Edited by Gerrit De Geest
Chapter 8: Discovery
[In: Volume 8, Chris William Sanchirico (ed) Procedural Law and Economics]
Discovery is the process by which parties obtain information from one another and from otherwise uncooperative third parties. The United States is unusual among the nations of the world for the extremely broad scope of discovery it allows (Subrin 1998). For example, the Federal Rules of Civil Procedure (FRCP) authorize discovery in federal court cases “regarding any matter, not privileged, that is relevant to the claim or defense of any party” and state that “relevant information need not be admissible at the trial if the discovery appears reasonably calculated to lead to the discovery of admissible evidence” (Rule 26). The FRCP also furnish a powerful set of discovery tools, including requests for production of documents (Rule 34); interrogatories (Rule 33); oral depositions (Rule 30); requests for admissions (Rule 36); and physical or mental examinations (Rule 35).
Broad discovery was first introduced into American litigation in 1938 as one of the most important innovations of the Federal Rules of Civil Procedure. Its purpose, as conceived by the FRCP drafters, was to resolve cases on the facts and evidence and avoid trial surprise (Subrin 1998). Broad discovery elicited few complaints for the first 40 years of its operation, but in the 1970s, judges, lawyers, and commentators began to criticize the system (Brazil 1978). Critics today complain that parties use broad discovery excessively and also sometimes abusively to impose costs on their opponents and leverage more favorable settlements.
Although there is anecdotal evidence to support these charges, reliable empirical evidence is limited (Easterbrook 1989; Mullenix 1994). According to several studies, serious discovery problems seem to be confined to a small set of complex, multi-party cases involving high stakes (Mullenix 1994; Willging et al. 1997; Kakalik et al. 1998; Garth 1998). For example, in a 1997 study based on a survey of about 1200 attorneys nationwide, the Federal Judicial Center found that “high levels of discovery problems and high expenses were more likely to occur in cases with high stakes, high levels of contentiousness, high levels of complexity, or p. 189high volumes of discovery activity” (Willging et al. 1997, at 2). Even so, this same study also found that investment in discovery constituted 50% of total litigation costs for those cases that had some discovery, and 90% of total costs for the top 5% of most expensive cases2 (Willging et al. 1997).
The following discussion reviews the economic literature on discovery. It first evaluates the benefits of discovery, then examines the costs, and concludes with a brief survey of some reform proposals.
2. The Social Benefits of Discovery
Proponents of broad discovery cite benefits for the quality of settlements and trial outcomes, for filing and other litigation incentives, and for incentives to comply with the substantive law. Before analyzing each of these benefits, it is first necessary to examine what, if any, information would be disclosed voluntarily in the absence of formal discovery. This initial inquiry defines the information-disclosure baseline against which the effects of discovery can be evaluated.
2.1. Voluntary Disclosure
Parties frequently share information voluntarily. For example, 46% of attorneys in the Federal Judicial Center’s study reported informal exchange in cases where no formal discovery or disclosure took place3 (Willging et al. 1997, at 13–14). The following discussion surveys the literature on voluntary disclosure, first relative to favorable information and then relative to unfavorable information.
2.1.1. Voluntary disclosure of favorable information
The economics literature contains numerous articles examining the voluntary revelation of private information in general (Milgrom and Roberts 1986), and the law-and-economics literature contains studies of voluntary disclosure incentives in litigation (Shavell 1989). One result is relatively obvious: a party will tend to disclose information favorable to its side early in the litigation, absent countervailing factors (Shavell 1989). Disclosing favorable information corrects for an opponent’s excessive optimism, which increases the likelihood of a successful settlement that is more favorable to the disclosing party. If the defendant discloses favorable information, p. 190for example, the plaintiff should reduce her estimate of likely trial success and thus her minimum demand. This drives the midpoint of the settlement range down, which, all other things equal, should yield a settlement more favorable to the defendant.
Three factors complicate this simple result. First, the opposing party must believe that the disclosure is truthful. However, disclosing parties have obvious incentives to exaggerate or misrepresent. Aware of these incentives, the opposing party will discount the truthfulness of disclosures he is unable to verify, and disclosing parties, who anticipate this response, will be less inclined to disclose when disclosure is costly (Hay 1994).
Second, in a world without formal discovery, parties might conceal favorable information to exploit the benefits of surprise at trial. But given the high costs of trial and the likelihood of settlement in any event, it is reasonable to suppose that the benefit of an early settlement will often exceed the expected benefit of trial surprise.
Third, the settlement benefits from voluntary disclosure depend on the receiving party processing information rationally. However, as the bounded rationality literature demonstrates, parties and their lawyers do not always behave with perfect rationality (Lowenstein et al. 1993; Korobkin and Guthrie 1994). For example, the self-serving bias causes parties to interpret information in ways that reinforce their pre-existing beliefs. Thus, a party who receives a disclosure might construe the information as less favorable to the disclosing party than a perfectly rational person would. This reduces the benefit from disclosure and thus the incentive to disclose in the first place.
Even so, it seems reasonable to assume that incentives to disclose favorable information voluntarily remain strong in a wide range of cases despite obstacles to verification, trial benefits from concealment, and bounded rationality effects.
2.1.2. Voluntary disclosure of unfavorable information
Although somewhat counterintuitive, there is reason to believe that a party will voluntarily disclose even unfavorable information when disclosure is needed to convince an opponent that the party is not as bad as the opponent assumes he is (Shavell 1989; Hay 1994). For example, an accomplice to a robbery and murder might confess to his role in the robbery in order to convince police that he did not commit the murder.
These disclosure incentives can produce an “unraveling effect,” whereby parties disclose all their private information, both favorable and unfavorable (Hay 1994). Consider the following simple example. Suppose that D has private information about liability. Suppose as well that the strength of a case can vary from 1 (the weakest for P) to 5 (the strongest for P) and p. 191that the different types are uniformly distributed over the interval. Assume P knows these facts but does not know D’s private information about liability in P’s particular case.
Given the uniform distribution, a P who knows nothing else will estimate the strength of her case as 3 (the mean of the uniform distribution). If D has information showing that the case is actually weaker than 3 (i.e., 1 or 2), D will disclose that information in order to convince P of that fact and obtain a better settlement. P knows that D will do this, so if D remains silent, P will infer that the case is a 3, 4 or 5 (since D would have disclosed otherwise). Thus, P will revise her estimate of case strength upward to 4 (the mean of the uniform distribution over the new interval [3, 5]). Anticipating this response from P, D will disclose its information if it shows that case strength is less than 4 (i.e., a 3). Thus, if D remains silent, P will infer the case is a 4 or 5 and revise her estimate of case strength upward to 4.5.
This dynamic repeats itself indefinitely until the only D remaining silent is the D in a case that is strongest for P. All the parties know this will happen, so in equilibrium the pool completely unravels and the defendant immediately reveals all his private information, favorable and unfavorable, in order to convince the plaintiff that the case is in fact weaker than the plaintiff thinks it is.
There are limits to unraveling (Shavell 1989; Hay 1994). For one thing, it works best only if the parties make truthful disclosures, but parties have obvious incentives to mislead. Although the litigation system includes sanctions and other mechanisms to encourage truthful disclosure and parties can sometimes detect deception on their own, there is still room to mislead. In addition, the costs of voluntary disclosure and trial limit the extent of unraveling. And in the absence of discovery, a privately informed party will not disclose information that the party knows will never be revealed at trial. Thus, the risk of deception, coupled with disclosure costs and other factors, reduces the efficacy of unraveling (Cooter and Rubinfeld 1994).
In sum, theory predicts that parties will sometimes disclose favorable and unfavorable information in the absence of discovery, and this prediction is confirmed by empirical evidence showing a significant amount of informal information exchange. The benefits of adding discovery must be evaluated against this baseline.
2.2. Benefits for Settlement
The available data show that less than 6% of filed cases in federal court are actually tried. Given this, the benefits of discovery for settlement loom particularly large. These benefits fall into two broad categories: increasing settlement rate and improving settlement quality.
2.2.1. p. 192Settlement rate
According to the standard argument, pre-trial discovery increases the settlement rate by reducing informational asymmetry (Cooter and Rubinfeld 1994). Formally, assume that P and D each have private information about the case. For example, P might have private information about damages (w) while D has private information about liability (p). Let x = pw be the expected trial award, and let xπ be plaintiff’s subjective estimate of x and xΔ be defendant’s subjective estimate of x. According to the standard settlement model, settlement is feasible only if xπ – xΔ ≤ cπ + cΔ, where cπ and cΔ are P’s and D’s expected costs of litigating the case through trial, respectively. Suppose the parties are mutually optimistic about x. This means xπ is high and xΔ is low, so the difference, xπ – xΔ, is large. The larger xπ – xΔ is, the more likely it is to exceed cπ + cΔ, rendering settlement impossible. Discovery corrects the informational asymmetry, which reduces xπ – xΔ and makes it more likely that the settlement feasibility condition, xπ – xΔ ≤ cπ + cΔ, is satisfied.
One problem with this standard argument is that it ignores the voluntary disclosure baseline. The change in the settlement rate with the addition of discovery depends in part on the mix of favorable and unfavorable information that would have been disclosed in the absence of discovery. For example, suppose the baseline is that parties voluntarily disclose favorable information but not much unfavorable information without discovery. Adding discovery could actually reduce the equilibrium settlement rate (Cooter and Rubinfeld 1994). Discovery will force the disclosure of information unfavorable to the disclosing party, which should make each side more optimistic about its chance of success compared to the baseline. As a result, xπ should increase and xΔ decrease, making it less likely that the settlement feasibility condition, xπ – xΔ ≤ cπ + cΔ, is satisfied. (This ignores discovery’s benefits for settlement quality, which are discussed in Section 2.2.2 below.)
A more complex analysis of voluntary disclosure incentives and the effect of adding discovery must take account of strategic effects. Shavell (1989), for example, models information revelation with privately informed plaintiffs and take-it-or-leave-it settlement offers made by defendants. Adding discovery (which Shavell assumes is costless) reduces the frequency of trials (which occur when some plaintiffs are unable to reveal their private information credibly) and also assures that all plaintiffs who can credibly disclose do so and obtain settlements equal to the expected value of their claims (rather than the mean value over a non-disclosing group).4
p. 193Sobel (1989) models discovery with two-sided incomplete information and lets the plaintiff make a take-it-or-leave-it counteroffer following the defendant’s initial settlement offer. Unlike Shavell, Sobel assumes that disclosure is costly for the disclosing party. The result is that defendants have no incentive to reveal their private information (since disclosure is costly and the plaintiff can capture the entire settlement surplus with her final offer), and some cases go to trial. The benefit of discovery is that it reduces the equilibrium probability of trial by forcing disclosure.5
Farmer and Pecorino (2005) expand on these models. They assume that the plaintiff has private information and can make a voluntary disclosure. If the plaintiff does not voluntarily disclose, the defendant can take discovery. They also assume that voluntary disclosure is costly for the disclosing party and that discovery is costly for both sides. They model both a screening game (defendant makes the offer) and a signaling game (plaintiff makes the offer). In the screening game, discovery takes place in equilibrium when it is not too costly; otherwise, voluntary disclosure takes place. Still, it is the credible threat of mandatory discovery that makes the plaintiff voluntarily disclose. In the signaling model, discovery never takes place, but voluntary disclosure does. As a result, the discovery option reduces trial frequency and thus trial costs. (Of course, it adds discovery and voluntary disclosure costs, which are considered below.)
Other game-theoretic models of the discovery process yield varying results depending on the model’s specifications and the solution concept employed (see, e.g., Mnookin and Wilson 1998). However, all confirm the potential benefit of discovery in increasing the settlement rate and reducing the frequency of trial.
2.2.2. Settlement quality
According to the standard argument, discovery improves settlement quality by reducing informational asymmetry and producing settlements that reflect the true merits of the case (Cooter and Rubinfeld 1994). What is less obvious is that discovery need not actually take place in order to reap at least some of these benefits.
To illustrate, assume that the defendant has private information which, if revealed, would help the plaintiff prove liability at trial and that the defendant would not disclose this information voluntarily. With p. 194discovery available to the plaintiff, defendant must consider that the information will be revealed if the case goes to trial. As for the plaintiff, she does not know the precise information that the defendant possesses, but she does know the kind of information defendants of the same type usually possess and can estimate the expected value of that information for her case.
The defendant anticipates that the plaintiff will make this estimate, so the defendant increases its maximum offer above where it would have been without the availability of discovery. So too, the plaintiff increases its minimum demand (at least if the expected cost of discovery to obtain the information is not too high). Assuming that litigation costs are equal across the party line, the likely result is a larger settlement for the plaintiff when discovery is available. More important, the settlement will be closer to the expected trial judgment with complete information (which can be considered an ideal) because the prospect of discovery forces the defendant to take account of its private information when bargaining (although in a limited way).6 To repeat, it is the credible threat of discovery that produces this result even if no discovery actually takes place.
The result can change, however, if the costs of discovery are distributed unequally between the parties. For example, if the plaintiff can credibly p. 195threaten to seek discovery that would be much more costly for the defendant to provide than for the plaintiff to seek (and the defendant has no credible threat to reciprocate in kind), then the prospect of costly discovery skewed against the defendant could lead to a settlement skewed in that same direction.
Thus, introducing formal discovery has two countervailing effects. First, it reduces informational asymmetry, which can improve settlement quality. Second, it sometimes creates a cost asymmetry across the party line, which can reduce settlement quality.
2.3. Benefits for Trial Outcomes
Proponents of broad discovery often cite its value in producing better trial verdicts, understood as verdicts more in line with the parties’ substantive entitlements. This is especially true if discovery forces disclosure of information that would never be revealed otherwise (Hay 1994). Cooter and Rubinfeld model this effect by focusing on the distribution of trial error (Cooter and Rubinfeld 1994, at 446). Let x* denote the accurate trial outcome when the court has complete information about the facts and the law. Cooter and Rubinfeld suppose that the court observes x* with some error, ε. The court’s observation is unbiased if E(ε) = 0. In this model, pre-trial discovery can be socially beneficial in two possible ways: (1) by reducing bias in the ε distribution, that is, by shifting E(ε) closer to 0, and (2) by reducing variance in the ε distribution.
There are two ways that discovery achieves these benefits. First, it changes the timing of disclosure. Before the Federal Rules of Civil Procedure instituted broad discovery, much information was revealed for the first time at trial, and this gave the opposing party very little time to prepare a response. By requiring disclosure before trial, discovery avoids surprise and allows each side to prepare more effectively. Even so, more time to prepare does not always improve the outcome. As many have noted, advanced preparation sometimes produces equivocal and even deceptive answers under circumstances where a spontaneous question would have elicited a more direct and truthful response. For discovery to improve outcomes by shifting disclosure to an earlier stage, it must be the case that the value of preparation exceeds the value of spontaneity.
The second way that discovery reduces bias and variance is by forcing disclosure of information that would never be revealed at trial (Hay 1994). The magnitude of this benefit depends on how much information is actually disclosed through discovery and how much more is disclosed than would be voluntarily revealed without discovery. These points are discussed in Section 2.2 above.
2.4. p. 196Benefits for Filing Incentives
At first glance, it might seem obvious that discovery would filter out many weak and frivolous suits. If the plaintiff knows that the weak merits of her case will be revealed before trial, she should assign a very low probability to success and refrain from filing suit in the first place. The problem with this analysis is that it assumes discovery actually takes place and succeeds in revealing the plaintiff’s private information to the defendant. But parties sometimes settle before the discovery stage is complete. Indeed, one reason why frivolous suits succeed is that the plaintiff can leverage the high cost of discovery to obtain a pre-discovery settlement (Bone 2003).
2.5. Benefits for Primary Activity Incentives
The availability of discovery can also have beneficial effects on incentives in the real world outside the courtroom. One way it does this is by improving the accuracy of trial verdicts and aligning settlements more closely with the expected trial value of the underlying claims. More accurate outcomes improve compliance with the substantive law.
Discovery can affect primary activity incentives in another, more direct way. Consider a rational party, D, choosing whether to comply with the substantive law. D knows that she might be sued whether she complies or not, but that she is more likely to be sued and pay damages if she does not comply. The cost of discovery itself adds an extra inducement to compliance by boosting the expected cost of a lawsuit. Moreover, suppose that the cost of discovery is higher when D does not comply with the substantive law than when D does (but is sued anyway). For example, a noncomplying D might fight hard to prevent discovery of damaging evidence. This discovery cost differential adds to compliance incentives. Noncompliance under these conditions produces higher discovery costs in two ways: by increasing the likelihood of suit (and thus the likelihood of having to invest in discovery) and by increasing the cost of discovery conditional on suit being filed.
However, the precise relationship between discovery and deterrence is more complicated. More discovery does not necessarily produce greater deterrence. For one thing, the deterrent effect of additional discovery depends on the cost of precautions necessary to comply with the substantive law (Hay 1994, at 502–9). Moreover, a system of broad discovery that deters when it is actually used might not deter if parties routinely settle before the discovery stage. This is important because broader discovery is more costly and thus more likely to pressure settlements in advance. When discovery is not actually undertaken, guilty defendants can pool with innocents and receive the same settlement in equilibrium (Hay 1994, at 513).
p. 197Finally, even if broader discovery enhances deterrence, it might not be socially optimal if its costs are too high. These costs include the expense of those extra precautions that added deterrence induces plus the cost of the broader discovery itself.
3. The Social Costs of Discovery
Discovery is extremely costly. The Federal Judicial Center study found that discovery consumed about half of total litigation costs for the 85% of sampled cases that had some significant discovery and about 90% of total costs for the top 5% of most expensive cases (Willging et al. 1997, at 3–4). Discovery costs appear to be particularly high in large, multi-party lawsuits with high stakes. The large amount of private information makes discovery vital and the high stakes invite strategic abuse (Willging et al. 1997; Garth 1998). Moreover, these costs have increased markedly with the advent of electronic discovery.
Concerns about the cost of discovery take two forms. First, some critics focus on party incentives to engage in excessive discovery, defined as discovery beyond the point where marginal benefit equals marginal cost. Second, some critics focus on party incentives to engage in abusive discovery, defined as discovery aimed not at obtaining information, but at gaining a strategic advantage by threatening the imposition of costs.
Before discussing each of these concerns, it is important to emphasize that the costs of discovery must be evaluated relative to the costs of information acquisition in the absence of discovery. This is important because parties without access to discovery might use more expensive self-help techniques to obtain the information they need, such as hacking into computers, relying on spies, bribing employees, and the like. Firms faced with these risks are likely to adopt countermeasures to deter disclosure. The result can be an escalating and wasteful “arms race” if those seeking information use increasingly sophisticated and costly acquisition techniques and those possessing information respond with sophisticated and costly countermeasures. Formal discovery reduces the reliance on self-help and thus prevents the arms race.
3.1. Excessive Discovery
The incentive to engage in excessive discovery results from the fact that the party requesting discovery does not have to pay the costs of responding. Cooter and Rubinfeld (1994) analyze this externality and argue for making the requesting party pay the responding party’s costs of response beyond a certain threshold level. Hay (1994) criticizes the Cooter-Rubinfeld argument by pointing out that discovery beyond the point where marginal private cost exceeds marginal benefit might sometimes be desirable from p. 198a social point of view. The reason is that private parties do not internalize all the benefits and costs of litigation (Shavell 1982), and in particular all the deterrence benefits and public costs of discovery. In other words, discovery that is excessive in terms of private costs and benefits might not be excessive in terms of social costs and benefits.
Other factors besides cost externalization contribute to excessive discovery. For example, lawyers hired on a fee-for-services basis have an incentive to run up discovery costs in order to pad their fees (Frankel 1993, at 258–9). Also, risk-averse young associates in large law firms, who are often left to handle discovery on their own, have incentives to err on the side of too much discovery out of fear that they might miss something significant and incur a partner’s wrath.
3.2. Abusive Discovery
Discovery is abusive when it is conducted not for its informational value, but rather for its strategic value in imposing costs on an opposing party and leveraging a more favorable settlement (Easterbrook 1989; Setear 1989). Several scholars model discovery abuse as a Prisoners’ Dilemma (PD) game (Setear 1989; Gilson and Mnookin 1994; Bone 2003). In a PD game, each side engages in abusive discovery out of fear that the other side will do so and they will end up a “sucker.”
More specifically, if it costs the requesting party less to make the request than it costs the responding party to respond, the requesting party can gain an advantage by serving a discovery request. If the responding party chooses to settle rather than respond, the resulting settlement will be skewed in the requesting party’s favor. If the responding party simply ignores an abusive request, the requesting party can file a motion to compel, and assuming a sufficiently high probability of a (mistaken) grant, the responding party will take the motion seriously. Thus, an abusive discovery request has teeth insofar as it can be backed up by a credible threat to compel, and in that case the responding party has an incentive to settle rather than incur the high cost of a response.
In the PD game, both sides make abusive discovery requests in order not to be left at a cost disadvantage in settlement. These strategies, however, leave the parties in the same position relative to settlement if the strategies cancel one another out. However, when the discovery requests are actually enforced (which must happen sometimes in order to have a credible threat), the parties end up worse off than they would have been if they both had avoided abusive discovery completely.
The PD game has some plausibility, but it might not be the best way to model abusive discovery. Sanchirico (2007) argues, following Katz (1988), that parties are not likely to engage in mutual aggression in the way the p. 199PD game predicts. Instead, one side is likely to respond to aggression with aggression (strategically complement), while the other side is likely to respond with retreat (strategically substitute). As a result, one side acts more aggressively than the other in equilibrium. These equilibrium strategies make intuitive sense. When A adopts a strategy of retreat in the face of aggression, B will act less aggressively, since A’s retreat reduces the marginal benefit to B of additional aggression. And less aggression by B makes A better off. Shepherd’s empirical study supports the existence of this equilibrium (Shepherd 1999). He finds that defendants respond to discovery aggression by acting aggressively and that plaintiffs respond by retreating.
This equilibrium does not generate the same level of wasteful litigation costs as the equilibrium of the PD model. It is possible, however, that the equilibrium systematically skews settlement in favor of the more aggressive party, thereby distorting the incentive effects of the substantive law.
4. Discovery Reforms
The current system relies primarily on sanctions to control discovery costs, but critics complain that obstacles to enforcement and judicial reluctance to sanction render sanctions less than optimally effective. The following discussion briefly discusses some other methods.
4.1. Mandatory Disclosure
In a system of mandatory disclosure, parties are required at the outset of the litigation to disclose certain core information without a formal discovery request, such as key documents and witness testimony.7 Proponents argue that mandatory disclosure saves the cost of discovery requests and reduces the need to rely on formal discovery later in the suit. Critics argue that formal discovery will be used anyway. Each party knows that the other party is not likely to disclose all required information, especially information that is particularly damaging. As a result, the parties will use formal discovery to check the completeness of the mandatory disclosures (Issacharoff and Loewenstein 1995).
Proponents of mandatory disclosure also cite its benefit in encouraging early settlement by forcing information exchange at the beginning of the lawsuit. Critics argue that reducing the need for formal discovery p. 200reduces the settlement surplus, which in turn reduces the incentive to settle (Issacharoff and Loewenstein 1995). This effect offsets to some extent the settlement benefits of early information exchange.
4.2. Discovery Limits
Another way to limit discovery costs is to limit the number of discovery requests. Quantitative limits can make parties better off by facilitating pre-commitment and avoiding the perverse effects of the Prisoners’ Dilemma (Bone 2007, at 2006–11). To do so effectively, discovery limits should often be strict rather than presumptive.8 Strict limits support pre-commitment more strongly and avoid the strategic abuse that presumptive limits invite. In a system of presumptive limits, parties can take discovery up to the limit and then threaten to drag their opponents through a costly battle to obtain more.
Strict limits, however, suffer from the same defects as strict rules more generally. A strict limit is set based on the needs of the average case and as such can get discovery wrong for cases that depart from the average. For example, a strict limit will furnish less discovery than is optimal for particularly complex and information-dense cases and too much discovery for particularly simple and information-sparse cases. Thus, the objective from an economic perspective must be to devise a system of discovery limits that optimally balances the benefits and costs.
Moss (2007) examines a different approach to limiting discovery, so-called proportionality limits that authorize the trial judge to deny additional discovery when the marginal costs exceed the marginal benefits.9 Using the federal e-discovery rules as his principal target, he argues that judges will find it difficult to strike the optimal cost-benefit balance without the information that discovery would reveal.10
4.3. p. 201Cost Shifting
Cooter and Rubinfeld (1994) argue that insofar as excessive discovery is an externality problem, the solution is to make the requesting party pay the opponent’s response costs. They add a requirement that the judge approve the reasonableness of the response costs in order to prevent the responding party from running up those costs.
There is, however, a problem with a simple cost-shifting rule. When private information is asymmetrically distributed, the party with less information has greater need for discovery and as a result will incur higher discovery costs, especially when it must pay for response costs as well as request costs. This cost asymmetry can produce settlements ex ante that are skewed against the party with less private information and a greater need for discovery. Cooter and Rubinfeld (1994) recognize this problem and try to deal with it by proposing a two-stage cost-shifting rule. Their proposal makes the responding party pay its response costs up to a certain point and then shifts those costs to the requesting party beyond that point. The switching point is set to roughly equalize the discovery costs for both sides.
There are problems with the switching-point proposal and with cost shifting more generally. Still, an approach that combines rule-based limits on discovery with some version of cost-shifting holds promise and is worth exploring further.
This chapter is based on the treatment of discovery in Bone (2003) at 200–31.
It is important to note that a low rate of observed discovery does not necessarily indicate the absence of discovery problems, even in those cases with little discovery. Little or no discovery can be part of a settlement equilibrium supported by credible threats to engage in abusive discovery.
In addition, 62% of attorneys reported some informal exchange of information in cases where formal discovery or disclosure also took place.
One interesting feature of the model is that the beneficial effect on the trial rate results not from parties disclosing and settling when they would otherwise have gone to trial. Rather, those forced to disclose in the model are parties who would have settled anyway.
In Sobel’s model, discovery also benefits plaintiffs by producing higher settlement offers compared to a regime with no disclosure at all.
Suppose the plaintiff’s estimate of the probability of establishing liability at trial without the private information is p. Suppose that the probability of success increases to p + α with the private information (i.e., α measures the value of the private information). If w is the expected trial award conditional on establishing liability, then (p + α)w is the ideal settlement. Let c be the cost to each side of going to trial without any discovery and let c + k be the cost to each side of going to trial with discovery (i.e., k is the cost of discovery for each side). Since the plaintiff has a credible threat to take discovery and obtain the private information, the defendant knows that if the case does not settle, the information will be revealed and its expected loss from trial will be (p + α) w + c + k. The plaintiff does not know the private information but can take an expectation over all possible private information scenarios. Suppose that the plaintiff estimates its trial success at p + β, where β is the plaintiff’s estimate of the expected value of the private information (β < α). Since we assume that costs are the same for both sides, the plaintiff’s minimum demand will be (p + α)w – c – k. Assuming that the settlement bargaining process does not signal defendant’s private information in any way, the defendant, while willing to offer up to (p + α)w + c + k if necessary, will behave as though its maximum offer is (p + β)w + c + k, aware that the plaintiff does not know any better. (This assumes no unraveling.) Now let’s calculate the likely settlements with and without discovery assuming equal bargaining power (so the Nash bargaining solution predicts an equal division of the surplus). With discovery, the settlement will be (p + β)w. Without discovery, the settlement will be pw. Recall that the ideal settlement is (p + α)w. Obviously, the settlement with discovery is closer to the ideal than the settlement without.
For example, Rule 26(a) of the Federal Rules of Civil Procedure imposes a duty to disclose certain categories of information, but limits the duty to information favorable to the disclosing party. Parties are then free to conduct formal discovery later in the litigation.
The Federal Rules of Civil Procedure, for example, employ presumptive limits, such as a presumptive limit of ten depositions and a presumptive limit of 25 interrogatories. A presumptive limit can be overcome by showing that more discovery is needed.
Rule 26(b)(2)(C)(iii) of the Federal Rules of Civil Procedure gives the judge power to deny burdensome discovery requests when the marginal costs of the additional discovery exceed the marginal benefits for the case.
The Federal Rules of Civil Procedure were amended in 2006 to add new provisions regulating discovery of material in electronic digital format. Rule 26(b)(2)(B) states that a party need not furnish discovery of electronic information in response to a discovery request when that information is not reasonably accessible because retrieval involves undue burden or cost. However, the judge can allow such discovery if the party seeking it shows good cause considering, among other things, the cost-benefit proportionality standard.