Credibility of Management Forecast Disaggregation: International Evidence

Transcription

1 Credibility of Management Forecast Disaggregation: International Evidence Wenjing Li Jinan University Jeff Ng The Chinese University of Hong Kong Albert Tsang York University Oktay Urcan University of Illinois at Urbana-Champaign August 2016 *We acknowledge the helpful comments of seminar participants at the 2014 Global Issues in Accounting Conference hosted by the University of Chicago Booth School of Business and the University of North Carolina Kenan-Flagler School of Business, the 9 th Annual Conference of Asia-Pacific Financial Markets, the 2015 Financial Accounting and Reporting Section Midyear Meeting, the 2015 European Accounting Association Annual Congress, the 2015 American Accounting Association Annual Meeting, and the 2015 Auckland Finance Meeting. 1

2 Abstract Using a comprehensive sample hand-collected from the original texts of management earnings forecasts from 28 countries, we examine the role of forecast disaggregation (i.e., management earnings forecasts that contain projections of the bottom-line net income number as well as forecasts of other key earnings performance measure) on the perceived credibility of management forecasts in countries around the world. We find that disaggregated earnings forecasts are perceived to be more informative especially in countries with a strong legal environment. However, in contrast to prior studies which suggest a positive relation between the perceived informativeness of financial information and voluntary disclosures, we find that managers are less inclined to issue disaggregated earnings forecasts in countries with a strong legal environment. Additional analyses reveal that disaggregated earnings forecasts are associated with lower earnings management and facilitate better incorporation of future earnings information into current stock prices. These results are more pronounced in countries with a strong legal environment. Overall, our results suggest that the strength of a country s legal environment can enhance the credibility of disaggregated earnings forecasts by constraining earnings manipulation and accelerating incorporation of earnings information in stock prices. Keywords: forecast disaggregation, legal environment, credibility, earnings manipulation 2

3 1. Introduction Management earnings forecasts represent one of the key voluntary disclosure mechanisms through which firms communicate their private information to market participants (see Hirst et al for a review of management earnings forecasts literature). However, management earnings forecasts are voluntary, tend to be future orientated, are subject to limited reporting guidelines, and potentially suffer from credibility concerns. Therefore, the effectiveness of management earnings forecasts in reducing information asymmetry between firms and investors depends crucially on the perceived credibility and informativeness of forecasts (Ball et al. 2012). Earnings forecast disaggregation (i.e., earnings forecasts that contain projections of various key performance measures such as sales, EBITDA, and operating income in addition to the forecast of net income) 1 is one of the major characteristics of management forecasts (Rogers and Stocken 2005; Lansford et al. 2013; Chen et al. 2009). 2 Hutton et al. (2003), Hirst et al. (2007), Lansford et al. (2013), and Merkley et al. (2013) suggest that a commitment to forecast disaggregation allows managers to credibly disclose private information, which in turn increases the credibility of management forecasts. However, other studies, such as Han and Wild (1991), and Chen et al. (2009) find that disaggregated forecasts do not necessarily increase the credibility of management forecasts. Differences in the legal and regulatory environment across countries have long been recognized as important determinants of a country s information environment (e.g., Ball et 1 Throughout this paper, we use forecast disaggregation to refer to both managers choice of forecasting multiple performance items and the degree to which the forecasts are disaggregated, and forecast items to refer to the specific accounting performance measures predicted in management forecasts. 2 Consistent with these studies, we find that nearly half of the management forecasts in our international management forecasts sample are disaggregated. 3

4 al. 2000; Bushman et al. 2004). Prior research suggests that the legal and regulatory environment not only influences the ability of management forecasts to reduce the information asymmetry between insiders and capital market participants (Baginski et al. 2002; Hirst et al. 2008; Beyer et al. 2010), but also managers voluntary disclosure decisions. One shortcoming of the prior literature examining the credibility of management forecast disaggregation is that most of these studies are conducted in the U.S., a country known for its strong legal environment, making it difficult to disentangle the effect of forecast disaggregation from the legal environment. In this study, we re-examine the role of disaggregation on credibility of management forecasts using an international sample through the lens of country-level legal environment. Theoretically, it is not clear ex ante whether, and to what extent, forecast disaggregation affects the perceived credibility of management forecasts in countries with different legal and regulatory environment. Hirst et al. (2007) suggest that forecast disaggregation could play a more important role in signaling the credibility of managers private information disclosure when managers incentives to manipulate earnings are high. Presumably, managers could have stronger incentives to manipulate earnings in countries where investors can exert greater pressure on firms managers to meet or beat their own forecasts, such as in countries with a stronger legal and regulatory environment. 34 On the other hand, Leuz et al. (2003) find lower earnings management in countries with a strong 3 In line with this argument, DeFond and Hung (2004) show a stronger association between CEO turnover and poor performance in countries with strong legal environment. The availability of class action lawsuits in different countries (Leuz 2010) also suggests the possible variation in incentives of meeting managers own forecasts issued previously across countries through earnings manipulations. 4 In fact, survey evidence indicates that 69% of portfolio/fund managers and analysts believe that providing earnings forecasts increases future earnings management (AIMR 2003). Further supporting this view, Krishnan et al. (2012) find a positive association between issuance of management earnings forecasts and audit fees suggesting that auditors consider management s forecast behavior to be of higher business risk via greater potential of earnings manipulation. 4

5 legal and regulatory environment. Similarly, DeFond et al. (2007) show that earnings announcements are more informative in countries with strong investor protection. Therefore, to the extent that investors tend to be less concerned about managers expropriating behaviour and the credibility of financial disclosures provided by managers voluntarily, commitment to disaggregated forecasts might play little role in enhancing the perceived informativeness of earnings forecasts in countries with a strong legal environment. 5 Moreover, prior research predicts that committing to a higher level of corporate transparency could play a less important role in reducing information asymmetry in countries with strong legal institutions (Lang et al. 2012). Following this view, one would also predict a weaker association between committing to disaggregated forecasts and stock market reaction in countries with a stronger legal and regulatory environment. As such, to the extent that managers incentives to manage earnings, particularly following the issuance of management earnings forecasts, vary with the legal and regulatory environment of a country, we expect an international sample with differing strength of capital market regulations and legal environment to provide a cleaner setting to examine the importance of disaggregated management forecasts in signaling the credibility of management forecasts. To examine our research questions, we create a comprehensive international sample of standalone management earnings forecasts (i.e., earnings forecasts that are not bundled with any other types of corporate announcement such as earnings announcements) 6 issued 5 Further supporting this view, Ball et al. (2012) argue and show that the quality of a firm s mandatory disclosure could lend credibility to its voluntary disclosure and thus mandatory and voluntary disclosures are complementary. 6 Since the issuance of earnings forecasts that are bundled with earnings announcements has become increasingly common in recent years (Rogers and Van Buskirk 2013), we exclude all the bundled earnings forecasts from our sample and focus only on standalone earnings forecasts issued by firms. In robustness tests, we repeat all of our tests on the full sample without excluding management earnings forecasts issued concurrent with other announcements (i.e., bundled earnings forecasts) and find our conclusion to be unchanged. 5

6 by firms from 28 countries. Specifically, we construct this dataset by hand-collecting detailed management forecast disaggregation information including (1) whether a forecast is disaggregated, (2) the level of forecast disaggregation, and (3) each performance item forecasted from the original texts/narratives of management earnings forecasts provided by the S&P Capital IQ (CIQ) database. 7 After controlling for an array of firm-, industry-, and forecast-level variables that might affect the informativeness of management forecasts, our results consistently find a positive association between forecast disaggregation and the informativeness of management forecasts, as measured by the absolute value of the two-day cumulative market-adjusted returns surrounding the management forecast release date. This finding thus provides empirical support for Hirst et al. (2007) and suggests that investors from around the world consider forecast disaggregation to be credibility enhancing. More importantly, we find that relative to disaggregated earnings forecasts issued in countries with a weak legal environment, disaggregated earnings forecasts issued in countries with a strong legal environment are associated with greater stock market reactions. We next investigate managers incentives to provide disaggregated management forecasts as a function of a country s legal environment. We document that a strong legal environment appears to discourage firms from providing disaggregated earnings forecasts, despite the greater price informativeness associated with disaggregated earnings forecasts issued by firms in such countries. This finding is surprising because Ball et al. (2012) and Lennox and Park (2006) suggest that firms have greater incentives to provide voluntary disclosures when investors perceive such disclosures to be more informative. However, 7 Due to the substantial time and efforts involved in reading and coding each management earnings forecast text issued by firms covered by CIQ, our sample period ends at

7 our findings are consistent with the theoretical argument by Dutta and Gigler (2002) that forecast disaggregation plays an important role in enhancing the perceived credibility of management earnings forecasts, particularly in instances where managers tend to have higher incentives to meet their own forecasts through earnings manipulation by constraining the potential of future earnings management. Earnings forecasts vary not only in the likelihood and level of forecast disaggregation, but also in the choice of forecasted items. In additional analysis, we further examine whether the perceived informativeness of disaggregated earnings forecasts varies with the choice of performance items included in the forecast while controlling for the level of forecast disaggregation. Our results show that among the key performance items commonly bundled with earnings forecasts (such as revenues, operating earnings before interest, income taxes, depreciation, and amortization (i.e., EBITDA), operating income before income taxes (i.e., operating income), income before income taxes, and income before extraordinary items and discontinued operations), earnings forecasts bundled with the projection of future revenues appear to be valued greater by investors than forecasts bundled with other common performance items, especially in countries with a strong legal environment. Given that revenues are more difficult to manipulate than other performance items (Ertimur et al. 2003; Ghosh et al. 2005), our evidence once again supports the conjecture that by constraining the potential of earnings management through providing revenue forecasts, disaggregated earnings forecasts play an effective role in enhancing the credibility of management earnings forecasts. We next investigate the channel through which forecast disaggregation improves informativeness of management earnings forecasts. In particular, we directly test the 7

8 possible role of disaggregated forecasts on future earnings manipulation. Consistent with the argument that providing more disaggregated earnings forecasts has the potential to constrain future earnings manipulation, we find a negative association between forecast disaggregation and future earnings management, especially in countries with a strong legal environment. Furthermore, we also find that disaggregated earnings forecasts appear to accelerate the incorporation of future earnings information into current stock price more in countries with a strong legal environment. Taken together, our evidence provides support for the conjecture that disaggregated earnings forecasts constrain the potential that managers might manipulate earnings. Our study advances the literature in several ways. First, management earnings forecasts are an important form of corporate voluntary disclosure that can mitigate capital market resource misallocation by reducing information asymmetry between firm insiders and investors (Coller and Yohn 1997; Healy and Palepu 2001; Hirst et al. 2008; Beyer et al. 2010; Shroff et al. 2013). Although there is a large body of research examining the determinants of, and market reactions, to management earnings forecasts in the U.S., 8 limited empirical evidence exists for management earnings forecast practices and their consequences in an international setting (Hirst et al. 2008). 9 8 For example, prior studies document that management earnings forecasts have the potential to affect stock prices (Pownall et al. 1993), analysts forecasts (Baginski and Hassell 1990), cost of capital (Frankel et al.1995; Shroff et al. 2013), and firms expected litigation costs (Skinner 1994; Kasznik and Lev 1995). 9 A few notable exceptions include Baginski et al. (2002) who compare management forecasts between U.S. and Canadian firms, and Kato et al. (2009) who examine management earnings forecasts in Japan where management forecasts are effectively mandated. More recently, Cao et al. (2016) examine the international differences in the effect of management earnings forecasts on the cost of equity capital across 31 countries and show the ultimate effect of management earnings forecasts in a country varies with a number of countrylevel institutional factors. However, no existing study, to the best of our knowledge, examines the role of cross-country variation in legal environment on the relation between management forecasts disaggregation (or different performance measure forecasted by managers) and the perceived informativeness of management earnings forecasts, as we do. 8

9 In addition, while theory suggests that disaggregated earnings forecasts that constrain managers in terms of opportunities for subsequent earnings management enhance the credibility of earnings forecasts (Dutta and Gigler, 2002), empirical evidence on the role of forecast disaggregation in enhancing the credibility of management earnings forecasts based on U.S. data is mixed. By taking advantage of the cross-country differences in legal environment which presumably has an effect on opportunities and incentives of earnings manipulation, our study extends earnings forecast disaggregation studies conducted in the U.S. (e.g., Han and Wild 1991; Hirst et al. 2007; Chen et al. 2009; Lansford et al. 2013; Merkley et al. 2013) to an international setting and sheds light on the heterogeneity of forecast disaggregation practices across countries around the world. Third, the effect of legal environment on corporate voluntary disclosure behaviour, such as management earnings forecasts has been recognized by both academics and regulators (e.g., Kasznik and Lev 1995; Baginski et al. 2002; Rogers and Van Buskirk 2009; Beyer et al. 2010). Existing studies suggest that differences in legal environment play an important role in shaping firms accounting and disclosure practices (Ball 2001; Bushman et al. 2004; Bushman and Piotroski 2006). However, the evidence to date is limited and inconclusive as to whether and how differences in legal environment across countries affect firms disclosure decisions and the importance of their disclosures. For example, Lang et al. (2012) find that firm-level transparency matters the most in countries with poor investor protection. In contrast, Hansen et al. (2015) show that greater firm-level transparency is positively associated with foreign capital only in countries with strong investor protection. Using the CIFAR disclosure scores as a measure of corporate 9

10 disclosure level, Francis et al. (2005) suggest that the incentives and effectiveness of firms disclosures are independent of country-level institutional factors. Our findings suggest that a country s legal environment is a major factor contributing both to the strengthened effectiveness enhancing the credibility of management earnings forecasts and to the reduced incentives to provide disaggregated earnings forecasts. These findings are consistent with the argument that corporate financial reporting and disclosure practices are shaped by both firms disclosure incentives and constraints (e.g., Beyer et al. 2010). In addition, our results also suggest that absent strong management incentives to build a reputation for credible communication of forwardlooking information and institutional factors to enhance its credibility, disaggregated earnings forecasts per se may have limited capital market consequences. Given the importance of voluntary disclosures to the functioning of capital markets (Jennings 1987; Mercer 2004; Gu and Li 2007), our study thus has important practical implications for both corporate managers and investors. 10 The remainder of this study is organized as follows. In Section 2, we review the related literature and develop our hypotheses. We discuss our data and descriptive statistics in Section 3. Section 4 provides details of empirical methodology. The results are discussed in Section 5, and Section 6 concludes the paper. 2. Literature Review and Hypotheses Development 10 For example, a better understanding of the variation in the credibility-enhancing effect of forecast disaggregation in different countries could help managers to make forecast decisions that are more likely to optimize the value of their forecasts and maximize their firms capital market benefits. 10

11 Previous studies suggest that a key factor motivating firms voluntary disclosures, such as the issuance of management earnings forecasts, is to develop a reputation for accurate and transparent reporting (Graham et al. 2005). An important mechanism through which managers can signal or enhance the credibility of their voluntary earnings forecasts is a commitment to forecast disaggregation, i.e. provide the forecasts of other major line items reported on the income statement in addition to a forecast of the bottom-line net income numbers (Hirst et al. 2007; Hirst et al. 2008). Using an experimental setting, Hirst et al. (2007) argue and show that by issuing disaggregated earnings forecasts that precommit managers to a specific path to achieve their earnings target, managers can mitigate investors skepticism and concern regarding the potential of self-interest associated with their forecasts, which in turn increases investors perceived credibility of the forecasts. However, there is mixed empirical evidence on the importance of forecast disaggregation in the perceived credibility of earnings forecasts. On the one hand, Hutton et al. (2003) document that verifiable forward-looking information about earnings components increase credibility of good news management forecasts. Lansford et al. (2013) find that forecast disaggregation leads to more timely analysts forecast revisions and a larger reduction in analysts disagreement than aggregated forecasts, suggesting that forecast disaggregation enhances a firm s information environment. Merkley et al. (2013) document that forecast disaggregation increases analysts sensitivity and reliance on the earnings news in management forecasts suggesting that analysts consider disaggregated forecasts more credible than aggregated forecasts Earlier studies, such as Trueman (1986) argues that disaggregated earnings forecasts that contain supplemental information may signal that managers have better information or superior forecasting ability. 11

12 On the other hand, Han and Wild (1991) compare the stock market reaction to earnings forecasts released together with revenue forecasts to stand-alone earnings forecasts issued without revenue forecasts and document that stand-alone earnings forecasts are more price informative than those bundled with revenue forecasts. Similarly, Chen et al. (2009) find that in some cases (e.g., bad news forecasts), disaggregated earnings forecasts are significantly less accurate and more optimistic than aggregated forecasts suggesting that disaggregated earnings forecasts are associated with lower information quality than aggregated earnings forecasts. Together, these findings challenge the view that providing disaggregated forecasts could signal higher credibility of management earnings forecasts. Given the opposite views and inconsistent conclusions on the credibility of disaggregated earnings forecasts in the literature, we present the first hypothesis formally as follows: Hypothesis 1: Management earnings forecast disaggregation has no impact on management forecast informativeness. All the studies mentioned above are conducted in a single country setting (i.e., U.S.) which is known by its strong legal environment which likely imposes high pressure on managers to meet previously issued management earnings forecasts. This might result in higher credibility of disaggregated earnings forecasts (e.g., due to high litigation risk) and lower quality of disaggregated forecasts (e.g., due to high potential of earnings management to meet such earnings forecasts). Although existing studies suggest the importance of legal environment in affecting managers voluntary disclosure incentives and also on the informativeness of voluntary disclosure, evidence from a single country 12

13 provides limited insight on whether and how the legal environment of a country could have an effect on firms voluntary disclosure behavior. As such, in this study, we examine the question of whether and how disaggregated management earnings forecasts could have an effect on forecast credibility using an international setting because institutional factors that can potentially influence the role of forecast disaggregation can vary substantially across countries. We next examine the impact of country-level legal environment on the relationship between disaggregation and management earnings forecast credibility. On the one hand, in countries with strong legal environment, disaggregated earnings forecasts could be perceived by investors to be more informative because issuing more disaggregated earnings forecasts in such countries could impose a higher constraint on managers flexibility for subsequent earnings manipulation. This line of reasoning follows the argument that disaggregated earnings forecasts signal managers commitment to higher financial reporting quality through lower opportunities for subsequent earnings management which could, in turn, be perceived by investors to be more informative (Dutta and Gigler 2002; Hirst et al. 2007). On the other hand, given the existence of better information environment and higher quality of mandatory disclosure in countries characterized by strong legal environment, the alternative prediction is that forecast disaggregation could play limited role in enhancing the perceived credibility of voluntary disclosures in such countries. This argument is in line with existing theory which suggests that a better information environment arising from higher mandatory disclosure requirements could limit the ability of additional disclosure 13

14 to reduce information asymmetry (e.g., Yohn 1998; DeFond et al. 2007). Following these views, we present the second hypothesis formally as follows: Hypothesis 2: A country s legal environment has no effect on the informativeness of disaggregated earnings forecasts in that country. Perceived higher informativeness of voluntary disclosure has the potential to increase managers incentives to issue disaggregated earnings forecasts (Ball et al. 2012; Lennox and Park 2006). However, higher costs associated with issuing misleading management forecasts and the constrained managerial ability for earning manipulation, especially in countries with a strong legal environment, could discourage managers from issuing disaggregated forecasts. 12 Rogers and Stocken (2005) argue and find that the likelihood of issuing self-serving forecasts could be moderated by investors ability to detect misrepresentation. This finding suggests a lower incentive to provide disaggregated earnings forecasts to enhance the credibility of managers earnings forecasts in countries with strong legal environment. In sum, whether and how a country s legal environment affects managers incentives to issue disaggregated earnings forecasts are empirical questions. Based on the above discussions, we state our third hypothesis as follows: Hypothesis 3: A country s legal environment has no effect on the likelihood of managers issuing disaggregated earnings forecasts in that country. 3. Sample and Descriptive Statistics We obtain a comprehensive international sample of management earnings forecasts data from S&P Capital IQ (CIQ hereafter). This database provides the original text / 12 For example, prior studies suggest that fear of potential legal liability deters managers from making management forecasts (Baginski et al. 2002; Hirst et al. 2008; Beyer et al. 2010). 14

15 narrative of management forecasts for firms across a large number of countries/regions starting from the first year for which CIQ provides comprehensive coverage of international firms, following its acquisition by McGraw Hill. According to CIQ, the raw text forecasts are extracted from various sources such as newspapers, regulatory filings, subscriptions, and transaction announcements. We exclude management forecasts that are bundled with any other type of corporate announcement such as earnings announcements (Rogers and Van Buskirk 2013) and observations with missing firm- or country-level control variables. We further exclude Japan because management forecasts in Japan are de facto mandatory (Kato et al. 2009). Our final sample for the 2004 to 2009 period consists of 28 home countries, 13 representing 4,830 unique firms and 13,062 individual unbundled management earnings forecasts. 14 To obtain detailed information on levels of earnings forecast disaggregation and disaggregated forecast items, we manually identify and collect all of the performance measures projected in each forecast. We start with performance measures identified by Barton et al. (2010) as likely to be important to investors globally, namely (1) Sales, (2) EBITDA (operating earnings before interest, income taxes, depreciation, and amortization), (3) OPINC (operating income before income taxes), (4) IBTAX (income before income taxes), (5) IBXIDO (income before extraordinary items and discontinued operations), and (6) NI (net income). All management earnings forecasts that contain the projections of the bottom-line net income number as well as forecasts of other key earnings performance 13 In our study, we define a firm s home country based on the country of the firm s primary stock listing. Our results are robust to alternate definitions of home countries, including defining a firm s home country based on a firm s headquarters or incorporation country. 14 As the data-collection process requires extensive resources and effort, our sample ends in Examples of disaggregated forecasts can be found in the Appendix 1. 15

16 measures identified above are defined as disaggregated earnings forecasts. Accordingly, we define an indicator variable, D_Disag, which takes the value of one for management earnings forecasts that are disaggregated, and zero otherwise. In addition, we also define the total number of unique performance measures projected in each forecast as NumItems, where a larger value indicates a more disaggregated earnings forecast. 15 Panel A of Table 1 provides the descriptive statistics of our variables of interest by country. For our full sample, the average NumItems is 1.52, suggesting that many earnings forecasts contain more than one performance measure. The worldwide average of D_Disag is 47.2 percent, indicating that nearly half of the earnings forecasts worldwide are disaggregated. From columns 1 and 2, we observe that 6,487 (1,979) forecasts (forecasting firms) are from the U.S., representing 49.7 percent (41 percent) of the global total. 16 Other well-represented countries in our sample include Australia (10.6 percent), the U.K. (5 percent), Germany (3.8 percent), and Hong Kong (3.3 percent). The average level of forecast disaggregation (NumItems in column 3) in each country ranges from 1.13 (South Africa) to 1.86 (Greece) and the average percentage of forecasts that are disaggregated (D_Disag in column 4) ranges from 11.8 (South Africa) to 65 (Finland). Columns 5 to 9 in Panel A of Table 1 show the likelihood that an individual performance measure is included in an earnings forecast in each country. Consistent with prior management forecasts studies, the results indicate sales is a commonly forecasted 15 Forecasts often include several related forecast items (e.g., earnings, earnings per share, and earnings growth). Because the underlying performance measure of such forecast items is the same (i.e., earnings), these are coded as one unique item. Unsurprisingly, there is a large variation in management forecasts practices across firms from around the world. For example, some forecasts contain projections of balance sheet items such as liabilities and future leverage in addition to income statement items. Since our interest is in management earnings forecast disaggregation, we only retain and code the six income statement items in our NumItems measure. 16 In robustness checks, we exclude forecasts made by U.S. firms from the sample and our results remain unchanged (see section 5.5.6). 16

17 performance measure around the world. Specifically, we find that on average 43.7 percent of net income forecasts also include sales. Due to their low forecast frequency, we combine EBITDA, OPINC, IBTAX, and IBXIDO, i.e., the middle four income related items reported on firms income statements, into a single measure called MID4 and find that 8 percent of management forecasts include at least one of these items. 17 Panel B of Table 1 reports the descriptive statistics of major forecast variables by industry. The computers industry is heavily represented, with 1,772 forecasts (13.6 percent) and with the highest average level of forecast disaggregation measured by the number of items included in each forecasts (each forecast contains 1.87 items on average). This finding is consistent with Gu and Li (2007) who suggest that investors have more credibility concerns regarding the voluntary disclosures made by high-tech firms. Other well-represented industries include financial (9.2 percent) and services (7.6 percent). The variation in the likelihood that various performance measures are included in a forecast across industries is also notable. Although approximately 79 percent of forecasts from firms in the computers industry include sales projections, only 9.5 and 12.6 percent of such forecasts are issued by firms in the utilities and financial industries, respectively. Firms in the utilities industry have the highest likelihood of forecasting EBITDA (7.6 percent). Panel C of Table 1 reports the descriptive statistics by year and indicate that our sample observations are evenly distributed over our sample period. We report the descriptive statistics for our test variables, other forecast properties, and other control variables in Table 2. In our sample, approximately 10 percent of forecasts 17 These overall statistics are in line with those documented by prior studies. For example, Hirst et al. (2007, page 814, footnote 2) show that in the U.S. about 71 percent of forecasts contain earnings and revenue forecasts and 29 percent of forecasts contain forecasts of other line items. 17

18 project a loss (F_Loss). The majority of forecasts are fairly precise, containing a range or point forecast. 11 percent of forecasts include an external or internal attribution (i.e., managers use either external or internal reasons to explain their forecasts). On average, forecasts are issued half a year in advance (178 days) and about half (46 percent) of all forecasts contain good news. We include descriptive statistics for forecast error (F_Error), calculated as the scaled absolute percentage difference between forecasted performance and actual performance, but note that this measure requires the forecast to be a point or closed-range estimate, reducing the sample size from 13,062 observations to 6,586. Our sample firms have average total assets of US$1,236 million, are followed by 14 analysts, are audited by Big 4 firms 74 percent of the time, and have average institutional ownership of 44 percent, suggesting that our sample firms are large. 4. Research Design 4.1 Informativeness of Management Earnings Forecast Disaggregation To test our first hypothesis (H1) on the informativeness of disaggregated management earnings forecast, we estimate the following OLS regression model: AAAAAAAAAAAA_1 = ββ 0 + ββ 1 FFFFFFFFFFFFFFFF DDDDDDDDDDDDDDDDDDDDDDDDDDDD + OOOOheeee FFFFFFFFFFFFFFFF PPPPPPPPPPPPPPPPPPPP + CCCCCCCCCCCCCC VVVVVVVVVVVVVVVVVV + CCCCCCCCCCCCyy, YYYYYYYY, aaaaaa IIIIIIIIIIIIIIII IIIIIIIIIIIIIIIIIIII + eeeeeeeeee (1) In equation (1), ABSCAR_1 is the absolute value of the cumulative market-adjusted return including the forecast date and the day following. We estimate Forecast Disaggregation with one of two variables, D_Disag, an indicator variable that takes the value of one if a management forecast includes multiple performance measures, and zero otherwise, or 18

19 NumItems, a count variable indicating how many performance items are included in each forecast (i.e., the degree of forecast disaggregation). Because this model is conducted at the forecast-level, it is important to control for the forecast attributes of each forecast. We include several forecast attributes in equation (1) to control for differences in forecast characteristics including whether a forecast predicts a loss (F_Loss), the degree of forecast precision (F_Prec), whether a forecast contains either an internal or external attribution (F_Attr), the timeliness of a forecast (F_Time), and whether a forecast is perceived to contain good news (F_Good), estimated by the two-day market adjusted return of the forecast. 18 We control for variables identified in prior studies which potentially influence the informativeness of management earnings forecasts, including the log of total assets to control for firm size (Assets), the number of analysts following a firm to control for overall information environment (Analysts), whether a firm has a Big 4 auditor to control for audit quality (Big4), the percentage holding of institutional investors (IO), whether a firm is in the high tech industry (HighTech), whether a firm reports a loss (Loss), the number of exchanges on which a stock is listed (StkExch), whether a firm is cross-listed in the U.S. as an ADR (ADR), the standard deviation of reported earnings per share scaled by the average total assets per share to control for earnings volatility (EarnVol), the level of a firm s assetscaled total accruals to control for financial opacity (Accruals), the level of capital expenditures less operating cash flow scaled by the total capital expenditures to control for dependence on external financing (ExtDep), the industry average likelihood of issuing 18 As a robustness check, we also include a control for the forecast error (F_Error) calculated as the deviation between forecasted performance and actual performance, for the subset of forecasts where this variable is available. The inferences from our results, untabulated, remain unchanged. 19

20 disaggregated forecasts to control for industry-level demand for disaggregated information (Likelihood), and the sales-based Herfindahl index multiplied by negative one to control for industry-level competition (Competition). We provide detailed definitions of these variables in Appendix 2. We estimate equation (1) at the forecast-level and include country, year, and industry indicators in the model and cluster all of the standard errors by both firm and year Country-level Legal Environment and the Stock Market Reaction to Forecast Disaggregation To test our second hypothesis (H2) on the informativeness of management earnings forecast disaggregation across countries, we estimate the following OLS regression model: AAAAAAAAAAAA_1 = ββ 0 + ββ 1 FFFFFFFFFFFFFFFF DDDDDDDDDDDDDDDDDDDDDDDDDDDD + ββ 2 LLLLLLLLLL EEEEEEEEEEEEEEEEEEEEEE + ββ 3 FFFFFFFFFFFFFFFF DDDDDDDDDDDDDDDDDDDDDDDDDDDD LLLLLLLLLL EEEEEEEEEEEEEEEEEEEEEE + OOOOheeee FFFFFFFFFFFFFFFF PPPPPPPPPPPPPPPPPPPP + CCCCCCCCCCCCCC VVVVVVVVVVVVVVVVVV + CCCCCCCCCCCCCC, YYYYYYYY, aaaaaa IIIIIIIIIIIIIIII IIIIIIiiiiiiiiiiiiii + eeeeeeeeee. (2) In equation (2), ABSCAR_1 and Forecast Disaggregation are defined as in equation (1). Our variable of interest is ββ 3 which estimates the relation between the interaction of forecast disaggregation variables and legal environment variables and the stock market reaction to forecasts. We define a country s legal environment, Legal_Avg, as the average of five proxies for a country s legal environment: (1) H_Disclose, an indicator variable that takes the value of one if the disclosure requirements level in a country is above the countrylevel median, and zero otherwise (Frost et al. 2006); (2) H_IP, an indicator variable that takes the value of one if the level of investor protection in a country is above the country- 19 For robustness, we also cluster the standard errors by both country and year, or by both industry and year. In all of these settings, the results are qualitatively similar. 20

21 level median, and zero otherwise (La Porta et al. 2006); (3) H_Pub_Enf, an indicator variable that takes the value of one if the level of public enforcement of securities regulation in a country is above the country-level median, and zero otherwise (La Porta et al. 2006); (4) H_Priv_Enf, an indicator variable that takes the value of one if the level of private enforcement of securities litigation in a country is above the country-level median, and zero otherwise (La Porta et al. 2006); and (5) H_Enforce, an indicator variable that takes the value of one if the level of accounting enforcement is above the country-level median, and zero otherwise (Brown et al. 2014). 20 Other control variables are the same as defined in equation (1). 4.3 Country-Level Legal Environment and Forecast Disaggregation To test our third hypothesis (H3), we examine whether the legal environment of a country is associated with firms decisions to issue disaggregated earnings forecasts by estimating the following regression model: FFFFFFFFFFFFFFFF DDDDDDDDDDDDDDDDDDDDDDDDDDDD = ββ 0 + ββ 1 LLLLLLLLLL EEEEEEEEEEoonnnnnnnnnn + CCCCCCCCCCCCCC VVVVVVVVVVVVVVVVVV + YYYYYYYY aaaaaa IIIIIIIIIIIIIIII IIIIIIIIIIIIIIIIIIII + eeeeeeeeee (3) In equation (3), β1 is our coefficient of interest and estimates the relation between a country s legal environment and firms forecast disaggregation decisions. As in equation (2), we use D_Disag and NumItems as proxies for Forecast Disaggregation. Models with D_Disag (NumItems) as the dependent variable are estimated using logistic (Poisson) 20 All five of the country-level legal environment proxies used in our study are highly correlated with each other, with Pearson coefficients range from 0.26 to As such, we estimate the legal environment using an average of the five measures. Our inferences remain unchanged if include the legal environment variables individually in the regressions as discussed in section below. 21

22 regressions. We use the same legal environment variables and control variables as in equation (1). 5. Empirical Results 5.1 Informativeness of Disaggregated Management Earnings Forecasts Univariate Analysis Our first hypothesis (H1) examines the relation between management earnings forecast disaggregation and forecast informativeness. We start by tabulating the mean ABSCAR of management forecasts by whether a forecast is disaggregated (D_Disag) and by how many items are included in a forecast (NumItems) in Panel A of Table 3. We estimate three values of ABSCAR; namely, the ABSCAR from the forecast announcement date to day +1, +7, and +30, which we denote ABSCAR_1, ABSCAR_7, and ABSCAR_30, respectively. We also report the difference in ABSCAR between disaggregated (D_Disag=1) and aggregated (D_Disag=0) forecasts and indicate whether such differences are significant (based on a t-test of the difference in means) across the two groups. The results shown in Table 3 Panel A indicate that disaggregated forecasts have stronger stock market reactions than forecasts that are not disaggregated. Evidence from Table 3 Panel A does not indicate that forecasting additional items beyond the second is more informative Cross-sectional Regression Analysis Next, we formally investigate H1 and estimate the relation between D_Disag (NumItems) and ABSCAR_1 using multivariate regressions based on equation (1). These results are reported in models 1 and 2 in Panel B of Table 3. Consistent with the argument that forecast disaggregation enhances the perceived credibility of forecasts (Hirst et al. 22

23 2007), we find a consistently positive coefficient on forecast disaggregation after controlling for other forecast attributes. We document that issuing a disaggregated forecast (forecasting an additional item) increases ABSCAR_1 by 0.45 percent (0.34 percent). These magnitudes are economically meaningful given that mean (median) ABSCAR_1 in our sample is 5.72 (3.20) percent Legal Environment and Informativeness of Disaggregated Management Earnings Forecasts Univariate Analysis Legal Environment on Forecast Informativeness Our second hypothesis (H2) examines whether the stock market reaction associated with disaggregated management earnings forecasts varies a country s legal environment. We first tabulate the difference in informativeness of forecasts issued by firms from countries with more or less stringent legal environments. As shown in Panel A of Table 4, management earnings forecasts are consistently perceived by stock market participants to be more informative when issued by firms from countries with a strong legal environment Cross-Sectional Regression Analysis We next formally test H2 by estimating Equation (2). These results are tabulated in models 1 and 2 in Panel B of Table 4. We find a positive coefficient on the interaction term between forecast disaggregation (D_Disag or NumItems) and the legal environment. Interestingly, the main effect on our D_Disag / NumItems variable is negative, suggesting that disaggregation by itself does not increase forecast informativeness in countries with weak legal environments. As such, our findings support the argument that a strong legal 21 We also estimate equation (1) using ABSCAR_7 and ABSCAR_30. These untabulated results are all qualitatively similar to those reported in Table 3. 23

24 environment could play an important role in the credibility-enhancing effect of the disaggregated forecasts Country-Level Legal Environment and Forecast Disaggregation Univariate Analysis Panel A of Table 5 presents the univariate analysis of the relation between the country-level legal environment variables and forecast disaggregation. In particular, it tabulates the likelihood of management forecasts including more than one forecast item (D_Disag) and the country-level average number of items included within a forecast (NumItems) by the countries level of legal environment partitioned at the median. The results show that across three of our five legal environment proxies (IP, Pub_Enf, and Priv_Enf), firms are significantly less likely to issue disaggregated forecasts in countries with stronger legal environments. Thus, the univariate results provide preliminary evidence that strong legal environments deter firms from providing disaggregated forecasts Regression Results Panel B of Table 5 report the regression estimates of equation (3) to formally test H3. Our results show that the likelihood of firms issuing disaggregated forecasts (D_Disag) and the degree of forecast disaggregation (NumItems) are negatively and significantly related to our proxy for a strong legal environment. In terms of economic significance, the results shown in Panel B of Table 5 suggest that a strong legal environment reduces the likelihood of firms issuing a disaggregated forecast by about 60.3 percent on average and 22 In robustness tests, we also examine the association between ABSCAR and forecast disaggregation on two subsamples, partitioned by country-level legal environment, and find results consistent with the findings on the interaction variable. For example, the estimated coefficient on D_Disag is (with t-value = 4.38) for countries with high Enforce, whereas it is (with t-value = -0.92) for countries with low Enforce. We find similar result when forecast disaggregation is measured by NumItems. 24

25 reduces the likelihood that firms include an additional item in their earnings forecast by 16.4 percent. 23 The estimated effect of several other control variables on D_Disag and NumItems are also worth noting. Firms with more volatile earnings (EarnVol) and those from industries where a larger proportion of firms provide disaggregated forecasts (Likelihood) are more likely to issue disaggregated forecasts or issue forecasts with higher levels of disaggregation. We find that firm size (Assets), firms audited by a Big4 auditor, and the number of stock exchanges on which a company is listed (StkExch) are significantly negatively related to forecast disaggregation. This evidence suggests that firms with richer information environments tend to issue less disaggregated management forecasts. We also find that analyst following (Analyst) and institutional ownership (IO) are positive and significant suggesting that analysts cover, and institutions invest in, firms with better management disclosure policies. 5.4 Additional Analysis and Robustness Tests Separate Legal Environment Variables Our primary analyses of H2 and H3 examine the impact of Legal_Avg on the informativeness of management earnings forecasts and firms likelihood to issue disaggregated management earnings forecasts. As an alternate specification, we separately examine the relation between each of our five legal environment variables (H_Disclose, H_IP, H_Pub_Enf, H_Priv_Enf, and H_Enforce) and the informativeness of, and likelihood that managers issue, disaggregated earnings forecasts. These results, 23 Economic significance is estimated by exponentiating our coefficients of interest and subtracting one, which estimates the change in likelihood that a firm will issue a disaggregated forecast (forecast an additional forecast item) from the logistic (Poisson) models. 25

26 untabulated, are similar to those reported in Tables 4 and 5 and consistently show that disaggregated management earnings forecasts are more informative when issued by firms from countries with strong legal environments, but that managers in such countries are less likely to issue such forecasts Informativeness of Management Earnings Forecasts with other Forecast Items Our first hypothesis examines the informativeness of disaggregated management earnings forecasts. In addition to choosing to issue disaggregated forecasts, firms can also choose the items to include in their earnings forecasts. We examine whether the choice of forecast items to include in a forecast is also related to its informativeness by replacing Forecast Disaggregation in Equation 1 with indicator variables for whether a forecast includes sales (D_Sales) and the four performance items between sales and net income in the income statement (D_Mid4). These results are reported in model 3 of Table 3 Panel B. We find a positive and significant coefficient on both D_Sales and D_Mid4, suggesting that investors generally find sales forecasts and forecasts including the performance items between sales and net income to be informative Country-level Legal Environment and Forecast Item Choice Our second hypothesis examines whether country-level legal environment impacts the informativeness of disaggregated management earnings forecasts. We do not have much guidance from prior studies on how the legal environment of a country would affect the informativeness of management forecasts containing different accounting performance measures. For completeness, however, we further examine the relation between legal environment and the informativeness of earnings forecasts containing Sales or Mid4 by estimating the following equation. 26