Payoff Matrices:

When an MEF mechanism exists, the manager and the auditor each have four choices. Panel A of Table 1 reports the payoff matrixes that the subjects face in the experiment.¹¹ Given that the manager makes a high forecast (i.e., (1)-(a) and (2)-(a) of Panel A) or a low forecast (i.e., (1)-(b) and (2)-(b) of Panel A), his choice pairs show in the columns, with two investment choices interacting with two reporting decisions. The auditor’s choice pairs are presented in the rows, with two audit effort levels interacting with two independence choices.

8In the remaining of this paper, I will use “misreporting” (“honest reporting”) and “compromise independence” (“remain independent”) interchangeably to describe the auditor’s accepting (rejecting) the bonus and issuing a favorable (unfavorable) audit report.

9Schwartz (1997) and King and Schwartz (2000) provide analytical and experimental evidence that a strict regime together with a damage measure that is independent of the investment level can induce socially optimal auditor effort and investment. I do not consider this issue explicitly; instead, I focus on the interplay between auditor independence and different regulation systems, and on manager’s earnings forecast, investment, and reporting decisions. Therefore, I assume the auditor’s technical and misreporting penalties are independent of the manager’s investment.

10I measure the forecast error by the difference between the forecasted and audited (rather than realized) earnings levels because the realized earnings level is the manager’s private information that is unobservable to the investors. In practice, the investors can only observe the publicly announced earnings numbers that have been audited by the auditor.

11These matrixes are calculated based on a set of parameters. These parameters do not necessarily represent real-world situations. Rather, they are chosen to (a) enable the manager’s and the auditor’s payoffs to be separated across choice options, thereby increasing the internal validity of the experiment (King 2002), and (b) facilitate the comparisons among equilibria in forming testable hypotheses (to be discussed in Section 2.3). See Appendix for details of selected parameters.

Both players’ choice options lead to sixteen possible “investment-report/effort-independence” combinations, as depicted in (1) and (2) under each legal regime. Panel B of Table 1 is similar to Panel A except that no MEF mechanism exists, thus reducing the manager’s decisions to only investment choices and reporting. All payoffs are measured by a notional currency called Experimental Dollars (EDs).

[Insert Table 1 here]

A regulation system comprises an MEF mechanism imposed on the manager (exists, denoted by MEF, or does not exist, denoted by NO) and a legal regime imposed on the auditor (negligence, denoted by NE, or strict, denoted by ST). Therefore, there are four possible MEF/regime combinations, as shown in (1) and (2) of Panels A and B. I construct the payoff matrixes by incorporating the following behavioral tensions in the experiment.

To the manager, even though choosing a low investment minimizes his effort cost, a low investment is more likely to produce a LOW realized earnings level. This possibility increases the manager’s probability of receiving a low earnings audit report, leading to forecast error costs and a low salary. The manager can offer the auditor a bonus, but the auditor may reject it. To the auditor, choosing low effort minimizes his effort cost but also increases the likelihood of obtaining inaccurate audit evidence, leading to a higher probability of committing a technical audit failure. Even though the auditor obtains correct evidence to support a LOW realized earnings level, he may still accept the manager’s bonus and misreport. This possibility increases the auditor’s probability of committing an independence audit failure. Under either audit failure, the auditor knows that his probability of being held liable is determined by his effort choice and legal regime. Once the auditor is held liable, he has to pay either a technical or a misreporting penalty.

In practice, C_High^LowEarnings_Forecast can be regarded as the manager’s forecast error cost associated with litigation and reputation loss, which is usually high. In contrast, a firm’s stock price usually declines when its manager forecasts a low earnings level. Although the firm’s stock price may rebound following a high earnings audit report, whether a firm’s stock price will increase or decrease following a low earnings audit report is unclear. In fact, whether a “net”

increase or decrease in stock price occurs when the manager makes a low earnings forecast and the auditor issues a high or low earnings report is not a major concern in my experiment. A more critical attribute that should be considered is that, given that the manager makes a low earnings forecast, the “net” stock price change following a

9

high earnings audit report is better than that following a low earnings audit report. For experimental purposes, I regard C_Low^High_Forecast^Earnings and C_Low^Low_Forecast^Earnings as decreases in a firm’s equity value attributable to a manager’s low earnings forecast and assume that the “net” stock price decrease following a high earnings audit report is smaller than that following a low earnings audit report. Thus, I choose three parameters for the three forecast error costs such that

Earnings igh H

Forecast

CLow < C_Low^Low_Forecast^Earnings < C_High^LowEarnings_Forecast. 2.3 Equilibria and Economic Hypotheses:

Table 1 reports the numerical equilibria with bolded labels from (#A) to (#I). Suppose an MEF mechanism exists and the auditor is subjected to an NE regime. Part (1) of Panel A indicates that subjects face two sub-games. In the first sub-game wherein the manager chooses to make a high earnings forecast, two competing equilibria exist. Equilibrium (#A) predicts that the manager will choose a high investment and honestly report Low when the realized earnings level is LOW. The auditor will respond by choosing a low effort and remaining

independent. The intuition of this equilibrium is as follows. Because the auditor decides to be independent (which reduces the independence audit failure to zero), the manager has to credibly report realized earnings.

Because the manager is mandated to make an earnings forecast and there are costs associated with forecast errors, the manager’s choice of a high investment not only increases the probability of obtaining a HIGH realized earnings level (which minimizes forecast error costs) but also increases the probability of receiving a high earnings audit report (which ensures that the manager earn a high salary). The manager chooses to make a high earnings forecast to ensure that he earns the forecast benefit associated with such a forecast. Thus, forecast accuracy increases. Because the auditor will not commit an independence audit failure and he knows that his ex ante expected probability of committing a technical audit failure decreases due to the manager’s high investment

choice, he will choose a low effort to minimize effort cost.

In contrast, equilibrium (#B) predicts that the manager will choose a low investment and misreport; the auditor will choose a high effort and compromise his independence. The auditor decides to misreport because the probabilities of being held liable are smaller under the NE regime when fraudulent attestations occur.

Because the manager expects that the auditor will misreport, the manager knows that he will earn a high salary and receive the forecast benefit if he forecasts a high earnings level. Therefore, the manager makes a high

earnings forecast, chooses the low investment, and misreports the true realized earnings. In contrast, because the auditor anticipates that the manager will choose a low investment and misreport, the auditor’s choice of high effort not only reduces the probability of committing a technical audit failure but also reduces the probability of being held liable.^12,13

In the second sub-game wherein the manager chooses to make a low earnings forecast, the two competing equilibria are the same as those when the manager makes a high earnings forecast. However, the underlying incentives are different. For example, when in equilibrium (#A), the manager chooses a high investment to increase the probability of minimizing forecast error cost ^C_High^LowEarnings_Forecast and to earn the forecast benefit. He makes the same decision in equilibrium (#C) to minimize forecast error costs associated with a low earnings forecast (note that C_Low^High_Forecast^Earnings <C_Low^Low_Forecast^Earnings ). Likewise, whereas the manager earns both a high salary and the forecast benefit in equilibrium (#B) due to the auditor’s lack of independence, he can only earn a high salary in equilibrium (#D). Thus, the manager has stronger incentive to choose a low investment to minimize effort cost.

Although each sub-game has two competing equilibria, equilibrium (#B) dominates the other three equilibria for the overall game because the manager is the first mover in the experiment and equilibrium (#B) provides him with the highest expected payoff (i.e., 13,858 EDs). Formally, I hypothesize the following:

HYPOTHESIS 1A: Under the MEF_NE regulation system,

(1) the manager more frequently makes a high earnings forecast, chooses a low investment, and misreports the low realized earnings level, and

(2) the auditor more frequently chooses a high audit effort and compromises independence.

12Because the auditor does not know the manager’s forecast and investment decisions and the realized earnings at the beginning of the game, his ex ante probability of committing either type of audit failure is uncertain. Therefore, the auditor’s ex ante best strategy is to choose a high effort to reduce a technical audit failure, even though his equilibrium strategy is to always reports favorably regardless of the audit evidence. Thus, (#B) is a feasible equilibrium to the auditor.

13In practice, an independence audit failure generally constitutes a fraud and when the auditor commits fraud he should be strictly liable regardless of the effort level he chooses. Therefore, the NE regime appears not applicable to the independence audit failure. This concern can be addressed by two explanations. First, the auditor’s misreporting may not be discovered until the economy worsens and investment failures occur (Business Week 2002). The key point is that fraud resulting from an independence audit failure does not automatically trigger an auditor’s legal liability. An independence audit failure usually becomes public either through whistle-blowing or by a client’s announcement of a restatement or bankruptcy. Second, in real litigation against the auditors, an independence audit failure is relatively more difficult to identify and determine than a technical audit failure given the unavailability of and difficulty in gathering convincing evidence that can prove the auditor’s violation of independence (Spellmire et al. 1993). Therefore, an auditor who misreports has the chance to escape litigation under the NE regime.

11

When the auditor is subjected to the ST regime, part (2) of Panel A shows that both sub-games have equal equilibria (#E) and (#F). Notably, these two equilibria are also the same as equilibria (#A) and (#C). Because the auditor faces much higher probabilities of being held liable under the ST regime, equilibria (#B) and (#D) cannot be sustained. Therefore, only one equilibrium exists in each sub-game. Again, equilibrium (#E) dominates equilibrium (#F) for the overall game because the manager’s expected payoff is higher (i.e., 13,340 EDs). Thus, I test the following hypothesis:

HYPOTHESIS 1B: Under the MEF_ST regulation system,

(1) the manager more frequently makes a high earnings forecast, chooses a high investment, and honestly reports the low realized earnings level, and

(2) the auditor more frequently chooses a low audit effort and remains independent.

When no MEF exists,¹⁴ Panel B of Table 1 shows that two competing equilibria exist under the NE regime and one equilibrium exists under the ST regime. Note that equilibria (#G) and (#H) are the same as equilibria (#A) and (#B) except that the manager only has to determine his investment choice (which affects his probability of earning a certain salary level and the auditor’s probability of committing a technical audit failure) and reporting decision. The forecast benefit and costs associated with forecast errors do not exist. Similarly, equilibrium (#I) is the same as equilibrium (#E). Because equilibrium (#H) dominates equilibrium (#G), I test the following two hypotheses:

HYPOTHESIS 2A: Under the NO_NE regulation system,

(1) the manager more frequently chooses a low investment and misreports the low realized earnings level, and

(2) the auditor more frequently chooses a high audit effort and compromises independence.

HYPOTHESIS 2B: Under the NO_ST regulation system,

(1) the manager more frequently chooses a high investment and honestly reports the low realized earnings level, and

(2) the auditor more frequently chooses a low audit effort and remains independent.

14I focus on the “no earnings forecast” scenario instead of the “voluntary earnings forecast” scenario for two reasons. First, the “no earnings forecasts” scenario provides a pure benchmark based on which I can accentuate the importance of MEF. Second, an examination of the “voluntary earnings forecast” scenario requires a consideration of certain benefits (e.g., preempt bad news surprises, attract new capital, reduce cost of capital) in the analyses, which not only complicates the experiment, but also blurs the focus of this study. See, for example, Sengupta (1998) and Skinner (1994, 1997) for discussions of these benefits.

Table 1 implies a trade-off between the manager’s and the auditor’s behavior. For example, equilibrium (#E) indicates that, when an MEF mechanism exists, the ST regime motivates greater auditor independence and induces the manager to make a high earnings forecast and choose a high investment more often, despite the fact that audit effort also decreases. This trade-off contrasts with the result of Schwartz (1997) that, under the ST regime, the social optimal level of audit effort is attainable if the regulator chooses a proper damage award. This difference arises because our studies focus on different incentive problems. Schwartz (1997) examines the auditor’s incentive problem with respect to effort and the investor’s incentive problem with respect to investment. The regulator can eliminate the investor’s incentive problem by adopting a damage award that is independent of the investor’s investment. Therefore, the regulator’s problem reduces to the choice of an appropriate damage award that can induce the auditor to exert the socially optimal effort. Different from her study, my study investigates the manager’s incentive problems in forecasting, investment, and reporting, and the auditor’s incentive problems in effort and independence. Both players have incentive problems that cannot be eliminated because they are strategically dependent.

An important regulatory question that has not been explored in the literature is whether an MEF mechanism is effective in motivating managers’ and auditors’ desirable behavior. To address this issue, I intentionally choose the parameters to be able to compare equilibria with and without MEF. As shown in Panels A and B of Table 1, equilibria (#E) and (#I) make the same predictions under the ST regime. The only difference is that equilibrium (#E) involves MEF whereas equilibrium (#I) does not. From the manager’s perspective, I predict that the manager will chooses a high investment more often under the MEF_ST system because doing so increases the manager’s probability of minimizing his forecast error cost and earning the forecast benefit associated with a high earnings forecast. These two incentives do not exist under the NO_ST system. Following this reasoning, I also predict that the manager will make honest reporting more often under the MEF_ST system because the manager has a higher probability of obtaining a HIGH realized earnings level. Behaviorally and economically, the manager has no incentive to misreport a HIGH realized earnings level. From the auditor’s perspective, he will remain independent more often under the MEF_ST system because the manager is more likely to choose a high investment to legally meet the earnings forecast and honestly report realized earnings.

Therefore, the bonus payment tends to be unnecessary in helping the manager receive a high earnings audit

13

report. Because the auditor’s honest reporting reduces his probability of committing an independence audit failure to zero, and because the manager’s undertaking of a high investment reduces the auditor’s probability of committing a technical audit failure, the auditor will choose low effort more often to minimize his cost.

Based on these discussions, I examine the effectiveness of MEF by testing the following hypothesis (in alternative form):

HYPOTHESIS 3: Under the ST regime, the existence of MEF is effective in that:

(1) the manager more frequently chooses a high investment and makes honest reporting under the MEF_ST system than under the NO_ST system, and

(2) the auditor more frequently chooses a low audit effort and remains independent under the MEF_ST system than under the NO_ST system.

I do not compare MEF_NE and NO_NE because equilibria (#B) and (#H) predict that the manager will choose a low investment and misreport and that the auditor will compromise independence. To securities regulators, these decisions are detrimental to the capital markets and investors, making the comparison between MEF_NE and NO_NE meaningless.