Customer-Supplier Relationships and Management Earnings Forecasts

Transcription

1 Customer-Supplier Relationships and Management Earnings Forecasts Abstract This paper examines whether customer base composition, i.e., whether a firm s major customers comprise of government entities or publicly traded companies in the U.S., affects the properties of supplier s management earnings forecasts. Using a sample of 1,168 management earnings forecasts from 1998 through 2014, we find that firms whose major customers are government entities, i.e., government suppliers, issue more precise management earnings forecasts than firms whose major customers are public companies, i.e., corporate suppliers. Moreover, when managers disclose new material information to the market, government suppliers issue more accurate earnings forecasts than corporate suppliers. Finally, earnings forecasts issued by government suppliers have greater price impact than those issued by corporate suppliers, but only when companies miss analysts consensus forecast. Collectively, the results demonstrate that having government as a major customer enables supplier firms to issue high-quality earnings forecasts, thus supporting the government effect hypothesis more so than the information spillover hypothesis. Our focus on the dichotomous customer types provides new insight on the impact of government contract on private sector disclosure quality. Key Words: Customer-supplier relationship; Government suppliers; Information spillover; Management earnings forecasts JEL Classifications: G14, D80, M41 1

2 Customer-Supplier Relationships and Management Earnings Forecasts 1. Introduction Customers and suppliers have strong economic links. Economic theory and the supply chain literature demonstrate that information complementarities exist not only among firms in the same industry but also between customers and suppliers (Foster 1981; Olsen and Dietrich, 1985; Hertzel et al., 2008; Pandit et al., 2011). Information about a customer firm is useful in assessing the business performance and prospects of its suppliers (Patatoukas 2012; Guan, et al., 2015; Luo and Nagarajan, 2015). A customer s public disclosure thus constitutes a valuable source of information about the suppliers. Existing research on the information spillover effect along the supply chain does not distinguish between different customer types. A stream of recent studies start to pay attention to the role of the U.S. government as a major customer in supplier firms operating performance and information environment (Cohen and Malloy, 2016; Cohen and Li, 2016a, 2016b; Cohen et al., 2016). We extend the literature by investigating whether different customer base compositions, specifically whether having government or publicly traded companies as major customers, affect the quality of information that suppliers provide to the capital market via management earnings forecasts (MEFs). We examine MEFs because they are important voluntary disclosures through which managers communicate with market participants. Numerous studies on MEFs have been done in the past decades (for example, Healy and Palepu, 2001; Hirst et al., 2008; Leuz and Wysocki, 2016). Despite the extensive research in this area, forecast characteristics appear to be the least understood component of earnings forecasts both in terms of theory and empirical research even though it is the component over which managers have the most control (Hirst et al., 2008). By focusing on the dichotomous types of major customers, we explore two channels the information spillover channel and the government effect channel through which customersupplier relationships affect the characteristics of MEFs. We exploit a financial reporting standard, SFAS No. 131, that requires firms to report the sales to and the identity of any customer that comprises more than 10% of a firm s consolidated sales 2

3 revenues (so-called major customer) if the loss of a customer would have a material adverse impact on the firm. 1 Using this data, we are able to identify firms with different customer base composition. We explore the difference between two types of firms: firms whose major customer are exclusively U.S. federal, state, and/or local governments, which we label government suppliers, versus firms whose major customers are exclusively publicly traded firms, which we label corporate suppliers. A primary advantage of our setting is that it allows us to examine two channels through which customers of different types affect the properties of suppliers management forecasts. The first channel is the information spillover effect that prior literature has shown to exist between customers and suppliers. Since publicly traded firms make both mandatory and voluntary disclosure to the public and are analyzed by financial intermediaries such as analysts, there is a great amount of public information about these firms. To the extent that a customer's disclosures often contain useful information about its suppliers' business performance and prospects, market participants can benefit from the information spillover effect and know more about the suppliers. Unlike the corporate sector, governments do not have equity stakeholders but, instead, receive funding from federal or state governments plus tax revenues levied on taxpayers. Since neither funding party expects a direct financial return on their contribution, there is a lower level of demand for high-quality financial reports from government entities to assess their economic performance (e.g., Copley et al., 1997; Dewyer and Wilson, 1989; Pinnuck and Potter, 2009). Consequently, although the U.S. government entities are subject to reporting standards issued by the Government Accounting Standard Board (GASB), which is the equivalent of the Financial Accounting Standard board (FASB), their disclosure is much less frequent and less timely than those made by public firms. 2 The usefulness and relevance of financial statements depends on the 1 Ellis et al. (2012) provide a summary of the evolution of disclosure rules regarding customer information in their Appendix A. 2 In the U.S., the use of accrual accounting by local government is mandated by state or is at the discretion of the local government (Baber and Gore, 2008; Gore 2004). Prior research suggests that municipal governments are often very slow in producing and disclosing financial statements, with the average time for filing compulsory statements taking over twice as long as the SEC-mandated time for publicly traded corporations. There are typically no governmental rules or explicit penalties associated with delayed reports (Henke and Maher 2016). A global survey by PwC reveals that 63% of central governments publish their financial statements within six months of the closing dates, 37% of governments do not; 90% of governments have their accounts audited, but 19% of them do not make the auditor s report publicly available (PwC 2013). The U.S. Securities and Exchange Commission (SEC) noted that municipal 3

4 timeliness of the information provided to decision-makers. Therefore, compared to corporate sector, we assume that there is much less public information disclosed by the government that is useful and relevant for investors to assess financial performance of government suppliers. Following this line of reasoning, we posit that management forecasts issued by government suppliers would be less precise, less accurate, and have greater price impact than those issued by corporate suppliers. The second channel is through the business transactions with the U.S. government. A stream of recent working papers (e.g., Cohen and Malloy, 2016; Cohen and Li, 2016a, 2016b; Cohen, et al., 2016) examine the role of government as a major customer in affecting suppliers capital investment, operating performance, cash holdings, information environment, and loan contracts. The U.S. government as a major customer is unique in many aspects. For example, government procurement is subject to Federal Acquisition Regulation (FAR) and often affected by political and social welfare considerations. More importantly, the federal government has power in taxation and law enforcement, with lower solvency and bankruptcy risk than average customers do. Cohen and Malloy (2016) document an adverse effect of having government as a major customer. They show that government-linked firms invest less in physically and intellectual capital and have lower future sales growth. To the contrary, Cohen and Li (2016a) provide evidence that relative to corporate customers, government customers help suppliers in achieving higher efficiencies in asset turnover and customer-specific SG&A investments, which in turn results in lower operational uncertainty. Cohen et al. (2016) provide evidence that, when the borrowing firm reports a major government customer, a loan contract contains significantly fewer covenants. Collective evidence in the above papers suggest that having business relationship with the government makes a difference and results in a distinct government effect, which is opposite to the information spillover effect. Thus, we posit that MEFs issued by government suppliers are more precise, more accurate and have less price impact than those issued by corporate suppliers. We study three characteristics of MEFs forecast precision, forecast accuracy, and price impact, because they are commonly studied in the MEF literature and managers have different degrees of bond investors are often not afforded access to the same timely financial information as investors in other U.S. capital markets (SEC, 2012). 4

5 control over them. First, forecast precision is largely under managers control at the time when the forecast is made public. Second, forecast accuracy depends on actual earnings that is released after the forecast. Though managers know better than outsiders do, they also face uncertainty as to what will be the actual earnings. Third, price impact depends on investors inference of new information in management forecast and market conditions. We expect that the differences in MEFs between government suppliers and corporate suppliers are greater in the characteristic over which managers have stronger control. Our sample consists of 1,168 management earnings forecasts spanning from 1998 to In order to better capture the information content contained in a management forecast, we develop a classification scheme that is independent of the IBES code for information content. As discussed in details in Section 4.1, certain IBES codes tend to misclassify MEFs that meet, beat, or miss analyst consensus forecast. We show that our classification of MEFs provides a finer and more accurate proxy for the information content in a management forecast. Our empirical results from both univariate and multivariate analyses show that MEFs issued by government suppliers tend to have higher quality than those issued by corporate suppliers. First, we find that government suppliers issue more precise MEFs than corporate suppliers do. Second, government suppliers issue more accurate MEFs than corporate suppliers do. However, when MEFs merely match analysts consensus forecast, the difference in accuracy between the two groups of suppliers disappears. Third, MEFs issued by government suppliers have greater price impact than those issued by corporate suppliers, but only when they miss analysts consensus forecast. Collectively, our evidence is consistent with the government effect hypothesis more so than the information spillover hypothesis. Our study makes several contributions. First, we contribute to the supply chain literature by identifying customer type as a unique factor in affecting supplier firms voluntary disclosure quality. While investors of corporate suppliers can, in general, enjoy the benefit of information spillover via public information disclosed by customers, perhaps somewhat surprisingly, we show that government suppliers tend to issue higher quality MEFs than corporate suppliers do. We 5

6 interpret the results as evidence that government as a unique and reliable customer has some positive influence on the quality of information that suppliers provide to the capital market. The message from prior research examining the impact of government spending is mixed: some studies document an adverse impact of government as a major customer on firm s incentive to invest and compete (Cohen and Malloy, 2016), while others document the positive effects on suppliers operational uncertainty and loan contract (Cohen and Li, 2016a; Cohen et al., 2016). Our study adds to this line of research by showing the positive impact of government contract on private sector disclosure quality. Second, our paper builds on and contributes to the extensive literature on management earnings forecasts. By classifying government versus corporate suppliers and investigating the difference in the characteristics of MEFs issued by the two groups of firms, our paper answers previous researchers calls for more research on the characteristics of management earnings forecasts (Hirst et al., 2008) and more research into market-wide effects of disclosure and reporting activities (Leuz and Wysocki, 2016). Regarding the latter call for research, prior research in this area focus on information spillover effects of earnings announcement, restatements, misreporting, and investment decisions, etc. Recent work by Shroff et al. (2017) further our understanding of disclosure externalities by showing that information environment of peer firms is negatively associated with a firm s cost of equity capital when there is less publicly available firm-specific information. Nonetheless, information externalities on the quality of management forecasts is relatively underexplored. By showing that government suppliers issue MEFs that are more precise and more accurate than corporate suppliers, we provide interesting insights that having sales transactions with the U.S. government brings greater benefit to suppliers management forecast quality than having sales transactions with firms in the corporate sector. The remainder of the paper is organized as follows. Section 2 discusses related literature and develops hypotheses. Section 3 describes the sample and variable measurement. Section 4 presents the empirical results and Section 5 concludes. 2. Related literature and hypothesis development 6

7 2.1 Related literature Customer-supplier relationship and information spillover Economic theory suggests that economic links between one firm and another can generate information spillover across these firms. The earlier literature examines information transfer among firms in the same industry (e.g., Baginski 1987; Foster 1981; Han and Wild 1990). This literature provides evidence that investors value a firm based on information disclosed by other firms in the same industry. In addition to sharing industry-related common shocks, firms are also economically linked through customer-supplier relationships. A growing supply chain literature examines how the customersupplier relationships influence firms economic fundamentals, stock market valuation, operating and financing policies, financial reporting and disclosure practice, etc. For example, Olsen and Dietrich (1985) find that retailers monthly sales report significantly affect the stock price of their suppliers. Raman and Shahrur (2008) provide evidence that earnings management is used opportunistically to influence the perception of suppliers or customers about the firm s prospects. Hertzel et al. (2008) and Hertzel and Officer (2012) show that customer s bankruptcy filings and financial distress lead to measureable wealth losses to suppliers. Patatoukas (2012) documents a positive association between customer-base concentration and accounting rate of return. Wang (2012) finds that firms relying more on customer-supplier relationships pay significant lower dividends and are less likely to adopt dividend-smoothing policies. Itzkowitz (2013) shows that firms' cash holdings increase in the strength of customer relationship. Cohen and Li (2014) find that the amount of sales between a firm and the U.S. government affects its cash holdings. The strong economic links between customers and suppliers lead researchers to examine the information transfer effects in stock market. Cohen and Frazzini (2008) find that customers' monthly stock returns predict their suppliers' future returns and attribute this predictability to investors reacting to customer news with delay. Menzly and Ozbas (2010) also find that stock returns on economically related suppliers and customers cross-predict each other and the return cross-predictability weakens as the level of analyst coverage and institutional ownership increases. Pandit et al. (2011) report that customers' quarterly earnings announcements have an externality effect on suppliers' stock prices and identify the economic factors that influence the magnitude of 7

8 this information externality. Cheng and Eshleman (2014) show that suppliers stock prices overreact to customers earnings announcements and the overreaction is partially corrected when suppliers announce their own earnings. Using four measures of customer s information quality, including customer s earnings forecast precision, earnings quality, coverage by analysts and credit rating agencies, Radhakrishnan et al. (2014) find that the quality of information provided by the customers to the capital market has a spillover effect on the input market, that is, to help suppliers improve operating performance. Most of the studies discussed above on information transfer do not distinguish between different types of customers. In this study, we aim to extend this line of inquiry and ask whether customer base composition, i.e., whether a supplier firm s major customers are government entities or publicly traded companies, affects the supplier s disclosure quality. Our paper is more closely related to a stream of recent working papers that focus on the effect of customer base composition on suppliers operating performance, cash holdings, information environment, and loan contract, etc. For example, Cohen and Malloy (2016) show that government-linked firms invest less in physically and intellectual capital and have lower future sales growth. Cohen and Li (2016a) provide evidence that relative to corporate customers, government customers help suppliers in achieving higher efficiencies in asset turnover and customer-specific SG&A investments, which results in lower operational uncertainty. Cohen and Li (2016b) find that firms that have the U.S. government as a major customer holder smaller amount of cash and have less volatile future earnings. Cohen (2016) find strong evidence that, when the borrowing firm reports a major government customer, a loan contract contains significantly fewer covenants. In contrast, they do not find such an effect for a firm that has a major corporate customer. Our paper extends the line of research by highlighting the impact of having government as major customer on supplier firms voluntary disclosure quality Management earnings forecasts We focus on the effect of different customer types on supplier s MEFs because MEFs are important voluntary disclosures through which managers communicate with market participants. In a review paper, Beyer et al. (2010) show that, over the period between 1994 and 2007, MEFs provide, on average, almost 55% of accounting-based information that is incorporated into stock price annually. Numerous studies on MEFs have been done in the past several decades (Healy and 8

9 Palepu, 2001; Hirst et al., 2008; Leuz and Wysocki, 2016). One stream of studies investigate managers incentives of issuing MEFs such as the reduction in information asymmetry (e.g., Verrecchia, 2001), reduction in litigation costs (e.g., Skinner, 1994), and compensation-related incentives (e.g., Aboody and Kasznik, 2000). Chen et al. (2011) provide evidence that once a firm starts to issue an earnings forecast, it is costly to stop the issuance of MEFs. Some other studies of MEFs examine its consequences on stock price reaction, market liquidity, information asymmetry among investors, cost of capital, litigation risk, analyst and investor behavior, and reputation (Hirst et al., 2008). Most of these consequences are directly or indirectly related to capital market responses to management earnings forecasts. Only a few studies examine whether and how MEFs change managers behavior. Kasznik (1999) find that managers use discretionary accruals to revise earnings upward to meet their own forecasts. Motivated by recent practitioners concerns about managerial short-termism, Call et al. (2014) investigate and show that the issuance of short-term quarterly earnings guidance is associated with less, rather than more, earnings management. Despite the vast literature on antecedents and consequences of MEFs, relatively fewer studies examine the characteristics of MEFs. As pointed out by Hirst et al. (2008), Forecast characteristics appear to be the least understood component of earnings forecasts both in terms of theory and empirical research even though it is the component over which managers have the most control. In addition, few studies examine MEFs in a supply chain setting. An exception is Chen and Wang (2015), in which the authors show that the likelihood of a supplier issuing management forecasts and the subsequent stock price reaction are related to the news content of customers disclosure. Our study adds to the existing literature by showing that suppliers firms with different customer base composition, i.e., having government versus corporations as major customer, exhibit different characteristics of MEFs. 2.2 Hypothesis development We consider a unique setting where we can identify major customers of different types and develop two competing but not mutually exclusive hypotheses: the information spillover hypothesis versus the government effect hypothesis. Publicly listed companies are required by security 9

10 regulators and stock exchanges to make periodic disclosures to the public, e.g., quarterly and annual financial reports. In contrast, government entities are not subject to such disclosure requirements and have much lesser concern over asymmetric information with business partners (Chaney et al., 2011). If customers' disclosures affect suppliers MEFs, we expect that corporate suppliers differ from government suppliers in the characteristics of their management forecasts. On one hand, if a supplier s major customers are pubic companies, market participants can benefit more from the information spillover effect. This happens because customers public disclosures constitute a valuable source of information about its supplier. Since a supplier usually obtains a large percentage of sales from its major customers, market participants can learn useful information about the supplier s business performance and future prospects from customers disclosures and update their assessment of the supplier s value accordingly. When investors know more about the supplier, its managers have incentives to improve forecast precision and accuracy because inadequate disclosures potentially increase litigation risk (e.g., Kasznik, 1999; Skinner, 1994). On the other hand, if a supplier s major customers are exclusively government entities who are not subject to financial reporting regulations and therefore do not disclose information to the public regularly, market participants will benefit much less on the information spillover effect between customers and suppliers. Following this line of argument, we predict that corporate suppliers tend to have high quality MEFs than government suppliers. We examine three dimension of MEF quality precision, accuracy, and price impact. We state our first hypothesis the information spillover hypothesis as follows: Hypothesis 1 (Information Spillover Hypothesis): Management forecasts of corporate suppliers are more precise, more accurate, but have a smaller price impact than those of government suppliers. Although Hypothesis 1 suggests that having the government as a sole major customer can negatively affect forecast characteristics, current literature provides counter evidence that it can be beneficial for suppliers to sell to government entities. Government as a customer is unique in many 10

11 ways. The government's objectives are different from corporations, and thus its decisions are often affected by non-business considerations such as social welfare, environmental awareness, political lobbying and campaign donations, etc. (Goldman et al., 2013). Government agencies follow a standard procurement process regulated by the Federal Acquisition Regulation (FAR), and have much lesser concern over asymmetric information with business partners (Chaney et al., 2011). Taking these differences into account, Cohen and Li (2016a) provide evidence that relative to corporate customers, government customers help suppliers in achieving higher efficiencies in asset turnover and customer-specific SG&A investments, which results in lower operational uncertainty. Cohen and Li (2016b) find that firms that have the U.S. government as a major customer hold smaller amount of cash and have less volatile future earnings. Cohen et al. (2016) find strong evidence that, when the borrowing firm reports a major government customer, a loan contract contains significantly fewer covenants. In contrast, they do not find such an effect for a firm that has a major corporate customer. This suggests that lenders benefit from having government as a reliable and stable customer. Collectively, this line of studies lead us to believe that if a supplier s major customer is the government instead of commercial customers, the supplier will have lower operational uncertainty and managers will be better able to issue earnings forecasts that are more precise and accurate. Thus, we state our second hypothesis the government effect hypothesis as follows. Hypothesis 2 (Government Effect Hypothesis): Management forecasts of government suppliers are more precise, more accurate, but have a smaller price impact than those of corporate suppliers. Hypotheses 1 and 2 are not mutually exclusive. It is an empirical question whether information spillover effect or government effect dominates. Below, we discuss further our predictions on the three dimension of MEFs that we focus on. On the precision: The information spillover effect suggests that corporate customers disclose information through which investors can learn more about the suppliers current and future performance. To the extent that capital market can play a more efficient price discovery role when there is more information available in the public domain (e.g., Shroff et al., 2017), one would expect corporate suppliers management forecasts to be more precise than those issued by 11

12 government suppliers. On the other hand, having the government as the sole major customer helps to generate more stable income streams for suppliers, hence one would expect government suppliers management forecasts to be more precise than those issued by corporate suppliers. On the accuracy: It is reasonable to argue that relative to forecast precision, managers have less control over the accuracy of their forecast. After a manager releases an earnings forecast, unexpected events may cause the actual earnings to deviate from the value that the manager predicts. Since government suppliers have less operational uncertainty, their MEFs will be more accurate than those issued by corporate suppliers. On the price impact: It is intuitive to argue that among the three characteristics, managers have the least control on the price impact of their forecasts. Relative to government suppliers, we expect corporate suppliers to experience a smaller price impact because there is more information available about the firm (via the spillover effect of its customer s disclosure) that will be incorporated into the price. Earnings forecasts issued by suppliers management may not appear as a big surprise to investors if they already acquire value-relevant information elsewhere. 3. Data and variable definition We collect supplier-customer relationship data from the Compustat Segment database. The Financial Accounting Standards Board s (FASB) new segment reporting standard was issued in June 1997 and became effective for financial statements for periods beginning after December 15, 1997 (e.g., Berger and Hann, 2003). Hence, we begin our sample period from the fiscal year ending in The Compustat Segment database classifies each customer under seven categories: GOVDOM, GOVSTATE, GOVLOC, COMPANY, GEOREG, MARKET, and GOVFRN. We classify a supplier as corporate supplier if the firm reports only sales to major customers in the COMPANY category in three years consecutively. We classify a supplier as government supplier if the firm reports only sales to government customers in the GOVDOM, GOVSTATE and GOVLOC categories in three years consecutively. Thus, our sample includes only suppliers that do not switch from one type to another. For example, if a supplier reports any sales to a non- 12

13 government customer in any given year, it drops out of our sample of government suppliers in that year and onward. The three-year restriction ensures that the customer type of a supplier is stable over an extended period and thus our empirical analysis is not contaminated by suppliers switching from one type to another. We obtain management forecasts of quarterly earnings from the IBES Earnings Guidance database. For a government supplier whose major customers are exclusively government entities in the three consecutive fiscal years, t-2, t-1 and t, we collect the quarterly earnings forecasts that the supplier released within the fiscal years t and t+1. For a corporate supplier whose major customers are exclusively public companies in the three consecutive fiscal years, t-2, t-1 and t, we collect the quarterly earnings forecasts that the supplier released within the fiscal years t and t+1. If a supplier issues multiple forecasts of the same quarter s earnings, we keep the last forecast before the relevant earnings announcement. For each management earnings forecast (MEF), we calculate its precision, accuracy, and price impact. We follow prior studies to measure forecast precision as the negative of the difference between the upper and the lower end of a range forecast, scaled by the beginning-of-quarter stock price, or zero for a point forecast. We measure forecast accuracy by the absolute forecast error, that is, the absolute value of the difference between actual earnings and the forecast value, scaled by the beginning-of-quarter stock price. For a range forecast, the forecast value is equal to its midpoint. We measure the price impact by the market adjusted cumulative abnormal returns in the four-day [-2, 1] window around the forecast release day (i.e., day 0). In addition, we measure the information content in a management forecast by the difference between the forecast value and the median of analysts earnings forecasts as of the day prior to the release of the management forecast. We also measure several firm characteristics, such as firm size (SIZE), book-to-market ratio (BMR), return on assets (ROA), and stock return volatility (RETVOL). SIZE is equal to the logarithm of market capitalization at the beginning of current quarter. BMR is the book-to-market-equity ratio at the beginning of current quarter. ROA is the return on assets for the previous fiscal year. RETVOL is equal to the standard deviation of daily stock returns in the past 12 months before current quarter. 13

14 We obtain financial statement data from Compustat, stock prices and returns from CRSP, and analyst forecasts from the IBES Detailed Analyst Forecast database. Our sample includes management forecasts that have non-missing values for all of the above variables. The final sample includes 764 management earnings forecasts issued by 133 corporate suppliers and 404 management earnings forecasts issued by 75 government suppliers. The Appendix lists the 208 suppliers company names and their industry distribution. We follow the Fama-French 48 industry classification. The 75 government suppliers fall into 20 industries, while the 133 corporate suppliers are from 26 industries. The Business Services industry has the largest number of government suppliers, whereas the Electronic Equipment industry has the largest number of corporate suppliers. We winsorize each variable at the 1% and 99% percentiles of the whole sample of 1,168 forecasts. 4. Empirical results 4.1. Univariate comparison between corporate suppliers and government suppliers We expect that the information spillover effect is stronger on corporate suppliers management forecasts while the government effect is stronger on government suppliers management forecasts. By examining the differences in the characteristics of management forecasts between corporate suppliers and government suppliers, we expect to find the empirical evidence for or against these two hypotheses. In this section, we compare the two groups of suppliers with regards to three forecast characteristics, namely, forecast precision, forecast accuracy, and price impact. As discussed earlier, managers have different level of control over the three characteristics, with the strongest control over forecast precision, followed by forecast accuracy, and then price impact. Panel A of Table 1 reports the mean, median, and standard deviation of variables that capture three aspects for corporate suppliers and government suppliers, separately. We observe several significant differences in management earnings forecast properties, forecast surprises, as well as firm characteristics between the two groups. The first set of differences exist in the forecast properties. Specifically, government suppliers issue more precise forecasts than corporate suppliers do. The mean (median) forecast precision of government suppliers management 14

15 earnings forecasts (MEFs) is ( ), whereas the mean (median) forecast precision of corporate suppliers MEFs is only ( ). Statistical tests show that both mean and median are significantly different between the two groups of suppliers. 3 The evidence is consistent with the government effect hypothesis, which predicts that government suppliers have less operational uncertainty and hence managers are able to provide more precise forecasts. [Table 1 is about here.] Moreover, government suppliers issue more accurate forecasts than corporate suppliers do. The mean (median) forecast accuracy of government suppliers MEFs is (0.013), whereas the mean (median) forecast accuracy of corporate suppliers MEFs is (0.011). Both mean and median are significantly different between the two groups of suppliers. Again, the evidence is supportive of the government effect hypothesis. However, the difference in forecast accuracy between the two groups is much smaller than the difference in forecast precision. This is likely due to the stronger influence managers have on forecast precision than on forecast accuracy. Finally, there is no significant difference in the price impact of management forecasts between corporate suppliers and government suppliers. The mean (median) cumulative abnormal return is -0.8% (0.1%) in the four-day window around the release of government suppliers MEFs, whereas the mean (median) cumulative abnormal return is -1.3% (-0.7%) around the release of corporate suppliers MEFs. The abnormal returns have a very large standard deviation, 10.5% for government suppliers and 10.9% for corporate suppliers. The second difference between the two groups exists in the information content of the MEFs, measured by the difference between the forecast value and the median analysts forecast, which we label the MEF surprise. The mean (median) surprise in corporate suppliers MEFs is $ ($-0.013), whereas the mean (median) surprise in government suppliers MEFs is $ ($-0.005). The negative surprise is consistent with the well-documented phenomenon that 3 The F-test shows that the two groups have significantly different standard deviations in forecast precision. We use the t-statistic with the Satterthwaite adjustment for unequal variance to test whether the two groups have the same mean. We use the Wilcoxon statistics with normal approximation to test whether the two groups have the same median. 15

16 managers tend to issue warnings if their earnings fall below expectations (e.g., Kasznik and Lev, 1995). Relative to corporate suppliers, government suppliers have a statistically significantly smaller shortfall from forecasted earnings, which is consistent with a lower level operational uncertainty in government suppliers. The third set of differences lie in other firm characteristics such as firm size, profitability, bookto-market ratio, and return volatility. Government suppliers are more profitable than corporate suppliers. The mean return on assets (ROA) is -0.5% for corporate suppliers, but is positive at 4.2% for government suppliers. Corporate suppliers have a much larger standard deviation in profitability than government suppliers, with the median ROA of 3.7% and 4.9%, respectively. Both the book-to-market ratio (BMR) and stock return volatility are significantly higher for corporate suppliers than for government suppliers. The mean daily return volatility (RETVOL) is 3.3% for corporate suppliers, but 3.0% for government suppliers. However, there is no significant difference in firm size between the two groups of suppliers. The average market capitalization of both groups of suppliers is about $950 million. Panel B of Table 1 reports the correlation coefficients of these variables for the whole sample. The correlation coefficients show some patterns. First, forecast precision and forecast accuracy have a large negative correlation coefficient of , which suggests that more precise forecasts tend to have higher level of accuracy. Second, forecast precision has relatively large positive correlation with ROA and SIZE, but large negative correlation with BMR and RETVOL. The evidence suggests that larger and more profitable suppliers issue more precise forecasts, while suppliers with higher book-to-market ratio and higher return volatility have less precise forecasts. Third, forecast accuracy has relatively large negative correlation with ROA and SIZE, but large positive correlation with BMR and RETVOL. The evidence suggests that larger and more profitable suppliers have more accurate forecasts, while suppliers with higher book-to-market ratio and higher return volatility issue less accurate forecasts. Moreover, there is a large positive correlation between price impact and the MEF surprise. A management forecast of a large surprise causes price to move in the same direction by a large amount. 16

17 Management issue earnings forecast to communicate with equity market participants. They have the tendency to issue warnings if their earnings fall below market expectations. The forecasts that carry positive news are likely to have different characteristics than negative forecasts. We group MEFs into three subsamples according to the MEF surprise: the MEFs with a surprise less than or equal to $-0.01, the MEFs with a surprise greater than $-0.01 and less than $0.01, and the MEFs with a surprise greater than or equal to $0.01. The MEF surprise is measured as the difference between the forecast value and the median of analyst forecasts as of the day prior to the MEF release day. IBES provides a classification code to indicate the information content in a management forecast. The code takes four values: 1 for the MEF missing analyst consensus forecast, 2 for the MEF beating analyst consensus forecast, 3 for the MEF matching analyst consensus forecast, and 6 for unclassified MEFs. However, the IBES code does not provide a clear indication of information content. Table 2 shows that our classification based on the MEF surprise has an advantage over the IBES codes. Panel A in Table 2 reports the number of MEFs that are classified in subsamples according to both the IBES code and the MEF surprise. While IBES code 1 is highly consistent in capturing MEFs that miss analyst consensus forecast, Codes 3 and 6 do a poor job in capturing the information content in management forecasts. Specifically, out of the 160 MEFs that have the IBES code equal to 6, we find that 64 have negative MEF surprise, 51 match the median analyst forecast, and 45 have positive MEF surprise. Similarly, out of the 445 MEFs with the IBES code 3, 121 have negative surprise, 101 have positive surprise, and only 223 match the median analyst forecast. The number of MEFs with negative surprise doubles that of MEFs with positive surprise. Based on the above observation, we believe that compared to IBES code classification, our classification of the MEF surprise provides a finer and more accurate proxy for the information content of a management forecast. [Table 2 is about here.] 17

18 To gain further insight, we report the mean and median surprises of MEFs in each subsample in Panel B of Table 2. For the 1,168 MEFs, the mean (median) surprise is $ ( ). For the 588 MEFs with negative surprise, the mean (median) surprise is $ ( ). For the 295 MEFs with positive surprise, the mean (median) surprise is $ (0.0300). For the 285 MEFs that match the consensus analyst forecast, the mean (median) surprise is $ (0). Taken together, subsample descriptive statistics in Table 2 reveal that the MEF surprise does a better job than the IBES code in separating MEFs into groups according to the information content. In the following analysis, we study the three subsamples grouped by the MEF surprise. Within each subsample, we compare corporate suppliers MEFs with government suppliers MEFs. The three panels in Table 3 report comparison results for the three subsamples, separately. [Table 3 is about here.] In Panel A of Table 3, we examine the differences in three MEF characteristics between corporate suppliers and government suppliers for the subsample of negative MEF surprise. The two groups of suppliers have similar firm characteristics. They are only marginally different in the median of profitability (ROA), book-to-market ratio, and daily return volatility. The average firm size of corporate suppliers is $961 million while the average firm size of government suppliers is $706 million. Turning to forecast properties, government suppliers have significantly more precise and more accurate forecasts than corporate suppliers. The magnitude of negative surprise is much greater for corporate suppliers than that for government suppliers. The mean (median) surprise of corporate suppliers forecasts is $ ($-0.065), whereas the mean (median) surprise of government suppliers forecasts is $ ($-0.043). However, there is no significant difference in the price impact between the two groups of suppliers. In Panel B of Table 3, we examine the differences between corporate suppliers and government suppliers for the subsample of MEFs that match the median analysts forecast. The two groups of suppliers appear to differ in several aspects. Government suppliers are more profitable, have a smaller book-to-market ratio, larger firm size and lower stock return volatility than corporate suppliers. Again, they have significantly more precise forecasts than corporate suppliers do. 18

19 However, there is no significant difference in forecast accuracy between the two groups. Both corporate suppliers and government suppliers have, on average, positive abnormal return in the four-day window. Government suppliers have marginally larger price impact than corporate suppliers, despite the fact that there is no significant difference in the MEF surprise between the two groups. In Panel C of Table 3, we examine the differences between corporate suppliers and government suppliers for the subsample of positive news MEFs. Government suppliers are significantly more profitable and have a significantly smaller book-to-market ratio than corporate suppliers. The two groups are similar in firm size and stock return volatility. Government suppliers have significantly more precise and more accurate forecasts than corporate suppliers. However, there is no significant difference in the price impact, nor in the MEF surprise, between the two groups. In summary, the univariate comparison shows that government suppliers issue more precise and more accurate management forecasts than corporate suppliers. The difference in the price impact between the two groups of suppliers varies across the subsamples of MEFs Regression analysis of forecast characteristics In the following analysis, we test our hypotheses in a regression framework that controls for the differences in firm characteristics among suppliers. In view of the differences between the subsamples of MEFs in Table 3, we carry out the regression analysis for the three subsamples of MEFs separately. First, we estimate the following regression model of forecast precision Precision = α + β GOV + θ X + ε, (1) where the dependent variable, forecast precision, is equal to the negative of the difference between the upper and the lower end of a range forecast, scaled by beginning-of-quarter stock price, or zero for a point forecast. The independent variable GOV is an indicator variable that equals one if the forecast is issued by a government supplier, and X represents the vector of control variables that includes firm size (SIZE), book-to-market ratio (BMR), return on assets (ROA), and stock return volatility (RETVOL). Table 4 reports the results of estimating the regression for the three subsamples of MEFs, separately. All of the coefficients in Table 4 are multiplied by a factor of one thousand. 19

20 [Table 4 is about here.] The coefficient of the variable GOV is positive and statistically significant in all of the three subsamples. After taking into consideration factors that may influence forecast precision, we find that government suppliers issue significantly more precise forecasts than corporate suppliers. The evidence is consistent with the government effect hypothesis, which suggests that having government as the exclusive customers reduce operational uncertainty and hence enable government suppliers to issue more precise forecasts. Turning to control variables, both ROA and SIZE have significantly positive coefficients in the three subsamples, suggesting that profitable and large suppliers tend to have higher forecast precision. Although the coefficients of BMR and RETVOL have the negative sign in the three subsamples, the coefficient of BMR is statistically significant only in the negative surprise subsample, while the coefficient of RETVOL is statistically significant only in the positive surprise subsample. Next, we estimate the following regression model of forecast accuracy Accuracy = α + β GOV + θ X + ε, (2) where the dependent variable, forecast accuracy, is equal to the absolute value of the difference between actual earnings and management forecast, scaled by beginning-of-quarter stock price. The independent variable GOV is an indicator variable that equals one if the forecast is issued by a government supplier, and X represents the vector of control variables that includes firm size (SIZE), book-to-market ratio (BMR), return on assets (ROA), and stock return volatility (RETVOL). Table 5 reports the results of estimating the regression for the three subsamples of MEFs, separately. [Table 5 is about here.] The coefficient of the variable GOV is negative in all of the three subsamples, statistically significant in both the negative surprise subsample and the positive surprise subsample, but 20

21 insignificant in the match subsample. After taking into consideration factors that may influence forecast accuracy, we find that government suppliers issue significantly more accurate forecasts than corporate suppliers when the forecasts convey new material information to the public. The evidence is consistent with the government effect hypothesis, which suggests that government suppliers have a lower level of operational uncertainty and are able to issue more accurate forecasts. In the match sample, when suppliers forecasts do not carry new information, there is no significant difference in forecast accuracy between the two groups. Turning to control variables, BMR has significantly positive coefficients in the three subsamples, suggesting that suppliers with higher book-to-market ratio tend to have lower forecast accuracy. The coefficient of ROA is negative and statistically significant only in the negative surprise subsample. SIZE is marginally significant in both negative and positive surprise subsamples while RETVOL is only marginally significant in the positive surprise subsample. At last, we estimate the following regression of the price impact Impact = α + β Surprise + γ GOV + η GOV*Surprise + θ X + ε, (3) where the dependent variable is the price impact of a supplier's forecast, that is, the market adjusted cumulative abnormal return in the four-day [-2, 1] window around the forecast's release day (i.e., day 0). The independent variable GOV is an indicator variable defined the same as before. The other independent variable, Surprise, is equal to the difference between the forecast value and the median of analyst forecasts as of the day prior to the MEF release day. We also include the interaction term, GOV*Surprise, in the regression of the price impact because the results in Table 3 suggest that the price impact is related to the information content in MEFs. Moreover, we want to include the interaction term to test whether the price response to the information content in MEFs differs between government suppliers and corporate suppliers. As before, the vector of control variables X includes firm size (SIZE), book-to-market ratio (BMR), return on assets (ROA), and stock return volatility (RETVOL). Table 6 reports the results of estimating the regression for the three subsamples of MEFs, separately. [Table 6 is about here.] 21