Misreporting Corporate Performance

Transcription

1 ast revision: January 23 Misreporting Corporate Perormance ucian Arye Bebchuk arvard aw School and NBER Oren Bar-Gill arvard Society o Fellows (bargill@law.harvard.edu We are grateul to Allen Ferrell, Jesse Fried and Oliver art or their help and to the John M. Olin Center or aw, Economics, and Business at arvard aw School or its inancial support.

2 Misreporting Corporate Perormance Abstract This paper develops a model o the causes and consequences o misreporting o corporate perormance. Misreporting in our model covers all actions, whether legal or illegal, that enable managers o irms with low longterm value to make statements that mimic the statements made by irms with high long-term value. We model why and when managers o public companies will choose to misreport and to invest in creating opportunities or misreporting. Even managers who cannot sell their shares in the short run might misreport in order to improve the terms upon which the company would be able to issue equity to inance new projects or stock acquisitions. When managers are ree to sell some o their holdings in the short-run, incentives to misreport and the incidence o misreporting increase to an extent depending on the raction o their holdings that managers may sell and on their ability to sell without the market knowing about it. Investments in misreporting have real economic costs and lead to distortions in capital raising decisions, with irms that misreport raising too much equity and irms that do not misreport raising too little. ax accounting and legal rules increase investments in opportunities to misreport and the incidence o misreporting and, as a result, reduce ex ante share value. Our analysis provides a range o testable predictions concerning the periods, industries, and type o irms in which misreporting is likely to occur. The analysis also has implications or corporate governance and executive compensation. JE classiication: G3, K22, M4 Keywords: Asymmetric inormation, acquisitions, corporate governance, disclosure, myopia, short-termism, executive compensation, stock options, insider trading, accounting, inancial statements, earnings management, auditor, inancial reporting.

3 1. Introduction Recent events have directed much public attention to the reporting made to investors and the market by publicly traded companies. While the number o earnings restatements by publicly traded irms was on average 49 per year rom 199 to 1997, such restatements numbered 92 in 1997, 15 in 1999, and 225 in 21. From January 1997 through June 22, about 1 percent o all listed companies announced at least one restatement. 1 This paper investigates the causes and consequences o corporate misreporting. We deine misreporting as any action that enables low-value companies to report the same earnings, or revenues, or other dimensions o corporate perormance, as high-value companies report. This category includes, o course, cases in which companies get auditors to approve statements that are inconsistent with accounting standards. But it also includes cases in which companies with low long-term prospects take actions that are completely legal, being within accepted accounting and legal standards, to be able to make statements about the company s current perormance that make it impossible or investors to distinguish the company rom companies with high long-term prospects. We develop a model o why and when misreporting occurs and the distortions associated with misreporting. The model has positive implications concerning the circumstances in which misreporting is more and less likely to occur. The model also has implications or corporate governance. It identiies the eiciency costs o misreporting and how these costs depend on corporate governance arrangements. Why and when would managers take costly actions to be able to engage in misreporting i long-term cash lows are eventually going to be revealed so that the market cannot be ultimately ooled? To investigate this phenomenon, we study a ive-stage model. First, in the ex ante stage, managers decide how much to invest in creating opportunities to engage in short-term misreporting in the event that they subsequently learn that the irm is o a low-value type. Second, there is a stage in which managers learn inormation that suggests whether the expected long-term value o the irm s current projects is low or high. In this stage, managers are required (or at least may choose to make statements to the market about short-term perormance. Managers o highvalue irms are always able to report positive short-term results. Managers o low-value irms might sometimes be able to make such statements, say by 1 See Moriarty and ivingston (21, General Accounting Oice (22. 1

4 engaging in earnings management or creative accounting, with the probability o their being in a position to do so depending on their earlier investments in creating opportunities to misreport. We reer to the making o such statements by low-value irms as misreporting even when there is nothing illegal or alse in the statements made. Third, the learning o inormation and disclosure decisions stage is ollowed by market trading in the irm s shares. Fourth, the irm operates and might have an opportunity to engage in a new project or acquisition that would require raising additional capital. In the ith and inal stage, payos rom the irm s initial projects, and whatever additional project i any was added in the ourth stage, are realized and distributed. Our model is one in which markets engage in rational pricing. Investors do not know what managers private inormation is, but they are aware o the possibility that managers will misreport, and they draw rational inerences rom whatever they know o the managers actions as to what the managers likely inormation is. We start with the case in which managers may not sell their own shares in the intermediate trading stage and must keep all o their shares until the inal period. Even in this case, where managers must keep their shares until values are ully realized, and in which managers interests are aligned with those o long-term shareholders, managers might have an incentive to invest in creating opportunities to misreport and to misreporting when they can do so. When an opportunity to raise equity to inance an acquisition or a new project arises, misreporting enables the managers who learn that the irm s current projects are o low value to pool themselves with irms with existing projects o high-value and in this way improve the terms upon which they would be able to raise capital. The initial shareholders (including the managers o irms with existing projects whose value is lower than estimated by the market will be made better o i they will be able to obtain new capital on avorable terms, i.e., or a price exceeding what the outside investors are getting and the initial shareholders are giving up. This motive or misreporting might be especially important in circumstances in which irms are engaged in a series o stock-inance acquisitions. Whereas misreporting might take place even when managers must keep their shares until the inal realization period, the beneits to managers rom misreporting, and the incidence o misreporting, increase in the case in which managers may sell some o their shares in the intermediate trading period. When managers who learn that their projects have a low expected value can 2

5 and do misreport, the misreporting would enable them to gain by selling shares during the intermediate period. The selling o shares by managers o Enron enabled them to get out with substantial value beore the market learned that the irm s value was substantially lower than expected. The extent to which managers would gain rom misreporting, and thus the increased incentive to do so, would depend on the raction o the managers shares that they are permitted to sell. Thus, the greater the reedom o managers to sell their holdings in the short-run, the greater the incentive to misreport. The extent to which managerial short-run selling can increase their gains rom misreporting depends also on whether the market can tell whether such selling is taking place. The market s ability to identiy managerial selling in turn depends on whether and how requently sales must be disclosed, the existence o trading windows, and the amount o shares managers might seek to sell relative to the ordinary volume o trading arising rom liquidity selling. When the market knows or suspects that the managers are selling, the market price will decline to relect an increased probability that the irm s managers have been misreporting; this would reduce the proits that misreporting managers would be able to make by selling their shares in the intermediate trading period. owever, as long as the market would not be able to tell whether known or suspected sales by managers are due to managers knowledge o negative inormation or due to managers liquidity needs, allowing managers to sell shares in the intermediate trading period will increase their gain rom misreporting and their incentive to invest in creating opportunities to misreport. One eature o our model is that the incidence o misreporting is endogenously determined and is a unction o ex ante investments (such as in how operations and inancial structures are set. These investments are inluenced by various parameters o the irm, the industry, and the managers compensation package, which aect the potential beneits rom misreporting. These investments are also inluenced by the technology o misreporting, which is a unction o legal and accounting rules (and the implementing institutional inrastructure that are in place. Our model has signiicant descriptive implications. The model can help shed light on the increase in misreporting in the second hal o the 9 s. It identiies several actors that might have played a role in this increase. To begin, consider the many companies whose value was based primarily on uture growth opportunities. For these companies, the dierence between 3

6 good and bad inormation about current operations (or example, about current revenues, which has little direct eect on the irm s value, could still lead to a large dierence in the market s estimate o the irm s uture growth and business opportunities and thus o the expected uture value o the irm s projects. Furthermore, the 9 s were a period in which managers opportunities to sell large amounts o shares in the short-run was substantial. Speciically, the use o large stock option grants in compensation schemes became more common, and in the case o many high-tech irms, managers had some signiicant pre-ipo holdings which they could sell to the market. Finally, because o reductions in the potential liability o auditors and the structure o auditor services, auditors became more likely to acquiesce. As a result, the likelihood that a irm was able to misreport or any given level o prior investment in creating opportunities to misreport -- went up. 2 Our analysis provides a wide range o testable predictions concerning the circumstances in terms o the period, the industry, the irm, and the managers payos -- in which misreporting is more and less likely to occur. Throughout, we identiy relationships between the likelihood o misreporting and such parameters. Some o these relationships are consistent with already existing evidence, and others could be tested by uture empirical work Our model also has corporate governance and policy implications. The analysis highlights that the phenomenon o misreporting does not have only distributive consequences but also gives rise to potentially signiicant eiciency costs. One type o cost arises rom deadweight investments in creating opportunities to misreport and pool one s irm with irms o higher value. Another important cost arises rom distorting the allocative role o capital markets. When some low value irms can misreport and thereby pool themselves with high-value irms, the inancing and investment decisions o both types o irms will be distorted. In the pooling equilibrium caused by misreporting, high-value irms will be cross-subsidizing those low-value irms whose managers will misreport. Because o this compelled crosssubsidization, high-value irms might orgo some eicient projects to avoid the need to raise capital, whereas some low-value irms that misreport might 2 Coee (22 provides a compelling account o the legal and industry developments that led auditors to be more likely to go along with managers (misreporting decisions. 4

7 raise equity even when they do not have eicient projects. 3 As a result, there will be under-investment by irms that do not misreport and over-investment by irms that do. The existence o such real economic costs can be expected to reduce ex ante share value. Thus, arrangements that encourage misreporting can have a negative eect on share value. In our model, lax rules, which make it more likely that a irm will be able to misreport given its investment in creating opportunities to do so, are shown to increase the level o such investments and the incidence o misreporting. An important role o rules and institutions in our context is not simply to penalize some instances o misreporting ex post but also to aect the ex ante set o misreporting opportunities. Our analysis supports the calls by participants in public policy discussions (see, e.g., evitt (1998 or rigorous accounting standards and implementing institutions, which can narrow irms degrees o reedom to engage in misreporting. Our analysis also shows how the incidence o misreporting and its associated costs can be inluenced by the design o corporate governance arrangements. We ind that any increase in the raction o managers holdings which they may sell beore the inal period will increase incentives to invest in creating opportunities to misreport and the incidence o misreporting. We also ind that arrangements that eliminate or reduce the ability o managers to sell without the market s knowledge such as the recent legislative mandate that all trading by managers be very promptly disclosed can reduce but not eliminate the adverse incentives created by managers reedom to unload holdings in the short-run. It is worth noting, however, that the incentives to invest in misreporting would decrease, but not be ully eliminated by arrangements that require or encourage managers to keep their shares or the long haul. Thus, while such arrangements might be helpul in addressing the problem o misreporting, their adoption would not make the problem (and thus the accounting and legal measures that can constrain it irrelevant. Our analysis is organized as ollows. Section 2 presents our ramework o analysis. Section 3 analyzes reporting and investment decisions in the case in which managers must keep their shares until the inal realization period. Section 4 analyzes these decisions in the case in which managers are permitted to sell at least some o their shares beore the inal period. Section 5 discusses related literature, and Section 6 concludes. 3 As will be discussed below (see remark 2 ollowing proposition 3, these results are consistent with recent empirical evidence documenting the real eects o earnings manipulations (Polk & Sapienza (22. 5

8 2. Framework o Analysis 2.1 Sequence o Events The sequence o events in the model is as ollows: T=: Initial situation with (initially identical publicly traded irms each run by a manager. T=1: Managers may invest in creating opportunities or uture misreporting o corporate perormance. T=2: earning o inormation and disclosure decisions managers learn inormation pertaining to the companies expected inal cash lows and make disclosures about the irm s current perormance. T=3: Market trading. T=4: Investment and inancing decisions the company may have an opportunity or a potentially beneicial project that would require raising additional capital. T=5: Realization o payos. The initial situation Creating opportunities to misreport earning o inormation and disclosure decisions Market trading Financing and investment decisions Payos realized T Fig. 1: Sequence o Events We now speciy the assumptions regarding each one o the six stages. 2.2 T=: Initial Situation At T=, all companies are publicly traded, and each is run by a manager. Without loss o generality, we assume that at T= each company has one issued share that is held by initial shareholders including the company s manager. The manager holds a raction β o the company s stock. (The results would be essentially the same assuming that the manager has an option to purchase a raction β o the company s stock. The manager is assumed to be cash-constrained and thus cannot purchase additional equity. 6

9 We shall abstract rom other incentives that the manger might have (e.g., due to the threat o a control contest and will assume that the manager will be making decisions or the irm in all o the model s periods. We will urther assume that the manager s interest in enhanced share value comes solely rom the speciied holding o shares. We initially assume that, due to legal or contractual constraints, managers may not sell shares at the T=3 market trading stage. Thereore, the manager s objective is to maximize the price o the company s stock at the inal period. This assumption will be relaxed in Section 3, which will allow or managerial selling o shares in the intermediate trading period. The T= value o a company, which is also the T= market price o the company s share, is denoted by P. 2.3 T=1: Creating Opportunities or Misreporting Investments in creating opportunities to misreport i.e., to make positive disclosures even when long-term prospects are unavorable -- are + determined endogenously. At T=1 the manager invests C R in creating opportunities or uture misreporting o unavorable inormation concerning corporate perormance. 4 Speciically, given an investment C at T=1, i at T=2 the manager learns negative inormation, she will be able to misreport with ~ ~ ~ probability δ = λ δ ( C. We assume that λ, δ ' ( C >, δ ' ' ( C < and ~ δ ' (. 5 ~ The unction λ δ ( C represents the various legal and institutional actors in the economy that deine the available misreporting technology. Many actors inluence this technology, including legal rules, accounting standards and conventions, the structure o the accounting industry and accounting services (which in turn aect the ability o managers to inluence 4 We assume that C is not observable to the market (although the market will be able to anticipate the level o C at equilibrium. Otherwise, managers would be able to eectively commit to truthul reporting simply by setting C =. Indeed, i C were observable and veriiable, we would expect shareholders (or those who took the irm public to require managers (through corporate charters or contracts with the managers not to invest in creation o opportunities to misreport. 5 An alternative assumption would be that C can be invested ater the company s type is revealed (at T=2. Similar results will hold under this alternative assumption. Fischer and Verrecchia (2 study a model where misreporting is costly (they study a singleperiod model, so the question when the cost o distorting the report is incurred does not come up. 7

10 auditors to go along, and the intensity o outside monitoring (by analysts, plainti lawyers, the inancial press, etc. A higher λ represents a more lax environment that makes it easier to misreport. 6 Theoretically, the investment C may create opportunities or misreporting o avorable inormation as well. Moreover, managers can perhaps invest in creating additional opportunities or misreporting C g avorable inormation. owever, it can be readily shown that managers will never misreport, announcing that their inormation is negative, when in act it is positive. It ollows that managers will not have any reason to invest in creating opportunities to hide avorable inormation. Thereore, without loss o generality, we assume that the investment C creates only opportunities to misreport unavorable inormation. 2.4 T=2: earning o Inormation and Disclosure Decisions At T=2 the manager o each company learns whether the company s existing project is promising or not. Speciically, the manager will learn whether the company is a low-value, type company, or a high-value, type company. The expected value o the inal (T=5 payo rom the irm s existing assets is V or type irms and V or type irms, with V > V. We denote the dierence between the two expected values as V = V V. The probability that the manager will learn that the irm is type (type is θ ( 1 θ. Managers inormation about the irm s type is unobservable (even to the irm s auditors, and managers are not required to make any statements about this inormation. Managers, however, are required to make statements at T=2 about the company s current perormance. The statements might concern current igures pertaining to revenues, costs, earnings, and the like, as well as current actions such as the establishment o strategic or other relations, and so orth. For concreteness we assume that managers are required to state an earnings 6 We shall, or simplicity o exposition, assume that the level o λ is given by the environment and is not a parameter chosen by the irm itsel. O course, although the environment clearly plays a key role in shaping the scope o misreporting opportunities, companies might be able to adopt observable arrangements that inluence the level o λ. Our model can be easily extended to the case in which the environment deines a range o values or λ rom which irms can choose. In our model, i companies could lower λ in an observable ashion, it would be ex ante optimal or shareholders to do so. Accordingly, one can simply interpret the analysis below assuming that the λ it uses is the lowest possible given the legal and institutional environment. 8

11 igure (e.g. in the company s inancial statements, E { E, E }, where E > E. We denote by E the dierence between low and high earnings ( E = E E. We also denote by µ the ratio V E, i.e., the ratio between the dierence in value between type and type projects and the dierence between the two possible earning igures. We assume that, i the company is type, then managers will always be able to report at T=2 a high earnings igure,. I the company is type, its manager will always be able to report a low earnings igure,. owever, a manager o a type irm might also be able to report a high earnings igure. Speciically, there is a probability δ -- where δ is determined by the T=1 investments as already described -- that the manager o a type irm will be able to report a high earnings igure. It is worth noting that all o our results also apply to the case in which disclosure at T=2 is voluntary. All the irms that announce high earning igures in the mandatory disclosure case will also elect to make such statements in the voluntary disclosure case, and all the irms that announce low earning igures in the mandatory disclosure case will be silent in the voluntary disclosure case. The discussion below will use the mandatory disclosure model, but we will on occasion note the particular voluntary disclosure interpretation o our results. Investors at T=2 only observe the manager s statement about current perormance. They make whatever inerences can be rationally drawn rom the manager s announcement. The T=2 market price o the company s share will be denoted by 2 P. 2.5 T=3: Market Trading At T=3 market trading occurs, because some shareholders must sell or 3 liquidity reasons, and a price P is set or the company s stock. Given our current assumption that legal or contractual restrictions prevent the manager rom trading at T=3, the T=3 price cannot relect any new inormation, i.e. 3 2 P = P. 2.6 T=4: Investment and Financing Decisions At T=4, the manager might learn -- with probability γ -- o a potential new and proitable project. This project requires an investment K. Our analysis can be viewed as covering both the case in which this capital is 9 E E

12 needed to build the new project rom scratch and the case in which this capital is needed to acquire another company. The project will increase the inal cash low by K + R, where R is distributed over R + ( =, with a positive and continuous pd (R and a cd (R. et [ F E( R R = denote the average return o the new project. Note that or now we are assuming that the new project, i it emerges, would be an eicient one, with the uncertainty being only about its proitability, but we shall drop this assumption and allow or ineicient projects in Section 3.6. The manager knows R at T=4. The market, however, knows only the distribution (R. In the event that the company elects to raise capital, the market will also make whatever inerences can be drawn rom the manager s T=2 announcement and the T=4 decision to raise capital. The T=4 market 4 price o the company s stock is denoted by P. We shall or now assume that capital can be raised only by issuing equity; Section 3.7 will consider the case o issuing debt. In the case o inancing an acquisition, the new equity might be given directly to the shareholders o the acquired company or it might be sold to third parties and the cash obtained orm them given to the selling shareholders. We shall denote by αˆ the number o new shares that would have to be sold to raise K through issuing new equity. Selling αˆ shares would involve giving up a ˆ α raction α = o the company s T=5 total cash lows, and it will be 1+ ˆ α convenient to use α, rather than αˆ, in the mathematical derivations. et α ( E and α ( E denote the ractions o the company that will need to be sold in order to raise K when managers announce E and E, respectively. 2.7 T=5: Realization o Payos At T=5, all cash lows are realized. The company s initial project will produce cash lows o V + εo, where V { V,V } and ε O is a random zeromean noise. The company s new project, i one was undertaken at T=4, will produce cash lows o K + R + ε N, where ε N is a random zero-mean noise. I at T=1 the company invested in creating opportunities to misreport its earnings at T=2, cash lows are reduced by the cost C o doing so. The inal T=5 stock price is denoted by P. Note that the presence o noise implies that it is not possible to iner clearly rom a company s T=5 cash lows whether or not misreporting took 1

13 place at T=2. When a company reported high earnings at T=2, a relatively low value at T=5 could be due to an unavorable realization o uncertainty rather than to misreporting at T=2. O course, while the model assumes that whether misreporting occurred is not directly observable, in reality ex post investigations sometimes unearth evidence that misreporting took place. We shall assume or simplicity that no ex post penalties will be imposed at T=5. 7 Our model, however, can be easily extended to the case in which misreporting is penalized ex post with some probability. In such a case, misreporting will take place only i the beneits rom it, as derived below, exceed the expected sanction. The results presented below are qualitatively robust to such an extension (adding this threshold condition or misreporting. 3. Reporting and Investment Decisions As is conventional, we solve the model by backward induction, starting with the T=4 inancing and investment decisions. We irst examine how decisions at this stage would be made i no misreporting took place earlier. We then study how these decisions would be made in the presence o misreporting. 3.1 Financing and Investment Decisions Without Misreporting Consider irst the benchmark case in which no company can misreport, so that at T=2 companies report E and companies report E. In this case, since R, all companies that ace a new project will issue equity and raise K to und it. Assuming no misreporting, let α α α α ( and ( E E denote the ractions o the company that need to be sold in order to raise K when managers announce E and E, respectively. Speciically, in order to raise K: companies will sell a raction α ( + K + R K ; and V = α o the company such that 7 We assume that neither legal penalties nor extra-legal (or reputational market penalties will be imposed. In act, absent actual raud, when misreporting entails legal manipulation o the earnings igure within the boundaries set by the pertaining accounting standards, there is no threat o legal penalties. Still, in a model with multiple reporting periods, reputational penalties may inluence managers reporting decisions. 11

14 companies will sell a raction ( + K + R K V = α. Clearly, α o the company such that α > α, i.e. companies will have to sell a larger raction o their T=5 cash lows in order to und the new project. In the no misreporting case, the expected inal T=5 per-share prices or and companies are E( P = ( α ( V + K + R E( P = ( α ( V + K R + 1 and 1, respectively. 8 In this model, these will be the manager s (per-share payos, depending on her company s type. Thereore, the manager will always sell equity to inance the new project. A manager acing a below average new project, i.e. R < R, will clearly sell equity to inance this new project. This manager enjoys both the positive revenues rom the new project and a cross-subsidization eect that lowers the number o shares that must be sold to raise K. A manager acing an above average new project, i.e. R > R, will also sell equity to inance the new project, since the high revenues more than oset the cross-subsidization eect Financing and Investment Decisions with Misreporting In our model, some irms might have an opportunity to misreport at T=2. When companies with an opportunity to misreport mimic companies and announce, the market cannot distinguish between these E two types o companies. Consequently, a single pooling price is set or all the companies that announce. et α α denote the raction o the E ( P E E T=5 cash lows that a company that announces will have to sell in order to und a new project when (some companies misreport earnings. Speciically, in order to raise K, managers must sell a raction α P o the company such that α Π K, where Π is the expected value o a company P = conditional on the act that the company reported This expected value is given by: E and is selling equity. 8 The expected T=5 cash lows o the company are V + K + R or companies and V + K + R or companies. To get the per-share market price, we divide these values by the number o outstanding shares, 1+ αˆ. Recall that α = ˆ α ( 1+ ˆ α, which implies 1 + ˆ α = 1 ( 1 α. 9 Formally, the manager will sell equity i and only i ( 1 α ( V + K + R > V, where α satisies α ( V + K + R = K R K R ( V + R. This condition can be rewritten (ater some rearranging as >, which implies that the manager will sell equity or all R. 12

15 Π = V [ R ] ρ, ( θ δ Pr( sell ρ V + K + θ δ Pr( sell R + (1 θ Pr( sell where Pr( sell is the probability that an company that misreports and has a new project sells equity, Pr( sell is the probability that an company that has a new project sells equity, ρ = θ δ Pr ( sell + (1 θ Pr( sell is the overall probability that a company that has a project and reported E sells equity, R is the expected value o the proit rom an company s new project conditional on the company announcing and selling equity, and R is the expected value o the proit rom an company s new project conditional on the company selling equity. To proceed, we need to derive the probabilities Pr( sell and Pr( sell and to identiy the circumstances in which each type o company sells equity. This is done in the ollowing proposition. Proposition 1: When companies that can misreport do so at T=2, then, in the event that a proitable new project emerges at T=4 - (i Managers o companies, both those that misreported and those that did not, will always sell equity to und the project. (ii Managers o companies will sell equity i and only i the proitability o the new project exceeds a threshold equation: θ δ R + (1 θ Rˆ θ δ + (1 θ E Rˆ, which is deined by the ollowing R ( R dr θ δ V ( 1 F( Rˆ V ˆ R + K + Rˆ =. Remark: The intuition or this result, whose proo is provided in the Appendix, is as ollows: In the no misreporting case, we have seen that managers will always sell equity to inance new projects. Introducing misreporting by companies adds a cross-subsidization eect, which stems rom the pooling between companies and companies that misreport earnings. This cross-subsidization will make misreporting companies all the more eager to sell equity in order to inance their new projects, because they will now need to sell ewer shares at T=4 (compared with the case in which there is no misreporting. The same cross-subsidization eect might prevent companies rom pursuing new projects, since they will now need to sell more shares at T=4 in 13

16 order to raise K. Speciically, companies acing a proitable, yet insuiciently attractive project, i.e. a new project with R < ˆ, will orgo the new project. From proposition 1, we know that companies that misreport earnings ρ = θ δ + ( 1 θ Pr sell. will always sell equity, i.e. ( sell 1 Pr =. Thereore, ( Since the proportion o companies among companies that announce sell equity is o central importance, we deine companies sell equity, i.e. i = and ˆ R E and ω θ δ ρ. Also, i all R ( sell 1 Pr =, we deine θ δ ω =. Using these deinitions, we can state the ollowing θ δ + (1 θ Corollary. Corollary 1 (Eiciency Costs: As long as the dierence between and types satisies V > R ω, then - (i Some Companies will not inance and invest in new eicient projects. (ii The likelihood that an company will not inance and invest in a new eicient project that it aces is increasing in the threshold value Rˆ, which Rˆ in turn is an increasing unction o V and µ : Rˆ > and >. 1 V µ ( (iii Among companies that announce, companies that misreport earnings will be more likely to raise capital. Remark 1 (intuition: The intuition or this result, whose proo is provided in the Appendix, is as ollows: (i In the symmetric inormation case, absent cross-subsidization o companies, companies will always sell equity to pursue the new project i it emerges. Misreporting introduces the cross-subsidization eect, which imposes an additional cost, ω V, on companies that sell equity. When this cross-subsidization cost is suiciently small ( ω V < R, then companies will always sell equity even in the presence o misreporting. owever, when the cross-subsidization eect is signiicant (speciically, i ω V > R, companies will sell equity only when acing a new project that is suiciently proitable, namely when R > ˆ. 1 For this part o the corollary it is necessary to assume that ( precise condition is provided in the proo in the Appendix. 14 E R ˆ is not too large (the R

17 (ii When the cross-subsidization eect is signiicant, i.e. when ω V > R, companies will orgo eicient projects with R < ˆ. Thereore, the likelihood that an company will not inance and invest in new eicient projects is increasing in Rˆ. The threshold value, Rˆ, depends on the magnitude o the crosssubsidization loss that companies must bear i they choose to sell equity, as measured by ω V. When the dierence between the expected value o the initial projects o and companies, V, is larger, the cross-subsidization eect is also larger. Put dierently, since V = µ E, when the impact o misreported earnings on the estimate o the T=5 inal value (as measured by µ is greater, the cross-subsidization eect is larger. Note that, when ω V > R and ˆ >, companies might not raise capital to und eicient projects. Speciically they will not undertake projects with R, ˆ. Thereore, The possibility o misreporting might lead to ( R allocative ineiciency, generating a real economic cost. E R (iii When ω V > R, each company that misreports and announces will sell equity. On the other hand, among the companies, which all also announce project, with E R > E Rˆ, only companies acing a suiciently proitable new, will sell equity. ence, among companies that announce, those that misreport earnings are subsequently more likely to sell equity. R Remark 2 (empirical implications: Corollary 1 provides us with the ollowing testable predictions or uture empirical work: (i Companies that restate earnings, or are otherwise ound to have misreported, are more likely to have subsequently sold equity. This prediction is consistent with recent empirical evidence documented by ang and undholm (2. 11 (ii The greater the magnitude o the misreporting o earnings, or the more signiicant the misreporting in terms o its implication or the expected inal value, the more likely it is to be ollowed by an equity sale Palmrose & Scholz (2 and Palmrose, Richardson & Scholz (21 collect data on restatements by companies. This type o data can be used to test the predictions derived rom our theoretical model. 12 When V (or µ are higher, the threshold Rˆ is higher, which means that ewer companies sell equity. Since companies always sell equity, i ewer companies sell equity, then rom the pool o companies that announce and sell equity, the share o 15 E

18 3.3 The Reporting Decision Ater analyzing the T=4 inancing and investment decisions, we now move one step backwards in time, and solve or the T=2 decision o companies whether to misreport earnings. 13 Proposition 2: In the unique equilibrium, all companies that can misreport at T=2 will elect to do so. Remark 1 (intuition: The intuition or this result, whose detailed proo is omitted, is as ollows: The existing shareholders o an company clearly gain rom announcing E reports (rather than E E. Whatever the other companies report, i an company the market will assign a larger probability that the company is o type, as compared to the case in which the company announces. ence, misreporting will reduce the cost o raising capital. And, since misreporting is costless at T=2 (the cost C o creating opportunities to misreport is sunk at this stage, misreporting would be a dominant strategy or the managers o any company that can misreport. Note that, given that managers have opportunities to misreport, shareholders o companies will beneit rom their managers doing so; this is the case even though, as will be discussed (see proposition 4, shareholders will be better o ex ante under a regime that provides managers with ewer opportunities to misreport. Remark 2 (the gain rom misreporting: As noted above, the gain rom misreporting derives rom the more avorable terms or raising equity i.e., rom having to sell ewer shares to inance the new project. In particular, without misreporting an company will have to sell a raction α o the company such that α ( + K + R K, leaving it with an expected value o ( α ( + K R V + V = 1. In contrast, an company that misreports will need to sell only a raction α P o the company, deined by α P Π = K, leaving the 1 α V + K + R. Thereore, initial shareholders with an expected value o ( ( P E misreporting companies increases. Consequently, the correlation between misreporting and selling equity increases. 13 Since we are currently assuming that managers cannot sell stock at T=3, we can skip period 3. 16

19 recalling that a new project will emerge with probability γ, the gain rom misreporting is ( α ( V + K R G = γ α. P + The gain rom misreporting is increasing in the probability that a new project will become available, γ. Also, the gain rom misreporting is decreasing in α P, or equivalently is increasing in Π, the expected value o a company that announces and sells equity. Since Π is decreasing in δ, the E gain rom misreporting is also decreasing in δ. Intuitively, when the level o misreporting is higher, the market will know that among companies announcing there are more companies. Consequently, the market will E oer a lower price per-share or companies that announces E, reducing the gain rom misreporting. Finally, the gain rom misreporting is increasing in V (or µ. When the dierence in value between companies and companies is greater, companies have more to gain rom pooling with companies. 3.4 Creating Opportunities to Misreport At T=1 managers decide how much to invest in creating opportunities to misreport earnings. The equilibrium level o this investment decision is characterized in the ollowing proposition. * Proposition 3: In the unique equilibrium, all companies invest C at T=1 in * ~ * creating opportunities to misreport, where is deined by λ δ ' C * G δ C =. The level o investment, C ( ( ( 1 * * C, and as a result the overall level o misreporting, ( C δ, is - (i increasing in the laxity o the legal and accounting environment, as measured by λ ; (ii increasing in the probability that a company will ace a new project, γ ; (iii increasing in the magnitude o the dierence in value between and companies, V, and thus in the signiicance o the misreporting or the expected inal value, as measured by µ. Remark 1 (intuition: The intuition or this result, whose proo is provided in the Appendix, is as ollows: A single manager has no inluence on the overall level o misreporting, * and thereore takes ( C δ as given in her T=1 decision concerning how much 17

20 to invest in creating opportunities to misreport. The manager thereore will * λ ~ δ ' C G δ C exceeds the increase C as long as the marginal beneit o ( ( ( marginal cost: 14 (i When the legal and accounting environment is more lax, the marginal beneit o investment in creating opportunities to misreport in terms o the increased probability o being able to misreport is larger. Consequently, managers will invest more in creating opportunities to misreport. This result implies that, by reducing the laxity o the legal and accounting environment, we may be able to lower investment in creating opportunities to misreport and in turn the incidence o misreporting. (ii As explained in section 3.3, the gain rom misreporting is increasing in the probability that a new project will emerge. Thereore, when the probability γ increases, managers will invest more in creating opportunities to misreport. (iii As explained in section 3.3, the gain rom misreporting is increasing in the dierence between the value o and companies, V, i.e., in the signiicance o the misreporting or the estimated expected inal value, µ. Thereore, when V and opportunities to misreport. µ are higher, managers will invest more in creating Remark 2 (empirical implications: The results stated in proposition 3 provide the ollowing testable predictions: (i The result stated in part (i o proposition 3 is consistent with empirical evidence regarding the positive eects o switching to more strict accounting standards (see euz and Verrecchia (2. 15 It is also consistent with evidence indicating that managers take advantage o minimal disclosure requirements to engage in earnings management (see obo & Zhou ( * Given δ ( C ~ δ '( C G( δ ( C * = 1 * and consequently o δ ( C must induce the aggregate level o misreporting, ( C ence the condition: ~ * λ δ '( C G( δ ( C * = 1. Since both δ ( C and G ( ( C, there is a unique investment level, C, that satisies the FOC, λ. At equilibrium, the many individual managerial choices o C, δ. δ are decreasing in C, the condition λ ~ δ '( C * G( δ ( C * = 1 deines a unique level o investment in creating opportunities to misreport. 15 euz and Verrecchia (2 document the lower bid-ask spreads and higher trade volumes enjoyed by German irms that switched rom the German reporting regime to an international reporting regime (IAS or U.S. GAAP. 18

21 Furthermore, since the eectiveness o any set o legal or accounting standards and practices varies across industries, proposition 3(i suggests cross-sectional variations in the level o misreporting. This result is consistent with recent empirical evidence documenting more severe real eects o earnings manipulation in R&D intensive companies where existing accounting standards provide or only limited transparency (see Polk & Sapienza (22 and Aboody and ev (2. (ii The result stated in part (ii o proposition 3 suggests the ollowing testable predictions: (a Cross-sectionally, in industries where companies are likely to ace new opportunities that require additional capital, misreporting o earnings is more likely to occur. (b Comparing dierent time periods, in periods when more companies ace such new opportunities, misreporting o earnings is more likely to occur. Relatedly, in periods when there are many equity oerings, misreporting o earnings is more likely to occur. (iii The result stated in part (iii o proposition 3 provides the ollowing testable predictions: (a Cross-sectionally, in industries where the impact o misreported earnings on the estimated inal value (as measured by µ is greater, misreporting is more likely to occur. This prediction is consistent with the evidence that inormation asymmetries are especially large in R&D intensive industries (see Aboody and ev (2. In such industries, misreporting is more likely to occur. (b Comparing dierent time periods, in periods when managerial misreporting has a large impact (as measured by µ on the estimated value o the initial projects, misreporting is more likely to occur. 3.5 The Ex Ante Cost o Misreporting We can now state the magnitude o the eiciency cost generated by misreporting. Proposition 4: Misreporting generates an expected eiciency cost o: * Φ = C + γ 1 θ Pr R < Rˆ E R R < ˆ. ( ( ( R 19

22 Thereore, with misreporting, the ex ante T= value is reduced by Φ to V = V θ V + γ R Φ. The eiciency cost, Φ, and thus the reduction in ex ante value, are - (i increasing in the laxity o the legal and accounting environment, as measured by λ ; (ii increasing in the probability that a company will ace a new project, γ ; (iii increasing in the magnitude o the dierence between and companies, V, and thus in the signiicance o the misreporting or the expected inal value, as measured by µ. Remark 1 (intuition: The intuition or this result, whose detailed proo is omitted, is as ollows: Misreporting leads to two types o eiciency costs, which are relected in the two elements o Φ : * (1 The deadweight cost o creating opportunities to misreport, C. (2 Ineicient investment decisions: With probability γ the company aces a new eicient project. owever, i this company is o type (the probability o high value is 1 θ, and the new project is not suiciently proitable Pr R < ˆ, then this eicient project will not be pursued, leading to an ( ( R expected loss o E( R Rˆ R <. As stated in proposition 3, an increase in the laxity o the legal and accounting environment, in the probability that a company will ace a new project, in the magnitude o misreporting and in the signiicance o the misreporting or the expected inal value all lead to a higher investment in creating opportunities to misreport. These actors thus increase both the irst element o Φ and the second element o Φ. When the level o misreporting is higher, companies will be more reluctant to raise capital in order to inance a new project. These companies will invest in the new project only i its returns are especially high. Formally, a higher level o misreporting increases the threshold Rˆ, and with it the raction o projects that will be orgone and the expected loss rom orgoing such projects. Remark 2 (empirical implications: The results stated in proposition 4 provide the ollowing testable predictions: (i Companies located in countries with a legal and accounting environment that constrains misreporting to a greater extent will enjoy a 2

23 higher Tobin s Q. 16 Similarly, changes in the legal and accounting environment that make it more diicult to create opportunities to misreport will increase share value. (ii To the extent that companies can through private action reduce the level o the parameter λ in an observable way, such reductions would raise ex ante share value. This conclusion is consistent with the results in Fishman and agerty (1989, 199 and Dye (199 (see also Verrecchia (21, sec. 4 and the reerences he cites that commitments to improved disclosure raise ex ante share value. 3.6 Ineicient Projects We have shown that the possibility o misreporting has real economic costs. In particular, companies might not pursue eicient projects. In a more general model, misreporting leads to a second eiciency cost that companies might pursue ineicient projects. To show this point, we now extend the basic model to allow or ineicient projects. In particular, we introduce the possibility o ineicient projects with ~ ~ R = R ( R >. We assume that, in the event that a new project emerges, the ~ project will be ineicient (i.e. R = R with probability p, and with probability + 1 p the project will be eicient, with R being distributed as beore over R according to the pd (. The equilibrium in this extension, which allows or ineicient projects, is as ollows. Proposition 5: In the case in which the new project might be ineicient, i V is suiciently large, or equivalently i µ is suiciently large, then - (i companies that misreport and announce will invest in ineicient (ii projects. companies will pursue ewer eicient projects, as compared to the case in which all new projects are eicient. The threshold or investing is Rˆ characterized by Rˆ > and ~ >. p R Remark 1 (intuition: The intuition or this result, whose proo is provided in the Appendix, is as ollows: E 16 See euz and Verrecchia (2 or some preliminary evidence consistent with this prediction. 21

24 (i When acing an ineicient project, managers o companies that misreport might still choose to sell equity. While the initial shareholders o the company (including the manager will bear part o the loss produced by the ineicient project, they also will gain rom the cross-subsidization eect involved by raising equity while pooling with companies. When the latter eect dominates, managers o companies that misreport will sell equity even when acing an ineicient project. As V increases, the crosssubsidization gain increases as well, making an equity oering more appealing. (ii When companies that misreport pursue ineicient projects, the magnitude o the cross-subsidy eect increases. Consequently, companies will be more reluctant to sell equity, i.e., will sell equity only when the proitability o the new project exceeds a higher threshold. The magnitude o the cross-subsidy eect is increasing in the raction o projects that are ineicient, p, and in the magnitude o the ineiciency o these projects, R ~. Consequently, a higher p and a higher R ~ decrease the raction o eicient projects that companies will choose to undertake. (iii Note that the possibility o ineicient new projects introduces a orm o ineiciency that was absent rom the basic model. When the dierence between the value o the initial projects o and companies, V, is suiciently large, and thus the cross-subsidization eect is suiciently large, companies will pursue ineicient projects. Equivalently, when the impact o managerial misreporting on the estimated inal value o the irm s initial projects is suiciently large (i.e., i µ is suiciently large, the crosssubsidization eect will be suiciently large, and companies will pursue ineicient projects. Thus, once ineicient projects are introduced into the model, misreporting generates yet another orm o allocative ineiciency resulting rom the undertaking o ineicient projects. Remark 2 (testable predictions: The result stated in proposition 5 implies that ineicient projects projects that destroy rather than enhance value -- will more likely be undertaken by companies that misreported prior to undertaking the project. Recent evidence on the poor long-term perormance o companies that engage in earnings management either at the IPO stage (Teoh et al. (1998a or prior to a seasoned equity oering (Teoh et al. (1998b - is consistent with this prediction. 22