1
Risk Assessment of Investors: Comprehensive Income
Reporting
Josha Sellam, 10425004 Master Accountancy & Control
University of Amsterdam Email: Josha1991@hotmail.com Supervised by: Dr R. Felleg University of Amsterdam Email: R.Felleg@uva.nl Thesis June 20, 2016 Word count: 13,362
Amsterdam Business School
2
Statement of Originality
This document is written by student Josha Sellam, who declares, he takes full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Abstract
This study provides new insight on the influence of reporting location of comprehensive income on the risk assessment of investors for on a random sample of 100 firms for the period 2006-2015. I examine if Accounting Standards Update 2011-5 (ASU 2011-5), which eliminates the possibility of reporting OCI in a statement of changes in equity and makes it mandatory to report OCI in a performance statement has an effect on the risk assessment of investors. I argue that firms that already reported OCI in a performance statement before ASU 2011-5 experience no effect on the risk assessment of investors and that firms that reported OCI in a statement of changes in equity experience a negative effect on the risk assessment of investors. My results show that firms that already reported OCI in a performance statement before ASU 2011-5 experience no effect on the risk assessment of investors. However, my results also show that firms that reported OCI in a statement of changes in equity before ASU 2011-5 experience no negative effect on the risk assessment of investors like I predicted, but a positive effect. Furthermore, I find no significant evidence that OCI is related with share price. This is not surprising, because share price is influenced by a lot of other factors which could have let to these results.
Keywords: Reporting location, Comprehensive income, OCI, Volatility, Risk assessment, Investors, Accounting Standards Update 2011-5
4 Table of content
1. Introduction ... 5
2. Literature Review ... 8
2.1 Comprehensive Income ... 8
2.2 Risk Relevance and Income Reporting ... 9
2.3 Hypotheses ... 13
3. Sample and Methodology ... 15
3.1 Sample ... 15
3.2 Methodology ... 16
4. Results ... 20
4.1 Income volatility ... 21
4.2 Association between income volatility and market risk ... 25
4.3 Association between income volatility and stock price ... 32
4.4 Robustness tests ... 36
4.5 Hypotheses results ... 37
5. Conclusion ... 39
5 1. Introduction
Comprehensive income reporting has been a controversial topic for the last decades. From 1997 it became mandatory to report comprehensive income in a full set of general-purpose financial statements. The FASB issued the Statement of Financial Accounting Standards No. 130 (SFAS 130) ‘reporting of comprehensive income’, which established standards for the reporting and display of comprehensive income and its components. The controversy around comprehensive income is related to volatility and the perception of increased risk. The FASB describes comprehensive income as follows: A measure of all changes in equity of an entity that result from recognized transactions and other economic events of the period other than transactions with owners in their capacity as owners (FASB, 1997). SFAS 130 defines comprehensive income as the sum of net income and other comprehensive income, where other comprehensive income consists of revenues, expenses, gains and losses that are excluded from net income and are consistent with the classifications needed. SFAS 130 names three options for reporting OCI: in an income statement, in a separate statement or in the statement of changes in equity. From now on the options income statement and separate statement are referred to as performance statement.
Prior literature has shown that the risk assessment of investors for a firm is influenced by the volatility of an income measure. Bamber et al. (2010) state that if changes in OCI are shown in the income statement this will negatively affect the share price of that company, because OCI is volatile which increases risk for investors. Koonce et al. (2005) show that financial statement users perceive uncontrollable items as increasing risk, thus the relatively uncontrollable nature of OCI items are likely to exacerbate users’ perceptions of firm risk which lowers share price. This was also the biggest argument against mandating firms to report comprehensive income in a performance statement. OCI is transitory and therefore more volatile than net income, which increases investors’ assessment of firm risk. However,
6
the reporting location is also off influence in the risk assessment of investors. Hirschleifer and Teoh (2003) argue that more salient reporting of volatile, transitory and incomplete information like reporting OCI in a performance statement will increase users’ appraisal of the volatility of the firm’s performance than firms that report OCI in a statement of changes in equity. They argue that users are better able to incorporate information, which is more salient and fail to fully incorporate relevant information that is less visible. Based on the argumentation of Hirschleifer and Teoh (2003) can assume that if managers report comprehensive income in a more salient way, users will put more weight on comprehensive income as a performance measure than if comprehensive income is reported in a less salient way, which will also result in users assessing the firm’s performance as more volatile. These arguments show that the risk assessment of investors is higher for firms that report OCI in a performance statement than firms that report OCI in a statement of changes in equity.
In 2011 the FASB issued Accounting Standards Update 2011-5 (ASU 2011-5) ‘presentation of comprehensive income’. ASU 2011-5 eliminated the option of reporting OCI in a statement of changes in equity and made it mandatory to report OCI in a performance statement. The FASB issued ASU-2011-5 to increase transparency of the figures presented by companies to increase decision usefulness for the assessment of comprehensive income by the investors.
The objective of this paper is to examine what kind of effect ASU 2011-5 had on the risk assessment of investors and if this update in fact increased decision usefulness for the assessment of comprehensive income by investors. The effects on risk assessment by investors after issuing ASU 2011-5 is examined for the two types of reporting OCI under SFAS 130. Therefore the following research question is formulated:
7
The research question is being investigated in an archival, quantitative study. Based on prior literature I hypothesize that ASU 2011-5 has no effect on the risk assessment of investors for companies that reported other comprehensive income in a performance statement under SFAS 130 and has a negative effect on the risk assessment of investors for companies that reported other comprehensive income in a statement of changes in equity under SFAS 130.
I test my expectations on a random sample of 100 NASDAQ listed firms during the period 2006-2015, where 2006-2010 is the period of reporting OCI under SFAS 130 and 2011-2015 is the period of reporting OCI after ASU 2011-5 is issued. Consistent with my first hypothesis I find that firms that reported OCI in a performance statement under SFAS 130 experience no effect on the risk assessment of investors. However, for my second hypothesis I find that firms, which reported OCI in a statement of changes in equity under SFAS 130 experience a positive effect on the risk assessment of investors when they change their reporting location of OCI to a performance statement under ASU 2011-5. This result is contradictory to prior literature and also inconsistent with my second hypothesis, which is therefore rejected.
My paper contributes to existing literature by examining the effects of reporting location on the risk assessment of investors for comprehensive income reporting. Previous literature has examined these topics separately from each other but never in a combined research project. Because of ASU 2011-5 the effects from comprehensive income reporting location on the risk assessment of investors is even more relevant. The accounting update makes it possible to examine what kind of effect the update of comprehensive income reporting has on the risk assessment of investors for companies that used different reporting locations of comprehensive income under the old standard SFAS 130.
8
This paper is organized in the following chapters. First, chapter two provides the literature review and the hypotheses development. In Chapter three, I discuss the sample selection procedures and methodology. In chapter four, I present the descriptive statistics, the results from the regression models and the results for the hypotheses. And finally, chapter five provides the conclusion.
2. Literature Review 2.1 Comprehensive Income
In 1997, the Financial Accounting Standards Board (FASB) issued the Statement of Financial Accounting Standards No. 130 (SFAS 130) (1997), ‘Reporting Comprehensive Income’. SFAS 130 established standards for the reporting and display of comprehensive income and its components in a full set of general-purpose financial statements. SFAS 130 shows that all items are required to be recognized under accounting standards as components of comprehensive income be reported in a financial statement that is displayed with the same importance as other financial statements (FASB, 1997).
The FASB describes comprehensive income as follows: A measure of all changes in equity of an entity that result from recognized transactions and other economic events of the period other than transactions with owners in their capacity as owners (FASB, 1997). Comprehensive income is defined by SFAS 130 as the sum of net income and other comprehensive income. Other comprehensive income (OCI) consists of revenues, expenses, gains and losses that are excluded from net income and are consistent with one of four classifications: [1] foreign currency translation adjustments, [2] available-for-sale marketable securities adjustments, [3] minimum required pension liability adjustments, and [4] adjustments on derivative securities that qualify for cash flow or foreign currency hedge accounting treatment (Chambers et al., 2007). Hodder (2006) states that comprehensive
9
income is a more complete measure of firm performance than net income because comprehensive income also includes fair value changes.
SFAS 130 names three options for reporting OCI: [1] in an income statement that includes both the component and totals of net income and comprehensive income, [2] in a separate statement of comprehensive income that begins with net income, reports each component of other comprehensive income and ends with total comprehensive income, and [3] in the statement of changes in equity.
SFAS 130 lasted until 2011 when the FASB issued Accounting Standards Update 2011-5 (ASU 2011-5) ‘Presentation of Comprehensive Income’. ASU 2011-5 eliminates option three of SFAS 130, thus the option of reporting OCI in the statement of changes in equity was eliminated and it was required to report comprehensive income in a performance statement. FASB issued ASU-2011-5 to increase transparency of the figures presented by companies to increase decision usefulness for the assessment of comprehensive income by the investors.
Therefore, under SFAS 130 firms were able to report OCI in a statement of changes in equity and a performance statement. This lasted until 2011, when ASU 2011-5 was issued and the option of reporting OCI in a statement of changes in equity was eliminated.
2.2 Risk Relevance and Income Reporting
Bamber et al. (2010) state that rational investors fully process all the information regardless of its location, which means that the location of reporting comprehensive income would not matter for investors. The FASB, however, stated a preference for the first and second option of OCI reporting, so they stated a preference for reporting OCI in a performance statement and not in the statement of changes in equity. Managers also think the location of reporting OCI matters, because although the numbers are presented in the same, most of the managers choose to present OCI in the statement of changes in equity. So
10
although FASB stated a preference for reporting OCI in a performance statement, many companies ignored that preference and kept reporting OCI in the statement of changes in equity. Many managers believed reporting comprehensive income in a salient performance statement would lead investors and other stakeholders to increase their assessments of the volatility of the firm’s performance, which would lead to a decreasing share price. Therefore Bamber et al. (2010) hypothesize that managers with more powerful equity-based incentives and managers with less job security would be less likely to report comprehensive income in a performance statement. Evidence that managers’ personal equity-based incentives and job security concerns are associated with comprehensive income reporting location choices complements the Lee et al. (2006) evidence that cherry-picking sales of AFS securities is associated with reporting OCI in a statement of changes in equity in the insurance industry. Lee et al. (2006) provide evidence that insurers who report comprehensive income in a statement of equity are more likely to smooth earnings by cherry-picking realized gains and losses on available-for-sale securities. But Chambers et al. (2007) have a counterargument and find evidence that investors put more weight on OCI reported in a statement of changes in shareholders’ equity then OCI reported in a performance statement.
The greatest part of OCI consists of unrealized gains and losses relating to investments, foreign currency fluctuations, and derivate hedges (Chambers et al., 2007). Meaning that OCI consists of transitory income and is very volatile. This was also an argument against the issuance of SFAS 130. Most of OCI are either transitory or derive from noisy market price movements that do not truly reflect fundamental changes in a firm’s assets and liabilities (Black, 2016). Ohlson (1999) states that there are three properties of which at least two must exist before a financial statement item is truly transitory: [1] Inability to predict itself, [2] Irrelevance for forecasting next-period abnormal net comprehensive income, and [3] Value irrelevance. The following papers have examined these properties and most of them
11
find evidence that supports the notion that comprehensive income is transitory. [1] Jones and Smith (2011) find evidence which indicates that comprehensive income and OCI are less predictable than net income. [2] Barton et al. (2010) measure eight performance measures to predict operating cash flows and find that comprehensive income has the lowest ability to predict operating cash flows. [3] Dhaliwal et al. (1999) show in their paper that comprehensive income has less explanatory power on share prices than does net income. However Goncharov and Hodgson (2011) find that OCI and comprehensive income are in fact price-relevant, but still less price-relevant than net income. In addition, Kanagaretnam et al. (2009) find evidence that OCI is significantly correlated with stock prices. So point 1 and 2 are proven and point 3 has mixed results. However, this is enough to mark OCI as transitory. (Black, 2016) This indicates that OCI is a volatile income measure and hard to predict for future assessments which increases the risk factor for investors, and this risk factor bares weight in the valuation of the share price.
Bamber et al. (2010) state that if changes in OCI are shown in the income statement this will negatively affect the share price of that company, because OCI is volatile. Meaning an increased risk assessment of investors. Ryan (2012, p. 2) has the following definition of risk: “a random variation in firms’ future economic performance given currently available information”. Koonce et al. (2005) show that financial statement users perceive uncontrollable items as increasing risk, so the relatively uncontrollable nature of OCI items likely exacerbate users’ perceptions of firm risk which lowers share price. This was the biggest argument against mandating firms to report comprehensive income in a performance statement, OCI is transitory and therefore more volatile than net income, which increases investors’ assessment of firm risk. In fact 34 percent of the comment letters on FASB’s exposure draft (1996) argue that comprehensive income is more volatile than net income, and therefore might give a wrong presentation of firm performance to investors (Yen et al., 2007).
12
Hirschleifer and Teoh (2003) argue that more salient reporting of volatile, transitory and incomplete information will increase users’ appraisal of the volatility of the firm’s performance. They argue that users are better able to incorporate information that is more salient and fail to fully incorporate relevant information that is less visible. Based on the argumentation of Hirschleifer and Teoh we can conclude that if managers report comprehensive income in a more salient way, users will put more weight on comprehensive income as a performance measure than if comprehensive income is reported in a less salient way, which will also result in users assessing the firm’s performance as more volatile. This is shown through the study of Yen et al. (2007), where they find that over 80 percent of the comment letters on the 1996 Exposure Draft expressed concern about the reaction of financial statement users to mandatory performance reporting. Almost half of the comment letters expressed concerns over the volatility of comprehensive income. Also a lot of the comment letters stated that performance reporting would increase the perceived risk by investors and therefore companies would change their operations to reduce the volatility of comprehensive income to eliminate the negative user reactions (Yen, Hirst, & Hopkins, 2007).
Different papers also show that financial statement users assess a firms’ performance as more volatile when comprehensive income is reported in a performance statement, so being more salient. Maines and McDaniel (2000, p. 179) state the following: “investors’ judgments of corporate and management performance reflect the volatility of comprehensive income only when it is presented in a statement of comprehensive income”. In addition, Hirst and Hopkins (1998) state that half of the experienced financial analyst participants of their experiment did not notice the term comprehensive income when this was reported in the statement of equity. Therefore the evidence of these papers suggests that managers believe the location of reporting comprehensive income affect investors’ perception.
13
Furthermore, banks criticized the proposal of the FASB for comprehensive income measurement and reporting in the mid-1990s because it would fail to fully reflect the risk-management activities of the banks, it would increase the volatility in comprehensive income, and increase the perceived risk assessment of investors (Hodder et al., 2006). Respondents to the FASB’s exposure draft (1996) argued that the items, which are identified as other comprehensive income are not related to the actual firm performance and therefore to include them in a performance statement would be confusing and misleading to investors (FASB, 1997, par. 60). However, Black (2016) states that researchers assume investors incorporate risk-relevant information into share prices, and that these share prices represent investors’ future cash flows.
Finally, the paper of Schaberl et al. (2015) examines whether OCI becomes more value relevant after the Accounting Standards Update in 2011. Their results show that OCI is more relevant when it is reported in a less salient way and that the value relevance decreases when the place is changed to a more salient way. These results contradict the FASB’s expectation and prior archival evidence that I explained above.
2.3 Hypotheses
Based on the literature described in the previous chapters, I formulate two hypotheses. Prior literature shows that OCI is a more volatile and transitory income measure than net income and therefore hard to predict, which increases the risk assessment of investors. Bamber et al. (2010) and Koonce et al. (2005) show that the volatile nature of OCI leads to a higher risk assessment of investors. And Kanagaretnam et al. (2009) find evidence that OCI is significantly correlated with stock price, therefore an increasing risk assessment of investors will also result in a decreasing share price. Also of influence on the risk assessment of investors is the reporting location. A more salient way of reporting increases the risk assessment even more compared to a less salient way of reporting (Hirschleifer & Teoh,
14
2003). Maines and McDaniel (2000) even state that investor assessment of a firm’ performance only reflect the volatility of comprehensive income when it is reported in a performance statement. Therefore I hypothesize that ASU 2011-5 will have no effect on the risk assessment of investors for companies that already reported comprehensive income in a more salient way like a performance statement under SFAS 130. Because they already reported OCI in a performance statement and therefore don’t need to change anything in their reporting format, there will be no change in the risk assessment of investors.
H1
Accounting Standards Update 2011-5 has no effect on the risk assessment of investors for companies that reported other comprehensive income in a performance statement under SFAS 130.
But for the firms that reported OCI in a statement of changes in equity under SFAS 130 I hypothesize that ASU 2011-5 will have a negative effect on the risk assessment of investors. This is because they reported OCI in a less salient way which they need to change to a more salient way of reporting. This will lead investors and other stakeholders to increase their risk assessments of the firm, which could potentially lead to a decreasing share price. H2
Accounting Standards Update 2011-5 has a negative effect on the risk assessment of investors for companies that reported other comprehensive income in a statement of changes in equity under SFAS 130.
15 3. Sample and Methodology
3.1 Sample
My sample comes from Compustat where I collected two sample periods: pre-update 2011-5 2005-2010 and post-update 2011-5 2011-2015. I have two samples of five years each because there were five full years after ASU 2011-5 was issued. I merged these sample periods together to get one sample for the entire period, which was 17,852 companies and 123,877 firm year observations and I dropped all the duplicates in the sample. To account for these two periods I included a dummy variable for the period of reporting, so 0 for pre-update which is 2006-2010 and 1 for post-update which is 2011-2015. Then I collected a sample from CRSP for the variable Beta. I also merged this sample with the sample from Compustat. Table 1 describes the sample selection and the steps of dropping observations to come to the final sample for my tests.
Table 1
Sample and elimination of observations Sample
Firms
Observations Eliminations Entire sample from Compustat and CRSP 17,852 123,877
Dropping duplicates 111,700 -12,177
Dropping stock exchange codes other than Nasdaq (14) 29,922 -81,778
Dropping SIC 4900-4999 and 6000-6999 22,284 -7,638
Dropping missing observations 13,549 -8,735
Dropping firms with missing years 8,801 -4,748
Final sample 880 8,801
Random selection sample 100 1,000
The variable Beta was only available for firms that were Nasdaq listed or listed on the New York Stock Exchange (NYSE). I collected both NASDAQ and NYSE Beta variables and merged these in my sample. However my sample only contained full period samples for firms that were NASDAQ listed, the NYSE firms all has missing observations which resulted in incomplete period samples. therefore the only full period samples that remained were firms that are NASDAQ listed, which is stock exchange code 14. Therefore I have to drop all the firms that have different stock exchange codes. I also add the risk-free rate of return (one
16
month treasury bill rate) from Dataset List: Fama French, Momentum, and Liquidity. I use the risk-free rate of return at the end of the year, thus month 12 of each year. And finally I add the share prices of the firms from CRSP and because I only use annual data for my sample I use the year-end share price of every firm.
Consistent with prior studies, I eliminate utilities (SIC codes 4900–4999) and financial firms (SIC codes 6000–6999) because they are subject to additional regulation (Schaberl & Victoravich, 2015). After dropping firm years with missing observations for all the variables. In order to run my tests correctly, every firm needs to have the full period of observations. So 5 years pre-update and 5 years post-update, therefore I drop the firms that have missing years. Leading to a final sample of 880 firms and 8801 firm years. To complete my sample I have to add the reporting location of comprehensive income for all the firms manually. Consistent with Schaberl and Victoravich (2015), I use the K-10 forms from the SEC website just like. This data is obtained manually, wherefore I slimmed down my sample to make it more manageable. Which is done by random selection in order to maintain a reliable sample. Before I made the random selection I winsorized the data on the 1 percent and 99 percent level in order to loose extreme outliers which might distort the sample. I let Stata make a random selection of 100 companies and 1,000 company years out of the remainder of the companies and observations in my sample. For the rest of the paper I will use the random sample for all the tests.
3.2 Methodology
In this section I describe the methodology that is used to test the hypotheses for this quantitative archival study. Given the nature of the panel data, which is not normally distributed I conduct a Generalized Least Squares (GLS) regression model. This model is a random effects model, which assumes no fixed effects. These fixed effects in the dataset are eliminated after the data is clustered by year.
17
In my paper I test if the risk assessment of investors is affected by the reporting location of OCI, therefore I check if the income measures (comprehensive income and net income) have any effect on the risk assessment. I will do this in steps. First, I check for the effects on market risk (total risk and systematic risk) and then I check if these income measures have an effect on share price. In order to determine whether the volatility of NI and CI is captured by market risk I follow Khan and Bradbury (2014) and estimate the following regression models:
MRj = 0 + 1DEj + 2CFj + 3NIj + 6DUM_RL + 7DUM_PER + j (1) MRj = 0 + 1DEj + 2CFj + 4CIj + 6DUM_RL + 7DUM_PER + j (2) MRj = 0 + a1DEj + 2CFj + 3NIj + 5(CIj – NIj) + 6DUM_RL + 7DUM_PER + j (3)
In these regressions MR is the market risk measure for firm j. Two measures of market risk are used as a proxy for the risk assessment of investors: stock return volatility and Beta. Stock return volatility measures total risk and Beta measures systematic risk. Stock return volatility is the standard deviation of annual raw returns over the period 2006-2015. NIj and
CIj are volatility measures of NI and CI. These variables are estimated using the standard deviation of annual data over the period 2006-2015 (Khan & Bradbury, 2014). I also include two dummy variables, one for the type of reporting a company uses pre and post ASU 2011-5, which will be 1 if the firm avoids reporting comprehensive income in a performance report and instead reports in a statement of equity, and 0 if a firm reports comprehensive income in a performance report pre ASU 2011-5. Showing the effects ASU 2011-5 has on both types of reporting comprehensive income. And the second is a dummy variable for the period of reporting, so 0 for pre ASU 2011-5, which is 2006-2010 and 1 for post ASU 2011-5, which is 2011-2015.
Debt-to-equity ratio (DE) and operating cash flow-to-current liability ratio (CF) are accounting risk-based measures which are included to eliminate other accounting variables
18
which might have an impact on market risk and distort the results. By including these accounting risk variables I follow Khan and Bradbury (2014) who use prior research that has examined the relationship between accounting-based risk measures and beta (e.g. Beaver et al., 1970; Hamada, 1972; Bowman, 1979; Goh & Emanuel, 1981).
Eqs. (1) and (2) test if the income volatility measures are associated with market risk after checking for the accounting risk variables DE and CF. If that is the case and income volatility is related to market risk then 3 and 4 should be positive and significant. Eq. (3) tests whether the incremental volatility of CI (over NI) is associated with market risk. If so, then a5 should be positive and significant. For the Dummy variables DUM_RL and DUM_PER I expect to find negative significant coefficients for 6 and 7 in models (1), (2) and (3) (market risk), which would indicate that firms that reported OCI in a statement of changes in equity experience higher market risk post-ASU 2011-5 than pre-ASU 2011-5 and that firms that already reported OCI in a performance statement experience no effect on market risk. These results would indicate that both hypotheses are accepted.
To test whether volatility of income is an element of risk that decreases share prices, I adopt the Hodder et al. (2006) model, which is a simplified version of the residual-income model (Ohlson, 1995). Eq. (4) is the benchmark model before introducing any income volatility measures, on a pooled sample over 2006-2015:
Pj = 0 + 1BVEj + 2AEj + 8DUM_RL + 9DUM_PER + j (4) Where, DUM_RL and DUM_PER are previously defined.
In this equation, P is the price per share for firm j at the end of t; BVE is the book value of equity per share at the end of year t; and AE is the abnormal earnings per share for period t, which is used as a proxy for expected future abnormal earnings. I measure abnormal earnings as current period earnings (scaled by the number of shares outstanding) less the product of the risk-free rate of return at the beginning of year t, times book value per share at
19
the beginning of year t. I use the risk-free rate for this computation to allow coefficient estimates to capture the effects of risk. Hodder et al. (2006, p. 358) states that the coefficient estimate on BVE (1) should equal one, but omitted variables (e.g., off-balance sheet net asset values and gains or losses that have not yet been realized or recognized) may induce this coefficient estimate to deviate from one. I expect the coefficient on AE (2) to be positive, reflecting market capitalization of abnormal earnings. Because the simplified model only includes abnormal earnings for the current period (rather than all future periods). This coefficient should reflect the cross-sectional mean share-price effects of persistence, risk, growth, and other information in abnormal earnings (Hodder et al., 2006). For the dummy variables DUM_RL and DUM_PER I expect to find positive significant coefficients for 8 and 9 which would indicate that firms that reported OCI in a performance statement experienced less market capitalization than firms that reported OCI in a statement of changes in equity. This would indicate that firms that changed their reporting location from a statement of changes in equity to a performance statement experienced a decrease in share price and that the firms that already reported OCI in a performance statement experience nog effect on share price. Therefore both hypotheses would be accepted.
And to examine whether the capital market prices reflect income volatility, I interact each accounting risk measure with abnormal earnings and estimate the following models (Khan & Bradbury, 2014):
Pj = 0 + 1BVEj + 2AEj + 3(DEj x AEj) + 4(CFj x AEj) + 5(NIj x AEj) + 8DUM_RL +
9DUM_PER + j (5)
Pj = 0 + 1BVEj + 2AEj + 3(DEj x AEj) + 4(CFj x AEj) + 6(CIj x AEj) + 8DUM_RL +
9DUM_PER + j (6)
Pj = 0 + 1BVEj + 2AEj + 3(DEj x AEj) + 4(CFj x AEj) + 5(NIj x AEj) + 7[(CIj - NIj) x
20
Where, P, BVE, AE, DE, CF, NI,CI, DUM_RL and DUM_PER are previously defined. I interact the accounting-based risk measures (DE and CF) with abnormal earnings and predict negative coefficients for 3 and 4, suggesting the market assigns a lower capitalization multiple to the abnormal earnings of firms with higher accounting risk. I use model (7) to investigate whether the incremental volatility of CI (over NI) is priced. A significant negative coefficient for 7 indicates the marginal volatility of comprehensive income (over net income) is associated with elements of risk that are priced by the market. For the dummy variables DUM_RL and DUM_PER I expect to find positive significant coefficients for 8 and 9 in models (5), (6) and (7) which would indicate that firms that reported OCI in a performance statement are associated with elements of risk that are priced by the market. This would indicate that firms that report OCI in a performance statement experience a higher risk assessment of investors than firms that report OCI in a statement of changes in equity. Therefore both hypotheses would be accepted.
4. Results
This chapter describes the outcomes of the regression analyses. The chapter is divided into three parts: income volatility, association between income volatility and market risk and association between income volatility and share price. These tests will show if the income volatility measures are in fact associated with market risk and share price. So first I will run tests and discuss the results from the regression models. Then I will use those results to answer the two hypotheses. And I will finish with the robustness test.
21 4.1 Income volatility
Table 2
Descriptive Statistics for Income Measures and Selected Components of Income (n=100)
Variable Mean Std. Dev. Min 1st. Quart Median 3rd Quart Max
NI 0.0287 0.2004 -0.0443 0.0000 0.0001 0.0021 1.1111 FCT -0.0001 0.0018 -0.0055 0.0000 0.0000 0.0000 0.0043 DGL 0.0000 0.0156 -0.0664 0.0000 0.0450 0.0010 0.1169 EPB 0.0007 0.0036 0.0000 0.0000 0.0000 0.0002 0.0132 SGL 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Other 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 OCI -0.0007 0.0189 -0.0660 0.0000 0.0000 0.0000 0.0387 CI 0.0288 0.2064 -0.0530 0.0000 0.0001 0.0021 1.127
NI is net income. FCT is foreign currency translation adjustments. DGL is derivatives gains/losses. EPB is employee pension benefits. SGL is securities gains/losses. Other is any other adjustment to arrive at CI. OCI is total other comprehensive income and CI is comprehensive income, where all variables have been scaled by opening market equity.
Table 2 provides descriptive statistics for the income measures net income (NI), comprehensive income (CI) and the components of other comprehensive income (OCI). These variables are scaled as a percentage of opening market value of equity. The statistics in table 2 indicate that the pooled 10-year means (medians) of NI and CI have indistinguishable magnitudes: 0.0287 (0.0001) for NI versus 0.0288 (0.0001) for CI. However the mean (median) of OCI -0.0007 (0.0000) differs substantially from NI and CI, which indicates that the magnitude of OCI is substantially smaller. This is a result of the low values from the OCI components. And compared to the papers of Khan and Bradbury (2014) and Hodder et al. (2006) the values of the OCI components from my sample are considerably lower.
22
Figure 1
Time Series sample period of Income Measures
NI is cross-sectional mean of net income scaled by the opening market value of equity each year and CI is cross-sectional mean of comprehensive income scaled by the opening market value of equity each year.
Figure 1 shows the time series of the income measures NI and CI through the sample period 2006-2015. It shows that CI is greater than NI in 2009 and 2010 and lower in 2006, 2010, 2013 and 2015 but overall the mean values of NI and CI are indistinguishable over the sample period just like table 2 shows.
Table 3
Descriptive statistics and comparative analyses of measures of income volatility
Variable Mean Std. Dev. Min 1st. Quart Median 3rd Quart Max Panel A: Descriptive statistics for firm-specific measures of income volatility (i.e. standard deviation over the time period (2006-2015) (n=100)
NI 0.0338 0.1339 0.0000 0.0004 0.0027 0.0108 0.8998
CI 0.0349 0.1428 0.0000 0.0004 0.0027 0.0106 0.9747
Panel B: Descriptive statistics of standard deviation ratio: standard deviation of CI/standard deviation of NI (n=100) CI/NI 1.0325 1.0665 0.0000 1.0000 1.0000 0.9815 1.0832 0,0000 0,0100 0,0200 0,0300 0,0400 0,0500 0,0600 0,0700 0,0800 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Pooled over time Per ce n t NI CI
23
Table 3, continued
Comparisons Count
Percent of total firms Panel C: Firm-specific test to compare measures of income volatility over the period
2006-2015 (n=100)
CI/NI>1 48 48%
CI/NI=1 2 2%
CI/NI<1 50 50%
Panel D: Statistical test of standard deviation ratio
Wilcoxon-signed rank test:
p-Value(one tailed) 0.2858
Estimated median 1.067
NI is the firm-specific standard deviation of NI measured over the period 2006-2015 CI is the firm-specific standard deviation of CI measured over the period 2006-2015
NI and CI are net income and comprehensive income, each scaled by the opening market value of equity.
To measure the volatility of the income measures the descriptive statistics of the firm-specific standard deviations (income volatility measures) of NI and CI across the 100 firms over the years 2006-2015 are reported in Table 3 Panel A and are graphically outlined in Figure 2. The firms experience a standard deviation in net income of 0.0338 and a standard deviation in comprehensive income of 0.0349. The left side of Figure 2 shows the means of the firm-specific standard deviations of the two income measures, which are also shown in Table 3 Panel A. The middle part of Figure 2 shows the annual standard deviations for the firms over the entire sample period and the right part shows the pooled standard deviations over the entire sample period. Figure 2 shows that the volatility of the income measures net income and comprehensive income is almost indistinguishable over the sample period. Firm specific and pooled over time standard deviations are almost the same and the annual observations of standard deviation are for two of the years higher for comprehensive income, for three years higher for net income and in the five remaining years are indistinguishable. These results indicate that comprehensive income is more volatile than net income, which is shown through the higher mean standard deviation of comprehensive income and the higher max standard deviation in table 3. The standard deviation ratio in Panel B shows that the mean of CI to NI is 1.0325, which means that comprehensive income is 3.25% more volatile
24
than net income. This is also shown in figure 2 where the standard deviations between 2009 and 2011 have bigger changes in value than does net income.
Figure 2
Time series of standard deviations of income measures
Figure 2 shows the Standard deviations of income measures NI and CI in cross-section and over time
(n=100). This figure shows the standard deviations of NI and CI scaled by the opening market value of equity each year and computed over the entire sample period 2006-2015. The left side shows the means of the firm specific standard deviations and the middle of the figure shows the cross-sectional average standard deviations for each year of the sample period and the right side shows the pooled standard deviation over the entire sample period.
I Test NI and CI for normality with the Jarque-Bera test for normality. For a variable to be normally distributed the P-value needs to exceed 0.10 to not reject the null hypothesis, that the variable is normally distributed. The P-values of the Jarque-Bera test for normality NI and CI are both 0.000, indicating that the null hypothesis is rejected and the variables are not normally distributed. However there are no assumptions made based on the distribution of NI and CI since the GLS regression is not very sensitive to variables which are not normally distributed.
However the results from Table 3 Panel C don’t back this up. These results show that for the 100 firms from my sample, volatility of CI is greater than NI for 48 (48%) firms and lower for 50 (50%) firms. And there are 2 firms where the volatility of CI is the same as NI.
0,0000 0,0500 0,1000 0,1500 0,2000 0,2500 0,3000 0,3500 0,4000 0,4500
Firm specific 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Pooled over time
Stan d ar d D e vi ation NI CI
25
This shows that more firms experience higher volatility in net income than in comprehensive income. The explanation for the results from Table 3, Panel A in contrast to the results from Panel C is that the volatility of CI is larger for the companies that experience higher volatility in CI than NI than for the companies that experience higher volatility in NI than CI.
I use the Wilcoxon-signed rank test to test the equality of CI and NI. The results from this test are shown in Table 3, Panel D. The null hypothesis is that both distributions are the same. The results in Table 3 show that the p-value is 0.2858, which means that the null hypothesis can’t be rejected and that the distributions of CI and NI are the same.
4.2 Association between income volatility and market risk
Table 4, Panel A contains the descriptive statistics for the market-based (proxy for the risk assessment of investors) and accounting-based risk measures for my sample of firms for the period 2006-2015. The mean (median) for the volatility of stock returns is 0.5365 (0.4301) and the mean (median) for beta is 0.9729 (0.9468). The mean (median) of beta is below the market average beta of 1.000, which is an indication that the sample firms are on average less volatile than the market. The min and max of Beta show that the spread of volatility in the sample is big, there are firms that are not volatile at all but there are also firms that are way more volatile than the market average. The mean (median) of DE 2.4522 (2.8352), means that the mean (median) of debt-to-equity is approximately 245.2% (283.5%). And the mean (median) of CF is 0.4107 (0.4702), which means that the cash flow-to-current liabilities is approximately 41.1% (47%).
I Test SR, Beta, DE and CF for normality with the Jarque-Bera test for normality. For a variable to be normally distributed the P-value needs to exceed 0.10 to not reject the null hypothesis that the variable is normally distributed. The P-values of the Jarque-Bera test for normality for the variables SR, Beta, DE and CF are all 0.000, indicating that the null hypothesis is rejected and the variables are not normally distributed. There are no assumptions
26
made based on the distribution of SR, Beta, DE and CF since the GLS regression is not very sensitive to variables, which are not normally distributed.
Table 4
Descriptive statistics and correlation matrix of risk measures
Variable Mean Std. Dev. Min 1st. Quart Median 3rd Quart Max
Panel A: Descriptive statistics of risk measures (n=100)
SR 0.5365 0.3941 0.0594 0.2785 0.4301 0.6485 1.8826
Beta 0.9729 0.5904 -0.1429 0.5785 0.9468 1.3303 2.5376
DE 2.4522 2.5908 -4.3738 0.9617 2.8352 4.1973 8.0785
CF 0.4107 1.2002 -4.6778 0.1134 0.4702 0.8273 3.6868
NI CI SR Beta DE CF DUM_RL DUM_PER
Panel B: Pearson correlation matrix among risk measures (n=100) NI 1.000 CI 0.9978*** 1.000 SR 0.6717*** 0.7193*** 1.000 Beta -0.1846*** -0.1808*** -0.0858*** 1.000 DE 0.0050 -0.0019 -0.0684 0.0319 1.000 CF -0.1305*** -0.1109*** -0.1545*** -0.0239 0.0221 1.000 DUM_RL 0.0300 0.304 0.0259 0.1082*** 0.0653 -0.0002 1.000 DUM_PER 0.0000 0.0000 0.0000 -0.0495 0.0376 0.0237 -0.7996*** 1.000 *. **. *** Denotes p < .10, p < .05 and p < .01, respectively, all two-tailed.
NI is the firm-specific standard deviation of NI, each year measured over the period 2006-2015. CI is the firm-specific standard deviation of CI, each year measured over the period 2006-2015. NI and CI are net income and comprehensive income, each scaled by the opening market value of equity.
SR is firm-specific standard deviation of average annual stock returns, each year measured over the period 2006-2015.
Beta is beta, measured over the period 2006-2015. DE is the debt-to-equity ratio.
CF is the operating cash flow-to-current liabilities ratio. DUM_RL is a dummy variable for reporting location DUM_PER is a dummy variable for period
Table 4, Panel B contains the Pearson correlation statistics among the income volatility measures and the market-based risk measures. This matrix calculates the separate correlation coefficients between the variables used in the GLS regression. Issues regarding multi-collinearity of variables could arise when the correlation coefficients between two independent variables exceeds 0.8. This could disturb the data because one independent variable could be linearly predicted by the other independent variable, which could make the data unreliable. The income volatility measures NI and CI are highly correlated as
27
expected, comprehensive income consist of net income and the remainder is other comprehensive income. Therefore it is obvious that NI and CI are highly correlated (0.9978). But, these variables are not examined in the same regression model, therefore the multi-collinearity between NI and CI can be ignored.
The same applies for firm-specific standard deviation of average annual stock returns (SR) with the income volatility measures NI and CI. Investors use NI and CI in their assessment of a firm’s performance and future performance, which will result in an increase or decrease in the share price. So income measures are expected to positively and significantly correlate with stock return volatility. Suggesting that the income volatility measures are influenced by risk factors that relate to market risk. More specifically it shows that volatility in reported income and volatility in returns are the main measures of total risk. This confirms that the income measures NI and CI capture risk factors that relate to volatility in stock returns. These variables need to be checked for multi-collinearity because they are examined in the same regression model. The income volatility measures have a negative significant correlation with Beta, which is a measure of systematic risk, and indicates that the income volatility measures capture risk factors that relate to market risk and more specifically systematic risk.
The accounting-risk based measures DE and CF show very small correlations with the volatility measures, where DE doesn’t show any significant correlation and CF very small positive significant correlation. This does not match the results from Khan and Bradbury (2014, p. 82), who state the following: the negative correlation between accounting volatility and the accounting-based risk measures is consistent with debt levels and liquidity being determined simultaneously with the firms expected income volatility.
The last high significant correlation is between the dummies reporting location and period. This was expected because after the Accounting Standards Update 2011-5 it was
28
mandatory for all firms to report comprehensive income in a performance statement. Thus all the firms that reported comprehensive income in the statement of changes of equity before the update, changed their reporting location and the other firms kept reporting comprehensive income in the same reporting location. These variables need to be checked for multi-collinearity because they are examined in the same regression model.
For the variables with high significant correlations that are examined in the same regression model I perform a multi-collinearity check. I will perform a variance inflation factor (VIF) test which checks if these high correlations lead to multi-collinearity in the GLS regression. Which means that one independent variable could be considered as a linear combination of another independent variable. Multi-collinearity issues occur when VIF values exceed 5. Table 5 shows the results from the VIF test and indicates that there are no VIF values which exceed 5, therefore there are no multi-collinearity issues in the GLS regression.
Table 5
Multi-collinearity among independent variables (n=100)
Model 1 Model 2 Model 3
Variabel VIF Tolerance VIF Tolerance VIF Tolerance
DE 1.04 0.961538 1.04 0.961538 1.04 0.961538 CF 1.04 0.961538 1.04 0.961538 1.04 0.961538 NI 1.03 0.970874 - - 1.76 0.568182 CI - - 1.03 0.970276 - - CI - NI - - - - 1.72 0.581395 DUM_RL 2.70 0.37037 2.70 0.37037 2.70 0.37037 DUM_PER 2.73 0.3663 2.73 0.3663 2.73 0.3663 NI is the firm-specific standard deviation of NI each year measured over the period 2006-2015. CI is the firm-specific standard deviation of CI each year measured over the period 2006-2015. NI and CI are net income and comprehensive income, each scaled by the opening market value of equity.
DE is the debt-to-equity ratio.
CF is the operating cash flow-to-current liabilities ratio. DUM_RL is a dummy variable for reporting location DUM_PER is a dummy variable for period
29
Table 6
Regressions of Market Risk Measures on Income Volatility Measures and Other Accounting-based Risk Measures
Model Intercept DE CF NI CI NI-CI DUM_RL DUM_PER Wald chi² Adj-R² Panel A: Coefficients (z-statistics) from regression of volatility of stock returns (SR) on income volatility
and accounting-based risk measures (n=100)
(1a) 0.5088 0.0353 -0.1347 0.0043 - - -0.0239 -0.0034 134.74*** 0.1635 (11.60)*** (6.35)*** (-9.32)*** (3.47)*** (-0.50) (-0.07) (2a) 0.5077 0.0354 -0.1340 - 0.0044 - -0.0238 -0.0038 138.02*** 0.1669 (11.60)*** (6.37)*** (-9.30)*** (3.85)*** (-0.49) (-0.08) (3a) 0.5149 0.0343 -0.1355 -0.0008 - 0.0728 -0.0216 -0.0008 164.56*** 0.1928 (11.95)*** (6.27)*** (-9.55)*** (-0.51) (4.99)*** (-0.46) (0.02)
Panel B: Coefficients (z-statistics) from regression of beta on income volatility and accounting-based risk measures (n=100)
(1b) 0.9325 0.0284 -0.0165 -0.0081 - - 0.1485 0.0106 36.04*** 0.0477 (10.39)*** (3.26)*** (-0.73) (-4.17)*** (1.97)** (0.10) (2b) 0.9313 0.0285 -0.0161 - -0.0074 - 0.1483 0.0100 35.40*** 0.0469 (10.36)*** (3.27)*** (-0.71) (-4.10)*** (1.97)** (0.09) (3b) 0.9340 0.0279 -0.0161 -0.0093 - 0.0175 0.1493 0.0113 36.92*** 0.0469 (11.60)*** (3.19)*** (-3.67) (-3.67)*** (0.75) (1.98)** (0.12)
NI is the firm-specific standard deviation of NI, each year measured over the period 2006-2015. NI is the firm-specific standard deviation of CI, each year measured over the period 2006-2015. NI and CI are net income and comprehensive income, each scaled by the opening market value of equity.
SR is firm-specific standard deviation of average annual stock returns, each year measured over the period 2006-2015.
Beta is beta, measured over the period 2006-2015. DE is the debt-to-equity ratio.
CF is the operating cash flow-to-current liabilities ratio. DUM_RL is a dummy variable for reporting location DUM_PER is a dummy variable for period
MRj = 0 + 1DEj + 2CFj + 3NIj + 6DUM_RL + 7DUM_PER + j (1) MRj = 0 + 1DEj + 2CFj + 4CIj + 6DUM_RL + 7DUM_PER + j (2) MRj = 0 + a1DEj + 2CFj + 3NIj + 5(CIj – NIj) + 6DUM_RL + 7DUM_PER + j (3) *. **. *** Denotes p < .10. p < .05 and p < .01. respectively, all two-tailed.
Tabel 6 reports regression models (1), (2) and (3). Models (1) and (2) test if any of the two income volatility measures (NI and CI) provide incremental risk-relevant information beyond the accounting risk variables DE and CF. If that is the case then 3 and 4 should be positive and significant. Panel A tests stock return volatility, which is a measure of total risk and Panel B tests Beta, which is a measure of systematic risk. The coefficients in Panel A for
30
NI (3) and CI (4) are indeed positive and significant at the 5% level, with a p-value of 0.000 and 0.001. Indicating that both income volatility measures provide incremental risk-relevant information and in this case more specifically total risk. The models show an overall adjusted R-squared of 0.1635 and 0.1669, indicating that the models explain 16.35% and 16.69% of the total variance. These values are slightly lower than in earlier studies but high enough to draw conclusions from the regression models.
Model (3) tests whether incremental volatility in comprehensive income provides any market-risk-relevant information beyond the volatility in net income and the accounting risk variables DE and CF. If so, then 5 should be positive and significant. The results from model (3) in Panel A indicate that incremental volatility in comprehensive income indeed provides significant risk-relevant information for total risk in returns, beyond net income volatility. DE and CF. 5 Is significant at the 1% level with a p value of 0.000 and the model shows an overall adjusted R-squared of 0.1928, indicating that the model explains 19.28% of the total variance. This is also lower than the overall adjusted R-squared of comparable studies but high enough to draw conclusions from the regression models. To answer the hypotheses I incorporate two dummy variables, DUM_RL tests if reporting location is of influence on total risk. However, the coefficients of DUM_RL are not significant which means that reporting location of OCI has no influence on total risk. DUM_PER tests if the reporting period is of influence on total risk. The coefficients from DUM_PER are not significant, indicating that the period of reporting has no influence on total risk.
The coefficients in Panel B for models (1) and (2). NI (3) and CI (4) are not positively significant but negatively significant at the 1% level. Meaning that neither income volatility measures provide incremental systematic risk-relevant information. The results from model (3) in Panel B do not provide significant results for 5, which indicates that incremental volatility in comprehensive income does not provide systematic risk-relevant
31
information. Therefore all three models in Panel B fail to provide systematic risk-relevant information for the income volatility measures.
However, to answer the hypotheses, DUM_RL tests if reporting location is of influence on systematic risk. The coefficients of DUM_RL are significant for all three models at the 5% level which means that reporting location is of influence on systematic risk. And because the coefficients are positive it means that the firms with 0 as a dummy for reporting location have lower systematic risk. DUM_PER is not significant which means that the period of reporting no influence has systematic risk, so pre, or post-SFAS130 comprehensive income reporting has no influence on market risk. These two results mean that firms that reported OCI in a performance statement under SFAS130 had lower systematic risk than firms that reported OCI in a statement of changes in equity. Firms that reported OCI in a statement of changes in equity pre-SFAS130 and which changed the reporting location of OCI because of ASU 2011-5 experienced a decrease in systematic risk. These results are substantiated by the Wald Chi2 test, which tests if the dummy variables DUM_RL and DUM_PER create a statistically significant improvement to the fit of the model. The null hypothesis is that both the coefficients for the variables DUM_RL and DUM_PER are simultaneously equal to zero. Table 7 shows the results of the Wald Chi² test on models (1), (2) and (3). These results indicate that the null hypothesis for model (1b), (2b) and (3b) is rejected with a P-value of 0.10, indicating that the coefficients DUM_RL and DUM_PER are not simultaneously equal to zero, meaning that including these variables creates a statistically significant improvement in the fit of the model. However for the models (1a), (2a) and (3a) the null hypothesis can’t be rejected, meaning that the variables DUM_RL and DUM_PER don’t improve the model.
32
Table 7
Testing the effect of the dummy variables DUM_RL and DUM_PER on the dependent variables Stock Return Volatility and Beta
Model Chi² P-value
Wald Chi² test (two sided)
1a 0.52 0.7700 2a 0.50 0.7785 3a 0.60 0.7427 1b 5.11 0.0775 2b 5.10 0.0781 3a 5.84 0.0539
DUM_RL is a dummy variable for reporting location DUM_PER is a dummy variable for period
The adjusted R-squared of the three models in Table 6 Panel B shows a very small value, which indicates that it would be doubtful to draw conclusions from these regression models.. These adjusted R-squared values also don’t correspond with the adjusted R-squared values from the papers of Hodder et al. (2006) or Khan and Bradbury (2014). Because of the low adjusted R-squared I look at the probability Wald Chi² statistic in the GLS regression output. This statistic tests the overall significance of the regression. The probability Wald Chi² statistic for model (1b), (2b) and (3b) is 0.0000 for all the models which means that the null hypothesis is rejected. The null hypothesis is that all of the regression coefficients are equal to zero and that nothing is going on. These results indicate that it is perfectly fine to draw conclusions from these regression models.
4.3 Association between income volatility and stock price
Table 8. Panel A contains the descriptive statistics for share price, book value, abnormal earnings and the interactions terms with the variables from chapters 4.1 and 4.2. Share price shows a mean (median) of 30.5597 (13.8950). The mean (median) for book value of equity per share is 2.5199 (2.1700) and the mean (median) for abnormal earnings per share is 3.6800 (3.5763).
I test P, BVE and AE for normality, with the Jarque-Bera test for normality. For a variable to be normally distributed, the P-value needs to exceed 0.10 to reject the null
33
hypothesis that the variable normally distributes. The P-values of the Jarque-Bera test for normality for the variables P and BVE are both 0.000, indicating that the null hypothesis is rejected and the variables are not normally distributed. The P-value for AE exceeds 0.10 (0.8277) and is therefore, assumed to be normally distributed. There are no assumptions based on the distribution of P and BVE since the GLS regression is not very sensitive to variables, which are not normally distributed.
Table 8
Descriptive statistics and regressions of stock price on book equity abnormal earnings and accounting risk measures.
Variable Mean Std. Dev. Min 1st. Quart Median 3rd Quart Max Panel A: Descriptive statistics for the regression variables
(n=100) P 30.5597 78.5325 0.5025 5.1700 13.8950 34.5300 318.1300 BVE 2.5199 33.8957 -13.8838 1.2630 2.1700 4.0123 27.7796 AE 3.6800 1.2175 0.9867 2.9030 3.5763 4.4134 6.6273 DE x AE 2.2440 1.1160 -1.1264 1.7203 2.5336 3.0336 3.9067 CF x AE 1.5607 4.1132 -13.8685 0.3427 1.5466 3.2591 12.8434 sni x AE -0.2288 2.2106 -5.0917 -1.9513 -0.1318 1.1211 5.3622 sci x AE -0.2266 2.2051 -5.0692 -2.0074 -0.1098 1.1187 5.3772
Model Intercept BVE AE DE x AE CF x AE NI x AE CI x AE
(CI-NI)
x AE DUM_RL DUM_PER Wald chi² Adj-R²
Panel B: Parameter estimates (n=100; z-statistics in parentheses)
(4) -12.7854 4.1215 2.1355 - - - -6.4986 1.8999 342.93*** 0.2533 (-1.43) (17.92)*** (1.21) (-0.88) (0.26) (5) -18.3309 4.8465 -3.6296 8.3804 3.3661 7.0238 - - -9.9596 0.5363 191.79*** 0.2458 (-1.15) (12.28)*** (-0.74) (1.56) (3.03)*** (3.97)*** (-0.80) (0.04) (6) -18.2868 4.8390 -3.6746 8.3756 3.3481 - 6.9092 - -9.8071 0.7196 191.17*** 0.2452 (-1.14) (12.26)*** (-0.75) (1.56) (3.01)*** (3.91)*** (-0.79) (0.06) (7) 38.8865 6.6845 -27.2632 24.4603 5.8722 3.9274 - 4.4289 -21.2757 -4.9265 132.47*** 0.3062 (1.22) (9.79)*** (-2.36)** (2.11)** (3.25)*** (0.81) (1.32) (-1.02) (-0.25)
P is price per share, I measure this variable over the period 2006-2015.
BE is book value of equity per share, I measure this variable over the period 2006-2015.
AE is abnormal earnings per share, which I compute as reported earnings per share for the year t minus the risk-free rate of return at the beginning of year t times the book value of equity per share at the beginning of year t. I measure this variable over the period 2006-2015.
DE is the debt-to-equity ratio. I measure this ratio over the period 2006-2015
CF is the operating cash flow-to-current liabilities ratio. I measure this ratio over the period 2006-2015.
NI and CI are the firm-specific standard deviations of annual net income and comprehensive income, scaled by the opening market value of equity. I measure these standard deviations over the period 2006-2015.
DUM_RL is a dummy variable for reporting location DUM_PER is a dummy variable for period
34
Panel B reports the regression analysis of firms’ stock price on the book value of equity, abnormal earnings, interaction of accounting risk measures (i.e. DE and CF) with abnormal earnings and interaction of income volatility measures with abnormal earnings pooled over the period 2006-2015.
Pj = 0 + 1BVEj + 2AEj + j (4)
Pj = 0 + 1BVEj + 2AEj + 3(DEj x AEj) + 4(CFj x AEj) + 5(NIj x AEj) + j (5) Pj = 0 + 1BVEj + 2AEj + 3(DEj x AEj) + 4(CFj x AEj) + 6(CIj x AEj) + j (6) Pj = 0 + 1BVEj + 2AEj + 3(DEj x AEj) + 4(CFj x AEj) + 5(NIj x AEj) + 7[(CIj - NIj) x AEj] + j (7)
*, **, *** Denotes p < .10, p < .05 and p < .01, respectively, all two-tailed.
Table 8, Panel B reports regression models (4), (5), (6) and (7). Model (4) tests whether volatility of income is an element of risk that decreases share prices. The coefficient
1 (BVE) in model (4) is expected to be equal to 1. However off-balance-sheet net asset values and gains or losses that have not been realized or recognized may cause this coefficient to vary from 1. For AE I expect a positive coefficient 2, which shows that the variable reflects market capitalization of abnormal earnings. Model (5) and (6) test the effects on share price by interacting the income volatility measures CI and NI separately to abnormal earnings, which show the effect of the income volatility measure on abnormal earnings which effects share price. Therefore I expect a significant negative coefficient for 5/6. Suggesting that the income volatility measure captures elements of risk that are negatively priced in the capital markets after controlling the accounting risk variables DE and CF. These accounting risk variables are also interacted with abnormal earnings to check for the effects of accounting risk on share price. I predict significant negative coefficients for 3/4, which suggests that the market has a lower valuation for the abnormal earnings of firms with higher accounting risks. And model (7) tests if the incremental volatility of CI (over NI) affects share price. I predict a significant negative coefficient for 7, which indicates that the incremental volatility of CI (over NI) is associated with elements of risk and therefore priced by the market.
35
The models in Table 8 Panel B show an adjusted R-squared coefficient higher than 0.2452, which indicates that the models explain at least 24,52% of the total variance. This value is lower than in comparable studies, however it is high enough to draw conclusions from the regression models. The results in Table 6 panel B show that the BVE coefficients are significantly positive for all four models just as I expected. However, all the coefficients are four to six time the value expected which means that BVE has a higher positive effect on share price than predicted. The coefficients for AE are not significant in models (4), (5) and (6) and are negatively significant in model (7), indicating that the variable abnormal earnings don’t reflect any market capitalization in model (4), (5) and (6) which is shown in the share price. And abnormal earnings reflect negative market capitalization in model (6). The coefficients 5/6 in model (5) and (6) are not negatively significant as I expected but positively significant, so the income volatility measures don’t capture any elements of risk that are priced by the capital market and shown in share price. The coefficients 3/4 for the accountings risk variables DE and CF are not negatively significant as I expected, in fact 3/4 are positive and 4 is also significant, so these are contrasting results. The contrasting results indicate that the market doesn’t assign a lower value to the abnormal earnings of firms with higher accounting risks. On the contrary they even give a significantly higher value to the abnormal earnings for the cash flow-to-current liabilities ratio.
The dummies reporting location and period are not significant in models (4), (5), (6) and (7), indicating that reporting location of OCI and the period of reporting have no significant influence on market capitalization and share price. These results are substantiated by the Wald Chi2 test, which tests if the dummy variables DUM_RL and DUM_PER create a statistically significant improvement to the fit of the model. The null hypothesis is that both the coefficients for the variables DUM_RL and DUM_PER are simultaneously equal to zero. Table 9 shows the results of the Wald Chi² test on model (4), (5), (6) and (7). These results
