Political pressures and the evolution of disclosure regulation

Abstract

This paper examines the process that drives the formation and evolution of disclosure regulations. In equilibrium, changes in the regulation depend on the status quo, standard-setters’ political accountability and underlying objectives, and the cost and benefits of disclosure to reporting entities. Excessive political accountability need not implement the regulation preferred by diversified investors. Political pressures slow standard-setting and, if the standard-setter prefers high levels of disclosure, induce regulatory cycles characterized by long phases of increasing disclosure requirements followed by a sudden deregulation.

Should standard-setters be accountable to the public and its elected representatives? Historically, the question has been divisive. On the one hand, many standard-setters argue against political interference; Dennis Beresford, a former chairman of the Financial Accounting Standard Board (FASB), notes that members of Congress often strongly oppose certain FASB positions during congressional hearings: “The FASB often is on the defensive because these hearings are generally convened when certain companies, industry associations, or others allege that pending FASB positions will cause serious economic harm if adopted as final accounting standards” (Beresford 2001).¹ On the other hand, government regulators have often argued that standard-setting should be subject to high levels of political oversight (Zeff 2003). Consistent with this view, the current institutional environment provides the means for law-makers to immediately override any accounting standard. This environment is different from other policy choices such as judiciary rulings (e.g., the Supreme Court) or monetary policy (e.g., the Federal Reserve).

Resolving this debate is difficult because, for the most part, the economic consequences of political accountability are understudied. This paper proposes to speak to this debate by examining the costs and benefits of political oversight on the regulation of accounting standards. We examine this question within the general paradigm of accountability in government and refer to accountability as a political process that restrains the actions of a regulator (Laffont and Tirole 1991; Maskin and Tirole 2004). As in this literature, we ask whether political accountability will effectively discipline a regulator to implement a social objective.

To analyze the consequences of accountability, our theoretical model incorporates the following principal elements.

Reporting motives. Managers have private information about future cash flows and prefer standards that maximize the (short-term) stock price after disclosures have been made. Managers can report information voluntarily but cannot withhold information in a manner that violates the disclosure regulation. Managers’ preference over regulations thus depends on their private information: managers tend to prefer regulations in which (a) they have discretion to withhold their own information (since they can always disclose voluntarily) but (b) other firms observing comparably less favorable information must report their information.

Political accountability. Managers can collectively prevent a new standard from being implemented by standard-setters, and, when this occurs, managers can impose a preferred alternative. Standard-setters are more accountable when fewer managers can block a new standard. However, as we show in the model, the standard preferred by most managers is typically not the standard that maximizes welfare. We also assume that standard-setters are not perfectly benevolent (welfare-maximizing) and thus political accountability has a purpose. For example, the concepts statements of the FASB emphasize promoting transparency but provide little room for deliberation economic consequences or welfare. Hence, by design, the current institution may prefer more transparency than is socially desirable.²

Evolution of standards. Accounting regulations are evolved institutions that dynamically change over time.³ Reporting managers consider their preferred alternative relative to the status quo if the current regulation remains in place. As a result, the status quo determines whether managers will attempt to block a new regulation and which new standards are politically feasible. With each new round of standard-setting, the status quo evolves, leading to implications about the dynamics of regulations. In the model, these dynamics can converge to a long-term stable standard in which the issues are permanently settled or feature regulatory cycles whose general patterns are analyzed.

We present below an overview of the model. The economy proceeds through time and features successive generations of managers who, as is usual in this literature, receive private information prior to the realization of final cash flows, which they may disclose prior to a sale.

There are two channels through which managers disclose information. First, a regulation requires a disclosure of certain events. Second, for events that are not covered by the standard, managers may disclose voluntarily. Then, managers make a productive decision and choose whether to (a) liquidate early and distribute the proceeds to current shareholders or (b) continue and sell the firm at the expected cash flow conditional on all disclosures, if any.

The disclosure regulation is selected as the outcome of a political game between managers and the standard-setter. In each period, there is a status quo representing the standard in the previous period. The standard-setter makes a new regulation proposal, and managers can strategically decide whether to oppose the proposal. The proposal fails whenever there is too much opposition. Then, the standard-setter loses control over the agenda, and the new regulation is chosen by a regulator maximizing approval over the status quo. The implemented regulation endogenously evolves over time because the political opposition to a new standard is a function of the status quo.

The primary result that emerges from the analysis is that political accountability does not necessarily work to direct the standard-setter toward stable welfare-maximizing regulations. Instead, excessive accountability can destabilize the standard-setting process into recurring regulatory cycles. This situation occurs specifically when a standard-setter desiring high levels of disclosure is subject to high levels of accountability. Regulatory cycles proceed along two phases. In the first phase, increasingly comprehensive disclosure rules are imposed over time, starting from an unregulated economy and evolving toward increased disclosure requirements. Evolution is slow, especially when political accountability constrains the standard-setter to increase disclosure requirements in small steps to offset political opposition. In the second phase, the current regulation reaches a turning point where most firms are required to disclose and force the standard-setter to cut back on disclosure. What follows is a quick deregulation. Then, the new standard moves to relatively low levels of disclosure and the next regulatory cycle begins.

The economic intuition for regulatory cycles is as follows. In this model, disclosure requirements are optimally over unfavorable events (e.g., an asset impairment) because these events are not reported voluntarily for individual reporting purposes and result in reduced aggregate economic efficiency.⁴ In the first phase of evolution, the standard-setter increases transparency by requiring that relatively unfavorable events be subject to a disclosure requirement. Managers newly subject to mandatory disclosure relative to the prior status quo oppose the loss of discretion, and therefore a politically accountable standard-setter cannot increase disclosure requirements too quickly without losing control of the proposal process.

Over time, the status quo evolves with increasingly favorable events becoming subject to the disclosure requirement. Eventually most of the firms are subject to disclosure requirements, and the second phase of evolution begins. At this turning point, the status quo is no longer the alternative collectively preferred by managers because a disclosing firm is always weakly better off when retaining the discretion to withhold information. Nor is a small decrease in disclosure requirements possible because such a new regulation would be opposed by all remaining nondisclosers under the status quo, because their market price would decrease. Hence, the solution at this stage is an abrupt reduction in disclosure requirements, supported by the largest fraction of firms that disclosed under the status quo but do not under the new regulation.

A standard-setter who prefers low levels of transparency might not reach the second phase, in which case the regulatory process will attain a long-term stable regulation. Within our model assumptions, the second phase is not attained if the standard-setter maximizes the average market price. Under this scenario, the standard will converge to the price-maximizing disclosure requirement in the long run but convergence is slower when the standard-setter is more accountable. We further show that political accountability is entirely ineffective at disciplining a standard-setter preferring lower disclosure requirements than the regulation maximizing the market price.

The benefits of some independence from political pressures by policy-making bodies such as the Federal Reserve or the Supreme Court has been the subject of a large literature in institutional economics (Kydland and Prescott 1977; Gely and Spiller 1990). However, this literature does not examine debates that pertain specifically to accounting regulations. Several recent empirical studies provide evidence that firms pressure regulators strategically, in response to the perceived market consequences of regulation proposals (Chan et al. 2006; Hochberg et al. 2009; Allen and Ramanna 2013). While these studies have made researchers aware of the key role of political pressures, they are descriptive and do not test predictions about the effects of political activism on disclosure standards.

Relating to these issues, a strand of the literature analyzes the influence of various parties in the standard-setting process (Amershi et al. 1982; Fields and King 1996); our research focus here is different in that we take influence as the starting point and study how it may affect regulatory choice.⁵ Our study also complements a recent literature on institutional design in accounting, which discusses how certain characteristics of the institution affect policy choices. The broad implications of the consolidating standard-setting into a single body are discussed by Dye and Sunder (2001), Basu and Waymire (2008) and Bertomeu and Cheynel (2013). These studies find various benefits in multipolar standard-setting institutions in which market forces will push for more efficient standards. At the other side of this debate, Ray (2012) examines the potential learning cost of having multiple standards, and Friedman and Heinle (2014) show that multiple standards magnify the social costs of corporate lobbying.

Section 1 of the paper presents the basic model and some preliminary results. Section 2 provides an analysis of managers’ preferences and the disclosure rule that will be instituted if the standard-setter loses control of the agenda. The standard-setter’s strategy for keeping control of the agenda appears in Sect. 3, and the evolution of disclosure rules over time is discussed in Sect. 4. Section 5 discusses the effects of relaxing the model’s assumptions, and Sect. 6 concludes. The appendices provide a table of notation (Table 1), proofs, and further analysis of the design of disclosure regulations.

1 Model and preliminaries

The economy unfolds over an infinite time horizon, with periods indexed by $t\ge 0$, and is populated by successive generations of standard-setters and atomistic firms that deliver their cash flow at the end of the period. Each firm has been initially financed with equity, and some of this equity is owned by the manager (possibly as part of a compensation arrangement), while the remaining portion is held by diversified investors. The timeline of each period contains the following events, as illustrated in Fig. 1.

Fig. 1

Model timeline

At date $t.1$, managers receive private information about an end-of-period cash flow. For now, we assume that all managers are informed and further considerations in the case of some uninformed managers are delayed until the main results are presented. Each firm has an i.i.d. cash flow $v$ that is distributed according to a uniform distribution with support on $[0,1]$.

At date $t.2$, a new regulatory process begins. We focus here on regulations described by a threshold $A$ such that events with $v<A$ must be disclosed. This restriction is with some loss of generality as it takes as a given the use of impairment-based rules (widespread in accounting), such as lower-of-cost-or-market, impairments of long-lived assets or other-than-temporary losses, advance recognitions of loss-making sales, etc.⁶ We show in Appendix 2 that this form of mandatory disclosure maximizes popularity because, in our model, favorable events are already disclosed voluntarily.

Denote $A_{t-1}$ as the status quo, defined as the regulation implemented in the previous period and beginning with no disclosure $A_0=0$. The regulatory process takes place over two stages, on which we elaborate in Fig. 2.

Fig. 2

Regulatory process

At stage $t.2(a)$, the standard-setter makes a proposal $A$ (e.g., an exposure draft). We endow the standard-setter with a single-peaked preference ${\mathcal {U}}(A)$ with a maximum at $A^* \in (0,1)$.⁷ Managers are empowered to vote for their firm and may oppose the proposal. We capture their influence by a function $Opp(A,A_{t-1})$, defined as the fraction of managers who are strictly worse off under $A$ than they would be if the proposal were to fail. This construct intends to capture several venues through which, in practice, corporate lobbies can oppose a new regulation, such as comment letters or congressional hearings. Note that this function will be solved for by backward induction, as we assume that managers have rational expectations about what standard will pass at $t.2(b)$ if the standard-setter’s proposal fails.

Note that we do not model supporters for a new standard at this stage because, in practice, comment letters and congressional hearings overwhelmingly focus on groups that have grievances against a new proposed regulation (Beresford 2001; Zeff 2005). If $Opp(A,A_{t-1})\le \alpha$, the proposal passes and $A_t=A$ is implemented. The parameter $\alpha \in [0,1]$ captures the political accountability of the standard-setter: from $\alpha =1$ such that any proposal passes to $\alpha =0$ such that unanimity is required. Managers’ votes are not observable to investors.

If $Opp(A,A_{t-1})> \alpha$, the proposal is rejected and stage $t.2(b)$ begins in which the standard-setter no longer controls the agenda. Then, a new proposal is made by an office-driven bureaucrat or politician (such as a congressional subcommittee), who has a greater chance of staying in office or being elected when the proposal is more popular.⁸ Formally, we define the popularity of a regulation $A\ne A_{t-1}$ as $Net(A,A_{t-1})$, indicating the difference between the fraction of firms strictly better-off and the fraction of firms strictly worse off under $A$ versus the status quo $A_{t-1}$. $Net(A,A_{t-1})$ has a discontinuity at $A=A_{t-1}$, so we extend this function by continuity and set $Net(A_{t-1},A_{t-1})=\limsup _{A\rightarrow A_{t-1}} Net(A,A_{t-1})$. Put differently, this means that $A=A_{t-1}$ also refers to an infinitesimally small change to the status quo; in the limit, we interpret this situation as one in which the status quo is maintained.⁹

We assume that the most popular regulation $A_t \in argmax_A Net(A,A_{t-1})$ is implemented. For later use, we define the $Pop(A_{t-1})$ as the most popular regulation, where $Pop(A_{t-1})\in \hbox { argmax}_A\, Net(A,A_{t-1})$. This function affects managers’ opposition to the standard-setter’s proposals, and therefore the standard-setter’s choice of proposed disclosure rules. Except for the knife-edge case of $A_{t}=\max (1/2,4c/(4c+1))$, the most popular regulation is unique; however, should this knife-edge occur, we can select the solution closest to the status quo.

At date $t.3$, managers learn a liquidation cash flow $\theta$, drawn from an i.i.d. uniform distribution with support on $[0,1]$. If the firm is liquidated, the end-of-period expected cash flow $v$ is forfeited, and $\theta$ is distributed with no further need for disclosure. If the firm continues, the payoff $\theta$ is forfeited. This step is not critical for the main analysis, and the results are unchanged if the information has no productive purpose. The assumption serves to illustrate the economic distortions created by the political process in an environment where the price-maximizing regulation might feature some nonzero level of regulated disclosure.

At date $t.4$, managers make their disclosures, which we denote $d^{t}(v)\in \{v,ND\}$. If $v< A_t$, a disclosure is mandatory and $d^t(v)=v$. If $v\ge A_t$, the firm can withhold information and choose $d^t(v)=ND$ or disclose $d^t(v)=v$ voluntarily. There is a cost $c>0$ when making a mandatory or a voluntary disclosure. Hence, we assume that the same underlying technology is used in both disclosure channels; for example, the cost may represent a formal audit or leakages of proprietary information. Nondisclosers must also certify that $v>A_t$, which we assume entails a cost $A_t c$ that is linear in the probability of the event “$v\le A_t$.” As more values of $v$ are required to be disclosed, the greater the cost to certify that non-disclosure is appropriate. To avoid straightforward environments in which the standard-setter can pass any policy, we assume that $\alpha$ is not too large relative to the cost, that is, $\alpha \le \overline{\alpha }=\min (c,2c/(4c+1))$.

At date $t.4$, managers sell their shares in a competitive market. Conditional on a public disclosure $x\in [0,1]\cup \{ND\}$, investors price the firm at the expected cash flow minus the disclosure cost if any:

$$\begin{aligned} P^t(x)={\mathbb {E}}(v|d^t(v)=x)-1_{d^t(v)\ne ND}c-1_{d^t(v)=ND}A_tc, \end{aligned}$$

(1.1)

where $1_{d^t(v)\ne ND}$ (resp., $1_{d^t(v)=ND}$) is an indicator function equal to 1 if a disclosure is made (resp., if no disclosure is made) and zero otherwise.

In what follows, let $\tau ^t$ represent the threshold above which the event $v$ would be disclosed voluntarily if it were not subject to mandatory disclosure. As is well known (Jovanovic 1982; Verrecchia 1983), this voluntary disclosure threshold threshold is determined by the point at which a firm is indifferent between a voluntary disclosure and a non-disclosure,¹⁰

$$\begin{aligned} \tau ^t-c=P^t(ND)=\frac{A_t+\tau ^t}{2}-A_t c. \end{aligned}$$

(1.2)

Solving this equation, the voluntary disclosure threshold is given by

$$\begin{aligned} \tau ^t=A_t + 2 c (1 - A_t). \end{aligned}$$

(1.3)

As is entirely intuitive, increasing the mandatory disclosure threshold $A_t$ increases market expectations and thus also increases the voluntary disclosure threshold. We also note that the fraction of disclosing firms $1-(\tau ^t-A_t)$ increases when the mandatory disclosure threshold $A_t$ is increased.

Substituting (1.3) into (1.2) to derive

$$\begin{aligned} P^t(ND)=(1-2c)A_t+c. \end{aligned}$$

(1.4)

This implies that a firm, as long as it remains a nondiscloser, obtains a higher market price when the mandatory disclosure threshold $A_t$ is increased. The next lemma summarizes these observations, which are used throughout our analysis.¹¹

Lemma 1.1

The probability of disclosure and the non-disclosure market price are increasing in the disclosure threshold $A_t$ .

2 Popularity over the status quo

We solve the model by backward induction in period $t$ and first analyze stage t.2(b) of the regulatory process, that is, after the proposal made by the standard-setter fails. At this point, managers select the most popular regulation $A$ against the status quo $A_{t-1}$, with voluntary disclosure thresholds $\tau ^A$ and $\tau ^{t-1}$, respectively. We consider next several scenarios for the choice of $A$.

The first scenario involves a new regulation $A$ such that the non-disclosure price increases relative to the status quo $A_{t-1}$. For this to hold, the regulation $A$ must feature more mandatory disclosure than the status quo, implying $A> A_{t-1}$. Below, we analyze the preference of a manager with continuation value $v$.

• (a) Disclosers under both regulations, with $v\notin [A_{t-1},\tau ^{A})$ or $v\in [\tau ^{t-1},A)$, are indifferent;

• (b) disclosers under $A$ but not $A_{t-1}$, with $v\in [A_{t-1},\min (A,\tau ^{t-1}))$, prefer the status quo because, they retain and exercise the option to withhold information;

• (c) nondisclosers under $A$, with $v\in [A,\tau ^A]$ prefer $A,$ because they achieve a higher non-disclosure price.

These three regions are represented in Fig. 3. The net popularity of $A$ over $A_{t-1}$ is thus given by the fraction of shaded firms in region (c) minus the fraction of striped firms in region (b). As $A$ moves away from $A_{t-1}$, the number of firms in favor of the change (region a) decreases, and the number of firms that prefer the status quo (region b) increases. Therefore, the maximal net popularity is achieved by a regulation with $A$ set arbitrarily close to $A_{t-1}$, which achieves the objective of increasing the non-disclosure price with a minimal fraction of new disclosers (that oppose). As noted earlier, we interpret this situation in the limit as one in which the status quo does not materially change. Note also that, at the point of standard-setting, firms do not yet know their liquidation payoff $\theta$: they may either liquidate or continue, whichever yields the highest payoff. Given that liquidation implies a fixed payoff that does not depend on the standard, they always prefer standards that yield a higher continuation price.

Fig. 3

Preference for an alternative threshold disclosure $A$ versus a status quo disclosure threshold $A_{t-1}<A$

Things are different with a decrease in mandatory disclosure. When a new policy reduces the disclosure threshold strictly below the status quo, it is opposed by all nondisclosers. However, the policy also tends to be supported by all firms that had to disclose under the status quo but no longer have to disclose (see Fig. 4).

As $A<A^{t-1}$ is decreased, this new policy turns more disclosers into nondisclosers (shaded area in Fig. 4) and receives more support, while the opposing firms (striped area in Fig. 4) are constant. Indeed, the most preferred decrease in mandatory disclosure is one that features the greatest probability of non-disclosure for previously disclosing firms, which in our case corresponds to a complete removal of any mandatory disclosure $A=0$.

Fig. 4

Preference for an alternative threshold disclosure $A$ versus a status quo disclosure threshold $A_{t-1}>A$

In summary, the most popular reporting alternative will be one of two options—either maintain the status quo or do away with the mandatory disclosure altogether and return to an unregulated environment. Nondisclosers vote as a block and play a key role in this result. Specifically, complete deregulation maximizes the fraction of new nondisclosers (relative to the status quo), while a small increase in the regulation $A\approx A_{t-1}$, maximizes the fraction of nondisclosers with the constraint of increasing the non-disclosure price. In the next proposition, we compare the relative popularity of each of these alternatives.

Proposition 2.1

Let $\hat{A}=\max (1/2,4c/(4c+1))$ .

• (i) If $A_{t-1}\le {\hat{A}}$ (low levels of disclosure), the most popular standard is the status quo $Pop(A_{t-1})=A_{t-1}$ .

• (ii) If $A_{t-1}> {\hat{A}}$ (high levels of disclosure), the most popular standard is no disclosure $Pop(A_{t-1})=0$ .

Proposition 2.1 describes the main economic drivers of our study. status quo nondisclosers tend to support increases in the disclosure threshold, while status quo disclosers tend to support reducing disclosure requirements. As a result, nondisclosers form a majority when the status quo features low levels of disclosure. However, if the status quo features sufficiently high levels of disclosure, disclosers become more numerous than nondisclosers, and the alternative preferred by managers shifts from maintaining the status quo to no disclosure.

Corollary 2.1

The threshold ${\hat{A}}$ on the status quo, above which no disclosure becomes the most popular option, is nondecreasing in $c$ .

The comparative statics in the cost $c$ may seem counterintuitive to the extent that, intuitively, one could expect less disclosure to become more appealing in the presence of greater cost. On the contrary, a greater disclosure cost shifts the threshold ${\hat{A}}$ to the right, and therefore no disclosure $A=0$ is collectively preferred over a smaller set of status quo standards when $c$ increases.

To understand this property, recall that the political process does not directly weight the expected market price as an objective so that the relevant argument is not that it is socially desirable to reduce disclosure in the presence of higher cost. Instead, the key argument is that status quo nondisclosers—which benefit from higher market prices—are the group that typically supports more disclosure. Hence, an increase in the size of the non-disclosure group tends to increase the demand for more mandatory disclosure. Within this logic, a greater disclosure cost will reduce the amount of voluntary disclosure implying, for any status quo, an increase in the size of the non-disclosure group.

3 The standard-setter’s proposal

We analyze next the standard-setter’s proposal stage at $t.2(a)$. This stage is composed of two decision nodes. First, at $t.2(a.i)$, the standard-setter issues a new regulation $A$. Since, in this model, the standard-setter would never propose a regulation that is certain to fail, the proposal can be restricted to satisfy $Opp(A,A_{t-1})\le \alpha$. Second, at $t.2(a.ii)$, managers decide whether to oppose $A$, expecting that, if $A$ fails, the most popular standard $Pop(A_{t-1})$ will be implemented (as described in Proposition 2.1). We will show that a status quo could never reach above the level preferred by the standard-setter $A^*$, so we save space by focusing here on the case of $A_{t-1}\in [0,A^*]$.

Again, we proceed by backward induction to derive the opposition at $t.2(a.ii)$. Because $A_{t-1}\in [0,A^*]$, the standard-setter wishes to increase the disclosure threshold. There are two cases to consider, illustrated in Fig. 5. If $A_{t-1} \le {\hat{A}}$, the status quo will be maintained if the proposal fails. Therefore, all nondisclosers under $A_{t-1}$ oppose any $A$ that would remove their discretion to withhold: opposition increases if the standard-setters’ proposal requires more events to be disclosed. If, on the other hand, $A_{t-1}>{\hat{A}}$, the economy will be unregulated if the standard-setter’s proposal fails. Therefore, all managers who would not disclose in the unregulated environment tend to oppose a proposal in which they must disclose.

Fig. 5

Opposition to standard-setter’s proposal $A$

The next proposition formalizes the political tension faced faced by the standard-setter. The more the standard-setter wishes to increase mandatory disclosure, the more managers begin opposing the proposal. Put differently, the analysis demonstrates that high levels of political accountability slow down standard-setting.

Proposition 3.1

For a given status quo $A_{t-1}\le A^*$ , the standard-setter implements a new regulation $A_{t}=\min (A^*,Pop(A_{t-1})+\alpha )$ . This disclosure threshold is increasing in the disclosure cost $c$ , decreasing in the political accountability $1-\alpha$ and, as long as $A_t<{\hat{A}}$ , increasing in the status quo $A_{t-1}$ . Furthermore, $A_{t}<A_{t-1}$ (i.e., the disclosure threshold is reduced) if and only if $A_{t-1}>{\hat{A}}$ .

As long as the status quo is not too large (below ${\hat{A}}$), managers refer to the status quo as the most preferred regulation. The standard-setter can spend up to $\alpha$ in political capital to increase the policy above the status quo. However, when the turning point ${\hat{A}}$ is passed, the manager-preferred regulation reverts to the unregulated economy $(A=0)$, and therefore the standard-setter can only increase the disclosure threshold relative to this new benchmark. As a result, the standard-setter must concede a reduction in mandatory disclosure to $A=\alpha$ under the threat that, doing otherwise, the proposal would be rejected, and lead to an entirely unregulated economy.¹² While, in the model, the economy never reaches a state of complete deregulation, the regulation $A_t=\alpha$ may feature very low levels of mandatory disclosure in cases where political accountability is very high.

Some simple comparative statics follow. If the standard-setter pushes for more disclosure or has more political independence, a greater level of mandatory disclosure will be required in the exposure draft. In the extreme case in which $\alpha \approx 0$ is very low, the standard-setter increases mandatory disclosure by a very small increment and cannot implement any major piece of legislation (unless $A>{\hat{A}}$, and the new legislation moves toward deregulation). When the cost of disclosure increases, more nondisclosers support the status quo, thus helping the standard-setter increase mandatory disclosure.

We conclude this section by examining a scenario in which, for an exogenous (unmodelled) reason, the status quo is greater than the standard-setter’s preferred threshold $A^*$. As an example, $A_{t-1}$ may be greater than $A^*$ if the default standard for a new transaction has branched out from some other standard, or there may be a structural break in the cost of disclosures (e.g., change in legal systems, more information technology), or a change in the preferences of the standard-setter or the constituencies it represents. Since this case is identical to the previous setting if $Pop(A_{t-1})=0$, we focus here on $Pop(A_{t-1})=A_{t-1}$ (or $A<{\hat{A}}$).

While the standard-setter will now want to decrease the disclosure threshold, doing so can be problematic. As shown earlier, decreases in the threshold are opposed by all nondisclosers, and therefore any $A<A_{t-1}$ generates an opposition given by:

$$\begin{aligned} Opp(A,A_{t-1})=\tau ^{t-1}-A_{t-1}. \end{aligned}$$

(3.1)

When $A_{t-1}$ is not too large, this term can be greater than $\alpha$, and therefore a standard-setter subject to high levels of accountability cannot pass any decrease in the disclosure threshold, even if she wishes to do so. This observation stands in contrast with increases in the disclosure threshold, in which some small increase relative to $Pop(A_{t-1})$ may generally be passed.

Proposition 3.2

Suppose that $A_{t-1}\in (A^*,{\hat{A}})$ (the status quo implies more disclosure than preferred by the standard-setter).

• (i) If $\alpha \ge \tau ^{t-1}-A_{t-1}$ , the standard-setter implements $A_t=A^*$ .

• (ii) Otherwise, the standard-setter maintains the status quo at $A_t=A_{t-1}$ .

In summary, high political accountability joint with a status quo featuring high disclosure levels creates a political standstill. Because of the pressure by status quo nondisclosers, the standard-setter cannot reduce the amount of disclosure. Then, the equilibrium level of disclosure may remain at levels that the standard-setter views as excessive but is politically unable to change.

4 Evolution of mandatory disclosure

We now use the predictions obtained in each period $t$ to examine the dynamics of disclosure regulations. The sequence of regulatory outcomes is denoted $\{A_t\}$ with initial condition $A_0=0$ and the updating rule described in Proposition 3.1.

Several scenarios may occur. One scenario is that the standard-setter does not wish to implement too much disclosure $A^*\le {\hat{A}}$. Then, the deregulation region “$A_{t-1}\ge {\hat{A}}$” is never reached, and the standard-setter can always attain the preferred policy. If political accountability is high, reaching $A^*$ is a slow process that requires many periods of regulation.

A second scenario is that $A^{*}>{\hat{A}}$ if, for example, the standard-setter has a preference for high levels of transparency. Then, the standard-setter will increase the threshold gradually, until $A_{t-1}>{\hat{A}}$ is reached. Then, the economy reverts to being (nearly) unregulated and a new cycle begins.

Proposition 4.1

The regulatory process $\{A_t\}$ has the following properties:

• (i) If the standard-setter prefers low levels of disclosure (i.e., $A^*\le {\hat{A}}$ ), $A_t=min(\alpha t,A^*)$ is increasing in $t$ and converges to $A^*$ .

• (ii) If the standard-setter prefers high levels of disclosure (i.e., $A^*>{\hat{A}}$ ), $A_t$ is nonmonotonic and features cycles of length $k=[{\hat{A}}/\alpha ]+1$ , decreasing in $\alpha$ , whereby for any $n\ge 0$ and $t\in [1,k]$ , $A_{n k+t}=A_t=min(\alpha t,A^*)$ .

Figure 6 illustrates a regulatory process for each scenario. The standard-setter pushes toward $A^*$, increasing the threshold by $\alpha$ in each period. If this is sufficient to attain the standard-setter’s preferred regulation $A^*$, as on the left side of the figure, the regulatory process settles for the long run. On the right-hand side, an example is given in which $A^*>{\hat{A}}$. When the process reaches above ${\hat{A}}$, the regulation will revert back to $A_1$. For instance, a period $t-1$ standard-setter may have been able to set the standard at $A_{t-1}>{\hat{A}}$. But then the period $t$ standard-setter inherits a standard that cannot be sustained. The most popular disclosure rule among firms is no disclosure at all, and the period $t$ standard-setter must limit the reduction in disclosure by proposing $A_1$.

Fig. 6

The regulatory process: convergence versus cycles

Another point worth emphasizing is that cycles can be very long if $\alpha$ is low. In particular, deregulation will be more intense when it follows after a longer period of increased regulation and when the standard-setter increased the regulation only slowly (since the deregulation that occurs at the end of each cycle is to $A_1=\alpha$).

A closer inspection of the threshold ${\hat{A}}$ will reveal an important fact about cycles in our model. Although conceptual statements do not explicitly mention price-maximization as an objective of standard-setting, a useful welfare benchmark arises if the standard-setter simply acts in the best interest of diversified investors $A^*=A^{fb}$ where $A^{fb}$ maximizes the expected value of a firm. Formally, $A^{fb}$ is the mandatory disclosure threshold that maximizes the expected value (over $v$ and $\theta$) of the maximum of the liquidation value $\theta$ and the continuation value, which is either equal to $P^t(ND)$ or $P^t(v)=v-c$ depending on the disclosure,

$$\begin{aligned} A^{fb}= argmax_{A_t} \quad {\mathbb {E}}(\max (\theta ,1_{v\in [A_t,\tau ^t]}P^t(ND)+1_{v\notin [A_t,\tau ^t]}P^t(v)). \end{aligned}$$

(4.1)

Corollary 4.1

The first-best disclosure threshold $A^{fb}$ is always lower than ${\hat{A}}$ . Hence, a standard-setter that maximizes value to investors, with $A^*=A^{fb}$ , will not induce regulatory cycles.

Our results thus suggest that a standard-setting body that is primarily controlled by diversified investors provides an additional side benefit to the regulatory process. This standard-setter will not issue standards that will be later rescinded during regulatory cycles.

Observe that Proposition 4.1 is derived under the (debatable) assumption of a myopic standard-setter. However, myopic standard-setting is suboptimal for a patient standard-setter if excessive increases in the regulation trigger cycles. Fortunately, the argument can be easily extended to a scenario in which the standard-setter has a multi-period objective function. Assume that the standard-setter has a separable utility function at date $t$ given by $U^t$ where:

$$\begin{aligned} U^t=\sum _{t'=t}^{+\infty }\beta ^{t'-t} {\mathcal {U}}(A_t), \end{aligned}$$

where $\beta \in [0,1)$ is the standard-setter’s discount rate with $\beta =0$ corresponding to the myopic standard-setter discussed in the baseline model.

The case in which $A^*\le {\hat{A}}$ is straightforward. As shown in Proposition 4.1 [case (i)], this is a situation in which the sequence of policies $\{A_t\}$ chosen by the myopic standard-setter converges to $A^*$. Since the myopic standard-setter already increases $A_t$ as much as political pressures allow it each period, these policy choices remain optimal for any discount factor $\beta$.

If $A^*>{\hat{A}}$, a different course of action might be optimal as a forward-looking standard-setter may strategically avoid cycles. To begin with, note that the forward-looking standard-setter will still propose $A_{t+1}=A_{t}+\alpha$ as long as $A_t+\alpha <{\hat{A}}$ and the status quo that would start a cycle is not yet reached. Things are different when the critical status quo $A_t$ is reached such that $A_t\le {\hat{A}}$ but $A_t+\alpha >{\hat{A}}$. At this point, the forward-looking standard-setter must make a choice over two possible options: (a) implement $A_{t+1}=min(A^*,A_{t}+\alpha )$ and trigger a cycle in the next period, or (b) implement $A_{t+1}={\hat{A}}$ and stabilize the regulation at ${\hat{A}}<A^*$ for all future periods.

Proposition 4.2

Let $\Lambda$ be defined as:

$$\begin{aligned} \Lambda ={\mathcal {U}}(\min (A^*,\alpha k))-{\mathcal {U}}({\hat{A}})+\sum _{n=1}^{k-1}\beta ^n ({\mathcal {U}}(\alpha n)-{\mathcal {U}}({\hat{A}})). \end{aligned}$$

(4.2)

• (i) If $A^*\le {\hat{A}}$ or $\Lambda >0$ , the standard-setting dynamics will be identical to the baseline in Proposition 4.1.

• (ii) Otherwise, the standard-setter will implement $A_{t+1}=min(A_{t}+\alpha ,{\hat{A}})$ , and the policy will always stabilize at ${\hat{A}}$ in the long run.

A forward-looking standard-setter will evaluate the current benefit of passing a high policy $A_t>{\hat{A}}$ against the future losses caused by the regulatory cycle. This may imply that an intermediate policy set at ${\hat{A}}$ becomes attractive. In this case, the standard-setter does not achieve her preferred policy $A^*$ even in the long run.

Note that, while an impatient standard-setter never stabilizes, a fully patient standard-setter (when $\beta$ converges to one) may also opt not to stabilize. For example, if ${\mathcal {U}}(min(A_s,\alpha k),\gamma )-{\mathcal {U}}({\hat{A}},\gamma )$ is large, $\Lambda >0$ will be positive for any discount factor. On the other hand, stabilization is optimal if the cost of triggering a new cycle is large enough. This only occurs when $\alpha$ is small, so that, once a new cycle begins, it takes a large number of periods to increase the policy toward ${\hat{A}}$. Hence, a forward-looking standard-setter with high levels of political accountability generally tends to favor stabilization.

5 Discussion

In the preceding sections, we analyzed the model under several assumptions that allow us to lay out the core intuitions in their simplest form. We develop here some further discussion points that are relevant in richer economic environments.

Uninformed participants. In the baseline model, firms that can participate in the political process must be endowed with information, so that uninformed managers (or diversified investors) may only be represented via the standard-setter’s actions and preferences. The model has similar dynamics if we assume that there is a proportion of firms active in the political process that are uninformed. In this case, uninformed managers vote as a group in favor of standards closer to $A^{fb}$. In turn, this tends to cause an interval of standards located around $A^{fb}$ where the policy can settle. Then, the standard-setter can no longer increase the threshold because doing so would be opposed by all uninformed managers. Hence, when the probability of not being informed is large enough, the policy may not settle at $A^*$ or cycle but instead will settle at some level between $A^{fb}$ and $A^*$.

Distributional assumptions. The main result on cycles is robust to a more general specification of the cash flow distribution. Specifically, even if distributions are not uniformly distributed, the opposition to a standard will increase as the standard-setter elevates the proposal too far above the status quo (i.e., nondisclosers oppose new requirements in which they have to disclose) and a decrease in the disclosure threshold must be large enough so that enough firms that no longer disclose under the new standard support it. Nevertheless, a few observations in the model are specific to the uniform. First, the fact that the standard would increase by fixed increments is specific to the flat density of the uniform distribution; under other bell-shaped distributions, for example, the threshold would increase by fixed “probability mass” increments, faster in the tails where the density is thin and few firms oppose and slower near the mean where more firms oppose. Second, the disclosure threshold falls toward no mandatory disclosure when a new cycle begins under the uniform distribution; it will fall by a large amount as well with more general distributions but only up to the level that would maximize the total fraction of nondisclosers. In general, this level need not be complete deregulation because no disclosure might entail a significant amount of voluntary disclosure. Third, in the baseline model, the level that maximizes the market price is always below the cycling threshold. This may or may not be true for more general distributions, and, in particular for distributions that are skewed, the cycling region may even be reached before the ex ante preferred is reached.

Other real effects. We have focused on a simple liquidation decision as the real effect but the results would be similar if we assumed a post-disclosure real effect since the main argument follows from only two forces, both of which would still hold with real effects, that: (1) firms forced to disclose are weakly worse off since they could do so voluntarily, and (2) nondisclosers benefit with a standard that features a higher threshold $A$. To the extent that a general model would change the distribution of cash flows, many of the incremental forces with production would be similar to those with general distributions, as discussed above.

Time-varying environment. As in any model featuring multi-period dynamics, we have focused the baseline on the main variable of interest, the disclosure threshold as a moving part. Similar predictions can be inferred from the analysis for various shocks to fundamentals, and we discuss a few. If, for example, the quality of projects were to vary, then there would be more demand to reduce the disclosure threshold during periods with fewer high-quality projects (recessions) and, vice versa, demand to increase the disclosure threshold during periods with fewer low-quality projects (expansions), as in Bertomeu and Magee (2011). As another possibility, the political independence of the standard-setter $\alpha$ might randomly change across periods, possibly in tandem with changes in fundamentals. This would cause the standard-setter to possibly attain the preferred level $A^*$ during periods of high independence only to trigger deregulation during a period of low independence.

Proposal game. In the baseline model, we assume that, once a standard-setter’s proposal fails, a new regulator makes the most popular proposal. The conceptual results would be similar if, at this stage, we assume that a new proposal is selected from the set of proposals that obtain a majority $M=\{A:Net(A,A_{t-1})\ge .5\}$ according to some decision rule (provided that, if this set includes a sufficiently small subset of values $A>A_{t-1}$, the decision rule must be below $A_{t-1}$). For example, another possibility would be to use a Baron and Ferejohn (1989) random proposer game (i.e., a standard proposer is drawn randomly and can make a proposal in $M$). This type of model would feature stochastic dynamics, as the identity of the proposer would vary, with possibly random increases and a random date of a fallback to a lower threshold. Such a one-step random proposer game, by partly taking away agenda-setting power from the standard-setter, would also tend to make regulatory cycles more likely (and less predictable).¹³

6 Concluding remarks

Financial reporting standard setters strive to achieve a balance between independent assessment of the benefits of reporting changes and the variety of viewpoints presented by interested parties. For instance, the FASB (2009, p. 2) describes the following as one of its precepts:¹⁴

To weigh carefully the views of its constituents in developing concepts and standards: However, the ultimate determinant of concepts and standards must be the Board’s judgment, based on research, public input, and careful deliberation about the usefulness of the resulting information.

Notwithstanding standard-setters” objective of independence, there are times when standard setting bodies are subject to political pressure and when that pressure affects the standards that are adopted. Zeff (2005) chronicles the political forces that have affected U.S. GAAP, from allowing LIFO inventory accounting to accounting for the investment tax credit to the expensing of employee stock options. Beresford (2001) describes the U.S. congressional activities surrounding the accounting for acquisitions, and he recounts the pressures encountered by the FASB from companies and from members of Congress. He concludes: “Congressional oversight is an essential part of our society and our economic environment. Although we may disagree with the motives of some of the parties who avail themselves of this opportunity, few of us favor a system where a group like the FASB is accountable to no one.”

How might political pressures affect the evolution of accounting standards? Distinctive to our approach is to place the standard-setting institution as a strategic agent subject to objectives and constraints: regulation emerges endogenously as a result of trade-offs between meeting those objectives and responding to opportunistic political pressures. Reporting firms always have the option to disclose voluntarily, so they oppose any requirements that decrease their discretion. Increases in required disclosure proceed more slowly when the standard-setter is less politically influential or when greater disclosure costs imply greater political resistance by reporting firms. In addition, there is a critical point in the disclosure regulation at which the reporting firms prefer to eliminate all regulation, perhaps forcing a fallback to low disclosure requirements. Such regulatory cycles, when they occur, would take the form of steady increases in disclosure, punctuated by bursts of deregulation. We hope that examining the economic forces at play provides one first step furthering the understanding of accounting regulation, and that future research in this domain will extend this paradigm to other dimensions of accounting regulation.

Appendices

Appendix 1: Omitted proofs

Proof of Lemma 1.1

Let $A_t$ be the implemented regulation at date $t$. The probability of disclosure $p_d$ is given by:

$$\begin{aligned} p_d^t=A+1-2 c(1-A_t)-A_t=2 c A_t + 1- 2 c. \end{aligned}$$

It follows that $p_d^t$ is increasing in $A_t$. Solving for the non-disclosure price $P^t(ND)$,

$$\begin{aligned} P^t(ND)=\frac{1}{2}(A_t+\tau ^t)-c A_t=A_t+c(1-A_t)-c A_t=(1-2c)A_t+c. \end{aligned}$$

This function is increasing $A_t$. $\square$

Table 1 Main notations

Proof of Proposition 2.1

We know from the analysis in text that we need to compare $Net(A_{t-1},A_{t-1})=\lim _{\epsilon \rightarrow 0^+}Net(A_{t-1}+\epsilon ,A_{t-1})$ to $Net(0,A_{t-1})$. Note that $Net(A_{t-1},A_{t-1})-Net(0,A_{t-1})$ is strictly decreasing in $A_{t-1}$ so that there exists a threshold ${\hat{A}}$ such that $Pop(A_{t-1})=1_{A_{t-1}\le {\hat{A}}}A_{t-1}$. We determine this threshold next as $Net({\hat{A}},{\hat{A}})=Net(0,{\hat{A}})$.

Case 1 Suppose that ${\hat{A}}\le 2c$.

$$\begin{aligned} Net({\hat{A}},{\hat{A}})&= Net(0,{\hat{A}}), \\ (1-{\hat{A}})2c&= {\hat{A}}-(1-{\hat{A}})2c,\\ 4 c&= {\hat{A}}(4 c +1), \\ \frac{4 c}{4 c+1}&= {\hat{A}}. \end{aligned}$$

Verifying that ${\hat{A}}=4c/(4c+1)\le 2c$ requires that $c\ge 1/4$.

Case 2 Suppose ${\hat{A}}> 2c$.

$$\begin{aligned} Net({\hat{A}},{\hat{A}})&= Net(0,{\hat{A}}) \nonumber \\ (1-{\hat{A}})2c&= 2c-(1-{\hat{A}})2c \nonumber \\ 1/2&= {\hat{A}}. \end{aligned}$$

(7.1)

For $1/2>2c$, one must have that $c<1/4$.

In summary, we have demonstrated that ${\hat{A}}=\max (1/2,4c/(4c+1))$. $\square$

Proof of Proposition 3.1

Recall that we focus here on $A_{t-1}\le A^*$ so that the standard-setter prefers the maximal feasible regulation, up to $A^*$. Define $A_{max}$ as the maximum regulation that would pass and let us solve for $A_{max}$.

Suppose that $A_{t-1}\le {\hat{A}}$. Then, the most popular regulation is $Pop(A_{t-1})=A_{t-1}$. It follows that for any proposed policy $A>A_{t-1}$,

$$\begin{aligned} Opp(A,A_{t-1})=\min (2c (1-A_{t-1}),A-A_{t-1}). \end{aligned}$$

Because $\alpha \le 2c(1-{\hat{A}})$, $Opp(A_{max},A_{t-1})<2c(1-A_{t-1})$. Therefore, $A_{max}=\min (1,A_{t-1}+\alpha )$.

Suppose that $A_{t-1}>{\hat{A}}$. Then, the most popular regulation is $Pop(A_{t-1})=0$. It follows that for any proposed policy $A>0$,

$$\begin{aligned} Opp(A,A_{t-1})=min(2c,A) \end{aligned}$$

Therefore, $A_{max}$ is given by $A_{max}=\alpha$.

It then follows that the standard-setter’s optimal proposal (which passes) is:

$$\begin{aligned} A_{t}= \min (Pop(A_{t-1})+\alpha ,A^*)=1_{A_{t-1}\le {\hat{A}}} min(A^*,A_{t-1}+\alpha )+1_{A_{t-1}>{\hat{A}}}min(A^*,\alpha ). \end{aligned}$$

Note that $A_{t}$ is increasing in $Pop(A_{t-1})$, $\alpha$ and $A^*$. $\square$

Proof of Proposition 3.2

There are two cases to consider, depending on whether $A_{t-1}\le {\hat{A}}$ (case 1) or $A_{t-1}>{\hat{A}}$ (case 2).

Case 1 If $A_{t-1}\le {\hat{A}}$, the policy that passes if the standard-setter’s proposal fails is the status quo $A_{t-1}$. It follows that all nondisclosing managers oppose any decrease in $A$, and therefore (any) $A<A_{t-1}$ can be passed if and only if $\alpha \ge (1-A_{t-1})2c$. Since $\alpha \le \overline{\alpha }\le (1-{\hat{A}})2c \le (1-A_{t-1})2c$, it follows that no policy $A<A_{t-1}$ can be passed.

Case 2 If $A_{t-1}> {\hat{A}}$, the policy that passes if the standard-setter’s proposal fails is no disclosure. It follows that the standard-setter can pass up to $A_{max}=\alpha$. This implies that $A_{t}=\min (A^*,\alpha ).$ $\square$

Proof of Corollary 4.1

Let $EP(A_t)$ be defined as the expected surplus conditional on an implemented regulation $A_t$.

$$\begin{aligned} EP(A_t)= \underbrace{\int _{0}^{A_t} \int _{0}^1 \max (\theta ,v-c) d\theta dv}_{K_1}+\underbrace{\int _{A_t}^{{\tau ^t}}\int _{0}^1 \max (\theta ,P^t(ND)) d\theta dv}_{K_2}+\underbrace{\int _{{\tau ^t}}^{1}\int _{0}^1 \max (\theta ,v-c)d\theta dv}_{K_3}. \end{aligned}$$

Examining each term in the above expression,

$$\begin{aligned} \frac{\partial K_1}{\partial A_t}&= \int _0^1 \max (\theta ,A_t-c)d \theta \\&= 1_{A_t\le c}\int _0^1 \theta d\theta +1_{A_t>c} \int _0^1 \max (\theta ,A_t-c)d \theta \\&= 1_{A_t\le c}\frac{1}{2} +1_{A_t>c}\left( \int _0^{A_t-c} (A_t-c)d\theta +\int _{A_t-\theta }^1 \theta d\theta \right) \\&= 1_{A_t\le c}\frac{1}{2} +1_{A_t>c}\left( (A_t-c)^2 +\frac{1}{2}-\frac{1}{2}(A_t-c)^2\right) \\&= 1_{A_t\le c}\frac{1}{2} +1_{A_t>c}\left( (A_t-c)^2 +\frac{1}{2}-\frac{1}{2}(A_t-c)^2 \right) \frac{1}{2} (A_t-c)^2+\frac{1}{2}\\&= 1_{A_t\le c}\frac{1}{2} +1_{A_t>c}\left( \frac{A_t^2}{2}-c A_t +\frac{1}{2}+\frac{c^2}{2}\right) \\&= \frac{1}{2}+1_{A_t>c} \frac{1}{2}(A_t-c)^2. \end{aligned}$$

Next,

$$\begin{aligned} K_2&= \int _{A_t}^{{\tau ^t}} \int _{0}^1 \max (\theta ,c+A_t(1-2c)) d\theta dv\\&= 2 c (1-A_t) \int _{0}^1 \max (\theta ,c+A_t(1-2c)) d \theta \\&= 2 c (1-A_t)\left( \int _0^{c+A_t(1-2c)} (c+A_t(1-2c)) d \theta +\int _{c+A_t(1-2c)}^1 \theta d\theta \right) \\&= 2 c (1-A_t)\left( (c+A_t(1-2c))^2+\frac{1}{2}-\frac{1}{2}(c+A_t(1-2c))^2\right) \\&= c(1-A_t) \left( 1+(c+A_t(1-2c))^2\right) \\ \frac{\partial K_2}{\partial A_t}&= -3 A_t^2 (1 - 2 c)^2 c + 2 A_t c (1 - 6 c + 8 c^2) - c (1 - c (2 - 5 c)). \end{aligned}$$

And, similarly,

$$\begin{aligned} \frac{\partial K_3}{\partial A_t}&= -\frac{\partial \tau ^t}{\partial A_t}\int _0^1 \max (\theta ,\tau ^t-c)d\theta \\&= -(1-2c)\int _0^1 \max (\theta ,A_t (1-2c)+ c)d\theta \\&= -(1-2 c)\frac{1}{2}((c+A_t(1-2c))^2+ 1) \\&= -\frac{1}{2} A_t^2 (1 - 2c)^3 - A_t (1 - 2c)^2 c - \frac{1}{2}(1 - 2c) (c^2 + 1) \\&= -\frac{1}{2} (1 - 2 c) (1 + (c + A_t (1 - 2 c))^2). \end{aligned}$$

Then,

$$\begin{aligned} EP'(A_t)=A_t^2\frac{1}{2} (1_{A_t > c} - (1 - 2c)^2 (1 + 4c)) + A_t c (1 - 4c (2 - 3c) - 1_{A_t > c}c ) + \frac{1}{2}c^2 (3 + 1_{A_t > c} - 8c). \end{aligned}$$

Case 1 Assume that $c<1/4$.

Consider $A_t\in (0,c)$. In this region, $EP'(.)$ is inverse U-shaped with:

$$\begin{aligned} EP'(0)&= \frac{1}{2}c^2 (3-8c)>0 \\ EP'(c)&= 2 (1 - c)^2 c^2 (1 - 4c)>0. \end{aligned}$$

It follows that $EP'(.)>0$ on $(0,c)$, and therefore $A^{fb}\ge c$.

Consider next $A_t\in [c, 1)$. In this region, $EP'(.)$ is U-shaped with:

$$\begin{aligned} EP'(1)&= 0 \\ EP''(1)&= 4 (1-c)c^2. \end{aligned}$$

Note that $A_t=1$ satisfies the first-order condition for an optimum but is not the desired solution, as $EP''(1)>0$ implies that it is a local minimum.

As $EP'(.)$ is a quadratic U-shaped function, we know that $EP'(.)$ decreases then increases on $(c,1)$, and therefore there is a unique solution in $(c,1)$ that satisfies $EP'(A^{fb})=0$. Factorizing the polynomial $EP'(.)$ by observing that one of its roots is $A_t=1$,

$$\begin{aligned} EP'(A_t)=2 (1 - A_t)c^2 (1 - 2c - A_t (3 - 4c)). \end{aligned}$$

The second root (which is the only root that satisfies the second-order condition) is then given by:

$$\begin{aligned} A^{fb}=\frac{1-2c}{3-4c}. \end{aligned}$$

Case 2 Assume that $c \in [1/4, 3/8)$.

Consider $A_t\in (0,c]$. In this region, $EP'(.)$ is inverse U-shaped with:

$$\begin{aligned} EP'(0)&= \frac{1}{2}c^2 (3-8c)>0 \\ EP'(c)&= 2 (1 - c)^2 c^2 (1 - 4c)<0. \end{aligned}$$

It follows that $EP'(.)$ has a unique root on $(0,c)$, which satisfies the second-order condition (i.e., $EP''<0$). Consider next $A_t\in (c, 1)$. In this region, $\sigma '(.)$ is U-shaped with (as before) $EP'(1)=0$. It follows that $EP'<0$ for any $A_t\in (c,1)$.

Therefore, the policy $A^{fb}$ is in $(0,c)$ and, solving $EP'(A^{fb})=0$, is given by the equation below.

$$\begin{aligned} A^{fb}= \frac{c (8c - 3)}{8c^2 - 2c - 1}. \end{aligned}$$

Case 3 Assume that $c\ge 3/8$.

$$\begin{aligned} EP'(0)-EP'(c)=\frac{1}{2}c^2 (1 - 2c) (2c (7-4c)-1)>0 \end{aligned}$$

This implies that $EP'<0$ on $A_t\in (0,c)$. As in case 2, $EP'<0$ on $(c,1)$ and $A^{fb}=0$. $\square$

Proof of Proposition 4.2

The case with $A^*\le {\hat{A}}$ is already explained in text so that let us assume here that $A^*>{\hat{A}}$. The forward-looking standard-setter will implement $A_{t}=A_{t-1}+\alpha$ as long as $A_{t-1}+\alpha \le {\hat{A}}$ and, when $k$ such that $A_{k-1}+\alpha > {\hat{A}}$ is reached, may either set $A_{k}=\min (A^*,A_{k-1}+\alpha )$ (in which case the regulatory dynamics will be identical to the baseline) or $A_{t}={\hat{A}}$ for any $t\ge k$.

Define $U_{cycle}$ as the surplus when the standard-setter chooses to cycle (first option) and $U_{stab}$ as the surplus when the standard-setter chooses to stabilize at ${\hat{A}}$ (second option). Let us define $k$ as the duration of a cycle if the first option is chosen, where [.] indicates the integer part.

A cycling policy visits states $\alpha , 2\alpha ,\ldots , \min (A^*,A_{k-1}+\alpha )$ and repeats, which implies that:

$$\begin{aligned} U_{cycle}=\frac{1}{1-\beta ^{k}}({\mathcal {U}}(min(A^*,\alpha k)) +\sum _{n=1}^{k-1}\beta ^n {\mathcal {U}}(\alpha n)). \end{aligned}$$

In the equation above, the payoff obtained along one cycle ${\mathcal {U}}(min(A^*,\alpha k))+\sum _{n=1}^{k-1}\beta ^n {\mathcal {U}}(\alpha n)$ is repeated as a perpetuity with a discount rate $\beta ^{k}$ given that each cycle lasts for $k$ periods.

On the other hand, stabilizing the policy at ${\hat{A}}$ implies a constant surplus

$$\begin{aligned} U_{stab}=\frac{{\mathcal {U}}({\hat{A}})}{1-\beta }. \end{aligned}$$

It then follows that $U_{cycle}< U_{stab}$ if and only if

$$\begin{aligned} \Lambda =\underbrace{{\mathcal {U}}(\min (A^*,\alpha k))-{\mathcal {U}}({\hat{A}})}_{>0}+\sum _{n=1}^{k-1}\beta ^n \underbrace{({\mathcal {U}}(\alpha n)-{\mathcal {U}}({\hat{A}}))}_{<0}<0. \end{aligned}$$

The function $\Lambda$ is decreasing in $\beta$ and in $\alpha$. $\square$

Appendix 2: Other disclosure regulations

This appendix proves two claims. First, all managers weakly prefer regulations in which favorable events are not subject to mandatory disclosure (Lemma C.1). Second, threshold regulations in which events $v< A$ are subject to a mandatory disclosure maximize popularity (Lemma C.4).

We generalize the notations and assumptions used in the baseline model to describe non-interval type of disclosure rules. Define a standard as an indicator function h: $[0,1]\rightarrow \{0,1\}$ such that $h(v)=1$ (resp., $h(v)=0$) indicates that the event $v$ is not subject to mandatory disclosure (resp., must be disclosed). The voluntary disclosure threshold is denoted $\tau ^h$, and the non-disclosure price is denoted $P^h(ND)$.

Disclosing firms bear a cost $c>0$. Nondisclosers bear a cost $c\phi (P^h(ND))\ge 0$ where $P^h(ND)=\int _0^{\tau ^h} h(v)vdv/\int _0^{\tau ^h} h(v)dv$ is the gross non-disclosure price (excluding costs); $\phi (1)=1$; and $0 \le \phi '< 1/c$.¹⁵ As is well known, the voluntary disclosure threshold satisfies the following equation:

$$\begin{aligned} P^h(ND)- c \phi (P^h(ND))= \tau ^h-c. \end{aligned}$$

(8.1)

If there is more than one solution, we choose the highest solution because it is Pareto dominant from the perspective of managers. Note that we parameterize the cost in terms of the non-disclosure price that nests the baseline model (see footnote 13) and provides tractability to the model if the mandatory disclosure region features multiple disjoint sets.

We restrict the attention to regulations in which $ND^h=\{v:h(v)=1, v\le \tau ^h\}$ is empty or can be written as a finite union of closed intervals. The probability of non-disclosure is denoted $q^h=\int _0^{\tau ^h}h(s)ds$. In shorthand, denote $h_A$ for the function $h_A(v)=1-1_{v< A}$ (this is the baseline threshold regulation). With a slight abuse of notation, we use $P^A(ND)$ instead of $P^{h_A}(ND)$ and use this shorthand notation in other places where $h_A$ would appear as a superscript. All statements are made up to events with probability zero.

1.1 Non-disclosure of favorable events

This section demonstrates several observations that are useful in proving the main results.

Lemma C. 1

Let $h_1$ be such that $ND^h$ has a maximal nonempty interval $[x,y]$. Then, a standard $h_2$ such that $ND^{h_2}=ND^{h}\cup [x,\tau ^h]$ is weakly preferred by all managers, strictly so by managers with $v\in (y,\tau ^h]$.

Proof

This follows from the following comparison between $h_1$ and $h_2$: (a) managers with $v\notin ND^{h_2}$ are indifferent; (b) managers with $v \in (y,\tau ^h]$ (strictly) prefer $h_2$ since they could have disclosed voluntarily; (c) managers with $v\in ND^{h_1}$ prefer $h_2$ because they obtain a higher non-disclosure price under $h_2$. $\square$

Lemma C.1 implies that we can restrict the attention to regulations in which $\max ND^h=\tau ^h$. In particular, if $ND^h$ is an interval, it must have the threshold form $h_A$ for some $A$.

1.2 Popularity of threshold regulations

As we solve the model by backward induction, we initially examine the second phase of the regulatory game and derive the standard $h$ that is the most popular against an existing status quo $h_A$.

We first establish two preliminary lemmas.

Lemma C. 2

Let there be two standards $h$ and $h_A$. If $P^h(ND)\ge P^A(ND)$, then $q^{h}\le \tau ^{A}-A$ (strictly if $h\ne h_A$).

Proof

We solve for the standard that maximizes the probability of non-disclosure subject to $P^h(ND)\ge P^A(ND)$.

$$\begin{aligned} \max _{q^h,\tau ^h,h(.)} q^h \end{aligned}$$

s.t.

$$\begin{aligned}&P^h(ND)- c \phi (P^h(ND))=\tau ^h-c \quad (l_{\tau }) \\&\tau ^h-c\ge P^A(ND) \quad (l_A) \\&q^h=\int _0^{\tau ^h} h(v)dv \quad (l_q) \\&P^h(ND)=\frac{\int _0^{\tau ^h} v h(v)dv}{q^h}\quad (l_P) \end{aligned}$$

Differentiating the Lagrangian $L$:

$$\begin{aligned} \frac{\partial L}{\partial q^h}=1+l_q+l_P\frac{P^h(ND)}{q^h}=0 \end{aligned}$$

(8.2)

And, for $v<\tau ^h$,

$$\begin{aligned} \frac{\partial L}{\partial h(v)}=-l_q-l_P\frac{v}{q^h}=-l_P\frac{v}{q^h}-1-l_P \frac{P^h(ND)}{q^h} \end{aligned}$$

(8.3)

A standard $h(v)=0$ for all $v$ cannot be a solution (it achieves $q^h=0<q^A$); therefore $l_P<0$. This function intersects zero at most once, from below, implying that the solution has the form $h_{A'}$ where, as $q^{A'}$ is decreasing in $A'$, implies that the solution is $h_A$.

Lemma C. 3

For any standard $h\ne h_0$, $q^{h}<q^0$.

Proof

If $P^h(ND)>P^0(ND)$, this statement follows from Lemma C.2. If $P^h(ND)=P^0(ND)$, $\tau ^h=\tau ^0$, which implies that $q^h\ge q^0=\tau ^0$ with equality if and only if $h=h_0$. If $P^h(ND)<P^0(ND)$, $\tau ^h<\tau ^0$, which also clearly implies $q^{h}<q^0$. $\square$

As in the baseline model, denote the (net) popularity of a standard $h$ over $h_A$ by $Net(h,h_A)$.

Lemma C. 4

For any $A$, $h_0$ or $h_A$ maximizes popularity.

Proof

Consider a regulation $h$ in which $ND^h$ is composed of at least two disjoint intervals. We need to show that $Net(h,h_A)\le \max (q^0,q^A)$.

Case 1 Suppose that $P^h(ND)\ge P^A(ND)$. Lemma C.2 implies that $q^h<q^A$ and given that $Net(h,h_A)$ is bounded from above by $q^h$ (i.e., only nondisclosers under $h$ might prefer $h$), we know that $Net(h,h_A)\le q^A$.

Case 2 Suppose that $P^h(ND)<P^A(ND)$. Then:

$$\begin{aligned} Net(h,h_A)&= \int _0^{\min (A,\tau ^h)} h(v)dv-(\tau ^A-A)\\&\le \min (A,\tau ^0)-(\tau ^{A}-A)=Net(h_0,h_A) \quad (\hbox {by Lemma C.3}). \end{aligned}$$

It then follows that the regulations $h_0$ or $h_A$ maximize the function $Net(h,h_A)$.¹⁶ $\square$