The current study investigates how corporate governance ratings affect the certainty of buy-side analysts' earnings forecasts. Nineteen financial analysts from the United States (U.S.) and 17 from the United Kingdom (U.K.) participated in a 1 × 2 (corporate governance ratings: below or above industry average) within-participant experiment. We find that, on average, analysts exhibit more certainty in their range forecasts when the corporate governance rating is above average, relative to below average. We also observe a significant interaction between the corporate governance ratings and country, indicating that U.K. analysts exhibit stronger responses to a below average rating than U.S. analysts, while responses to an above average rating are not significantly different between the two countries. These results suggest a need to investigate cultural or other factors that can impede the seamless integration of national capital markets into a unified global financial network.
Data Availability: Available upon request.
Corporate governance has been long recognized as an important component of the relationship between investors and businesses (Jensen and Meckling 1976). This recognition became amplified in the early 2000s due to the effects of spectacular business failures, such as the demise of Enron and its accounting firm, Arthur Andersen (Gompers et al. 2003; Ashbaugh-Skaife et al. 2006). As a result, several third-party providers of corporate governance ratings have emerged in recent years (Knowledge@Emory 2003). While there is archival evidence indicating that corporate governance ratings are informative to investors (e.g., Gompers et al. 2003; Ashbaugh-Skaife et al. 2006), the multi-dimensional nature of the corporate governance construct and recent mixed findings in corporate governance research (e.g., Larcker et al. 2007) suggest that financial analysts may not find corporate governance ratings particularly useful. The current study examines the extent to which financial analysts impound corporate governance ratings into their earnings forecasts from an experimental perspective, as the precision, control, and randomization offered by experiments allows us to understand whether analysts indeed rely on corporate governance ratings, unconfounded by various information substitutes and complements, and unaffected by the multitude of correlated market factors present in archival research.
Analysts tend to specialize as buy-side and sell-side. Buy-side analysts reflect key sources of private intelligence to institutional investors (Gillian and Starks 2003), while sell-side analysts produce public reports in which important information about companies is summarized as a basis for stock recommendations such as strong buy, buy, hold, sell (Bradshaw 2004; Cowen et al. 2006). Due to participant availability, we use only buy-side analysts in our study and assume that sell-side analysts have incentives similar to buy-side analysts to use corporate governance information in their earnings forecast.
We use a 1 × 2 within-participant experiment, wherein we randomized corporate governance ratings as above or below industry average. Nineteen buy-side analysts from the United States (U.S.) and 17 from the United Kingdom (U.K.) participated in the experiment. The analysts first read background and financial information related to a U.S. case company, and then predicted future earnings in the form of a range forecast. Afterward, participants learned about the company's corporate governance ratings (above or below industry average) and were offered an opportunity to revise their initial forecasts. Overall, we hypothesize and find supporting evidence that pre-to-post treatment earnings forecast ranges narrow (the analysts become more certain in their forecasts) in the above average corporate governance ratings condition and widen (the analysts become less certain in their forecasts) in the below average condition.
We also investigate whether analysts from the U.S. and the U.K. respond differently to corporate governance ratings. Since prior research indicates that the U.S. and the U.K. have relatively similar legal and regulatory origins (e.g., Shleifer and Vishny 1997), one might expect no differences in their reactions to corporate governance ratings. However, a recent study by the Institute of Chartered Accountants of England and Wales (ICAEW 2007) questions the notion that the U.S. and the U.K. have similar legal and regulatory regimes; hence it is possible that analysts from the two countries will exhibit differential responses to corporate governance ratings. Our findings suggest that the response of U.K. analysts to a below average corporate governance rating is significantly greater than U.S. analysts, while the response to an above average corporate governance rating is not significantly different between the two countries. We also find that the U.K. analysts' a priori presumptive beliefs about the general state of corporate governance quality in U.S.-based firms is significantly higher than that of U.S. analysts' beliefs; hence, when the analysts learn about a below average corporate governance rating, the negative news holds more surprise for the U.K. analysts than the U.S. analysts, triggering a greater reaction.
This study complements and extends the corporate governance (e.g., Agrawal and Chadha 2005; Ashbaugh-Skaife et al. 2006) and analyst decision-making literatures (e.g., Bouwman et al. 1987; Hunton and McEwen 1999) in several ways. First, the current study offers complementary experimental evidence to literature streams that are dominated by archival investigations. Second, we find that analysts impound third-party corporate governance ratings into their earnings forecasts in a manner consistent with the direction of the ratings news; specifically, we find that above (below) average corporate governance ratings are associated with increased (decreased) forecast certainty. Finally, our study adds an international dimension to understanding financial analysts' earnings forecasts, as the results indicate that U.S. and U.K. analysts hold different a priori presumptive beliefs about the quality of corporate governance in U.S.-based companies. These country differences raise questions about the reliability and volatility of analysts' earnings forecasts in different countries where investors have access to the same information. Normatively, analysts' country of domicile should not affect how financial and nonfinancial information is impounded into earnings forecasts, yet the current study indeed highlights disparate effects. Researchers should further investigate the root cause of the effect of such differences in an attempt to obtain more consistency across global markets.
The next section provides the background literature, presents a study hypothesis, and proposes a research question. The subsequent two sections describe the research method and offer the research findings. The final section summarizes the study findings and discusses the implications.
BACKGROUND, HYPOTHESIS, AND RESEARCH QUESTION
Corporate governance reflects a collection of mechanisms put in place by a company to assure investors of proper stewardship of organizational resources. High-quality corporate governance is expected to help mitigate the agency problem by protecting investors against the possibility of management expropriating their funds (Shleifer and Vishny 1997).
Effects of Corporate Governance on Financial Reporting Quality and Firm Values
A large body of literature has established that poor corporate governance can be associated with myriad problems, such as financial restatements. Abbott et al. (2004) report a significant association between audit committee independence and financial restatements, and document that audit committees composed of at least one director with financial expertise are inversely related to incidences of financial restatement. Agrawal and Chadha (2005) suggest that financial restatement is less likely in companies that have an independent director with a background in accounting or finance on their audit committees.
A related line of corporate governance research finds an association between corporate governance and earnings quality. For instance, Bedard and Johnstone (2004) study the association among earnings manipulation risk, corporate governance, and audit effort, and find a negative relationship between the quality of corporate governance and earnings manipulation risk. Leuz et al. (2003) document an international aspect to the importance of corporate governance to investors, by examining the level of investor protections across 31 countries. They find that corporate governance and earnings quality are endogenously determined such that increases in earnings quality arise from stronger legal protections for investors. Similarly, Hermalin and Weisbach (1991) examine the association between board composition and firm ownership structure and firm performance. They find that Tobin's Q and before-tax profits increase when board ownership is below one percent, and decrease after a 20 percent threshold. Similarly, Cremers and Nair (2005) find that the market rewards good governance, as measured by the absence of poison pills with premiums between 10 percent and 15 percent, when public pension funds hold high-equity stakes.
Other studies investigate the effects of corporate governance on fraudulent financial reporting (Beasley 1996; Farber 2005), profitability, and firm valuation (Gompers et al. 2003; Cremers and Nair 2005). For example, Farber (2005) studies firms identified by the SEC as having been involved in fraudulent financial reporting and finds that in the year prior to the fraudulent reporting year, such firms tended to have weaker corporate governance; that is, fraudulent reporting firms had boards of directors with smaller percentages of outsiders, were less diligent (held fewer meetings), and were more likely to have a CEO who was also chairman of the board. Fraudulent firms were also less likely to be audited by one of the Big 4 audit firms. Further, when the fraudulent firms that Farber (2005) studied improved their governance to the level of control firms, three years after the fraud year, their stock prices improved after controlling for earnings performance. Similarly, Gompers et al. (2003) find that companies with stronger shareholder rights are more valuable, more profitable, generate stronger sales growth, and make fewer corporate acquisitions than firms with weaker shareholder rights.
Overall, the corporate governance literature suggests that corporate governance quality has a significant impact on a firm's market value, and thus should be of interest to investors. However, it is difficult and costly for investors to engage in assessing the quality of corporate governance in most firms (Duffy 2004). To fill this void in the marketplace, a number of third-party suppliers, including GovernanceMetrics International (GMI), Institutional Shareholder Services (ISS) (now RiskMetrics), Financial Times Stock Exchange ISS (FTSE-ISS), and Standard and Poor's (S&P), have begun providing corporate governance quality ratings in recent years (Knowledge@Emory 2003). While the corporate governance attributes incorporated into the rankings are fairly easy to identify (e.g., separation of CEO and chairman positions, ratio of independent to total directors), it is not always clear how the rating agencies combine these attributes to arrive at a final corporate governance rating, as each vendor argues that its method is proprietary.
Using principal components analysis to examine 14 dimensions of corporate governance, Larcker et al. (2007) argue that corporate governance constructs have little relation to accounting restatements and a mixed relation to earnings quality, measured by abnormal accruals. However, the authors find that the corporate governance constructs have some ability to explain future operating performance. Given the mixed findings from Larcker et al. (2007) on the usefulness of corporate governance for investors, it is not clear how professional financial analysts, arguably one of the most influential groups of infomediaries in the global capital market, will respond to corporate governance ratings offered by rating agencies.
Analysts' Judgments and Decisions
An extensive literature has developed on the decision-making processes of financial analysts, including how they use pro forma financial information (e.g., Frederickson and Miller 2004), impound accounting policy disclosures (e.g., Hope 2003), and strategically search for information (e.g., Bouwman et al. 1987; Hunton and McEwen 1999). Generally, the literature suggests that analysts use a diverse set of inputs, processes, and tools in making buy, sell, and hold decisions. For example, Frederickson and Miller (2004) find that analysts' stock valuations are not affected by pro forma financial information, whereas nonprofessional investors' valuations are influenced by this information. They attribute the difference between the two groups to analysts using well-defined valuation models based on earnings multiples and cash flows, while nonprofessional investors tend to rely on relatively simplistic decision heuristics. Hope (2003) examined cross-country disclosure of accounting policies in annual reports. He found that the forecast accuracy of analysts is positively associated with the quantity of disclosures and the level of enforcement in a country. His findings indicate that more disclosure of accounting policies in a country provides analysts with more information, helping reduce analysts' uncertainty, thereby increasing forecast accuracy in that country.
The managerial accounting literature suggests that firms that incorporate both nonfinancial and financial measures tend to outperform those using only financial measures (e.g., Said et al. 2003). The rationale for the superior performance is that the additional nonfinancial information provides a more comprehensive and valuable information set that improves decision making. In a more direct example, Vanstraelen et al. (2003) study the association between nonfinancial disclosures and the dispersion in analysts' earnings forecasts in three countries (Belgium, Germany, and the Netherlands), and find that the dispersion decreases in cases of higher levels of forward-looking nonfinancial information. In the context of analysts' decision making, we propose that corporate governance ratings hold the potential to provide valuable incremental nonfinancial information that enhances the quality of the analyst's information set.
Influence of Corporate Governance Ratings on Analysts' Earnings Forecasts
Whether corporate governance ratings are self-determined or third-party provided, corporate governance quality indicators hold the potential to provide valuable incremental information to market participants (DeFond 1992). As indicated by Willenborg (1999), in settings where there is significant information asymmetry, as is often the case with corporate governance quality, investors will often demand and rely on incremental information from a reliable independent third party. However, Larcker et al.'s (2007) mixed results regarding the usefulness of corporate governance ratings suggests that analysts may not find corporate governance ratings to be useful.
Two counter-arguments can potentially refute this suggestion. First, one facet of an analyst's responsibility is to corroborate information provided by corporate managers. While analysts typically have the skill set required to corroborate financial information, they may not be as well versed in evaluating corporate governance and other related nonfinancial information. Second, even when analysts evaluate corporate governance information, the numerous dimensions of corporate governance (e.g., compensation, shareholder rights, board independence, board diligence and skills, auditor quality) that have to be considered could mean that it is likely more efficient for analysts to use third-party corporate governance ratings, relative to deriving such quality assessments on their own. Examples of third-party vendors of corporate governance ratings include GovernanceMetrics International (GMI), Institutional Shareholder Services (ISS, now RiskMetrics), Financial Times Stock Exchange ISS (FTSE-ISS), and Standard & Poor's (S&P).1
Based on their review of the literature on the association between corporate governance and financial reporting quality, Cohen et al. (2004) suggest a positive association between corporate governance quality and the quality of financial reporting. However, Cohen et al.'s (2004) review does not include any literature on whether corporate governance ratings impact financial reporting quality. The primary corporate governance mechanisms examined in many of the studies reviewed include characteristics of the boards of directors, audit committee, the type of auditor, and the relationships among the audit committee, internal auditors, external auditors, the larger board of directors, and management.2
In summary, the literature generally suggests that positive attributes of important dimensions of corporate governance (e.g., board independence and audit committee effectiveness) are associated with higher reporting quality. This finding suggests that a positive composite rating of corporate governance quality should similarly be associated with higher reporting quality; furthermore, Cohen et al. (2004) argue that the predictability of earnings should be greater as well. Greater predictability, in turn, will result in higher forecast certainty (Bedard and Johnstone 2004). Hence, we offer the following hypothesis:3
Hypothesis: Analysts will exhibit less (more) certainty in their earnings forecasts when a reliable third-party provider of corporate governance ratings indicates that a firm's corporate governance quality is below (above) average.
Potential Country Differences
There is evidence that corporate governance is differentially important in common law (e.g., the U.S. and U.K.) and civil law (e.g., France) countries (Gillian and Starks 2003). The argument is that common law countries generally offer greater investor protection through specific rules and regulations (e.g., LaPorta et al. 1997, 1999, 2006); hence, investors in common law countries might rely less on corporate governance ratings than investors in civil law countries. LaPorta et al. (2006) document that stronger investor protection, achieved through higher levels of disclosure and private enforcement liability rules, are associated with better-developed securities markets. They find that countries with English legal origins or common law-based regimes offer stronger investor protections and have better-developed securities markets than countries with French legal origins or civil law-based regimes.
This evidence suggests that analysts from common law countries, such as the U.K. and the U.S., should not respond differentially to corporate governance ratings. However, other research indicates that even within common law countries, differences in regulatory regimes could impact analysts' judgments. For example, LaPorta et al. (1997) document that the U.S. has stronger investor protection (as measured by anti-director rights and rule of law) than the U.K. Among measures of investor protections are disclosure requirements, liability standard, and the extent to which criminal sanctions are available against the parties issuing securities. Bhattacharya et al. (2003) construct measures of earnings aggressiveness, loss avoidance, and earnings smoothing, as well as a composite measure of earnings opacity from the first three measures, and use these variables to evaluate earnings quality across 34 countries. Their descriptive statistics reveal that common law countries, such as the U.S. and the U.K., tend to have less earnings opacity than civil law countries. However, when comparing these two countries, the U.S. places in the first quartile on all measures (lower quartiles are less opaque), while the U.K. falls in the third quartile on earnings aggressiveness and the second quartile on the other metrics.
The stricter set of disclosure regulations, such as the Securities Act of 1933 and 1934, and more recently the Sarbanes-Oxley Act of 2002, tends to reduce information asymmetry between shareholders and management, as manifest in reduced cost of capital for non-firms cross-listing in the U.S. (Hail and Leuz 2009). The stricter regulatory and legal setting potentially mitigates demand for other nonfinancial information, such as corporate governance ratings. Correspondingly, it can be costly to provide additional information beyond what is required in the U.S. For example, a firm that voluntarily discloses information beyond what is required has to invest resources to ensure that such information is credible and that it can continue to meet the increased expectations going forward, else face potential litigation. In the extreme, as the potential cost of additional disclosure becomes prohibitively expensive, the U.S. market can approach a phenomenon known as “lemons equilibrium,” where no firm voluntarily discloses any additional information (Core 2001).
On the other hand, the U.K. capital market is relatively less regulated, which in turn can bring about a higher cost of capital. As a result, U.K. firms (and non-U.K. firms listing in the U.K.) likely have incentives to differentiate themselves from their peers, as the U.K. market approaches another phenomenon known as a “pooling equilibrium” (Core 2001; Fischer and Verrecchia 2000). In this circumstance, good news (bad news) signals emanating from voluntary disclosures or a credible third party about a firm's corporate governance quality can be incrementally beneficial (harmful) to the firm's valuation. This could partly explain why Meek et al. (1995) indicate that continental European companies tend to voluntarily disclose more nonfinancial information than their U.K. counterparts, who in turn often voluntarily disclose more nonfinancial information than their U.S. counterparts.
Hope (2003) compares analysts' forecast accuracy from 22 countries and shows that a composite measure of enforcement (composed of such variables as audit expenditures, and a measure of existence and enforcement of insider trading laws) is positively associated with forecast precision. Though the U.K. was closest to the U.S. on the composite enforcement measure (consistent with their similar legal origins), the U.K. score was nevertheless lower than the U.S. score. On a measure of flexibility in choosing accounting principles, the U.K. was found to be more flexible than the U.S.4
Finally, a recent study by the Institute of Chartered Accountants of England and Wales (ICAEW 2007) suggests that U.S. and U.K. analysts may react differently to the same corporate governance rating. The ICAEW study asserts that the view that the U.S. and the U.K. are very similar in investor protection is essentially a “myth” in that the U.S. system of regulations governing investor protection is rules-based, with great power residing with the board of directors. On the other hand, the study concludes that the U.K. system is more principles-based, where relatively more power resides with shareholders. For example, in the context of takeovers, ICAEW (2007) argues that in the U.S., ultimate power resides with the directors in that they can act in the “interest of the company” at the expense of shareholders, whereas in the U.K., ultimate power resides with shareholders. ICAEW (2007) further suggests that a likely reason for this difference is that shareholder engagement with the company in the U.K. is relatively more collegial because ownership is relatively more concentrated in size and geography compared to the U.S.
Taken as a whole, these results suggest that differences in regulatory environments and financial accounting regimes are associated with less earnings forecast precision from U.K., as compared to U.S., analysts. Thus, one might expect that third-party corporate governance ratings will yield more influence on U.K. analysts than U.S. analysts, as information of this nature can reduce market uncertainty to a greater extent in the U.K. than the U.S. However, when U.S. and U.K. analysts are asked to forecast earnings for the same U.S.-based company, it is not clear ex ante that the perceived differences between U.S. and U.K. regulatory and financial accounting environments will become manifest in the respective country's analysts' forecasts. If such differences are found, it would be a testament to how strong an influence the regulatory and financial reporting practices of a country can have on how investors impound reliable, independent corporate governance information into their investment decisions. We are not aware of any studies directly comparing U.S. and U.K. financial analysts' earnings forecasts in the context of corporate governance quality ratings; hence, we present the following exploratory research question:
Research Question: Will analysts from the U.S. and U.K. indicate differential certainty in their earnings forecasts when a reliable third-party provider of corporate governance ratings indicates that a firm's corporate governance quality is below (above) average?
The study's research design involves a 1 × 2 (corporate governance ratings: below or above industry average) within-participants randomized experiment.5 The study was administered on-site at two different mutual fund companies—one in the U.K. and the other in the U.S. The buy-side analysts were attending company-sponsored sessions. In the U.K., the analysts were receiving economic forecast outlooks from the treasury function of the U.K. government. In the U.S., the analysts were attending a training session that focused on leadership development.6 One of the researchers was allowed to administer the current study at the start of each session.
After reading and agreeing to the informed consent form, participants read selected earnings information for the first case company (randomly named ABC Company or XYZ Company).7 The earnings information included consensus analysts' forecasts, management's earnings guidance issued within a couple of weeks of publicly disclosing actual earnings for 2001 through 2006, and actual EPS numbers for 2001 through 2005. Two sets of earnings information were randomized among participants (see Appendix A). Differences among consensus forecasts, management guidance, and actual earnings varied widely so as to allow ample room for variation in the analysts' responses.8 After reading the earnings information, participants were asked to provide a range forecast of earnings for FY 2006 (lower-bound, mid-point, and upper-bound). This initial range forecast was designed to simply establish a baseline against which to compare subsequent revisions, as a way to control for individual variation. The corporate governance rating had not been introduced to participants at this point. This concluded Part I of the study.
The analysts were next handed Part II of the study, where they read the following:9
You have just learned that XYZ [ABC] Company has received a “below average” [“above average”] corporate governance rating from a reliable rating service. Corporate governance ratings are based on evaluation of a given company against its industry peers and other leading companies on such factors as board and management structure, disclosure practices and record, institutional share ownership, shareholder rights, management and board share ownership, size and type of auditor compensation, and executive compensation.
At this point, you have an opportunity, if you want, to change your earlier range forecast. If you do not wish to change your initial forecast, please enter NC for “no change” in the boxes below and then turn the page to Part III of the study.
After reading this information, the analysts either changed or maintained their initial forecasts.10
The analysts next moved to Part III of the study, where they read earnings information about a second company. The earnings information covered the same period as the first case company, but the amounts and year-to-year changes varied between the two case companies (see Appendix A). After reading the information, the analysts provided an initial range forecast. They were then handed Part IV of the study, wherein a corporate governance rating opposite from the first case was revealed. As with the first case company, the analysts were offered an opportunity to hold to or change their initial forecasts. The study administrator collected Parts I through IV of the study, and handed out Part V, which included manipulation check, debriefing, and demographic items.
The data allowed for calculation of initial (pre-range) and final (post-range) earnings forecast ranges (upper bound minus lower bound). Within each treatment (above average and below average corporate governance ratings), the dependent variable reflected the percentage change from the initial to the final range, determined as follows:
If the percentage change is negative (positive), the final range will narrow (widen) over the initial range, reflecting less (more) uncertainty in the earnings forecast.
Fifteen male and two female buy-side analysts from a top-10 U.K. mutual fund attended their company-sponsored session, and all volunteered to participate in the study. All were U.K. citizens and lifelong residents. Their mean (standard deviation) age and years of experience as analysts were 35.71 (7.00) and 9.76 (4.62), respectively. The U.S. sample included 17 male and two female buy-side analysts. They were among 21 analysts who attended the training session.12 All analysts were U.S. citizens and lifelong residents, and their mutual fund company is considered in the top 25 in the U.S., based on total investments. Their mean (standard deviation) age and years of experience as analysts were 34.00 (6.46) and 8.53 (4.38), respectively.13
Once all study materials were collected from the analysts, they responded to two manipulation check items. The first (second) item asked whether the corporate governance rating for ABC (XYZ) Company was above or below average. Since the case company names (ABC and XYZ) were randomized within-participant, the manipulation check items did not necessarily follow the manipulation order. All participants, except for one U.S. analyst, correctly answered the manipulation check items related to corporate governance ratings in accordance with the treatment.14
We also asked the analysts to provide the percentage of the U.S. firms listed on major U.S. stock exchanges in their portfolios, relative to the overall portfolio of firms that they analyze on a regular basis. We asked this question because the experimental materials described the case company as a firm that was listed on a major U.S. exchange. The mean percentage (standard deviation) of U.S. firms listed on major U.S. exchanges in the analysts' portfolios was 94.63 (5.88) and 90.83 (2.99) for the U.S. and U.K., respectively. While the means are significantly different (t = 2.49, p = 0.02), the mean percentage of U.S. firms that are analyzed by both U.S. and U.K. analysts are nevertheless quite high (over 90 percent).
Finally, using a 0–5 Likert scale (0 = never, 1 = very seldom, 5 = very often), we asked the analysts to provide the extent to which they forecast earnings during the normal course of their job. The purpose of asking this question was to ensure that the experimental task was appropriate for the participants. The relatively high overall mean (standard deviation) of 4.86 (0.35) indicates congruence between their experimental and professional tasks.15 Overall, the manipulation check results indicate that the participants understood the experimental treatments to which they were assigned and the experimental task was appropriate for the analysts.
Mean forecast lows, highs, and ranges are shown in Table 1; Panel A shows the combined U.S. and U.K. results, Panel B offers the U.K. results, and Panel C presents the U.S. results. Table 2 offers the overall percentage change in earnings forecast ranges from pre- to post-corporate governance ratings treatment. For U.K. analysts, there was a mean pre-to-post range widening of 45 percent in the below average condition; for the U.S. analysts, the corresponding mean widening was 6 percent, suggesting that a below average corporate governance rating yielded less certainty in the earnings forecasts. For above average corporate governance ratings, U.K. analysts contracted their range by −14 percent and U.S. analysts contracted their range by about −16 percent, indicating that an above average corporate governance rating resulted in greater certainty in the earnings forecast (Table 2).
The primary interest in this study is to investigate whether an analyst's earnings forecast range widens or contracts, conditioned on the quality of a firm's corporate governance ratings. However, it is also possible that some analysts may not just widen their forecast range when corporate governance ratings are below average and contract it when they are above average, but may shift their range upward or downward. We investigate this possibility, by grouping the results by country, by corporate governance ratings condition, and overall, and find that there was not a shift in the range, but rather a contraction when corporate governance ratings are above average, and a widening when corporate governance ratings are below average. The results of this analysis are presented in Table 1. Panel A contains the data for analysts from both countries combined, Panel B contains data for U.K. analysts, and Panel C contains data for U.S. analysts. This result strongly suggests that, as expected, analysts attended to the effect of corporate governance ratings on forecast certainty as intended.
A preliminary ANCOVA model (not tabulated) indicated that the following potential covariates were non-significant (p > 0.10): order of corporate governance ratings treatment (above average or below average), order of case company name (ABC or XYZ), order of earnings information set (one or two), percentage of U.S. firms in the analyst's portfolio, age, years of experience as an analyst, and gender. Hence, these variables are not included in subsequent analyses.
We assessed three more potential covariates related to the analysts' (1) familiarity with one or more corporate governance scoring agencies, (2) use of any corporate governance ratings when making earnings forecasts, and (3) extent to which their companies expect them to use corporate governance ratings when forecasting earnings (see Table 3 for wording and results). These items were considered as covariates because their inclusion could account for other potential explanations for the dependent variable responses, should they be significant. A preliminary ANCOVA model indicated insignificance (p > 0.20) for each variable; thus, these metrics were not included in the upcoming analyses.
The study hypothesis posits that an above (a below) average corporate governance rating will increase (decrease) earnings forecast certainty. As shown in Table 2, the mean pre-to-post percentage change in earnings forecast is +25 percent for the “below average” condition, which indicates a widening of final over initial ranges (less certainty); and the mean pre-to-post percentage change in earnings forecast is −15 percent for the “above average” treatment, which suggests a narrowing of final over initial ranges (more certainty). The means are directionally consistent with the hypothesis; this, coupled with a significant within-participant main effect (p < 0.01) for corporate governance ratings in the ANOVA model (Table 4, Panel A), provides support for the hypothesis.
The research question asks whether there will be any differences in forecast certainties between the U.S. and the U.K. analysts when the corporate governance ratings are above or below average. The significant effect (p < 0.01) of “corporate governance rating × country” in the within-participant model (Table 4, Panel A) and significant “country” effect (p < 0.01) in the between-participant model (Table 4, Panel B), suggest significant response differences between the two countries. Figure 1 illustrates the mean response patterns for the two countries.
To determine significant differences among means, we performed several multiple pair-wise comparisons using the Sheffe, Bonferroni, and Duncan procedures (alpha = 0.05). All three tests indicate, as shown in Figure 1, that U.K. analysts responded significantly stronger to the below average corporate governance ratings than U.S. analysts, and both U.K. and U.S. analysts responded similarly to the above average corporate governance ratings. Post-experimental debriefing questions can offer insight into psychological processes that underlie forecast certainty differences between the U.S. and U.K. analysts. The wording and results of such debriefing items assessed in the current study are shown in Table 5.16
The first item asks about the analysts' a priori presumption regarding the quality of corporate governance for U.S. firms listed on a major U.S. stock exchange, in the absence of independent verification thereof. Interestingly, the U.K. analysts presume that the quality of corporate governance is moderately high (mean = 3.94), while the U.S. analysts' presumption is somewhat low (mean = −2.36). Item two asks about the analysts' a priori presumption regarding the extent to which they believe the Sarbanes-Oxley Act (U.S. House of Representatives 2002) affects the quality of corporate governance for U.S. companies listed on U.S. stock exchanges. U.K. analysts believe that SOX (U.S. House of Representatives 2002) improves corporate governance quality (mean = 3.71) to a significantly greater extent than U.S. analysts (mean = 0.92). The different a priori presumptions of corporate governance quality and perceptions of SOX influence help explain the U.K. and U.S. analysts' differential responses to the treatments, as explained next.
Referring to Figure 1, when U.K. analysts learned that corporate governance was above (below) average, such information confirmed (disconfirmed) their a priori presumptions of corporate governance quality and SOX (U.S. House of Representatives 2002) beliefs, thus their response to the above average treatment was relatively weaker (absolute percentage change in range estimate = 14 percent) than the below average condition (percentage change in range estimate = 45 percent; t = 4.16, p < 0.01). Conversely, when U.S. analysts learned that corporate governance was above (below) average, such information disconfirmed (confirmed) their a priori corporate governance presumptions and SOX (U.S. House of Representatives 2002) beliefs, hence their response to the above average treatment was relatively stronger (absolute percentage change in range estimate = 16 percent) than the below average condition (percentage change in range estimate = 6 percent; t = 2.07, p = 0.04).
We averaged items one and two from Table 5 into a single index reflecting presumptive beliefs about corporate governance quality (r = 0.80, p < 0.01). We then performed mediator-moderator analyses as described in Baron and Kenny (1986), the results of which are shown in Table 6. The first three steps are necessary to establish mediation. In step one, the mediator (presumptive belief index) must be significantly related to the independent variable (corporate governance rating); however, in this instance, the relationship is not significant, suggesting that the presumptive belief index is not a mediator. Additional support for non-mediation is found in steps two and three, where the beta coefficient for corporate governance rating in step two is unchanged in step three, where the expectation for mediation is that this beta coefficient should be significantly lower. Referring to step four, corporate governance rating, the presumptive belief index, and the interaction of corporate governance rating and the presumptive belief index are all significant (p < 0.01). The significant interaction term suggests that the presumptive belief index moderates the relationship between corporate governance ratings and earnings forecast certainty, where the nature of the moderation can be visualized in Figure 1.
Since differences in a priori presumptions about the quality of corporate governance moderate the results, then one might expect that the initial earnings forecast ranges (before learning about the corporate governance rating) might indicate more certainty for U.K. analysts than U.S. analysts, as the U.K. analysts appear to hold a higher presumptive belief about the quality of corporate governance of U.S.-based companies than do the U.S. analysts. The mean (standard deviation) of the initial range for the U.K. analysts of 1.01 (0.47) was significantly smaller (t = 2.40, two-tailed p-value = 0.015) than the U.S. analysts' mean (standard deviation) of the initial range of 1.29 (0.49), suggesting that the U.K. analysts indeed were more certain in their initial forecasts, which is consistent with their stated a priori beliefs. Thus, to assess the incremental effect of the corporate governance ratings on earnings forecast certainty, it was necessary to standardize the final earnings forecast range (after learning about the corporate governance rating) by the initial range.
The next two debriefing items ask how the analysts' perceptions of the accuracy of a management-issued earnings forecast would change if they subsequently learned that the company received an above average (item 3) or below average (item 4) corporate governance rating from a reliable source. For the above average item, the analysts indicated that their perceptions would increase somewhat (overall mean = 1.67). The U.K. and U.S. sample mean responses were not significantly different, even though an above average rating would disconfirm the U.S. analysts' a priori presumptions and beliefs about relatively weak corporate governance quality; the non-significant difference is likely due to the U.S. analysts' relatively limited incorporation of corporate governance indicators in their decision-making processes (see Table 3, item 2). With regard to the below average item, both groups of analysts indicated a perception of less accuracy (overall mean = −2.17), and the U.K. mean (−3.00) was significantly lower than the U.S. mean (−1.42; p < 0.01), which is consistent with disconfirmation of the U.K. analysts' a priori presumption of relatively good corporate governance quality.
Items 5a and 5b assessed analysts' reactions to above and below average corporate governance ratings on stock recommendations. Both items are on a scale of +1 (strong buy) to −3 (sell). In the above average condition (5a), the U.K. analysts moved from a “buy” to a “strong buy” recommendation (0.82 > 0, t = 8.64, p < 0.01), as did the U.S. analysts (0.26 > 0, t = 2.53, p = 0.02); however, the U.K. response was significantly stronger than the U.S. response (t = 3.95, p < 0.01). In the below average condition (5b), the U.K. analysts moved from a “buy” to a “weak buy” recommendation (−0.76 < 0, t = 4.19, p < 0.01), as did the U.S. analysts (−1.11 < 0, t = 5.50, p < 0.01); however, the two means were not significantly different (t = 1.24, p = 0.22). Taken as a whole, responses to items 3 through 5b are consistent with the experimental results, thereby suggesting that the analysts' responses to the treatments were conscious. They are also consistent with analysts reacting more strongly to corporate governance ratings that are inconsistent with their a priori presumption about a firm's quality of corporate governance.
Debriefing items 6 and 7 explore the influence of corporate governance ratings on the analysts' perceptions of the reliability of public disclosures, other than earnings forecasts, issued by company management. If the rating is above average, perceived reliability increases (overall mean = 1.34), with no significant difference between U.S. and U.K. analysts. If the rating is below average, perceived reliability drops for U.S. analysts (−1.83) and declines significantly more for the U.K. analysts (−2.88). Again, these results are consistent with the experimental results and the analysts' initial presumptive beliefs about corporate governance quality.
The remaining debriefing items (8 through 10) assess the analysts' familiarity with and use of ratings from four major corporate governance-rating agencies. We do not describe each item due to space limitations, but the responses might be of interest to other corporate governance researchers. In summary, analysts are most familiar with and use Standard & Poor's (S&P) and Institutional Shareholder Services (ISS, now RiskMetrics) as corporate governance-rating agencies, with the U.K. analysts favoring ISS and U.S. analysts favoring S&P.
The current study provides evidence that buy-side financial analysts rely on corporate governance ratings issued by a reliable third party when forecasting earnings, as the experimental results indicate that a below (an above) average corporate governance rating decreases (increases) the analysts' certainty in their earnings forecasts. We further find that analysts place less weight on corporate governance ratings when the ratings are consistent with their a priori presumption about the quality of corporate governance of U.S. firms listed on a major U.S. stock exchange, and place more weight on corporate governance ratings when these ratings are inconsistent with their a priori presumption.
Experimental results and debriefing question analyses help answer the research question, asking whether analysts from the U.S. and the U.K. will exhibit differential forecast certainty when a corporate governance rating is above or below average. We estimate how analysts incorporate corporate governance ratings into their forecasts by analyzing the change in the forecast range before and after learning about the quality of corporate governance. A reduction in the pre-to-post forecast range is consistent with more forecast certainty and vice versa.
We find that the relative magnitude of responses to corporate governance ratings differs between the U.K. and the U.S. analysts in that U.K. analysts react stronger to a rating that is inconsistent with their a priori presumption of corporate governance quality than U.S. analysts. The relative under-reaction of U.S. analysts could be partially explained by their lower use of corporate governance ratings when forecasting earnings, as compared to U.K. analysts. Post-experiment analysis of country differences coupled with the debriefing items are insightful, as they point to the need for more culturally based research in behavioral accounting and finance.
Our study is limited in several ways. First, by necessity of time constraint, background information of the case companies was highly summarized. Had the analysts known more about the companies, perhaps the effect of corporate governance ratings on their forecast certainty would have been different, as they would have been in a better position to assess the inherent riskiness of the case companies. We did not collect data on the analysts' perceptions of company riskiness, which likely added statistical noise to our data and worked against the odds of finding a difference between the corporate governance rating treatments posited in our hypothesis. Second, while analysts' responses could have been affected by their individual confidence levels, scaling the revised forecasts by the initial forecasts helps to minimize the effect of differential individual confidence levels; as such, differences would comprise part of the forecast range dispersions. With regard to company riskiness perceptions and individual confidence differences, random assignment of participants to treatment conditions likely neutralized any systematic effects of these constructs on the study findings. The participating analysts were not randomly drawn from their companies, neither were the companies randomly drawn from all possible companies that employ buy-side analysts; thus, the research findings should be interpreted with this limitation in mind.
Finally, access to busy professionals such as financial analysts for data collection is based on their availability and interest in the subject matter of the study. While we were fortunate to have access to financial analysts in the U.K. and the U.S., our access was limited to buy-side analysts. These analysts primarily serve the interests of institutional investors, while sell-side analysts produce reports of stock recommendations such as strong buy, buy, hold, sell for various users. While we have no reason to believe that buy-side and sell-side analysts use corporate governance information differentially for their earnings forecasts, we nevertheless caution about the generalizability of our results to sell-side analysts. Future studies may be needed to investigate sell-side analysts' attendance to corporate governance in their earnings forecasts and stock recommendation.
The cultural aspect of the current study is also interesting and potentially insightful. In a review of the corporate governance and financial reporting literature, Cohen et al. (2004) suggested that cultural differences in how investors respond to various corporate governance mechanisms is largely ignored in the literature. Our results confirm that this is an area worthy of more research in the future. For instance, research aimed at developing a cultural theory of how and why analysts and investors from diverse legal and regulatory regimes respond differently to corporate governance quality could complement and extend the cultural framework provided by Hofstede (1991), as revised by House et al. (2004).
As a beginning step toward such theory development, a recent study by the Institute of Chartered Accountants of England and Wales (ICAEW, 2007) can help to partially explain the asymmetric influence of the corporate governance ratings on U.K. and U.S. analysts' earnings forecast certainty. The ICAEW argues that despite similar legal origins, the U.S. regulatory system is more rules-based, while the U.K. system is more principles-based. Thus, the U.S. analysts could have minimized or under-weighted their impounding of the corporate governance ratings into their earnings forecasts because there is no rule stating that they must consider such information. In contrast, the U.K. analysts might have considered the rating information more in the spirit of its content, than the rule of its inclusion. This principles-based versus rules-based explanation could be examined further in future research.
Finally, on December 7, 2006, the U.K. revised its corporate governance regulation through the new Companies Act. It represents the most extensive changes of U.K. corporate law in more than twenty years. While the purpose of the Companies Act (U.K. Parliament 2006) includes strengthening corporate governance, the early focus will be on bringing the U.K. in line with the European Union's takeover transparency directive, shareholder communications, and e-commerce. We believe that there will be opportunities to investigate the effects of the new Companies Act on closing the differences between financial analysts from the two countries. Nevertheless, due to cultural and educational differences (Diaz 2002), there is reason to believe that significant differences might continue to exist at some level.
We do not know, ex ante, how buy-side analysts assess corporate governance quality through private channels or derive corporate governance quality through self-determined means. In the current study, the analysts are asked to provide an earnings forecast related to a case company for which they have no other information about corporate governance quality except for the third-party corporate governance ratings experimental treatment. We recognize that the analysts' responses to the corporate governance ratings will reflect differences from their a priori presumption about corporate governance quality in general.
LaPorta et al. (1998) documented additional enforcement factors, such as judicial efficiency, rule of law, and anti-director rights. Neither LaPorta et al. (1998) nor Hope (2003) compare countries one-to-one. As a result we cannot make any inference about statistical significance of the difference in enforcement scores between the U.K. and the U.S.
The case company name (ABC or XYZ) and prior earnings information set (one and two) are also randomized among participants (see Appendix A for the two information sets) to preclude the possibility of a confounding order effect.
The researchers were not involved in either the information session held in the U.K. or the training session held in the U.S.
Pilot test participants (two senior buy-side analysts from the U.K. and one senior buy-side analyst from the U.S.) suggested that we describe the case companies as being located in the U.S. and listed on a major U.S. stock exchange. The U.K. pilot test analysts worried that the U.K. participants might respond differently to a European firm relative to a U.S. firm, thereby making it difficult to compare responses between the U.K. and U.S. The U.K. pilot test analysts also noted that the U.K. participants who were available to participate in the current study mostly analyze U.S. firms listed on major U.S. exchanges.
Pilot test participants indicated that the earnings information was easy to understand and would likely allow for necessary wide variation in the participants' task responses (i.e., range forecasts). Also, the analysts indicated that they could not discern any particular earnings pattern (e.g., upward or downward trend) in either set of earnings information (see Appendix A), which was intended by the researchers so as not to confound the earnings pattern with the corporate governance rating manipulations. To accomplish the “no pattern” earnings history, up and down movements in the year-to-year earnings information were randomized from 2001 through 2006 within both sets of earnings information.
The pilot test analysts agreed that we should not mention any particular corporate governance ratings provider; rather, we should refer generically to a “reliable rating agency.” The analyst also believed it necessary to include a brief description of the types of factors impounded into corporate governance ratings.
Participants were told not to change earlier responses as they progressed through the study materials. To help ensure compliance, participants used a blue ink pen throughout the experiment. This way, if they tried to change a prior response, such attempts would be indicated.
The item of interest in this study is more akin to the second moment than the first moment. We take the level of EPS as given and ask the question: How does the dispersion around the mean forecast change with the quality of corporate governance? The corporate governance literature that tests for the positive effects of corporate governance generally uses measures related to uncertainty about the firm's reports. For example, in a study that finds that high-quality corporate governance is associated with lower cost of capital (Ashbaugh-Skaife et al. 2006), the reduced cost of capital is associated with a reduction in uncertainty (a risk construct) about the firm, which is related to the second moment. Even for studies that test for an association with firm values (e.g., Cremers and Nair 2005), firm value itself is affected by a construct related to uncertainty or risk (i.e., the lower discount factor implicit in higher firm values), variance, or the second moment.
The two analysts who did not want to participate gave no reasons. They stayed in a lounge area outside the training room during administration of the study.
Between the U.K. and U.S. samples, mean age was statistically non-significant (t = 0.76, p = 0.45) as was mean years experience (t = 0.12, p = 0.91).
Upcoming testing was conducted with and without the U.S. analyst who incorrectly responded to the manipulation check item. The results were qualitatively and quantitatively similar in both analyses, although power was slightly weakened with the analyst included in the study. Since leaving the analyst in the sample served to bias findings against rejecting the null hypothesis, he was included in all upcoming analyses.
The U.K. and U.S. means were not significantly different (t = 1.59, p = 0.12).
Pilot test participants helped the researchers to decide the types of debriefing items that would offer insight into explaining why U.S. and U.K. analysts might respond differently to the experimental treatment (corporate governance rating: below or above average). The pilot test participants also helped the researchers considerably with item wording and scaling throughout the entire set of experimental materials.
Published Online: November 2011
This paper is based on the first author's doctoral dissertation at Bentley University, where Mohammad Abdolmohammadi (Chair), James Hunton, Ganesh Krishnamoorthy (External Reader), and William Read served on the dissertation committee. We are indebted to the following people for offering invaluable critiques at various stages of writing this paper: Jean Bedard, Ann Dzuranin, Christine Earley, Dorothy Feldman, Kelly Pope, Tracy Noga, Andrew Rosman, David Shwarzkopf, and Jay Thibodeau. We have also benefited from participant comments at The PhD Project Accounting Doctoral Students Association conference held in Chicago in August 2007; the Accounting, Behavior, and Organizations Section Research Conference held in October 2007 in Philadelphia; the University of Baltimore; Bryant University; the University of Central Florida; the University of Massachusetts Dartmouth; the University of New Hampshire; and the University of North Texas. We thank an anonymous reviewer at BRIA for a very constructive review.