Voluntary Disclosure and Market Competition: Theory and Evidence from The U.S. Services Sector

Transcription

1 Voluntary Disclosure and Market Competition: Theory and Evidence from The U.S. Services Sector G. Nathan Dong Eda Orhun April 20, 2016 Abstract This paper analyzes a firm s incentives to disclose private information about market demand and its cost when there is a potential entrant. There is a partially-pooling disclosure equilibrium in which high demand-high cost and low demand-high cost types are nontransparent in the case of risky debt issuance. We use a sample of U.S. service firms to test the theoretical predictions. Consistent with the model s implications, among low-debt service firms, those of high demand-high cost type are likely to avoid information disclosure, whereas among highdebt firms, those of high demand-high cost and low demand-high cost types are less likely to disclose private information. Keywords: Voluntary Disclosure; Market Competition; Analyst Forecast. JEL Classification: G14, L13, L8, M41. Columbia University, Dept. of Health, Policy and Management, 600 West 168th Street, New York, NY 10032, USA; gd2243@columbia.edu; Tel: (212) Zayed University, Khalifa City B, , Abu Dhabi, UAE; eda.orhun@zu.ac.ae. We would like to thank conference and seminar participants at 2014 AAA (Atlanta), 2013 FMA (Chicago), 2013 FMA European (Luxembourg), 2011 FMA Doctoral Student Consortium (Denver), 2011 AFFI (Paris), 2011 EFMA (Braga), 2011 WFC (Rhodes), Vienna Graduate School of Finance, and Vienna Graduate School of Economics, Zsuzsanna Fluck, Maarten Janssen, Michael Kirschenheiter, Jiro Kondo, Juliusz Radwanski, Leopold Sogner and Klaus Ritzberger, and Robert Webb for helpful discussions. 1

2 Intel cited weak demand from other computer makers for its cheeps, as well as softening prices... The Wall Street Journal, September 16th 1991 A Introduction In today s financial system firms accessing capital markets are required to follow mandatory disclosure rules set by regulatory institutions like the Securities and Exchange Commission (SEC). In addition to this, some firms but not all make voluntary disclosures such as management forecasts, press releases, etc. This action makes sense when one takes into consideration the informational asymmetry existing between managers and other stakeholders about value-relevant firm related information. Thus, these voluntary disclosures could be aiming at minimizing the lemons problem of Akerlof (1970) as much as possible. However, a puzzling empirical fact is that firms voluntary disclosure policies differ a lot in terms of the amount and types of disclosures supplied. While some firms are known for being transparent others are likely to be much more silent (Lang, 1998 and Watson et al., 2002). A challenging task for researchers is to determine and understand the underlying incentives behind these varying actions of voluntary disclosure. Several attempts have been made in different scenarios relating voluntary disclosure. This paper starts with studying the voluntary disclosure policy of an incumbent firm when there is a potential entrant into the market. The incumbent firm holds private information about the market demand and its own production cost. Both the market demand and incumbent s production cost can be high or low. The incumbent can choose a different disclosure policy depending on the combination of the demand and its own cost state. While a high (low) market demand increases (decreases) the entry probability, a low (high) cost incumbent discourages (encourages) entry. Here we will not model the possibility of the firm to lie about its type considering the fact of punishment if it is found out or available costless verification mechanism 2

3 which is consistent with the earlier literature. To understand this specific market situation, consider the case of Intel in 1980s before AMD entered the microprocessors business as a stand-alone entity and produced AM 386/40 first in In this regard, Intel, being a monopolist, kept control of all the connections in the computer producers industry and knew the exact demand of microchips from their side. In addition, Intel had the private information of its own production cost of a microchip. As expected, Intel had made several strategic product market related voluntary disclosures back then in the form of press releases before the entry of AMD into the market. In such a scenario, the analysis shows that there exists a unique fully revealing disclosure equilibrium in which every type of the incumbent, except the high demand-high cost type, is transparent. The high demand-high cost type is indifferent between a transparent and nontransparent disclosure policy in this equilibrium. The key point behind this result is that the high demandhigh cost type or the most favorable state from the viewpoint of the entrant is likely to be nontransparent in any equilibrium candidate. By committing to a nontransparent disclosure policy, the high demand-high cost type tries to reduce the entry probability. She would succeed in this attempt if any other type of incumbent, which is in a less favorable state for the entrant, also prefers to be nontransparent in equilibrium. However, it is not optimal for any other type of the incumbent to pool with the high demand-high cost type. Thus, all of the incumbent types disclose their private information in equilibrium due to the strong incentive of the high demand-high cost type to withhold information. This result is similar to the standard unravelling argument of Milgrom (1981) and Grossman (1981). Yet, the unique fully revealing equilibrium of the baseline model contradicts with the paradox that empirically some firms are transparent and others prefer to stay silent. Additional to being in contradiction with empirical observation, the statement that even the incumbent with the least prospects (low demand-high cost type) decides to reveal her type, is atypical. Particularly, this is very questionable if there exists a third party assessing the incumbent in terms of profitability like her financiers or the financial market. Taking 3

4 this effect into account, the baseline model is then extended to the case of simultaneous signalling to product and credit markets in which the incumbent issues risky debt. In such a scenario, since the riskier the incumbent appears from the viewpoint of the debtholders the more costly the debt issuance is, the low demand-high cost type would become very cautious to disclose her type. Additional to the earlier fully revealing equilibrium, this extended model yields an equilibrium in which high demand-high cost and low demand-high cost types do not disclose their types and hold the same level of debt whereas the other two types reveal their information. In this equilibrium, the low demand-high cost type prefers to pool with the high demand-high cost type, whose profits are higher than hers, because she can decrease her debt amount to be paid back. As before, the high demand-high cost type would like to camouflage due to the entrant s higher probability of entry. That is to say, these two types pool because the low demand-high cost type can deceive the debtholders and issue debt less costly and the high demand-high cost type succeeds in misleading the entrant about her true state. And, the other two types (low demand-low cost and high demand-low cost types) prefer to separate because pooling benefits them neither from the debt nor the entry side. The main contribution of this paper is to analyze the disclosure incentives of an incumbent firm when market demand (industry-wide) and cost (firm-specific) information co-exist. Although there has been extensive research about voluntary disclosure of demand or cost information in oligopoly markets, the simultaneous effect of demand and cost on voluntary disclosure has been neglected. Most models study the optimal disclosure policy of either demand or cost information in Cournot or Bertrand competition settings. Their results change depending on the assumed competition structure. However, the main result of the full-disclosure equilibrium of the present baseline model does not depend on any competition structure but is more general. Additionally, the previous models assume a binary decision of entry for the entrant. The entrant either enters or stays out depending on the disclosure decision of the incumbent. However, the current model assumes a random entry cost. The entrant enters into the market with some probability depending on her expected profit upon entry which is 4

5 contingent on the information disclosed. This also avoids the necessity of assuming fixed proprietary costs which is common in similar models. Earlier literature which is in close relation with the current paper is discussed more in detail in the next section. In addition, the baseline model is extended to consider the impact of the credit market on the voluntary disclosure decision which could not be ignored especially for firms with high amount of debt outstanding. For this reason, the paper analyzes the case of risky debt issuance by the incumbent where the debt level serves as an additional signal in the second part of the theory section. The extended model clearly illustrates that a second effect in addition to the entry effect exists in influencing the disclosure policy if the incumbent signals simultaneously to both parties. This second effect is the incumbent s own profitability which has an effect on her riskiness (cost of debt issuance) and in turn on her disclosure policy. With this new effect, incumbent types who were fully disclosing before get new incentives to pool with the high demand-high cost type. This is particularly true for the type with the lowest profits, the low demand-high cost type. This impact of (risky) debt issuance on disclosure policy in an entry game has not been much covered in the earlier literature. Last but not least, the findings of the model also hints some important public policy implications. The extended model demonstrates that there exists an equilibrium where the most favorable state from the viewpoint of the entrant (high demand-high cost) and the least profitable type (low demand-high cost) may pool together. As explained earlier, the reason for the existence of this equibrium is that the least profitable type benefits from the pooling by obtaining external financing less costly. In other words, the least profitable type prefers pooling with the black duck this time because she realizes that this is a way to deceive the debtholders and prevent their severe punishment. Obviously, this would affect the welfare of the investors negatively. In this regard, it could be advised that product market related voluntary disclosures should become instead mandatory if the policy-makers aim to maximize the welfare of the investors. 5

6 As highlighted earlier, our model provides some clear empirical predictions: Every incumbent type is expected to be transparent, except the high-demand/high-cost type, when signalling to the credit market is not in the picture which would be true for firms with no or low levels of debt. On the other hand, if signalling to the credit market arises due to existence of risky debt, partial disclosure in the product market where high-demand/high-cost and lowdemand/high-cost types pool is likely to occur. It is apparent that signalling to the credit market would be an important concern especially for firms with high levels of debt. In this respect, empirical section of the paper tests these predictions by using the panel data of U.S. services sector firms from which we generate two sub-samples with respect to the debt level of firms. The first sub-sample only includes firms with debt-to-assets ratio less than the 25 percentile value of the debt-to-assets ratio in each industry in each year whereas the second sub-sample includes firms whose debt-to-assets ratio lies between 50 and 100 percentile value of the debt-to-assets ratio in each industry in each year. We proxy for voluntary disclosure or transparency level of firms by using the dispersion of analysts forecasts for firm i in year t by exploiting the documented evidence of negative relationship between a firm s information disclosure policies and the dispersion among individual analyst forecasts (Lang and Lundholm 1996). The results of the regression analysis are in accordance with the model s predictions: The dispersion of analyst forecasts is highest for the high-demand/high-cost type which implies this type is less likely to reveal private information within the low-debt firms. Moreover, the low-demand/high-cost incumbent together with the high-demand/high-cost type is more likely to avoid voluntary disclosure within the sample of high-debt firms which is implied by statistically higher analyst forecast dispersion. The next section provides more detailed information regarding the earlier literature and restates the contributions of this paper. 6

7 A.1 Prior Research The earliest studies on voluntary disclosure are Grossman (1981) and Milgrom (1981). They analyze the disclosure policy when a firm is evaluated by the financial market. In this case, the only equilibrium is full disclosure. Afterwards, the optimal disclosure of market demand or cost information has been extensively examined beginning with Vives (1984) and Gal-or (1985, 1986). The main results are disclosure of (concealing) market demand is optimal with Bertrand (Cournot) competition. Conversely, disclosure of (concealing) cost information is optimal with Cournot (Bertrand) competition. Most related papers are Darrough and Stoughton (1990), Wagenhofer (1990), Hwang and Kirby (2000), Pae (2002) and Koessler (2008). Especially Darrough and Stoughton (1990) considers a similar problem by looking at the effect of voluntary disclosure on market entry and financial market valuation. However, there are some important differences between the current paper and Darrough and Stoughton (1990). Firstly, Darrough and Stoughton (1990) accounts only for one type of private information of the incumbent which they call favourable versus unfavourable information. They assume the private information of the incumbent to be favourable or unfavourable both for the entrant and the financial market at the same time. An example that would fit this criteria would be market demand information. However, private information on cost of production would not fit. Clearly, a high cost incumbent would mean unfavourable news for the financial market whereas favourable news for the incumbent. As Darrough and Stoughton (1990) state voluntary disclosure decision in this situation would be trivial if only cost information exists. However, voluntary disclosure decision becomes interesting when the incumbent holds private information both on market demand and production cost which is the main topic of this paper. Secondly, the results of Darrough and Stoughton (1990) depends on the cost of entry. Our model, on the other hand, assumes a random entry cost which generalises the setup and prevents the equilibrium results to depend on the fixed entry cost. Last but not least, debt level in our model works as an additional 7

8 signal together with the type of disclosure affecting the posterior beliefs of both the entrant and the debtholders. In other words, an incumbent type can separate itself from other types with its debt level even when she follows a nontransparent disclosure policy. In Darrough and Stoughton (1990), equity level does not function as an additional signal for the entrant and the financial market but it is only the type of disclosure affecting the posterior beliefs of the players as well as the stock price of the firm. Wagenhofer (1990) deals with a similar problem but considers continuous private (market demand) information of the incumbent which is more general than binary information structure of Darrough and Stoughton (1990). Hwang and Kirby (2000) study a game where firms in an oligopolistic market hold private information about their own cost levels and new entry to the market is also possible. Pae (2002) is the only paper which considers the disclosure of market demand and cost information simultaneously. He considers an incumbent firm which operates in the market as a monopolist and becomes privately informed about the market demand and its production cost in the first period. In the second period, a new uninformed firm enters the market. The incumbent prefers to fully disclose its private information in equilibrium like in the current model. The reason behind this result is the inter-temporal incentives of the incumbent. By disclosing all information, the incumbent nullifies its self-defeating inter-temporal incentives by which she tries to deceive the entrant about the true state. The present paper offers an alternative explanation for the full-disclosure equilibrium with a signalling game. As opposed to Pae (2002) who is restricted to the case of Cournot competition, the results of this paper hold in a more general market setup. Additionally, Pae (2002) assumes no entry cost for the entrant which implies the entrant always enters in the second period. This is also not the case in this model which assumes a random entry cost. Finally, Koessler (2008) questions a very similar topic from a different perspective: He compares public versus private information revelation of binary state to two different audiences with verifiability. The paper characterises the equilibrium depending on whether meetings take place publicly or privately. Koessler (2008) shows that when information revelation is made privately to each audience, one gets full disclosure whereas 8

9 when disclosure takes place publicly, no information is revealed by the sender. Our outcome may be interpreted as a specific example of their scenario by illustrating that less information revelation occurs by the incumbent with public communication to a potential competitor and to the credit market. This paper provides supporting evidence that in the presence of market imperfections the financing decisions of firms have real effects, a fundamental question originally proposed by Modigliani and Miller (1958). The paper advances the corporate financing literature on the interaction of product market decisions and the financing of firms initiated by Brander and Lewis (1986) and furthered by Gertner; Gibbons and Scharfstein (1988), Maksimovic (1988), Chevalier and Scharfstein (1996), Campello and Fluck (2006), among others by higlighting the real effect of capital structure on strategic voluntary disclosures related to product market. From these papers, Gertner; Gibbons and Scharfstein (1988) studies a firm s indirect information revelation to the product market through its capital structure, rather than direct and verifiable information disclosure. By using the prominent result of high-profit firms may separate from low-profit ones by issuing more debt from Myers and Majluf (1984), they illustrate that separating equilibrium is less likely to occur with simultaneous signalling to the capital and product markets. Their outcome is very much related and in accordance with the current model s implications. Pooling between high-demand/high-cost and low-demand/highcost types becomes likely when simultaneous effects of product and credit markets are taken into account. After having focused on the theoretical studies, we next review the empirical literature that tries to understand the effect of product market competition on voluntary disclosure. This literature has not yet agreed on a common answer but provides conflicting results regarding the relationship between competition and voluntary disclosure (Berger 2011). These conflicting results arise mostly due to data problems and their proxy measure for competition which is the industry concentration. Using industry concentration as the proxy for competition causes limitations due to missing data on private firms and unclarity whether concentrated 9

10 industries imply less or more competition. However, mostly accepted notion by the earlier literature is that highly concentrated industries are less competitive. Anyhow, while some studies like Harris (1998), Botosan and Stanford (2005), Bamber and Cheon (1998) find a negative relation between industry concentration and disclosure, others find either positive or statistically insignificant relations (e.g., Verrecchia and Weber (2006), Botosan and Harris (2000), Berger and Hann (2007), Bens, Berger, and Monahan (2011)). In addition, some recent studies like Ali, Klasa, and Yeung (2009) and Ali, Klasa, and Yeung (2014) prove that when missing data on private firms is included to obtain a more accurate measure of industry concentration, the results of Harris (1998) become statistically insignificant and the positive results of Verrecchia and Weber (2006) reverse. In short, this literature with many different answers is yet to be concluded. More closely related to our paper, Bhojraj, Blacconiere and D Souza (2004) study voluntary disclosure decisions of electric utility companies during the time period in when the industry is in transition to deregulation and competitive pricing. In such a scenario, there are three different audiences of voluntary disclosures which are regulators, capital market participants and product market competitors. Overall, they show that voluntary disclosure decisions are generally consistent with predicted incentives to affect the three important audiences. They show that (positive) voluntary disclosures decrease with the magnitude of utilities stranded costs where the stranded cost recovery issue is unresolved. This result is consistent with the incentive to prevent the regulators getting less inclined to allow full recovery of stranded costs after hearing the positive news about the company. Moreover, Bhojraj et al. (2004) find that while voluntary disclosures increase with capital market incentives (higher number of institutional investors and higher percentage of institutional holdings in the firm s equity), they decrease with product market competition incentives (future market demand). Incumbent firms that are in high demand markets would not make voluntary disclosures to deter new entrants especially after deregulation as it is also confirmed by our paper. In a similar fashion, Burks et al. (2013) ask the question how the voluntary disclosure decisions of 10

11 incumbent banks would change after the relaxation of interstate branching regulations under the Interstate Banking and Branching Efficiency Act (IBBEA). They provide evidence that the tone of press releases becomes less positive and more negative after entry barriers are abolished which is again in accordance with the entry deterrence motive. Thus, these two papers empirically confirms the entry deterrence motive. Our paper will add to these papers by looking at voluntary disclosure decisions of (low versus high-debt) firms with respect to the combination of their demand and cost states. To our best knowledge there is no earlier study looking at this question. A.2 The Model Assumptions Before going into the details of the theoretical model and its solution, this section will discuss some assumptions that we adopt and why these assumptions are not changing the main essence and the results of the study. The first assumption to be discussed why the model only considers either true revelation of the state or no disclosure by the incumbent. The alternative way of looking at the disclosure decision could have been the incumbent discloses nothing, only its cost state, only its demand state or both states. Adopting this way of disclosure decision would not change the uniqueness of the fully revealing disclosure equilibrium in the baseline model. The reasoning is as follows: L D, L C type would reveal its type completely as before whereas H D, L C type would reveal only its cost state. However, since L D, L C type has already revealed itself, the identity of H D, L C would become also unveiled. Similarly, L D, H C type would reveal only its demand state but again its identity would become also unveiled due to full revelation of L D, L C type. Finally, the identity of H D, H C type would be also detected independent of its disclosure policy since all the other types are already revealed. Another assumption of the model is having only two levels of market demand and incumbent s cost for simplicity. But the main result of the paper would go through if instead a continuous uncertainty space with respect to both market demand and cost is studied. The model would yield again a unique full disclosure equilibrium if the incumbent does not need any external financing since 11

12 there would exist only one state which is the best from the viewpoint of the entrant. In this state, there would be the highest market demand and the incumbent would have the highest cost. Again analogously, the fully revealing disclosure equilibrium would stop to be unique once the effect of signalling to the credit market is considered with risky debt issuance. The reason is that the incumbent s own profitability would create again a second force that may give incentives to other types to become nontransparent. The incumbent type with the least prospects, which would have the highest cost and face the lowest market demand, would get the strongest incentive to become nontransparent as before. Last but not least, the extended model with risky debt issuance assumes that the incumbent defaults if the entrant is in the market. This assumption could be easily relaxed such that while some incumbent types are always sound; some default upon entry. But since the ones who default first would be the least profitable ones, the main result of the model would still go through. B Theoretical Analysis B.1 The Model The model is a Bayesian game with signalling. There are two players. The first player is an incumbent firm which holds a monopoly position in a product market. The second player is an entrant firm which is willing to enter this market. In the beginning, nature chooses the type of the incumbent. There are four possible types of the incumbent that correspond to combinations of two different demand and cost levels. After observing her type, the incumbent determines her disclosure policy which can be disclosing the state (both cost and demand state) or witholding all information. The profit of the incumbent is affected by her disclosure policy through the potential entrant s decision. While a high (low) market demand increases (decreases) the entry probability, a low (high) cost incumbent discourages 12

13 (encourages) entry. The entrant disburses a random cost of K to enter the market with continuous cumulative distribution function F (K). This means that the entrant always enters into the market with some probability. An important assumption in this setting is that the entrant becomes informed about the state once she enters the market even when the incumbent has chosen non-transparency. Summarizing the timeline of events: First, nature chooses the incumbent s type. Then the incumbent determines her disclosure policy which can be either disclosing the state or withholding all information. Thus, disclosure policy of the incumbent is a binary decision. The variable t {0, 1} indicates the disclosure decision of the incumbent, with t = 1 standing for full disclosure. Following the incumbent s move, the entrant learns his entry cost drawn from F (K) and decides whether to enter or not. Finally, payoffs realize in the product market. Figure 1 summarizes the timeline of events. [Insert Figure 1 Here] The states or the types of the incumbent are denoted by ω where ω {1, 2, 3, 4}. Each state, from 1 to 4, occurs with probability P (ω) which are common knowledge and 4 ω=1 P (ω) = 1. Type 1 corresponds to the high demand-low cost state (H D, L C ) whereas type 2 corresponds to the high demand-high cost state (H D, H C ). Similarly, type 3 refers to the low demand-low cost state (L D, L C ), whereas type 4 corresponds to the low demand-high cost state (L D, H C ). The profit of the incumbent in state ω is denoted π I (ω) when the entrant enters and Π(ω), when she does not. Here, Π(ω) refers to the monopoly profit in state ω. The profit of the entrant in state ω is denoted π E (ω), when she is in the market. It is clear that the monopoly profit is larger than the duopoly profit at ω, Π(ω) > π I (ω). Additionally, it is possible to order monopoly (duopoly) profits in different states. The best possible state for the incumbent is the high demand-low cost state ω = 1 whereas, the worst possible state is low demand-high cost, ω = 4. Thus, the monopoly and duopoly profits of the incumbent can be ordered as π I (1) π I (2) π I (4) (Π(1) Π(2) 13

14 Π(4)) and π I (1) π I (3) π I (4) (Π(1) Π(3) Π(4)). However, it is not possible to make a comparison between π I (2) and π I (3) (Π(2) and Π(3)). In a similar fashion, the profits of the entrant can be ordered as π E (2) π E (1) π E (3) and π E (2) π E (4) π E (3). The entrant benefits from a high demand state but suffers from a low-cost competitor. Thus, one cannot compare the profits of π E (1) and π E (4) like in the case of the incumbent with π I (2) and π I (3). The complete ordering of the incumbent s and the entrant s profits depends on the marginal effects of cost and demand. Both players are risk neutral and maximize their expected profits. That is to say, the incumbent decides on her disclosure policy to maximize her expected payoff and the entrant determines her entry probability such that she enters the market only if she obtains positive payoff. Last but not least, an important assumption in the model is truthful and costless disclosure. It is reasonable to assume that firms truthfully disclose, especially if they are subject to auditing. B.2 Strategic Analysis Disclosure Decision Backwards induction is used in order to determine the Perfect Bayesian-Nash (PBE) equilibria of the game. First, the entry decision of the entrant in terms of probabilities is determined. The entrant knows the type of the incumbent with probability one when she encounters a transparent incumbent. In this situation, she enters into the market with probability F (π E (ω)) where ω {1, 2, 3, 4}. Hence, she enters only when her cost of entry is less than her profit. On the other hand, the entrant forms posterior beliefs according to Bayes rule at each information set where she encounters a non-transparent incumbent. The posterior beliefs are denoted by q(ω t = 0) where t = 0 indicates a nontransparent disclosure policy. Now, the entrant enters the market with probability F ( 4 ω=1 q(ω t = 0)π E(ω)). Accordingly, she takes into 14

15 consideration her expected payoff after seeing a non-transparent incumbent and enters the market when she obtains expected positive profit. Now turn to the incumbent s disclosure decision. Intuitively, the disclosure decision of the incumbent depends mainly on the entrant s decision. When the entrant is more likely to enter, the incumbent is more likely to conceal information. This fact creates strong incentive for the high demand-high cost type (ω = 2) to choose a nontransparent policy. This is because the entrant can obtain the highest possible profit when the incumbent is of type 2. On the other hand, the low demand-low cost type (ω = 3) is the least likely type to prefer a nontransparent policy since it is not a profitable state for the entrant. This intuition is formalized in the following proposition. Proposition 1 There exist a PBE where all types are transparent. In this equilibrium, only type 2 is indifferent between a transparent and nontransparent disclosure policy. The consistent equilibrium belief is q(2 t = 0) = 1. Proposition 1 shows the existence of a full disclosure equilibrium. Furthermore, the next proposition illustrates that this is the unique equilibrium of the game in pure strategies. The insight behind this result can be summarized as follows: Revealing her type is a dominant strategy for the low demand-low cost type (ω = 3) since it is the least profitable state for the entrant. Conversely, the high demand-high cost type (ω = 2) would like to withold the information especially when the other types are also nontransparent. By committing to a nontransparent disclosure policy, she tries to discourage entry since she is in the most promising state for the entrant. Anticipating this, the other types avoid to be pooled with the high demand-high cost type by disclosing their types. They prefer disclosure since getting pooled with the high demand-high cost type would lead to an increase in the entry probability compared to when they are transparent. Proposition 2 The equilibrium of Proposition 1 is the unique equilibrium in pure strategies. 15

16 Up to now, the solution of the disclosure policy game has offered only one answer. According to the model, all types of firms prefer to be transparent. However, casual observation suggests that a fully seperating equilibrium is not very common. This result reinforces the paradox that empirically, while some firms are transparent, others prefer to stay silent. This unrealistic aspect leads to questioning the assumptions of the model. In view of that, a firm is not solely evaluated by its competitors but usually there exists a third party assessing the incumbent in terms of profitability like her financiers or the financial market. If such a third party existed, the incumbent with the least prospects (the low demand-high cost type) would think twice before disclosing her type. This is due to the fact that the debtholders become claimants to the profit of the incumbent owing to the limited liability in case of a default. This implies that the riskier the incumbent is from the viewpoint of the debtholders the more costly the debt issuance is. This effect can prevent the low demand-high cost type from being transparent and provide incentives to pool with the high demand-high cost type who is less riskier from the viewpoint of debtholders. Taking this effect into account, the next section considers the case of risky debt issuance by the incumbent. B.3 The Model with Risky Debt Issuance The basic idea of the model is the same as before except that the incumbent needs to raise external financing denoted I in the form of standard debt with face value of D in order to start production. In such a situation of borrowing by the incumbent, consider the case that the debt is riskless. Since the incumbent is prosperous and can pay back her debt in this scenario, the debtholders are not concerned about profitability of the incumbent. That is also why the incumbent does not consider the debtholders when deciding about her disclosure policy. One can easily show that the fully revealing equilibrium of the original model continues to be unique when the incumbent issues riskless debt. But this result changes when debt becomes risky. The debtholders then become concerned about profitability of the incumbent because in some states the incumbent can default and instead of the face value of debt they obtain 16

17 incumbent s profit due to limited liability. In this respect, the debtholders would ask a higher face value of debt from an incumbent who is riskier or obtains lower profits. In return, the incumbent would take into consideration the debtholders attitude when determining her disclosure policy. The incumbent type who would be most concerned about this, is the one with the least prospects or the low demand-high cost type. In this regard, it is assumed that the incumbent defaults when she does not obtain the monopoly profit (Π(ω)) but the duopoly profit (π I (ω)). In other words, the incumbent defaults when the entrant enters the market and debtholders get the duopoly profit, π I (ω) after such a default. 1 To get such a default scheme, the external financing parameter I should lie in some interval such that min Π(ω) = Π(4) > I > max π I (ω) = π I (1) 2. The timeline of events changes as follows: First, nature chooses the incumbent s type as before and then incumbent becomes informed. Then again, the incumbent determines her disclosure policy which still has the same structure. Now, after deciding for disclosure policy, the incumbent offers a debt level of D in return for her borrowing I to the investors. Investors can say yes or no to D. This offered debt level D can also be observed by the entrant. Thus, debt level D serves as an additional common signal together with the disclosure policy, t {0, 1}. Following the incumbent s moves, the entrant learns her entry cost drawn from F (K) and decides whether to enter or not as in the original model. Finally, payoffs realize in the product market. Figure 2 summarizes the new timeline of events. [Insert Figure 2 Here] The property that the entrant can also observe the debt level D is natural when the debt in question is public debt. Thus, one can regard the debt here as a corporate bond. Debt Level 1 This assumption could be relaxed easily as discussed in the model assumptions section. 2 The first inequality should be satisfied such that I is sufficiently smaller than the left-hand side. The precise threshold is given by the participation constraint of the debtholders F (π E (ω))π I (ω) + (1 F (π E (ω))) min(d, Π(ω)) I from which I Π(ω) 1 υ where υ = 1 F (π E(ω)) π I (ω) I 1 F (π E (ω)) > 1 17

18 The debt level serves as an additional signal given that the incumbent preferred nontransparency in the earlier stage. Hence, the entrant and the debtholders form posterior beliefs according to Bayes rule after observing both the transparency policy and the offered debt level D. Define behavior strategies ξ ω and σ ω (D) for disclosure policy and choice of debt level, respectively. The posterior beliefs are equal to: q(ω t = 0, D) = (1 ξω)σω(d)p (ω). The P (t=0,d) debt level plays no signalling role after transparency, since the type of the incumbent is then revealed. After observing a transparent incumbent, the debtholders accept the debt offer D in state ω if, F (π E (ω))π I (ω) + (1 F (π E (ω)))d I. (1) This equation is the participation constraint of the debtholders and the incumbent would offer the minimum debt level D pushing the debtholders to the break even point. Thus, the face value of debt to be offered by a transparent incumbent in state ω is D(t = 1) = I F (π E(ω))π I (ω). (2) 1 F (π E (ω)) On the other hand, the participation constraint of the debtholders after observing a nontransparent incumbent reads as follows, F ( [ 4 4 ] [ q(ω t = 0, D)π E (ω)) q(ω t = 0, D)π I (ω) + 1 F ( ω=1 ω=1 ] 4 q(ω t = 0, D)π E (ω)) D I. Here, F ( 4 ω=1 q(ω t = 0, D)π E(ω)) is the entry probability as in the original model. Given that the entrant is in the market, the incumbent defaults and debtholders receive the expected duopoly profit of the incumbent. Debtholders should at least break even in expectation given the posterior beliefs. Then, the face value of debt to be offered by a nontransparent incumbent ω=1 (3) 18

19 is given by D(t = 0) = I F ( 4 ω=1 q(ω t = 0, D)π E(ω)) [ 4 ω=1 q(ω t = 0, D)π I(ω) ] [ 1 F ( 4 ω=1 q(ω t = 0, D)π E(ω)) ] (4) by turning the inequality sign in (12) into an equality. Equilibria This section shows that fully revealing disclosure equilibrium ceases to be unique when there is risky debt issuance. Risky debt creates a second effect additional to the original entry effect. The second effect is the incumbent s profitability. Without risky debt all incumbent types, except the type in the most profitable state for the entrant (the high demand-high cost type), preferred full disclosure. They avoided being pooled with the high demand-high cost type in order to minimize the entry probability. And, there was no device to assess the incumbent s profitability. This changes once the incumbent needs to borrow from a third party. Then, especially the incumbent type with the least expected profitability (low demand-high cost) would sometimes like to be nontransparent and camouflage herself in order to decrease the punishment by the debtholders, even at the cost of being pooled with the high demand-high cost type. In this respect, the next pages show the existence of an equilibrium in which the high demand-high cost and low demand-high cost types do not disclose their types and hold the same level of debt whereas the other two types reveal their information. This constitutes an additional equilibrium to the fully revealing disclosure equilibrium. Proposition 3 states the condition that the duopoly profit of the high demand-high cost type, π I (2) needs to satisfy for the existence of a fully revealing disclosure equilibrium. This condition is necessary to ensure that the low demand-low cost (ω = 3) and low demand-high cost (ω = 4) types do not obtain any benefits from being nontransparent. In other words, π I (2) should be small enough so that the gains of these two types due to obtainment of 19

20 debt less costly by pooling with the more profitable high demand-high cost (ω = 2) type are smaller than the loss due to more likely presence of the entrant. If the profit ordering of the incumbent is given such that π I (1) π I (3) π I (2) π I (4), then the low demand-low cost (ω = 3) type would have no gain by pooling, since she becomes more profitable than the high demand-high cost (ω = 2) type. In this case, the condition on π I (2) stemming from the incentive compatibility of the low demand-low cost (ω = 3) type becomes redundant. Proposition 3 There exists a PBE where all types are transparent and each holds a different level of debt which is given by equation 2 if, ([ ] [ π I (2) min 1 F (π E(3)) F (π E Π(3) + F (π E(3)) π (2)) F (π E (2)) I(3), 1 F (π E(4)) F (π E (2)) ] ) Π(4) + F (π E(4)) π F (π E (2)) I(4). Only type 2 is indifferent between a transparent and nontransparent disclosure policy. The consistent equilibrium belief is q(2 t = 0, I F (π E(2))π I (2) 1 F (π E ) = 1 for the equilibrium debt level of type (2)) 2. The next proposition determines the certain range in which the high demand-high cost type s profit (π I (2)) should be situated for the existence of a PBE where the high demandhigh cost (ω = 2) and the low demand-high cost (ω = 4) types pool. The conditions on π I (2) derive from incentive compatibility (IC) constraints of each type which take into consideration both debt and entry effects. For instance, IC constraint of type 4 imposes the following: π I (2) should be large enough that pooling is optimal for type 4 because her gain due to obtainment of debt less costly is larger than the loss due to higher entry likelihood. Depending on the complete profit ordering, the IC constraints of some types (1 and 3) may become redundant. Proposition 4 There exists a PBE where types 2 and 4 are nontransparent and offer the same level of debt, D, which is determined by equation (4) and equilibrium beliefs, given that π I (2) lies in a specific range determined as follows. The other two types are transparent and their debt level are given by equation (2). Equilibrium beliefs satisfy q(2 t = 0, D) + q(4 t = 0, D) = 1. 20

21 C Empirical Tests In this section we use regressions to test the model. Because the voluntary disclosure model is closely related to the study of market information intermediation, we interpret our empirical analysis as a step toward testing implications from the general class of market information production and acquisition models, as called for by Healy and Palepu (2001). 3 We begin by stating the empirical implications. C.1 Empirical Implications The best possible test of our model would compare the behavior of firms voluntarily disclosing private information to both the demands of their products and services and the costs of labor and materials. This comparison would allow one to know whether firms take the same actions to reveal their private information as others when they are facing different levels of demand for their products and services and cost of inputs. Unfortunately, we are not aware of a comprehensive data source containing all firms decisions to voluntarily disclose private information beyond what is required by the regulator. Prior studies have attempted to proxy voluntary disclosure by using survey rankings such as the AIMR scores (e.g., Brown and Hillegeist 2007, Lang and Lundholm 1993, 1996, Welker 1995, Healy et al. 1999, and Sengupta 1998), research-constructed indices of disclosure (e.g., Botosan 1997, Francis et al and Shalev 2009), measures from natural language processing technologies (e.g., Li 2008, 2010), and properties of the firms reported earnings (e.g., Dechow and Dichev 2002 and Francis et al. 2004). However, as Heitzman et al. (2010) and Beyer et al. (2010) point out, these measures capture both voluntary and mandatory disclosures; therefore, interpreting these measures as representing purely voluntary disclosures is problematic. Realistically, the use of information quality proxy always depends on the underlying characteristics and determinants of these; 3 Also see the discussion of empirical evidence in Core (2001). 21

22 hence, any proxy variable will not necessarily solve this problem, and again points to the need for an economic definition of earnings/information quality and direct derivation of measures from that definition. 4 In this paper, rather than relying on subjective measures of information quality (survey scores, discretionary choice of index members, word tone dictionary, etc.), we test the model s implications by observing whether there is a general agreement among analysts about the firm s future earnings prospects as reflected in the standard deviation of annual earnings forecasts. The empirical evidence of the negative relationship between a firm s information disclosure policies and the dispersion among individual analyst forecasts has been documented by, among others, Lang and Lundholm (1996). Given this relation, a reasonable assumption is that analyst forecast dispersion for a firm is larger when the firm chooses not to disclose private information, and vice versa. Therefore, the test of our model is to examine whether firms of different types (high demand-high cost, high demand-low-cost, low demand-low cost, or low demand-high cost) is associated with larger or smaller dispersion of analyst forecasts of annual earnings. Proposition 1 derives the Perfect Bayesian-Nash equilibrium that all types of firms have incentives to disclose their private information except the type 2 firm (high demand-high cost) which is indifferent to the voluntary disclosure policy. Thus, the model implies, for example, that the dispersion among analyst forecasts of annual earnings for a firm of either high demand-low cost, low demand-low cost, or low demand-high cost type will not depend on its disclosure practices, whereas it might be the case for a firm of high demand-high cost type. By examining the dispersions of analyst forecasts for different firm types, we might find some significant dispersion of earnings forecasts among firms of high demand-high cost type, but not among firms of other types. Similarly, Proposition 4 derives the Perfect Bayesian- Nash equilibrium that when debt is used to finance the business, type 1 (high demand-low 4 Also see the discussion of these issues, among others, in more detail in Core (2001) and Dechow et al. (2010). 22

23 cost) and type 3 (low demand-low cost) firms will disclose private information, whereas type 2 (high demand-high cost) and type 4 (low demand-high cost) firms will not. We summarize the hypotheses as follows: HYPOTHESES: All else equal, larger dispersion of analyst earnings forecasts could be observed for firms with 1. low levels of debt that are associated with high demand-high cost type. 2. high levels of debt that are associated with high demand-high cost and low demand-high cost types. C.2 Sample Data Implications from the theoretical model are tested with a collection of analyst forecasts of annual earning from the I/B/E/S database. The sample of earnings forecasts ranges from 1982 to We study the relationship between voluntary disclosure of private information and firm type by examining whether the dispersion of analysts forecast in a given industry is associated with the firms production and labor costs. Implicitly, we assume an efficient market of information production and acquisition in the sense that there will be no disagreement among financial analysts if the firm they are covering discloses all private information. The I/B/E/S (Institutional Brokers Estimate System) database contains earnings estimates for U.S. companies from 1976, and I/B/E/S History covers historical estimate records on over 45,000 companies across 70 markets. There are two versions of the I/B/E/S earnings estimate history (Summary and Detail), and in this paper we focus on the annual earning estimates among firms in the service sector in the Summary History dataset. The challenge in studying firms strategic decisions here is identifying and measuring the costs of production and labor. To do so, it is helpful to focus on a single sector, rather than comparing manufacturers to business service providers. In other words, we can better control for the impact 23

24 of sector characteristics on the perceptions of strategic issues (Yasai and Ardekani 1986). In addition, having a narrowed focus on a single sector can ensure a high level of internal validity. For example, Burks et al. (2013) examine the empirical relationship between market competition and voluntary disclosure in the banking industry; however their research focus is on identifying the causal effect rather than disentangling the underlying economic factors contributing to this relationship. We define the service firms in the U.S. based on the 49 Industry Portfolios on Kenneth French s web site, and list all industries descriptions of the service sector in Table 1. 5 [Insert Table 1 Here] C.3 Empirical Specification The dependent variable is the dispersion of analysts forecasts for firm i in year t. We use the standard deviation of annual earning estimates to proxy for the forecasts dispersion. We standardize the dispersion measure by dividing the value of standard deviation by the mean of all estimates for firm i in year t. The explanatory variables are described next. At the end of this section, we discuss regressions that correct the bias caused by the residuals that may be correlated across industries, and thereby verify that the correlation within an industry cluster does not distort the coefficient estimates of our OLS regressions. The model assumes two factors that can influence a firm s decision to voluntarily disclose its private information: cost and demand. We estimate a firm s cost as the sum of its Cost of Goods Sold (COGS) and Selling, General and Administrative Expenses (SG&A), both available from the firm s annual financial statements. In order to make the cost measure comparable across companies, we standardize it by dividing it by the firm s total revenue (sales) in that year. Similarly, we take the revenue divided by total assets to proxy for the demand of the firm s products and services. In addition, within each industry and year, we 5 Retrieved from Library/changes ind.html 24