Did the location update of comprehensive income information have a positive effect on analysts’ cash flow forecasts?

(1)

Did the location update of comprehensive income information have a positive effect on analysts’ cash flow forecasts?

Faculty of Economics and Business

Name: Maaike Elisabeth Maria Ruiter Student number: 10281347

MSc Thesis: MSc in Accountancy and Control – Specialization Accountancy Thesis supervisor: Dr R. (Réka) Felleg MSc

Date: 22 June 2015 Version: final draft Words: 12.509

(2)

1

Statement of originality

This document is written by student Maaike Ruiter, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are 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 content. Reporting comprehensive income information in a separate statements of equity and net income is not allowed anymore.

Abstract

Since 15 December 2010, the updated FAS 130 of the Financial Accounting Standards Board (FASB) is effective. The FAS 130 update requires that the total of comprehensive income, the components of net income, and the components of other comprehensive income are presented in either a single statement of comprehensive income or in two separate but consecutive statements of net income and comprehensive income. Reporting comprehensive income in two separate statements of equity and net income is not allowed anymore. The FASB updated FAS 130 to create more transparency to improve the decision making process for investors. I investigate the following research question: is there a change in analyst’s cash flow forecast error and forecast dispersion in the period from 1999-2012 and after the FAS 130 update based on the location of comprehensive income information? I complement Hirst and

Hopkins (1998)’ research with this research question by examining the period before and after the update of FAS 130 based on quantitative research, in contrast to experimentally

examining the effect of comprehensive income information located in the statement of comprehensive income and the statement of equity on analysts’ stock price forecasts. My results show, based on quantitative research on Standard and Poor (S&P) 500 data during the period from 1999-2014, that analyst’ forecast error and analysts’ forecast dispersion is bigger in the period after the FAS 130 update than in the period from 1999-2012, based on the location of comprehensive income. These results contradict the experimental results of Hirst and Hopkins (1998), who find that reporting comprehensive income information in a single statement of comprehensive income, reduces analysts’ stock price forecast error compared to two separate statements of equity and net income. The valuation theory evokes that the stock

(3)

2

price is the present value of all expected future cash flows from outcomes of the firm’s operating and investing decisions (Baginski & Wahlen, 2003). Thus, I would expect that the analysts’ cash flow forecast quality would decrease based on the FAS 130 update, which is not the case. The difference in results could be explained by the natural limited

generalizability of experimental research. Overall, my study suggests that the location update of comprehensive income information has a negative effect on analysts’ cash flow forecasts for large firms like the S&P 500 firms.

Since 1999, all public trading companies in the United States have to include comprehensive income in their financial statements. Financial Accounting Standard (FAS) number 130 allows companies to report about comprehensive income in either a statement of performance or in a statement of equity (Dehning & Ratliff, 2004, p. 228). A statement of performance can be presented in a single statement of comprehensive income, or in two separate statements of comprehensive income and net income. While the amount of net income, other

comprehensive income and comprehensive income are exactly calculated in the same manner, the reporting location choice will affect no more than where the comprehensive income data appear (Bamber et al., 2010). However, Bradshaw et al. (2010) conclude, as opposed to traditional models, that a distinct section for comprehensive income information will reduce the limitations in human processing of overly complicated information like financial

statements. Therefore, to create more transparency to improve the decision making process for investors, the Financial Accounting Standards Board (FASB) updated FAS 130 in June 2011. The update requires that the total of comprehensive income, the components of net income, and the components of other comprehensive income are presented in either a single statement of comprehensive income or in two separate but consecutive statements of net income and comprehensive income. Thus reporting comprehensive income information in a separate statements of equity and net income is not allowed anymore. In the period between 1999 and 2011 most firms reported comprehensive income in the statement of equity, in contrary to the preference of standard setters. In the period between 2012 and 2014 most firms of the S&P 500 firms chose to report comprehensive income in two separate consecutive statements of net income and comprehensive income. Thus, after the update it was no longer allowed to report comprehensive income in two separate statements of equity and net income and most firms changed their reporting location from two separate statements of equity and net income to two separate consecutive statements of net income and comprehensive income. However, it is unclear whether this update concerning the location of comprehensive income information improved the decision usefulness for analysts. Analysts who can make better decisions, based on the update of the location of comprehensive income information, might also be expected to make better forecasts. Therefore I examine the following research

question: Is there a change in analyst’s cash flow forecast error and forecast dispersion in the period from 1999-2012 and after the FAS 130 update based on the location of comprehensive income information?

(5)

4

My research is motivated by a response to calls for research as well as by regulatory interest. Schipper (1991) and Brown (1993) request for research which gives insight in the financial analysts’ decision making process, while financial analysts are a representative group of sophisticated investors to whom financial reporting should be (and is) addressed. Ramnath et al. (2008) mention that insights in the analyst decision making process are as relevant today as they were in 1992. With this research I give insights in whether the update of FAS 130 is an improvement for analysts’ decision making process. I use analyst' cash flow forecast error and cash flow dispersion, because the valuation theory evokes that the stock price is the present value of all expected future cash flows from outcomes of the firm’s

operating and investing decisions (Baginski & Wahlen, 2003). Therefore, the valuation theory suggests that a difference between analysts’ stock price forecasts and the actual stock price should be caused by a difference in their assessment of future cash flows. Thus, analysts’ cash flow forecasts directly influence analysts’ recommendation instead of for example earnings per share forecasts.

Prior research finds that differential accounting locations of Mandatory Redeemable Preferred Stock (MRPS) affect the stock price judgement of analysts (Hopkins, 1996). Analysts predicted a higher stock price when the MRPS was classified as liability compared to when the MRPS was classified as equity. In addition, Hirst and Hopkins (1998) find that reporting comprehensive income information in a single statement of comprehensive income, is effective in enhancing transparency of company’s earnings management activities and reduces analysts’ stock price forecast error to the same level as those observed for an identical company that is not engaged in earnings management. However, Hirst and Hopkins (1998) find that reporting comprehensive income and its components in two separate statements of equity and net income is not as successful as a single statement of comprehensive income in disclosing earnings management and reducing stock price forecast errors of analysts. I

complement Hirst and Hopkins’ research by using S&P 500 data instead of experimental data. This research can show the effect of the location of comprehensive income on analysts in the real world instead of an experimentally created case. Besides, the natural generalizability of this quantitative research is greater than for experimental research. Moreover, I complement Hirst and Hopkins’ research by examining the period before and after the update of FAS 130, in contrast to experimentally examining the effect of comprehensive income information located in different statements (a statement of comprehensive income or two separate statements of equity and net income) on analysts’ stock price forecasts. This research shows whether the update of FAS 130 really increased the decision usefulness of analysts. Finally, I

(6)

5

complement Hirst and Hopkins' (1998) research by measuring the forecast quality of analysts’ forecasts based on the forecast error and forecast dispersion compared to their research, where they only measured the forecast error. The forecast error reflects the common error, but may also reflect the idiosyncratic error which is incompletely diversified due to limited forecasts (Barron et al., 1998). On the other hand, the forecast dispersion reflects the idiosyncratic error (Barron et al., 1998). The forecast error and the forecast dispersion are the most common measures of the quality of analysts’ forecast (Ramnath et al., 2008) and they complement each other, which gives a more complete picture of the effect of the location of comprehensive income on analysts’ forecast quality. Thus, based on an academic point of view, this research will complement prior research by examining whether there is a change in analysts' cash flow forecast error and forecast dispersion in the period from 1999 to 2012 and after the FAS 130 update, based on the location of comprehensive income information.

Besides that, based on a social point of view, the research findings may add

knowledge to the Financial Accounting Standard Board (FASB). The mission of the FASB is to create and improve the financial accounting standards and to report standards that foster financial reporting by non governmental entities that provide useful information about their decisions to investors and other users of financial reports. Financial analysts are among the primary users of financial statements (Lang & Lundholm, 1996). This research increases the knowledge of the FASB about the association between decision usefulness and the update of FAS 130.

I test my expectations on the comprehensive income reporting choice made by the Standard and Poor (S&P) 500 firms during the 1999-2014 period. Not consistent with my hypotheses, analyst’ forecast error and analysts’ forecast dispersion is bigger in the period after the FAS 130 update than in the period from 1999-2012, based on the location of

comprehensive income information. To assess the robustness of my results I also examine the effect of the location of comprehensive income on all separate analysts’ forecasts. My

inferences also generalize to this sample. The results of the robustness test confirm the results that analysts’ cash flow forecast quality is reduced in the period after the update of FAS 130 compared to the in the period from 1999-2012, based on the location of comprehensive income information.

These results contradict the experimental results of Hirst and Hopkins (1998) who find that reporting comprehensive income information in a single statement of comprehensive income, reduces analysts’ stock price forecast error compared to two separate statement of equity and net income. The valuation theory evokes that the stock price is the present value of

(7)

6

all expected future cash flows from outcomes of the firm’s operating and investing decisions (Baginski & Wahlen, 2003). Thus, I would expect that the cash forecast would increase based on the FAS 130 update which is not the case. The difference in results could be explained by the natural limited generalizability of experimental research. Overall my study suggests that the location update of comprehensive income information has a negative effect on analysts’ cash flow forecasts for large firms like the S&P 500 firms.

The effect of the location of comprehensive income on analysts’ forecasts quality depends on the informativeness of the disclosure (Lang & Lundholm, 1996). This implies that the decreased forecast quality of analysts after the update is caused by a decreased level of informativeness of firm disclosures. The decreased level of informativeness can be caused by two reasons. Firstly, this could be caused by prohibiting firms to report comprehensive income information in two separate statements of equity and net income after the update of FAS 130. Secondly, this could be caused by the difference in reporting comprehensive income information in one statement or in two separate statements.

My research makes several contributions. The first contribution of my research is to provide empirical evidence on the research question: Is there a change in analysts' cash flow forecast error and forecast dispersion in the period from 1999 to 2012 and after the FAS 130 update, based on the location of comprehensive income information? This study explains that analysts' cash flow forecast error and forecast dispersion increases in the period after the FAS 130 update compared to the period from 1999 to 2012, based on the location of

comprehensive income information for large companies like S&P 500 firms. Besides, my results contribute with empirical evidence by contradicting the experimental results of Hirst and Hopkins (1998) who find that reporting comprehensive income information in a single statement of comprehensive income, reduces analysts’ forecast error compared to the two separate statement of equity and net income. The second contribution of my study is that it investigates the impact of accounting classification on the usefulness of the decisions that financial analysts make. Schipper (1991) and Brown (1993) did request for research which gives new insight in the financial analysts’ decision making process, while financial analysts are a representative group of sophisticated investors to whom financial reporting should be addressed. My research provides new insights in the financial analysts' decision making process by showing contradicting results to Hirst & Hopkins' (1998) research. My results show that by not allowing firms to report comprehensive income information in two separate statements of equity and net income, has a negative impact on the decision usefulness of financial analysts for large firms. Analysts who can make better decisions, based on the

(8)

7

update of the location of comprehensive income information, might also be expected to make better forecasts, which is not the case. Thus I conclude that analysts’ decision making for large firms is reduced by the update of FAS 130, which does not allow reporting of

comprehensive income in two separate statements of equity and net income anymore. Besides that, this research also provides descriptive guidance to standard-setting organizations FASB, who want to increase decision usefulness for investors, based on presentation of

comprehensive income. This research gives descriptive guidance to standard-setting organizations FASB about the opposite effect of the update of FAS 130. This quantitative research shows that the FAS 130 update has a negative impact on analysts’ decision

usefulness for large companies. The final contribution of my study is that it investigates the use of accounting information by financial analysts. Financial analysts are sophisticated investors, one would expect that financial analysts are not influenced by the use of a different location of the financial information. This research shows, compared to a small number of other studies, that financial analysts’ judgements are affected by the location of information.

The paper proceeds as follows: section two describes how and why the reporting accounting location and the financial analysts’ forecast of future cash flows are related. The hypotheses will also be formulated in this section, based on aforementioned relations. Section three explains the sample of observations, the empirical measures that proxy for financial analysts’ forecast quality, and explains the empirical model that will be used to test the hypotheses. Section four presents and explains the results in different tables. Finally, section five will discuss the results and is ended with a short conclusion.

2. Literature review

2.1 The reporting location update

Since 1999, all public trading companies in the United States have to include comprehensive income in their financial statements. Financial accounting standard number 130 allows companies to report comprehensive income in either a statement of performance or in the statement of equity (Dehning & Ratliff, 2004). A statement of performance can be presented in a single statement of comprehensive income, or in two separate statements of

comprehensive income and net income. Appendix 1 shows an example of a single statement of comprehensive income, appendix 2 shows an example of two separate statements of

(9)

8

comprehensive income and net income and appendix 3 shows an example of two separate statements of equity and net income. Comprehensive income is the sum of net income and other comprehensive income. Other comprehensive income includes gains and losses from adjustments to fair value of certain debt and equity securities, foreign currency translation adjustments, the unrecognized pension expenses, and several other types of unrealized gains and losses (Scott & William, 2014). When gains and losses from fair value adjustments are realized or amortised, they are reclassified to net income. The amount of net income, other comprehensive income and comprehensive income are calculated the same way, the reporting location choice based on FAS 130 will affect no more than where the comprehensive income data appear (Bamber et al., 2010).

In June 2011, the FASB updated FAS No. 130, with the aim to create more transparency, comparability and consistency to improve the decision making process for investors and to improve the understandability of comprehensive income and the relationships between the components of other comprehensive income and net income (FASB, 2011). The update requires that the total of comprehensive income, the components of net income, and the components of other comprehensive income, are presented in either a single statement of comprehensive income or in two separate but consecutive statements of net income and comprehensive income.

In the period between 1999 and 2012, most firms reported comprehensive income in two separate statements of equity and net income, in contrary to the preference of standard setters. In the period between 2012 and 2014, most of the S&P 500 firms chose to report comprehensive income in two separate consecutive statements of comprehensive income and net income. After the update it was no longer allowed to report comprehensive income in statement of equity and the most firms changed their reporting location from two separate statements of equity and net income to two separate consecutive statements of comprehensive income and net income.

2.2 The effect of the location of comprehensive income on analysts’ forecasts

Analysts forecast based on integration of different performance-relevant information contained the primary financial statements. The location where the performance-relevant information is reported in the financial statement is an extensively debated topic. Traditional models assume that financial analysts process all information at any location, they assume that it does not matter where comprehensive income information is reported in the financial

(10)

9

statements. However, Bradshaw et al. (2010) argue, as opposed to traditional models, that a distinct section for comprehensive income, like the statement of comprehensive income, will reduce the limitations in human processing of overly complicated information like financial statements. The FASB already stated in FAS 130 that they prefer to see an performance-like presentation (FASB, 1997), because they view it as the more transparent option. Managers do also think that the location of comprehensive income does matter. Managers say that

transparency is one of the fundamental factors for their disclosure decisions (Graham et al., 2005). You would expect that managers who think that the location of comprehensive income does not matter, would probably chose to report comprehensive income in the statement of performance instead of the statement of equity. However, in the period between 1999 and 2012 most firms reported comprehensive income in two separate statements of equity and net income, in contradiction to what the preference of standard setters is. Prior research also suggests that the location of comprehensive does matter, because firms are willing to incur additional direct cost and opportunity costs to manage the location. They are willing to incur costs to manage the location of hybrid securities on the balance sheet (Engle et al., 1999), the location of an item as short term obligation versus long term liability (Gramlich et al., 2001), and the location of an item as income for continuing operations versus income for

discontinued operations (Rapaccioli & Schiff, 1991).

Prior research suggests that the location of information within the financial statements affects analysts’ forecasts. Hopkins (1996) found for example that analysts predicted a higher stock price when the MRPS was classified as liability, compared to when the MRPS was classified as equity. Experimental research also indicates that when the information is reported in two separate statements of equity and net income, this has less impact on analysts’ forecasts than information reported in the statement of comprehensive income. Hirst and Hopkins (1998) show this by examining whether comprehensive income reporting based on FAS 130 influences financial analysts' stock price forecasts of the value of a firm that manages earnings its available-for-sale (AFS) marketable securities portfolio. AFS marketable securities are measured based on fair value. The unrealized gains and losses on marketable securities are reported directly in shareholders’ equity. Management can manage net income by timing the sale of AFS marketable securities, which is called cherry picking. Hirst and Hopkins (1998) find that reporting comprehensive income information in a single statement of comprehensive income is effective in enhancing transparency of a company’s earnings management activities and reduces analysts’ stock price forecast error to the same level as those observed for an identical company that is not engaged in earnings management.

(11)

10

Though, Hirst and Hopkins (1998) find that reporting comprehensive income and its components in two separate statements of equity and net income is not as successful as a single statement of comprehensive income in disclosing earnings management and reducing forecast errors of analysts. Hirst and Hopkins (1998) show, based on an additional analysis, that the difference in stock price forecast error between the statement of comprehensive income and the statement of equity is caused by a different belief of analysts about the likelihood of future earnings growth and the quality of financial reporting. Analysts who did not undo the earnings management effect on earnings, did have a favourable perception of the future cash flows of the firm. Firms probably know that analysts have a more favourable perception of future cash flows when they report comprehensive income in two separate statements of equity and net income, because firms with more opportunities to manage earnings, based on selectively selling AFS securities, more often choose to report comprehensive income in two separate statements of equity and net income instead of reporting it the statement of performance.

2.3 Analysts’ forecast quality

The forecast quality decreases when the differences between analysts’ forecasts and the actual value increases and when analysts’ forecasts differ more from other analysts’ forecasts. The most common measures of analysts’ forecast quality are forecast error and forecast dispersion (Ramnath et al., 2008). The effect of the location of comprehensive income on analysts’ forecasts quality depends on the informativeness of the disclosure (Lang & Lundholm, 1996). The forecast error reflects the common error, but may also reflect the idiosyncratic error which is incompletely diversified due to limited forecasts (Barron et al., 1998). A common error occurs from public information that analysts rely upon and an idiosyncratic error occurs from private information analysts rely upon (Barron et al., 1998). On the other hand, the forecast dispersion reflects the idiosyncratic error (Barron et al., 1998). Lang and Lundholm (1996) find that analysts will place less weight on private information as the informativeness of firm-provided disclosure increases.

2.4 How are analysts’ forecasts affected by location

(12)

11

According to Lipe (1998), the different beliefs of analysts about the likelihood of future earnings growth and the quality of financial reporting which affected the stock price error, could be caused by (a) the fact that financial statement users may not know where to find the items of comprehensive income, or (b) financial statement users may have difficulties in integrating these disparate pieces of comprehensive income information, and/or (c) financial statement users are affected by income statement effect. The income statement effect appears to increase the amount of attention afforded to comprehensive income information when it is included in the statement of performance (Lipe, 1998). B and c are probably the most likely factors of why financial analysts are influenced by location, because financial analysts are the most sophisticated group of investors who know where to find the information.

The proximity compatibility principle explains why sophisticated analysts may have difficulties in integrating disparate pieces of comprehensive income information. The proximity compatibility principle suggests that data values in one single format will support better information integration than more separable formats will (Carswell & Wickens, 1996). So, when analysts have to evaluate future cash flows of a firm based on performance-relevant information, performance will be enhanced when the information on for example

comprehensive income is displayed in a single statement of comprehensive income, instead of in two separate consecutive statement of comprehensive income and net income or two separate statements of equity and net income. Hodge et al. (2010) show in their research that non-professional investors' forecast error and forecast dispersion is smaller when they receive forecast-relevant cash flow information in a single display, in contrast to when they receive forecast-relevant cash flow information in a separate statement. Hirst and Hopkins (1998) confirm this experimentally for analysts by showing that analysts can reduce their stock price forecast error of firms who manage earnings, but report comprehensive income in a single statement of comprehensive income, to the same level as those observed for an identical firm which does not manage its earnings. But, analysts are not as successful in detecting earnings management when the performance-relevant information is separated by reporting

comprehensive income information in the statement of equity and net income. The cognitive load theory explains how memory limitations likely create the proximity compatibility principle. The cognitive load theory proposed that individuals have a limited (short term) memory making, which makes it difficult to integrate multiple pieces of information,

especially when the information is dispersed (Sweller, 1989). This is also likely the case when the comprehensive income information is presented on separate computer screens (Hodge et al., 2010).

(13)

12

The direct search strategy explains why sophisticated analysts appears to increase the amount of attention afforded to comprehensive income information when it is included in a statement of performance. The financial statement format appears to affect these analysts' judgments partly because of a failure to acquire information (Maines & McDaniel, 2000). Analysts’ judgements are based on their valuation models. They acquire relevant information from these models using a directed-search strategy, which leads them to skip back and forth among information in the financial statements (Hunton & McEwen 1997). Those analysts will only read the information which is relevant for their models. Brown (1997) explained that the statement of equity is perceived as less important by analysts compared to a statement of performance. This could be an explanation of why analysts in the experiment of Hirst and Hopkins (1998) may have ignored the statement of equity.

Maines and McDaniel (2000) identify a mediating effect on the relation between locating comprehensive income and the ability of financial analysts to make a better forecast about a firm. The mediating factor here is a type of business activity. The performance

judgement of financial analysts may depend on whether comprehensive income information is related to core- or non-core business activities. When the comprehensive income information is related to core business activities, the location of comprehensive income would probably have less impact. This would be the case, because the information is relevant to the exactness of their forecast. Financial analysts will skip back and forth among issues like core activities. Hirst and Hopkins (1998) confirm this by showing that when comprehensive income

information is related to non-core activities, the location of comprehensive income did have a negative impact on the forecast ability of the financial analysts.

2.5 Hypotheses

The effect of the location of comprehensive income on analysts’ forecasts depends on the informativeness of the disclosure (Lang and Lundholm, 1996). Informativeness can be defined as an increased knowledge or as dissipating ignorance.

Applying the proximity compatibility principle to the analysis of financial statements suggests that financial analysts will more effectively and efficiently integrate comprehensive income information when it is presented in a single statement of comprehensive income in contrast to when it is presented in two separate consecutive statements of comprehensive income and net income or two separate statements of equity and net income. One way to improve the proximity of comprehensive income information is to present the information of

(14)

13

comprehensive income in one statement, which is the case in the statement of comprehensive income instead of in two separate statements (the statement of comprehensive income and the statement of net income or the statement of equity and the statement of net income).

Thus, analysts’ knowledge about comprehensive income information is increased more by a single statement of comprehensive income than by two separate statements of comprehensive income and income or two separate statements of equity and net income, because analysts have less difficulty with integrating the comprehensive income information.

Besides that, applying the directed-search strategy to the analysis of financial

statements suggests that financial analysts will pay more attention to comprehensive income information when it is presented in a single statement of comprehensive income or two separate consecutive statements of comprehensive income and net income compared to when it is presented in two separate statements of equity and net income. One way to increase the attention from analysts to comprehensive income in the directed-search strategy, is to report the comprehensive income information in a statement of performance instead of in a

statement of equity, because the statement of equity is perceived by analysts as less important for their forecasting models. Thus, analysts dissipate less ignorance to comprehensive income information when the comprehensive information is reported in a single statement of

comprehensive income or by two separate statements of comprehensive income and income in contrast to two separate statements of equity and net income, because the statement of equity is perceived as less important for analysts' forecasting models.

Thus, theory suggests that an increased proximity of information and an improved location of comprehensive income information for the direct-search strategy by a single statement of comprehensive income or two separate consecutive statement of comprehensive income and net income in the period after the FAS 130 update, compared to two separate statements of equity and net income which were permitted in the period before the FAS 130 update, will lead to increased informativeness of disclosures and should increase the quality of the forecasts that rely on that information. Thus, I expect that analysts’ forecast error will be smaller after the update of FAS 130, based on the location of comprehensive income information. Accordingly, I test the following hypothesis:

H1: Analysts’ cash flow forecast error is smaller in the period after the FAS 130

update than in the period from 1999 to 2012, based on the location of comprehensive income information.

(15)

14

The effect of the location of comprehensive income on the dispersion of analysts’ forecasts also depends on differences in information (Lang & Lundholm, 1996). Analysts will place less weight on private information, based on the proximity principle and direct search principle, when the comprehensive income information is reported in a single statement of comprehensive or two separate consecutive statements of comprehensive income and net income instead of in a statement of equity and a statement of net income, because a single statement of comprehensive or two separate consecutive statements of comprehensive income and net income are more informative. I expect that analysts’ forecast dispersion will be smaller after the update of FAS130, based on the location of comprehensive income information Accordingly, I test the following hypothesis:

H2: Analysts’ cash flow forecast dispersion is smaller in the period after the FAS

130 update than in the period from 1999 to 2012, based on the location of comprehensive income information.

3. Methodology

3.1 Sample

Comprehensive income disclosure is not fully reported in Compustat or any other machine-readable database. Therefore, to manage my data collection efforts I selected a smaller but still representative sample of U.S. firms: the firms included in the S&P 500 index over the year 1999 till 2014 (Chambers et al., 2007). Since 1999, all public trading companies in the United States have to include comprehensive income in their financial statements in either a performance statement or in the statement of equity (Dehning & Ratliff, 2004). The update of FAS 130 was effective for all public trading companies in the United States from December 15th, 2011. Thus, my sample period spans sixteen years: thirteen years before and three years after the update of FAS 130. I obtained all the other financial statement data, annual fiscal stock price close data, total annual fiscal market value data and Standard Industry

Classification (SIC) data from Compustat. The yearly forecasted cash flow data, the yearly actual cash flow data, the number of analysts that did forecast cash flows and the forecast horizon data I obtained from I/B/E/S Detail History (US edition). I use 'one year ahead' cash

(16)

15

flow forecasts, which is in line with the research of DeFond and Hund (2002), because almost all I/B/E/S cash flow forecasts are annual. Finally I collected the daily return data just as Rayburn (1986) did from CRSP. I excluded observations from the sample with insufficient data in Compustat, I/B/E/S and CRSP databases which are needed to compute my variables. By excluding all firms which had no cash flow forecast, I control for other factors that might affect the availability of cash flow forecasts (DeFond & Hung, 2003).Finally, I

hand-collected the location of comprehensive income data from annual 10-K reports that were filed with the Securities and Exchange Commission (SEC). I hand-collected the data from the EDGAR database based on the firm’ CIK code and fiscal year. My final sample period spans sixteen years: thirteen years before and three years after the update of FAS 130 and is made up of 285 firms, with 3264 observations.

3.2 Empirical measures

Forecast quality

In this study I focus on two of the most common measures of analysts’ forecast quality: forecast error and forecast dispersion (Ramnath et al., 2008). I use analyst’ cash flow forecast error and cash flow dispersion, because the valuation theory evokes that the stock price is the present value of all expected future cash flows from outcomes of the firm’s operating and investing decisions (Baginski & Wahlen, 2003). Therefore, the valuation theory suggests that a difference between analysts’ stock price forecasts and the actual stock price should be caused by a difference in their assessment of future cash flows. Thus, analysts’ cash flow forecasts directly influence analysts’ recommendation instead of for example earnings per share forecasts who does not. Analysts’ forecast error (FERROR) is defined as the natural log of the absolute mean errors of all forecasts made in the year for cash flows, scaled by the stock price of the beginning of the year (Dhaliwal et al., 2012):

FERRORi,t = ln( i,t,j– ACFi,t]/Pi,t)

(1) The subscripts, i,t and j represent firm i, year t and forecast j. FCF is the analyst’ forecasted cash flow. ACF is the actual cash flow and P is the stock price at the beginning of the year. FCF and ACF are both collected from the I/B/E/S database to ensure consistency. The last cash flow forecast of each year is used to mitigate for stale forecasts (Dhaliwal, et al., 2012).

(17)

16

Secondly, I use analysts’ forecast dispersion as inverse measure of forecast accuracy.

Analysts’ forecast dispersion (FDISP) is defined as the natural log of the standard deviation of all forecasts made for cash flows (Irani & Karamanou, 2003):

FDISPi,t =ln( – ) (2)

AFCF is the average forecasted cash flow and N is the number of observations. 3.3 Empirical model

This study examines if analysts’ forecast error (H1) and forecast dispersion (H2) is smaller in the period after the FAS 130 update than in the period from 1999 to 2012, based on the location of comprehensive income information. I test my hypothesis by estimating the coefficients in the following two multiple regression models:

FERROR i,t = β0 + β1LOCPP + β2LNANA i,t + β3VOLA i,t + β4FHORi,t + β5SIZE i,t (3)

+ β6LOSSi,t + β7ACCRUALi,t + β8FIRM + β9INDUSTRTY + ε

FDISPi,t = β0 + β1LOCPP + β2LNANA i,t + β3VOLA i,t + β4FHORi,t + β5SIZE i,t (4)

+ β6LOSSi,t + β7ACCRUALi,t + β8FIRM + β9INDUSTRTY + ε

where:

LOCPP = 1 for the post-period from 2012 till 2014 and 0 for the pre-period

from 1999-2012.

LNANA = The natural log of the number of analysts that follows the company

through that fiscal year (Dhaliwal et al., 2012).

VOLA = The natural logarithm of the standard deviation of the daily fiscal year

stock return .RETURN =

(Berks & Demarzo’s,

2011), where R is the stock price return of the firm, is the average stock price return of the firm and N the number of observations. Rayburn (1986) also used daily stock return for his volatility measure and finds an association between cash flow and return.

(18)

17

FHOR = The median forecast horizon. A forecast horizon is the number of

days between the cash flow announcement day and the forecast date

(O’Brien 1990), .

SIZE = The natural log of a firm’s market value at the beginning of the year

(Dhaliwal et al, 2012).

LOSS = 1 if the firm reports a negative net income and 0 if the firm reports a

positive net income.

ACCRUAL = 1 if the firm reports higher average absolute accruals over the past three

years than the industry-year mean and 0 if the firm reports lower accruals than the industry-year mean. Accruals are calculated in line with Bhattacharya et al. (2003). Accruals are based on the following formula: ACCRUALi,t = (ΔCAi,t – ΔCLi,t – ΔCASHi,t + ΔSTDi,t – DEPi,t

+ ΔTPi,t)/TAi,t. Where ΔCA is the change in the firm’s total current

assets, ΔCL is the change in the firm’s total current liabilities, ΔCASH is the change in the firm’s cash, ΔSTD is the change in the firm ‘s current part of long term debt which is included in current liabilities, DEP is the firm’s depreciation and amortization expense, ΔTP is the change in the firm’s income taxes payable and ΔTA is the firm’s total asset at the end of the previous year.

FIRM = Firm is a dummy variable for each firm. I identify each firm based on

their CUSIP.

INDUSTRY = Industry is a dummy variable for each major economic sector. I identify

the economic sector by the standard industry code (SIC). I use one-digit SIC industry classifications.

My main variable of interest is LOCPP, this variable measures the location of comprehensive income information. My hypothesis predicts that LOCPP is negatively associated with FERROR and FDISP. Thus LOCPP, which is 1 after the update of FAS 130, should decrease the forecast error and forecast dispersion. I include a number of additional control variables, which are especially correlated with transparency of disclosures like Hope (2003) found, and firm-specific variables. Specifically I control for the number of analysts (LNANA), because forecast quality is associated with analysts, following (Kross et al., 1990). The forecast precision of analysts is a response to more competition (Lys & SOO, 1995). Analysts have more incentives to make an accurate forecast based on their common and idiosyncratic

(19)

18

information when they have more competition. I would expect that LNANA has a negative relationship with FERROR and FDISP. There is also controlled for volatility (VOLA). When the returns of a firm are more volatile, it is more complicated for analysts to make a forecast of the firm’s future cash flows based on both idiosyncratic and common information

(Dhaliwal et al., 2010). I expect that VOLA would have a positive association with FERROR and FDISP. In addition I control for analysts’ forecast horizon (FHOR). The number of days between the forecast and the annual report announcement are likely to affect the amount of information available for analysts (O’Brien, 1990). When the forecast horizon increases, there would be less common information available for analysts. Analysts should rely more on idiosyncratic information. The forecast dispersion increases when analysts rely more on idiosyncratic information, because analysts rely more on their own information, which provides more spread in forecasts. Besides, the forecast error increases when the

informativeness decreases. Thus, I expect that FHOR has a positive relation with FERROR and FDISP. Similar to research by Hope (2003) I control for firm size (SIZE), because firm size is a proxy of the general information environment.Choi et al. (2004) describe that the forecast accuracy is an increasing function of firm size. Hope (2003) shows that firm size is correlated with the transparency of disclosures. Analysts will make less forecast errors based on more transparent public information. Analysts' forecast dispersion is also lower with more transparent information, because they will put less weight on idiosyncratic information (Lang & Lundholm, 1996). Thus, I expect that SIZE has a negative relationship with FERROR and FDISP. Moreover, there is controlled for loss (LOSS). Das and Zhang (1998) have shown that the forecasts' quality for loss firms are significantly different to that of profit firms. It is more difficult for analysts to predict future cash flows when the firm generated a loss, because the earnings are more volatile, which creates more uncertainty in information (Hope, 2003). The informativeness of financial statements decreases with more uncertainty, that is why I expect a positive relation between LOSS and FERROR and FDISP. Further I control for accruals (ACCRUAL) just as (Dhaliwal et al., 2012) did, to control for firm’s financial opaqueness. Dhaliwal et al. (2012) demonstrate that analysts' forecasts are less accurate for financial opaque firms than for firms which are more transparent. Transparency increases the decision usefulness of the information, which decreases the forecast error. Besides, when the

information is more useful, analysts put less weight on idiosyncratic information, which decreases the forecast dispersion. The transparency of the variable accrual is higher when the dummy is 0. Thus, I expect that ACCRUAL is positively correlated with FERROR and FDISP. Finally I control for FIRM and INDUSTRY. I do not include additional time

(20)

19

dummies, because LOCPP already measures time. The model becomes too specified based on two time measures, which result in a battle for correlation.

4. Results

4.1 Descriptive statistics

In the period from 1999 till 2012 only 2 percent of my sample reports comprehensive income in a single statement of comprehensive income and only 22 percent of my sample reports comprehensive income in two separate statements of comprehensive income and net income compared to 79 percent of my sample that reports comprehensive income in two separate statements of equity and net income. Bamber et al. (2010) find in their sample of the S&P 500 in a period from 1999 to 2001, that only 19 percent of their sample reports comprehensive income in a statement of performance. This difference of two percent can be explained by firms who changed their reporting location between 2001 and 2012. In the period from 2012 till 2014 only 9 percent of my sample reports comprehensive income in a single statement of comprehensive income in contrast to 91 percent of my sample that reports comprehensive income in two separate statements of comprehensive income and net income. My sample of the S&P 500 firms is large, with a median (mean) market value of equity of $11.1 ($27.7) billion. Bamber et al (2010)’median (mean) market value of equity is $7.9 ($20.2) billion. This suggests that the market value of S&P 500 firms is increased during the period between 2001 and 2014. Table 1 displays the descriptive statistics. The standard deviation of size is decreased compared to the sample of Bamber et al. (2010), which is from 1.227 to 1.149. This suggest that S&P 500 firms have a higher market value, but the difference between the market values of the firms is smaller. Besides, the standard deviation is relatively bigger for the forecast dispersion than the forecast error. This implies that analyst’ forecast dispersion differs relatively more among analysts than analysts' forecast errors. Moreover, 10 percent of the S&P 500 firms report a loss and 7 percent of the firms report higher accruals over the past three years than the industry mean.

Table 2 reports correlations among the dependent and independent variables. Table 2 shows that LOCPP is positively associated with FERROR and FDISP. This suggests that forecast dispersion and forecast error increases after the update of FAS 130. Moreover,

(21)

20

TABLE 1

Descriptive statistics for independent variables

Std. Lower Upper

Variable Mean Dev. Quartile Median Quartile

FERROR -4,050 1,257 -4,847 -4,070 -3,250 FDISP -0,920 1,236 -1,632 -0,812 -0,131 LOCPP 0,236 0,425 0,000 0,000 0,000 LNANA 2,701 1,092 2,109 2,773 3,296 VOLA -4,011 0,454 -4,342 -4,029 -3,719 FHOR 169,139 64,989 133,000 182,000 200,000 SIZE 9,403 1,149 8,675 9,311 10,106 LOSS 0,100 0,294 0,000 0,000 0,000 ACCRUAL 0,070 0,258 0,000 0,000 0,000

LOCPP and LNANA are positively correlated. This suggests that after the update of FAS 130, the number of analysts who forecast cash flows increased. Finally, the forecast dispersion is higher for firms bigger in size, but decreases when more analysts follow the company, while the forecast error has the strongest correlation with losses. This implies that forecast error and forecast dispersion, two of the most common measures of analysts’ forecast quality, provide a more clear view of forecast quality when they are both measured.

4.2 Multivariate regression results

Agriculture, forestry and fishing is the benchmark industry in tables 3,4,7, whose effect is impounded in the intercept. The significance levels of table 3 and 4 are based on robust standard errors which are corrected for heteroscedasticity and dependence across companies. Besides, I have tested the multicollinearity of my main variables in table 3 and 4. I find that there is a multicollinearity problem at the mining industry, construction industry and

wholesale trade industry, which I have excluded from the regression. Table 3 shows the results of estimating the multivariate regression (3) in equation with controlling for residual industries (Bowen et al. 1995). Hypothesis 1 is confirmed when LOCPP is negatively associated with FERROR. However, LOCPP is positively associated with FERROR

(p≤0.000), which means that after the update of FAS 130, the forecast error of analysts’ cash flow forecasts significantly increases based on the location of comprehensive income

(22)

21

TABLE 2

Pearson correlations among variables

FERROR FDISP LOCPP LNANA VOLA FHOR SIZE LOSS ACCRUAL

LOCPP 0,062 0,145 1,000 (0,027) (0,000) LNANA 0,066 -0,397 0,204 1,000 (0,019) (0,000) (0,000) VOLA -0,013 -0,068 -0,021 0,028 1,000 (0,641) (0,018) (0,465) (0,328) FHOR 0,054 0,100 0,094 0,228 0,008 1,000 (0,058) (0,000) (0,001) (0,000) (0,768) SIZE -0,143 0,150 0,044 0,306 -0,061 -0,037 1,000 (0,000) (0,000) (0,118) (0,000) (0,031) (0,189) LOSS 0,115 0,036 -0,072 -0,070 0,019 0,049 0,017 1,000 (0,000) (0,209) (0,011) (0,014) (0,511) (0,082) (0,554) ACCRUAL 0,042 -0,034 0,129 -0,030 -0,033 0,047 -0,182 -0,040 1,000 (0,137) (0,232) (0,000) (0,293) (0,237) (0,096) (0,000) (0,152)

(23)

22 TABLE 3

Multiple regression of the effect of the location of comprehensive income on analysts’ forecasts error (FERROR) before and after the update

FERROR i,t = β0 + β1LOCPP + β2LNANA i,t + β3VOLA i,t + β4FHORi,t + β5SIZE i,t +β6LOSSi,t (3)

+ β7ACCRUALi,t + β8FIRM + β9INDUSTRTY + ε

Variable Predicted With Industry (N=239) Sign Controls LOCPP - 0,231*** (0,000) LNANA - 0,020 (0,469) VOLA + 0,003 (0,879) FHOR + 0,018 (0,298) SIZE - -0,424*** (0,000) LOSS + 0,306*** (0,000) ACCRUAL + -0,046 (0,579) Manufacturing ? -0,060* (0,065) Transportation, ? 0,051 Communication (0,208) and Utilities ? Retail trade ? -0,041 (0,216)

Finance, insurance, and ? 0,004

real estate (0,922) Services ? 0,004 (0,910) Intercept -4,482*** (0,000) Adjusted R2 0,474

There is controlled for firms by dummies. The two-tailed p-values are in parentheses. *, **, ***

Significant at 10%, 5%, 1% respectively based on two-tailed tests

(24)

23

compared to the period from 1999-2012. This implies that analysts are less able to forecast the future cash flow of a firm after the update of FAS 130 based on the location of comprehensive income. That LOCPP is positively associated, can be caused by two reasons. Firstly, this could be caused by prohibiting firms to report comprehensive income information in two separate statements of equity and net income after the update of FAS 130. Secondly, this could be caused by the difference in reporting comprehensive income information in one statement or in two separate statements. The S&P 500 firms that report comprehensive income information in one statement is increased from 2 percent to percent 9 percent.

Table 3 also displays with respect to the remaining control variables, that the number of analysts (LNANA) have a positive association with analysts’ cash flow forecast errors for large firms and ACCRUALS have a negative association with FERROR, but both are not significant. In addition, this research confirms the results of Choi et al. (2004) who describe that the forecast accuracy is an increasing function of firm size. This implies that the

informativeness increases when the firm size is bigger, because analysts make less forecast errors. Further, the manufacturing industry dummy is the only significant industry dummy, indicating that there is no systematic difference in analyst’ forecast errors between industries. However, several firm dummies are significant indicating that there is systematic difference in analysts’ cash flow forecast errors between firms.

Table 4 shows the results of estimating the multivariate regression (4) in equation with controlling for industry (Bowen et al. 1995). My second hypothesis is confirmed when

LOCPP is negatively associated with FDISP. However, LOCPP is positively associated with FDISP (p≤0.000), which means that after the update of FAS 130, the forecast dispersion of analysts’ cash flow forecasts significantly increases based on the location of comprehensive income information compared to the period from 1999-2012. This implies that an analyst‘ cash flow forecasts differs more among other analysts’ cash flow forecasts after the update of FAS 130 based on the location of comprehensive income. This result confirms that the

location update of comprehensive income information has a negative effect on analysts’ cash flow forecasts. The forecast dispersion reflects the idiosyncratic error (Barron et al., 1998). An idiosyncratic error occurs from private information analysts rely upon (Barron et al., 1998). Lang and Lundholm (1996) find that analysts will place less weight on private

information as the informativeness of firm-provided disclosure increases. This result implies that informativeness of the comprehensive income information is decreased after the update of FAS 130.

(25)

24 TABLE 4

Multiple regression of the effect of the location of comprehensive income on analysts’ forecasts dispersion (FDISP) before and after the update

FDISPi,t = β0 + β1LOCPP + β2LNANA i,t + β3VOLA i,t + β4FHORi,t + β5SIZE i,t +β6LOSSi,t (4)

+ β7ACCRUALi,t + β8FIRM + β9INDUSTRTY + ε

Variable Predicted With Industry (N=239) Sign Controls LOCPP - 0,206*** (0,000) LNANA - 0,634*** (0,000) VOLA + -0,014 (0,447) FHOR + -0,015 (0.407) SIZE - 0,412*** (0,000) LOSS + 0,341*** (0,000) ACCRUAL + -0,141* (0,079) Manufacturing ? -0,015 (0,627) Transportation, ? -0,005 Communication (0,891) and Utilities ? Retail trade ? 0,004 (0,899)

Finance, insurance, and ? -0,001

real estate (0,988) Services ? -0,020 (0,579) Intercept -1.702*** (0,000) Adjusted R2 0,571

(26)

25

Table 4 also displays with respect to the remaining control variables, the number of analysts (LNANA) has a positive association with analysts’ cash flow forecast dispersion for large firms. This contradicts the results of Lys and Soo (1995) who find that the forecast precession of analysts is a response to more competition. This result implies that when the number of analysts increases, the forecast dispersion also increases. The forecast dispersion reflects the idiosyncratic error (Barron et al., 1998). When the number of analysts who forecast the cash flows of a company increases, the number of analysts that have different idiosyncratic information also might increase. This can result in more variance in analysts’ cash flow. In addition, Hope (2003) shows that firm size is correlated with the transparency of disclosures. My research shows that analyst’ forecast dispersion is increased by size. This suggests that the relation between size and transparency of disclosures is not negative among only large firms. Moreover, the forecast dispersion is decreased by firms with more accruals over the past three years than the industry mean. This might imply that analysts have the same common sense about firms that manage their earnings above average compared to their

industry. Further, this research shows that the forecast dispersion is positively associated with the variable loss. This result confirms the research of Das and Zhang (1998). They have shown that the forecasts' quality for loss firms are significantly different to that of profit firms, because it is more difficult for analysts to predict future cash flows when the firm generated a loss, because the earnings are more volatile, which creates more uncertainty in information (Hope, 2003). Thus, this result suggests that analysts have less common sense about the future value when the firm reports a loss. Furthermore, the industry dummies are not significant, indicating that there is no systematic difference in analyst’ forecast errors between industries. However, several firm dummies are significant, indicating that there is systematic difference in analyst’s cash flow forecast dispersion between firms.

4.3 Robustness test

To help ensure that my results with the dependent variable forecast error are not simply an artefact of my prior sample, I use all the forecasts of analysts instead of the mean analysts’ forecast error. This sample consists of 285 firms, with 49552 observations. I also include a dummy variable for analysts to verify the robustness of my results. I examine my results by estimating the coefficients in following multiple regression:

AERROR i,t = β0 + β1LOCPP + β2LNANA i,t + β3VOLA i,t + β4FHORi,t + β5SIZE i,t (5)

(27)

26

Where:

AERROR = The natural log of the absolute difference between analysts’ forecasted

cash flows and the actual cash flows.

ANA = ANA is a dummy variable for each analyst. I indentify each analyst

based on their analyst code.

Table 5 displays the descriptive statistics. AERROR is -0.706, this implies that the mean forecast over sixteen years for all the companies differs $ -0.71 per forecast. This indicates that the average analyst has a positive view of firms’ future cash flows. The descriptive statistics in table 5 differ from the descriptive statistics in table 1 because these descriptive statistics are influenced by the number of analysts that forecasted the future cash flows of the firms.

TABLE 5

Descriptive statistics for independent variables

Std. Lower Upper

Variable Mean Dev. Quartile Median Quartile

AERROR -0.706 1,350 -1.470 -0.545 0.235 LOCPP 0,160 0,366 0.000 0.000 0.000 LNANA 3.798 1.023 3.000 3.663 4.710 VOLA -3.974 0,442 -4.268 -4.005 -3.705 FHOR 179.365 38.112 162.000 186.000 198.000 SIZE 9.755 1,255 8.919 9.683 10.457 LOSS 0,070 0,249 0.000 0.000 0.000 ACCRUAL 0.040 0.196 0.000 0.000 0.000 COMPRE 14.021 1.478 13.097 13.952 14.905

Table 6 reports correlations among the dependent and independent variables. Table 2 shows that LOCPP is positively associated with AERROR, just as FERROR and FDISP. This suggests that analysts’ individual forecast error increases after the update of FAS 130. Besides, the forecast horizon (FHOR) is positively associated with the number of analysts (LNANA). This implies that when there is more competition among analysts, analysts publish their cash flow forecasts earlier. Moreover, there is a positive relation between the number of analysts and the volatility, this suggest that more analysts make a cash flow forecast for more

(28)

27

TABLE 6

Pearson correlations among variables

AERROR LOCPP LNANA VOLA FHOR SIZE LOSS ACCRUAL

LOCPP 0,049 1,000 (0,000) LNANA 0,027 0,090 1,000 (0,000) (0,000) VOLA 0,052 -0,24 0,237 1,000 (0,000) (0,000) (0,000) FHOR -0,011 0,093 0,220 0,005 1,000 (0,011) (0,000) (0,000) (0,253) SIZE 0,060 0,101 0,089 -0,260 -0,083 1,000 (0,000) (0,000) (0,000) (0,000) (0,000) LOSS 0,035 -0,072 -0,057 0,027 0,032 0,028 1,000 (0,286) (0,018) (0,029) (0,497) (0,088) (0,598) ACCRUAL -0,032 0,072 -0,133 -0,049 -0,020 -0,195 -0,041 1,000 (0,000) (0,000) (0,000) (0,000) (0,000) (0,000) (0,159)

(29)

28

volatile firms. Finally, SIZE is negatively correlated with volatility (VOLA), which entails that when the firm becomes larger, the returns of the firm are less volatile.

Table 7 shows the results of estimating the multivariate regression (5) in equation with and controlling for industry. The significance levels of table 7 are based on robust standard errors which are corrected for heteroscedasticity and dependence across companies. Besides, I have tested the multicollinearity of my main variables in table 7. I find that there is no

multicollinearity problem when including all variables in the regression. Table 7 confirmed the results of my research by showing that LOCPP is positively associated with AERROR (p≤0.000), which means that after the update of FAS 130, the forecast error of analysts’ cash flow forecasts significantly increases based on the location of comprehensive income

compared to the period from 1999-2012. This result confirms that the location update of comprehensive income information has a negative effect on analysts’ cash flow forecasts quality. This result implies that analysts’ individual cash flow forecast error increases after the update of FAS 130 and that the informativeness of the firm‘ disclosure is decreased for

individual analysts after the update of FAS 130. Table 3 already showed that the informativeness of the firms’ disclosure is also decreased for the average analyst.

Table 7 also displays, with respect to the remaining control variables, that the number of analysts (LNANA) has a positive association with analysts’ individual cash flow forecast error. The results suggest that individual analysts forecast with more precision, based on a response to more competition. This confirms the results of Lys & Soo (1995) who find that the forecast precision of analysts is a response to more competition. Moreover, the size of the firm is, like equation 4, again positively correlated with analyst forecast quality. This suggests that the relation between size and transparency of disclosures is not negative among only large firms. This contradicts the results of Choi et al. (2004) who describe that the forecast accuracy is an increasing function of firm size. Further, the variable loss (LOSS), forecast horizon (FHOR) and volatility (VOLA) are positively associated with the individual cash flow forecast error, as expected. These variables increase the uncertainty of the firm disclosed information, which decreases analysts’ individual cash flow forecast accuracy. In addition, individual analyst’ forecast error is decreased by firms with more accruals over the past three years than the industry mean. This might imply that individual analysts have the same

common sense about firms that manage their earnings above average, compared to their

(30)

29

TABLE 7

Multiple regression of the effect of the location of comprehensive income on analysts’ individual forecasts error (AERROR) before and after the update

AERROR i,t = β0 + β1LOCPP + β2LNANA i,t + β3VOLA i,t + β4FHORi,t + β5SIZE i,t

+ β6LOSSi,t + β7ACCRUALi,t + β8ANA + β9FIRM + β10INDUSTRTY + ε (5)

Variable Predicted With Industry (N=239) Sign Controls LOCPP - 0,225*** (0,000) LNANA + 0,163*** (0,000) VOLA + 0,118*** (0,000) FHOR + 0,025*** (0,000) SIZE - 0,200*** (0,000) LOSS + 0,244*** (0,000) ACCRUAL + -0,325*** (0,000) Mining ? -0,153*** (0,000) Construction ? -0,643*** (0,000) Manufacturing ? -1,979*** (0,000) Transportation, ? -1,293*** Communication (0,000) and Utilities Whole saletrade ? 0,835*** (0,000) Retail trade ? -1,841*** (0,000)

(31)

30 TABLE 7 (continued)

Multiple regression of the effect of the location of comprehensive income on analysts’ individual forecasts error (AERROR) before and after the update

AERROR i,t = β0 + β1LOCPP + β2LNANA i,t + β3VOLA i,t + β4FHORi,t + β5SIZE i,t

+ β6LOSSi,t + β7ACCRUALi,t + β8ANA + β9FIRM + β10INDUSTRTY + ε (5)

Variable Predicted

With Industry

(N=239) Sign Controls

Finance, Insurance ? 0,151

and Real estate (0,117)

Services ? -0,248 (0,087) Intercept -0,306 (0,000) Adjusted R2222 _0,285

There is controlled for firm by dummies. The two-tailed p-values are in parentheses. *, **, ***

Significant at 10%, 5%, 1% respectively based on two-tailed tests

difference in analyst’ individual cash flow forecast errors between industries. Moreover, several firm dummies are significant, indicating that there is systematic difference in analysts’ individual cash flow forecast errors between firms. Only the wholesale trade industry is positively correlated with the individual cash flow forecast error compared to the agriculture, forestry and fishing benchmark industry. This implies that individual analyst have a harder time to forecast cash flows of the wholesale trade industry compared to other industries. Finally, several analyst dummies are significant, indicating that there is systematic difference in individual analysts’ cash flow forecast error between analysts.

(32)

31

5. Conclusion

My study provides insights in whether there is a change in analyst’s cash flow forecast error and forecast dispersion in the period from 1999-2012 and after the FAS 130 update based on the location of comprehensive income information. Since December 15th 2011, the updated FAS 130 of the Financial Accounting Standards Board (FASB) is effective. The FAS 130 update requires that the total of comprehensive income, the components of net income, and the components of other comprehensive income are presented in either a single statement of comprehensive income or in two separate but consecutive statements of net income and comprehensive income. Reporting comprehensive income in two separate statements of equity and net income is not allowed anymore. The FASB updated FAS 130 to create more transparency to improve the decision making process for investors.

To answer whether there is a change in analysts' forecast quality after the update compared to the period from 1999-2012, I draw on the following theoretical research: analyst forecast properties, proximity principle and directed search strategy. Analyst forecast

properties theory suggests that the effect of the location of comprehensive income on analysts’ forecast quality depends on the informativeness of the disclosure (Lang and

Lundholm, 1996). Informativeness can be defined as an increased knowledge or as dissipating ignorance. One way to increase the knowledge of the analysts, based on the proximity

principle, is to present the information of comprehensive income in one statement, instead of in two separate statements of comprehensive income and net income or in two separate

statements of equity and net income, because one statement makes it more easy for the analyst to integrate the comprehensive income information. One way to dissipate ignorance from analysts, based on the directed-search strategy, is to report the comprehensive income information in a single statement of comprehensive income or in two separate statements of comprehensive income and net income instead of in two separate statements of equity and net income. This is the case because the statement of equity is perceived as less important by analysts, compared to the performance statement for their forecasting models.

My empirical results do not support my hypotheses that analysts’ cash flow forecast dispersion and cash flow forecast error are smaller in the period after the FAS 130 update than in the period from 1999 to 2012, based on the location of comprehensive income information. My results do show, based on quantitative research on S&P 500 data during the period from 1999-2014, that analyst’ forecast error and analysts’ forecast dispersion are bigger in the