An Investigation of the Causal Effect of Voluntary Disclosure Quality on Cost of Equity Capital

Master's Thesis 2012 94 Pages

Economics - Finance




2.2.1 Benchmark analysis (Botosan, 1997)
2.2.2 Financial vs. social disclosure (Richardson & Welker, 2001)
2.2.3 Different disclosure types (Botosan & Plumlée, 2002)
2.2.4 Swiss capital market (Hail, 2002)
2.2.5 Longitudinal analysis (Gietzmann & Ireland, 2005)
2.2.6 Management earnings forecast (Baginski & Rakow, 2008)
2.2.7 Endogeneity bias (Larcker&Rusticus, 2010)
2.3.1 Earnings quality (Francis, Nanda & Olsson, 2008)
2.3.2 Operating cash flow (Cohen, 2008)
2.3.3 Bid-ask spreads (Welker, 1995)
2.3.4 Accounting choice (Leuz & Verrecchia, 2000)
2.3.5 Realized returns (Eugster & Wagner, 2011)

3.2.1 Definitions
3.2.2 The residual income valuation valuation model
3.2.3 The abnormal earnings growth valuation model
3.2.4 Advantages and disadvantages of the models
3.2.5 Cost of equity capital proxy
3.3.1 Proxy for disclosure quality
3.3.2 Theoretical foundations and hypothesis development Public and private disclosures and information processing Endogenous disclosure policy
3.3.3 Important firm characteristics Exogenous control variables Instrument candidates
3.3.4 Descriptive statistics
3.3.5 An instrumental variable approach

4.2.1 Preliminary stage results
4.2.2 2SLS first-stage estimation results
4.2.3 OLS and reduced form estimation results
4.3.1 Causality hypothesis
4.3.2 Endogeneity hypotheses
4.3.3 Alternative explanations
4.4.1 Market sectors
4.4.2 Analyst following




The relation between voluntary disclosure quality and cost of equity capital is widely known to be affected by self-selection. This study controls for potential endogeneity bias by adopting a two-stage least squares approach in a cross-sectional setting. Voluntary disclosure quality is proxied by the annual reports disclosure scores for a well-diversified sample of Swiss firms as developed by the Department of Banking and Finance of the University of Zurich. Further, an ex-ante cost of capital metric derived from the dividend discount model is used in this study. Empirical evidence shows that the association between voluntary disclosure quality and cost of equity differs with a firm's stock listing history. While the relation is predicted to be negative for firms at the IPO stage, it is likely reversed at some point in a firm's stock listing history. These results suggest that analysts' information processing activities negatively moderate the impact of voluntary disclosure quality on firm value. Importantly, the predicted interaction between voluntary disclosure quality and stock listing history remains significant when adjusting for endogeneity.


I. Problem

Corporate disclosure is essential for the proper functioning of capital markets. Managers, analysts and market specialists have expert knowledge about their firm, which they can use to deceive potential shareholders about the value of their investment project or expropriate resources provided by investors to the firm. This informational disadvantage creates information asymmetry between the firm and its investors. As a consequence of investors’ uncertainty about the quality of investment opportunities and lack of distrust towards management, investors charge an equity premium for providing capital to the firm. This additional equity premium causes firms with a high information asymmetry to have a ceteris paribus higher cost of equity capital than firms with a low information asymmetry. To the extent that a higher cost of equity capital is detrimental to the value of a firm, reducing information asymmetry is also in the interest of managers who aim at maximizing firm value. Management can reduce information asymmetry by providing higher quality disclosures of private information to the capital markets. However, increasing disclosure quality involves direct costs associated with the production of quality information and indirect costs arising from reduced incentives, litigation and proprietary costs for the firm (Core, 2001). Management thus faces a trade-off of reduced information asymmetry against direct and indirect costs from increased disclosure quality. As a result, increasing disclosure quality beyond the level of mandatory disclosures required by disclosure regulations may not always be optimal. In particular, information that unfavorably affects firm value is not disclosed to the capital markets (Verrecchia, 2001).

Since disclosure costs and the quality of accounting information are different across firms, the choice of disclosure level is unlikely to be a random event but rather depends on firm performance and governance (Healy & Palepu, 2001). As a result, the association between voluntary disclosure quality and cost of equity capital is most likely affected by self-selection bias. The direction of causality in documented associations using standard OLS regression techniques is therefore unclear, resulting in researchers' attempts to deal with the underlying endogeneity problem (Welker, 1995; Hail, 2002; Brown & Hillegeist, 2003). Recent studies analyzing the impact of corporate disclosure on firm value have increasingly made use of instrumental variables and fixed-effects estimation to mitigate endogeneity problems arising from unobserved firm heterogeneity and omitted variables bias (Nikolaev & Van Lent, 2005; Larcker & Rusticus, 2010). Nevertheless, empirical evidence on the sign and magnitude of the relation between voluntary disclosure quality and cost of capital has produced conflicting results so far, suggesting that these novel techniques are only partially successful in reducing endogeneity problems.

II. Purpose of the study

This study investigates the relation between voluntary disclosure quality and cost of equity capital in a cross-sectional setting. A two-stage least squares approach (2SLS) is adopted to adjust for endogeneity. The study attempts to close the extant research gap by evaluating potential instrument candidates that are exogenous to a firm's disclosure policy. Further, this study aims at identifying the driving force behind the likely variable association of voluntary disclosure quality with cost of equity. In particular, the impact of analyst following and analysts' behavior with respect to their familiarity with firms that have a long stock listing history is explored in this study. Assuming that the relation in question is effectively moderated by a firm’s stock listing history, an interaction between voluntary disclosure quality and stock listing history is included in the empirical disclosure model. Methodologically, the analysis thus also addresses the difficulties related to finding valid instruments in the presence of interaction effects with exogenous control variables. The study makes a substantial contribution to contemporaneous accounting research through justifying the choice of instrumental variables by economic theory and reporting specification tests for instrument weakness and relevance (Angrist & Krueger, 2001; Larcker & Rusticus, 2010).

III. Methods

Following Hail (2002) and Eugster & Wagner (2011), this study uses the annual value reporting rating conducted by the Department of Banking and Finance (DBF) as a measure of voluntary disclosure quality. Moreover, I adopt the mean implied cost of equity method as proposed in Hail & Leuz (2006) to calculate my ex-ante cost of equity capital proxy for a well-diversified sample of Swiss firms (Botosan et al., 2011). The study investigates a sample of 130 firms in nine different market sectors controlling for various firm characteristics such as stock listing history, firm size, financial leverage, profitability, and risks associated with growth opportunities and stock price volatility. To test the causality hypothesis, I develop an empirical disclosure model that includes a proxy for stock listing history as a moderator variable, thus analyzing the role of accumulated background information with respect to analysts' information processing activities.

Analysts are assumed to make increased use of accumulated background information about a firm to enhance their production of private information from publicly disclosed quality information with growing stock listing history. Since analysts’ information processing activities exacerbate information asymmetry but are positively associated with voluntary disclosure quality, they increasingly offset the direct negative effect of voluntary disclosure quality on cost of equity capital. As a result of the positive indirect effect of enhanced analysts’ information processing activities on information asymmetry becoming stronger, the net effect of voluntary disclosure quality on cost of equity is reversed at some point in stock listing history.

Apart from the causality hypothesis, I introduce two further hypotheses stating that self- selection likely causes OLS coefficients to be insignificant and biased towards zero (reverse causality hypothesis) and that the 2SLS model should outperform OLS by producing consistent coefficients in the presence of endogeneity (consistency hypothesis). These hypotheses are subsequently tested implementing contemporary model specification tests such as the Cragg-Donald (1993) F -statistic for weak instruments, the Stock-Wright (2000) S - statistic for instrument relevance and the Durbin-Wu-Hausman endogeneity test.

IV. Results

The multivariate regression results show that a significant relation between voluntary disclosure quality and cost of equity capital exists that strongly varies with my proxy for stock listing history. These findings are consistent with Botosan (1997), who reports a negative association between her voluntary disclosure measure and cost of equity capital only for firms with a low analyst following, and Kristandl & Bontis (2007), who document an unexpected positive relation for firms providing a relatively high level of historical information in their annual reports. Both, analyst following and the proportion of historical information provided in annual reports disclosures are likely to increase with a firm’s stock listing history.

Consistent with evidence from prior studies, firms with a relatively long stock listing history, a lower book-to-market ratio, higher capital intensity and lower risks associated with growth opportunities and price volatility have a lower cost of equity capital. Surprisingly, firm size, profitability and market leverage do not show any association with cost of equity capital regardless whether treating voluntary disclosure quality as endogenous or not.

The main conclusions are unchanged when adjusting for endogeneity, suggesting that stock listing history actually moderates the relation between voluntary disclosure quality and cost of equity capital. However, the results should be interpreted with caution since 2SLS does not improve over OLS and the study’s findings may not be generalizable to alternative measures of voluntary disclosure quality, different market sectors or investigation periods.


Does voluntary disclosure quality pay off? And if so, what are the driving forces behind the relationship of voluntary disclosure quality and the cost of equity capital? This study addresses these and other questions in the context of analyzing the determinants of the cost of equity capital for Swiss firms.

The cost of capital is a fundamental question for the decision-makers of firms (Easley & O’Hara 2004). When assessing the profitability of investment projects, the firm’s decision is based on the hurdle rate. The capital structure of a firm is directly influenced by the cost of equity capital. Therefore, the performance of a firm’s operations and the profitability of the firm as a whole critically depend on its cost of capital. A firm’s cost of capital primary reflects the market’s perception of the risk associated with investing in a firm’s security, which varies systematically across firms and industries (Gebhardt, Lee, & Swaminathan 2001). Managers can take active measures to mitigate the risk associated with investing in their firm’s stock by providing information to the capital markets. Corporate disclosure is an important source of information, since it helps the capital markets in allocating resources effectively. As Healy & Palepu (2001, p. 406) put it: “Corporate disclosure is critical for the functioning of an efficient capital market”. Investors need firm-specific information to assess the value of their investment (La Porta et al., 2000). This information has to be adequately provided by the lenders of investors' resources, which are represented by the firms issuing securities on the capital market (Rikanovic, 2005). Firms provide disclosure through regulated financial reports, including financial statements, footnotes, management analysis and discussion and other mandatory filings (Depoers, 2000; Healy & Palepu, 2001; Rikanovic, 2005). Accounting standards and disclosure regulations are commonly interpreted as legal requirements for firms to provide a certain level of disclosure quality regardless whether it is in the firm's particular interest to do so (Bloomfield & Wilks, 2000). Legal requirements provide investors with mandated disclosures satisfying their needs for information to protect themselves from expropriation through managers. La Porta et al. (2000) argue that investors pay more for securities if they are protected from expropriation, thus raising the incentives for managers and entrepreneurs to issue these securities. Conversely, investors demand a higher risk premium for investing in securities of firms that fail to mitigate the perceived risk of expropriation. Since disclosure requirements are designed to satisfy investors' need for information, mandated disclosures can be viewed as benefiting the shareholders rather than the firm. However, it can be attractive for firms to provide financial information beyond the legal requirements if the benefits from doing so outweigh the disadvantages associated with providing additional disclosures (Depoers, 2000). In their simplest form, disadvantages are measured in terms of disclosure costs generated by the preparation, presentation and communication of managers' private information to the public (Barnea, 2007). Moreover, firms may chose not to disclose more information than legally required if management considers this additional information as a potential source of risk for the publishing firm (Verrecchia, 1983; Lev & Penman, 1990). Risks inherent in disclosing private information include the loss of proprietary information, potential litigation cost and the reduction of competitive advantage (Lev, 1992; Lang & Lundholm, 1993).

Theory has generally proposed a negative association between the quality of information disclosed by a firm and its cost of equity capital (Dye, 2001; Healy & Palepu, 2001; Verrecchia, 2001; Easley & O’Hara, 2004). For this reason, prior literature investigating the incentives for managers to provide voluntary disclosures has focused on the quality of information (Verrecchia, 1983, 1990). From an economic standpoint, firms voluntarily disclose information only if information quality exceeds a certain threshold (Verrecchia 2001). Consequently, managers may withhold information that is unfavorable to the firm or to their own interests (Rikanovic, 2005, p. 37). However, if investors rationally interpret withheld information as unfavorable information, withholding bad news may have a negative impact on the current valuation of the firm (Verrecchia, 1983). Under rational expectations, it will thus be in the best interest of managers to disclose information however unfavorable it may be (Verrecchia, 2001; Rikanovic, 2005). Alternatively, managers may disclose adverse performance news promptly if they fear that failure to do so results in large earnings surprises and stock price declines on the subsequent earnings announcement day (Skinner, 1994). In these models, managers disclose unfavorable information to minimize the threat of litigation from disgruntled investors and to avoid reputational losses associated with failure to disclose bad news in a timely manner. Managers who have a reputation for withholding information may bear costs from reduced credibility of their disclosure activities (Skinner, 1994). However, firms may also decide to disclose unfavorable information and withhold favorable information to deter possible competitors from entering their industry (Evans & Sridhar, 2002). Concerns associated with the loss of proprietary information may thus impose a trade- off between voluntary disclosing bad news to discourage potential competitors from entering the market, and disclosing good news to benefit shareholders (Rikanovic, 2005, p. 38). Firms actually never choose full disclosure of information due to proprietary costs, which arise from disclosure of information that can damage the firm’s competitive position in the product markets1. In other words, firms may elect to withhold proprietary information despite the possibility that outside investors interpret withheld information rationally as being negative (Verrecchia, 2001; Rikanovic, 2005, p. 40).

These direct and indirect costs of disclosure have to be weighed against the benefits of providing voluntary disclosure. Broadly speaking, voluntary disclosures benefit the firms by reducing the cost of capital. Theoretical literature has identified three possible channels through which corporate disclosure is likely to reduce the cost of capital: (i) stock market liquidity, (ii) information intermediation and (iii) estimation risk (Rikanovic, 2005; p. 46, Botosan, 2006; Kristandl & Bontis, 2007).

The liquidity-based approach suggests that corporate disclosure reduces transaction cost associated with acquiring information about a firm's stock (Diamond & Verrecchia, 1991; Kim & Verrecchia, 1994; Verrecchia, 2001). Since transaction cost decreases stock market liquidity, enhanced disclosure of private information improves stock liquidity by reducing transaction cost. The liquidity-based approach relies on the semi-efficient market hypothesis that extant public information is already reflected in market prices (Barnea, 2007). Conversely, information that is held privately by the firm's management is not reflected in market prices until it is voluntarily disclosed to the market (Kim & Verrecchia, 1994). Therefore, voluntary disclosure of private information by corporate management essentially reduces information asymmetry among informed and uninformed investors (Diamond & Verrecchia, 1991; Verrecchia, 2001). To the extent that revealing private information to the capital markets enhances investors' confidence in the market valuation of a firm's stock, voluntary disclosure increases stock liquidity and thus reduces a firm's cost of equity capital. Recent studies extend the link between information asymmetry and the cost of capital (Easley & O'Hara, 2004). In these models, uninformed investors require compensation for expected losses resulting from transactions with informed investors. The additional compensation required by uninformed investors induces firms with a larger proportion of private information to have a higher cost of capital (Easley & O'Hara, 2004). Moreover, the demand for a firm's stock decreases with information asymmetry between investors and the firm, which reduces stock liquidity on capital markets. Given that firms disclosing less private information have a relatively high information asymmetry, voluntary disclosure helps to mitigate information asymmetry and increase investors' demand for a firm's stock, thus lowering its cost of capital (Botosan, 2006). In practice, enhanced voluntary disclosure of private information by managers is expected to lower the bid-ask spreads of the firm’s securities, which is tantamount to reducing its cost of capital (Amihud & Mendelson, 1986).

The information intermediation approach by Merton (1987) distinguishes itself from the liquidity-based approach by assuming that the level of disclosure is the same for all firms, but the information is unevenly distributed among investors. Informational differences across investors may result from a lack of access to corporate disclosure among potential investors in the investment community (Rikanovic, 2005, p. 87). In particular, firms may suffer from low investor cognizance due to a ‘home-bias’ of investors or a low corporate disclosure level (Bushee & Miller, 2009). Enhanced information intermediation as manifested in a higher quality of corporate disclosures reduces this information asymmetry through facilitating information acquisition by investors (Merton, 1987). To the extent that increased voluntary disclosure quality succeeds in mitigating informational differences across investors, it enables the firm to attract a larger investor base, thus ultimately reducing the cost of capital.

Theoretical literature relating to the liquidity-based and information intermediation approach explain the relationship between the cost of equity capital and corporate disclosure addressing two fundamental problems (Healy & Palepu, 2001). First, investors face an information problem due to informational differences and conflicting incentives between managers and investors. Since managers have typically better information about the value of business investment opportunities, they have incentives to overstate their value. There are several solutions to the information or “lemons” problem, which was first introduced by Akerlof (1970). Apart from optimal contracts between managers and investors, regulations that require managers to disclose their private information help resolve the information asymmetry problem (Healy & Palepu, 2001). Information intermediaries, such as financial analysts and rating agencies who engage in private information production, may create additional incentives for managers to disclose their superior information. Second, once business opportunities have been invested in, managers have an incentive to expropriate investors’ savings, thereby creating an agency problem (Healy & Palepu, 2001). The agency problem is closely related to managerial behavior and ownership structure of the firm (Jensen & Meckling, 1976). The solutions to the agency problem are generally similar to those that alleviate the information problem. Importantly, the board of directors can be interpreted as mechanism for monitoring and disciplining management on behalf of outside investors (Healy & Palepu, 2001). Other corporate control mechanisms effectively alleviating the agency problem include the managerial disclosure of private information (Rikanovic, 2005, p. 6).

The third approach addresses estimation risk as a principal component of the cost of capital. Investors have to estimate the parameters of a security’s required return distribution from historical information of firm performance available to them (Rikanovic, 2005, p. 88, Botosan, 2006). In general, estimation risk decreases with the amount of historical information of firm performance provided to investors (Coles & Loewenstein, 1988; Coles et al., 1995). Estimation risk implies uncertainty about firm performance and thus a higher expected cost of capital for low information firms if estimation risk is non-diversifiable (Barry & Brown, 1985; Botosan, 2006). Prior research yields evidence that reduced estimation risk through disclosure activities is reflected in stock returns (e.g. Healy et al., 1999; Clarkson et al., 1996). In particular, increased voluntary disclosure quality benefits investors by reducing the magnitude of periodic earnings surprises (Bushee & Noe, 2000). To the extent that higher voluntary disclosure quality reduces estimation risk, investors require a relatively lower risk premium for holding securities from high disclosure firms, thus decreasing their cost of capital (Barry & Brown 1985; Coles & Loewenstein, 1988; Coles et al., 1995).

Empirical studies investigating the relation between voluntary disclosure quality and cost of equity capital has provided conflicting results (Healy & Palepu, 2001). A substantial part of the empirical literature finds strong negative associations consistent with the notion that disclosure reduces both the information differences and incentive problems between investors and the firm (Welker, 1995; Leuz & Verrecchia, 2000; Hail, 2002; Rikanovic, 2005). Other studies fail to document a significant relation or find only partial evidence due to different disclosure settings and components (Richardson & Welker, 2001; Botosan & Plumlée, 2002; Kristandl & Bontis, 2007). In recent years, the possibility of endogeneity bias has received increasing attention as a means to explain why empirical findings are not consistent with theoretical predictions. It is commonly agreed upon that endogeneity biases the results from Ordinary Least Squares (OLS) regressions of cost of capital on voluntary disclosure quality and causes them to be inconsistent (Core, 2001; Healy & Palepu, 2001; Larcker & Rusticus, 2010). Since empirical findings from OLS regressions are thus likely difficult to interpret, the presence of endogeneity is assumed to be the primary source for the lack of empirical agreement on the sign of the relation (Nikolaev & Van Lent, 2005; Larcker & Rusticus, 2008).

The present study contributes to the existing literature addressing (i) the endogeneity problem in the relation between voluntary disclosure quality and the cost of equity capital and (ii) exploring possible reasons for the observed variation in the causal effect of voluntary disclosure quality on cost of equity. The innovation of the present study consists in applying a novel methodology for calculating an ex-ante disclosure proxy based on four different implied cost of equity capital models first established in Hail & Leuz (2006). Further, the present study goes beyond the previous analysis of the effect of voluntary disclosure quality on the cost of equity in Hail (2002) by providing a thorough analysis of endogeneity issues in a cross-sectional sample of Swiss firms. A comprehensive set of possible determinants of the cost of equity capital and instrument candidates for assessing voluntary disclosure quality are analyzed in a two-stage instrumental variables approach. The standard OLS approach is then compared with the applied two-stage least squares (2SLS) methodology using different model specification and endogeneity tests. To my knowledge, this is the first study relating to the effect of voluntary disclosure quality on the cost of capital of Swiss firms that provides a detailed analysis and comprehensive discussion of possible endogeneity bias in a cross- sectional framework.

The remainder of the present work is organized as follows: Section 2 gives an overview of the empirical literature in this research field so far. Empirical findings from selected studies investigating different disclosure scores in different markets are summarized and their potential approach to solving the endogeneity problem is highlighted. In Section 3, I first describe my sample selection procedure and briefly review the literature on the different valuation models before developing my own cost of equity proxy. Then, I define my identification strategy by describing the dependent and independent variables and develop my hypotheses. Disclosure proxy selection and descriptive statistics for sample firms are provided in this section. In Section 4, I present the empirical results and discuss my findings in order to evaluate my hypotheses. Sensitivity analyses are further conducted to check the robustness of the main results to alternative model specifications. Section 5 concludes.


2.1 General considerations

Investigating the association between the cost of equity capital and voluntary disclosure quality is complicated by the potentially endogenous relation between these two characteristics. In particular, managers of firms with a higher cost of capital may be more tempted to increase their voluntary disclosure quality because the expected benefits are higher than for firms with a lower cost of capital. Failure to account for this possible endogeneity can result in spurious results and misleading inferences for the researcher (Maddala, 1983). While former studies (e.g. Botosan, 1997; Gietzmann & Ireland, 2005) on the relationship between the cost of capital and voluntary disclosure quality completely neglect this issue, increasing effort is been spent to mitigate potential endogeneity bias in contemporaneous accounting research. For instance, Welker (1995) conducted one of the first studies in this particular field that recognized the need to incorporate this endogeneity into his research design. Most studies address the problem by applying an instrumental variable procedure to substitute for the common OLS technique in their primary analysis (e.g. Baginski & Rakow, 2008; Larcker & Rusticus, 2010). Brown & Hillegeist (2003) extend the task to control for endogeneity by employing a three-stage least squares (3SLS) procedure in their study on the relation between information asymmetry and cost of capital. Unfortunately, a situation similar to a perfect natural experiment, which would open the possibility to employ related methodologies such as regression discontinuity design (RDD), has not been available to date (Eugster & Wagner 2011, Angrist & Pischke 2008, p. 205). Despite these serious limitations to using alternative endogeneity approaches, some studies have tried to isolate the causal effects of endogenous disclosure policy changes in a time-series research design (Baginski & Rakow, 2008; Leuz & Verrecchia, 2000).

Numerous studies have been conducted since with different focus on endogeneity issues. Most studies analyze the U.S. or Canadian disclosure market (e.g. Welker, 1995; Botosan, 1997; Lang & Lundholm, 2000), while more recent studies direct their analysis at European markets such as Switzerland (Hail, 2002; Eugster & Wagner, 2011), Germany (Rikanovic, 2005), China (Zhang & Ding, 2006), or investigate the effects of disclosure on cost of capital across several countries (Francis et al. 2005; Kristandl & Bontis, 2007).

2.2 Direct approach

The direct approach includes studies that use both an implied cost of equity capital measure and a disclosure metric that directly measures the quality or level of corporate disclosures. Following the direct approach here also implies using an ex-ante or implied measure of the cost of equity capital. Indirect measures of the cost of equity capital are provided by bid-ask spreads (e.g. Welker, 1995) or ex-post realized returns (e.g. Healy, Hutton and Palepu, 1999; Eugster & Wagner, 2011). Studies taking an indirect approach with respect to the cost of equity capital are reviewed in Section 2.3.

There are several limitations to choosing a direct proxy for voluntary disclosure quality. Some prior studies use the Association for Investment Management and Research (AIMR)2 ratings of U.S. firm’s disclosure activities which include annual and quarterly reports information, and more diffuse disclosures arising from investor relations activities (Botosan, 2006; Leuz & Wysocki, 2006). These ratings capture the usefulness of disclosure information as perceived by analysts (Core, 2001). The AIMR ratings may thus be biased towards large and successful firms (Lang & Lundholm, 1993; Leuz & Wysocki, 2006). Other studies employ self- constructed disclosure measures of firm’s disclosure activities including management forecasts (Baginski & Rakow, 2008) or investor relations activities (Rikanovic, 2005)3. The problem with these disclosure scores is that they fail to capture other disclosure activities that can complement or substitute for written report disclosures. Moreover, these types of measures are limited due to the subjective selection and coding of the relevant disclosures (Leuz & Wysocki, 2006). Apart from causing selection bias, self-constructed disclosure measures can be limited by capturing the existence rather than the quality of particular disclosures.

In the following, I summarize selected studies which incorporate direct approaches in the analysis of the relation between voluntary disclosure and the cost of equity capital.

2.2.1 Benchmark analysis (Botosan, 1997)

In a seminal paper, Botosan (1997) examines the relationship between cost of equity capital and voluntary disclosure level for a sample of 122 manufacturing firms with considerable variation in firm size and analyst following. The analysis is restricted to the machinery industry because different industries may give separate weights to different patterns of disclosure. Since Botosan (1997) uses cross-sectional observations for fiscal year 1990 only, she needs to ensure sufficient variation in disclosure levels by choosing a self-constructed and more narrowly defined measure of disclosure level than given by the standard score. Moreover, Botosan (1997) cites empirical evidence that firms closely coordinate their disclosure policies across different media, although investor relations are somewhat weaker correlated with other publication disclosures (Lang & Lundholm, 1993). Her analysis is thus restricted on the annual report as a standardized means of corporate reporting, which serves as an acceptable proxy for the level of voluntary disclosure supplied by a firm across all disclosure channels (Botosan, 1997, p. 331). She includes the five different categories background information, summary of historical results, key non-financial statistical, projected information and management discussion and analysis, all of which are considered as valuable elements for investors and financial analysts in her disclosure measure. For the regression analysis, the sample is divided in two subsamples based on the median with respect to analyst following. Consistent with findings from prior literature, Botosan (1997) reports significantly higher average disclosure scores for firms with low analyst following, as compared to firms with high analyst following. Botosan (1997) obtains her final measure for voluntary disclosure level, the fractional rank (DRANK), by dividing the rank of the firm’s disclosure score (DSCORE) by the number of firms in the full sample. Her analysis adopts a finite four- period version of the accounting based valuation formula ultimately developed by Ohlson (1995) as a proxy for the cost of equity capital. Botosan (1997) reports a quantitatively high 20% for the mean cost of equity of the full sample, whereas firms with high analyst following yield lower values on average. Moreover, the reported difference in average cost of equity capital of roughly 4% between the two subsamples indicates a negative association between firm size and cost of equity capital. In fact, Botosan (1997) finds a statistically significant correlation between market value proxying firm size and analyst following of 0.819 using the Pearson correlation coefficient. Regarding firms with low analyst following, Botosan (1997) documents a negative relationship between the cost of equity capital and voluntary disclosure level. The magnitude of this effect is such that a one-unit increase in a firm’s disclosure score results in an average reduction in its cost of equity capital of 28 basis points (Botosan, 1997, p. 344). After controlling for the effects of firm size and market beta on disclosure level, this relationship is statistically significant. However, no association between the cost of equity capital and disclosure level was found for firms with high analyst following. These findings are reinforced by an alternative model specification including an interaction term between disclosure level and analyst following. While the coefficient on DRANK remains negative, the coefficient on the interaction term is significantly positive, indicating that the cost of equity decreases in disclosure level at a diminishing rate as analyst following increases. Botosan’s (1997) findings support the conjecture that firms with low analyst following communicate directly with investors through additional information provided by the management. For these firms, including voluntary disclosure types such as forecast information and key non-financial statistics in their annual report substantially reduces information risk. On the other hand, firms with high analyst following typically benefit from providing historical summary information that can be used in analyst forecasts. Put another way, firms with high analyst following tend to communicate with the market via analysts rather than other communication channels such as the annual report. Based on these results, Botosan (1997) hypothesizes that the annual report is an insufficient proxy for voluntary disclosure level when analyst following is high. In her robustness test, Botosan (1997) focuses on model specifications with alternative potential proxies for voluntary disclosure such as the fractional rank of analyst following or the number of Wall Street Journal articles published about a firm. Moreover, she addresses potential multicollinearity problems between the explaining variables included in her regressions, especially regarding the tight association between firm size and analyst following. However, Botosan (1997) restricts her analysis to the simple OLS framework without extending her analysis to alternative model specifications using instrumental variables or the differences-in- differences approach. Apparently, no particular attempt is made to solve the potentially inherent problem of endogeneity bias throughout her robustness checks.

2.2.2 Financial vs. social disclosure (Richardson & Welker, 2001)

Richardson & Welker (2001, hereafter: R&W) investigate the relationship between financial and social disclosure and the cost of equity capital for a sample of 225 firm-year observations from 87 Canadian firms operating in 9 different industry sectors. The authors argue that firms may benefit from enhanced disclosures about social activities to shareholders similarly as from disclosing financial information under the presumption that information about social activities is relevant for assessing the firm's prospects (R&W, p. 599). Besides the well- documented channels of reduced information asymmetry and estimation risk coming from higher financial disclosure, prior literature suggests that social disclosure might influence the cost of equity capital directly through investor preference effects (e.g. Richardson et al, 1999). That is, investors are willing to accept a lower expected return on investments by a company that engages in social activities for which those investors have an affinity. This notion is supported theoretically by literature on organizational behavior which suggests that corporate social responsibility has a legitimate aspect potentially improving the firm's future prospects. In practice, the emergence of Green Funds and Ethical Investigating document the existence of investor preference effects. Arguably, for social disclosures to actually decrease the cost of capital through investor preference effects, it is important that the firm credibly discloses information about social activities to shareholders. This may be indeed the case if a perceived positive relation between social performance and social disclosure exists, as is often assumed in the past literature.

R&W point out that mandatory disclosures in Canada are generally less comprehensive than in the U.S. due to less stringent regulatory requirements, which leads the authors to the conclusion that their sample of Canadian firms provide satisfactory variation in disclosed information. R&W use the assessments of the 1990, 1991 and 1992 annual reports of a broad cross-section of Canadian companies, which were jointly developed by the Society of Management Accountants of Canada (SMAC) and the University of Quebec at Montreal (UQAM), as their financial and social performance measures. Although the SMAC/UQAM disclosure scores are only available for a limited time series and are based on the judgments of less experienced raters than e.g. the AIMR ratings for U.S. companies, they allow the researchers to bypass possibly severe limitations on sample size as imposed by self-generated disclosure proxies (R&W, p. 601).

Following Botosan (1997) and Botosan & Plumlée (2002), R&W adopt an accounting based valuation model as presented in Gebhardt, Lee & Swaminathan (2001, hereafter: GLS) to obtain their estimates of the cost of equity capital. Similarly as in Hail (2002), their formula involves a reversion of the forecasted return on equity (ROE) beyond year 3 using a 9-years period with a linear fading rate to the industry average ROE. For the 224 observations in the overall sample, R&W obtain an average cost of equity capital (COST) estimate of 8.9%, ranging between 1.8% and 22.5%. R&W include the financial disclosure score (FDISC), the social disclosure score (SDISC), beginning-of-year market value of common equity in dollar millions (SIZE), analyst following (NANAL), debt-to-equity ratio (LEV) and return on equity (ROE) in their empirical analysis. Further, they classify about one third of their sample firms as coming from one of the environmentally/socially sensitive sectors known as the oil, gas and chemical industry or the mines, metals and forestry products industry (Richardson et al., 1999). The descriptive statistics comparing between the two groups reveal that environmentally sensitive industries have better financial and social disclosure scores and lower cost of equity capital estimates, but predominantly inferior financial performance than their counterparts in the non-sensitive industries do (R&W, p. 605).

R&W report that the coefficient on the regressors on FDISC is significantly negative, while the coefficient on SDISC is significantly positive on the 5% level or better in a multivariate OLS regression using industry adjusted data and controlling also for NANAL and LEV. They also investigate the interaction effect between FDISC and NANAL as described in Botosan (1997) and find that it significantly influences their cost of equity capital measure, COST, on the 5% level. Whereas the coefficient on FDISC is no more significant, the coefficient on SDISC keeps its positive sign and NANAL is still negatively related with COST. Moreover, R&W test their proposition that the economic situation of a firm mediates the relation between SDISC and COST in an analogous way as the statistical association between FDISC and COST is mediated by analyst following. To this aim, they add the interaction term DROE x SDISC, with DROE being a dummy variable which is set equal to one if ROE is above the industry sector median for year t and zero otherwise. This interaction term has a significantly negative impact on the cost of equity capital, while the statistical properties of the coefficients on the remaining determinants are unaffected by this change. R&W generalize this result by arguing that firms with poor performance face larger costs of equity capital irrespective whether this situation is firm-specific, industry-specific, or even brought about by an economic downturn to the firm. Although R&W analyze the effects of the potentially influential determinants SIZE, ROE, LEV and NANAL on the financial disclosure variable FDISC in an OLS model that is similar to the first-stage regression of an instrumental variable approach, the authors do not even mention the possibility of endogeneity bias. However, it should be noted that an IV approach is particularly worth testing if there is more than one potentially endogenous disclosure variable included in the regression. Regarding the social and financial disclosure measures FDISC and SDISC, R&W (p. 610) indicate that these variables could be estimated using two-stage least squares (2SLS) instead of OLS.

2.2.3 Different disclosure types (Botosan & Plumlée, 2002)

Botosan & Plumlée (2002, hereafter: B&P) extend the analysis in Botosan (1997) to investigate the effect of the timeliness of disclosures on the cost of capital. They use the scores of the industry subcommittees for the annual AIMR Reports as their measures for annual report disclosure and two further types of disclosure, namely other published reports and investor relations activities. While other published reports include more timely disclosures such as quarterly reports, press releases and newsletters, investor relations activities refer to direct contacts such as presentations to analysts. In comparison with Botosan (1997), they select a much broader sample of 3’618 firm-year observations from 43 industries, which includes 668 individual firms in the period from 1985/86 to 1995/96 (B&P, 2002, p. 24). They include the market beta, the natural logarithm of market value and the fractional disclosure rank in the analysis of the cost of equity capital. In their model, market beta is suggested to control for systematic risk, whereas market value is included to avoid a correlated omitted variable bias with both the expected cost of equity capital (CEQ) and disclosure level (B&P, p. 27). Using a four-period form of the dividend discount model, B&P estimate the implied cost of capital by taking the mean of the minimum and maximum long- run price forecasts published by Value Line. Forming median and mean of equity capital estimates, B&P observe an upward trend in their cost of equity capital measure which is reverted after reaching a peak of 22% in 1990. B&P use the fractional ranks of four different measures of disclosure, which proxy for the total (RTSCR), annual (RANLSCR), other publications (ROPBSCR), and investor relations (RINVSCR) disclosure. They report that their three types of disclosure are all correlated with each other, with the highest correlation of 0.634 between RANLSCR and ROPBSCR. In their univariate analysis of the association between CEQ and disclosure level, they find the expected significance and signs of the coefficients relating market beta and firm size with the implied cost of equity capital, respectively. They also demonstrate that their disclosure measures are not consistently correlated with CEQ, which explains the need for including additional control variables in their analysis. The authors then estimate five alternative models including different combinations of the disclosure measures as controls and find that the effect of the coefficient on total disclosure (RTSCR) is not statistically significant as contrary to their expectations (B&P, p. 35). However, they also document that the signs of the coefficients on RANLSCR and ROPBSCR are significantly negative and positive, respectively, implying that the association between disclosure and the implied cost of equity critically depends on the type of disclosure. In particular, the results suggest that the implied cost of equity capital increases with annual report disclosure but is negatively associated with a greater level of more timely disclosure. B&P find that providing greater annual report disclosure amounts to a 0.7 percentage point increase when moving from the least to the most forthcoming firm, whereas a greater other publications disclosure is associated with a 1.3 percentage point decrease. Contrary to Botosan (1997), these results generalize to firms with high analyst following in various industries monitored across an eleven-year time period. The authors also perform a sensitivity test to check for the possibility of correlated omitted variables bias that might occur due to the high correlation between RANLSCR and ROPBSCR in combination with the statistically strong impact of ROPBSCR on cost of equity capital. Estimating the regressions with only one alternative disclosure measure indeed yields insignificant results, which demonstrates the importance of controlling for ROPBSCR when assessing the association between RANLSCR and CEQ (B&P, p. 36). The authors then cite managerial and financial research sources that corroborate their findings that a greater level of timely disclosure may increase the cost of equity capital. They argue that providing greater timely disclosures leads investors to focus on short-term earnings news and attracts a higher proportion of transient institutional investors to the firm, which increases the volatility of stock returns. The detrimental impact of timely disclosures largely offsets the volatility-reducing effect of the trading activity of long-term investors, which is achieved through providing a higher overall disclosure quality (B&P, p. 37). In further sensitivity tests including also the book-to-market ratio and price momentum in their regression equations, B&P report no significant effect on the primarily drawn conclusions. Moreover, adding these variables increases the explanatory power of their full regression model from 6.3% to 18.7% despite not seemingly capturing information risk. Although B&P plausibly claim that they could not find a statistically significant association between total disclosure and the expected cost of equity capital in their eleven-year period analysis controlling for different disclosure types, it remains unclear whether their conclusions might be biased by endogeneity problems. In particular, they do not document any robustness checks that would be able to detect some possibly simultaneous relationships between the four alternative measures of disclosure and the cost of equity capital, respectively. Failing to check the results from multivariate OLS regression analysis for possible endogeneity can lead to erroneous conclusions regarding the statistical significance, magnitude or even sign of the coefficients (Larcker & Rusticus, 2010).

2.2.4 Swiss capital market (Hail, 2002)

Hail (2002) investigates the effect of the quality of corporate disclosures und the ex-ante cost of capital. He focuses on the Swiss market which exhibits a low mandatory level of disclosure compared to other disclosure environments such as the US market investigated in prior studies. Swiss firms are given considerable reporting discretion because the only requirement for stock listed companies is compliance with Swiss GAAP (Hail, 2002, p. 742). This implies that Swiss firms enjoy a high degree of freedom in choosing their voluntary disclosure policy, which leads to a substantial variability in the observed level of disclosures. Hail (2002) evaluates a cross-sectional sample of seventy-three non-financial companies using fractional ranks rather than absolute scores based on the observed quality of corporate disclosure. Hail (2002) adopts a disclosure index developed by the Department of Banking and Finance (DBF, formerly Swiss Banking Institute) of the University of Zurich in 2001, which is based on the amount of voluntary information provided by Swiss firms in their 1997 annual report to shareholders. This disclosure index is restricted to the 50 highest capitalized Swiss firms plus a randomly chosen subsample of 61 smaller and partly non-listed firms, which indicates a substantial cross-sectional variation in firm size (Hail, 2002, p. 752). His cost of capital measure is based on a version of the residual income model, in which the ex-ante cost of capital is computed as the internal rate of return that equates a firm’s intrinsic value to its current stock price. Similar to GLS, he uses a three-stage procedure with a short-term period of explicit earnings forecasts, a medium-term period with earnings forecasts derived by linearly fading return on book equity to the median market sector return on equity, and a long- term perpetuity assuming zero growth.

In his OLS regression analysis, Hail (2002, p. 758) captures the risk effects on the cost of equity capital by a firm’s market beta and market leverage. The influence of firm size on the cost of equity is accounted for by including the firm’s market capitalization. Hail (2002, p. 762) reports that the statistical relationship between his risk proxies and the cost of equity capital is significantly positive as expected. The fractional rank of disclosure enters the regression significantly negative after controlling for market beta, leverage and market value. The simple OLS regression of the fractional rank of disclosure on cost of equity yields an adjusted R -squared of 22%, whereas multiple OLS regression including the three control variables increases its maximum value to 38%. Treating the disclosure measure as exogenous, Hail's (2002) findings imply that the most forthcoming firms could reduce their cost of equity between 1.8% and 2.4% compared to the least forthcoming firms in the sample. In a specification test, he divides his entire sample of firms into two subsamples contingent on the median of analyst following to check whether his findings are affected by this variable. In contrast to Botosan (1997) however, he does not find any evidence for the association between voluntary disclosure and cost of equity capital to disappear for firms with high analyst following.

Hail (2002, p. 764) devotes a further specification test to the problem that firms might choose their reporting quality contingent on the costs and benefits associated with a higher disclosure level. Since this behavior of firms is likely to result in a self-selection bias, Hail (2002) controls for possible endogeneity by adopting a system of equations that treats disclosure level as endogenous. In the first 2SLS equation, he performs a regression of disclosure level on firm size, profitability level, stock listing status, financial leverage and audit firm size. In the second equation, he repeats the analysis of the cost of equity capital with the fitted values from the first-stage regression as instruments for disclosure quality. The coefficient of the disclosure measure becomes even more negative though exhibiting a somewhat higher p - value, which is however still significant on the 5%-level. Surprisingly, the OLS estimates thus rather understate than overestimate the effect of disclosure level on cost of equity capital. Hail’s (2002) specification test from applying 2SLS yields some evidence for the robustness of his results. However, these results should be interpreted with caution due to the small sample size and the predictive power of the instruments being rather weak as reported by Hail (2002, p. 764). Even though addressing the endogeneity problem by using the instrumental variable approach, Hail (2002) does not present any additional statistics for testing either the strength or the exogenous nature of the instruments. According to Larcker & Rusticus (2010), this might have been useful in determining the extent to which the 2SLS estimates are more reliable than their conventional OLS counterparts. Moreover, his analysis relies on a finite horizon version of the accounting-based valuation formula proposed by GLS as the only measure to compute the prospective cost of capital. This suggests that there is a substantial variation in his cost of equity capital measure representing the explanatory variable in both the OLS and the 2SLS regressions.

2.2.5 Longitudinal analysis (Gietzmann & Ireland, 2005)

Gietzmann & Ireland (2005, hereafter: G&I) explore the relationship between strategic disclosure and the expected cost of equity capital by focusing on timely disclosures of quoted firms in the UK. Following B&P, the authors claim that strategic timely disclosures are more relevant to investors and analysts than annual report disclosures. G&I argue that firms with actual earnings above forecast may prefer to increase their level of disclosure, whereas firms with actual earnings below forecast may opt for more aggressive accounting. Due to the expected interaction of firms’ reporting and accounting policies, G&I control for both level of disclosure and accounting policy choice in their analysis. G&I use a self-constructed variable measuring the quality rather than the quantity of timely disclosures for their sample of firms listed on the London Stock exchange, which announce their news via the Regulatory News Service (RNS). Since the listing rules of the London Stock Exchange requires its firms to disclose news that are likely to have a significant price effect first on the RNS, the quality of those disclosures can be characterized by their price sensitivity. The disclosure variable adopted by G&I thus captures ‘newsworthiness’ as a measure for quality rather than the absolute frequency of disclosed news.

G&I calculate the ex-ante cost of equity capital using a three-stage approach first described by GLS and subsequently adopted by Hail (2002). G&I regress their measure of ex-ante cost of equity capital on their within-sample ranked disclosure measure (DISCLOSE), the logarithmic market value (LNMV), market beta (BETA), the logarithmic book-to-market ratio (LNBMV) and debt-to-market value (LNDMV). Further, they add the number of analysts’ forecasts (ANALYSTS), the logarithmic dispersion in analysts’ forecasts (DISP), the mean expected long-term growth rate (LTG) and an indicator variable for post 1999 observation (YEAR) to the OLS regressions. G&I also distinguish between firms employing aggressive accounting from those employing a conservative accounting policy by means of the sign of estimated discretionary accruals. Adopting a modified Jones (1991) model, they define firms as aggressive if they have positive discretionary accruals, and conservative otherwise (Dechow & Dichev, 2002).

G&I focus on a sample of 301 firm-year observations in the IT industry, which they claim to yield sufficient variation in accounting policy choice. The original sample of IT companies with accounts and data in Thomson Reuters Datastream included 1’131 company-year observations, but observations with missing data and missing variables to calculate the cost of capital were removed from the analysis. In particular, G&I report the exclusion of 13 observations with negative CEQ values or large CEQ values above 100%. This procedure yields mean and median values of the cost of capital around 10%, with the dependent variable COST varying between 1% and 61%. G&I collect accounting and market data from Datastream for a maximum of 10 calendar years from 1 January, 1993 to December 31, 2002. They obtain disclosure data form FACTIVA for UK listed firms that are classified as operating within the IT sector at 31 December, 2002. Their disclosure measure DISCLOSE compares RNS disclosures to other disclosures including all other major newswire and national newspaper reports relating to the RNS announcements. DISCLOSE is thus defined as the fractional rank of the ratio of the number of non-RNS disclosures to RNS disclosures over the period of one year, at the end of which the market data is collected.

In contrast to B&P, G&I find a negative association between the cost of equity capital (COST) and the timely disclosure variable (DISCLOSE), which is significant on the 5% level without controlling for accounting choice. As expected, the relationship between COST and DISCLOSE remains unchanged when controlling for accounting choice. G&I attribute their contradictory findings to differences in the construction of the disclosure variable with respect to B&P. On the other hand, G&I observe that firms with aggressive accounting face higher costs of equity capital than firms making conservative accounting choices. When analyzing firms separately according to the different signs of their discretionary accruals, G&I find that the coefficient on DISCLOSE remains significant only for aggressive firms, i.e. for firms with positive discretionary accruals. In the overall analysis, the regressors DISP, LNDMV and YEAR are all positively related to COST. Therefore, more leveraged companies with a large dispersion in analysts’ forecasts or in less favorable economic conditions (post-1999) have higher cost of equity capital. LTG is the only overall significant variable that differs in the sign between aggressive and conservative firms, suggesting that the effect of growth expectations depends on accounting choice. Conservative firms face a decrease in their costs of equity capital with higher expected future growth, whereas this relationship is reversed for aggressive firms. The latter coincides with the findings in GLS and Cohen (2008), who both report a positive association of LTG with their measures of implied cost of equity capital in a multivariate cross-sectional regression setting.

Although G&I document some evidence for both a negative association of the cost of equity capital with timely disclosure and a positive association with accounting choice, they leave open related questions of endogeneity. Arguably, there is a strong concern for interdependence between lagged costs of equity capital and disclosures in a time-series setting as used by G&I. Further, G&I claim that disclosure and accounting choice are causally related to each other while treating disclosure as exogenous. However, it seems unlikely that firms choose their accounting policy independently of cost of equity capital considerations. Assuming the inherently endogenous nature of accounting policy to be true increases the probability that disclosure is also endogenous. Since theory suggests that accounting choice interacts with both cost of capital, and timely disclosure, a reverse causality between the dependent variable and the two explaining variables seems indeed plausible. Thus a robustness check using IV estimation instead of ordinary OLS would be highly recommendable.

2.2.6 Management earnings forecast (Baginski & Rakow, 2008)

Baginski & Rakow (2008, hereafter: B&R) investigate the association between management earnings forecast and the cost of equity capital. Their study substantially contributes to the extant research about the effect of voluntary disclosure on the cost of equity capital since management earnings forecasts is the type of disclosure most strongly influencing investors' payoff forecasting tasks (B&R, p. 3). From a theoretical perspective, management earnings forecasts are thus expected most likely to reduce information asymmetry and estimation risk. B&R develop a measure for management earnings forecasting behavior over a four-year period from 2001 to 2004, which closely follows the passage of legislation to regulate disclosure (Regulation Fair Disclosure; Reg FD). Their disclosure quality metric MFDiscPol captures the fact of supplying a forecast during a four year period as well as the frequency and the precision of quarterly management forecasts issued by a firm from 2001 to 2004 (B&R, p. 10). Similar as in Francis, Nanda & Olsson (2008), B&R believe that the construct validity of their disclosure quality metric is enhanced by using a multi-year determination of management forecasting policy. Moreover, they take the PEG method from Easton (2004) to estimate a firm’s cost of equity capital in their primary tests, in which they exclude the firms from the financial industry, as prior research has shown that the PEG measure may not be valid for these firms (Easton, 2004).

B&R choose the age of the firm, the number of shareholders and the number of analysts following the firm as their proxies for high voluntary disclosure quality. They use the firm's performance measured by the sign of the return on assets (ROA) as a categorical proxy for the incentive to supply management forecasts. B&R also include two further proxies for financing activities, namely capital intensity and the percentage change in common shares outstanding adjusted for stock splits during the period from 2000 to 2004, as instruments in their first- stage model. Following Cohen (2005), they also proxy for longer-run disclosure costs and benefits by using the Herfindahl-Hirschman Index at the end of 2000, in addition to two dummies capturing whether the firm operates in a regulated industry or is a member of the high-tech industry, respectively (B&R, p. 18). In the second stage, B&R regress their cost of capital measure COC on MFDiscPol and a set of control variables, where COC is calculated at the end of 2004 using the PEG method. They include the log of the market value of equity, the log of the book-to-market ratio and market beta to control for systematic risk. Consistent with the implications from Francis et al. (2008), B&R also control for earnings quality by including EarnVar, the standard deviation of earnings before extraordinary items scaled by total assets over the 1992-2004 period into their second-stage regressions.

B&R use a sample of 1’535 firms with a mean cost of equity capital of roughly 11% and a standard deviation of about half the mean. They include all nine instruments in the first-stage OLS regression without however adding the covariates EarnVar, market beta, market value and book-to-market ratio included in the second-stage regression. As pointed out by Angrist & Pischke (2008, p. 142), this variant of two-stage least squares procedure yields inconsistent estimates, which may lead to erroneous conclusions also with respect to their subsequent weak instruments specification tests4.


1 Proprietary information is defined as mainly non-financial information about a firm’s technology, strategy and operations (Evans & Sridhar, 2002; Rikanovic, 2005).

2 The AIMR has changed its name to CFA Institute. The disclosure ratings published by AIMR discontinued in 1997 (Core, 2001).

3 More recent studies have attempted to infer disclosure quality “indirectly” from the association between a firm’s operating cash flow and earnings (Cohen, 2008) or the quality of accounting information provided to outside investors by analyzing the properties of a firm’s reported earnings (Francis, Nanda & Olsson, 2008). These studies are briefly reviewed in Section 2.3.

4 Based on their spurious first-stage regression lacking the previously specified covariates, B&R find an R - squared of 13.2% and an F-statistic of 23.8. Comparing this figure with the corresponding benchmark for the necessary size of the F-statistic developed by Stock, Wright and Yogo (2002), they conclude that the weak instruments null hypothesis can be jointly rejected for their set of 9 instruments.


ISBN (eBook)
ISBN (Book)
File size
943 KB
Catalog Number
Institution / College
University of Zurich – Department of Banking and Finance
Voluntary disclosure quality Cost of Capital endogeneity bias Swiss capital market valuation models analyst following



Title: An Investigation of the Causal Effect of Voluntary Disclosure Quality on Cost of Equity Capital