32). 2 The three securities exchanges are NASDAQ OMX Stockholm (245 listed companies, see Nordic Growth Market (16 listed companies, see and Burgundy (0 listed companies, see and the four multilateral trading facilities are NASDAQ OMX Stockholm First North (112 listed companies, see Nordic Growth Market (18 listed companies, see AktieTorget (125 listed companies, see and Burgundy (0 listed companies, see burgundy.se/). The information from the market places was retrieved March 10,

16 non-financial firms we excluded financial and real estate companies. We excluded also firm-year observations when all components of other comprehensive income were zero. According to the rules of Stockholm Stock Exchange, the companies' financial reports shall be available at the company's website for at least five years after publication (NASDAQ OMX Stockholm, 2012:2, p ). We collected data accounting information from the companies annual reports that covers period We started from the year 2009 because the companies reported in accordance with IFRS have to present comprehensive income since January, 1 st To collect accounting data for the fiscal year 2011, we used both annual and year-end reports because some companies had not published their annual reports yet. The year-end reports are not audited but to ensure the reliability of the data we have compared information included in annual and year-end reports for previous years. We confirmed that information was identical. Finally, we have ended in 246 firm-year observations. We manually collected information about net income, components of other comprehensive income and book value of equity from the annual and year-end reports. Most of the information was in Swedish Crowns, but some information was in US dollars or Euros. When firms disclosed financial information in foreign currency we converted it to Swedish Crowns using the average exchange rate. To ensure that the market had the information from the annual reports or year-end reports, we used the share price and number of shares outstanding 3 months after the fiscal year end. From the corporate press releases we got to know that the most of the companies publish their annual reports and all companies publish their year-end reports to the end of March. We used the database Datastream to collect information about the market value and number of shares outstanding. The information was brought together in a spreadsheet in Excel. Before we ran the multiple regression analysis, we removed observations that were regarded as outliers. However, before the exclusion of the outliers we investigated whether these extreme values those were detected as outliers were wrongly recorded under the data collection process. Pallant (2010, p. 43) describes outliers as scores that are much below or well above other scores. If outliers are left in the sample, they can significantly bias the results of a regression analysis (Moore, 2001). Pallant (2010, p. 159) explains that outliers can be detected using different methods. We followed recommendations provided by Pallant (2010, p. 159) and excluded outliers that had Mahalanobis distances above the critical value. For the regression model 1, we have excluded outliers that had 15

17 Mahalanobis distances above 18,47 since this regression model contains 4 independent variables. Before the regression analysis according model 2, we excluded outliers those had Mahalanobis distances above 20,52 because this model includes 5 independent variables. Furthermore, we excluded outliers that had standardized residuals that were higher than 3,3 or less than 3,3. Totally, before regression analysis according model 1 we excluded 28 outliers 1 and 30 outliers before the regression according model 2. In general, these outliers could be interpreted as observations with extremely high and extremely low values of net income, book value, etc. These outliers included, for example, such companies as AstraZeneca, Stora Enso and Lundin Mining Corporation SDB, etc. In other words, for the companies those were identified as outliers there were large variations between net income, other comprehensive income components and other values. 3.2 Research design Value relevance of accounting information can be proved by statistical study of relationship between book value of equity and earnings as well as other relevant information to the stock prices (Ohlson, 1995). To test value relevance of comprehensive income, researchers commonly use statistical studies applying regression analysis (Kanagaretnam et. al., 2009; Pronobis and Zülch, 2011; Roberts and Wang, 2009). Multiple regression analysis allows to explore the relationship between one continuous dependent measure and a number of independent variables (Pallant, 2010). Research on value relevance of comprehensive income often follows the well-known theoretical model developed by Ohlson (1995) that explains investor s value of a firm as a function of a firm s book value and residual earnings (Kanagaretnam et. al., 2009). Since we aim to clarify if stock prices reflect the information recognized in components of other comprehensive income, research design follows the statistical study conducted by Kanagaretnam et. al. (2009). Kanagaretnam et. al. (2009) among other things investigated components of other comprehensive income as explanatory variables for stock prices. The models that are used in this part of the paper were developed by Kanagaretnam et. al. (2009, pp ). The regression models that the researchers developed are based on the valuation function of Ohlson (1995). Model 1: PRICE = b 0 + b 1 BVE_S + b 2 NI_S + b 3 HEDGE_S + b 4 SEC_S + b 5 FOREX_S+ έ 16

18 Where: PRICE = price per share 3 months after the end of fiscal year; BVE_S = book value of common equity at the end of the year deflated by the number of outstanding shares; NI_S = annual net income after extraordinary items and discontinued operations deflated by the number of outstanding shares; HEDGE_S = the change of the fair value of cash flow hedges for a year deflated by the number of outstanding shares; SEC_S = the change in fair value of available-for-sale financial instruments for a year deflated by the number of outstanding shares; FOREX_S = the change of accumulated foreign currency translation adjustment for a year deflated by the number of outstanding shares; έ = error term Use of such variable as book value of common equity is ambiguous, and; therefore, we exchange it for book value of equity similar to Ohlson (1995) and Graham, Lefanowicz, and Petrony (2003). Furthermore, we replace net income after extraordinary items and discontinued operations reported according to Canadian GAAP by net income before tax. It should be also noted that we have tried to do multiple regression using extended model 1 of Kanagaretnam et. al. (2009) and add the variable of actuarial gains and losses. However, this regression analysis indicated high multicollinearity; therefore, we have decided to modify this equation to ensure the reliability of regression results. Under the annual report study, we have found only 42 observations of such variable as the change in fair value of available-for-sale financial instruments. The number of observations is much less compared, for example, with such components of other comprehensive income as the change of the fair value of cash flow hedges and the change of accumulated foreign currency translation adjustment. Therefore, we conduct a bivariate regression between share price and changes in fair value of available-for-sale financial instruments to identify whether it can be used as explanatory variable for the share price. Regression equation is: 17

19 PRICE = b 0 + b 1 SEC_S + έ Table 3.1 presents the results of regression between share price and the change in fair value of available-for-sale financial instruments. Variables Unstandardized Standardized t Sig. Coefficients Coefficients B Std. Error Constant 125,366 16,563 7,569 0,000 SEC_S -9,928 44,808-0,035-0,222 0,826 Table 3.1 Regression between share price and the change of fair value of available-for-sale financial instruments. Pallant (2010, p. 161) says that standardized coefficients play an important role when comparing the variables. However, analysis of unstandardized coefficients is used when one tries to build a regression equation. Since we aim to compare variables not to build a regression equation, we use analysis of standardized regression coefficients in the study. According regression a standardized coefficient on the change of available-for-sale financial instruments was negative (-0,035) and not statistically significant. Significance means p-value that is also called as the significance level (Moore, 2001, p. 442). In the statistical sense significant does not suggest important. It actually means that something does not likely to happen by chance (Moore, 2001, p. 443). Freedman (2009, p. 70) recommends that if p is less than 0,1 (10%) one can say that the coefficient is statistically significant at 10% level, or barely significant. If p is less than 0,5 (5%) coefficient is statistically significant at 5% level, or statistically significant. If p is less than 0,01 (1%), then coefficient is statistically significant at 1% level, or highly significant (Freedman, 2009, p. 70). Since the variable does not have statistical significance it can be interpreted as if the variable is not making a statistically significant contribution to the equation (Pallant, 2010, p. 161). The variable of the change in fair value of available-for-sale financial instruments (SEC_S) is not statistically significant and, hence, we excluded it from the regression model 1. We wanted to extend the regression model developed by Kanagaretnam et. al. (2009); therefore, we collected data about the companies actuarial gains and losses. However, we have only 48 observations. Therefore, we make a regression between share price and actuarial gains and losses to find out whether this variable can contribute to the regression model. 18

20 Regression equation is: PRICE = b 0 + b 1 AGL_S + έ Where: AGL_S = actuarial gains and losses for a year deflated by the number of outstanding shares Table 3.2 presents the results of regression between share price and actuarial gains and losses. Variables Unstandardized Standardized t Sig. Coefficients Coefficients B Std. Error Constant 115,807 16,505 7,017 0,000 AGL_S -13,379 9,596-0,201-1,394 0,170 Table 3.2 Regression between share price and actuarial gains and losses. The coefficient on actuarial gains and losses is negative (-0,201) and is not statistically significant. Since this variable is not making a unique statistical contribution to the equation, it was not included in the model. We found extremely few observations (less than 10) for both changes in revaluation amount of property, plant and equipment (IAS 16) and changes in fair value of intangible assets (IAS 38), consequently these components of other comprehensive income could not be used as variables. This can be also interpreted as such components of other comprehensive income as change of the fair value of cash flow hedges, and the change of accumulated foreign currency translation adjustment are the most common for the largest and medium-sized Swedish companies. Consequently, the regression model 1 becomes: PRICE = b 0 + b 1 BVE_S+ b 2 NI_S + b 3 HEDGE_S + b 4 FOREX_S + έ Where: 19

21 PRICE = price per share 3 months after the end of fiscal year; ors: Elena Ringström BVE_S = book value of equity at the end of the year deflated by the number of outstanding shares; NI_S = net income before tax for the fiscal year deflated by the number of outstanding shares; HEDGE_S = the change of fair value of cash flow hedges for a fiscal year deflated by the number of outstanding shares; FOREX_S = the change of accumulated foreign currency translation adjustment for a year deflated by the number of outstanding shares; έ = error term Then Kanagaretnam et. al. (2009) added an indicator variable GAIN to their basic regression model to investigate how the market reacts to positive cash flow hedge numbers. GAIN is equal 1 if a company has profit in cash flow hedging and is equal 0 if a company has a loss in cash flow hedging. Then they introduced an interaction variable HEDGE_GAIN_S that is equal to GAIN*HEDGE_S (Kanagaretnam et. al., 2009, p. 356). Kanagaretnam et. al. (2009, p. 356) developed regression model 2 that is: PRICE = b 0 + b 1 BVE_S + b 2 NI_S + b 3 HEDGE_S + b 4 SEC_S + b 5 FOREX_S + b 6 GAIN + b 7 HEDGE_GAIN_S + έ The basic regression model 2: PRICE = b 0 + b 1 BVE_S + b 2 NI_S + b 3 HEDGE_S + b 4 FOREX_S + b 5 GAIN + b 6 HEDGE_GAIN_S + έ We have tried to do regression analysis using this model; however, Pearson s correlation between HEDGE_S and HEDGE_GAIN_S was 0,966. According to Pallant (2010, p. 151), a correlation that is equal to 0,9 or above can be a sign of multicollinearity. We also had a VIF value of HEDGE_S that was equal to 27,063 and a VIF value of HEDGE_GAIN_S that was equal to 24,349. Pallant (2010) mentions that a VIF value above 10 also is a sign of multicollinearity. Phenomenon of multicollinearity can occur when explanatory variables move too closely together that can make evaluation of their individual effects on dependent variable difficult or even impossible (Murray, 2006, p. 321). Furthermore, when one independent variable is a combination of 20

22 other independent variables, singularity can occur (Pallant, 2010, p. 151). In this model, the variable HEDGE_GAIN_S is a combination of other two independent variables such as HEDGE_S and GAIN. Pallant (2010, p. 151) says that before the regression analysis one should check for both multicollinearity and singularity because these phenomena unquestionably do not contribute to an appropriate regression model. Consequently, we have excluded variable HEDGE_S from the model 2 to avoid the possibility of multicollinearity or singularity. Finally, the regression model 2 becomes: PRICE = b 0 + b 1 BVE_S + b 2 NI_S + b 3 FOREX_S + b 4 GAIN + b 5 HEDGE_GAIN_S + έ 21

23 4. Empirical analysis 4.1 Regression model 1 Table 4.1 below presents descriptive statistics for the variables that we used in the study of association between share price and components of other comprehensive income. Variables Mean Std. N Deviation PRICE 108, , NI_S 8, , BVE_S 44, , FOREX_S - 0,6432 3, HEDGE_S 0,2506 1, Table 4.1 Descriptive statistics for the variables. Variables: PRICE = price per share 3 months after the end of fiscal year; NI_S = annual net income before tax deflated by the number of outstanding shares; BVE_S = book value of equity at the end of the year deflated by the number of outstanding shares; FOREX_S = the change of accumulated foreign currency translation adjustment for a year deflated by the number of outstanding shares; HEDGE_S = the change of the fair value of cash flow hedges for a year deflated by the number of outstanding shares. All variables are deflated by the number of shares outstanding. According to table 4.1 the mean of net income is equal 8,9943, which signals that the majority of non-financial companies in Large and Mid Cap were profitable over the period Most of the companies have negative value of such component of other comprehensive income as change of accumulated foreign currency translation adjustment. This component of other comprehensive income mainly includes changes in fair values of assets and liabilities in foreign subsidiaries (Lönnqvist, 2011). This detail that most of the companies have negative value of change of accumulated foreign currency translation adjustment under can be probably understood in the light of this fact that Swedish Crown was weak under 2009 and became stronger under However, the majority of the firms have positive change of such component as change of the fair value of cash flow hedges. This can be eventually understood in this way that the largest and the medium-sized Swedish companies listed on NASDAQ OMX Stockholm successfully used such risk management procedures as financial hedging under

24 Sig. (1-tailed) Pearson Correlation ors: Elena Ringström Then to determine the strength of relationship between the variables we define coefficients on Pearson correlation for pairs of variables that are presented in table 4.2. Variables PRICE NI_S BVE_S FOREX_S HEDGE_S PRICE 1,000 0,773 0,529-0,077-0,132 NI_S 0,773 1,000 0,815-0,121-0,258 BVE_S 0,529 0,815 1,000-0,018-0,164 FOREX_S -0,077-0,121-0,018 1,000-0,351 HEDGE_S -0,132-0,258-0,164-0,351 1,000 PRICE. 0,000 0,000 0,136 0,066 NI_S 0,000. 0,000 0,042 0,001 BVE_S 0,000 0,000. 0,397 0,031 FOREX_S 0,136 0,042 0,397. 0,000 HEDGE_S 0,066 0,001 0,031 0,000. Table 4.2 The strength of relationship between the variables. Pallant (2010, p. 134) describes that tree levels of strength when speaking about a correlation between variables. Correlation is small if the correlation coefficient is between 0,10 and 0,29. Correlation is medium if the correlation coefficient is between 0,30 and 0,49. Correlation is large if the correlation coefficient is between 0,50 and 1,0. Net income is highly correlated to the share price (0,773). Furthermore, this correlation coefficient is statistically significant at 1% level. Change of accumulated foreign currency translation adjustment (FOREX_S) has small negative correlation to the share price. However, the correlation coefficient is not statistically significant. Change of the fair value of cash flow hedges (HEDGE_S) has small negative correlation to the share price (-0,132) thus it is higher than the correlation between share price and change of accumulated foreign currency translation adjustment. It is essential to emphasize that the correlation coefficient between share price and change of the fair value of cash flow hedges is statistically significant at 10% level. Totally, any of variables have a correlation coefficient that is equal or higher than 0,9 that could be a sign of multicollinearity. To verify that results of regression are not biased by multicollinearity that probably not evident from the correlation table, we do collinearity test (Pallant, 2010, p. 158). This test allows identifying such values as tolerance and variance inflation factor (VIF) those can also indicate multicollinearity. Tolerance is a factor that shows the extent of variability of the particular independent variable that is not explained by the other independent variables. VIF is calculated as 1 divided by tolerance. Multicollinearity occurs when tolerance is less than 0,10 and VIF above 10. According to the test of 23